[llvm] r372092 - Patch from Phabricator
Roman Lebedev via llvm-commits
llvm-commits at lists.llvm.org
Tue Sep 17 02:48:51 PDT 2019
Was this meant to be committed?
The commit message is super unhelpful, consider reverting and
recommitting with a better message, referencing review.
On Tue, Sep 17, 2019 at 12:41 PM Luis Marques via llvm-commits
<llvm-commits at lists.llvm.org> wrote:
>
> Author: luismarques
> Date: Tue Sep 17 02:43:08 2019
> New Revision: 372092
>
> URL: http://llvm.org/viewvc/llvm-project?rev=372092&view=rev
> Log:
> Patch from Phabricator
>
> Modified:
> llvm/trunk/lib/Target/RISCV/RISCVSubtarget.h
> llvm/trunk/test/CodeGen/RISCV/add-before-shl.ll
> llvm/trunk/test/CodeGen/RISCV/addc-adde-sube-subc.ll
> llvm/trunk/test/CodeGen/RISCV/addcarry.ll
> llvm/trunk/test/CodeGen/RISCV/alloca.ll
> llvm/trunk/test/CodeGen/RISCV/alu64.ll
> llvm/trunk/test/CodeGen/RISCV/arith-with-overflow.ll
> llvm/trunk/test/CodeGen/RISCV/atomic-cmpxchg-flag.ll
> llvm/trunk/test/CodeGen/RISCV/atomic-cmpxchg.ll
> llvm/trunk/test/CodeGen/RISCV/atomic-rmw.ll
> llvm/trunk/test/CodeGen/RISCV/bare-select.ll
> llvm/trunk/test/CodeGen/RISCV/blockaddress.ll
> llvm/trunk/test/CodeGen/RISCV/bswap-ctlz-cttz-ctpop.ll
> llvm/trunk/test/CodeGen/RISCV/callee-saved-fpr32s.ll
> llvm/trunk/test/CodeGen/RISCV/callee-saved-fpr64s.ll
> llvm/trunk/test/CodeGen/RISCV/callee-saved-gprs.ll
> llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32-ilp32f-common.ll
> llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32-ilp32f-ilp32d-common.ll
> llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32.ll
> llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32d.ll
> llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32f-ilp32d-common.ll
> llvm/trunk/test/CodeGen/RISCV/calling-conv-lp64-lp64f-lp64d-common.ll
> llvm/trunk/test/CodeGen/RISCV/calling-conv-lp64.ll
> llvm/trunk/test/CodeGen/RISCV/calling-conv-rv32f-ilp32.ll
> llvm/trunk/test/CodeGen/RISCV/codemodel-lowering.ll
> llvm/trunk/test/CodeGen/RISCV/compress.ll
> llvm/trunk/test/CodeGen/RISCV/div.ll
> llvm/trunk/test/CodeGen/RISCV/double-arith.ll
> llvm/trunk/test/CodeGen/RISCV/double-bitmanip-dagcombines.ll
> llvm/trunk/test/CodeGen/RISCV/double-br-fcmp.ll
> llvm/trunk/test/CodeGen/RISCV/double-calling-conv.ll
> llvm/trunk/test/CodeGen/RISCV/double-convert.ll
> llvm/trunk/test/CodeGen/RISCV/double-fcmp.ll
> llvm/trunk/test/CodeGen/RISCV/double-imm.ll
> llvm/trunk/test/CodeGen/RISCV/double-intrinsics.ll
> llvm/trunk/test/CodeGen/RISCV/double-mem.ll
> llvm/trunk/test/CodeGen/RISCV/double-previous-failure.ll
> llvm/trunk/test/CodeGen/RISCV/double-select-fcmp.ll
> llvm/trunk/test/CodeGen/RISCV/double-stack-spill-restore.ll
> llvm/trunk/test/CodeGen/RISCV/float-arith.ll
> llvm/trunk/test/CodeGen/RISCV/float-br-fcmp.ll
> llvm/trunk/test/CodeGen/RISCV/float-convert.ll
> llvm/trunk/test/CodeGen/RISCV/float-fcmp.ll
> llvm/trunk/test/CodeGen/RISCV/float-imm.ll
> llvm/trunk/test/CodeGen/RISCV/float-intrinsics.ll
> llvm/trunk/test/CodeGen/RISCV/float-mem.ll
> llvm/trunk/test/CodeGen/RISCV/float-select-fcmp.ll
> llvm/trunk/test/CodeGen/RISCV/fp128.ll
> llvm/trunk/test/CodeGen/RISCV/frame-info.ll
> llvm/trunk/test/CodeGen/RISCV/get-setcc-result-type.ll
> llvm/trunk/test/CodeGen/RISCV/hoist-global-addr-base.ll
> llvm/trunk/test/CodeGen/RISCV/imm-cse.ll
> llvm/trunk/test/CodeGen/RISCV/imm.ll
> llvm/trunk/test/CodeGen/RISCV/indirectbr.ll
> llvm/trunk/test/CodeGen/RISCV/inline-asm-d-constraint-f.ll
> llvm/trunk/test/CodeGen/RISCV/inline-asm-f-constraint-f.ll
> llvm/trunk/test/CodeGen/RISCV/interrupt-attr-nocall.ll
> llvm/trunk/test/CodeGen/RISCV/legalize-fneg.ll
> llvm/trunk/test/CodeGen/RISCV/lsr-legaladdimm.ll
> llvm/trunk/test/CodeGen/RISCV/mem.ll
> llvm/trunk/test/CodeGen/RISCV/mem64.ll
> llvm/trunk/test/CodeGen/RISCV/mul.ll
> llvm/trunk/test/CodeGen/RISCV/remat.ll
> llvm/trunk/test/CodeGen/RISCV/rv64f-float-convert.ll
> llvm/trunk/test/CodeGen/RISCV/rv64i-complex-float.ll
> llvm/trunk/test/CodeGen/RISCV/rv64i-w-insts-legalization.ll
> llvm/trunk/test/CodeGen/RISCV/select-optimize-multiple.ll
> llvm/trunk/test/CodeGen/RISCV/setcc-logic.ll
> llvm/trunk/test/CodeGen/RISCV/shifts.ll
> llvm/trunk/test/CodeGen/RISCV/split-offsets.ll
> llvm/trunk/test/CodeGen/RISCV/umulo-128-legalisation-lowering.ll
> llvm/trunk/test/CodeGen/RISCV/vararg.ll
> llvm/trunk/test/CodeGen/RISCV/zext-with-load-is-free.ll
>
> Modified: llvm/trunk/lib/Target/RISCV/RISCVSubtarget.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/RISCV/RISCVSubtarget.h?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/RISCV/RISCVSubtarget.h (original)
> +++ llvm/trunk/lib/Target/RISCV/RISCVSubtarget.h Tue Sep 17 02:43:08 2019
> @@ -80,6 +80,7 @@ public:
> const SelectionDAGTargetInfo *getSelectionDAGInfo() const override {
> return &TSInfo;
> }
> + bool enableMachineScheduler() const override { return true; }
> bool hasStdExtM() const { return HasStdExtM; }
> bool hasStdExtA() const { return HasStdExtA; }
> bool hasStdExtF() const { return HasStdExtF; }
>
> Modified: llvm/trunk/test/CodeGen/RISCV/add-before-shl.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/add-before-shl.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/add-before-shl.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/add-before-shl.ll Tue Sep 17 02:43:08 2019
> @@ -96,36 +96,36 @@ define i128 @add_wide_operand(i128 %a) n
> ; RV32I-LABEL: add_wide_operand:
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: lw a2, 0(a1)
> -; RV32I-NEXT: srli a3, a2, 29
> -; RV32I-NEXT: lw a4, 4(a1)
> -; RV32I-NEXT: slli a5, a4, 3
> -; RV32I-NEXT: or a6, a5, a3
> -; RV32I-NEXT: srli a4, a4, 29
> -; RV32I-NEXT: lw a5, 8(a1)
> -; RV32I-NEXT: slli a3, a5, 3
> -; RV32I-NEXT: or a3, a3, a4
> +; RV32I-NEXT: lw a3, 4(a1)
> +; RV32I-NEXT: lw a6, 12(a1)
> +; RV32I-NEXT: lw a1, 8(a1)
> +; RV32I-NEXT: srli a5, a2, 29
> +; RV32I-NEXT: slli a4, a3, 3
> +; RV32I-NEXT: or a4, a4, a5
> +; RV32I-NEXT: srli a3, a3, 29
> +; RV32I-NEXT: slli a5, a1, 3
> +; RV32I-NEXT: or a3, a5, a3
> +; RV32I-NEXT: srli a1, a1, 29
> +; RV32I-NEXT: slli a5, a6, 3
> +; RV32I-NEXT: or a1, a5, a1
> ; RV32I-NEXT: slli a2, a2, 3
> +; RV32I-NEXT: lui a5, 128
> +; RV32I-NEXT: add a1, a1, a5
> ; RV32I-NEXT: sw a2, 0(a0)
> ; RV32I-NEXT: sw a3, 8(a0)
> -; RV32I-NEXT: sw a6, 4(a0)
> -; RV32I-NEXT: srli a2, a5, 29
> -; RV32I-NEXT: lw a1, 12(a1)
> -; RV32I-NEXT: slli a1, a1, 3
> -; RV32I-NEXT: or a1, a1, a2
> -; RV32I-NEXT: lui a2, 128
> -; RV32I-NEXT: add a1, a1, a2
> +; RV32I-NEXT: sw a4, 4(a0)
> ; RV32I-NEXT: sw a1, 12(a0)
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: add_wide_operand:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: slli a1, a1, 3
> ; RV64I-NEXT: srli a2, a0, 61
> +; RV64I-NEXT: slli a1, a1, 3
> ; RV64I-NEXT: or a1, a1, a2
> +; RV64I-NEXT: slli a0, a0, 3
> ; RV64I-NEXT: addi a2, zero, 1
> ; RV64I-NEXT: slli a2, a2, 51
> ; RV64I-NEXT: add a1, a1, a2
> -; RV64I-NEXT: slli a0, a0, 3
> ; RV64I-NEXT: ret
> %1 = add i128 %a, 5192296858534827628530496329220096
> %2 = shl i128 %1, 3
>
> Modified: llvm/trunk/test/CodeGen/RISCV/addc-adde-sube-subc.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/addc-adde-sube-subc.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/addc-adde-sube-subc.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/addc-adde-sube-subc.ll Tue Sep 17 02:43:08 2019
> @@ -20,9 +20,9 @@ define i64 @addc_adde(i64 %a, i64 %b) no
> define i64 @subc_sube(i64 %a, i64 %b) nounwind {
> ; RV32I-LABEL: subc_sube:
> ; RV32I: # %bb.0:
> +; RV32I-NEXT: sltu a4, a0, a2
> ; RV32I-NEXT: sub a1, a1, a3
> -; RV32I-NEXT: sltu a3, a0, a2
> -; RV32I-NEXT: sub a1, a1, a3
> +; RV32I-NEXT: sub a1, a1, a4
> ; RV32I-NEXT: sub a0, a0, a2
> ; RV32I-NEXT: ret
> %1 = sub i64 %a, %b
>
> Modified: llvm/trunk/test/CodeGen/RISCV/addcarry.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/addcarry.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/addcarry.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/addcarry.ll Tue Sep 17 02:43:08 2019
> @@ -10,17 +10,17 @@ define i64 @addcarry(i64 %x, i64 %y) nou
> ; RISCV32-LABEL: addcarry:
> ; RISCV32: # %bb.0:
> ; RISCV32-NEXT: mul a4, a0, a3
> -; RISCV32-NEXT: mulhu a5, a0, a2
> -; RISCV32-NEXT: add a4, a5, a4
> -; RISCV32-NEXT: sltu a6, a4, a5
> -; RISCV32-NEXT: mulhu a5, a0, a3
> -; RISCV32-NEXT: add a6, a5, a6
> -; RISCV32-NEXT: mulhu a5, a1, a2
> -; RISCV32-NEXT: add a7, a6, a5
> +; RISCV32-NEXT: mulhu a7, a0, a2
> +; RISCV32-NEXT: add a4, a7, a4
> ; RISCV32-NEXT: mul a5, a1, a2
> ; RISCV32-NEXT: add a6, a4, a5
> -; RISCV32-NEXT: sltu a4, a6, a4
> -; RISCV32-NEXT: add a4, a7, a4
> +; RISCV32-NEXT: sltu t0, a6, a4
> +; RISCV32-NEXT: sltu a4, a4, a7
> +; RISCV32-NEXT: mulhu a5, a0, a3
> +; RISCV32-NEXT: add a4, a5, a4
> +; RISCV32-NEXT: mulhu a5, a1, a2
> +; RISCV32-NEXT: add a4, a4, a5
> +; RISCV32-NEXT: add a4, a4, t0
> ; RISCV32-NEXT: mul a5, a1, a3
> ; RISCV32-NEXT: add a5, a4, a5
> ; RISCV32-NEXT: bgez a1, .LBB0_2
> @@ -31,13 +31,13 @@ define i64 @addcarry(i64 %x, i64 %y) nou
> ; RISCV32-NEXT: # %bb.3:
> ; RISCV32-NEXT: sub a5, a5, a0
> ; RISCV32-NEXT: .LBB0_4:
> +; RISCV32-NEXT: slli a1, a5, 30
> +; RISCV32-NEXT: srli a3, a6, 2
> +; RISCV32-NEXT: or a1, a1, a3
> +; RISCV32-NEXT: slli a3, a6, 30
> ; RISCV32-NEXT: mul a0, a0, a2
> ; RISCV32-NEXT: srli a0, a0, 2
> -; RISCV32-NEXT: slli a1, a6, 30
> -; RISCV32-NEXT: or a0, a1, a0
> -; RISCV32-NEXT: srli a1, a6, 2
> -; RISCV32-NEXT: slli a2, a5, 30
> -; RISCV32-NEXT: or a1, a2, a1
> +; RISCV32-NEXT: or a0, a3, a0
> ; RISCV32-NEXT: ret
> %tmp = call i64 @llvm.smul.fix.i64(i64 %x, i64 %y, i32 2);
> ret i64 %tmp;
>
> Modified: llvm/trunk/test/CodeGen/RISCV/alloca.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/alloca.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/alloca.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/alloca.ll Tue Sep 17 02:43:08 2019
> @@ -82,8 +82,7 @@ define void @alloca_callframe(i32 %n) no
> ; RV32I-NEXT: sw a1, 8(sp)
> ; RV32I-NEXT: addi a1, zero, 10
> ; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: addi a1, zero, 9
> -; RV32I-NEXT: sw a1, 0(sp)
> +; RV32I-NEXT: addi t0, zero, 9
> ; RV32I-NEXT: addi a1, zero, 2
> ; RV32I-NEXT: addi a2, zero, 3
> ; RV32I-NEXT: addi a3, zero, 4
> @@ -91,6 +90,7 @@ define void @alloca_callframe(i32 %n) no
> ; RV32I-NEXT: addi a5, zero, 6
> ; RV32I-NEXT: addi a6, zero, 7
> ; RV32I-NEXT: addi a7, zero, 8
> +; RV32I-NEXT: sw t0, 0(sp)
> ; RV32I-NEXT: call func
> ; RV32I-NEXT: addi sp, sp, 16
> ; RV32I-NEXT: addi sp, s0, -16
>
> Modified: llvm/trunk/test/CodeGen/RISCV/alu64.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/alu64.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/alu64.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/alu64.ll Tue Sep 17 02:43:08 2019
> @@ -123,8 +123,8 @@ define i64 @slli(i64 %a) nounwind {
> ;
> ; RV32I-LABEL: slli:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: slli a1, a1, 7
> ; RV32I-NEXT: srli a2, a0, 25
> +; RV32I-NEXT: slli a1, a1, 7
> ; RV32I-NEXT: or a1, a1, a2
> ; RV32I-NEXT: slli a0, a0, 7
> ; RV32I-NEXT: ret
> @@ -140,8 +140,8 @@ define i64 @srli(i64 %a) nounwind {
> ;
> ; RV32I-LABEL: srli:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: srli a0, a0, 8
> ; RV32I-NEXT: slli a2, a1, 24
> +; RV32I-NEXT: srli a0, a0, 8
> ; RV32I-NEXT: or a0, a0, a2
> ; RV32I-NEXT: srli a1, a1, 8
> ; RV32I-NEXT: ret
> @@ -157,8 +157,8 @@ define i64 @srai(i64 %a) nounwind {
> ;
> ; RV32I-LABEL: srai:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: srli a0, a0, 9
> ; RV32I-NEXT: slli a2, a1, 23
> +; RV32I-NEXT: srli a0, a0, 9
> ; RV32I-NEXT: or a0, a0, a2
> ; RV32I-NEXT: srai a1, a1, 9
> ; RV32I-NEXT: ret
> @@ -194,9 +194,9 @@ define i64 @sub(i64 %a, i64 %b) nounwind
> ;
> ; RV32I-LABEL: sub:
> ; RV32I: # %bb.0:
> +; RV32I-NEXT: sltu a4, a0, a2
> ; RV32I-NEXT: sub a1, a1, a3
> -; RV32I-NEXT: sltu a3, a0, a2
> -; RV32I-NEXT: sub a1, a1, a3
> +; RV32I-NEXT: sub a1, a1, a4
> ; RV32I-NEXT: sub a0, a0, a2
> ; RV32I-NEXT: ret
> %1 = sub i64 %a, %b
> @@ -218,13 +218,14 @@ define i64 @sll(i64 %a, i64 %b) nounwind
> ; RV32I-NEXT: mv a0, zero
> ; RV32I-NEXT: ret
> ; RV32I-NEXT: .LBB11_2:
> +; RV32I-NEXT: sll a1, a1, a2
> ; RV32I-NEXT: addi a3, zero, 31
> ; RV32I-NEXT: sub a3, a3, a2
> ; RV32I-NEXT: srli a4, a0, 1
> ; RV32I-NEXT: srl a3, a4, a3
> -; RV32I-NEXT: sll a1, a1, a2
> ; RV32I-NEXT: or a1, a1, a3
> -; RV32I-NEXT: sll a0, a0, a2
> +; RV32I-NEXT: sll a2, a0, a2
> +; RV32I-NEXT: mv a0, a2
> ; RV32I-NEXT: ret
> %1 = shl i64 %a, %b
> ret i64 %1
> @@ -304,13 +305,14 @@ define i64 @srl(i64 %a, i64 %b) nounwind
> ; RV32I-NEXT: mv a1, zero
> ; RV32I-NEXT: ret
> ; RV32I-NEXT: .LBB15_2:
> +; RV32I-NEXT: srl a0, a0, a2
> ; RV32I-NEXT: addi a3, zero, 31
> ; RV32I-NEXT: sub a3, a3, a2
> ; RV32I-NEXT: slli a4, a1, 1
> ; RV32I-NEXT: sll a3, a4, a3
> -; RV32I-NEXT: srl a0, a0, a2
> ; RV32I-NEXT: or a0, a0, a3
> -; RV32I-NEXT: srl a1, a1, a2
> +; RV32I-NEXT: srl a2, a1, a2
> +; RV32I-NEXT: mv a1, a2
> ; RV32I-NEXT: ret
> %1 = lshr i64 %a, %b
> ret i64 %1
> @@ -331,11 +333,11 @@ define i64 @sra(i64 %a, i64 %b) nounwind
> ; RV32I-NEXT: srai a1, a1, 31
> ; RV32I-NEXT: ret
> ; RV32I-NEXT: .LBB16_2:
> +; RV32I-NEXT: srl a0, a0, a2
> ; RV32I-NEXT: addi a3, zero, 31
> ; RV32I-NEXT: sub a3, a3, a2
> ; RV32I-NEXT: slli a4, a1, 1
> ; RV32I-NEXT: sll a3, a4, a3
> -; RV32I-NEXT: srl a0, a0, a2
> ; RV32I-NEXT: or a0, a0, a3
> ; RV32I-NEXT: sra a1, a1, a2
> ; RV32I-NEXT: ret
>
> Modified: llvm/trunk/test/CodeGen/RISCV/arith-with-overflow.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/arith-with-overflow.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/arith-with-overflow.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/arith-with-overflow.ll Tue Sep 17 02:43:08 2019
> @@ -10,17 +10,17 @@ declare {i32, i1} @llvm.usub.with.overfl
> define i1 @sadd(i32 %a, i32 %b, i32* %c) nounwind {
> ; RV32I-LABEL: sadd:
> ; RV32I: # %bb.0: # %entry
> -; RV32I-NEXT: add a3, a0, a1
> -; RV32I-NEXT: sw a3, 0(a2)
> -; RV32I-NEXT: addi a2, zero, -1
> -; RV32I-NEXT: slt a1, a2, a1
> -; RV32I-NEXT: slt a0, a2, a0
> -; RV32I-NEXT: slt a2, a2, a3
> -; RV32I-NEXT: xor a2, a0, a2
> -; RV32I-NEXT: xor a0, a0, a1
> -; RV32I-NEXT: seqz a0, a0
> -; RV32I-NEXT: snez a1, a2
> -; RV32I-NEXT: and a0, a0, a1
> +; RV32I-NEXT: addi a3, zero, -1
> +; RV32I-NEXT: slt a4, a3, a1
> +; RV32I-NEXT: slt a5, a3, a0
> +; RV32I-NEXT: xor a4, a5, a4
> +; RV32I-NEXT: seqz a4, a4
> +; RV32I-NEXT: add a1, a0, a1
> +; RV32I-NEXT: slt a0, a3, a1
> +; RV32I-NEXT: xor a0, a5, a0
> +; RV32I-NEXT: snez a0, a0
> +; RV32I-NEXT: and a0, a4, a0
> +; RV32I-NEXT: sw a1, 0(a2)
> ; RV32I-NEXT: ret
> entry:
> %x = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
> @@ -33,17 +33,17 @@ entry:
> define i1 @ssub(i32 %a, i32 %b, i32* %c) nounwind {
> ; RV32I-LABEL: ssub:
> ; RV32I: # %bb.0: # %entry
> -; RV32I-NEXT: sub a3, a0, a1
> -; RV32I-NEXT: sw a3, 0(a2)
> -; RV32I-NEXT: addi a2, zero, -1
> -; RV32I-NEXT: slt a1, a2, a1
> -; RV32I-NEXT: slt a0, a2, a0
> -; RV32I-NEXT: slt a2, a2, a3
> -; RV32I-NEXT: xor a2, a0, a2
> -; RV32I-NEXT: xor a0, a0, a1
> +; RV32I-NEXT: addi a3, zero, -1
> +; RV32I-NEXT: slt a4, a3, a1
> +; RV32I-NEXT: slt a5, a3, a0
> +; RV32I-NEXT: xor a4, a5, a4
> +; RV32I-NEXT: snez a4, a4
> +; RV32I-NEXT: sub a1, a0, a1
> +; RV32I-NEXT: slt a0, a3, a1
> +; RV32I-NEXT: xor a0, a5, a0
> ; RV32I-NEXT: snez a0, a0
> -; RV32I-NEXT: snez a1, a2
> -; RV32I-NEXT: and a0, a0, a1
> +; RV32I-NEXT: and a0, a4, a0
> +; RV32I-NEXT: sw a1, 0(a2)
> ; RV32I-NEXT: ret
> entry:
> %x = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
> @@ -57,8 +57,8 @@ define i1 @uadd(i32 %a, i32 %b, i32* %c)
> ; RV32I-LABEL: uadd:
> ; RV32I: # %bb.0: # %entry
> ; RV32I-NEXT: add a1, a0, a1
> -; RV32I-NEXT: sw a1, 0(a2)
> ; RV32I-NEXT: sltu a0, a1, a0
> +; RV32I-NEXT: sw a1, 0(a2)
> ; RV32I-NEXT: ret
> entry:
> %x = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
> @@ -72,8 +72,8 @@ define i1 @usub(i32 %a, i32 %b, i32* %c)
> ; RV32I-LABEL: usub:
> ; RV32I: # %bb.0: # %entry
> ; RV32I-NEXT: sub a1, a0, a1
> -; RV32I-NEXT: sw a1, 0(a2)
> ; RV32I-NEXT: sltu a0, a0, a1
> +; RV32I-NEXT: sw a1, 0(a2)
> ; RV32I-NEXT: ret
> entry:
> %x = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/atomic-cmpxchg-flag.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/atomic-cmpxchg-flag.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/atomic-cmpxchg-flag.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/atomic-cmpxchg-flag.ll Tue Sep 17 02:43:08 2019
> @@ -7,7 +7,6 @@
> ; higher bits were masked to zero for the comparison.
>
> define i1 @cmpxchg_i32_seq_cst_seq_cst(i32* %ptr, i32 signext %cmp,
> - i32 signext %val) nounwind {
> ; RV64IA-LABEL: cmpxchg_i32_seq_cst_seq_cst:
> ; RV64IA: # %bb.0: # %entry
> ; RV64IA-NEXT: .LBB0_1: # %entry
> @@ -22,6 +21,7 @@ define i1 @cmpxchg_i32_seq_cst_seq_cst(i
> ; RV64IA-NEXT: xor a0, a3, a1
> ; RV64IA-NEXT: seqz a0, a0
> ; RV64IA-NEXT: ret
> + i32 signext %val) nounwind {
> entry:
> %0 = cmpxchg i32* %ptr, i32 %cmp, i32 %val seq_cst seq_cst
> %1 = extractvalue { i32, i1 } %0, 1
>
> Modified: llvm/trunk/test/CodeGen/RISCV/atomic-cmpxchg.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/atomic-cmpxchg.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/atomic-cmpxchg.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/atomic-cmpxchg.ll Tue Sep 17 02:43:08 2019
> @@ -24,24 +24,24 @@ define void @cmpxchg_i8_monotonic_monoto
> ;
> ; RV32IA-LABEL: cmpxchg_i8_monotonic_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a4, a4, a3
> -; RV32IA-NEXT: andi a2, a2, 255
> -; RV32IA-NEXT: sll a2, a2, a3
> +; RV32IA-NEXT: sll a4, a4, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: andi a2, a2, 255
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB0_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a3, (a0)
> -; RV32IA-NEXT: and a5, a3, a4
> +; RV32IA-NEXT: lr.w a2, (a3)
> +; RV32IA-NEXT: and a5, a2, a4
> ; RV32IA-NEXT: bne a5, a1, .LBB0_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB0_1 Depth=1
> -; RV32IA-NEXT: xor a5, a3, a2
> +; RV32IA-NEXT: xor a5, a2, a0
> ; RV32IA-NEXT: and a5, a5, a4
> -; RV32IA-NEXT: xor a5, a3, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a2, a5
> +; RV32IA-NEXT: sc.w a5, a5, (a3)
> ; RV32IA-NEXT: bnez a5, .LBB0_1
> ; RV32IA-NEXT: .LBB0_3:
> ; RV32IA-NEXT: ret
> @@ -61,24 +61,24 @@ define void @cmpxchg_i8_monotonic_monoto
> ;
> ; RV64IA-LABEL: cmpxchg_i8_monotonic_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a4, a4, a3
> -; RV64IA-NEXT: andi a2, a2, 255
> -; RV64IA-NEXT: sllw a2, a2, a3
> +; RV64IA-NEXT: sllw a4, a4, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: andi a2, a2, 255
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB0_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a3, (a0)
> -; RV64IA-NEXT: and a5, a3, a4
> +; RV64IA-NEXT: lr.w a2, (a3)
> +; RV64IA-NEXT: and a5, a2, a4
> ; RV64IA-NEXT: bne a5, a1, .LBB0_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB0_1 Depth=1
> -; RV64IA-NEXT: xor a5, a3, a2
> +; RV64IA-NEXT: xor a5, a2, a0
> ; RV64IA-NEXT: and a5, a5, a4
> -; RV64IA-NEXT: xor a5, a3, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a2, a5
> +; RV64IA-NEXT: sc.w a5, a5, (a3)
> ; RV64IA-NEXT: bnez a5, .LBB0_1
> ; RV64IA-NEXT: .LBB0_3:
> ; RV64IA-NEXT: ret
> @@ -102,24 +102,24 @@ define void @cmpxchg_i8_acquire_monotoni
> ;
> ; RV32IA-LABEL: cmpxchg_i8_acquire_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a4, a4, a3
> -; RV32IA-NEXT: andi a2, a2, 255
> -; RV32IA-NEXT: sll a2, a2, a3
> +; RV32IA-NEXT: sll a4, a4, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: andi a2, a2, 255
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB1_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a3, (a0)
> -; RV32IA-NEXT: and a5, a3, a4
> +; RV32IA-NEXT: lr.w.aq a2, (a3)
> +; RV32IA-NEXT: and a5, a2, a4
> ; RV32IA-NEXT: bne a5, a1, .LBB1_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB1_1 Depth=1
> -; RV32IA-NEXT: xor a5, a3, a2
> +; RV32IA-NEXT: xor a5, a2, a0
> ; RV32IA-NEXT: and a5, a5, a4
> -; RV32IA-NEXT: xor a5, a3, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a2, a5
> +; RV32IA-NEXT: sc.w a5, a5, (a3)
> ; RV32IA-NEXT: bnez a5, .LBB1_1
> ; RV32IA-NEXT: .LBB1_3:
> ; RV32IA-NEXT: ret
> @@ -139,24 +139,24 @@ define void @cmpxchg_i8_acquire_monotoni
> ;
> ; RV64IA-LABEL: cmpxchg_i8_acquire_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a4, a4, a3
> -; RV64IA-NEXT: andi a2, a2, 255
> -; RV64IA-NEXT: sllw a2, a2, a3
> +; RV64IA-NEXT: sllw a4, a4, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: andi a2, a2, 255
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB1_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a3, (a0)
> -; RV64IA-NEXT: and a5, a3, a4
> +; RV64IA-NEXT: lr.w.aq a2, (a3)
> +; RV64IA-NEXT: and a5, a2, a4
> ; RV64IA-NEXT: bne a5, a1, .LBB1_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB1_1 Depth=1
> -; RV64IA-NEXT: xor a5, a3, a2
> +; RV64IA-NEXT: xor a5, a2, a0
> ; RV64IA-NEXT: and a5, a5, a4
> -; RV64IA-NEXT: xor a5, a3, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a2, a5
> +; RV64IA-NEXT: sc.w a5, a5, (a3)
> ; RV64IA-NEXT: bnez a5, .LBB1_1
> ; RV64IA-NEXT: .LBB1_3:
> ; RV64IA-NEXT: ret
> @@ -180,24 +180,24 @@ define void @cmpxchg_i8_acquire_acquire(
> ;
> ; RV32IA-LABEL: cmpxchg_i8_acquire_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a4, a4, a3
> -; RV32IA-NEXT: andi a2, a2, 255
> -; RV32IA-NEXT: sll a2, a2, a3
> +; RV32IA-NEXT: sll a4, a4, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: andi a2, a2, 255
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB2_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a3, (a0)
> -; RV32IA-NEXT: and a5, a3, a4
> +; RV32IA-NEXT: lr.w.aq a2, (a3)
> +; RV32IA-NEXT: and a5, a2, a4
> ; RV32IA-NEXT: bne a5, a1, .LBB2_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB2_1 Depth=1
> -; RV32IA-NEXT: xor a5, a3, a2
> +; RV32IA-NEXT: xor a5, a2, a0
> ; RV32IA-NEXT: and a5, a5, a4
> -; RV32IA-NEXT: xor a5, a3, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a2, a5
> +; RV32IA-NEXT: sc.w a5, a5, (a3)
> ; RV32IA-NEXT: bnez a5, .LBB2_1
> ; RV32IA-NEXT: .LBB2_3:
> ; RV32IA-NEXT: ret
> @@ -217,24 +217,24 @@ define void @cmpxchg_i8_acquire_acquire(
> ;
> ; RV64IA-LABEL: cmpxchg_i8_acquire_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a4, a4, a3
> -; RV64IA-NEXT: andi a2, a2, 255
> -; RV64IA-NEXT: sllw a2, a2, a3
> +; RV64IA-NEXT: sllw a4, a4, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: andi a2, a2, 255
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB2_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a3, (a0)
> -; RV64IA-NEXT: and a5, a3, a4
> +; RV64IA-NEXT: lr.w.aq a2, (a3)
> +; RV64IA-NEXT: and a5, a2, a4
> ; RV64IA-NEXT: bne a5, a1, .LBB2_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB2_1 Depth=1
> -; RV64IA-NEXT: xor a5, a3, a2
> +; RV64IA-NEXT: xor a5, a2, a0
> ; RV64IA-NEXT: and a5, a5, a4
> -; RV64IA-NEXT: xor a5, a3, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a2, a5
> +; RV64IA-NEXT: sc.w a5, a5, (a3)
> ; RV64IA-NEXT: bnez a5, .LBB2_1
> ; RV64IA-NEXT: .LBB2_3:
> ; RV64IA-NEXT: ret
> @@ -258,24 +258,24 @@ define void @cmpxchg_i8_release_monotoni
> ;
> ; RV32IA-LABEL: cmpxchg_i8_release_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a4, a4, a3
> -; RV32IA-NEXT: andi a2, a2, 255
> -; RV32IA-NEXT: sll a2, a2, a3
> +; RV32IA-NEXT: sll a4, a4, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: andi a2, a2, 255
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB3_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a3, (a0)
> -; RV32IA-NEXT: and a5, a3, a4
> +; RV32IA-NEXT: lr.w a2, (a3)
> +; RV32IA-NEXT: and a5, a2, a4
> ; RV32IA-NEXT: bne a5, a1, .LBB3_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB3_1 Depth=1
> -; RV32IA-NEXT: xor a5, a3, a2
> +; RV32IA-NEXT: xor a5, a2, a0
> ; RV32IA-NEXT: and a5, a5, a4
> -; RV32IA-NEXT: xor a5, a3, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a2, a5
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a3)
> ; RV32IA-NEXT: bnez a5, .LBB3_1
> ; RV32IA-NEXT: .LBB3_3:
> ; RV32IA-NEXT: ret
> @@ -295,24 +295,24 @@ define void @cmpxchg_i8_release_monotoni
> ;
> ; RV64IA-LABEL: cmpxchg_i8_release_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a4, a4, a3
> -; RV64IA-NEXT: andi a2, a2, 255
> -; RV64IA-NEXT: sllw a2, a2, a3
> +; RV64IA-NEXT: sllw a4, a4, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: andi a2, a2, 255
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB3_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a3, (a0)
> -; RV64IA-NEXT: and a5, a3, a4
> +; RV64IA-NEXT: lr.w a2, (a3)
> +; RV64IA-NEXT: and a5, a2, a4
> ; RV64IA-NEXT: bne a5, a1, .LBB3_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB3_1 Depth=1
> -; RV64IA-NEXT: xor a5, a3, a2
> +; RV64IA-NEXT: xor a5, a2, a0
> ; RV64IA-NEXT: and a5, a5, a4
> -; RV64IA-NEXT: xor a5, a3, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a2, a5
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a3)
> ; RV64IA-NEXT: bnez a5, .LBB3_1
> ; RV64IA-NEXT: .LBB3_3:
> ; RV64IA-NEXT: ret
> @@ -336,24 +336,24 @@ define void @cmpxchg_i8_release_acquire(
> ;
> ; RV32IA-LABEL: cmpxchg_i8_release_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a4, a4, a3
> -; RV32IA-NEXT: andi a2, a2, 255
> -; RV32IA-NEXT: sll a2, a2, a3
> +; RV32IA-NEXT: sll a4, a4, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: andi a2, a2, 255
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB4_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a3, (a0)
> -; RV32IA-NEXT: and a5, a3, a4
> +; RV32IA-NEXT: lr.w a2, (a3)
> +; RV32IA-NEXT: and a5, a2, a4
> ; RV32IA-NEXT: bne a5, a1, .LBB4_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB4_1 Depth=1
> -; RV32IA-NEXT: xor a5, a3, a2
> +; RV32IA-NEXT: xor a5, a2, a0
> ; RV32IA-NEXT: and a5, a5, a4
> -; RV32IA-NEXT: xor a5, a3, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a2, a5
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a3)
> ; RV32IA-NEXT: bnez a5, .LBB4_1
> ; RV32IA-NEXT: .LBB4_3:
> ; RV32IA-NEXT: ret
> @@ -373,24 +373,24 @@ define void @cmpxchg_i8_release_acquire(
> ;
> ; RV64IA-LABEL: cmpxchg_i8_release_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a4, a4, a3
> -; RV64IA-NEXT: andi a2, a2, 255
> -; RV64IA-NEXT: sllw a2, a2, a3
> +; RV64IA-NEXT: sllw a4, a4, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: andi a2, a2, 255
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB4_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a3, (a0)
> -; RV64IA-NEXT: and a5, a3, a4
> +; RV64IA-NEXT: lr.w a2, (a3)
> +; RV64IA-NEXT: and a5, a2, a4
> ; RV64IA-NEXT: bne a5, a1, .LBB4_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB4_1 Depth=1
> -; RV64IA-NEXT: xor a5, a3, a2
> +; RV64IA-NEXT: xor a5, a2, a0
> ; RV64IA-NEXT: and a5, a5, a4
> -; RV64IA-NEXT: xor a5, a3, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a2, a5
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a3)
> ; RV64IA-NEXT: bnez a5, .LBB4_1
> ; RV64IA-NEXT: .LBB4_3:
> ; RV64IA-NEXT: ret
> @@ -414,24 +414,24 @@ define void @cmpxchg_i8_acq_rel_monotoni
> ;
> ; RV32IA-LABEL: cmpxchg_i8_acq_rel_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a4, a4, a3
> -; RV32IA-NEXT: andi a2, a2, 255
> -; RV32IA-NEXT: sll a2, a2, a3
> +; RV32IA-NEXT: sll a4, a4, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: andi a2, a2, 255
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB5_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a3, (a0)
> -; RV32IA-NEXT: and a5, a3, a4
> +; RV32IA-NEXT: lr.w.aq a2, (a3)
> +; RV32IA-NEXT: and a5, a2, a4
> ; RV32IA-NEXT: bne a5, a1, .LBB5_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB5_1 Depth=1
> -; RV32IA-NEXT: xor a5, a3, a2
> +; RV32IA-NEXT: xor a5, a2, a0
> ; RV32IA-NEXT: and a5, a5, a4
> -; RV32IA-NEXT: xor a5, a3, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a2, a5
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a3)
> ; RV32IA-NEXT: bnez a5, .LBB5_1
> ; RV32IA-NEXT: .LBB5_3:
> ; RV32IA-NEXT: ret
> @@ -451,24 +451,24 @@ define void @cmpxchg_i8_acq_rel_monotoni
> ;
> ; RV64IA-LABEL: cmpxchg_i8_acq_rel_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a4, a4, a3
> -; RV64IA-NEXT: andi a2, a2, 255
> -; RV64IA-NEXT: sllw a2, a2, a3
> +; RV64IA-NEXT: sllw a4, a4, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: andi a2, a2, 255
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB5_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a3, (a0)
> -; RV64IA-NEXT: and a5, a3, a4
> +; RV64IA-NEXT: lr.w.aq a2, (a3)
> +; RV64IA-NEXT: and a5, a2, a4
> ; RV64IA-NEXT: bne a5, a1, .LBB5_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB5_1 Depth=1
> -; RV64IA-NEXT: xor a5, a3, a2
> +; RV64IA-NEXT: xor a5, a2, a0
> ; RV64IA-NEXT: and a5, a5, a4
> -; RV64IA-NEXT: xor a5, a3, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a2, a5
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a3)
> ; RV64IA-NEXT: bnez a5, .LBB5_1
> ; RV64IA-NEXT: .LBB5_3:
> ; RV64IA-NEXT: ret
> @@ -492,24 +492,24 @@ define void @cmpxchg_i8_acq_rel_acquire(
> ;
> ; RV32IA-LABEL: cmpxchg_i8_acq_rel_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a4, a4, a3
> -; RV32IA-NEXT: andi a2, a2, 255
> -; RV32IA-NEXT: sll a2, a2, a3
> +; RV32IA-NEXT: sll a4, a4, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: andi a2, a2, 255
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB6_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a3, (a0)
> -; RV32IA-NEXT: and a5, a3, a4
> +; RV32IA-NEXT: lr.w.aq a2, (a3)
> +; RV32IA-NEXT: and a5, a2, a4
> ; RV32IA-NEXT: bne a5, a1, .LBB6_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB6_1 Depth=1
> -; RV32IA-NEXT: xor a5, a3, a2
> +; RV32IA-NEXT: xor a5, a2, a0
> ; RV32IA-NEXT: and a5, a5, a4
> -; RV32IA-NEXT: xor a5, a3, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a2, a5
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a3)
> ; RV32IA-NEXT: bnez a5, .LBB6_1
> ; RV32IA-NEXT: .LBB6_3:
> ; RV32IA-NEXT: ret
> @@ -529,24 +529,24 @@ define void @cmpxchg_i8_acq_rel_acquire(
> ;
> ; RV64IA-LABEL: cmpxchg_i8_acq_rel_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a4, a4, a3
> -; RV64IA-NEXT: andi a2, a2, 255
> -; RV64IA-NEXT: sllw a2, a2, a3
> +; RV64IA-NEXT: sllw a4, a4, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: andi a2, a2, 255
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB6_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a3, (a0)
> -; RV64IA-NEXT: and a5, a3, a4
> +; RV64IA-NEXT: lr.w.aq a2, (a3)
> +; RV64IA-NEXT: and a5, a2, a4
> ; RV64IA-NEXT: bne a5, a1, .LBB6_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB6_1 Depth=1
> -; RV64IA-NEXT: xor a5, a3, a2
> +; RV64IA-NEXT: xor a5, a2, a0
> ; RV64IA-NEXT: and a5, a5, a4
> -; RV64IA-NEXT: xor a5, a3, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a2, a5
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a3)
> ; RV64IA-NEXT: bnez a5, .LBB6_1
> ; RV64IA-NEXT: .LBB6_3:
> ; RV64IA-NEXT: ret
> @@ -570,24 +570,24 @@ define void @cmpxchg_i8_seq_cst_monotoni
> ;
> ; RV32IA-LABEL: cmpxchg_i8_seq_cst_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a4, a4, a3
> -; RV32IA-NEXT: andi a2, a2, 255
> -; RV32IA-NEXT: sll a2, a2, a3
> +; RV32IA-NEXT: sll a4, a4, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: andi a2, a2, 255
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB7_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a3, (a0)
> -; RV32IA-NEXT: and a5, a3, a4
> +; RV32IA-NEXT: lr.w.aqrl a2, (a3)
> +; RV32IA-NEXT: and a5, a2, a4
> ; RV32IA-NEXT: bne a5, a1, .LBB7_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB7_1 Depth=1
> -; RV32IA-NEXT: xor a5, a3, a2
> +; RV32IA-NEXT: xor a5, a2, a0
> ; RV32IA-NEXT: and a5, a5, a4
> -; RV32IA-NEXT: xor a5, a3, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a2, a5
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a3)
> ; RV32IA-NEXT: bnez a5, .LBB7_1
> ; RV32IA-NEXT: .LBB7_3:
> ; RV32IA-NEXT: ret
> @@ -607,24 +607,24 @@ define void @cmpxchg_i8_seq_cst_monotoni
> ;
> ; RV64IA-LABEL: cmpxchg_i8_seq_cst_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a4, a4, a3
> -; RV64IA-NEXT: andi a2, a2, 255
> -; RV64IA-NEXT: sllw a2, a2, a3
> +; RV64IA-NEXT: sllw a4, a4, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: andi a2, a2, 255
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB7_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a3, (a0)
> -; RV64IA-NEXT: and a5, a3, a4
> +; RV64IA-NEXT: lr.w.aqrl a2, (a3)
> +; RV64IA-NEXT: and a5, a2, a4
> ; RV64IA-NEXT: bne a5, a1, .LBB7_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB7_1 Depth=1
> -; RV64IA-NEXT: xor a5, a3, a2
> +; RV64IA-NEXT: xor a5, a2, a0
> ; RV64IA-NEXT: and a5, a5, a4
> -; RV64IA-NEXT: xor a5, a3, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a2, a5
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a3)
> ; RV64IA-NEXT: bnez a5, .LBB7_1
> ; RV64IA-NEXT: .LBB7_3:
> ; RV64IA-NEXT: ret
> @@ -648,24 +648,24 @@ define void @cmpxchg_i8_seq_cst_acquire(
> ;
> ; RV32IA-LABEL: cmpxchg_i8_seq_cst_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a4, a4, a3
> -; RV32IA-NEXT: andi a2, a2, 255
> -; RV32IA-NEXT: sll a2, a2, a3
> +; RV32IA-NEXT: sll a4, a4, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: andi a2, a2, 255
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB8_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a3, (a0)
> -; RV32IA-NEXT: and a5, a3, a4
> +; RV32IA-NEXT: lr.w.aqrl a2, (a3)
> +; RV32IA-NEXT: and a5, a2, a4
> ; RV32IA-NEXT: bne a5, a1, .LBB8_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB8_1 Depth=1
> -; RV32IA-NEXT: xor a5, a3, a2
> +; RV32IA-NEXT: xor a5, a2, a0
> ; RV32IA-NEXT: and a5, a5, a4
> -; RV32IA-NEXT: xor a5, a3, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a2, a5
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a3)
> ; RV32IA-NEXT: bnez a5, .LBB8_1
> ; RV32IA-NEXT: .LBB8_3:
> ; RV32IA-NEXT: ret
> @@ -685,24 +685,24 @@ define void @cmpxchg_i8_seq_cst_acquire(
> ;
> ; RV64IA-LABEL: cmpxchg_i8_seq_cst_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a4, a4, a3
> -; RV64IA-NEXT: andi a2, a2, 255
> -; RV64IA-NEXT: sllw a2, a2, a3
> +; RV64IA-NEXT: sllw a4, a4, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: andi a2, a2, 255
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB8_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a3, (a0)
> -; RV64IA-NEXT: and a5, a3, a4
> +; RV64IA-NEXT: lr.w.aqrl a2, (a3)
> +; RV64IA-NEXT: and a5, a2, a4
> ; RV64IA-NEXT: bne a5, a1, .LBB8_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB8_1 Depth=1
> -; RV64IA-NEXT: xor a5, a3, a2
> +; RV64IA-NEXT: xor a5, a2, a0
> ; RV64IA-NEXT: and a5, a5, a4
> -; RV64IA-NEXT: xor a5, a3, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a2, a5
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a3)
> ; RV64IA-NEXT: bnez a5, .LBB8_1
> ; RV64IA-NEXT: .LBB8_3:
> ; RV64IA-NEXT: ret
> @@ -726,24 +726,24 @@ define void @cmpxchg_i8_seq_cst_seq_cst(
> ;
> ; RV32IA-LABEL: cmpxchg_i8_seq_cst_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a4, a4, a3
> -; RV32IA-NEXT: andi a2, a2, 255
> -; RV32IA-NEXT: sll a2, a2, a3
> +; RV32IA-NEXT: sll a4, a4, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: andi a2, a2, 255
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB9_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a3, (a0)
> -; RV32IA-NEXT: and a5, a3, a4
> +; RV32IA-NEXT: lr.w.aqrl a2, (a3)
> +; RV32IA-NEXT: and a5, a2, a4
> ; RV32IA-NEXT: bne a5, a1, .LBB9_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB9_1 Depth=1
> -; RV32IA-NEXT: xor a5, a3, a2
> +; RV32IA-NEXT: xor a5, a2, a0
> ; RV32IA-NEXT: and a5, a5, a4
> -; RV32IA-NEXT: xor a5, a3, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a2, a5
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a3)
> ; RV32IA-NEXT: bnez a5, .LBB9_1
> ; RV32IA-NEXT: .LBB9_3:
> ; RV32IA-NEXT: ret
> @@ -763,24 +763,24 @@ define void @cmpxchg_i8_seq_cst_seq_cst(
> ;
> ; RV64IA-LABEL: cmpxchg_i8_seq_cst_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a4, a4, a3
> -; RV64IA-NEXT: andi a2, a2, 255
> -; RV64IA-NEXT: sllw a2, a2, a3
> +; RV64IA-NEXT: sllw a4, a4, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: andi a2, a2, 255
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB9_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a3, (a0)
> -; RV64IA-NEXT: and a5, a3, a4
> +; RV64IA-NEXT: lr.w.aqrl a2, (a3)
> +; RV64IA-NEXT: and a5, a2, a4
> ; RV64IA-NEXT: bne a5, a1, .LBB9_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB9_1 Depth=1
> -; RV64IA-NEXT: xor a5, a3, a2
> +; RV64IA-NEXT: xor a5, a2, a0
> ; RV64IA-NEXT: and a5, a5, a4
> -; RV64IA-NEXT: xor a5, a3, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a2, a5
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a3)
> ; RV64IA-NEXT: bnez a5, .LBB9_1
> ; RV64IA-NEXT: .LBB9_3:
> ; RV64IA-NEXT: ret
> @@ -804,26 +804,26 @@ define void @cmpxchg_i16_monotonic_monot
> ;
> ; RV32IA-LABEL: cmpxchg_i16_monotonic_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a3, 16
> -; RV32IA-NEXT: addi a3, a3, -1
> -; RV32IA-NEXT: and a1, a1, a3
> -; RV32IA-NEXT: and a2, a2, a3
> -; RV32IA-NEXT: slli a4, a0, 3
> -; RV32IA-NEXT: andi a4, a4, 24
> -; RV32IA-NEXT: sll a3, a3, a4
> -; RV32IA-NEXT: sll a2, a2, a4
> -; RV32IA-NEXT: sll a1, a1, a4
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a4, 16
> +; RV32IA-NEXT: addi a4, a4, -1
> +; RV32IA-NEXT: sll a5, a4, a0
> +; RV32IA-NEXT: and a1, a1, a4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: and a2, a2, a4
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB10_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a3
> -; RV32IA-NEXT: bne a5, a1, .LBB10_3
> +; RV32IA-NEXT: lr.w a2, (a3)
> +; RV32IA-NEXT: and a4, a2, a5
> +; RV32IA-NEXT: bne a4, a1, .LBB10_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB10_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a2
> -; RV32IA-NEXT: and a5, a5, a3
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> -; RV32IA-NEXT: bnez a5, .LBB10_1
> +; RV32IA-NEXT: xor a4, a2, a0
> +; RV32IA-NEXT: and a4, a4, a5
> +; RV32IA-NEXT: xor a4, a2, a4
> +; RV32IA-NEXT: sc.w a4, a4, (a3)
> +; RV32IA-NEXT: bnez a4, .LBB10_1
> ; RV32IA-NEXT: .LBB10_3:
> ; RV32IA-NEXT: ret
> ;
> @@ -842,26 +842,26 @@ define void @cmpxchg_i16_monotonic_monot
> ;
> ; RV64IA-LABEL: cmpxchg_i16_monotonic_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a3, 16
> -; RV64IA-NEXT: addiw a3, a3, -1
> -; RV64IA-NEXT: and a1, a1, a3
> -; RV64IA-NEXT: and a2, a2, a3
> -; RV64IA-NEXT: slli a4, a0, 3
> -; RV64IA-NEXT: andi a4, a4, 24
> -; RV64IA-NEXT: sllw a3, a3, a4
> -; RV64IA-NEXT: sllw a2, a2, a4
> -; RV64IA-NEXT: sllw a1, a1, a4
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a4, 16
> +; RV64IA-NEXT: addiw a4, a4, -1
> +; RV64IA-NEXT: sllw a5, a4, a0
> +; RV64IA-NEXT: and a1, a1, a4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: and a2, a2, a4
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB10_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a3
> -; RV64IA-NEXT: bne a5, a1, .LBB10_3
> +; RV64IA-NEXT: lr.w a2, (a3)
> +; RV64IA-NEXT: and a4, a2, a5
> +; RV64IA-NEXT: bne a4, a1, .LBB10_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB10_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a2
> -; RV64IA-NEXT: and a5, a5, a3
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> -; RV64IA-NEXT: bnez a5, .LBB10_1
> +; RV64IA-NEXT: xor a4, a2, a0
> +; RV64IA-NEXT: and a4, a4, a5
> +; RV64IA-NEXT: xor a4, a2, a4
> +; RV64IA-NEXT: sc.w a4, a4, (a3)
> +; RV64IA-NEXT: bnez a4, .LBB10_1
> ; RV64IA-NEXT: .LBB10_3:
> ; RV64IA-NEXT: ret
> %res = cmpxchg i16* %ptr, i16 %cmp, i16 %val monotonic monotonic
> @@ -884,26 +884,26 @@ define void @cmpxchg_i16_acquire_monoton
> ;
> ; RV32IA-LABEL: cmpxchg_i16_acquire_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a3, 16
> -; RV32IA-NEXT: addi a3, a3, -1
> -; RV32IA-NEXT: and a1, a1, a3
> -; RV32IA-NEXT: and a2, a2, a3
> -; RV32IA-NEXT: slli a4, a0, 3
> -; RV32IA-NEXT: andi a4, a4, 24
> -; RV32IA-NEXT: sll a3, a3, a4
> -; RV32IA-NEXT: sll a2, a2, a4
> -; RV32IA-NEXT: sll a1, a1, a4
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a4, 16
> +; RV32IA-NEXT: addi a4, a4, -1
> +; RV32IA-NEXT: sll a5, a4, a0
> +; RV32IA-NEXT: and a1, a1, a4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: and a2, a2, a4
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB11_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a3
> -; RV32IA-NEXT: bne a5, a1, .LBB11_3
> +; RV32IA-NEXT: lr.w.aq a2, (a3)
> +; RV32IA-NEXT: and a4, a2, a5
> +; RV32IA-NEXT: bne a4, a1, .LBB11_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB11_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a2
> -; RV32IA-NEXT: and a5, a5, a3
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> -; RV32IA-NEXT: bnez a5, .LBB11_1
> +; RV32IA-NEXT: xor a4, a2, a0
> +; RV32IA-NEXT: and a4, a4, a5
> +; RV32IA-NEXT: xor a4, a2, a4
> +; RV32IA-NEXT: sc.w a4, a4, (a3)
> +; RV32IA-NEXT: bnez a4, .LBB11_1
> ; RV32IA-NEXT: .LBB11_3:
> ; RV32IA-NEXT: ret
> ;
> @@ -922,26 +922,26 @@ define void @cmpxchg_i16_acquire_monoton
> ;
> ; RV64IA-LABEL: cmpxchg_i16_acquire_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a3, 16
> -; RV64IA-NEXT: addiw a3, a3, -1
> -; RV64IA-NEXT: and a1, a1, a3
> -; RV64IA-NEXT: and a2, a2, a3
> -; RV64IA-NEXT: slli a4, a0, 3
> -; RV64IA-NEXT: andi a4, a4, 24
> -; RV64IA-NEXT: sllw a3, a3, a4
> -; RV64IA-NEXT: sllw a2, a2, a4
> -; RV64IA-NEXT: sllw a1, a1, a4
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a4, 16
> +; RV64IA-NEXT: addiw a4, a4, -1
> +; RV64IA-NEXT: sllw a5, a4, a0
> +; RV64IA-NEXT: and a1, a1, a4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: and a2, a2, a4
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB11_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a3
> -; RV64IA-NEXT: bne a5, a1, .LBB11_3
> +; RV64IA-NEXT: lr.w.aq a2, (a3)
> +; RV64IA-NEXT: and a4, a2, a5
> +; RV64IA-NEXT: bne a4, a1, .LBB11_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB11_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a2
> -; RV64IA-NEXT: and a5, a5, a3
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> -; RV64IA-NEXT: bnez a5, .LBB11_1
> +; RV64IA-NEXT: xor a4, a2, a0
> +; RV64IA-NEXT: and a4, a4, a5
> +; RV64IA-NEXT: xor a4, a2, a4
> +; RV64IA-NEXT: sc.w a4, a4, (a3)
> +; RV64IA-NEXT: bnez a4, .LBB11_1
> ; RV64IA-NEXT: .LBB11_3:
> ; RV64IA-NEXT: ret
> %res = cmpxchg i16* %ptr, i16 %cmp, i16 %val acquire monotonic
> @@ -964,26 +964,26 @@ define void @cmpxchg_i16_acquire_acquire
> ;
> ; RV32IA-LABEL: cmpxchg_i16_acquire_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a3, 16
> -; RV32IA-NEXT: addi a3, a3, -1
> -; RV32IA-NEXT: and a1, a1, a3
> -; RV32IA-NEXT: and a2, a2, a3
> -; RV32IA-NEXT: slli a4, a0, 3
> -; RV32IA-NEXT: andi a4, a4, 24
> -; RV32IA-NEXT: sll a3, a3, a4
> -; RV32IA-NEXT: sll a2, a2, a4
> -; RV32IA-NEXT: sll a1, a1, a4
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a4, 16
> +; RV32IA-NEXT: addi a4, a4, -1
> +; RV32IA-NEXT: sll a5, a4, a0
> +; RV32IA-NEXT: and a1, a1, a4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: and a2, a2, a4
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB12_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a3
> -; RV32IA-NEXT: bne a5, a1, .LBB12_3
> +; RV32IA-NEXT: lr.w.aq a2, (a3)
> +; RV32IA-NEXT: and a4, a2, a5
> +; RV32IA-NEXT: bne a4, a1, .LBB12_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB12_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a2
> -; RV32IA-NEXT: and a5, a5, a3
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> -; RV32IA-NEXT: bnez a5, .LBB12_1
> +; RV32IA-NEXT: xor a4, a2, a0
> +; RV32IA-NEXT: and a4, a4, a5
> +; RV32IA-NEXT: xor a4, a2, a4
> +; RV32IA-NEXT: sc.w a4, a4, (a3)
> +; RV32IA-NEXT: bnez a4, .LBB12_1
> ; RV32IA-NEXT: .LBB12_3:
> ; RV32IA-NEXT: ret
> ;
> @@ -1002,26 +1002,26 @@ define void @cmpxchg_i16_acquire_acquire
> ;
> ; RV64IA-LABEL: cmpxchg_i16_acquire_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a3, 16
> -; RV64IA-NEXT: addiw a3, a3, -1
> -; RV64IA-NEXT: and a1, a1, a3
> -; RV64IA-NEXT: and a2, a2, a3
> -; RV64IA-NEXT: slli a4, a0, 3
> -; RV64IA-NEXT: andi a4, a4, 24
> -; RV64IA-NEXT: sllw a3, a3, a4
> -; RV64IA-NEXT: sllw a2, a2, a4
> -; RV64IA-NEXT: sllw a1, a1, a4
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a4, 16
> +; RV64IA-NEXT: addiw a4, a4, -1
> +; RV64IA-NEXT: sllw a5, a4, a0
> +; RV64IA-NEXT: and a1, a1, a4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: and a2, a2, a4
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB12_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a3
> -; RV64IA-NEXT: bne a5, a1, .LBB12_3
> +; RV64IA-NEXT: lr.w.aq a2, (a3)
> +; RV64IA-NEXT: and a4, a2, a5
> +; RV64IA-NEXT: bne a4, a1, .LBB12_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB12_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a2
> -; RV64IA-NEXT: and a5, a5, a3
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> -; RV64IA-NEXT: bnez a5, .LBB12_1
> +; RV64IA-NEXT: xor a4, a2, a0
> +; RV64IA-NEXT: and a4, a4, a5
> +; RV64IA-NEXT: xor a4, a2, a4
> +; RV64IA-NEXT: sc.w a4, a4, (a3)
> +; RV64IA-NEXT: bnez a4, .LBB12_1
> ; RV64IA-NEXT: .LBB12_3:
> ; RV64IA-NEXT: ret
> %res = cmpxchg i16* %ptr, i16 %cmp, i16 %val acquire acquire
> @@ -1044,26 +1044,26 @@ define void @cmpxchg_i16_release_monoton
> ;
> ; RV32IA-LABEL: cmpxchg_i16_release_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a3, 16
> -; RV32IA-NEXT: addi a3, a3, -1
> -; RV32IA-NEXT: and a1, a1, a3
> -; RV32IA-NEXT: and a2, a2, a3
> -; RV32IA-NEXT: slli a4, a0, 3
> -; RV32IA-NEXT: andi a4, a4, 24
> -; RV32IA-NEXT: sll a3, a3, a4
> -; RV32IA-NEXT: sll a2, a2, a4
> -; RV32IA-NEXT: sll a1, a1, a4
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a4, 16
> +; RV32IA-NEXT: addi a4, a4, -1
> +; RV32IA-NEXT: sll a5, a4, a0
> +; RV32IA-NEXT: and a1, a1, a4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: and a2, a2, a4
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB13_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a3
> -; RV32IA-NEXT: bne a5, a1, .LBB13_3
> +; RV32IA-NEXT: lr.w a2, (a3)
> +; RV32IA-NEXT: and a4, a2, a5
> +; RV32IA-NEXT: bne a4, a1, .LBB13_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB13_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a2
> -; RV32IA-NEXT: and a5, a5, a3
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> -; RV32IA-NEXT: bnez a5, .LBB13_1
> +; RV32IA-NEXT: xor a4, a2, a0
> +; RV32IA-NEXT: and a4, a4, a5
> +; RV32IA-NEXT: xor a4, a2, a4
> +; RV32IA-NEXT: sc.w.rl a4, a4, (a3)
> +; RV32IA-NEXT: bnez a4, .LBB13_1
> ; RV32IA-NEXT: .LBB13_3:
> ; RV32IA-NEXT: ret
> ;
> @@ -1082,26 +1082,26 @@ define void @cmpxchg_i16_release_monoton
> ;
> ; RV64IA-LABEL: cmpxchg_i16_release_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a3, 16
> -; RV64IA-NEXT: addiw a3, a3, -1
> -; RV64IA-NEXT: and a1, a1, a3
> -; RV64IA-NEXT: and a2, a2, a3
> -; RV64IA-NEXT: slli a4, a0, 3
> -; RV64IA-NEXT: andi a4, a4, 24
> -; RV64IA-NEXT: sllw a3, a3, a4
> -; RV64IA-NEXT: sllw a2, a2, a4
> -; RV64IA-NEXT: sllw a1, a1, a4
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a4, 16
> +; RV64IA-NEXT: addiw a4, a4, -1
> +; RV64IA-NEXT: sllw a5, a4, a0
> +; RV64IA-NEXT: and a1, a1, a4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: and a2, a2, a4
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB13_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a3
> -; RV64IA-NEXT: bne a5, a1, .LBB13_3
> +; RV64IA-NEXT: lr.w a2, (a3)
> +; RV64IA-NEXT: and a4, a2, a5
> +; RV64IA-NEXT: bne a4, a1, .LBB13_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB13_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a2
> -; RV64IA-NEXT: and a5, a5, a3
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> -; RV64IA-NEXT: bnez a5, .LBB13_1
> +; RV64IA-NEXT: xor a4, a2, a0
> +; RV64IA-NEXT: and a4, a4, a5
> +; RV64IA-NEXT: xor a4, a2, a4
> +; RV64IA-NEXT: sc.w.rl a4, a4, (a3)
> +; RV64IA-NEXT: bnez a4, .LBB13_1
> ; RV64IA-NEXT: .LBB13_3:
> ; RV64IA-NEXT: ret
> %res = cmpxchg i16* %ptr, i16 %cmp, i16 %val release monotonic
> @@ -1124,26 +1124,26 @@ define void @cmpxchg_i16_release_acquire
> ;
> ; RV32IA-LABEL: cmpxchg_i16_release_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a3, 16
> -; RV32IA-NEXT: addi a3, a3, -1
> -; RV32IA-NEXT: and a1, a1, a3
> -; RV32IA-NEXT: and a2, a2, a3
> -; RV32IA-NEXT: slli a4, a0, 3
> -; RV32IA-NEXT: andi a4, a4, 24
> -; RV32IA-NEXT: sll a3, a3, a4
> -; RV32IA-NEXT: sll a2, a2, a4
> -; RV32IA-NEXT: sll a1, a1, a4
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a4, 16
> +; RV32IA-NEXT: addi a4, a4, -1
> +; RV32IA-NEXT: sll a5, a4, a0
> +; RV32IA-NEXT: and a1, a1, a4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: and a2, a2, a4
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB14_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a3
> -; RV32IA-NEXT: bne a5, a1, .LBB14_3
> +; RV32IA-NEXT: lr.w a2, (a3)
> +; RV32IA-NEXT: and a4, a2, a5
> +; RV32IA-NEXT: bne a4, a1, .LBB14_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB14_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a2
> -; RV32IA-NEXT: and a5, a5, a3
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> -; RV32IA-NEXT: bnez a5, .LBB14_1
> +; RV32IA-NEXT: xor a4, a2, a0
> +; RV32IA-NEXT: and a4, a4, a5
> +; RV32IA-NEXT: xor a4, a2, a4
> +; RV32IA-NEXT: sc.w.rl a4, a4, (a3)
> +; RV32IA-NEXT: bnez a4, .LBB14_1
> ; RV32IA-NEXT: .LBB14_3:
> ; RV32IA-NEXT: ret
> ;
> @@ -1162,26 +1162,26 @@ define void @cmpxchg_i16_release_acquire
> ;
> ; RV64IA-LABEL: cmpxchg_i16_release_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a3, 16
> -; RV64IA-NEXT: addiw a3, a3, -1
> -; RV64IA-NEXT: and a1, a1, a3
> -; RV64IA-NEXT: and a2, a2, a3
> -; RV64IA-NEXT: slli a4, a0, 3
> -; RV64IA-NEXT: andi a4, a4, 24
> -; RV64IA-NEXT: sllw a3, a3, a4
> -; RV64IA-NEXT: sllw a2, a2, a4
> -; RV64IA-NEXT: sllw a1, a1, a4
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a4, 16
> +; RV64IA-NEXT: addiw a4, a4, -1
> +; RV64IA-NEXT: sllw a5, a4, a0
> +; RV64IA-NEXT: and a1, a1, a4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: and a2, a2, a4
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB14_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a3
> -; RV64IA-NEXT: bne a5, a1, .LBB14_3
> +; RV64IA-NEXT: lr.w a2, (a3)
> +; RV64IA-NEXT: and a4, a2, a5
> +; RV64IA-NEXT: bne a4, a1, .LBB14_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB14_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a2
> -; RV64IA-NEXT: and a5, a5, a3
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> -; RV64IA-NEXT: bnez a5, .LBB14_1
> +; RV64IA-NEXT: xor a4, a2, a0
> +; RV64IA-NEXT: and a4, a4, a5
> +; RV64IA-NEXT: xor a4, a2, a4
> +; RV64IA-NEXT: sc.w.rl a4, a4, (a3)
> +; RV64IA-NEXT: bnez a4, .LBB14_1
> ; RV64IA-NEXT: .LBB14_3:
> ; RV64IA-NEXT: ret
> %res = cmpxchg i16* %ptr, i16 %cmp, i16 %val release acquire
> @@ -1204,26 +1204,26 @@ define void @cmpxchg_i16_acq_rel_monoton
> ;
> ; RV32IA-LABEL: cmpxchg_i16_acq_rel_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a3, 16
> -; RV32IA-NEXT: addi a3, a3, -1
> -; RV32IA-NEXT: and a1, a1, a3
> -; RV32IA-NEXT: and a2, a2, a3
> -; RV32IA-NEXT: slli a4, a0, 3
> -; RV32IA-NEXT: andi a4, a4, 24
> -; RV32IA-NEXT: sll a3, a3, a4
> -; RV32IA-NEXT: sll a2, a2, a4
> -; RV32IA-NEXT: sll a1, a1, a4
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a4, 16
> +; RV32IA-NEXT: addi a4, a4, -1
> +; RV32IA-NEXT: sll a5, a4, a0
> +; RV32IA-NEXT: and a1, a1, a4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: and a2, a2, a4
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB15_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a3
> -; RV32IA-NEXT: bne a5, a1, .LBB15_3
> +; RV32IA-NEXT: lr.w.aq a2, (a3)
> +; RV32IA-NEXT: and a4, a2, a5
> +; RV32IA-NEXT: bne a4, a1, .LBB15_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB15_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a2
> -; RV32IA-NEXT: and a5, a5, a3
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> -; RV32IA-NEXT: bnez a5, .LBB15_1
> +; RV32IA-NEXT: xor a4, a2, a0
> +; RV32IA-NEXT: and a4, a4, a5
> +; RV32IA-NEXT: xor a4, a2, a4
> +; RV32IA-NEXT: sc.w.rl a4, a4, (a3)
> +; RV32IA-NEXT: bnez a4, .LBB15_1
> ; RV32IA-NEXT: .LBB15_3:
> ; RV32IA-NEXT: ret
> ;
> @@ -1242,26 +1242,26 @@ define void @cmpxchg_i16_acq_rel_monoton
> ;
> ; RV64IA-LABEL: cmpxchg_i16_acq_rel_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a3, 16
> -; RV64IA-NEXT: addiw a3, a3, -1
> -; RV64IA-NEXT: and a1, a1, a3
> -; RV64IA-NEXT: and a2, a2, a3
> -; RV64IA-NEXT: slli a4, a0, 3
> -; RV64IA-NEXT: andi a4, a4, 24
> -; RV64IA-NEXT: sllw a3, a3, a4
> -; RV64IA-NEXT: sllw a2, a2, a4
> -; RV64IA-NEXT: sllw a1, a1, a4
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a4, 16
> +; RV64IA-NEXT: addiw a4, a4, -1
> +; RV64IA-NEXT: sllw a5, a4, a0
> +; RV64IA-NEXT: and a1, a1, a4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: and a2, a2, a4
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB15_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a3
> -; RV64IA-NEXT: bne a5, a1, .LBB15_3
> +; RV64IA-NEXT: lr.w.aq a2, (a3)
> +; RV64IA-NEXT: and a4, a2, a5
> +; RV64IA-NEXT: bne a4, a1, .LBB15_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB15_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a2
> -; RV64IA-NEXT: and a5, a5, a3
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> -; RV64IA-NEXT: bnez a5, .LBB15_1
> +; RV64IA-NEXT: xor a4, a2, a0
> +; RV64IA-NEXT: and a4, a4, a5
> +; RV64IA-NEXT: xor a4, a2, a4
> +; RV64IA-NEXT: sc.w.rl a4, a4, (a3)
> +; RV64IA-NEXT: bnez a4, .LBB15_1
> ; RV64IA-NEXT: .LBB15_3:
> ; RV64IA-NEXT: ret
> %res = cmpxchg i16* %ptr, i16 %cmp, i16 %val acq_rel monotonic
> @@ -1284,26 +1284,26 @@ define void @cmpxchg_i16_acq_rel_acquire
> ;
> ; RV32IA-LABEL: cmpxchg_i16_acq_rel_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a3, 16
> -; RV32IA-NEXT: addi a3, a3, -1
> -; RV32IA-NEXT: and a1, a1, a3
> -; RV32IA-NEXT: and a2, a2, a3
> -; RV32IA-NEXT: slli a4, a0, 3
> -; RV32IA-NEXT: andi a4, a4, 24
> -; RV32IA-NEXT: sll a3, a3, a4
> -; RV32IA-NEXT: sll a2, a2, a4
> -; RV32IA-NEXT: sll a1, a1, a4
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a4, 16
> +; RV32IA-NEXT: addi a4, a4, -1
> +; RV32IA-NEXT: sll a5, a4, a0
> +; RV32IA-NEXT: and a1, a1, a4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: and a2, a2, a4
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB16_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a3
> -; RV32IA-NEXT: bne a5, a1, .LBB16_3
> +; RV32IA-NEXT: lr.w.aq a2, (a3)
> +; RV32IA-NEXT: and a4, a2, a5
> +; RV32IA-NEXT: bne a4, a1, .LBB16_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB16_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a2
> -; RV32IA-NEXT: and a5, a5, a3
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> -; RV32IA-NEXT: bnez a5, .LBB16_1
> +; RV32IA-NEXT: xor a4, a2, a0
> +; RV32IA-NEXT: and a4, a4, a5
> +; RV32IA-NEXT: xor a4, a2, a4
> +; RV32IA-NEXT: sc.w.rl a4, a4, (a3)
> +; RV32IA-NEXT: bnez a4, .LBB16_1
> ; RV32IA-NEXT: .LBB16_3:
> ; RV32IA-NEXT: ret
> ;
> @@ -1322,26 +1322,26 @@ define void @cmpxchg_i16_acq_rel_acquire
> ;
> ; RV64IA-LABEL: cmpxchg_i16_acq_rel_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a3, 16
> -; RV64IA-NEXT: addiw a3, a3, -1
> -; RV64IA-NEXT: and a1, a1, a3
> -; RV64IA-NEXT: and a2, a2, a3
> -; RV64IA-NEXT: slli a4, a0, 3
> -; RV64IA-NEXT: andi a4, a4, 24
> -; RV64IA-NEXT: sllw a3, a3, a4
> -; RV64IA-NEXT: sllw a2, a2, a4
> -; RV64IA-NEXT: sllw a1, a1, a4
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a4, 16
> +; RV64IA-NEXT: addiw a4, a4, -1
> +; RV64IA-NEXT: sllw a5, a4, a0
> +; RV64IA-NEXT: and a1, a1, a4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: and a2, a2, a4
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB16_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a3
> -; RV64IA-NEXT: bne a5, a1, .LBB16_3
> +; RV64IA-NEXT: lr.w.aq a2, (a3)
> +; RV64IA-NEXT: and a4, a2, a5
> +; RV64IA-NEXT: bne a4, a1, .LBB16_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB16_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a2
> -; RV64IA-NEXT: and a5, a5, a3
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> -; RV64IA-NEXT: bnez a5, .LBB16_1
> +; RV64IA-NEXT: xor a4, a2, a0
> +; RV64IA-NEXT: and a4, a4, a5
> +; RV64IA-NEXT: xor a4, a2, a4
> +; RV64IA-NEXT: sc.w.rl a4, a4, (a3)
> +; RV64IA-NEXT: bnez a4, .LBB16_1
> ; RV64IA-NEXT: .LBB16_3:
> ; RV64IA-NEXT: ret
> %res = cmpxchg i16* %ptr, i16 %cmp, i16 %val acq_rel acquire
> @@ -1364,26 +1364,26 @@ define void @cmpxchg_i16_seq_cst_monoton
> ;
> ; RV32IA-LABEL: cmpxchg_i16_seq_cst_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a3, 16
> -; RV32IA-NEXT: addi a3, a3, -1
> -; RV32IA-NEXT: and a1, a1, a3
> -; RV32IA-NEXT: and a2, a2, a3
> -; RV32IA-NEXT: slli a4, a0, 3
> -; RV32IA-NEXT: andi a4, a4, 24
> -; RV32IA-NEXT: sll a3, a3, a4
> -; RV32IA-NEXT: sll a2, a2, a4
> -; RV32IA-NEXT: sll a1, a1, a4
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a4, 16
> +; RV32IA-NEXT: addi a4, a4, -1
> +; RV32IA-NEXT: sll a5, a4, a0
> +; RV32IA-NEXT: and a1, a1, a4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: and a2, a2, a4
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB17_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a3
> -; RV32IA-NEXT: bne a5, a1, .LBB17_3
> +; RV32IA-NEXT: lr.w.aqrl a2, (a3)
> +; RV32IA-NEXT: and a4, a2, a5
> +; RV32IA-NEXT: bne a4, a1, .LBB17_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB17_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a2
> -; RV32IA-NEXT: and a5, a5, a3
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> -; RV32IA-NEXT: bnez a5, .LBB17_1
> +; RV32IA-NEXT: xor a4, a2, a0
> +; RV32IA-NEXT: and a4, a4, a5
> +; RV32IA-NEXT: xor a4, a2, a4
> +; RV32IA-NEXT: sc.w.aqrl a4, a4, (a3)
> +; RV32IA-NEXT: bnez a4, .LBB17_1
> ; RV32IA-NEXT: .LBB17_3:
> ; RV32IA-NEXT: ret
> ;
> @@ -1402,26 +1402,26 @@ define void @cmpxchg_i16_seq_cst_monoton
> ;
> ; RV64IA-LABEL: cmpxchg_i16_seq_cst_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a3, 16
> -; RV64IA-NEXT: addiw a3, a3, -1
> -; RV64IA-NEXT: and a1, a1, a3
> -; RV64IA-NEXT: and a2, a2, a3
> -; RV64IA-NEXT: slli a4, a0, 3
> -; RV64IA-NEXT: andi a4, a4, 24
> -; RV64IA-NEXT: sllw a3, a3, a4
> -; RV64IA-NEXT: sllw a2, a2, a4
> -; RV64IA-NEXT: sllw a1, a1, a4
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a4, 16
> +; RV64IA-NEXT: addiw a4, a4, -1
> +; RV64IA-NEXT: sllw a5, a4, a0
> +; RV64IA-NEXT: and a1, a1, a4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: and a2, a2, a4
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB17_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a3
> -; RV64IA-NEXT: bne a5, a1, .LBB17_3
> +; RV64IA-NEXT: lr.w.aqrl a2, (a3)
> +; RV64IA-NEXT: and a4, a2, a5
> +; RV64IA-NEXT: bne a4, a1, .LBB17_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB17_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a2
> -; RV64IA-NEXT: and a5, a5, a3
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> -; RV64IA-NEXT: bnez a5, .LBB17_1
> +; RV64IA-NEXT: xor a4, a2, a0
> +; RV64IA-NEXT: and a4, a4, a5
> +; RV64IA-NEXT: xor a4, a2, a4
> +; RV64IA-NEXT: sc.w.aqrl a4, a4, (a3)
> +; RV64IA-NEXT: bnez a4, .LBB17_1
> ; RV64IA-NEXT: .LBB17_3:
> ; RV64IA-NEXT: ret
> %res = cmpxchg i16* %ptr, i16 %cmp, i16 %val seq_cst monotonic
> @@ -1444,26 +1444,26 @@ define void @cmpxchg_i16_seq_cst_acquire
> ;
> ; RV32IA-LABEL: cmpxchg_i16_seq_cst_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a3, 16
> -; RV32IA-NEXT: addi a3, a3, -1
> -; RV32IA-NEXT: and a1, a1, a3
> -; RV32IA-NEXT: and a2, a2, a3
> -; RV32IA-NEXT: slli a4, a0, 3
> -; RV32IA-NEXT: andi a4, a4, 24
> -; RV32IA-NEXT: sll a3, a3, a4
> -; RV32IA-NEXT: sll a2, a2, a4
> -; RV32IA-NEXT: sll a1, a1, a4
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a4, 16
> +; RV32IA-NEXT: addi a4, a4, -1
> +; RV32IA-NEXT: sll a5, a4, a0
> +; RV32IA-NEXT: and a1, a1, a4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: and a2, a2, a4
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB18_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a3
> -; RV32IA-NEXT: bne a5, a1, .LBB18_3
> +; RV32IA-NEXT: lr.w.aqrl a2, (a3)
> +; RV32IA-NEXT: and a4, a2, a5
> +; RV32IA-NEXT: bne a4, a1, .LBB18_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB18_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a2
> -; RV32IA-NEXT: and a5, a5, a3
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> -; RV32IA-NEXT: bnez a5, .LBB18_1
> +; RV32IA-NEXT: xor a4, a2, a0
> +; RV32IA-NEXT: and a4, a4, a5
> +; RV32IA-NEXT: xor a4, a2, a4
> +; RV32IA-NEXT: sc.w.aqrl a4, a4, (a3)
> +; RV32IA-NEXT: bnez a4, .LBB18_1
> ; RV32IA-NEXT: .LBB18_3:
> ; RV32IA-NEXT: ret
> ;
> @@ -1482,26 +1482,26 @@ define void @cmpxchg_i16_seq_cst_acquire
> ;
> ; RV64IA-LABEL: cmpxchg_i16_seq_cst_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a3, 16
> -; RV64IA-NEXT: addiw a3, a3, -1
> -; RV64IA-NEXT: and a1, a1, a3
> -; RV64IA-NEXT: and a2, a2, a3
> -; RV64IA-NEXT: slli a4, a0, 3
> -; RV64IA-NEXT: andi a4, a4, 24
> -; RV64IA-NEXT: sllw a3, a3, a4
> -; RV64IA-NEXT: sllw a2, a2, a4
> -; RV64IA-NEXT: sllw a1, a1, a4
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a4, 16
> +; RV64IA-NEXT: addiw a4, a4, -1
> +; RV64IA-NEXT: sllw a5, a4, a0
> +; RV64IA-NEXT: and a1, a1, a4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: and a2, a2, a4
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB18_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a3
> -; RV64IA-NEXT: bne a5, a1, .LBB18_3
> +; RV64IA-NEXT: lr.w.aqrl a2, (a3)
> +; RV64IA-NEXT: and a4, a2, a5
> +; RV64IA-NEXT: bne a4, a1, .LBB18_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB18_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a2
> -; RV64IA-NEXT: and a5, a5, a3
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> -; RV64IA-NEXT: bnez a5, .LBB18_1
> +; RV64IA-NEXT: xor a4, a2, a0
> +; RV64IA-NEXT: and a4, a4, a5
> +; RV64IA-NEXT: xor a4, a2, a4
> +; RV64IA-NEXT: sc.w.aqrl a4, a4, (a3)
> +; RV64IA-NEXT: bnez a4, .LBB18_1
> ; RV64IA-NEXT: .LBB18_3:
> ; RV64IA-NEXT: ret
> %res = cmpxchg i16* %ptr, i16 %cmp, i16 %val seq_cst acquire
> @@ -1524,26 +1524,26 @@ define void @cmpxchg_i16_seq_cst_seq_cst
> ;
> ; RV32IA-LABEL: cmpxchg_i16_seq_cst_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a3, 16
> -; RV32IA-NEXT: addi a3, a3, -1
> -; RV32IA-NEXT: and a1, a1, a3
> -; RV32IA-NEXT: and a2, a2, a3
> -; RV32IA-NEXT: slli a4, a0, 3
> -; RV32IA-NEXT: andi a4, a4, 24
> -; RV32IA-NEXT: sll a3, a3, a4
> -; RV32IA-NEXT: sll a2, a2, a4
> -; RV32IA-NEXT: sll a1, a1, a4
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a3, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a4, 16
> +; RV32IA-NEXT: addi a4, a4, -1
> +; RV32IA-NEXT: sll a5, a4, a0
> +; RV32IA-NEXT: and a1, a1, a4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: and a2, a2, a4
> +; RV32IA-NEXT: sll a0, a2, a0
> ; RV32IA-NEXT: .LBB19_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a3
> -; RV32IA-NEXT: bne a5, a1, .LBB19_3
> +; RV32IA-NEXT: lr.w.aqrl a2, (a3)
> +; RV32IA-NEXT: and a4, a2, a5
> +; RV32IA-NEXT: bne a4, a1, .LBB19_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB19_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a2
> -; RV32IA-NEXT: and a5, a5, a3
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> -; RV32IA-NEXT: bnez a5, .LBB19_1
> +; RV32IA-NEXT: xor a4, a2, a0
> +; RV32IA-NEXT: and a4, a4, a5
> +; RV32IA-NEXT: xor a4, a2, a4
> +; RV32IA-NEXT: sc.w.aqrl a4, a4, (a3)
> +; RV32IA-NEXT: bnez a4, .LBB19_1
> ; RV32IA-NEXT: .LBB19_3:
> ; RV32IA-NEXT: ret
> ;
> @@ -1562,26 +1562,26 @@ define void @cmpxchg_i16_seq_cst_seq_cst
> ;
> ; RV64IA-LABEL: cmpxchg_i16_seq_cst_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a3, 16
> -; RV64IA-NEXT: addiw a3, a3, -1
> -; RV64IA-NEXT: and a1, a1, a3
> -; RV64IA-NEXT: and a2, a2, a3
> -; RV64IA-NEXT: slli a4, a0, 3
> -; RV64IA-NEXT: andi a4, a4, 24
> -; RV64IA-NEXT: sllw a3, a3, a4
> -; RV64IA-NEXT: sllw a2, a2, a4
> -; RV64IA-NEXT: sllw a1, a1, a4
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a3, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a4, 16
> +; RV64IA-NEXT: addiw a4, a4, -1
> +; RV64IA-NEXT: sllw a5, a4, a0
> +; RV64IA-NEXT: and a1, a1, a4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: and a2, a2, a4
> +; RV64IA-NEXT: sllw a0, a2, a0
> ; RV64IA-NEXT: .LBB19_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a3
> -; RV64IA-NEXT: bne a5, a1, .LBB19_3
> +; RV64IA-NEXT: lr.w.aqrl a2, (a3)
> +; RV64IA-NEXT: and a4, a2, a5
> +; RV64IA-NEXT: bne a4, a1, .LBB19_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB19_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a2
> -; RV64IA-NEXT: and a5, a5, a3
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> -; RV64IA-NEXT: bnez a5, .LBB19_1
> +; RV64IA-NEXT: xor a4, a2, a0
> +; RV64IA-NEXT: and a4, a4, a5
> +; RV64IA-NEXT: xor a4, a2, a4
> +; RV64IA-NEXT: sc.w.aqrl a4, a4, (a3)
> +; RV64IA-NEXT: bnez a4, .LBB19_1
> ; RV64IA-NEXT: .LBB19_3:
> ; RV64IA-NEXT: ret
> %res = cmpxchg i16* %ptr, i16 %cmp, i16 %val seq_cst seq_cst
> @@ -2173,12 +2173,13 @@ define void @cmpxchg_i64_acquire_monoton
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi sp, sp, -16
> ; RV32I-NEXT: sw ra, 12(sp)
> +; RV32I-NEXT: mv a5, a4
> ; RV32I-NEXT: sw a2, 4(sp)
> ; RV32I-NEXT: sw a1, 0(sp)
> ; RV32I-NEXT: mv a1, sp
> +; RV32I-NEXT: addi a4, zero, 2
> ; RV32I-NEXT: mv a2, a3
> -; RV32I-NEXT: mv a3, a4
> -; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: mv a5, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> ; RV32I-NEXT: lw ra, 12(sp)
> @@ -2189,12 +2190,13 @@ define void @cmpxchg_i64_acquire_monoton
> ; RV32IA: # %bb.0:
> ; RV32IA-NEXT: addi sp, sp, -16
> ; RV32IA-NEXT: sw ra, 12(sp)
> +; RV32IA-NEXT: mv a5, a4
> ; RV32IA-NEXT: sw a2, 4(sp)
> ; RV32IA-NEXT: sw a1, 0(sp)
> ; RV32IA-NEXT: mv a1, sp
> -; RV32IA-NEXT: mv a2, a3
> -; RV32IA-NEXT: mv a3, a4
> ; RV32IA-NEXT: addi a4, zero, 2
> +; RV32IA-NEXT: mv a2, a3
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: mv a5, zero
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> ; RV32IA-NEXT: lw ra, 12(sp)
> @@ -2233,13 +2235,14 @@ define void @cmpxchg_i64_acquire_acquire
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi sp, sp, -16
> ; RV32I-NEXT: sw ra, 12(sp)
> +; RV32I-NEXT: mv a6, a4
> ; RV32I-NEXT: sw a2, 4(sp)
> ; RV32I-NEXT: sw a1, 0(sp)
> ; RV32I-NEXT: mv a1, sp
> -; RV32I-NEXT: mv a2, a3
> -; RV32I-NEXT: mv a3, a4
> ; RV32I-NEXT: addi a4, zero, 2
> ; RV32I-NEXT: addi a5, zero, 2
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: mv a3, a6
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> ; RV32I-NEXT: lw ra, 12(sp)
> ; RV32I-NEXT: addi sp, sp, 16
> @@ -2249,13 +2252,14 @@ define void @cmpxchg_i64_acquire_acquire
> ; RV32IA: # %bb.0:
> ; RV32IA-NEXT: addi sp, sp, -16
> ; RV32IA-NEXT: sw ra, 12(sp)
> +; RV32IA-NEXT: mv a6, a4
> ; RV32IA-NEXT: sw a2, 4(sp)
> ; RV32IA-NEXT: sw a1, 0(sp)
> ; RV32IA-NEXT: mv a1, sp
> -; RV32IA-NEXT: mv a2, a3
> -; RV32IA-NEXT: mv a3, a4
> ; RV32IA-NEXT: addi a4, zero, 2
> ; RV32IA-NEXT: addi a5, zero, 2
> +; RV32IA-NEXT: mv a2, a3
> +; RV32IA-NEXT: mv a3, a6
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> ; RV32IA-NEXT: lw ra, 12(sp)
> ; RV32IA-NEXT: addi sp, sp, 16
> @@ -2293,12 +2297,13 @@ define void @cmpxchg_i64_release_monoton
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi sp, sp, -16
> ; RV32I-NEXT: sw ra, 12(sp)
> +; RV32I-NEXT: mv a5, a4
> ; RV32I-NEXT: sw a2, 4(sp)
> ; RV32I-NEXT: sw a1, 0(sp)
> ; RV32I-NEXT: mv a1, sp
> -; RV32I-NEXT: mv a2, a3
> -; RV32I-NEXT: mv a3, a4
> ; RV32I-NEXT: addi a4, zero, 3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: mv a5, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> ; RV32I-NEXT: lw ra, 12(sp)
> @@ -2309,12 +2314,13 @@ define void @cmpxchg_i64_release_monoton
> ; RV32IA: # %bb.0:
> ; RV32IA-NEXT: addi sp, sp, -16
> ; RV32IA-NEXT: sw ra, 12(sp)
> +; RV32IA-NEXT: mv a5, a4
> ; RV32IA-NEXT: sw a2, 4(sp)
> ; RV32IA-NEXT: sw a1, 0(sp)
> ; RV32IA-NEXT: mv a1, sp
> -; RV32IA-NEXT: mv a2, a3
> -; RV32IA-NEXT: mv a3, a4
> ; RV32IA-NEXT: addi a4, zero, 3
> +; RV32IA-NEXT: mv a2, a3
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: mv a5, zero
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> ; RV32IA-NEXT: lw ra, 12(sp)
> @@ -2353,13 +2359,14 @@ define void @cmpxchg_i64_release_acquire
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi sp, sp, -16
> ; RV32I-NEXT: sw ra, 12(sp)
> +; RV32I-NEXT: mv a6, a4
> ; RV32I-NEXT: sw a2, 4(sp)
> ; RV32I-NEXT: sw a1, 0(sp)
> ; RV32I-NEXT: mv a1, sp
> -; RV32I-NEXT: mv a2, a3
> -; RV32I-NEXT: mv a3, a4
> ; RV32I-NEXT: addi a4, zero, 3
> ; RV32I-NEXT: addi a5, zero, 2
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: mv a3, a6
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> ; RV32I-NEXT: lw ra, 12(sp)
> ; RV32I-NEXT: addi sp, sp, 16
> @@ -2369,13 +2376,14 @@ define void @cmpxchg_i64_release_acquire
> ; RV32IA: # %bb.0:
> ; RV32IA-NEXT: addi sp, sp, -16
> ; RV32IA-NEXT: sw ra, 12(sp)
> +; RV32IA-NEXT: mv a6, a4
> ; RV32IA-NEXT: sw a2, 4(sp)
> ; RV32IA-NEXT: sw a1, 0(sp)
> ; RV32IA-NEXT: mv a1, sp
> -; RV32IA-NEXT: mv a2, a3
> -; RV32IA-NEXT: mv a3, a4
> ; RV32IA-NEXT: addi a4, zero, 3
> ; RV32IA-NEXT: addi a5, zero, 2
> +; RV32IA-NEXT: mv a2, a3
> +; RV32IA-NEXT: mv a3, a6
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> ; RV32IA-NEXT: lw ra, 12(sp)
> ; RV32IA-NEXT: addi sp, sp, 16
> @@ -2413,12 +2421,13 @@ define void @cmpxchg_i64_acq_rel_monoton
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi sp, sp, -16
> ; RV32I-NEXT: sw ra, 12(sp)
> +; RV32I-NEXT: mv a5, a4
> ; RV32I-NEXT: sw a2, 4(sp)
> ; RV32I-NEXT: sw a1, 0(sp)
> ; RV32I-NEXT: mv a1, sp
> -; RV32I-NEXT: mv a2, a3
> -; RV32I-NEXT: mv a3, a4
> ; RV32I-NEXT: addi a4, zero, 4
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: mv a5, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> ; RV32I-NEXT: lw ra, 12(sp)
> @@ -2429,12 +2438,13 @@ define void @cmpxchg_i64_acq_rel_monoton
> ; RV32IA: # %bb.0:
> ; RV32IA-NEXT: addi sp, sp, -16
> ; RV32IA-NEXT: sw ra, 12(sp)
> +; RV32IA-NEXT: mv a5, a4
> ; RV32IA-NEXT: sw a2, 4(sp)
> ; RV32IA-NEXT: sw a1, 0(sp)
> ; RV32IA-NEXT: mv a1, sp
> -; RV32IA-NEXT: mv a2, a3
> -; RV32IA-NEXT: mv a3, a4
> ; RV32IA-NEXT: addi a4, zero, 4
> +; RV32IA-NEXT: mv a2, a3
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: mv a5, zero
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> ; RV32IA-NEXT: lw ra, 12(sp)
> @@ -2473,13 +2483,14 @@ define void @cmpxchg_i64_acq_rel_acquire
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi sp, sp, -16
> ; RV32I-NEXT: sw ra, 12(sp)
> +; RV32I-NEXT: mv a6, a4
> ; RV32I-NEXT: sw a2, 4(sp)
> ; RV32I-NEXT: sw a1, 0(sp)
> ; RV32I-NEXT: mv a1, sp
> -; RV32I-NEXT: mv a2, a3
> -; RV32I-NEXT: mv a3, a4
> ; RV32I-NEXT: addi a4, zero, 4
> ; RV32I-NEXT: addi a5, zero, 2
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: mv a3, a6
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> ; RV32I-NEXT: lw ra, 12(sp)
> ; RV32I-NEXT: addi sp, sp, 16
> @@ -2489,13 +2500,14 @@ define void @cmpxchg_i64_acq_rel_acquire
> ; RV32IA: # %bb.0:
> ; RV32IA-NEXT: addi sp, sp, -16
> ; RV32IA-NEXT: sw ra, 12(sp)
> +; RV32IA-NEXT: mv a6, a4
> ; RV32IA-NEXT: sw a2, 4(sp)
> ; RV32IA-NEXT: sw a1, 0(sp)
> ; RV32IA-NEXT: mv a1, sp
> -; RV32IA-NEXT: mv a2, a3
> -; RV32IA-NEXT: mv a3, a4
> ; RV32IA-NEXT: addi a4, zero, 4
> ; RV32IA-NEXT: addi a5, zero, 2
> +; RV32IA-NEXT: mv a2, a3
> +; RV32IA-NEXT: mv a3, a6
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> ; RV32IA-NEXT: lw ra, 12(sp)
> ; RV32IA-NEXT: addi sp, sp, 16
> @@ -2533,12 +2545,13 @@ define void @cmpxchg_i64_seq_cst_monoton
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi sp, sp, -16
> ; RV32I-NEXT: sw ra, 12(sp)
> +; RV32I-NEXT: mv a5, a4
> ; RV32I-NEXT: sw a2, 4(sp)
> ; RV32I-NEXT: sw a1, 0(sp)
> ; RV32I-NEXT: mv a1, sp
> -; RV32I-NEXT: mv a2, a3
> -; RV32I-NEXT: mv a3, a4
> ; RV32I-NEXT: addi a4, zero, 5
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: mv a5, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> ; RV32I-NEXT: lw ra, 12(sp)
> @@ -2549,12 +2562,13 @@ define void @cmpxchg_i64_seq_cst_monoton
> ; RV32IA: # %bb.0:
> ; RV32IA-NEXT: addi sp, sp, -16
> ; RV32IA-NEXT: sw ra, 12(sp)
> +; RV32IA-NEXT: mv a5, a4
> ; RV32IA-NEXT: sw a2, 4(sp)
> ; RV32IA-NEXT: sw a1, 0(sp)
> ; RV32IA-NEXT: mv a1, sp
> -; RV32IA-NEXT: mv a2, a3
> -; RV32IA-NEXT: mv a3, a4
> ; RV32IA-NEXT: addi a4, zero, 5
> +; RV32IA-NEXT: mv a2, a3
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: mv a5, zero
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> ; RV32IA-NEXT: lw ra, 12(sp)
> @@ -2593,13 +2607,14 @@ define void @cmpxchg_i64_seq_cst_acquire
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi sp, sp, -16
> ; RV32I-NEXT: sw ra, 12(sp)
> +; RV32I-NEXT: mv a6, a4
> ; RV32I-NEXT: sw a2, 4(sp)
> ; RV32I-NEXT: sw a1, 0(sp)
> ; RV32I-NEXT: mv a1, sp
> -; RV32I-NEXT: mv a2, a3
> -; RV32I-NEXT: mv a3, a4
> ; RV32I-NEXT: addi a4, zero, 5
> ; RV32I-NEXT: addi a5, zero, 2
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: mv a3, a6
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> ; RV32I-NEXT: lw ra, 12(sp)
> ; RV32I-NEXT: addi sp, sp, 16
> @@ -2609,13 +2624,14 @@ define void @cmpxchg_i64_seq_cst_acquire
> ; RV32IA: # %bb.0:
> ; RV32IA-NEXT: addi sp, sp, -16
> ; RV32IA-NEXT: sw ra, 12(sp)
> +; RV32IA-NEXT: mv a6, a4
> ; RV32IA-NEXT: sw a2, 4(sp)
> ; RV32IA-NEXT: sw a1, 0(sp)
> ; RV32IA-NEXT: mv a1, sp
> -; RV32IA-NEXT: mv a2, a3
> -; RV32IA-NEXT: mv a3, a4
> ; RV32IA-NEXT: addi a4, zero, 5
> ; RV32IA-NEXT: addi a5, zero, 2
> +; RV32IA-NEXT: mv a2, a3
> +; RV32IA-NEXT: mv a3, a6
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> ; RV32IA-NEXT: lw ra, 12(sp)
> ; RV32IA-NEXT: addi sp, sp, 16
> @@ -2653,13 +2669,14 @@ define void @cmpxchg_i64_seq_cst_seq_cst
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi sp, sp, -16
> ; RV32I-NEXT: sw ra, 12(sp)
> +; RV32I-NEXT: mv a6, a4
> ; RV32I-NEXT: sw a2, 4(sp)
> ; RV32I-NEXT: sw a1, 0(sp)
> ; RV32I-NEXT: mv a1, sp
> -; RV32I-NEXT: mv a2, a3
> -; RV32I-NEXT: mv a3, a4
> ; RV32I-NEXT: addi a4, zero, 5
> ; RV32I-NEXT: addi a5, zero, 5
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: mv a3, a6
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> ; RV32I-NEXT: lw ra, 12(sp)
> ; RV32I-NEXT: addi sp, sp, 16
> @@ -2669,13 +2686,14 @@ define void @cmpxchg_i64_seq_cst_seq_cst
> ; RV32IA: # %bb.0:
> ; RV32IA-NEXT: addi sp, sp, -16
> ; RV32IA-NEXT: sw ra, 12(sp)
> +; RV32IA-NEXT: mv a6, a4
> ; RV32IA-NEXT: sw a2, 4(sp)
> ; RV32IA-NEXT: sw a1, 0(sp)
> ; RV32IA-NEXT: mv a1, sp
> -; RV32IA-NEXT: mv a2, a3
> -; RV32IA-NEXT: mv a3, a4
> ; RV32IA-NEXT: addi a4, zero, 5
> ; RV32IA-NEXT: addi a5, zero, 5
> +; RV32IA-NEXT: mv a2, a3
> +; RV32IA-NEXT: mv a3, a6
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> ; RV32IA-NEXT: lw ra, 12(sp)
> ; RV32IA-NEXT: addi sp, sp, 16
>
> Modified: llvm/trunk/test/CodeGen/RISCV/atomic-rmw.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/atomic-rmw.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/atomic-rmw.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/atomic-rmw.ll Tue Sep 17 02:43:08 2019
> @@ -21,23 +21,23 @@ define i8 @atomicrmw_xchg_i8_monotonic(i
> ;
> ; RV32IA-LABEL: atomicrmw_xchg_i8_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB0_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> +; RV32IA-NEXT: lr.w a4, (a2)
> ; RV32IA-NEXT: add a5, zero, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB0_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xchg_i8_monotonic:
> @@ -52,23 +52,23 @@ define i8 @atomicrmw_xchg_i8_monotonic(i
> ;
> ; RV64IA-LABEL: atomicrmw_xchg_i8_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB0_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> +; RV64IA-NEXT: lr.w a4, (a2)
> ; RV64IA-NEXT: add a5, zero, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB0_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xchg i8* %a, i8 %b monotonic
> ret i8 %1
> @@ -87,23 +87,23 @@ define i8 @atomicrmw_xchg_i8_acquire(i8*
> ;
> ; RV32IA-LABEL: atomicrmw_xchg_i8_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB1_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> +; RV32IA-NEXT: lr.w.aq a4, (a2)
> ; RV32IA-NEXT: add a5, zero, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB1_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xchg_i8_acquire:
> @@ -118,23 +118,23 @@ define i8 @atomicrmw_xchg_i8_acquire(i8*
> ;
> ; RV64IA-LABEL: atomicrmw_xchg_i8_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB1_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> +; RV64IA-NEXT: lr.w.aq a4, (a2)
> ; RV64IA-NEXT: add a5, zero, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB1_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xchg i8* %a, i8 %b acquire
> ret i8 %1
> @@ -153,23 +153,23 @@ define i8 @atomicrmw_xchg_i8_release(i8*
> ;
> ; RV32IA-LABEL: atomicrmw_xchg_i8_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB2_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> +; RV32IA-NEXT: lr.w a4, (a2)
> ; RV32IA-NEXT: add a5, zero, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB2_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xchg_i8_release:
> @@ -184,23 +184,23 @@ define i8 @atomicrmw_xchg_i8_release(i8*
> ;
> ; RV64IA-LABEL: atomicrmw_xchg_i8_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB2_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> +; RV64IA-NEXT: lr.w a4, (a2)
> ; RV64IA-NEXT: add a5, zero, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB2_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xchg i8* %a, i8 %b release
> ret i8 %1
> @@ -219,23 +219,23 @@ define i8 @atomicrmw_xchg_i8_acq_rel(i8*
> ;
> ; RV32IA-LABEL: atomicrmw_xchg_i8_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB3_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> +; RV32IA-NEXT: lr.w.aq a4, (a2)
> ; RV32IA-NEXT: add a5, zero, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB3_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xchg_i8_acq_rel:
> @@ -250,23 +250,23 @@ define i8 @atomicrmw_xchg_i8_acq_rel(i8*
> ;
> ; RV64IA-LABEL: atomicrmw_xchg_i8_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB3_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> +; RV64IA-NEXT: lr.w.aq a4, (a2)
> ; RV64IA-NEXT: add a5, zero, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB3_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xchg i8* %a, i8 %b acq_rel
> ret i8 %1
> @@ -285,23 +285,23 @@ define i8 @atomicrmw_xchg_i8_seq_cst(i8*
> ;
> ; RV32IA-LABEL: atomicrmw_xchg_i8_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB4_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> +; RV32IA-NEXT: lr.w.aqrl a4, (a2)
> ; RV32IA-NEXT: add a5, zero, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB4_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xchg_i8_seq_cst:
> @@ -316,23 +316,23 @@ define i8 @atomicrmw_xchg_i8_seq_cst(i8*
> ;
> ; RV64IA-LABEL: atomicrmw_xchg_i8_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB4_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> +; RV64IA-NEXT: lr.w.aqrl a4, (a2)
> ; RV64IA-NEXT: add a5, zero, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB4_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xchg i8* %a, i8 %b seq_cst
> ret i8 %1
> @@ -351,23 +351,23 @@ define i8 @atomicrmw_add_i8_monotonic(i8
> ;
> ; RV32IA-LABEL: atomicrmw_add_i8_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB5_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> +; RV32IA-NEXT: lr.w a4, (a2)
> ; RV32IA-NEXT: add a5, a4, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB5_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_add_i8_monotonic:
> @@ -382,23 +382,23 @@ define i8 @atomicrmw_add_i8_monotonic(i8
> ;
> ; RV64IA-LABEL: atomicrmw_add_i8_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB5_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> +; RV64IA-NEXT: lr.w a4, (a2)
> ; RV64IA-NEXT: add a5, a4, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB5_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw add i8* %a, i8 %b monotonic
> ret i8 %1
> @@ -417,23 +417,23 @@ define i8 @atomicrmw_add_i8_acquire(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_add_i8_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB6_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> +; RV32IA-NEXT: lr.w.aq a4, (a2)
> ; RV32IA-NEXT: add a5, a4, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB6_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_add_i8_acquire:
> @@ -448,23 +448,23 @@ define i8 @atomicrmw_add_i8_acquire(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_add_i8_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB6_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> +; RV64IA-NEXT: lr.w.aq a4, (a2)
> ; RV64IA-NEXT: add a5, a4, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB6_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw add i8* %a, i8 %b acquire
> ret i8 %1
> @@ -483,23 +483,23 @@ define i8 @atomicrmw_add_i8_release(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_add_i8_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB7_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> +; RV32IA-NEXT: lr.w a4, (a2)
> ; RV32IA-NEXT: add a5, a4, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB7_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_add_i8_release:
> @@ -514,23 +514,23 @@ define i8 @atomicrmw_add_i8_release(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_add_i8_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB7_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> +; RV64IA-NEXT: lr.w a4, (a2)
> ; RV64IA-NEXT: add a5, a4, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB7_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw add i8* %a, i8 %b release
> ret i8 %1
> @@ -549,23 +549,23 @@ define i8 @atomicrmw_add_i8_acq_rel(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_add_i8_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB8_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> +; RV32IA-NEXT: lr.w.aq a4, (a2)
> ; RV32IA-NEXT: add a5, a4, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB8_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_add_i8_acq_rel:
> @@ -580,23 +580,23 @@ define i8 @atomicrmw_add_i8_acq_rel(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_add_i8_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB8_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> +; RV64IA-NEXT: lr.w.aq a4, (a2)
> ; RV64IA-NEXT: add a5, a4, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB8_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw add i8* %a, i8 %b acq_rel
> ret i8 %1
> @@ -615,23 +615,23 @@ define i8 @atomicrmw_add_i8_seq_cst(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_add_i8_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB9_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> +; RV32IA-NEXT: lr.w.aqrl a4, (a2)
> ; RV32IA-NEXT: add a5, a4, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB9_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_add_i8_seq_cst:
> @@ -646,23 +646,23 @@ define i8 @atomicrmw_add_i8_seq_cst(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_add_i8_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB9_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> +; RV64IA-NEXT: lr.w.aqrl a4, (a2)
> ; RV64IA-NEXT: add a5, a4, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB9_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw add i8* %a, i8 %b seq_cst
> ret i8 %1
> @@ -681,23 +681,23 @@ define i8 @atomicrmw_sub_i8_monotonic(i8
> ;
> ; RV32IA-LABEL: atomicrmw_sub_i8_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB10_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> +; RV32IA-NEXT: lr.w a4, (a2)
> ; RV32IA-NEXT: sub a5, a4, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB10_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_sub_i8_monotonic:
> @@ -712,23 +712,23 @@ define i8 @atomicrmw_sub_i8_monotonic(i8
> ;
> ; RV64IA-LABEL: atomicrmw_sub_i8_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB10_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> +; RV64IA-NEXT: lr.w a4, (a2)
> ; RV64IA-NEXT: sub a5, a4, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB10_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw sub i8* %a, i8 %b monotonic
> ret i8 %1
> @@ -747,23 +747,23 @@ define i8 @atomicrmw_sub_i8_acquire(i8*
> ;
> ; RV32IA-LABEL: atomicrmw_sub_i8_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB11_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> +; RV32IA-NEXT: lr.w.aq a4, (a2)
> ; RV32IA-NEXT: sub a5, a4, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB11_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_sub_i8_acquire:
> @@ -778,23 +778,23 @@ define i8 @atomicrmw_sub_i8_acquire(i8*
> ;
> ; RV64IA-LABEL: atomicrmw_sub_i8_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB11_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> +; RV64IA-NEXT: lr.w.aq a4, (a2)
> ; RV64IA-NEXT: sub a5, a4, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB11_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw sub i8* %a, i8 %b acquire
> ret i8 %1
> @@ -813,23 +813,23 @@ define i8 @atomicrmw_sub_i8_release(i8*
> ;
> ; RV32IA-LABEL: atomicrmw_sub_i8_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB12_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> +; RV32IA-NEXT: lr.w a4, (a2)
> ; RV32IA-NEXT: sub a5, a4, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB12_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_sub_i8_release:
> @@ -844,23 +844,23 @@ define i8 @atomicrmw_sub_i8_release(i8*
> ;
> ; RV64IA-LABEL: atomicrmw_sub_i8_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB12_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> +; RV64IA-NEXT: lr.w a4, (a2)
> ; RV64IA-NEXT: sub a5, a4, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB12_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw sub i8* %a, i8 %b release
> ret i8 %1
> @@ -879,23 +879,23 @@ define i8 @atomicrmw_sub_i8_acq_rel(i8*
> ;
> ; RV32IA-LABEL: atomicrmw_sub_i8_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB13_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> +; RV32IA-NEXT: lr.w.aq a4, (a2)
> ; RV32IA-NEXT: sub a5, a4, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB13_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_sub_i8_acq_rel:
> @@ -910,23 +910,23 @@ define i8 @atomicrmw_sub_i8_acq_rel(i8*
> ;
> ; RV64IA-LABEL: atomicrmw_sub_i8_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB13_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> +; RV64IA-NEXT: lr.w.aq a4, (a2)
> ; RV64IA-NEXT: sub a5, a4, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB13_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw sub i8* %a, i8 %b acq_rel
> ret i8 %1
> @@ -945,23 +945,23 @@ define i8 @atomicrmw_sub_i8_seq_cst(i8*
> ;
> ; RV32IA-LABEL: atomicrmw_sub_i8_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB14_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> +; RV32IA-NEXT: lr.w.aqrl a4, (a2)
> ; RV32IA-NEXT: sub a5, a4, a1
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB14_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_sub_i8_seq_cst:
> @@ -976,23 +976,23 @@ define i8 @atomicrmw_sub_i8_seq_cst(i8*
> ;
> ; RV64IA-LABEL: atomicrmw_sub_i8_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB14_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> +; RV64IA-NEXT: lr.w.aqrl a4, (a2)
> ; RV64IA-NEXT: sub a5, a4, a1
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB14_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw sub i8* %a, i8 %b seq_cst
> ret i8 %1
> @@ -1011,17 +1011,17 @@ define i8 @atomicrmw_and_i8_monotonic(i8
> ;
> ; RV32IA-LABEL: atomicrmw_and_i8_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: not a3, a3
> +; RV32IA-NEXT: andi a1, a1, 255
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: or a1, a3, a1
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoand.w a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: amoand.w a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_and_i8_monotonic:
> @@ -1036,17 +1036,17 @@ define i8 @atomicrmw_and_i8_monotonic(i8
> ;
> ; RV64IA-LABEL: atomicrmw_and_i8_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: not a3, a3
> +; RV64IA-NEXT: andi a1, a1, 255
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: or a1, a3, a1
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoand.w a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: amoand.w a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw and i8* %a, i8 %b monotonic
> ret i8 %1
> @@ -1065,17 +1065,17 @@ define i8 @atomicrmw_and_i8_acquire(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_and_i8_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: not a3, a3
> +; RV32IA-NEXT: andi a1, a1, 255
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: or a1, a3, a1
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoand.w.aq a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: amoand.w.aq a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_and_i8_acquire:
> @@ -1090,17 +1090,17 @@ define i8 @atomicrmw_and_i8_acquire(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_and_i8_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: not a3, a3
> +; RV64IA-NEXT: andi a1, a1, 255
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: or a1, a3, a1
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoand.w.aq a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: amoand.w.aq a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw and i8* %a, i8 %b acquire
> ret i8 %1
> @@ -1119,17 +1119,17 @@ define i8 @atomicrmw_and_i8_release(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_and_i8_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: not a3, a3
> +; RV32IA-NEXT: andi a1, a1, 255
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: or a1, a3, a1
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoand.w.rl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: amoand.w.rl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_and_i8_release:
> @@ -1144,17 +1144,17 @@ define i8 @atomicrmw_and_i8_release(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_and_i8_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: not a3, a3
> +; RV64IA-NEXT: andi a1, a1, 255
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: or a1, a3, a1
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoand.w.rl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: amoand.w.rl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw and i8* %a, i8 %b release
> ret i8 %1
> @@ -1173,17 +1173,17 @@ define i8 @atomicrmw_and_i8_acq_rel(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_and_i8_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: not a3, a3
> +; RV32IA-NEXT: andi a1, a1, 255
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: or a1, a3, a1
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoand.w.aqrl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: amoand.w.aqrl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_and_i8_acq_rel:
> @@ -1198,17 +1198,17 @@ define i8 @atomicrmw_and_i8_acq_rel(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_and_i8_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: not a3, a3
> +; RV64IA-NEXT: andi a1, a1, 255
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: or a1, a3, a1
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoand.w.aqrl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: amoand.w.aqrl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw and i8* %a, i8 %b acq_rel
> ret i8 %1
> @@ -1227,17 +1227,17 @@ define i8 @atomicrmw_and_i8_seq_cst(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_and_i8_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: not a3, a3
> +; RV32IA-NEXT: andi a1, a1, 255
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: or a1, a3, a1
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoand.w.aqrl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: amoand.w.aqrl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_and_i8_seq_cst:
> @@ -1252,17 +1252,17 @@ define i8 @atomicrmw_and_i8_seq_cst(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_and_i8_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: not a3, a3
> +; RV64IA-NEXT: andi a1, a1, 255
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: or a1, a3, a1
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoand.w.aqrl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: amoand.w.aqrl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw and i8* %a, i8 %b seq_cst
> ret i8 %1
> @@ -1281,24 +1281,24 @@ define i8 @atomicrmw_nand_i8_monotonic(i
> ;
> ; RV32IA-LABEL: atomicrmw_nand_i8_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB20_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> +; RV32IA-NEXT: lr.w a4, (a2)
> ; RV32IA-NEXT: and a5, a4, a1
> ; RV32IA-NEXT: not a5, a5
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB20_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_nand_i8_monotonic:
> @@ -1313,24 +1313,24 @@ define i8 @atomicrmw_nand_i8_monotonic(i
> ;
> ; RV64IA-LABEL: atomicrmw_nand_i8_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB20_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> +; RV64IA-NEXT: lr.w a4, (a2)
> ; RV64IA-NEXT: and a5, a4, a1
> ; RV64IA-NEXT: not a5, a5
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB20_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw nand i8* %a, i8 %b monotonic
> ret i8 %1
> @@ -1349,24 +1349,24 @@ define i8 @atomicrmw_nand_i8_acquire(i8*
> ;
> ; RV32IA-LABEL: atomicrmw_nand_i8_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB21_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> +; RV32IA-NEXT: lr.w.aq a4, (a2)
> ; RV32IA-NEXT: and a5, a4, a1
> ; RV32IA-NEXT: not a5, a5
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB21_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_nand_i8_acquire:
> @@ -1381,24 +1381,24 @@ define i8 @atomicrmw_nand_i8_acquire(i8*
> ;
> ; RV64IA-LABEL: atomicrmw_nand_i8_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB21_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> +; RV64IA-NEXT: lr.w.aq a4, (a2)
> ; RV64IA-NEXT: and a5, a4, a1
> ; RV64IA-NEXT: not a5, a5
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB21_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw nand i8* %a, i8 %b acquire
> ret i8 %1
> @@ -1417,24 +1417,24 @@ define i8 @atomicrmw_nand_i8_release(i8*
> ;
> ; RV32IA-LABEL: atomicrmw_nand_i8_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB22_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> +; RV32IA-NEXT: lr.w a4, (a2)
> ; RV32IA-NEXT: and a5, a4, a1
> ; RV32IA-NEXT: not a5, a5
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB22_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_nand_i8_release:
> @@ -1449,24 +1449,24 @@ define i8 @atomicrmw_nand_i8_release(i8*
> ;
> ; RV64IA-LABEL: atomicrmw_nand_i8_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB22_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> +; RV64IA-NEXT: lr.w a4, (a2)
> ; RV64IA-NEXT: and a5, a4, a1
> ; RV64IA-NEXT: not a5, a5
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB22_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw nand i8* %a, i8 %b release
> ret i8 %1
> @@ -1485,24 +1485,24 @@ define i8 @atomicrmw_nand_i8_acq_rel(i8*
> ;
> ; RV32IA-LABEL: atomicrmw_nand_i8_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB23_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> +; RV32IA-NEXT: lr.w.aq a4, (a2)
> ; RV32IA-NEXT: and a5, a4, a1
> ; RV32IA-NEXT: not a5, a5
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB23_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_nand_i8_acq_rel:
> @@ -1517,24 +1517,24 @@ define i8 @atomicrmw_nand_i8_acq_rel(i8*
> ;
> ; RV64IA-LABEL: atomicrmw_nand_i8_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB23_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> +; RV64IA-NEXT: lr.w.aq a4, (a2)
> ; RV64IA-NEXT: and a5, a4, a1
> ; RV64IA-NEXT: not a5, a5
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB23_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw nand i8* %a, i8 %b acq_rel
> ret i8 %1
> @@ -1553,24 +1553,24 @@ define i8 @atomicrmw_nand_i8_seq_cst(i8*
> ;
> ; RV32IA-LABEL: atomicrmw_nand_i8_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a3, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB24_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> +; RV32IA-NEXT: lr.w.aqrl a4, (a2)
> ; RV32IA-NEXT: and a5, a4, a1
> ; RV32IA-NEXT: not a5, a5
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB24_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_nand_i8_seq_cst:
> @@ -1585,24 +1585,24 @@ define i8 @atomicrmw_nand_i8_seq_cst(i8*
> ;
> ; RV64IA-LABEL: atomicrmw_nand_i8_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a3, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB24_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> +; RV64IA-NEXT: lr.w.aqrl a4, (a2)
> ; RV64IA-NEXT: and a5, a4, a1
> ; RV64IA-NEXT: not a5, a5
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB24_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw nand i8* %a, i8 %b seq_cst
> ret i8 %1
> @@ -1621,13 +1621,13 @@ define i8 @atomicrmw_or_i8_monotonic(i8
> ;
> ; RV32IA-LABEL: atomicrmw_or_i8_monotonic:
> ; RV32IA: # %bb.0:
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoor.w a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoor.w a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_or_i8_monotonic:
> @@ -1642,13 +1642,13 @@ define i8 @atomicrmw_or_i8_monotonic(i8
> ;
> ; RV64IA-LABEL: atomicrmw_or_i8_monotonic:
> ; RV64IA: # %bb.0:
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoor.w a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoor.w a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw or i8* %a, i8 %b monotonic
> ret i8 %1
> @@ -1667,13 +1667,13 @@ define i8 @atomicrmw_or_i8_acquire(i8 *%
> ;
> ; RV32IA-LABEL: atomicrmw_or_i8_acquire:
> ; RV32IA: # %bb.0:
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoor.w.aq a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoor.w.aq a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_or_i8_acquire:
> @@ -1688,13 +1688,13 @@ define i8 @atomicrmw_or_i8_acquire(i8 *%
> ;
> ; RV64IA-LABEL: atomicrmw_or_i8_acquire:
> ; RV64IA: # %bb.0:
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoor.w.aq a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoor.w.aq a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw or i8* %a, i8 %b acquire
> ret i8 %1
> @@ -1713,13 +1713,13 @@ define i8 @atomicrmw_or_i8_release(i8 *%
> ;
> ; RV32IA-LABEL: atomicrmw_or_i8_release:
> ; RV32IA: # %bb.0:
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoor.w.rl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoor.w.rl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_or_i8_release:
> @@ -1734,13 +1734,13 @@ define i8 @atomicrmw_or_i8_release(i8 *%
> ;
> ; RV64IA-LABEL: atomicrmw_or_i8_release:
> ; RV64IA: # %bb.0:
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoor.w.rl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoor.w.rl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw or i8* %a, i8 %b release
> ret i8 %1
> @@ -1759,13 +1759,13 @@ define i8 @atomicrmw_or_i8_acq_rel(i8 *%
> ;
> ; RV32IA-LABEL: atomicrmw_or_i8_acq_rel:
> ; RV32IA: # %bb.0:
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoor.w.aqrl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoor.w.aqrl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_or_i8_acq_rel:
> @@ -1780,13 +1780,13 @@ define i8 @atomicrmw_or_i8_acq_rel(i8 *%
> ;
> ; RV64IA-LABEL: atomicrmw_or_i8_acq_rel:
> ; RV64IA: # %bb.0:
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoor.w.aqrl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoor.w.aqrl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw or i8* %a, i8 %b acq_rel
> ret i8 %1
> @@ -1805,13 +1805,13 @@ define i8 @atomicrmw_or_i8_seq_cst(i8 *%
> ;
> ; RV32IA-LABEL: atomicrmw_or_i8_seq_cst:
> ; RV32IA: # %bb.0:
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoor.w.aqrl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoor.w.aqrl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_or_i8_seq_cst:
> @@ -1826,13 +1826,13 @@ define i8 @atomicrmw_or_i8_seq_cst(i8 *%
> ;
> ; RV64IA-LABEL: atomicrmw_or_i8_seq_cst:
> ; RV64IA: # %bb.0:
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoor.w.aqrl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoor.w.aqrl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw or i8* %a, i8 %b seq_cst
> ret i8 %1
> @@ -1851,13 +1851,13 @@ define i8 @atomicrmw_xor_i8_monotonic(i8
> ;
> ; RV32IA-LABEL: atomicrmw_xor_i8_monotonic:
> ; RV32IA: # %bb.0:
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoxor.w a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoxor.w a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xor_i8_monotonic:
> @@ -1872,13 +1872,13 @@ define i8 @atomicrmw_xor_i8_monotonic(i8
> ;
> ; RV64IA-LABEL: atomicrmw_xor_i8_monotonic:
> ; RV64IA: # %bb.0:
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoxor.w a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoxor.w a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xor i8* %a, i8 %b monotonic
> ret i8 %1
> @@ -1897,13 +1897,13 @@ define i8 @atomicrmw_xor_i8_acquire(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_xor_i8_acquire:
> ; RV32IA: # %bb.0:
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoxor.w.aq a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoxor.w.aq a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xor_i8_acquire:
> @@ -1918,13 +1918,13 @@ define i8 @atomicrmw_xor_i8_acquire(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_xor_i8_acquire:
> ; RV64IA: # %bb.0:
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoxor.w.aq a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoxor.w.aq a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xor i8* %a, i8 %b acquire
> ret i8 %1
> @@ -1943,13 +1943,13 @@ define i8 @atomicrmw_xor_i8_release(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_xor_i8_release:
> ; RV32IA: # %bb.0:
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoxor.w.rl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoxor.w.rl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xor_i8_release:
> @@ -1964,13 +1964,13 @@ define i8 @atomicrmw_xor_i8_release(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_xor_i8_release:
> ; RV64IA: # %bb.0:
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoxor.w.rl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoxor.w.rl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xor i8* %a, i8 %b release
> ret i8 %1
> @@ -1989,13 +1989,13 @@ define i8 @atomicrmw_xor_i8_acq_rel(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_xor_i8_acq_rel:
> ; RV32IA: # %bb.0:
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoxor.w.aqrl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoxor.w.aqrl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xor_i8_acq_rel:
> @@ -2010,13 +2010,13 @@ define i8 @atomicrmw_xor_i8_acq_rel(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_xor_i8_acq_rel:
> ; RV64IA: # %bb.0:
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoxor.w.aqrl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoxor.w.aqrl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xor i8* %a, i8 %b acq_rel
> ret i8 %1
> @@ -2035,13 +2035,13 @@ define i8 @atomicrmw_xor_i8_seq_cst(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_xor_i8_seq_cst:
> ; RV32IA: # %bb.0:
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoxor.w.aqrl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoxor.w.aqrl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xor_i8_seq_cst:
> @@ -2056,13 +2056,13 @@ define i8 @atomicrmw_xor_i8_seq_cst(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_xor_i8_seq_cst:
> ; RV64IA: # %bb.0:
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoxor.w.aqrl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoxor.w.aqrl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xor i8* %a, i8 %b seq_cst
> ret i8 %1
> @@ -2079,33 +2079,33 @@ define i8 @atomicrmw_max_i8_monotonic(i8
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 24
> -; RV32I-NEXT: srai s0, a1, 24
> +; RV32I-NEXT: lbu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 24
> +; RV32I-NEXT: srai s0, a0, 24
> ; RV32I-NEXT: addi s3, sp, 11
> ; RV32I-NEXT: j .LBB35_2
> ; RV32I-NEXT: .LBB35_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB35_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> +; RV32I-NEXT: sb a1, 11(sp)
> ; RV32I-NEXT: mv a0, s1
> ; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a3, zero
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB35_4
> +; RV32I-NEXT: lb a1, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB35_4
> ; RV32I-NEXT: .LBB35_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 24
> -; RV32I-NEXT: srai a1, a1, 24
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: blt s0, a1, .LBB35_1
> +; RV32I-NEXT: slli a0, a1, 24
> +; RV32I-NEXT: srai a0, a0, 24
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: blt s0, a0, .LBB35_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB35_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB35_1
> ; RV32I-NEXT: .LBB35_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -2116,32 +2116,32 @@ define i8 @atomicrmw_max_i8_monotonic(i8
> ;
> ; RV32IA-LABEL: atomicrmw_max_i8_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 24
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: addi a3, zero, 255
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 24
> ; RV32IA-NEXT: srai a1, a1, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 24
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB35_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a4, a1, .LBB35_3
> +; RV32IA-NEXT: lr.w a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a3, a1, .LBB35_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB35_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB35_3: # in Loop: Header=BB35_1 Depth=1
> -; RV32IA-NEXT: sc.w a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB35_1
> +; RV32IA-NEXT: sc.w a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB35_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_max_i8_monotonic:
> @@ -2154,33 +2154,33 @@ define i8 @atomicrmw_max_i8_monotonic(i8
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 56
> -; RV64I-NEXT: srai s0, a1, 56
> +; RV64I-NEXT: lbu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 56
> +; RV64I-NEXT: srai s0, a0, 56
> ; RV64I-NEXT: addi s3, sp, 7
> ; RV64I-NEXT: j .LBB35_2
> ; RV64I-NEXT: .LBB35_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB35_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> +; RV64I-NEXT: sb a1, 7(sp)
> ; RV64I-NEXT: mv a0, s1
> ; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB35_4
> +; RV64I-NEXT: lb a1, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB35_4
> ; RV64I-NEXT: .LBB35_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 56
> -; RV64I-NEXT: srai a1, a1, 56
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB35_1
> +; RV64I-NEXT: slli a0, a1, 56
> +; RV64I-NEXT: srai a0, a0, 56
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: blt s0, a0, .LBB35_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB35_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB35_1
> ; RV64I-NEXT: .LBB35_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -2191,32 +2191,32 @@ define i8 @atomicrmw_max_i8_monotonic(i8
> ;
> ; RV64IA-LABEL: atomicrmw_max_i8_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 56
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: addi a3, zero, 255
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 56
> ; RV64IA-NEXT: srai a1, a1, 56
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 56
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB35_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a4, a1, .LBB35_3
> +; RV64IA-NEXT: lr.w a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a3, a1, .LBB35_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB35_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB35_3: # in Loop: Header=BB35_1 Depth=1
> -; RV64IA-NEXT: sc.w a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB35_1
> +; RV64IA-NEXT: sc.w a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB35_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw max i8* %a, i8 %b monotonic
> ret i8 %1
> @@ -2233,33 +2233,33 @@ define i8 @atomicrmw_max_i8_acquire(i8 *
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 24
> -; RV32I-NEXT: srai s0, a1, 24
> +; RV32I-NEXT: lbu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 24
> +; RV32I-NEXT: srai s0, a0, 24
> ; RV32I-NEXT: addi s3, sp, 11
> ; RV32I-NEXT: j .LBB36_2
> ; RV32I-NEXT: .LBB36_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB36_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a1, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 2
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB36_4
> +; RV32I-NEXT: lb a1, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB36_4
> ; RV32I-NEXT: .LBB36_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 24
> -; RV32I-NEXT: srai a1, a1, 24
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: blt s0, a1, .LBB36_1
> +; RV32I-NEXT: slli a0, a1, 24
> +; RV32I-NEXT: srai a0, a0, 24
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: blt s0, a0, .LBB36_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB36_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB36_1
> ; RV32I-NEXT: .LBB36_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -2270,32 +2270,32 @@ define i8 @atomicrmw_max_i8_acquire(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_max_i8_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 24
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: addi a3, zero, 255
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 24
> ; RV32IA-NEXT: srai a1, a1, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 24
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB36_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a4, a1, .LBB36_3
> +; RV32IA-NEXT: lr.w.aq a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a3, a1, .LBB36_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB36_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB36_3: # in Loop: Header=BB36_1 Depth=1
> -; RV32IA-NEXT: sc.w a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB36_1
> +; RV32IA-NEXT: sc.w a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB36_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_max_i8_acquire:
> @@ -2308,33 +2308,33 @@ define i8 @atomicrmw_max_i8_acquire(i8 *
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 56
> -; RV64I-NEXT: srai s0, a1, 56
> +; RV64I-NEXT: lbu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 56
> +; RV64I-NEXT: srai s0, a0, 56
> ; RV64I-NEXT: addi s3, sp, 7
> ; RV64I-NEXT: j .LBB36_2
> ; RV64I-NEXT: .LBB36_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB36_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a1, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB36_4
> +; RV64I-NEXT: lb a1, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB36_4
> ; RV64I-NEXT: .LBB36_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 56
> -; RV64I-NEXT: srai a1, a1, 56
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB36_1
> +; RV64I-NEXT: slli a0, a1, 56
> +; RV64I-NEXT: srai a0, a0, 56
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: blt s0, a0, .LBB36_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB36_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB36_1
> ; RV64I-NEXT: .LBB36_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -2345,32 +2345,32 @@ define i8 @atomicrmw_max_i8_acquire(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_max_i8_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 56
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: addi a3, zero, 255
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 56
> ; RV64IA-NEXT: srai a1, a1, 56
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 56
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB36_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a4, a1, .LBB36_3
> +; RV64IA-NEXT: lr.w.aq a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a3, a1, .LBB36_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB36_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB36_3: # in Loop: Header=BB36_1 Depth=1
> -; RV64IA-NEXT: sc.w a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB36_1
> +; RV64IA-NEXT: sc.w a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB36_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw max i8* %a, i8 %b acquire
> ret i8 %1
> @@ -2387,33 +2387,33 @@ define i8 @atomicrmw_max_i8_release(i8 *
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 24
> -; RV32I-NEXT: srai s0, a1, 24
> +; RV32I-NEXT: lbu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 24
> +; RV32I-NEXT: srai s0, a0, 24
> ; RV32I-NEXT: addi s3, sp, 11
> ; RV32I-NEXT: j .LBB37_2
> ; RV32I-NEXT: .LBB37_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB37_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> +; RV32I-NEXT: sb a1, 11(sp)
> +; RV32I-NEXT: addi a3, zero, 3
> ; RV32I-NEXT: mv a0, s1
> ; RV32I-NEXT: mv a1, s3
> -; RV32I-NEXT: addi a3, zero, 3
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB37_4
> +; RV32I-NEXT: lb a1, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB37_4
> ; RV32I-NEXT: .LBB37_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 24
> -; RV32I-NEXT: srai a1, a1, 24
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: blt s0, a1, .LBB37_1
> +; RV32I-NEXT: slli a0, a1, 24
> +; RV32I-NEXT: srai a0, a0, 24
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: blt s0, a0, .LBB37_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB37_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB37_1
> ; RV32I-NEXT: .LBB37_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -2424,32 +2424,32 @@ define i8 @atomicrmw_max_i8_release(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_max_i8_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 24
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: addi a3, zero, 255
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 24
> ; RV32IA-NEXT: srai a1, a1, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 24
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB37_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a4, a1, .LBB37_3
> +; RV32IA-NEXT: lr.w a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a3, a1, .LBB37_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB37_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB37_3: # in Loop: Header=BB37_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB37_1
> +; RV32IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB37_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_max_i8_release:
> @@ -2462,33 +2462,33 @@ define i8 @atomicrmw_max_i8_release(i8 *
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 56
> -; RV64I-NEXT: srai s0, a1, 56
> +; RV64I-NEXT: lbu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 56
> +; RV64I-NEXT: srai s0, a0, 56
> ; RV64I-NEXT: addi s3, sp, 7
> ; RV64I-NEXT: j .LBB37_2
> ; RV64I-NEXT: .LBB37_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB37_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> +; RV64I-NEXT: sb a1, 7(sp)
> +; RV64I-NEXT: addi a3, zero, 3
> ; RV64I-NEXT: mv a0, s1
> ; RV64I-NEXT: mv a1, s3
> -; RV64I-NEXT: addi a3, zero, 3
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB37_4
> +; RV64I-NEXT: lb a1, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB37_4
> ; RV64I-NEXT: .LBB37_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 56
> -; RV64I-NEXT: srai a1, a1, 56
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB37_1
> +; RV64I-NEXT: slli a0, a1, 56
> +; RV64I-NEXT: srai a0, a0, 56
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: blt s0, a0, .LBB37_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB37_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB37_1
> ; RV64I-NEXT: .LBB37_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -2499,32 +2499,32 @@ define i8 @atomicrmw_max_i8_release(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_max_i8_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 56
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: addi a3, zero, 255
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 56
> ; RV64IA-NEXT: srai a1, a1, 56
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 56
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB37_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a4, a1, .LBB37_3
> +; RV64IA-NEXT: lr.w a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a3, a1, .LBB37_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB37_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB37_3: # in Loop: Header=BB37_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB37_1
> +; RV64IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB37_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw max i8* %a, i8 %b release
> ret i8 %1
> @@ -2541,33 +2541,33 @@ define i8 @atomicrmw_max_i8_acq_rel(i8 *
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 24
> -; RV32I-NEXT: srai s0, a1, 24
> +; RV32I-NEXT: lbu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 24
> +; RV32I-NEXT: srai s0, a0, 24
> ; RV32I-NEXT: addi s3, sp, 11
> ; RV32I-NEXT: j .LBB38_2
> ; RV32I-NEXT: .LBB38_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB38_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a1, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 4
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB38_4
> +; RV32I-NEXT: lb a1, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB38_4
> ; RV32I-NEXT: .LBB38_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 24
> -; RV32I-NEXT: srai a1, a1, 24
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: blt s0, a1, .LBB38_1
> +; RV32I-NEXT: slli a0, a1, 24
> +; RV32I-NEXT: srai a0, a0, 24
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: blt s0, a0, .LBB38_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB38_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB38_1
> ; RV32I-NEXT: .LBB38_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -2578,32 +2578,32 @@ define i8 @atomicrmw_max_i8_acq_rel(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_max_i8_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 24
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: addi a3, zero, 255
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 24
> ; RV32IA-NEXT: srai a1, a1, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 24
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB38_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a4, a1, .LBB38_3
> +; RV32IA-NEXT: lr.w.aq a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a3, a1, .LBB38_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB38_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB38_3: # in Loop: Header=BB38_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB38_1
> +; RV32IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB38_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_max_i8_acq_rel:
> @@ -2616,33 +2616,33 @@ define i8 @atomicrmw_max_i8_acq_rel(i8 *
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 56
> -; RV64I-NEXT: srai s0, a1, 56
> +; RV64I-NEXT: lbu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 56
> +; RV64I-NEXT: srai s0, a0, 56
> ; RV64I-NEXT: addi s3, sp, 7
> ; RV64I-NEXT: j .LBB38_2
> ; RV64I-NEXT: .LBB38_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB38_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a1, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB38_4
> +; RV64I-NEXT: lb a1, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB38_4
> ; RV64I-NEXT: .LBB38_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 56
> -; RV64I-NEXT: srai a1, a1, 56
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB38_1
> +; RV64I-NEXT: slli a0, a1, 56
> +; RV64I-NEXT: srai a0, a0, 56
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: blt s0, a0, .LBB38_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB38_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB38_1
> ; RV64I-NEXT: .LBB38_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -2653,32 +2653,32 @@ define i8 @atomicrmw_max_i8_acq_rel(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_max_i8_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 56
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: addi a3, zero, 255
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 56
> ; RV64IA-NEXT: srai a1, a1, 56
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 56
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB38_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a4, a1, .LBB38_3
> +; RV64IA-NEXT: lr.w.aq a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a3, a1, .LBB38_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB38_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB38_3: # in Loop: Header=BB38_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB38_1
> +; RV64IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB38_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw max i8* %a, i8 %b acq_rel
> ret i8 %1
> @@ -2695,33 +2695,33 @@ define i8 @atomicrmw_max_i8_seq_cst(i8 *
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 24
> -; RV32I-NEXT: srai s0, a1, 24
> +; RV32I-NEXT: lbu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 24
> +; RV32I-NEXT: srai s0, a0, 24
> ; RV32I-NEXT: addi s3, sp, 11
> ; RV32I-NEXT: j .LBB39_2
> ; RV32I-NEXT: .LBB39_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB39_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a1, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 5
> ; RV32I-NEXT: addi a4, zero, 5
> +; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB39_4
> +; RV32I-NEXT: lb a1, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB39_4
> ; RV32I-NEXT: .LBB39_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 24
> -; RV32I-NEXT: srai a1, a1, 24
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: blt s0, a1, .LBB39_1
> +; RV32I-NEXT: slli a0, a1, 24
> +; RV32I-NEXT: srai a0, a0, 24
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: blt s0, a0, .LBB39_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB39_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB39_1
> ; RV32I-NEXT: .LBB39_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -2732,32 +2732,32 @@ define i8 @atomicrmw_max_i8_seq_cst(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_max_i8_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 24
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: addi a3, zero, 255
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 24
> ; RV32IA-NEXT: srai a1, a1, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 24
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB39_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a4, a1, .LBB39_3
> +; RV32IA-NEXT: lr.w.aqrl a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a3, a1, .LBB39_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB39_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB39_3: # in Loop: Header=BB39_1 Depth=1
> -; RV32IA-NEXT: sc.w.aqrl a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB39_1
> +; RV32IA-NEXT: sc.w.aqrl a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB39_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_max_i8_seq_cst:
> @@ -2770,33 +2770,33 @@ define i8 @atomicrmw_max_i8_seq_cst(i8 *
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 56
> -; RV64I-NEXT: srai s0, a1, 56
> +; RV64I-NEXT: lbu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 56
> +; RV64I-NEXT: srai s0, a0, 56
> ; RV64I-NEXT: addi s3, sp, 7
> ; RV64I-NEXT: j .LBB39_2
> ; RV64I-NEXT: .LBB39_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB39_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a1, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB39_4
> +; RV64I-NEXT: lb a1, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB39_4
> ; RV64I-NEXT: .LBB39_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 56
> -; RV64I-NEXT: srai a1, a1, 56
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB39_1
> +; RV64I-NEXT: slli a0, a1, 56
> +; RV64I-NEXT: srai a0, a0, 56
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: blt s0, a0, .LBB39_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB39_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB39_1
> ; RV64I-NEXT: .LBB39_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -2807,32 +2807,32 @@ define i8 @atomicrmw_max_i8_seq_cst(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_max_i8_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 56
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: addi a3, zero, 255
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 56
> ; RV64IA-NEXT: srai a1, a1, 56
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 56
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB39_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a4, a1, .LBB39_3
> +; RV64IA-NEXT: lr.w.aqrl a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a3, a1, .LBB39_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB39_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB39_3: # in Loop: Header=BB39_1 Depth=1
> -; RV64IA-NEXT: sc.w.aqrl a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB39_1
> +; RV64IA-NEXT: sc.w.aqrl a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB39_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw max i8* %a, i8 %b seq_cst
> ret i8 %1
> @@ -2849,33 +2849,33 @@ define i8 @atomicrmw_min_i8_monotonic(i8
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 24
> -; RV32I-NEXT: srai s0, a1, 24
> +; RV32I-NEXT: lbu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 24
> +; RV32I-NEXT: srai s0, a0, 24
> ; RV32I-NEXT: addi s3, sp, 11
> ; RV32I-NEXT: j .LBB40_2
> ; RV32I-NEXT: .LBB40_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB40_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> +; RV32I-NEXT: sb a1, 11(sp)
> ; RV32I-NEXT: mv a0, s1
> ; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a3, zero
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB40_4
> +; RV32I-NEXT: lb a1, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB40_4
> ; RV32I-NEXT: .LBB40_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 24
> -; RV32I-NEXT: srai a1, a1, 24
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bge s0, a1, .LBB40_1
> +; RV32I-NEXT: slli a0, a1, 24
> +; RV32I-NEXT: srai a0, a0, 24
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bge s0, a0, .LBB40_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB40_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB40_1
> ; RV32I-NEXT: .LBB40_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -2886,32 +2886,32 @@ define i8 @atomicrmw_min_i8_monotonic(i8
> ;
> ; RV32IA-LABEL: atomicrmw_min_i8_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 24
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: addi a3, zero, 255
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 24
> ; RV32IA-NEXT: srai a1, a1, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 24
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB40_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a1, a4, .LBB40_3
> +; RV32IA-NEXT: lr.w a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a1, a3, .LBB40_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB40_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB40_3: # in Loop: Header=BB40_1 Depth=1
> -; RV32IA-NEXT: sc.w a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB40_1
> +; RV32IA-NEXT: sc.w a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB40_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_min_i8_monotonic:
> @@ -2924,33 +2924,33 @@ define i8 @atomicrmw_min_i8_monotonic(i8
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 56
> -; RV64I-NEXT: srai s0, a1, 56
> +; RV64I-NEXT: lbu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 56
> +; RV64I-NEXT: srai s0, a0, 56
> ; RV64I-NEXT: addi s3, sp, 7
> ; RV64I-NEXT: j .LBB40_2
> ; RV64I-NEXT: .LBB40_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB40_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> +; RV64I-NEXT: sb a1, 7(sp)
> ; RV64I-NEXT: mv a0, s1
> ; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB40_4
> +; RV64I-NEXT: lb a1, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB40_4
> ; RV64I-NEXT: .LBB40_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 56
> -; RV64I-NEXT: srai a1, a1, 56
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB40_1
> +; RV64I-NEXT: slli a0, a1, 56
> +; RV64I-NEXT: srai a0, a0, 56
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bge s0, a0, .LBB40_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB40_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB40_1
> ; RV64I-NEXT: .LBB40_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -2961,32 +2961,32 @@ define i8 @atomicrmw_min_i8_monotonic(i8
> ;
> ; RV64IA-LABEL: atomicrmw_min_i8_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 56
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: addi a3, zero, 255
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 56
> ; RV64IA-NEXT: srai a1, a1, 56
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 56
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB40_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a1, a4, .LBB40_3
> +; RV64IA-NEXT: lr.w a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a1, a3, .LBB40_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB40_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB40_3: # in Loop: Header=BB40_1 Depth=1
> -; RV64IA-NEXT: sc.w a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB40_1
> +; RV64IA-NEXT: sc.w a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB40_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw min i8* %a, i8 %b monotonic
> ret i8 %1
> @@ -3003,33 +3003,33 @@ define i8 @atomicrmw_min_i8_acquire(i8 *
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 24
> -; RV32I-NEXT: srai s0, a1, 24
> +; RV32I-NEXT: lbu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 24
> +; RV32I-NEXT: srai s0, a0, 24
> ; RV32I-NEXT: addi s3, sp, 11
> ; RV32I-NEXT: j .LBB41_2
> ; RV32I-NEXT: .LBB41_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB41_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a1, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 2
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB41_4
> +; RV32I-NEXT: lb a1, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB41_4
> ; RV32I-NEXT: .LBB41_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 24
> -; RV32I-NEXT: srai a1, a1, 24
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bge s0, a1, .LBB41_1
> +; RV32I-NEXT: slli a0, a1, 24
> +; RV32I-NEXT: srai a0, a0, 24
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bge s0, a0, .LBB41_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB41_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB41_1
> ; RV32I-NEXT: .LBB41_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -3040,32 +3040,32 @@ define i8 @atomicrmw_min_i8_acquire(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_min_i8_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 24
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: addi a3, zero, 255
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 24
> ; RV32IA-NEXT: srai a1, a1, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 24
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB41_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a1, a4, .LBB41_3
> +; RV32IA-NEXT: lr.w.aq a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a1, a3, .LBB41_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB41_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB41_3: # in Loop: Header=BB41_1 Depth=1
> -; RV32IA-NEXT: sc.w a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB41_1
> +; RV32IA-NEXT: sc.w a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB41_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_min_i8_acquire:
> @@ -3078,33 +3078,33 @@ define i8 @atomicrmw_min_i8_acquire(i8 *
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 56
> -; RV64I-NEXT: srai s0, a1, 56
> +; RV64I-NEXT: lbu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 56
> +; RV64I-NEXT: srai s0, a0, 56
> ; RV64I-NEXT: addi s3, sp, 7
> ; RV64I-NEXT: j .LBB41_2
> ; RV64I-NEXT: .LBB41_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB41_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a1, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB41_4
> +; RV64I-NEXT: lb a1, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB41_4
> ; RV64I-NEXT: .LBB41_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 56
> -; RV64I-NEXT: srai a1, a1, 56
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB41_1
> +; RV64I-NEXT: slli a0, a1, 56
> +; RV64I-NEXT: srai a0, a0, 56
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bge s0, a0, .LBB41_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB41_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB41_1
> ; RV64I-NEXT: .LBB41_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -3115,32 +3115,32 @@ define i8 @atomicrmw_min_i8_acquire(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_min_i8_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 56
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: addi a3, zero, 255
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 56
> ; RV64IA-NEXT: srai a1, a1, 56
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 56
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB41_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a1, a4, .LBB41_3
> +; RV64IA-NEXT: lr.w.aq a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a1, a3, .LBB41_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB41_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB41_3: # in Loop: Header=BB41_1 Depth=1
> -; RV64IA-NEXT: sc.w a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB41_1
> +; RV64IA-NEXT: sc.w a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB41_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw min i8* %a, i8 %b acquire
> ret i8 %1
> @@ -3157,33 +3157,33 @@ define i8 @atomicrmw_min_i8_release(i8 *
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 24
> -; RV32I-NEXT: srai s0, a1, 24
> +; RV32I-NEXT: lbu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 24
> +; RV32I-NEXT: srai s0, a0, 24
> ; RV32I-NEXT: addi s3, sp, 11
> ; RV32I-NEXT: j .LBB42_2
> ; RV32I-NEXT: .LBB42_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB42_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> +; RV32I-NEXT: sb a1, 11(sp)
> +; RV32I-NEXT: addi a3, zero, 3
> ; RV32I-NEXT: mv a0, s1
> ; RV32I-NEXT: mv a1, s3
> -; RV32I-NEXT: addi a3, zero, 3
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB42_4
> +; RV32I-NEXT: lb a1, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB42_4
> ; RV32I-NEXT: .LBB42_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 24
> -; RV32I-NEXT: srai a1, a1, 24
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bge s0, a1, .LBB42_1
> +; RV32I-NEXT: slli a0, a1, 24
> +; RV32I-NEXT: srai a0, a0, 24
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bge s0, a0, .LBB42_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB42_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB42_1
> ; RV32I-NEXT: .LBB42_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -3194,32 +3194,32 @@ define i8 @atomicrmw_min_i8_release(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_min_i8_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 24
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: addi a3, zero, 255
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 24
> ; RV32IA-NEXT: srai a1, a1, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 24
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB42_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a1, a4, .LBB42_3
> +; RV32IA-NEXT: lr.w a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a1, a3, .LBB42_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB42_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB42_3: # in Loop: Header=BB42_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB42_1
> +; RV32IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB42_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_min_i8_release:
> @@ -3232,33 +3232,33 @@ define i8 @atomicrmw_min_i8_release(i8 *
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 56
> -; RV64I-NEXT: srai s0, a1, 56
> +; RV64I-NEXT: lbu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 56
> +; RV64I-NEXT: srai s0, a0, 56
> ; RV64I-NEXT: addi s3, sp, 7
> ; RV64I-NEXT: j .LBB42_2
> ; RV64I-NEXT: .LBB42_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB42_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> +; RV64I-NEXT: sb a1, 7(sp)
> +; RV64I-NEXT: addi a3, zero, 3
> ; RV64I-NEXT: mv a0, s1
> ; RV64I-NEXT: mv a1, s3
> -; RV64I-NEXT: addi a3, zero, 3
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB42_4
> +; RV64I-NEXT: lb a1, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB42_4
> ; RV64I-NEXT: .LBB42_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 56
> -; RV64I-NEXT: srai a1, a1, 56
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB42_1
> +; RV64I-NEXT: slli a0, a1, 56
> +; RV64I-NEXT: srai a0, a0, 56
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bge s0, a0, .LBB42_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB42_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB42_1
> ; RV64I-NEXT: .LBB42_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -3269,32 +3269,32 @@ define i8 @atomicrmw_min_i8_release(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_min_i8_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 56
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: addi a3, zero, 255
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 56
> ; RV64IA-NEXT: srai a1, a1, 56
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 56
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB42_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a1, a4, .LBB42_3
> +; RV64IA-NEXT: lr.w a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a1, a3, .LBB42_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB42_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB42_3: # in Loop: Header=BB42_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB42_1
> +; RV64IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB42_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw min i8* %a, i8 %b release
> ret i8 %1
> @@ -3311,33 +3311,33 @@ define i8 @atomicrmw_min_i8_acq_rel(i8 *
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 24
> -; RV32I-NEXT: srai s0, a1, 24
> +; RV32I-NEXT: lbu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 24
> +; RV32I-NEXT: srai s0, a0, 24
> ; RV32I-NEXT: addi s3, sp, 11
> ; RV32I-NEXT: j .LBB43_2
> ; RV32I-NEXT: .LBB43_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB43_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a1, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 4
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB43_4
> +; RV32I-NEXT: lb a1, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB43_4
> ; RV32I-NEXT: .LBB43_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 24
> -; RV32I-NEXT: srai a1, a1, 24
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bge s0, a1, .LBB43_1
> +; RV32I-NEXT: slli a0, a1, 24
> +; RV32I-NEXT: srai a0, a0, 24
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bge s0, a0, .LBB43_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB43_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB43_1
> ; RV32I-NEXT: .LBB43_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -3348,32 +3348,32 @@ define i8 @atomicrmw_min_i8_acq_rel(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_min_i8_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 24
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: addi a3, zero, 255
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 24
> ; RV32IA-NEXT: srai a1, a1, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 24
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB43_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a1, a4, .LBB43_3
> +; RV32IA-NEXT: lr.w.aq a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a1, a3, .LBB43_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB43_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB43_3: # in Loop: Header=BB43_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB43_1
> +; RV32IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB43_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_min_i8_acq_rel:
> @@ -3386,33 +3386,33 @@ define i8 @atomicrmw_min_i8_acq_rel(i8 *
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 56
> -; RV64I-NEXT: srai s0, a1, 56
> +; RV64I-NEXT: lbu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 56
> +; RV64I-NEXT: srai s0, a0, 56
> ; RV64I-NEXT: addi s3, sp, 7
> ; RV64I-NEXT: j .LBB43_2
> ; RV64I-NEXT: .LBB43_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB43_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a1, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB43_4
> +; RV64I-NEXT: lb a1, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB43_4
> ; RV64I-NEXT: .LBB43_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 56
> -; RV64I-NEXT: srai a1, a1, 56
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB43_1
> +; RV64I-NEXT: slli a0, a1, 56
> +; RV64I-NEXT: srai a0, a0, 56
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bge s0, a0, .LBB43_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB43_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB43_1
> ; RV64I-NEXT: .LBB43_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -3423,32 +3423,32 @@ define i8 @atomicrmw_min_i8_acq_rel(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_min_i8_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 56
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: addi a3, zero, 255
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 56
> ; RV64IA-NEXT: srai a1, a1, 56
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 56
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB43_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a1, a4, .LBB43_3
> +; RV64IA-NEXT: lr.w.aq a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a1, a3, .LBB43_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB43_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB43_3: # in Loop: Header=BB43_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB43_1
> +; RV64IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB43_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw min i8* %a, i8 %b acq_rel
> ret i8 %1
> @@ -3465,33 +3465,33 @@ define i8 @atomicrmw_min_i8_seq_cst(i8 *
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 24
> -; RV32I-NEXT: srai s0, a1, 24
> +; RV32I-NEXT: lbu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 24
> +; RV32I-NEXT: srai s0, a0, 24
> ; RV32I-NEXT: addi s3, sp, 11
> ; RV32I-NEXT: j .LBB44_2
> ; RV32I-NEXT: .LBB44_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB44_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a1, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 5
> ; RV32I-NEXT: addi a4, zero, 5
> +; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB44_4
> +; RV32I-NEXT: lb a1, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB44_4
> ; RV32I-NEXT: .LBB44_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 24
> -; RV32I-NEXT: srai a1, a1, 24
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bge s0, a1, .LBB44_1
> +; RV32I-NEXT: slli a0, a1, 24
> +; RV32I-NEXT: srai a0, a0, 24
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bge s0, a0, .LBB44_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB44_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB44_1
> ; RV32I-NEXT: .LBB44_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -3502,32 +3502,32 @@ define i8 @atomicrmw_min_i8_seq_cst(i8 *
> ;
> ; RV32IA-LABEL: atomicrmw_min_i8_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 24
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: addi a4, zero, 255
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: addi a3, zero, 255
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 24
> ; RV32IA-NEXT: srai a1, a1, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 24
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB44_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a1, a4, .LBB44_3
> +; RV32IA-NEXT: lr.w.aqrl a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a1, a3, .LBB44_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB44_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB44_3: # in Loop: Header=BB44_1 Depth=1
> -; RV32IA-NEXT: sc.w.aqrl a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB44_1
> +; RV32IA-NEXT: sc.w.aqrl a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB44_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_min_i8_seq_cst:
> @@ -3540,33 +3540,33 @@ define i8 @atomicrmw_min_i8_seq_cst(i8 *
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 56
> -; RV64I-NEXT: srai s0, a1, 56
> +; RV64I-NEXT: lbu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 56
> +; RV64I-NEXT: srai s0, a0, 56
> ; RV64I-NEXT: addi s3, sp, 7
> ; RV64I-NEXT: j .LBB44_2
> ; RV64I-NEXT: .LBB44_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB44_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a1, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB44_4
> +; RV64I-NEXT: lb a1, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB44_4
> ; RV64I-NEXT: .LBB44_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 56
> -; RV64I-NEXT: srai a1, a1, 56
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB44_1
> +; RV64I-NEXT: slli a0, a1, 56
> +; RV64I-NEXT: srai a0, a0, 56
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bge s0, a0, .LBB44_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB44_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB44_1
> ; RV64I-NEXT: .LBB44_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -3577,32 +3577,32 @@ define i8 @atomicrmw_min_i8_seq_cst(i8 *
> ;
> ; RV64IA-LABEL: atomicrmw_min_i8_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 56
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: addi a4, zero, 255
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: addi a3, zero, 255
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 56
> ; RV64IA-NEXT: srai a1, a1, 56
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 56
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB44_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a1, a4, .LBB44_3
> +; RV64IA-NEXT: lr.w.aqrl a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a1, a3, .LBB44_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB44_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB44_3: # in Loop: Header=BB44_1 Depth=1
> -; RV64IA-NEXT: sc.w.aqrl a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB44_1
> +; RV64IA-NEXT: sc.w.aqrl a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB44_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw min i8* %a, i8 %b seq_cst
> ret i8 %1
> @@ -3617,33 +3617,33 @@ define i8 @atomicrmw_umax_i8_monotonic(i
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: andi s0, a1, 255
> -; RV32I-NEXT: addi s3, sp, 11
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lbu a3, 0(a0)
> +; RV32I-NEXT: mv s3, a1
> +; RV32I-NEXT: andi s1, a1, 255
> +; RV32I-NEXT: addi s2, sp, 11
> ; RV32I-NEXT: j .LBB45_2
> ; RV32I-NEXT: .LBB45_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB45_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a3, 11(sp)
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: mv a3, zero
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB45_4
> +; RV32I-NEXT: lb a3, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB45_4
> ; RV32I-NEXT: .LBB45_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: andi a1, a0, 255
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bltu s0, a1, .LBB45_1
> +; RV32I-NEXT: andi a0, a3, 255
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a0, .LBB45_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB45_2 Depth=1
> -; RV32I-NEXT: mv a2, s2
> +; RV32I-NEXT: mv a2, s3
> ; RV32I-NEXT: j .LBB45_1
> ; RV32I-NEXT: .LBB45_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -3654,27 +3654,27 @@ define i8 @atomicrmw_umax_i8_monotonic(i
> ;
> ; RV32IA-LABEL: atomicrmw_umax_i8_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a6, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB45_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: and a3, a4, a6
> +; RV32IA-NEXT: lr.w a4, (a6)
> +; RV32IA-NEXT: and a2, a4, a3
> ; RV32IA-NEXT: mv a5, a4
> -; RV32IA-NEXT: bgeu a3, a1, .LBB45_3
> +; RV32IA-NEXT: bgeu a2, a1, .LBB45_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB45_1 Depth=1
> ; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> +; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: .LBB45_3: # in Loop: Header=BB45_1 Depth=1
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB45_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umax_i8_monotonic:
> @@ -3685,33 +3685,33 @@ define i8 @atomicrmw_umax_i8_monotonic(i
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: andi s0, a1, 255
> -; RV64I-NEXT: addi s3, sp, 7
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lbu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: andi s1, a1, 255
> +; RV64I-NEXT: addi s2, sp, 7
> ; RV64I-NEXT: j .LBB45_2
> ; RV64I-NEXT: .LBB45_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB45_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a3, 7(sp)
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB45_4
> +; RV64I-NEXT: lb a3, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB45_4
> ; RV64I-NEXT: .LBB45_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: andi a1, a0, 255
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s0, a1, .LBB45_1
> +; RV64I-NEXT: andi a0, a3, 255
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a0, .LBB45_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB45_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB45_1
> ; RV64I-NEXT: .LBB45_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -3722,27 +3722,27 @@ define i8 @atomicrmw_umax_i8_monotonic(i
> ;
> ; RV64IA-LABEL: atomicrmw_umax_i8_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a6, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB45_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: and a3, a4, a6
> +; RV64IA-NEXT: lr.w a4, (a6)
> +; RV64IA-NEXT: and a2, a4, a3
> ; RV64IA-NEXT: mv a5, a4
> -; RV64IA-NEXT: bgeu a3, a1, .LBB45_3
> +; RV64IA-NEXT: bgeu a2, a1, .LBB45_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB45_1 Depth=1
> ; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> +; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: .LBB45_3: # in Loop: Header=BB45_1 Depth=1
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB45_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umax i8* %a, i8 %b monotonic
> ret i8 %1
> @@ -3757,33 +3757,33 @@ define i8 @atomicrmw_umax_i8_acquire(i8
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: andi s0, a1, 255
> -; RV32I-NEXT: addi s3, sp, 11
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lbu a3, 0(a0)
> +; RV32I-NEXT: mv s3, a1
> +; RV32I-NEXT: andi s1, a1, 255
> +; RV32I-NEXT: addi s2, sp, 11
> ; RV32I-NEXT: j .LBB46_2
> ; RV32I-NEXT: .LBB46_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB46_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a3, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 2
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB46_4
> +; RV32I-NEXT: lb a3, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB46_4
> ; RV32I-NEXT: .LBB46_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: andi a1, a0, 255
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bltu s0, a1, .LBB46_1
> +; RV32I-NEXT: andi a0, a3, 255
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a0, .LBB46_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB46_2 Depth=1
> -; RV32I-NEXT: mv a2, s2
> +; RV32I-NEXT: mv a2, s3
> ; RV32I-NEXT: j .LBB46_1
> ; RV32I-NEXT: .LBB46_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -3794,27 +3794,27 @@ define i8 @atomicrmw_umax_i8_acquire(i8
> ;
> ; RV32IA-LABEL: atomicrmw_umax_i8_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a6, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB46_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a3, a4, a6
> +; RV32IA-NEXT: lr.w.aq a4, (a6)
> +; RV32IA-NEXT: and a2, a4, a3
> ; RV32IA-NEXT: mv a5, a4
> -; RV32IA-NEXT: bgeu a3, a1, .LBB46_3
> +; RV32IA-NEXT: bgeu a2, a1, .LBB46_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB46_1 Depth=1
> ; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> +; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: .LBB46_3: # in Loop: Header=BB46_1 Depth=1
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB46_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umax_i8_acquire:
> @@ -3825,33 +3825,33 @@ define i8 @atomicrmw_umax_i8_acquire(i8
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: andi s0, a1, 255
> -; RV64I-NEXT: addi s3, sp, 7
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lbu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: andi s1, a1, 255
> +; RV64I-NEXT: addi s2, sp, 7
> ; RV64I-NEXT: j .LBB46_2
> ; RV64I-NEXT: .LBB46_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB46_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a3, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB46_4
> +; RV64I-NEXT: lb a3, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB46_4
> ; RV64I-NEXT: .LBB46_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: andi a1, a0, 255
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s0, a1, .LBB46_1
> +; RV64I-NEXT: andi a0, a3, 255
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a0, .LBB46_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB46_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB46_1
> ; RV64I-NEXT: .LBB46_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -3862,27 +3862,27 @@ define i8 @atomicrmw_umax_i8_acquire(i8
> ;
> ; RV64IA-LABEL: atomicrmw_umax_i8_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a6, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB46_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a3, a4, a6
> +; RV64IA-NEXT: lr.w.aq a4, (a6)
> +; RV64IA-NEXT: and a2, a4, a3
> ; RV64IA-NEXT: mv a5, a4
> -; RV64IA-NEXT: bgeu a3, a1, .LBB46_3
> +; RV64IA-NEXT: bgeu a2, a1, .LBB46_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB46_1 Depth=1
> ; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> +; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: .LBB46_3: # in Loop: Header=BB46_1 Depth=1
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB46_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umax i8* %a, i8 %b acquire
> ret i8 %1
> @@ -3897,33 +3897,33 @@ define i8 @atomicrmw_umax_i8_release(i8
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: andi s0, a1, 255
> -; RV32I-NEXT: addi s3, sp, 11
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lbu a3, 0(a0)
> +; RV32I-NEXT: mv s3, a1
> +; RV32I-NEXT: andi s1, a1, 255
> +; RV32I-NEXT: addi s2, sp, 11
> ; RV32I-NEXT: j .LBB47_2
> ; RV32I-NEXT: .LBB47_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB47_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a3, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 3
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB47_4
> +; RV32I-NEXT: lb a3, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB47_4
> ; RV32I-NEXT: .LBB47_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: andi a1, a0, 255
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bltu s0, a1, .LBB47_1
> +; RV32I-NEXT: andi a0, a3, 255
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a0, .LBB47_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB47_2 Depth=1
> -; RV32I-NEXT: mv a2, s2
> +; RV32I-NEXT: mv a2, s3
> ; RV32I-NEXT: j .LBB47_1
> ; RV32I-NEXT: .LBB47_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -3934,27 +3934,27 @@ define i8 @atomicrmw_umax_i8_release(i8
> ;
> ; RV32IA-LABEL: atomicrmw_umax_i8_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a6, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB47_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: and a3, a4, a6
> +; RV32IA-NEXT: lr.w a4, (a6)
> +; RV32IA-NEXT: and a2, a4, a3
> ; RV32IA-NEXT: mv a5, a4
> -; RV32IA-NEXT: bgeu a3, a1, .LBB47_3
> +; RV32IA-NEXT: bgeu a2, a1, .LBB47_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB47_1 Depth=1
> ; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> +; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: .LBB47_3: # in Loop: Header=BB47_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB47_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umax_i8_release:
> @@ -3965,33 +3965,33 @@ define i8 @atomicrmw_umax_i8_release(i8
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: andi s0, a1, 255
> -; RV64I-NEXT: addi s3, sp, 7
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lbu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: andi s1, a1, 255
> +; RV64I-NEXT: addi s2, sp, 7
> ; RV64I-NEXT: j .LBB47_2
> ; RV64I-NEXT: .LBB47_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB47_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a3, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 3
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB47_4
> +; RV64I-NEXT: lb a3, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB47_4
> ; RV64I-NEXT: .LBB47_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: andi a1, a0, 255
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s0, a1, .LBB47_1
> +; RV64I-NEXT: andi a0, a3, 255
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a0, .LBB47_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB47_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB47_1
> ; RV64I-NEXT: .LBB47_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -4002,27 +4002,27 @@ define i8 @atomicrmw_umax_i8_release(i8
> ;
> ; RV64IA-LABEL: atomicrmw_umax_i8_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a6, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB47_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: and a3, a4, a6
> +; RV64IA-NEXT: lr.w a4, (a6)
> +; RV64IA-NEXT: and a2, a4, a3
> ; RV64IA-NEXT: mv a5, a4
> -; RV64IA-NEXT: bgeu a3, a1, .LBB47_3
> +; RV64IA-NEXT: bgeu a2, a1, .LBB47_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB47_1 Depth=1
> ; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> +; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: .LBB47_3: # in Loop: Header=BB47_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB47_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umax i8* %a, i8 %b release
> ret i8 %1
> @@ -4037,33 +4037,33 @@ define i8 @atomicrmw_umax_i8_acq_rel(i8
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: andi s0, a1, 255
> -; RV32I-NEXT: addi s3, sp, 11
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lbu a3, 0(a0)
> +; RV32I-NEXT: mv s3, a1
> +; RV32I-NEXT: andi s1, a1, 255
> +; RV32I-NEXT: addi s2, sp, 11
> ; RV32I-NEXT: j .LBB48_2
> ; RV32I-NEXT: .LBB48_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB48_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a3, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 4
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB48_4
> +; RV32I-NEXT: lb a3, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB48_4
> ; RV32I-NEXT: .LBB48_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: andi a1, a0, 255
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bltu s0, a1, .LBB48_1
> +; RV32I-NEXT: andi a0, a3, 255
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a0, .LBB48_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB48_2 Depth=1
> -; RV32I-NEXT: mv a2, s2
> +; RV32I-NEXT: mv a2, s3
> ; RV32I-NEXT: j .LBB48_1
> ; RV32I-NEXT: .LBB48_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -4074,27 +4074,27 @@ define i8 @atomicrmw_umax_i8_acq_rel(i8
> ;
> ; RV32IA-LABEL: atomicrmw_umax_i8_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a6, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB48_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a3, a4, a6
> +; RV32IA-NEXT: lr.w.aq a4, (a6)
> +; RV32IA-NEXT: and a2, a4, a3
> ; RV32IA-NEXT: mv a5, a4
> -; RV32IA-NEXT: bgeu a3, a1, .LBB48_3
> +; RV32IA-NEXT: bgeu a2, a1, .LBB48_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB48_1 Depth=1
> ; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> +; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: .LBB48_3: # in Loop: Header=BB48_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB48_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umax_i8_acq_rel:
> @@ -4105,33 +4105,33 @@ define i8 @atomicrmw_umax_i8_acq_rel(i8
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: andi s0, a1, 255
> -; RV64I-NEXT: addi s3, sp, 7
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lbu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: andi s1, a1, 255
> +; RV64I-NEXT: addi s2, sp, 7
> ; RV64I-NEXT: j .LBB48_2
> ; RV64I-NEXT: .LBB48_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB48_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a3, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB48_4
> +; RV64I-NEXT: lb a3, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB48_4
> ; RV64I-NEXT: .LBB48_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: andi a1, a0, 255
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s0, a1, .LBB48_1
> +; RV64I-NEXT: andi a0, a3, 255
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a0, .LBB48_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB48_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB48_1
> ; RV64I-NEXT: .LBB48_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -4142,27 +4142,27 @@ define i8 @atomicrmw_umax_i8_acq_rel(i8
> ;
> ; RV64IA-LABEL: atomicrmw_umax_i8_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a6, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB48_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a3, a4, a6
> +; RV64IA-NEXT: lr.w.aq a4, (a6)
> +; RV64IA-NEXT: and a2, a4, a3
> ; RV64IA-NEXT: mv a5, a4
> -; RV64IA-NEXT: bgeu a3, a1, .LBB48_3
> +; RV64IA-NEXT: bgeu a2, a1, .LBB48_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB48_1 Depth=1
> ; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> +; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: .LBB48_3: # in Loop: Header=BB48_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB48_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umax i8* %a, i8 %b acq_rel
> ret i8 %1
> @@ -4177,33 +4177,33 @@ define i8 @atomicrmw_umax_i8_seq_cst(i8
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: andi s0, a1, 255
> -; RV32I-NEXT: addi s3, sp, 11
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lbu a3, 0(a0)
> +; RV32I-NEXT: mv s3, a1
> +; RV32I-NEXT: andi s1, a1, 255
> +; RV32I-NEXT: addi s2, sp, 11
> ; RV32I-NEXT: j .LBB49_2
> ; RV32I-NEXT: .LBB49_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB49_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a3, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 5
> ; RV32I-NEXT: addi a4, zero, 5
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB49_4
> +; RV32I-NEXT: lb a3, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB49_4
> ; RV32I-NEXT: .LBB49_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: andi a1, a0, 255
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bltu s0, a1, .LBB49_1
> +; RV32I-NEXT: andi a0, a3, 255
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a0, .LBB49_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB49_2 Depth=1
> -; RV32I-NEXT: mv a2, s2
> +; RV32I-NEXT: mv a2, s3
> ; RV32I-NEXT: j .LBB49_1
> ; RV32I-NEXT: .LBB49_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -4214,27 +4214,27 @@ define i8 @atomicrmw_umax_i8_seq_cst(i8
> ;
> ; RV32IA-LABEL: atomicrmw_umax_i8_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a6, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB49_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV32IA-NEXT: and a3, a4, a6
> +; RV32IA-NEXT: lr.w.aqrl a4, (a6)
> +; RV32IA-NEXT: and a2, a4, a3
> ; RV32IA-NEXT: mv a5, a4
> -; RV32IA-NEXT: bgeu a3, a1, .LBB49_3
> +; RV32IA-NEXT: bgeu a2, a1, .LBB49_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB49_1 Depth=1
> ; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> +; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: .LBB49_3: # in Loop: Header=BB49_1 Depth=1
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB49_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umax_i8_seq_cst:
> @@ -4245,33 +4245,33 @@ define i8 @atomicrmw_umax_i8_seq_cst(i8
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: andi s0, a1, 255
> -; RV64I-NEXT: addi s3, sp, 7
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lbu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: andi s1, a1, 255
> +; RV64I-NEXT: addi s2, sp, 7
> ; RV64I-NEXT: j .LBB49_2
> ; RV64I-NEXT: .LBB49_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB49_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a3, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB49_4
> +; RV64I-NEXT: lb a3, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB49_4
> ; RV64I-NEXT: .LBB49_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: andi a1, a0, 255
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s0, a1, .LBB49_1
> +; RV64I-NEXT: andi a0, a3, 255
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a0, .LBB49_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB49_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB49_1
> ; RV64I-NEXT: .LBB49_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -4282,27 +4282,27 @@ define i8 @atomicrmw_umax_i8_seq_cst(i8
> ;
> ; RV64IA-LABEL: atomicrmw_umax_i8_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a6, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB49_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV64IA-NEXT: and a3, a4, a6
> +; RV64IA-NEXT: lr.w.aqrl a4, (a6)
> +; RV64IA-NEXT: and a2, a4, a3
> ; RV64IA-NEXT: mv a5, a4
> -; RV64IA-NEXT: bgeu a3, a1, .LBB49_3
> +; RV64IA-NEXT: bgeu a2, a1, .LBB49_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB49_1 Depth=1
> ; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> +; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: .LBB49_3: # in Loop: Header=BB49_1 Depth=1
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB49_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umax i8* %a, i8 %b seq_cst
> ret i8 %1
> @@ -4317,33 +4317,33 @@ define i8 @atomicrmw_umin_i8_monotonic(i
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: andi s0, a1, 255
> -; RV32I-NEXT: addi s3, sp, 11
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lbu a3, 0(a0)
> +; RV32I-NEXT: mv s3, a1
> +; RV32I-NEXT: andi s1, a1, 255
> +; RV32I-NEXT: addi s2, sp, 11
> ; RV32I-NEXT: j .LBB50_2
> ; RV32I-NEXT: .LBB50_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB50_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a3, 11(sp)
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: mv a3, zero
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB50_4
> +; RV32I-NEXT: lb a3, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB50_4
> ; RV32I-NEXT: .LBB50_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: andi a1, a0, 255
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bgeu s0, a1, .LBB50_1
> +; RV32I-NEXT: andi a0, a3, 255
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a0, .LBB50_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB50_2 Depth=1
> -; RV32I-NEXT: mv a2, s2
> +; RV32I-NEXT: mv a2, s3
> ; RV32I-NEXT: j .LBB50_1
> ; RV32I-NEXT: .LBB50_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -4354,27 +4354,27 @@ define i8 @atomicrmw_umin_i8_monotonic(i
> ;
> ; RV32IA-LABEL: atomicrmw_umin_i8_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a6, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB50_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: and a3, a4, a6
> +; RV32IA-NEXT: lr.w a4, (a6)
> +; RV32IA-NEXT: and a2, a4, a3
> ; RV32IA-NEXT: mv a5, a4
> -; RV32IA-NEXT: bgeu a1, a3, .LBB50_3
> +; RV32IA-NEXT: bgeu a1, a2, .LBB50_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB50_1 Depth=1
> ; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> +; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: .LBB50_3: # in Loop: Header=BB50_1 Depth=1
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB50_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umin_i8_monotonic:
> @@ -4385,33 +4385,33 @@ define i8 @atomicrmw_umin_i8_monotonic(i
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: andi s0, a1, 255
> -; RV64I-NEXT: addi s3, sp, 7
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lbu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: andi s1, a1, 255
> +; RV64I-NEXT: addi s2, sp, 7
> ; RV64I-NEXT: j .LBB50_2
> ; RV64I-NEXT: .LBB50_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB50_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a3, 7(sp)
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB50_4
> +; RV64I-NEXT: lb a3, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB50_4
> ; RV64I-NEXT: .LBB50_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: andi a1, a0, 255
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s0, a1, .LBB50_1
> +; RV64I-NEXT: andi a0, a3, 255
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a0, .LBB50_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB50_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB50_1
> ; RV64I-NEXT: .LBB50_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -4422,27 +4422,27 @@ define i8 @atomicrmw_umin_i8_monotonic(i
> ;
> ; RV64IA-LABEL: atomicrmw_umin_i8_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a6, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB50_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: and a3, a4, a6
> +; RV64IA-NEXT: lr.w a4, (a6)
> +; RV64IA-NEXT: and a2, a4, a3
> ; RV64IA-NEXT: mv a5, a4
> -; RV64IA-NEXT: bgeu a1, a3, .LBB50_3
> +; RV64IA-NEXT: bgeu a1, a2, .LBB50_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB50_1 Depth=1
> ; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> +; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: .LBB50_3: # in Loop: Header=BB50_1 Depth=1
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB50_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umin i8* %a, i8 %b monotonic
> ret i8 %1
> @@ -4457,33 +4457,33 @@ define i8 @atomicrmw_umin_i8_acquire(i8
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: andi s0, a1, 255
> -; RV32I-NEXT: addi s3, sp, 11
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lbu a3, 0(a0)
> +; RV32I-NEXT: mv s3, a1
> +; RV32I-NEXT: andi s1, a1, 255
> +; RV32I-NEXT: addi s2, sp, 11
> ; RV32I-NEXT: j .LBB51_2
> ; RV32I-NEXT: .LBB51_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB51_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a3, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 2
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB51_4
> +; RV32I-NEXT: lb a3, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB51_4
> ; RV32I-NEXT: .LBB51_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: andi a1, a0, 255
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bgeu s0, a1, .LBB51_1
> +; RV32I-NEXT: andi a0, a3, 255
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a0, .LBB51_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB51_2 Depth=1
> -; RV32I-NEXT: mv a2, s2
> +; RV32I-NEXT: mv a2, s3
> ; RV32I-NEXT: j .LBB51_1
> ; RV32I-NEXT: .LBB51_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -4494,27 +4494,27 @@ define i8 @atomicrmw_umin_i8_acquire(i8
> ;
> ; RV32IA-LABEL: atomicrmw_umin_i8_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a6, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB51_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a3, a4, a6
> +; RV32IA-NEXT: lr.w.aq a4, (a6)
> +; RV32IA-NEXT: and a2, a4, a3
> ; RV32IA-NEXT: mv a5, a4
> -; RV32IA-NEXT: bgeu a1, a3, .LBB51_3
> +; RV32IA-NEXT: bgeu a1, a2, .LBB51_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB51_1 Depth=1
> ; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> +; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: .LBB51_3: # in Loop: Header=BB51_1 Depth=1
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB51_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umin_i8_acquire:
> @@ -4525,33 +4525,33 @@ define i8 @atomicrmw_umin_i8_acquire(i8
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: andi s0, a1, 255
> -; RV64I-NEXT: addi s3, sp, 7
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lbu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: andi s1, a1, 255
> +; RV64I-NEXT: addi s2, sp, 7
> ; RV64I-NEXT: j .LBB51_2
> ; RV64I-NEXT: .LBB51_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB51_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a3, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB51_4
> +; RV64I-NEXT: lb a3, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB51_4
> ; RV64I-NEXT: .LBB51_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: andi a1, a0, 255
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s0, a1, .LBB51_1
> +; RV64I-NEXT: andi a0, a3, 255
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a0, .LBB51_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB51_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB51_1
> ; RV64I-NEXT: .LBB51_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -4562,27 +4562,27 @@ define i8 @atomicrmw_umin_i8_acquire(i8
> ;
> ; RV64IA-LABEL: atomicrmw_umin_i8_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a6, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB51_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a3, a4, a6
> +; RV64IA-NEXT: lr.w.aq a4, (a6)
> +; RV64IA-NEXT: and a2, a4, a3
> ; RV64IA-NEXT: mv a5, a4
> -; RV64IA-NEXT: bgeu a1, a3, .LBB51_3
> +; RV64IA-NEXT: bgeu a1, a2, .LBB51_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB51_1 Depth=1
> ; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> +; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: .LBB51_3: # in Loop: Header=BB51_1 Depth=1
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB51_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umin i8* %a, i8 %b acquire
> ret i8 %1
> @@ -4597,33 +4597,33 @@ define i8 @atomicrmw_umin_i8_release(i8
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: andi s0, a1, 255
> -; RV32I-NEXT: addi s3, sp, 11
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lbu a3, 0(a0)
> +; RV32I-NEXT: mv s3, a1
> +; RV32I-NEXT: andi s1, a1, 255
> +; RV32I-NEXT: addi s2, sp, 11
> ; RV32I-NEXT: j .LBB52_2
> ; RV32I-NEXT: .LBB52_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB52_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a3, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 3
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB52_4
> +; RV32I-NEXT: lb a3, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB52_4
> ; RV32I-NEXT: .LBB52_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: andi a1, a0, 255
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bgeu s0, a1, .LBB52_1
> +; RV32I-NEXT: andi a0, a3, 255
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a0, .LBB52_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB52_2 Depth=1
> -; RV32I-NEXT: mv a2, s2
> +; RV32I-NEXT: mv a2, s3
> ; RV32I-NEXT: j .LBB52_1
> ; RV32I-NEXT: .LBB52_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -4634,27 +4634,27 @@ define i8 @atomicrmw_umin_i8_release(i8
> ;
> ; RV32IA-LABEL: atomicrmw_umin_i8_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a6, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB52_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: and a3, a4, a6
> +; RV32IA-NEXT: lr.w a4, (a6)
> +; RV32IA-NEXT: and a2, a4, a3
> ; RV32IA-NEXT: mv a5, a4
> -; RV32IA-NEXT: bgeu a1, a3, .LBB52_3
> +; RV32IA-NEXT: bgeu a1, a2, .LBB52_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB52_1 Depth=1
> ; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> +; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: .LBB52_3: # in Loop: Header=BB52_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB52_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umin_i8_release:
> @@ -4665,33 +4665,33 @@ define i8 @atomicrmw_umin_i8_release(i8
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: andi s0, a1, 255
> -; RV64I-NEXT: addi s3, sp, 7
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lbu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: andi s1, a1, 255
> +; RV64I-NEXT: addi s2, sp, 7
> ; RV64I-NEXT: j .LBB52_2
> ; RV64I-NEXT: .LBB52_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB52_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a3, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 3
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB52_4
> +; RV64I-NEXT: lb a3, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB52_4
> ; RV64I-NEXT: .LBB52_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: andi a1, a0, 255
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s0, a1, .LBB52_1
> +; RV64I-NEXT: andi a0, a3, 255
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a0, .LBB52_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB52_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB52_1
> ; RV64I-NEXT: .LBB52_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -4702,27 +4702,27 @@ define i8 @atomicrmw_umin_i8_release(i8
> ;
> ; RV64IA-LABEL: atomicrmw_umin_i8_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a6, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB52_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: and a3, a4, a6
> +; RV64IA-NEXT: lr.w a4, (a6)
> +; RV64IA-NEXT: and a2, a4, a3
> ; RV64IA-NEXT: mv a5, a4
> -; RV64IA-NEXT: bgeu a1, a3, .LBB52_3
> +; RV64IA-NEXT: bgeu a1, a2, .LBB52_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB52_1 Depth=1
> ; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> +; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: .LBB52_3: # in Loop: Header=BB52_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB52_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umin i8* %a, i8 %b release
> ret i8 %1
> @@ -4737,33 +4737,33 @@ define i8 @atomicrmw_umin_i8_acq_rel(i8
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: andi s0, a1, 255
> -; RV32I-NEXT: addi s3, sp, 11
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lbu a3, 0(a0)
> +; RV32I-NEXT: mv s3, a1
> +; RV32I-NEXT: andi s1, a1, 255
> +; RV32I-NEXT: addi s2, sp, 11
> ; RV32I-NEXT: j .LBB53_2
> ; RV32I-NEXT: .LBB53_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB53_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a3, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 4
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB53_4
> +; RV32I-NEXT: lb a3, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB53_4
> ; RV32I-NEXT: .LBB53_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: andi a1, a0, 255
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bgeu s0, a1, .LBB53_1
> +; RV32I-NEXT: andi a0, a3, 255
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a0, .LBB53_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB53_2 Depth=1
> -; RV32I-NEXT: mv a2, s2
> +; RV32I-NEXT: mv a2, s3
> ; RV32I-NEXT: j .LBB53_1
> ; RV32I-NEXT: .LBB53_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -4774,27 +4774,27 @@ define i8 @atomicrmw_umin_i8_acq_rel(i8
> ;
> ; RV32IA-LABEL: atomicrmw_umin_i8_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a6, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB53_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a3, a4, a6
> +; RV32IA-NEXT: lr.w.aq a4, (a6)
> +; RV32IA-NEXT: and a2, a4, a3
> ; RV32IA-NEXT: mv a5, a4
> -; RV32IA-NEXT: bgeu a1, a3, .LBB53_3
> +; RV32IA-NEXT: bgeu a1, a2, .LBB53_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB53_1 Depth=1
> ; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> +; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: .LBB53_3: # in Loop: Header=BB53_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB53_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umin_i8_acq_rel:
> @@ -4805,33 +4805,33 @@ define i8 @atomicrmw_umin_i8_acq_rel(i8
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: andi s0, a1, 255
> -; RV64I-NEXT: addi s3, sp, 7
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lbu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: andi s1, a1, 255
> +; RV64I-NEXT: addi s2, sp, 7
> ; RV64I-NEXT: j .LBB53_2
> ; RV64I-NEXT: .LBB53_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB53_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a3, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB53_4
> +; RV64I-NEXT: lb a3, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB53_4
> ; RV64I-NEXT: .LBB53_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: andi a1, a0, 255
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s0, a1, .LBB53_1
> +; RV64I-NEXT: andi a0, a3, 255
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a0, .LBB53_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB53_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB53_1
> ; RV64I-NEXT: .LBB53_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -4842,27 +4842,27 @@ define i8 @atomicrmw_umin_i8_acq_rel(i8
> ;
> ; RV64IA-LABEL: atomicrmw_umin_i8_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a6, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB53_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a3, a4, a6
> +; RV64IA-NEXT: lr.w.aq a4, (a6)
> +; RV64IA-NEXT: and a2, a4, a3
> ; RV64IA-NEXT: mv a5, a4
> -; RV64IA-NEXT: bgeu a1, a3, .LBB53_3
> +; RV64IA-NEXT: bgeu a1, a2, .LBB53_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB53_1 Depth=1
> ; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> +; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: .LBB53_3: # in Loop: Header=BB53_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB53_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umin i8* %a, i8 %b acq_rel
> ret i8 %1
> @@ -4877,33 +4877,33 @@ define i8 @atomicrmw_umin_i8_seq_cst(i8
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lbu a0, 0(a0)
> -; RV32I-NEXT: andi s0, a1, 255
> -; RV32I-NEXT: addi s3, sp, 11
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lbu a3, 0(a0)
> +; RV32I-NEXT: mv s3, a1
> +; RV32I-NEXT: andi s1, a1, 255
> +; RV32I-NEXT: addi s2, sp, 11
> ; RV32I-NEXT: j .LBB54_2
> ; RV32I-NEXT: .LBB54_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB54_2 Depth=1
> -; RV32I-NEXT: sb a0, 11(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sb a3, 11(sp)
> ; RV32I-NEXT: addi a3, zero, 5
> ; RV32I-NEXT: addi a4, zero, 5
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_1
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lb a0, 11(sp)
> -; RV32I-NEXT: bnez a1, .LBB54_4
> +; RV32I-NEXT: lb a3, 11(sp)
> +; RV32I-NEXT: bnez a0, .LBB54_4
> ; RV32I-NEXT: .LBB54_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: andi a1, a0, 255
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bgeu s0, a1, .LBB54_1
> +; RV32I-NEXT: andi a0, a3, 255
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a0, .LBB54_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB54_2 Depth=1
> -; RV32I-NEXT: mv a2, s2
> +; RV32I-NEXT: mv a2, s3
> ; RV32I-NEXT: j .LBB54_1
> ; RV32I-NEXT: .LBB54_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -4914,27 +4914,27 @@ define i8 @atomicrmw_umin_i8_seq_cst(i8
> ;
> ; RV32IA-LABEL: atomicrmw_umin_i8_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> ; RV32IA-NEXT: addi a3, zero, 255
> -; RV32IA-NEXT: sll a6, a3, a2
> +; RV32IA-NEXT: sll a3, a3, a0
> ; RV32IA-NEXT: andi a1, a1, 255
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB54_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV32IA-NEXT: and a3, a4, a6
> +; RV32IA-NEXT: lr.w.aqrl a4, (a6)
> +; RV32IA-NEXT: and a2, a4, a3
> ; RV32IA-NEXT: mv a5, a4
> -; RV32IA-NEXT: bgeu a1, a3, .LBB54_3
> +; RV32IA-NEXT: bgeu a1, a2, .LBB54_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB54_1 Depth=1
> ; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> +; RV32IA-NEXT: and a5, a5, a3
> ; RV32IA-NEXT: xor a5, a4, a5
> ; RV32IA-NEXT: .LBB54_3: # in Loop: Header=BB54_1 Depth=1
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB54_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a2
> +; RV32IA-NEXT: srl a0, a4, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umin_i8_seq_cst:
> @@ -4945,33 +4945,33 @@ define i8 @atomicrmw_umin_i8_seq_cst(i8
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lbu a0, 0(a0)
> -; RV64I-NEXT: andi s0, a1, 255
> -; RV64I-NEXT: addi s3, sp, 7
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lbu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: andi s1, a1, 255
> +; RV64I-NEXT: addi s2, sp, 7
> ; RV64I-NEXT: j .LBB54_2
> ; RV64I-NEXT: .LBB54_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB54_2 Depth=1
> -; RV64I-NEXT: sb a0, 7(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sb a3, 7(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_1
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lb a0, 7(sp)
> -; RV64I-NEXT: bnez a1, .LBB54_4
> +; RV64I-NEXT: lb a3, 7(sp)
> +; RV64I-NEXT: bnez a0, .LBB54_4
> ; RV64I-NEXT: .LBB54_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: andi a1, a0, 255
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s0, a1, .LBB54_1
> +; RV64I-NEXT: andi a0, a3, 255
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a0, .LBB54_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB54_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB54_1
> ; RV64I-NEXT: .LBB54_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -4982,27 +4982,27 @@ define i8 @atomicrmw_umin_i8_seq_cst(i8
> ;
> ; RV64IA-LABEL: atomicrmw_umin_i8_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> ; RV64IA-NEXT: addi a3, zero, 255
> -; RV64IA-NEXT: sllw a6, a3, a2
> +; RV64IA-NEXT: sllw a3, a3, a0
> ; RV64IA-NEXT: andi a1, a1, 255
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB54_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV64IA-NEXT: and a3, a4, a6
> +; RV64IA-NEXT: lr.w.aqrl a4, (a6)
> +; RV64IA-NEXT: and a2, a4, a3
> ; RV64IA-NEXT: mv a5, a4
> -; RV64IA-NEXT: bgeu a1, a3, .LBB54_3
> +; RV64IA-NEXT: bgeu a1, a2, .LBB54_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB54_1 Depth=1
> ; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> +; RV64IA-NEXT: and a5, a5, a3
> ; RV64IA-NEXT: xor a5, a4, a5
> ; RV64IA-NEXT: .LBB54_3: # in Loop: Header=BB54_1 Depth=1
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB54_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a2
> +; RV64IA-NEXT: srlw a0, a4, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umin i8* %a, i8 %b seq_cst
> ret i8 %1
> @@ -5021,24 +5021,24 @@ define i16 @atomicrmw_xchg_i16_monotonic
> ;
> ; RV32IA-LABEL: atomicrmw_xchg_i16_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB55_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> +; RV32IA-NEXT: lr.w a3, (a2)
> ; RV32IA-NEXT: add a5, zero, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB55_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xchg_i16_monotonic:
> @@ -5053,24 +5053,24 @@ define i16 @atomicrmw_xchg_i16_monotonic
> ;
> ; RV64IA-LABEL: atomicrmw_xchg_i16_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB55_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> +; RV64IA-NEXT: lr.w a3, (a2)
> ; RV64IA-NEXT: add a5, zero, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB55_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xchg i16* %a, i16 %b monotonic
> ret i16 %1
> @@ -5089,24 +5089,24 @@ define i16 @atomicrmw_xchg_i16_acquire(i
> ;
> ; RV32IA-LABEL: atomicrmw_xchg_i16_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB56_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> +; RV32IA-NEXT: lr.w.aq a3, (a2)
> ; RV32IA-NEXT: add a5, zero, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB56_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xchg_i16_acquire:
> @@ -5121,24 +5121,24 @@ define i16 @atomicrmw_xchg_i16_acquire(i
> ;
> ; RV64IA-LABEL: atomicrmw_xchg_i16_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB56_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> +; RV64IA-NEXT: lr.w.aq a3, (a2)
> ; RV64IA-NEXT: add a5, zero, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB56_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xchg i16* %a, i16 %b acquire
> ret i16 %1
> @@ -5157,24 +5157,24 @@ define i16 @atomicrmw_xchg_i16_release(i
> ;
> ; RV32IA-LABEL: atomicrmw_xchg_i16_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB57_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> +; RV32IA-NEXT: lr.w a3, (a2)
> ; RV32IA-NEXT: add a5, zero, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB57_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xchg_i16_release:
> @@ -5189,24 +5189,24 @@ define i16 @atomicrmw_xchg_i16_release(i
> ;
> ; RV64IA-LABEL: atomicrmw_xchg_i16_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB57_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> +; RV64IA-NEXT: lr.w a3, (a2)
> ; RV64IA-NEXT: add a5, zero, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB57_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xchg i16* %a, i16 %b release
> ret i16 %1
> @@ -5225,24 +5225,24 @@ define i16 @atomicrmw_xchg_i16_acq_rel(i
> ;
> ; RV32IA-LABEL: atomicrmw_xchg_i16_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB58_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> +; RV32IA-NEXT: lr.w.aq a3, (a2)
> ; RV32IA-NEXT: add a5, zero, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB58_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xchg_i16_acq_rel:
> @@ -5257,24 +5257,24 @@ define i16 @atomicrmw_xchg_i16_acq_rel(i
> ;
> ; RV64IA-LABEL: atomicrmw_xchg_i16_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB58_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> +; RV64IA-NEXT: lr.w.aq a3, (a2)
> ; RV64IA-NEXT: add a5, zero, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB58_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xchg i16* %a, i16 %b acq_rel
> ret i16 %1
> @@ -5293,24 +5293,24 @@ define i16 @atomicrmw_xchg_i16_seq_cst(i
> ;
> ; RV32IA-LABEL: atomicrmw_xchg_i16_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB59_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> +; RV32IA-NEXT: lr.w.aqrl a3, (a2)
> ; RV32IA-NEXT: add a5, zero, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB59_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xchg_i16_seq_cst:
> @@ -5325,24 +5325,24 @@ define i16 @atomicrmw_xchg_i16_seq_cst(i
> ;
> ; RV64IA-LABEL: atomicrmw_xchg_i16_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB59_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> +; RV64IA-NEXT: lr.w.aqrl a3, (a2)
> ; RV64IA-NEXT: add a5, zero, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB59_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xchg i16* %a, i16 %b seq_cst
> ret i16 %1
> @@ -5361,24 +5361,24 @@ define i16 @atomicrmw_add_i16_monotonic(
> ;
> ; RV32IA-LABEL: atomicrmw_add_i16_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB60_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: add a5, a4, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: lr.w a3, (a2)
> +; RV32IA-NEXT: add a5, a3, a1
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB60_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_add_i16_monotonic:
> @@ -5393,24 +5393,24 @@ define i16 @atomicrmw_add_i16_monotonic(
> ;
> ; RV64IA-LABEL: atomicrmw_add_i16_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB60_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: add a5, a4, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: lr.w a3, (a2)
> +; RV64IA-NEXT: add a5, a3, a1
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB60_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw add i16* %a, i16 %b monotonic
> ret i16 %1
> @@ -5429,24 +5429,24 @@ define i16 @atomicrmw_add_i16_acquire(i1
> ;
> ; RV32IA-LABEL: atomicrmw_add_i16_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB61_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: add a5, a4, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: lr.w.aq a3, (a2)
> +; RV32IA-NEXT: add a5, a3, a1
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB61_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_add_i16_acquire:
> @@ -5461,24 +5461,24 @@ define i16 @atomicrmw_add_i16_acquire(i1
> ;
> ; RV64IA-LABEL: atomicrmw_add_i16_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB61_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: add a5, a4, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: lr.w.aq a3, (a2)
> +; RV64IA-NEXT: add a5, a3, a1
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB61_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw add i16* %a, i16 %b acquire
> ret i16 %1
> @@ -5497,24 +5497,24 @@ define i16 @atomicrmw_add_i16_release(i1
> ;
> ; RV32IA-LABEL: atomicrmw_add_i16_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB62_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: add a5, a4, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: lr.w a3, (a2)
> +; RV32IA-NEXT: add a5, a3, a1
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB62_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_add_i16_release:
> @@ -5529,24 +5529,24 @@ define i16 @atomicrmw_add_i16_release(i1
> ;
> ; RV64IA-LABEL: atomicrmw_add_i16_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB62_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: add a5, a4, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: lr.w a3, (a2)
> +; RV64IA-NEXT: add a5, a3, a1
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB62_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw add i16* %a, i16 %b release
> ret i16 %1
> @@ -5565,24 +5565,24 @@ define i16 @atomicrmw_add_i16_acq_rel(i1
> ;
> ; RV32IA-LABEL: atomicrmw_add_i16_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB63_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: add a5, a4, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: lr.w.aq a3, (a2)
> +; RV32IA-NEXT: add a5, a3, a1
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB63_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_add_i16_acq_rel:
> @@ -5597,24 +5597,24 @@ define i16 @atomicrmw_add_i16_acq_rel(i1
> ;
> ; RV64IA-LABEL: atomicrmw_add_i16_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB63_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: add a5, a4, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: lr.w.aq a3, (a2)
> +; RV64IA-NEXT: add a5, a3, a1
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB63_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw add i16* %a, i16 %b acq_rel
> ret i16 %1
> @@ -5633,24 +5633,24 @@ define i16 @atomicrmw_add_i16_seq_cst(i1
> ;
> ; RV32IA-LABEL: atomicrmw_add_i16_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB64_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV32IA-NEXT: add a5, a4, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: lr.w.aqrl a3, (a2)
> +; RV32IA-NEXT: add a5, a3, a1
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB64_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_add_i16_seq_cst:
> @@ -5665,24 +5665,24 @@ define i16 @atomicrmw_add_i16_seq_cst(i1
> ;
> ; RV64IA-LABEL: atomicrmw_add_i16_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB64_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV64IA-NEXT: add a5, a4, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: lr.w.aqrl a3, (a2)
> +; RV64IA-NEXT: add a5, a3, a1
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB64_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw add i16* %a, i16 %b seq_cst
> ret i16 %1
> @@ -5701,24 +5701,24 @@ define i16 @atomicrmw_sub_i16_monotonic(
> ;
> ; RV32IA-LABEL: atomicrmw_sub_i16_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB65_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: sub a5, a4, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: lr.w a3, (a2)
> +; RV32IA-NEXT: sub a5, a3, a1
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB65_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_sub_i16_monotonic:
> @@ -5733,24 +5733,24 @@ define i16 @atomicrmw_sub_i16_monotonic(
> ;
> ; RV64IA-LABEL: atomicrmw_sub_i16_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB65_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: sub a5, a4, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: lr.w a3, (a2)
> +; RV64IA-NEXT: sub a5, a3, a1
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB65_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw sub i16* %a, i16 %b monotonic
> ret i16 %1
> @@ -5769,24 +5769,24 @@ define i16 @atomicrmw_sub_i16_acquire(i1
> ;
> ; RV32IA-LABEL: atomicrmw_sub_i16_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB66_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: sub a5, a4, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: lr.w.aq a3, (a2)
> +; RV32IA-NEXT: sub a5, a3, a1
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB66_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_sub_i16_acquire:
> @@ -5801,24 +5801,24 @@ define i16 @atomicrmw_sub_i16_acquire(i1
> ;
> ; RV64IA-LABEL: atomicrmw_sub_i16_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB66_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: sub a5, a4, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: lr.w.aq a3, (a2)
> +; RV64IA-NEXT: sub a5, a3, a1
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB66_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw sub i16* %a, i16 %b acquire
> ret i16 %1
> @@ -5837,24 +5837,24 @@ define i16 @atomicrmw_sub_i16_release(i1
> ;
> ; RV32IA-LABEL: atomicrmw_sub_i16_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB67_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: sub a5, a4, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: lr.w a3, (a2)
> +; RV32IA-NEXT: sub a5, a3, a1
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB67_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_sub_i16_release:
> @@ -5869,24 +5869,24 @@ define i16 @atomicrmw_sub_i16_release(i1
> ;
> ; RV64IA-LABEL: atomicrmw_sub_i16_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB67_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: sub a5, a4, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: lr.w a3, (a2)
> +; RV64IA-NEXT: sub a5, a3, a1
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB67_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw sub i16* %a, i16 %b release
> ret i16 %1
> @@ -5905,24 +5905,24 @@ define i16 @atomicrmw_sub_i16_acq_rel(i1
> ;
> ; RV32IA-LABEL: atomicrmw_sub_i16_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB68_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: sub a5, a4, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: lr.w.aq a3, (a2)
> +; RV32IA-NEXT: sub a5, a3, a1
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB68_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_sub_i16_acq_rel:
> @@ -5937,24 +5937,24 @@ define i16 @atomicrmw_sub_i16_acq_rel(i1
> ;
> ; RV64IA-LABEL: atomicrmw_sub_i16_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB68_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: sub a5, a4, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: lr.w.aq a3, (a2)
> +; RV64IA-NEXT: sub a5, a3, a1
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB68_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw sub i16* %a, i16 %b acq_rel
> ret i16 %1
> @@ -5973,24 +5973,24 @@ define i16 @atomicrmw_sub_i16_seq_cst(i1
> ;
> ; RV32IA-LABEL: atomicrmw_sub_i16_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB69_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV32IA-NEXT: sub a5, a4, a1
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: lr.w.aqrl a3, (a2)
> +; RV32IA-NEXT: sub a5, a3, a1
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB69_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_sub_i16_seq_cst:
> @@ -6005,24 +6005,24 @@ define i16 @atomicrmw_sub_i16_seq_cst(i1
> ;
> ; RV64IA-LABEL: atomicrmw_sub_i16_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB69_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV64IA-NEXT: sub a5, a4, a1
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: lr.w.aqrl a3, (a2)
> +; RV64IA-NEXT: sub a5, a3, a1
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB69_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw sub i16* %a, i16 %b seq_cst
> ret i16 %1
> @@ -6041,18 +6041,18 @@ define i16 @atomicrmw_and_i16_monotonic(
> ;
> ; RV32IA-LABEL: atomicrmw_and_i16_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: not a2, a2
> -; RV32IA-NEXT: or a1, a2, a1
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoand.w a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a3
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: not a4, a4
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: or a1, a4, a1
> +; RV32IA-NEXT: amoand.w a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_and_i16_monotonic:
> @@ -6067,18 +6067,18 @@ define i16 @atomicrmw_and_i16_monotonic(
> ;
> ; RV64IA-LABEL: atomicrmw_and_i16_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: not a2, a2
> -; RV64IA-NEXT: or a1, a2, a1
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoand.w a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a3
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: not a4, a4
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: or a1, a4, a1
> +; RV64IA-NEXT: amoand.w a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw and i16* %a, i16 %b monotonic
> ret i16 %1
> @@ -6097,18 +6097,18 @@ define i16 @atomicrmw_and_i16_acquire(i1
> ;
> ; RV32IA-LABEL: atomicrmw_and_i16_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: not a2, a2
> -; RV32IA-NEXT: or a1, a2, a1
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoand.w.aq a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a3
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: not a4, a4
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: or a1, a4, a1
> +; RV32IA-NEXT: amoand.w.aq a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_and_i16_acquire:
> @@ -6123,18 +6123,18 @@ define i16 @atomicrmw_and_i16_acquire(i1
> ;
> ; RV64IA-LABEL: atomicrmw_and_i16_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: not a2, a2
> -; RV64IA-NEXT: or a1, a2, a1
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoand.w.aq a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a3
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: not a4, a4
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: or a1, a4, a1
> +; RV64IA-NEXT: amoand.w.aq a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw and i16* %a, i16 %b acquire
> ret i16 %1
> @@ -6153,18 +6153,18 @@ define i16 @atomicrmw_and_i16_release(i1
> ;
> ; RV32IA-LABEL: atomicrmw_and_i16_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: not a2, a2
> -; RV32IA-NEXT: or a1, a2, a1
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoand.w.rl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a3
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: not a4, a4
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: or a1, a4, a1
> +; RV32IA-NEXT: amoand.w.rl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_and_i16_release:
> @@ -6179,18 +6179,18 @@ define i16 @atomicrmw_and_i16_release(i1
> ;
> ; RV64IA-LABEL: atomicrmw_and_i16_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: not a2, a2
> -; RV64IA-NEXT: or a1, a2, a1
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoand.w.rl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a3
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: not a4, a4
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: or a1, a4, a1
> +; RV64IA-NEXT: amoand.w.rl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw and i16* %a, i16 %b release
> ret i16 %1
> @@ -6209,18 +6209,18 @@ define i16 @atomicrmw_and_i16_acq_rel(i1
> ;
> ; RV32IA-LABEL: atomicrmw_and_i16_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: not a2, a2
> -; RV32IA-NEXT: or a1, a2, a1
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoand.w.aqrl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a3
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: not a4, a4
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: or a1, a4, a1
> +; RV32IA-NEXT: amoand.w.aqrl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_and_i16_acq_rel:
> @@ -6235,18 +6235,18 @@ define i16 @atomicrmw_and_i16_acq_rel(i1
> ;
> ; RV64IA-LABEL: atomicrmw_and_i16_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: not a2, a2
> -; RV64IA-NEXT: or a1, a2, a1
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoand.w.aqrl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a3
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: not a4, a4
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: or a1, a4, a1
> +; RV64IA-NEXT: amoand.w.aqrl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw and i16* %a, i16 %b acq_rel
> ret i16 %1
> @@ -6265,18 +6265,18 @@ define i16 @atomicrmw_and_i16_seq_cst(i1
> ;
> ; RV32IA-LABEL: atomicrmw_and_i16_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: not a2, a2
> -; RV32IA-NEXT: or a1, a2, a1
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoand.w.aqrl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a3
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: not a4, a4
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: or a1, a4, a1
> +; RV32IA-NEXT: amoand.w.aqrl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_and_i16_seq_cst:
> @@ -6291,18 +6291,18 @@ define i16 @atomicrmw_and_i16_seq_cst(i1
> ;
> ; RV64IA-LABEL: atomicrmw_and_i16_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: not a2, a2
> -; RV64IA-NEXT: or a1, a2, a1
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoand.w.aqrl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a3
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: not a4, a4
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: or a1, a4, a1
> +; RV64IA-NEXT: amoand.w.aqrl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw and i16* %a, i16 %b seq_cst
> ret i16 %1
> @@ -6321,25 +6321,25 @@ define i16 @atomicrmw_nand_i16_monotonic
> ;
> ; RV32IA-LABEL: atomicrmw_nand_i16_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB75_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a1
> +; RV32IA-NEXT: lr.w a3, (a2)
> +; RV32IA-NEXT: and a5, a3, a1
> ; RV32IA-NEXT: not a5, a5
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB75_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_nand_i16_monotonic:
> @@ -6354,25 +6354,25 @@ define i16 @atomicrmw_nand_i16_monotonic
> ;
> ; RV64IA-LABEL: atomicrmw_nand_i16_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB75_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a1
> +; RV64IA-NEXT: lr.w a3, (a2)
> +; RV64IA-NEXT: and a5, a3, a1
> ; RV64IA-NEXT: not a5, a5
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB75_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw nand i16* %a, i16 %b monotonic
> ret i16 %1
> @@ -6391,25 +6391,25 @@ define i16 @atomicrmw_nand_i16_acquire(i
> ;
> ; RV32IA-LABEL: atomicrmw_nand_i16_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB76_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a1
> +; RV32IA-NEXT: lr.w.aq a3, (a2)
> +; RV32IA-NEXT: and a5, a3, a1
> ; RV32IA-NEXT: not a5, a5
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB76_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_nand_i16_acquire:
> @@ -6424,25 +6424,25 @@ define i16 @atomicrmw_nand_i16_acquire(i
> ;
> ; RV64IA-LABEL: atomicrmw_nand_i16_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB76_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a1
> +; RV64IA-NEXT: lr.w.aq a3, (a2)
> +; RV64IA-NEXT: and a5, a3, a1
> ; RV64IA-NEXT: not a5, a5
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB76_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw nand i16* %a, i16 %b acquire
> ret i16 %1
> @@ -6461,25 +6461,25 @@ define i16 @atomicrmw_nand_i16_release(i
> ;
> ; RV32IA-LABEL: atomicrmw_nand_i16_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB77_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a1
> +; RV32IA-NEXT: lr.w a3, (a2)
> +; RV32IA-NEXT: and a5, a3, a1
> ; RV32IA-NEXT: not a5, a5
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB77_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_nand_i16_release:
> @@ -6494,25 +6494,25 @@ define i16 @atomicrmw_nand_i16_release(i
> ;
> ; RV64IA-LABEL: atomicrmw_nand_i16_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB77_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a1
> +; RV64IA-NEXT: lr.w a3, (a2)
> +; RV64IA-NEXT: and a5, a3, a1
> ; RV64IA-NEXT: not a5, a5
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB77_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw nand i16* %a, i16 %b release
> ret i16 %1
> @@ -6531,25 +6531,25 @@ define i16 @atomicrmw_nand_i16_acq_rel(i
> ;
> ; RV32IA-LABEL: atomicrmw_nand_i16_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB78_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a1
> +; RV32IA-NEXT: lr.w.aq a3, (a2)
> +; RV32IA-NEXT: and a5, a3, a1
> ; RV32IA-NEXT: not a5, a5
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB78_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_nand_i16_acq_rel:
> @@ -6564,25 +6564,25 @@ define i16 @atomicrmw_nand_i16_acq_rel(i
> ;
> ; RV64IA-LABEL: atomicrmw_nand_i16_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB78_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a1
> +; RV64IA-NEXT: lr.w.aq a3, (a2)
> +; RV64IA-NEXT: and a5, a3, a1
> ; RV64IA-NEXT: not a5, a5
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB78_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw nand i16* %a, i16 %b acq_rel
> ret i16 %1
> @@ -6601,25 +6601,25 @@ define i16 @atomicrmw_nand_i16_seq_cst(i
> ;
> ; RV32IA-LABEL: atomicrmw_nand_i16_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a2, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB79_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV32IA-NEXT: and a5, a4, a1
> +; RV32IA-NEXT: lr.w.aqrl a3, (a2)
> +; RV32IA-NEXT: and a5, a3, a1
> ; RV32IA-NEXT: not a5, a5
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: and a5, a5, a2
> -; RV32IA-NEXT: xor a5, a4, a5
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV32IA-NEXT: bnez a5, .LBB79_1
> ; RV32IA-NEXT: # %bb.2:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_nand_i16_seq_cst:
> @@ -6634,25 +6634,25 @@ define i16 @atomicrmw_nand_i16_seq_cst(i
> ;
> ; RV64IA-LABEL: atomicrmw_nand_i16_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a2, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB79_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV64IA-NEXT: and a5, a4, a1
> +; RV64IA-NEXT: lr.w.aqrl a3, (a2)
> +; RV64IA-NEXT: and a5, a3, a1
> ; RV64IA-NEXT: not a5, a5
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: and a5, a5, a2
> -; RV64IA-NEXT: xor a5, a4, a5
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a2)
> ; RV64IA-NEXT: bnez a5, .LBB79_1
> ; RV64IA-NEXT: # %bb.2:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw nand i16* %a, i16 %b seq_cst
> ret i16 %1
> @@ -6671,15 +6671,15 @@ define i16 @atomicrmw_or_i16_monotonic(i
> ;
> ; RV32IA-LABEL: atomicrmw_or_i16_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoor.w a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoor.w a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_or_i16_monotonic:
> @@ -6694,15 +6694,15 @@ define i16 @atomicrmw_or_i16_monotonic(i
> ;
> ; RV64IA-LABEL: atomicrmw_or_i16_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoor.w a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoor.w a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw or i16* %a, i16 %b monotonic
> ret i16 %1
> @@ -6721,15 +6721,15 @@ define i16 @atomicrmw_or_i16_acquire(i16
> ;
> ; RV32IA-LABEL: atomicrmw_or_i16_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoor.w.aq a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoor.w.aq a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_or_i16_acquire:
> @@ -6744,15 +6744,15 @@ define i16 @atomicrmw_or_i16_acquire(i16
> ;
> ; RV64IA-LABEL: atomicrmw_or_i16_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoor.w.aq a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoor.w.aq a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw or i16* %a, i16 %b acquire
> ret i16 %1
> @@ -6771,15 +6771,15 @@ define i16 @atomicrmw_or_i16_release(i16
> ;
> ; RV32IA-LABEL: atomicrmw_or_i16_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoor.w.rl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoor.w.rl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_or_i16_release:
> @@ -6794,15 +6794,15 @@ define i16 @atomicrmw_or_i16_release(i16
> ;
> ; RV64IA-LABEL: atomicrmw_or_i16_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoor.w.rl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoor.w.rl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw or i16* %a, i16 %b release
> ret i16 %1
> @@ -6821,15 +6821,15 @@ define i16 @atomicrmw_or_i16_acq_rel(i16
> ;
> ; RV32IA-LABEL: atomicrmw_or_i16_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoor.w.aqrl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoor.w.aqrl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_or_i16_acq_rel:
> @@ -6844,15 +6844,15 @@ define i16 @atomicrmw_or_i16_acq_rel(i16
> ;
> ; RV64IA-LABEL: atomicrmw_or_i16_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoor.w.aqrl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoor.w.aqrl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw or i16* %a, i16 %b acq_rel
> ret i16 %1
> @@ -6871,15 +6871,15 @@ define i16 @atomicrmw_or_i16_seq_cst(i16
> ;
> ; RV32IA-LABEL: atomicrmw_or_i16_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoor.w.aqrl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoor.w.aqrl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_or_i16_seq_cst:
> @@ -6894,15 +6894,15 @@ define i16 @atomicrmw_or_i16_seq_cst(i16
> ;
> ; RV64IA-LABEL: atomicrmw_or_i16_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoor.w.aqrl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoor.w.aqrl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw or i16* %a, i16 %b seq_cst
> ret i16 %1
> @@ -6921,15 +6921,15 @@ define i16 @atomicrmw_xor_i16_monotonic(
> ;
> ; RV32IA-LABEL: atomicrmw_xor_i16_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoxor.w a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoxor.w a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xor_i16_monotonic:
> @@ -6944,15 +6944,15 @@ define i16 @atomicrmw_xor_i16_monotonic(
> ;
> ; RV64IA-LABEL: atomicrmw_xor_i16_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoxor.w a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoxor.w a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xor i16* %a, i16 %b monotonic
> ret i16 %1
> @@ -6971,15 +6971,15 @@ define i16 @atomicrmw_xor_i16_acquire(i1
> ;
> ; RV32IA-LABEL: atomicrmw_xor_i16_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoxor.w.aq a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoxor.w.aq a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xor_i16_acquire:
> @@ -6994,15 +6994,15 @@ define i16 @atomicrmw_xor_i16_acquire(i1
> ;
> ; RV64IA-LABEL: atomicrmw_xor_i16_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoxor.w.aq a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoxor.w.aq a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xor i16* %a, i16 %b acquire
> ret i16 %1
> @@ -7021,15 +7021,15 @@ define i16 @atomicrmw_xor_i16_release(i1
> ;
> ; RV32IA-LABEL: atomicrmw_xor_i16_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoxor.w.rl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoxor.w.rl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xor_i16_release:
> @@ -7044,15 +7044,15 @@ define i16 @atomicrmw_xor_i16_release(i1
> ;
> ; RV64IA-LABEL: atomicrmw_xor_i16_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoxor.w.rl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoxor.w.rl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xor i16* %a, i16 %b release
> ret i16 %1
> @@ -7071,15 +7071,15 @@ define i16 @atomicrmw_xor_i16_acq_rel(i1
> ;
> ; RV32IA-LABEL: atomicrmw_xor_i16_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoxor.w.aqrl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoxor.w.aqrl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xor_i16_acq_rel:
> @@ -7094,15 +7094,15 @@ define i16 @atomicrmw_xor_i16_acq_rel(i1
> ;
> ; RV64IA-LABEL: atomicrmw_xor_i16_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoxor.w.aqrl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoxor.w.aqrl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xor i16* %a, i16 %b acq_rel
> ret i16 %1
> @@ -7121,15 +7121,15 @@ define i16 @atomicrmw_xor_i16_seq_cst(i1
> ;
> ; RV32IA-LABEL: atomicrmw_xor_i16_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> -; RV32IA-NEXT: amoxor.w.aqrl a0, a1, (a0)
> -; RV32IA-NEXT: srl a0, a0, a2
> +; RV32IA-NEXT: andi a2, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: amoxor.w.aqrl a1, a1, (a2)
> +; RV32IA-NEXT: srl a0, a1, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_xor_i16_seq_cst:
> @@ -7144,15 +7144,15 @@ define i16 @atomicrmw_xor_i16_seq_cst(i1
> ;
> ; RV64IA-LABEL: atomicrmw_xor_i16_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> -; RV64IA-NEXT: amoxor.w.aqrl a0, a1, (a0)
> -; RV64IA-NEXT: srlw a0, a0, a2
> +; RV64IA-NEXT: andi a2, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: amoxor.w.aqrl a1, a1, (a2)
> +; RV64IA-NEXT: srlw a0, a1, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw xor i16* %a, i16 %b seq_cst
> ret i16 %1
> @@ -7169,33 +7169,33 @@ define i16 @atomicrmw_max_i16_monotonic(
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 16
> -; RV32I-NEXT: srai s0, a1, 16
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 16
> +; RV32I-NEXT: srai s0, a0, 16
> ; RV32I-NEXT: addi s3, sp, 10
> ; RV32I-NEXT: j .LBB90_2
> ; RV32I-NEXT: .LBB90_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB90_2 Depth=1
> -; RV32I-NEXT: sh a0, 10(sp)
> +; RV32I-NEXT: sh a1, 10(sp)
> ; RV32I-NEXT: mv a0, s1
> ; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a3, zero
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 10(sp)
> -; RV32I-NEXT: bnez a1, .LBB90_4
> +; RV32I-NEXT: lh a1, 10(sp)
> +; RV32I-NEXT: bnez a0, .LBB90_4
> ; RV32I-NEXT: .LBB90_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 16
> -; RV32I-NEXT: srai a1, a1, 16
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: blt s0, a1, .LBB90_1
> +; RV32I-NEXT: slli a0, a1, 16
> +; RV32I-NEXT: srai a0, a0, 16
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: blt s0, a0, .LBB90_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB90_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB90_1
> ; RV32I-NEXT: .LBB90_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -7206,33 +7206,33 @@ define i16 @atomicrmw_max_i16_monotonic(
> ;
> ; RV32IA-LABEL: atomicrmw_max_i16_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 16
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: lui a4, 16
> -; RV32IA-NEXT: addi a4, a4, -1
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 16
> ; RV32IA-NEXT: srai a1, a1, 16
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 16
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB90_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a4, a1, .LBB90_3
> +; RV32IA-NEXT: lr.w a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a3, a1, .LBB90_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB90_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB90_3: # in Loop: Header=BB90_1 Depth=1
> -; RV32IA-NEXT: sc.w a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB90_1
> +; RV32IA-NEXT: sc.w a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB90_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_max_i16_monotonic:
> @@ -7245,33 +7245,33 @@ define i16 @atomicrmw_max_i16_monotonic(
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 48
> -; RV64I-NEXT: srai s0, a1, 48
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 48
> +; RV64I-NEXT: srai s0, a0, 48
> ; RV64I-NEXT: addi s3, sp, 6
> ; RV64I-NEXT: j .LBB90_2
> ; RV64I-NEXT: .LBB90_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB90_2 Depth=1
> -; RV64I-NEXT: sh a0, 6(sp)
> +; RV64I-NEXT: sh a1, 6(sp)
> ; RV64I-NEXT: mv a0, s1
> ; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 6(sp)
> -; RV64I-NEXT: bnez a1, .LBB90_4
> +; RV64I-NEXT: lh a1, 6(sp)
> +; RV64I-NEXT: bnez a0, .LBB90_4
> ; RV64I-NEXT: .LBB90_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 48
> -; RV64I-NEXT: srai a1, a1, 48
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB90_1
> +; RV64I-NEXT: slli a0, a1, 48
> +; RV64I-NEXT: srai a0, a0, 48
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: blt s0, a0, .LBB90_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB90_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB90_1
> ; RV64I-NEXT: .LBB90_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -7282,33 +7282,33 @@ define i16 @atomicrmw_max_i16_monotonic(
> ;
> ; RV64IA-LABEL: atomicrmw_max_i16_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 48
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: lui a4, 16
> -; RV64IA-NEXT: addiw a4, a4, -1
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 48
> ; RV64IA-NEXT: srai a1, a1, 48
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 48
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB90_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a4, a1, .LBB90_3
> +; RV64IA-NEXT: lr.w a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a3, a1, .LBB90_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB90_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB90_3: # in Loop: Header=BB90_1 Depth=1
> -; RV64IA-NEXT: sc.w a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB90_1
> +; RV64IA-NEXT: sc.w a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB90_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw max i16* %a, i16 %b monotonic
> ret i16 %1
> @@ -7325,33 +7325,33 @@ define i16 @atomicrmw_max_i16_acquire(i1
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 16
> -; RV32I-NEXT: srai s0, a1, 16
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 16
> +; RV32I-NEXT: srai s0, a0, 16
> ; RV32I-NEXT: addi s3, sp, 10
> ; RV32I-NEXT: j .LBB91_2
> ; RV32I-NEXT: .LBB91_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB91_2 Depth=1
> -; RV32I-NEXT: sh a0, 10(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sh a1, 10(sp)
> ; RV32I-NEXT: addi a3, zero, 2
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 10(sp)
> -; RV32I-NEXT: bnez a1, .LBB91_4
> +; RV32I-NEXT: lh a1, 10(sp)
> +; RV32I-NEXT: bnez a0, .LBB91_4
> ; RV32I-NEXT: .LBB91_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 16
> -; RV32I-NEXT: srai a1, a1, 16
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: blt s0, a1, .LBB91_1
> +; RV32I-NEXT: slli a0, a1, 16
> +; RV32I-NEXT: srai a0, a0, 16
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: blt s0, a0, .LBB91_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB91_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB91_1
> ; RV32I-NEXT: .LBB91_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -7362,33 +7362,33 @@ define i16 @atomicrmw_max_i16_acquire(i1
> ;
> ; RV32IA-LABEL: atomicrmw_max_i16_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 16
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: lui a4, 16
> -; RV32IA-NEXT: addi a4, a4, -1
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 16
> ; RV32IA-NEXT: srai a1, a1, 16
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 16
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB91_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a4, a1, .LBB91_3
> +; RV32IA-NEXT: lr.w.aq a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a3, a1, .LBB91_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB91_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB91_3: # in Loop: Header=BB91_1 Depth=1
> -; RV32IA-NEXT: sc.w a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB91_1
> +; RV32IA-NEXT: sc.w a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB91_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_max_i16_acquire:
> @@ -7401,33 +7401,33 @@ define i16 @atomicrmw_max_i16_acquire(i1
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 48
> -; RV64I-NEXT: srai s0, a1, 48
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 48
> +; RV64I-NEXT: srai s0, a0, 48
> ; RV64I-NEXT: addi s3, sp, 6
> ; RV64I-NEXT: j .LBB91_2
> ; RV64I-NEXT: .LBB91_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB91_2 Depth=1
> -; RV64I-NEXT: sh a0, 6(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sh a1, 6(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 6(sp)
> -; RV64I-NEXT: bnez a1, .LBB91_4
> +; RV64I-NEXT: lh a1, 6(sp)
> +; RV64I-NEXT: bnez a0, .LBB91_4
> ; RV64I-NEXT: .LBB91_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 48
> -; RV64I-NEXT: srai a1, a1, 48
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB91_1
> +; RV64I-NEXT: slli a0, a1, 48
> +; RV64I-NEXT: srai a0, a0, 48
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: blt s0, a0, .LBB91_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB91_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB91_1
> ; RV64I-NEXT: .LBB91_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -7438,33 +7438,33 @@ define i16 @atomicrmw_max_i16_acquire(i1
> ;
> ; RV64IA-LABEL: atomicrmw_max_i16_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 48
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: lui a4, 16
> -; RV64IA-NEXT: addiw a4, a4, -1
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 48
> ; RV64IA-NEXT: srai a1, a1, 48
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 48
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB91_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a4, a1, .LBB91_3
> +; RV64IA-NEXT: lr.w.aq a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a3, a1, .LBB91_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB91_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB91_3: # in Loop: Header=BB91_1 Depth=1
> -; RV64IA-NEXT: sc.w a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB91_1
> +; RV64IA-NEXT: sc.w a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB91_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw max i16* %a, i16 %b acquire
> ret i16 %1
> @@ -7481,33 +7481,33 @@ define i16 @atomicrmw_max_i16_release(i1
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 16
> -; RV32I-NEXT: srai s0, a1, 16
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 16
> +; RV32I-NEXT: srai s0, a0, 16
> ; RV32I-NEXT: addi s3, sp, 10
> ; RV32I-NEXT: j .LBB92_2
> ; RV32I-NEXT: .LBB92_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB92_2 Depth=1
> -; RV32I-NEXT: sh a0, 10(sp)
> +; RV32I-NEXT: sh a1, 10(sp)
> +; RV32I-NEXT: addi a3, zero, 3
> ; RV32I-NEXT: mv a0, s1
> ; RV32I-NEXT: mv a1, s3
> -; RV32I-NEXT: addi a3, zero, 3
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 10(sp)
> -; RV32I-NEXT: bnez a1, .LBB92_4
> +; RV32I-NEXT: lh a1, 10(sp)
> +; RV32I-NEXT: bnez a0, .LBB92_4
> ; RV32I-NEXT: .LBB92_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 16
> -; RV32I-NEXT: srai a1, a1, 16
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: blt s0, a1, .LBB92_1
> +; RV32I-NEXT: slli a0, a1, 16
> +; RV32I-NEXT: srai a0, a0, 16
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: blt s0, a0, .LBB92_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB92_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB92_1
> ; RV32I-NEXT: .LBB92_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -7518,33 +7518,33 @@ define i16 @atomicrmw_max_i16_release(i1
> ;
> ; RV32IA-LABEL: atomicrmw_max_i16_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 16
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: lui a4, 16
> -; RV32IA-NEXT: addi a4, a4, -1
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 16
> ; RV32IA-NEXT: srai a1, a1, 16
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 16
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB92_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a4, a1, .LBB92_3
> +; RV32IA-NEXT: lr.w a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a3, a1, .LBB92_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB92_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB92_3: # in Loop: Header=BB92_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB92_1
> +; RV32IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB92_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_max_i16_release:
> @@ -7557,33 +7557,33 @@ define i16 @atomicrmw_max_i16_release(i1
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 48
> -; RV64I-NEXT: srai s0, a1, 48
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 48
> +; RV64I-NEXT: srai s0, a0, 48
> ; RV64I-NEXT: addi s3, sp, 6
> ; RV64I-NEXT: j .LBB92_2
> ; RV64I-NEXT: .LBB92_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB92_2 Depth=1
> -; RV64I-NEXT: sh a0, 6(sp)
> +; RV64I-NEXT: sh a1, 6(sp)
> +; RV64I-NEXT: addi a3, zero, 3
> ; RV64I-NEXT: mv a0, s1
> ; RV64I-NEXT: mv a1, s3
> -; RV64I-NEXT: addi a3, zero, 3
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 6(sp)
> -; RV64I-NEXT: bnez a1, .LBB92_4
> +; RV64I-NEXT: lh a1, 6(sp)
> +; RV64I-NEXT: bnez a0, .LBB92_4
> ; RV64I-NEXT: .LBB92_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 48
> -; RV64I-NEXT: srai a1, a1, 48
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB92_1
> +; RV64I-NEXT: slli a0, a1, 48
> +; RV64I-NEXT: srai a0, a0, 48
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: blt s0, a0, .LBB92_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB92_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB92_1
> ; RV64I-NEXT: .LBB92_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -7594,33 +7594,33 @@ define i16 @atomicrmw_max_i16_release(i1
> ;
> ; RV64IA-LABEL: atomicrmw_max_i16_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 48
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: lui a4, 16
> -; RV64IA-NEXT: addiw a4, a4, -1
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 48
> ; RV64IA-NEXT: srai a1, a1, 48
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 48
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB92_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a4, a1, .LBB92_3
> +; RV64IA-NEXT: lr.w a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a3, a1, .LBB92_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB92_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB92_3: # in Loop: Header=BB92_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB92_1
> +; RV64IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB92_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw max i16* %a, i16 %b release
> ret i16 %1
> @@ -7637,33 +7637,33 @@ define i16 @atomicrmw_max_i16_acq_rel(i1
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 16
> -; RV32I-NEXT: srai s0, a1, 16
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 16
> +; RV32I-NEXT: srai s0, a0, 16
> ; RV32I-NEXT: addi s3, sp, 10
> ; RV32I-NEXT: j .LBB93_2
> ; RV32I-NEXT: .LBB93_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB93_2 Depth=1
> -; RV32I-NEXT: sh a0, 10(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sh a1, 10(sp)
> ; RV32I-NEXT: addi a3, zero, 4
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 10(sp)
> -; RV32I-NEXT: bnez a1, .LBB93_4
> +; RV32I-NEXT: lh a1, 10(sp)
> +; RV32I-NEXT: bnez a0, .LBB93_4
> ; RV32I-NEXT: .LBB93_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 16
> -; RV32I-NEXT: srai a1, a1, 16
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: blt s0, a1, .LBB93_1
> +; RV32I-NEXT: slli a0, a1, 16
> +; RV32I-NEXT: srai a0, a0, 16
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: blt s0, a0, .LBB93_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB93_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB93_1
> ; RV32I-NEXT: .LBB93_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -7674,33 +7674,33 @@ define i16 @atomicrmw_max_i16_acq_rel(i1
> ;
> ; RV32IA-LABEL: atomicrmw_max_i16_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 16
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: lui a4, 16
> -; RV32IA-NEXT: addi a4, a4, -1
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 16
> ; RV32IA-NEXT: srai a1, a1, 16
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 16
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB93_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a4, a1, .LBB93_3
> +; RV32IA-NEXT: lr.w.aq a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a3, a1, .LBB93_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB93_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB93_3: # in Loop: Header=BB93_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB93_1
> +; RV32IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB93_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_max_i16_acq_rel:
> @@ -7713,33 +7713,33 @@ define i16 @atomicrmw_max_i16_acq_rel(i1
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 48
> -; RV64I-NEXT: srai s0, a1, 48
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 48
> +; RV64I-NEXT: srai s0, a0, 48
> ; RV64I-NEXT: addi s3, sp, 6
> ; RV64I-NEXT: j .LBB93_2
> ; RV64I-NEXT: .LBB93_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB93_2 Depth=1
> -; RV64I-NEXT: sh a0, 6(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sh a1, 6(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 6(sp)
> -; RV64I-NEXT: bnez a1, .LBB93_4
> +; RV64I-NEXT: lh a1, 6(sp)
> +; RV64I-NEXT: bnez a0, .LBB93_4
> ; RV64I-NEXT: .LBB93_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 48
> -; RV64I-NEXT: srai a1, a1, 48
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB93_1
> +; RV64I-NEXT: slli a0, a1, 48
> +; RV64I-NEXT: srai a0, a0, 48
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: blt s0, a0, .LBB93_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB93_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB93_1
> ; RV64I-NEXT: .LBB93_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -7750,33 +7750,33 @@ define i16 @atomicrmw_max_i16_acq_rel(i1
> ;
> ; RV64IA-LABEL: atomicrmw_max_i16_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 48
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: lui a4, 16
> -; RV64IA-NEXT: addiw a4, a4, -1
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 48
> ; RV64IA-NEXT: srai a1, a1, 48
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 48
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB93_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a4, a1, .LBB93_3
> +; RV64IA-NEXT: lr.w.aq a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a3, a1, .LBB93_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB93_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB93_3: # in Loop: Header=BB93_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB93_1
> +; RV64IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB93_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw max i16* %a, i16 %b acq_rel
> ret i16 %1
> @@ -7793,33 +7793,33 @@ define i16 @atomicrmw_max_i16_seq_cst(i1
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 16
> -; RV32I-NEXT: srai s0, a1, 16
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 16
> +; RV32I-NEXT: srai s0, a0, 16
> ; RV32I-NEXT: addi s3, sp, 10
> ; RV32I-NEXT: j .LBB94_2
> ; RV32I-NEXT: .LBB94_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB94_2 Depth=1
> -; RV32I-NEXT: sh a0, 10(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sh a1, 10(sp)
> ; RV32I-NEXT: addi a3, zero, 5
> ; RV32I-NEXT: addi a4, zero, 5
> +; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 10(sp)
> -; RV32I-NEXT: bnez a1, .LBB94_4
> +; RV32I-NEXT: lh a1, 10(sp)
> +; RV32I-NEXT: bnez a0, .LBB94_4
> ; RV32I-NEXT: .LBB94_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 16
> -; RV32I-NEXT: srai a1, a1, 16
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: blt s0, a1, .LBB94_1
> +; RV32I-NEXT: slli a0, a1, 16
> +; RV32I-NEXT: srai a0, a0, 16
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: blt s0, a0, .LBB94_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB94_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB94_1
> ; RV32I-NEXT: .LBB94_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -7830,33 +7830,33 @@ define i16 @atomicrmw_max_i16_seq_cst(i1
> ;
> ; RV32IA-LABEL: atomicrmw_max_i16_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 16
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: lui a4, 16
> -; RV32IA-NEXT: addi a4, a4, -1
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 16
> ; RV32IA-NEXT: srai a1, a1, 16
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 16
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB94_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a4, a1, .LBB94_3
> +; RV32IA-NEXT: lr.w.aqrl a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a3, a1, .LBB94_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB94_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB94_3: # in Loop: Header=BB94_1 Depth=1
> -; RV32IA-NEXT: sc.w.aqrl a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB94_1
> +; RV32IA-NEXT: sc.w.aqrl a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB94_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_max_i16_seq_cst:
> @@ -7869,33 +7869,33 @@ define i16 @atomicrmw_max_i16_seq_cst(i1
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 48
> -; RV64I-NEXT: srai s0, a1, 48
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 48
> +; RV64I-NEXT: srai s0, a0, 48
> ; RV64I-NEXT: addi s3, sp, 6
> ; RV64I-NEXT: j .LBB94_2
> ; RV64I-NEXT: .LBB94_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB94_2 Depth=1
> -; RV64I-NEXT: sh a0, 6(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sh a1, 6(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 6(sp)
> -; RV64I-NEXT: bnez a1, .LBB94_4
> +; RV64I-NEXT: lh a1, 6(sp)
> +; RV64I-NEXT: bnez a0, .LBB94_4
> ; RV64I-NEXT: .LBB94_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 48
> -; RV64I-NEXT: srai a1, a1, 48
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB94_1
> +; RV64I-NEXT: slli a0, a1, 48
> +; RV64I-NEXT: srai a0, a0, 48
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: blt s0, a0, .LBB94_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB94_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB94_1
> ; RV64I-NEXT: .LBB94_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -7906,33 +7906,33 @@ define i16 @atomicrmw_max_i16_seq_cst(i1
> ;
> ; RV64IA-LABEL: atomicrmw_max_i16_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 48
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: lui a4, 16
> -; RV64IA-NEXT: addiw a4, a4, -1
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 48
> ; RV64IA-NEXT: srai a1, a1, 48
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 48
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB94_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a4, a1, .LBB94_3
> +; RV64IA-NEXT: lr.w.aqrl a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a3, a1, .LBB94_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB94_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB94_3: # in Loop: Header=BB94_1 Depth=1
> -; RV64IA-NEXT: sc.w.aqrl a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB94_1
> +; RV64IA-NEXT: sc.w.aqrl a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB94_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw max i16* %a, i16 %b seq_cst
> ret i16 %1
> @@ -7949,33 +7949,33 @@ define i16 @atomicrmw_min_i16_monotonic(
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 16
> -; RV32I-NEXT: srai s0, a1, 16
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 16
> +; RV32I-NEXT: srai s0, a0, 16
> ; RV32I-NEXT: addi s3, sp, 10
> ; RV32I-NEXT: j .LBB95_2
> ; RV32I-NEXT: .LBB95_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB95_2 Depth=1
> -; RV32I-NEXT: sh a0, 10(sp)
> +; RV32I-NEXT: sh a1, 10(sp)
> ; RV32I-NEXT: mv a0, s1
> ; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a3, zero
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 10(sp)
> -; RV32I-NEXT: bnez a1, .LBB95_4
> +; RV32I-NEXT: lh a1, 10(sp)
> +; RV32I-NEXT: bnez a0, .LBB95_4
> ; RV32I-NEXT: .LBB95_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 16
> -; RV32I-NEXT: srai a1, a1, 16
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bge s0, a1, .LBB95_1
> +; RV32I-NEXT: slli a0, a1, 16
> +; RV32I-NEXT: srai a0, a0, 16
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bge s0, a0, .LBB95_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB95_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB95_1
> ; RV32I-NEXT: .LBB95_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -7986,33 +7986,33 @@ define i16 @atomicrmw_min_i16_monotonic(
> ;
> ; RV32IA-LABEL: atomicrmw_min_i16_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 16
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: lui a4, 16
> -; RV32IA-NEXT: addi a4, a4, -1
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 16
> ; RV32IA-NEXT: srai a1, a1, 16
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 16
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB95_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a1, a4, .LBB95_3
> +; RV32IA-NEXT: lr.w a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a1, a3, .LBB95_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB95_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB95_3: # in Loop: Header=BB95_1 Depth=1
> -; RV32IA-NEXT: sc.w a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB95_1
> +; RV32IA-NEXT: sc.w a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB95_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_min_i16_monotonic:
> @@ -8025,33 +8025,33 @@ define i16 @atomicrmw_min_i16_monotonic(
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 48
> -; RV64I-NEXT: srai s0, a1, 48
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 48
> +; RV64I-NEXT: srai s0, a0, 48
> ; RV64I-NEXT: addi s3, sp, 6
> ; RV64I-NEXT: j .LBB95_2
> ; RV64I-NEXT: .LBB95_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB95_2 Depth=1
> -; RV64I-NEXT: sh a0, 6(sp)
> +; RV64I-NEXT: sh a1, 6(sp)
> ; RV64I-NEXT: mv a0, s1
> ; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 6(sp)
> -; RV64I-NEXT: bnez a1, .LBB95_4
> +; RV64I-NEXT: lh a1, 6(sp)
> +; RV64I-NEXT: bnez a0, .LBB95_4
> ; RV64I-NEXT: .LBB95_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 48
> -; RV64I-NEXT: srai a1, a1, 48
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB95_1
> +; RV64I-NEXT: slli a0, a1, 48
> +; RV64I-NEXT: srai a0, a0, 48
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bge s0, a0, .LBB95_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB95_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB95_1
> ; RV64I-NEXT: .LBB95_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -8062,33 +8062,33 @@ define i16 @atomicrmw_min_i16_monotonic(
> ;
> ; RV64IA-LABEL: atomicrmw_min_i16_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 48
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: lui a4, 16
> -; RV64IA-NEXT: addiw a4, a4, -1
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 48
> ; RV64IA-NEXT: srai a1, a1, 48
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 48
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB95_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a1, a4, .LBB95_3
> +; RV64IA-NEXT: lr.w a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a1, a3, .LBB95_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB95_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB95_3: # in Loop: Header=BB95_1 Depth=1
> -; RV64IA-NEXT: sc.w a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB95_1
> +; RV64IA-NEXT: sc.w a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB95_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw min i16* %a, i16 %b monotonic
> ret i16 %1
> @@ -8105,33 +8105,33 @@ define i16 @atomicrmw_min_i16_acquire(i1
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 16
> -; RV32I-NEXT: srai s0, a1, 16
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 16
> +; RV32I-NEXT: srai s0, a0, 16
> ; RV32I-NEXT: addi s3, sp, 10
> ; RV32I-NEXT: j .LBB96_2
> ; RV32I-NEXT: .LBB96_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB96_2 Depth=1
> -; RV32I-NEXT: sh a0, 10(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sh a1, 10(sp)
> ; RV32I-NEXT: addi a3, zero, 2
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 10(sp)
> -; RV32I-NEXT: bnez a1, .LBB96_4
> +; RV32I-NEXT: lh a1, 10(sp)
> +; RV32I-NEXT: bnez a0, .LBB96_4
> ; RV32I-NEXT: .LBB96_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 16
> -; RV32I-NEXT: srai a1, a1, 16
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bge s0, a1, .LBB96_1
> +; RV32I-NEXT: slli a0, a1, 16
> +; RV32I-NEXT: srai a0, a0, 16
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bge s0, a0, .LBB96_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB96_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB96_1
> ; RV32I-NEXT: .LBB96_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -8142,33 +8142,33 @@ define i16 @atomicrmw_min_i16_acquire(i1
> ;
> ; RV32IA-LABEL: atomicrmw_min_i16_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 16
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: lui a4, 16
> -; RV32IA-NEXT: addi a4, a4, -1
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 16
> ; RV32IA-NEXT: srai a1, a1, 16
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 16
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB96_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a1, a4, .LBB96_3
> +; RV32IA-NEXT: lr.w.aq a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a1, a3, .LBB96_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB96_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB96_3: # in Loop: Header=BB96_1 Depth=1
> -; RV32IA-NEXT: sc.w a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB96_1
> +; RV32IA-NEXT: sc.w a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB96_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_min_i16_acquire:
> @@ -8181,33 +8181,33 @@ define i16 @atomicrmw_min_i16_acquire(i1
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 48
> -; RV64I-NEXT: srai s0, a1, 48
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 48
> +; RV64I-NEXT: srai s0, a0, 48
> ; RV64I-NEXT: addi s3, sp, 6
> ; RV64I-NEXT: j .LBB96_2
> ; RV64I-NEXT: .LBB96_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB96_2 Depth=1
> -; RV64I-NEXT: sh a0, 6(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sh a1, 6(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 6(sp)
> -; RV64I-NEXT: bnez a1, .LBB96_4
> +; RV64I-NEXT: lh a1, 6(sp)
> +; RV64I-NEXT: bnez a0, .LBB96_4
> ; RV64I-NEXT: .LBB96_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 48
> -; RV64I-NEXT: srai a1, a1, 48
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB96_1
> +; RV64I-NEXT: slli a0, a1, 48
> +; RV64I-NEXT: srai a0, a0, 48
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bge s0, a0, .LBB96_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB96_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB96_1
> ; RV64I-NEXT: .LBB96_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -8218,33 +8218,33 @@ define i16 @atomicrmw_min_i16_acquire(i1
> ;
> ; RV64IA-LABEL: atomicrmw_min_i16_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 48
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: lui a4, 16
> -; RV64IA-NEXT: addiw a4, a4, -1
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 48
> ; RV64IA-NEXT: srai a1, a1, 48
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 48
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB96_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a1, a4, .LBB96_3
> +; RV64IA-NEXT: lr.w.aq a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a1, a3, .LBB96_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB96_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB96_3: # in Loop: Header=BB96_1 Depth=1
> -; RV64IA-NEXT: sc.w a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB96_1
> +; RV64IA-NEXT: sc.w a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB96_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw min i16* %a, i16 %b acquire
> ret i16 %1
> @@ -8261,33 +8261,33 @@ define i16 @atomicrmw_min_i16_release(i1
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 16
> -; RV32I-NEXT: srai s0, a1, 16
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 16
> +; RV32I-NEXT: srai s0, a0, 16
> ; RV32I-NEXT: addi s3, sp, 10
> ; RV32I-NEXT: j .LBB97_2
> ; RV32I-NEXT: .LBB97_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB97_2 Depth=1
> -; RV32I-NEXT: sh a0, 10(sp)
> +; RV32I-NEXT: sh a1, 10(sp)
> +; RV32I-NEXT: addi a3, zero, 3
> ; RV32I-NEXT: mv a0, s1
> ; RV32I-NEXT: mv a1, s3
> -; RV32I-NEXT: addi a3, zero, 3
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 10(sp)
> -; RV32I-NEXT: bnez a1, .LBB97_4
> +; RV32I-NEXT: lh a1, 10(sp)
> +; RV32I-NEXT: bnez a0, .LBB97_4
> ; RV32I-NEXT: .LBB97_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 16
> -; RV32I-NEXT: srai a1, a1, 16
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bge s0, a1, .LBB97_1
> +; RV32I-NEXT: slli a0, a1, 16
> +; RV32I-NEXT: srai a0, a0, 16
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bge s0, a0, .LBB97_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB97_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB97_1
> ; RV32I-NEXT: .LBB97_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -8298,33 +8298,33 @@ define i16 @atomicrmw_min_i16_release(i1
> ;
> ; RV32IA-LABEL: atomicrmw_min_i16_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 16
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: lui a4, 16
> -; RV32IA-NEXT: addi a4, a4, -1
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 16
> ; RV32IA-NEXT: srai a1, a1, 16
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 16
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB97_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a1, a4, .LBB97_3
> +; RV32IA-NEXT: lr.w a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a1, a3, .LBB97_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB97_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB97_3: # in Loop: Header=BB97_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB97_1
> +; RV32IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB97_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_min_i16_release:
> @@ -8337,33 +8337,33 @@ define i16 @atomicrmw_min_i16_release(i1
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 48
> -; RV64I-NEXT: srai s0, a1, 48
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 48
> +; RV64I-NEXT: srai s0, a0, 48
> ; RV64I-NEXT: addi s3, sp, 6
> ; RV64I-NEXT: j .LBB97_2
> ; RV64I-NEXT: .LBB97_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB97_2 Depth=1
> -; RV64I-NEXT: sh a0, 6(sp)
> +; RV64I-NEXT: sh a1, 6(sp)
> +; RV64I-NEXT: addi a3, zero, 3
> ; RV64I-NEXT: mv a0, s1
> ; RV64I-NEXT: mv a1, s3
> -; RV64I-NEXT: addi a3, zero, 3
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 6(sp)
> -; RV64I-NEXT: bnez a1, .LBB97_4
> +; RV64I-NEXT: lh a1, 6(sp)
> +; RV64I-NEXT: bnez a0, .LBB97_4
> ; RV64I-NEXT: .LBB97_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 48
> -; RV64I-NEXT: srai a1, a1, 48
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB97_1
> +; RV64I-NEXT: slli a0, a1, 48
> +; RV64I-NEXT: srai a0, a0, 48
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bge s0, a0, .LBB97_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB97_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB97_1
> ; RV64I-NEXT: .LBB97_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -8374,33 +8374,33 @@ define i16 @atomicrmw_min_i16_release(i1
> ;
> ; RV64IA-LABEL: atomicrmw_min_i16_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 48
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: lui a4, 16
> -; RV64IA-NEXT: addiw a4, a4, -1
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 48
> ; RV64IA-NEXT: srai a1, a1, 48
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 48
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB97_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a1, a4, .LBB97_3
> +; RV64IA-NEXT: lr.w a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a1, a3, .LBB97_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB97_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB97_3: # in Loop: Header=BB97_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB97_1
> +; RV64IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB97_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw min i16* %a, i16 %b release
> ret i16 %1
> @@ -8417,33 +8417,33 @@ define i16 @atomicrmw_min_i16_acq_rel(i1
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 16
> -; RV32I-NEXT: srai s0, a1, 16
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 16
> +; RV32I-NEXT: srai s0, a0, 16
> ; RV32I-NEXT: addi s3, sp, 10
> ; RV32I-NEXT: j .LBB98_2
> ; RV32I-NEXT: .LBB98_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB98_2 Depth=1
> -; RV32I-NEXT: sh a0, 10(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sh a1, 10(sp)
> ; RV32I-NEXT: addi a3, zero, 4
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 10(sp)
> -; RV32I-NEXT: bnez a1, .LBB98_4
> +; RV32I-NEXT: lh a1, 10(sp)
> +; RV32I-NEXT: bnez a0, .LBB98_4
> ; RV32I-NEXT: .LBB98_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 16
> -; RV32I-NEXT: srai a1, a1, 16
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bge s0, a1, .LBB98_1
> +; RV32I-NEXT: slli a0, a1, 16
> +; RV32I-NEXT: srai a0, a0, 16
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bge s0, a0, .LBB98_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB98_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB98_1
> ; RV32I-NEXT: .LBB98_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -8454,33 +8454,33 @@ define i16 @atomicrmw_min_i16_acq_rel(i1
> ;
> ; RV32IA-LABEL: atomicrmw_min_i16_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 16
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: lui a4, 16
> -; RV32IA-NEXT: addi a4, a4, -1
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 16
> ; RV32IA-NEXT: srai a1, a1, 16
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 16
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB98_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a1, a4, .LBB98_3
> +; RV32IA-NEXT: lr.w.aq a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a1, a3, .LBB98_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB98_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB98_3: # in Loop: Header=BB98_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB98_1
> +; RV32IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB98_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_min_i16_acq_rel:
> @@ -8493,33 +8493,33 @@ define i16 @atomicrmw_min_i16_acq_rel(i1
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 48
> -; RV64I-NEXT: srai s0, a1, 48
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 48
> +; RV64I-NEXT: srai s0, a0, 48
> ; RV64I-NEXT: addi s3, sp, 6
> ; RV64I-NEXT: j .LBB98_2
> ; RV64I-NEXT: .LBB98_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB98_2 Depth=1
> -; RV64I-NEXT: sh a0, 6(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sh a1, 6(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 6(sp)
> -; RV64I-NEXT: bnez a1, .LBB98_4
> +; RV64I-NEXT: lh a1, 6(sp)
> +; RV64I-NEXT: bnez a0, .LBB98_4
> ; RV64I-NEXT: .LBB98_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 48
> -; RV64I-NEXT: srai a1, a1, 48
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB98_1
> +; RV64I-NEXT: slli a0, a1, 48
> +; RV64I-NEXT: srai a0, a0, 48
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bge s0, a0, .LBB98_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB98_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB98_1
> ; RV64I-NEXT: .LBB98_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -8530,33 +8530,33 @@ define i16 @atomicrmw_min_i16_acq_rel(i1
> ;
> ; RV64IA-LABEL: atomicrmw_min_i16_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 48
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: lui a4, 16
> -; RV64IA-NEXT: addiw a4, a4, -1
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 48
> ; RV64IA-NEXT: srai a1, a1, 48
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 48
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB98_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a1, a4, .LBB98_3
> +; RV64IA-NEXT: lr.w.aq a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a1, a3, .LBB98_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB98_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB98_3: # in Loop: Header=BB98_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB98_1
> +; RV64IA-NEXT: sc.w.rl a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB98_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw min i16* %a, i16 %b acq_rel
> ret i16 %1
> @@ -8573,33 +8573,33 @@ define i16 @atomicrmw_min_i16_seq_cst(i1
> ; RV32I-NEXT: sw s3, 12(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: slli a1, a1, 16
> -; RV32I-NEXT: srai s0, a1, 16
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: slli a0, s2, 16
> +; RV32I-NEXT: srai s0, a0, 16
> ; RV32I-NEXT: addi s3, sp, 10
> ; RV32I-NEXT: j .LBB99_2
> ; RV32I-NEXT: .LBB99_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB99_2 Depth=1
> -; RV32I-NEXT: sh a0, 10(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sh a1, 10(sp)
> ; RV32I-NEXT: addi a3, zero, 5
> ; RV32I-NEXT: addi a4, zero, 5
> +; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 10(sp)
> -; RV32I-NEXT: bnez a1, .LBB99_4
> +; RV32I-NEXT: lh a1, 10(sp)
> +; RV32I-NEXT: bnez a0, .LBB99_4
> ; RV32I-NEXT: .LBB99_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: slli a1, a0, 16
> -; RV32I-NEXT: srai a1, a1, 16
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bge s0, a1, .LBB99_1
> +; RV32I-NEXT: slli a0, a1, 16
> +; RV32I-NEXT: srai a0, a0, 16
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bge s0, a0, .LBB99_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB99_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB99_1
> ; RV32I-NEXT: .LBB99_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -8610,33 +8610,33 @@ define i16 @atomicrmw_min_i16_seq_cst(i1
> ;
> ; RV32IA-LABEL: atomicrmw_min_i16_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: slli a2, a0, 3
> -; RV32IA-NEXT: andi a2, a2, 24
> -; RV32IA-NEXT: addi a3, zero, 16
> -; RV32IA-NEXT: sub a6, a3, a2
> -; RV32IA-NEXT: lui a4, 16
> -; RV32IA-NEXT: addi a4, a4, -1
> -; RV32IA-NEXT: sll a7, a4, a2
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a7, a3, a0
> ; RV32IA-NEXT: slli a1, a1, 16
> ; RV32IA-NEXT: srai a1, a1, 16
> -; RV32IA-NEXT: sll a1, a1, a2
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: sll a1, a1, a0
> +; RV32IA-NEXT: addi a4, zero, 16
> +; RV32IA-NEXT: sub a4, a4, a0
> ; RV32IA-NEXT: .LBB99_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a5, (a0)
> -; RV32IA-NEXT: and a4, a5, a7
> -; RV32IA-NEXT: mv a3, a5
> -; RV32IA-NEXT: sll a4, a4, a6
> -; RV32IA-NEXT: sra a4, a4, a6
> -; RV32IA-NEXT: bge a1, a4, .LBB99_3
> +; RV32IA-NEXT: lr.w.aqrl a5, (a6)
> +; RV32IA-NEXT: and a3, a5, a7
> +; RV32IA-NEXT: mv a2, a5
> +; RV32IA-NEXT: sll a3, a3, a4
> +; RV32IA-NEXT: sra a3, a3, a4
> +; RV32IA-NEXT: bge a1, a3, .LBB99_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB99_1 Depth=1
> -; RV32IA-NEXT: xor a3, a5, a1
> -; RV32IA-NEXT: and a3, a3, a7
> -; RV32IA-NEXT: xor a3, a5, a3
> +; RV32IA-NEXT: xor a2, a5, a1
> +; RV32IA-NEXT: and a2, a2, a7
> +; RV32IA-NEXT: xor a2, a5, a2
> ; RV32IA-NEXT: .LBB99_3: # in Loop: Header=BB99_1 Depth=1
> -; RV32IA-NEXT: sc.w.aqrl a3, a3, (a0)
> -; RV32IA-NEXT: bnez a3, .LBB99_1
> +; RV32IA-NEXT: sc.w.aqrl a2, a2, (a6)
> +; RV32IA-NEXT: bnez a2, .LBB99_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a5, a2
> +; RV32IA-NEXT: srl a0, a5, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_min_i16_seq_cst:
> @@ -8649,33 +8649,33 @@ define i16 @atomicrmw_min_i16_seq_cst(i1
> ; RV64I-NEXT: sd s3, 8(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: slli a1, a1, 48
> -; RV64I-NEXT: srai s0, a1, 48
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: slli a0, s2, 48
> +; RV64I-NEXT: srai s0, a0, 48
> ; RV64I-NEXT: addi s3, sp, 6
> ; RV64I-NEXT: j .LBB99_2
> ; RV64I-NEXT: .LBB99_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB99_2 Depth=1
> -; RV64I-NEXT: sh a0, 6(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sh a1, 6(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 6(sp)
> -; RV64I-NEXT: bnez a1, .LBB99_4
> +; RV64I-NEXT: lh a1, 6(sp)
> +; RV64I-NEXT: bnez a0, .LBB99_4
> ; RV64I-NEXT: .LBB99_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: slli a1, a0, 48
> -; RV64I-NEXT: srai a1, a1, 48
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB99_1
> +; RV64I-NEXT: slli a0, a1, 48
> +; RV64I-NEXT: srai a0, a0, 48
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bge s0, a0, .LBB99_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB99_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB99_1
> ; RV64I-NEXT: .LBB99_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -8686,33 +8686,33 @@ define i16 @atomicrmw_min_i16_seq_cst(i1
> ;
> ; RV64IA-LABEL: atomicrmw_min_i16_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: slli a2, a0, 3
> -; RV64IA-NEXT: andi a2, a2, 24
> -; RV64IA-NEXT: addi a3, zero, 48
> -; RV64IA-NEXT: sub a6, a3, a2
> -; RV64IA-NEXT: lui a4, 16
> -; RV64IA-NEXT: addiw a4, a4, -1
> -; RV64IA-NEXT: sllw a7, a4, a2
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a7, a3, a0
> ; RV64IA-NEXT: slli a1, a1, 48
> ; RV64IA-NEXT: srai a1, a1, 48
> -; RV64IA-NEXT: sllw a1, a1, a2
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: sllw a1, a1, a0
> +; RV64IA-NEXT: addi a4, zero, 48
> +; RV64IA-NEXT: sub a4, a4, a0
> ; RV64IA-NEXT: .LBB99_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a5, (a0)
> -; RV64IA-NEXT: and a4, a5, a7
> -; RV64IA-NEXT: mv a3, a5
> -; RV64IA-NEXT: sll a4, a4, a6
> -; RV64IA-NEXT: sra a4, a4, a6
> -; RV64IA-NEXT: bge a1, a4, .LBB99_3
> +; RV64IA-NEXT: lr.w.aqrl a5, (a6)
> +; RV64IA-NEXT: and a3, a5, a7
> +; RV64IA-NEXT: mv a2, a5
> +; RV64IA-NEXT: sll a3, a3, a4
> +; RV64IA-NEXT: sra a3, a3, a4
> +; RV64IA-NEXT: bge a1, a3, .LBB99_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB99_1 Depth=1
> -; RV64IA-NEXT: xor a3, a5, a1
> -; RV64IA-NEXT: and a3, a3, a7
> -; RV64IA-NEXT: xor a3, a5, a3
> +; RV64IA-NEXT: xor a2, a5, a1
> +; RV64IA-NEXT: and a2, a2, a7
> +; RV64IA-NEXT: xor a2, a5, a2
> ; RV64IA-NEXT: .LBB99_3: # in Loop: Header=BB99_1 Depth=1
> -; RV64IA-NEXT: sc.w.aqrl a3, a3, (a0)
> -; RV64IA-NEXT: bnez a3, .LBB99_1
> +; RV64IA-NEXT: sc.w.aqrl a2, a2, (a6)
> +; RV64IA-NEXT: bnez a2, .LBB99_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a5, a2
> +; RV64IA-NEXT: srlw a0, a5, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw min i16* %a, i16 %b seq_cst
> ret i16 %1
> @@ -8730,33 +8730,33 @@ define i16 @atomicrmw_umax_i16_monotonic
> ; RV32I-NEXT: sw s4, 8(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s4, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: lui a1, 16
> -; RV32I-NEXT: addi s0, a1, -1
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: lui a0, 16
> +; RV32I-NEXT: addi s0, a0, -1
> ; RV32I-NEXT: and s1, s2, s0
> ; RV32I-NEXT: addi s3, sp, 6
> ; RV32I-NEXT: j .LBB100_2
> ; RV32I-NEXT: .LBB100_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB100_2 Depth=1
> -; RV32I-NEXT: sh a0, 6(sp)
> +; RV32I-NEXT: sh a1, 6(sp)
> ; RV32I-NEXT: mv a0, s4
> ; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a3, zero
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 6(sp)
> -; RV32I-NEXT: bnez a1, .LBB100_4
> +; RV32I-NEXT: lh a1, 6(sp)
> +; RV32I-NEXT: bnez a0, .LBB100_4
> ; RV32I-NEXT: .LBB100_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: and a1, a0, s0
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bltu s1, a1, .LBB100_1
> +; RV32I-NEXT: and a0, a1, s0
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bltu s1, a0, .LBB100_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB100_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB100_1
> ; RV32I-NEXT: .LBB100_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s4, 8(sp)
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> @@ -8768,28 +8768,28 @@ define i16 @atomicrmw_umax_i16_monotonic
> ;
> ; RV32IA-LABEL: atomicrmw_umax_i16_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a6, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB100_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: and a2, a4, a6
> -; RV32IA-NEXT: mv a5, a4
> +; RV32IA-NEXT: lr.w a3, (a6)
> +; RV32IA-NEXT: and a2, a3, a4
> +; RV32IA-NEXT: mv a5, a3
> ; RV32IA-NEXT: bgeu a2, a1, .LBB100_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB100_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> -; RV32IA-NEXT: xor a5, a4, a5
> +; RV32IA-NEXT: xor a5, a3, a1
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> ; RV32IA-NEXT: .LBB100_3: # in Loop: Header=BB100_1 Depth=1
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB100_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umax_i16_monotonic:
> @@ -8803,33 +8803,33 @@ define i16 @atomicrmw_umax_i16_monotonic
> ; RV64I-NEXT: sd s4, 16(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s4, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: lui a1, 16
> -; RV64I-NEXT: addiw s0, a1, -1
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: lui a0, 16
> +; RV64I-NEXT: addiw s0, a0, -1
> ; RV64I-NEXT: and s1, s2, s0
> ; RV64I-NEXT: addi s3, sp, 14
> ; RV64I-NEXT: j .LBB100_2
> ; RV64I-NEXT: .LBB100_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB100_2 Depth=1
> -; RV64I-NEXT: sh a0, 14(sp)
> +; RV64I-NEXT: sh a1, 14(sp)
> ; RV64I-NEXT: mv a0, s4
> ; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 14(sp)
> -; RV64I-NEXT: bnez a1, .LBB100_4
> +; RV64I-NEXT: lh a1, 14(sp)
> +; RV64I-NEXT: bnez a0, .LBB100_4
> ; RV64I-NEXT: .LBB100_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: and a1, a0, s0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s1, a1, .LBB100_1
> +; RV64I-NEXT: and a0, a1, s0
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bltu s1, a0, .LBB100_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB100_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB100_1
> ; RV64I-NEXT: .LBB100_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s4, 16(sp)
> ; RV64I-NEXT: ld s3, 24(sp)
> ; RV64I-NEXT: ld s2, 32(sp)
> @@ -8841,28 +8841,28 @@ define i16 @atomicrmw_umax_i16_monotonic
> ;
> ; RV64IA-LABEL: atomicrmw_umax_i16_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a6, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB100_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: and a2, a4, a6
> -; RV64IA-NEXT: mv a5, a4
> +; RV64IA-NEXT: lr.w a3, (a6)
> +; RV64IA-NEXT: and a2, a3, a4
> +; RV64IA-NEXT: mv a5, a3
> ; RV64IA-NEXT: bgeu a2, a1, .LBB100_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB100_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> -; RV64IA-NEXT: xor a5, a4, a5
> +; RV64IA-NEXT: xor a5, a3, a1
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> ; RV64IA-NEXT: .LBB100_3: # in Loop: Header=BB100_1 Depth=1
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB100_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umax i16* %a, i16 %b monotonic
> ret i16 %1
> @@ -8880,33 +8880,33 @@ define i16 @atomicrmw_umax_i16_acquire(i
> ; RV32I-NEXT: sw s4, 8(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s4, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: lui a1, 16
> -; RV32I-NEXT: addi s0, a1, -1
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: lui a0, 16
> +; RV32I-NEXT: addi s0, a0, -1
> ; RV32I-NEXT: and s1, s2, s0
> ; RV32I-NEXT: addi s3, sp, 6
> ; RV32I-NEXT: j .LBB101_2
> ; RV32I-NEXT: .LBB101_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB101_2 Depth=1
> -; RV32I-NEXT: sh a0, 6(sp)
> -; RV32I-NEXT: mv a0, s4
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sh a1, 6(sp)
> ; RV32I-NEXT: addi a3, zero, 2
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s4
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 6(sp)
> -; RV32I-NEXT: bnez a1, .LBB101_4
> +; RV32I-NEXT: lh a1, 6(sp)
> +; RV32I-NEXT: bnez a0, .LBB101_4
> ; RV32I-NEXT: .LBB101_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: and a1, a0, s0
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bltu s1, a1, .LBB101_1
> +; RV32I-NEXT: and a0, a1, s0
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bltu s1, a0, .LBB101_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB101_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB101_1
> ; RV32I-NEXT: .LBB101_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s4, 8(sp)
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> @@ -8918,28 +8918,28 @@ define i16 @atomicrmw_umax_i16_acquire(i
> ;
> ; RV32IA-LABEL: atomicrmw_umax_i16_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a6, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB101_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a2, a4, a6
> -; RV32IA-NEXT: mv a5, a4
> +; RV32IA-NEXT: lr.w.aq a3, (a6)
> +; RV32IA-NEXT: and a2, a3, a4
> +; RV32IA-NEXT: mv a5, a3
> ; RV32IA-NEXT: bgeu a2, a1, .LBB101_3
> -; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB101_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> -; RV32IA-NEXT: xor a5, a4, a5
> +; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB101_1 Depth=1
> +; RV32IA-NEXT: xor a5, a3, a1
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> ; RV32IA-NEXT: .LBB101_3: # in Loop: Header=BB101_1 Depth=1
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB101_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umax_i16_acquire:
> @@ -8953,33 +8953,33 @@ define i16 @atomicrmw_umax_i16_acquire(i
> ; RV64I-NEXT: sd s4, 16(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s4, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: lui a1, 16
> -; RV64I-NEXT: addiw s0, a1, -1
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: lui a0, 16
> +; RV64I-NEXT: addiw s0, a0, -1
> ; RV64I-NEXT: and s1, s2, s0
> ; RV64I-NEXT: addi s3, sp, 14
> ; RV64I-NEXT: j .LBB101_2
> ; RV64I-NEXT: .LBB101_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB101_2 Depth=1
> -; RV64I-NEXT: sh a0, 14(sp)
> -; RV64I-NEXT: mv a0, s4
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sh a1, 14(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s4
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 14(sp)
> -; RV64I-NEXT: bnez a1, .LBB101_4
> +; RV64I-NEXT: lh a1, 14(sp)
> +; RV64I-NEXT: bnez a0, .LBB101_4
> ; RV64I-NEXT: .LBB101_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: and a1, a0, s0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s1, a1, .LBB101_1
> +; RV64I-NEXT: and a0, a1, s0
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bltu s1, a0, .LBB101_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB101_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB101_1
> ; RV64I-NEXT: .LBB101_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s4, 16(sp)
> ; RV64I-NEXT: ld s3, 24(sp)
> ; RV64I-NEXT: ld s2, 32(sp)
> @@ -8991,28 +8991,28 @@ define i16 @atomicrmw_umax_i16_acquire(i
> ;
> ; RV64IA-LABEL: atomicrmw_umax_i16_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a6, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB101_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a2, a4, a6
> -; RV64IA-NEXT: mv a5, a4
> +; RV64IA-NEXT: lr.w.aq a3, (a6)
> +; RV64IA-NEXT: and a2, a3, a4
> +; RV64IA-NEXT: mv a5, a3
> ; RV64IA-NEXT: bgeu a2, a1, .LBB101_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB101_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> -; RV64IA-NEXT: xor a5, a4, a5
> +; RV64IA-NEXT: xor a5, a3, a1
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> ; RV64IA-NEXT: .LBB101_3: # in Loop: Header=BB101_1 Depth=1
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB101_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umax i16* %a, i16 %b acquire
> ret i16 %1
> @@ -9030,33 +9030,33 @@ define i16 @atomicrmw_umax_i16_release(i
> ; RV32I-NEXT: sw s4, 8(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s4, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: lui a1, 16
> -; RV32I-NEXT: addi s0, a1, -1
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: lui a0, 16
> +; RV32I-NEXT: addi s0, a0, -1
> ; RV32I-NEXT: and s1, s2, s0
> ; RV32I-NEXT: addi s3, sp, 6
> ; RV32I-NEXT: j .LBB102_2
> ; RV32I-NEXT: .LBB102_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB102_2 Depth=1
> -; RV32I-NEXT: sh a0, 6(sp)
> +; RV32I-NEXT: sh a1, 6(sp)
> +; RV32I-NEXT: addi a3, zero, 3
> ; RV32I-NEXT: mv a0, s4
> ; RV32I-NEXT: mv a1, s3
> -; RV32I-NEXT: addi a3, zero, 3
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 6(sp)
> -; RV32I-NEXT: bnez a1, .LBB102_4
> +; RV32I-NEXT: lh a1, 6(sp)
> +; RV32I-NEXT: bnez a0, .LBB102_4
> ; RV32I-NEXT: .LBB102_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: and a1, a0, s0
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bltu s1, a1, .LBB102_1
> +; RV32I-NEXT: and a0, a1, s0
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bltu s1, a0, .LBB102_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB102_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB102_1
> ; RV32I-NEXT: .LBB102_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s4, 8(sp)
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> @@ -9068,28 +9068,28 @@ define i16 @atomicrmw_umax_i16_release(i
> ;
> ; RV32IA-LABEL: atomicrmw_umax_i16_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a6, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB102_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: and a2, a4, a6
> -; RV32IA-NEXT: mv a5, a4
> +; RV32IA-NEXT: lr.w a3, (a6)
> +; RV32IA-NEXT: and a2, a3, a4
> +; RV32IA-NEXT: mv a5, a3
> ; RV32IA-NEXT: bgeu a2, a1, .LBB102_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB102_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> -; RV32IA-NEXT: xor a5, a4, a5
> +; RV32IA-NEXT: xor a5, a3, a1
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> ; RV32IA-NEXT: .LBB102_3: # in Loop: Header=BB102_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB102_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umax_i16_release:
> @@ -9103,33 +9103,33 @@ define i16 @atomicrmw_umax_i16_release(i
> ; RV64I-NEXT: sd s4, 16(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s4, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: lui a1, 16
> -; RV64I-NEXT: addiw s0, a1, -1
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: lui a0, 16
> +; RV64I-NEXT: addiw s0, a0, -1
> ; RV64I-NEXT: and s1, s2, s0
> ; RV64I-NEXT: addi s3, sp, 14
> ; RV64I-NEXT: j .LBB102_2
> ; RV64I-NEXT: .LBB102_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB102_2 Depth=1
> -; RV64I-NEXT: sh a0, 14(sp)
> +; RV64I-NEXT: sh a1, 14(sp)
> +; RV64I-NEXT: addi a3, zero, 3
> ; RV64I-NEXT: mv a0, s4
> ; RV64I-NEXT: mv a1, s3
> -; RV64I-NEXT: addi a3, zero, 3
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 14(sp)
> -; RV64I-NEXT: bnez a1, .LBB102_4
> +; RV64I-NEXT: lh a1, 14(sp)
> +; RV64I-NEXT: bnez a0, .LBB102_4
> ; RV64I-NEXT: .LBB102_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: and a1, a0, s0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s1, a1, .LBB102_1
> +; RV64I-NEXT: and a0, a1, s0
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bltu s1, a0, .LBB102_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB102_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB102_1
> ; RV64I-NEXT: .LBB102_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s4, 16(sp)
> ; RV64I-NEXT: ld s3, 24(sp)
> ; RV64I-NEXT: ld s2, 32(sp)
> @@ -9141,28 +9141,28 @@ define i16 @atomicrmw_umax_i16_release(i
> ;
> ; RV64IA-LABEL: atomicrmw_umax_i16_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a6, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB102_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: and a2, a4, a6
> -; RV64IA-NEXT: mv a5, a4
> +; RV64IA-NEXT: lr.w a3, (a6)
> +; RV64IA-NEXT: and a2, a3, a4
> +; RV64IA-NEXT: mv a5, a3
> ; RV64IA-NEXT: bgeu a2, a1, .LBB102_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB102_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> -; RV64IA-NEXT: xor a5, a4, a5
> +; RV64IA-NEXT: xor a5, a3, a1
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> ; RV64IA-NEXT: .LBB102_3: # in Loop: Header=BB102_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB102_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umax i16* %a, i16 %b release
> ret i16 %1
> @@ -9180,33 +9180,33 @@ define i16 @atomicrmw_umax_i16_acq_rel(i
> ; RV32I-NEXT: sw s4, 8(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s4, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: lui a1, 16
> -; RV32I-NEXT: addi s0, a1, -1
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: lui a0, 16
> +; RV32I-NEXT: addi s0, a0, -1
> ; RV32I-NEXT: and s1, s2, s0
> ; RV32I-NEXT: addi s3, sp, 6
> ; RV32I-NEXT: j .LBB103_2
> ; RV32I-NEXT: .LBB103_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB103_2 Depth=1
> -; RV32I-NEXT: sh a0, 6(sp)
> -; RV32I-NEXT: mv a0, s4
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sh a1, 6(sp)
> ; RV32I-NEXT: addi a3, zero, 4
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s4
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 6(sp)
> -; RV32I-NEXT: bnez a1, .LBB103_4
> +; RV32I-NEXT: lh a1, 6(sp)
> +; RV32I-NEXT: bnez a0, .LBB103_4
> ; RV32I-NEXT: .LBB103_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: and a1, a0, s0
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bltu s1, a1, .LBB103_1
> +; RV32I-NEXT: and a0, a1, s0
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bltu s1, a0, .LBB103_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB103_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB103_1
> ; RV32I-NEXT: .LBB103_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s4, 8(sp)
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> @@ -9218,28 +9218,28 @@ define i16 @atomicrmw_umax_i16_acq_rel(i
> ;
> ; RV32IA-LABEL: atomicrmw_umax_i16_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a6, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB103_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a2, a4, a6
> -; RV32IA-NEXT: mv a5, a4
> +; RV32IA-NEXT: lr.w.aq a3, (a6)
> +; RV32IA-NEXT: and a2, a3, a4
> +; RV32IA-NEXT: mv a5, a3
> ; RV32IA-NEXT: bgeu a2, a1, .LBB103_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB103_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> -; RV32IA-NEXT: xor a5, a4, a5
> +; RV32IA-NEXT: xor a5, a3, a1
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> ; RV32IA-NEXT: .LBB103_3: # in Loop: Header=BB103_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB103_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umax_i16_acq_rel:
> @@ -9253,33 +9253,33 @@ define i16 @atomicrmw_umax_i16_acq_rel(i
> ; RV64I-NEXT: sd s4, 16(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s4, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: lui a1, 16
> -; RV64I-NEXT: addiw s0, a1, -1
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: lui a0, 16
> +; RV64I-NEXT: addiw s0, a0, -1
> ; RV64I-NEXT: and s1, s2, s0
> ; RV64I-NEXT: addi s3, sp, 14
> ; RV64I-NEXT: j .LBB103_2
> ; RV64I-NEXT: .LBB103_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB103_2 Depth=1
> -; RV64I-NEXT: sh a0, 14(sp)
> -; RV64I-NEXT: mv a0, s4
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sh a1, 14(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s4
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 14(sp)
> -; RV64I-NEXT: bnez a1, .LBB103_4
> +; RV64I-NEXT: lh a1, 14(sp)
> +; RV64I-NEXT: bnez a0, .LBB103_4
> ; RV64I-NEXT: .LBB103_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: and a1, a0, s0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s1, a1, .LBB103_1
> +; RV64I-NEXT: and a0, a1, s0
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bltu s1, a0, .LBB103_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB103_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB103_1
> ; RV64I-NEXT: .LBB103_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s4, 16(sp)
> ; RV64I-NEXT: ld s3, 24(sp)
> ; RV64I-NEXT: ld s2, 32(sp)
> @@ -9291,28 +9291,28 @@ define i16 @atomicrmw_umax_i16_acq_rel(i
> ;
> ; RV64IA-LABEL: atomicrmw_umax_i16_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a6, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB103_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a2, a4, a6
> -; RV64IA-NEXT: mv a5, a4
> +; RV64IA-NEXT: lr.w.aq a3, (a6)
> +; RV64IA-NEXT: and a2, a3, a4
> +; RV64IA-NEXT: mv a5, a3
> ; RV64IA-NEXT: bgeu a2, a1, .LBB103_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB103_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> -; RV64IA-NEXT: xor a5, a4, a5
> +; RV64IA-NEXT: xor a5, a3, a1
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> ; RV64IA-NEXT: .LBB103_3: # in Loop: Header=BB103_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB103_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umax i16* %a, i16 %b acq_rel
> ret i16 %1
> @@ -9330,33 +9330,33 @@ define i16 @atomicrmw_umax_i16_seq_cst(i
> ; RV32I-NEXT: sw s4, 8(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s4, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: lui a1, 16
> -; RV32I-NEXT: addi s0, a1, -1
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: lui a0, 16
> +; RV32I-NEXT: addi s0, a0, -1
> ; RV32I-NEXT: and s1, s2, s0
> ; RV32I-NEXT: addi s3, sp, 6
> ; RV32I-NEXT: j .LBB104_2
> ; RV32I-NEXT: .LBB104_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB104_2 Depth=1
> -; RV32I-NEXT: sh a0, 6(sp)
> -; RV32I-NEXT: mv a0, s4
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sh a1, 6(sp)
> ; RV32I-NEXT: addi a3, zero, 5
> ; RV32I-NEXT: addi a4, zero, 5
> +; RV32I-NEXT: mv a0, s4
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 6(sp)
> -; RV32I-NEXT: bnez a1, .LBB104_4
> +; RV32I-NEXT: lh a1, 6(sp)
> +; RV32I-NEXT: bnez a0, .LBB104_4
> ; RV32I-NEXT: .LBB104_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: and a1, a0, s0
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bltu s1, a1, .LBB104_1
> +; RV32I-NEXT: and a0, a1, s0
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bltu s1, a0, .LBB104_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB104_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB104_1
> ; RV32I-NEXT: .LBB104_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s4, 8(sp)
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> @@ -9368,28 +9368,28 @@ define i16 @atomicrmw_umax_i16_seq_cst(i
> ;
> ; RV32IA-LABEL: atomicrmw_umax_i16_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a6, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB104_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV32IA-NEXT: and a2, a4, a6
> -; RV32IA-NEXT: mv a5, a4
> +; RV32IA-NEXT: lr.w.aqrl a3, (a6)
> +; RV32IA-NEXT: and a2, a3, a4
> +; RV32IA-NEXT: mv a5, a3
> ; RV32IA-NEXT: bgeu a2, a1, .LBB104_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB104_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> -; RV32IA-NEXT: xor a5, a4, a5
> +; RV32IA-NEXT: xor a5, a3, a1
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> ; RV32IA-NEXT: .LBB104_3: # in Loop: Header=BB104_1 Depth=1
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB104_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umax_i16_seq_cst:
> @@ -9403,33 +9403,33 @@ define i16 @atomicrmw_umax_i16_seq_cst(i
> ; RV64I-NEXT: sd s4, 16(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s4, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: lui a1, 16
> -; RV64I-NEXT: addiw s0, a1, -1
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: lui a0, 16
> +; RV64I-NEXT: addiw s0, a0, -1
> ; RV64I-NEXT: and s1, s2, s0
> ; RV64I-NEXT: addi s3, sp, 14
> ; RV64I-NEXT: j .LBB104_2
> ; RV64I-NEXT: .LBB104_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB104_2 Depth=1
> -; RV64I-NEXT: sh a0, 14(sp)
> -; RV64I-NEXT: mv a0, s4
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sh a1, 14(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s4
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 14(sp)
> -; RV64I-NEXT: bnez a1, .LBB104_4
> +; RV64I-NEXT: lh a1, 14(sp)
> +; RV64I-NEXT: bnez a0, .LBB104_4
> ; RV64I-NEXT: .LBB104_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: and a1, a0, s0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s1, a1, .LBB104_1
> +; RV64I-NEXT: and a0, a1, s0
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bltu s1, a0, .LBB104_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB104_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB104_1
> ; RV64I-NEXT: .LBB104_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s4, 16(sp)
> ; RV64I-NEXT: ld s3, 24(sp)
> ; RV64I-NEXT: ld s2, 32(sp)
> @@ -9441,28 +9441,28 @@ define i16 @atomicrmw_umax_i16_seq_cst(i
> ;
> ; RV64IA-LABEL: atomicrmw_umax_i16_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a6, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB104_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV64IA-NEXT: and a2, a4, a6
> -; RV64IA-NEXT: mv a5, a4
> +; RV64IA-NEXT: lr.w.aqrl a3, (a6)
> +; RV64IA-NEXT: and a2, a3, a4
> +; RV64IA-NEXT: mv a5, a3
> ; RV64IA-NEXT: bgeu a2, a1, .LBB104_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB104_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> -; RV64IA-NEXT: xor a5, a4, a5
> +; RV64IA-NEXT: xor a5, a3, a1
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> ; RV64IA-NEXT: .LBB104_3: # in Loop: Header=BB104_1 Depth=1
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB104_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umax i16* %a, i16 %b seq_cst
> ret i16 %1
> @@ -9480,33 +9480,33 @@ define i16 @atomicrmw_umin_i16_monotonic
> ; RV32I-NEXT: sw s4, 8(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s4, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: lui a1, 16
> -; RV32I-NEXT: addi s0, a1, -1
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: lui a0, 16
> +; RV32I-NEXT: addi s0, a0, -1
> ; RV32I-NEXT: and s1, s2, s0
> ; RV32I-NEXT: addi s3, sp, 6
> ; RV32I-NEXT: j .LBB105_2
> ; RV32I-NEXT: .LBB105_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB105_2 Depth=1
> -; RV32I-NEXT: sh a0, 6(sp)
> +; RV32I-NEXT: sh a1, 6(sp)
> ; RV32I-NEXT: mv a0, s4
> ; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a3, zero
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 6(sp)
> -; RV32I-NEXT: bnez a1, .LBB105_4
> +; RV32I-NEXT: lh a1, 6(sp)
> +; RV32I-NEXT: bnez a0, .LBB105_4
> ; RV32I-NEXT: .LBB105_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: and a1, a0, s0
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bgeu s1, a1, .LBB105_1
> +; RV32I-NEXT: and a0, a1, s0
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bgeu s1, a0, .LBB105_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB105_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB105_1
> ; RV32I-NEXT: .LBB105_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s4, 8(sp)
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> @@ -9518,28 +9518,28 @@ define i16 @atomicrmw_umin_i16_monotonic
> ;
> ; RV32IA-LABEL: atomicrmw_umin_i16_monotonic:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a6, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB105_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: and a2, a4, a6
> -; RV32IA-NEXT: mv a5, a4
> +; RV32IA-NEXT: lr.w a3, (a6)
> +; RV32IA-NEXT: and a2, a3, a4
> +; RV32IA-NEXT: mv a5, a3
> ; RV32IA-NEXT: bgeu a1, a2, .LBB105_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB105_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> -; RV32IA-NEXT: xor a5, a4, a5
> +; RV32IA-NEXT: xor a5, a3, a1
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> ; RV32IA-NEXT: .LBB105_3: # in Loop: Header=BB105_1 Depth=1
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB105_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umin_i16_monotonic:
> @@ -9553,33 +9553,33 @@ define i16 @atomicrmw_umin_i16_monotonic
> ; RV64I-NEXT: sd s4, 16(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s4, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: lui a1, 16
> -; RV64I-NEXT: addiw s0, a1, -1
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: lui a0, 16
> +; RV64I-NEXT: addiw s0, a0, -1
> ; RV64I-NEXT: and s1, s2, s0
> ; RV64I-NEXT: addi s3, sp, 14
> ; RV64I-NEXT: j .LBB105_2
> ; RV64I-NEXT: .LBB105_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB105_2 Depth=1
> -; RV64I-NEXT: sh a0, 14(sp)
> +; RV64I-NEXT: sh a1, 14(sp)
> ; RV64I-NEXT: mv a0, s4
> ; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 14(sp)
> -; RV64I-NEXT: bnez a1, .LBB105_4
> +; RV64I-NEXT: lh a1, 14(sp)
> +; RV64I-NEXT: bnez a0, .LBB105_4
> ; RV64I-NEXT: .LBB105_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: and a1, a0, s0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s1, a1, .LBB105_1
> +; RV64I-NEXT: and a0, a1, s0
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bgeu s1, a0, .LBB105_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB105_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB105_1
> ; RV64I-NEXT: .LBB105_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s4, 16(sp)
> ; RV64I-NEXT: ld s3, 24(sp)
> ; RV64I-NEXT: ld s2, 32(sp)
> @@ -9591,28 +9591,28 @@ define i16 @atomicrmw_umin_i16_monotonic
> ;
> ; RV64IA-LABEL: atomicrmw_umin_i16_monotonic:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a6, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB105_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: and a2, a4, a6
> -; RV64IA-NEXT: mv a5, a4
> +; RV64IA-NEXT: lr.w a3, (a6)
> +; RV64IA-NEXT: and a2, a3, a4
> +; RV64IA-NEXT: mv a5, a3
> ; RV64IA-NEXT: bgeu a1, a2, .LBB105_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB105_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> -; RV64IA-NEXT: xor a5, a4, a5
> +; RV64IA-NEXT: xor a5, a3, a1
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> ; RV64IA-NEXT: .LBB105_3: # in Loop: Header=BB105_1 Depth=1
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB105_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umin i16* %a, i16 %b monotonic
> ret i16 %1
> @@ -9630,33 +9630,33 @@ define i16 @atomicrmw_umin_i16_acquire(i
> ; RV32I-NEXT: sw s4, 8(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s4, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: lui a1, 16
> -; RV32I-NEXT: addi s0, a1, -1
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: lui a0, 16
> +; RV32I-NEXT: addi s0, a0, -1
> ; RV32I-NEXT: and s1, s2, s0
> ; RV32I-NEXT: addi s3, sp, 6
> ; RV32I-NEXT: j .LBB106_2
> ; RV32I-NEXT: .LBB106_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB106_2 Depth=1
> -; RV32I-NEXT: sh a0, 6(sp)
> -; RV32I-NEXT: mv a0, s4
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sh a1, 6(sp)
> ; RV32I-NEXT: addi a3, zero, 2
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s4
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 6(sp)
> -; RV32I-NEXT: bnez a1, .LBB106_4
> +; RV32I-NEXT: lh a1, 6(sp)
> +; RV32I-NEXT: bnez a0, .LBB106_4
> ; RV32I-NEXT: .LBB106_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: and a1, a0, s0
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bgeu s1, a1, .LBB106_1
> +; RV32I-NEXT: and a0, a1, s0
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bgeu s1, a0, .LBB106_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB106_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB106_1
> ; RV32I-NEXT: .LBB106_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s4, 8(sp)
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> @@ -9668,28 +9668,28 @@ define i16 @atomicrmw_umin_i16_acquire(i
> ;
> ; RV32IA-LABEL: atomicrmw_umin_i16_acquire:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a6, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB106_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a2, a4, a6
> -; RV32IA-NEXT: mv a5, a4
> +; RV32IA-NEXT: lr.w.aq a3, (a6)
> +; RV32IA-NEXT: and a2, a3, a4
> +; RV32IA-NEXT: mv a5, a3
> ; RV32IA-NEXT: bgeu a1, a2, .LBB106_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB106_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> -; RV32IA-NEXT: xor a5, a4, a5
> +; RV32IA-NEXT: xor a5, a3, a1
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> ; RV32IA-NEXT: .LBB106_3: # in Loop: Header=BB106_1 Depth=1
> -; RV32IA-NEXT: sc.w a5, a5, (a0)
> +; RV32IA-NEXT: sc.w a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB106_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umin_i16_acquire:
> @@ -9703,33 +9703,33 @@ define i16 @atomicrmw_umin_i16_acquire(i
> ; RV64I-NEXT: sd s4, 16(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s4, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: lui a1, 16
> -; RV64I-NEXT: addiw s0, a1, -1
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: lui a0, 16
> +; RV64I-NEXT: addiw s0, a0, -1
> ; RV64I-NEXT: and s1, s2, s0
> ; RV64I-NEXT: addi s3, sp, 14
> ; RV64I-NEXT: j .LBB106_2
> ; RV64I-NEXT: .LBB106_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB106_2 Depth=1
> -; RV64I-NEXT: sh a0, 14(sp)
> -; RV64I-NEXT: mv a0, s4
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sh a1, 14(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s4
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 14(sp)
> -; RV64I-NEXT: bnez a1, .LBB106_4
> +; RV64I-NEXT: lh a1, 14(sp)
> +; RV64I-NEXT: bnez a0, .LBB106_4
> ; RV64I-NEXT: .LBB106_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: and a1, a0, s0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s1, a1, .LBB106_1
> +; RV64I-NEXT: and a0, a1, s0
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bgeu s1, a0, .LBB106_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB106_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB106_1
> ; RV64I-NEXT: .LBB106_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s4, 16(sp)
> ; RV64I-NEXT: ld s3, 24(sp)
> ; RV64I-NEXT: ld s2, 32(sp)
> @@ -9741,28 +9741,28 @@ define i16 @atomicrmw_umin_i16_acquire(i
> ;
> ; RV64IA-LABEL: atomicrmw_umin_i16_acquire:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a6, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB106_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a2, a4, a6
> -; RV64IA-NEXT: mv a5, a4
> +; RV64IA-NEXT: lr.w.aq a3, (a6)
> +; RV64IA-NEXT: and a2, a3, a4
> +; RV64IA-NEXT: mv a5, a3
> ; RV64IA-NEXT: bgeu a1, a2, .LBB106_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB106_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> -; RV64IA-NEXT: xor a5, a4, a5
> +; RV64IA-NEXT: xor a5, a3, a1
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> ; RV64IA-NEXT: .LBB106_3: # in Loop: Header=BB106_1 Depth=1
> -; RV64IA-NEXT: sc.w a5, a5, (a0)
> +; RV64IA-NEXT: sc.w a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB106_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umin i16* %a, i16 %b acquire
> ret i16 %1
> @@ -9780,33 +9780,33 @@ define i16 @atomicrmw_umin_i16_release(i
> ; RV32I-NEXT: sw s4, 8(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s4, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: lui a1, 16
> -; RV32I-NEXT: addi s0, a1, -1
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: lui a0, 16
> +; RV32I-NEXT: addi s0, a0, -1
> ; RV32I-NEXT: and s1, s2, s0
> ; RV32I-NEXT: addi s3, sp, 6
> ; RV32I-NEXT: j .LBB107_2
> ; RV32I-NEXT: .LBB107_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB107_2 Depth=1
> -; RV32I-NEXT: sh a0, 6(sp)
> +; RV32I-NEXT: sh a1, 6(sp)
> +; RV32I-NEXT: addi a3, zero, 3
> ; RV32I-NEXT: mv a0, s4
> ; RV32I-NEXT: mv a1, s3
> -; RV32I-NEXT: addi a3, zero, 3
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 6(sp)
> -; RV32I-NEXT: bnez a1, .LBB107_4
> +; RV32I-NEXT: lh a1, 6(sp)
> +; RV32I-NEXT: bnez a0, .LBB107_4
> ; RV32I-NEXT: .LBB107_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: and a1, a0, s0
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bgeu s1, a1, .LBB107_1
> +; RV32I-NEXT: and a0, a1, s0
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bgeu s1, a0, .LBB107_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB107_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB107_1
> ; RV32I-NEXT: .LBB107_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s4, 8(sp)
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> @@ -9818,28 +9818,28 @@ define i16 @atomicrmw_umin_i16_release(i
> ;
> ; RV32IA-LABEL: atomicrmw_umin_i16_release:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a6, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB107_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w a4, (a0)
> -; RV32IA-NEXT: and a2, a4, a6
> -; RV32IA-NEXT: mv a5, a4
> +; RV32IA-NEXT: lr.w a3, (a6)
> +; RV32IA-NEXT: and a2, a3, a4
> +; RV32IA-NEXT: mv a5, a3
> ; RV32IA-NEXT: bgeu a1, a2, .LBB107_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB107_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> -; RV32IA-NEXT: xor a5, a4, a5
> +; RV32IA-NEXT: xor a5, a3, a1
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> ; RV32IA-NEXT: .LBB107_3: # in Loop: Header=BB107_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB107_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umin_i16_release:
> @@ -9853,33 +9853,33 @@ define i16 @atomicrmw_umin_i16_release(i
> ; RV64I-NEXT: sd s4, 16(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s4, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: lui a1, 16
> -; RV64I-NEXT: addiw s0, a1, -1
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: lui a0, 16
> +; RV64I-NEXT: addiw s0, a0, -1
> ; RV64I-NEXT: and s1, s2, s0
> ; RV64I-NEXT: addi s3, sp, 14
> ; RV64I-NEXT: j .LBB107_2
> ; RV64I-NEXT: .LBB107_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB107_2 Depth=1
> -; RV64I-NEXT: sh a0, 14(sp)
> +; RV64I-NEXT: sh a1, 14(sp)
> +; RV64I-NEXT: addi a3, zero, 3
> ; RV64I-NEXT: mv a0, s4
> ; RV64I-NEXT: mv a1, s3
> -; RV64I-NEXT: addi a3, zero, 3
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 14(sp)
> -; RV64I-NEXT: bnez a1, .LBB107_4
> +; RV64I-NEXT: lh a1, 14(sp)
> +; RV64I-NEXT: bnez a0, .LBB107_4
> ; RV64I-NEXT: .LBB107_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: and a1, a0, s0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s1, a1, .LBB107_1
> +; RV64I-NEXT: and a0, a1, s0
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bgeu s1, a0, .LBB107_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB107_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB107_1
> ; RV64I-NEXT: .LBB107_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s4, 16(sp)
> ; RV64I-NEXT: ld s3, 24(sp)
> ; RV64I-NEXT: ld s2, 32(sp)
> @@ -9891,28 +9891,28 @@ define i16 @atomicrmw_umin_i16_release(i
> ;
> ; RV64IA-LABEL: atomicrmw_umin_i16_release:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a6, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB107_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w a4, (a0)
> -; RV64IA-NEXT: and a2, a4, a6
> -; RV64IA-NEXT: mv a5, a4
> +; RV64IA-NEXT: lr.w a3, (a6)
> +; RV64IA-NEXT: and a2, a3, a4
> +; RV64IA-NEXT: mv a5, a3
> ; RV64IA-NEXT: bgeu a1, a2, .LBB107_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB107_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> -; RV64IA-NEXT: xor a5, a4, a5
> +; RV64IA-NEXT: xor a5, a3, a1
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> ; RV64IA-NEXT: .LBB107_3: # in Loop: Header=BB107_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB107_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umin i16* %a, i16 %b release
> ret i16 %1
> @@ -9930,33 +9930,33 @@ define i16 @atomicrmw_umin_i16_acq_rel(i
> ; RV32I-NEXT: sw s4, 8(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s4, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: lui a1, 16
> -; RV32I-NEXT: addi s0, a1, -1
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: lui a0, 16
> +; RV32I-NEXT: addi s0, a0, -1
> ; RV32I-NEXT: and s1, s2, s0
> ; RV32I-NEXT: addi s3, sp, 6
> ; RV32I-NEXT: j .LBB108_2
> ; RV32I-NEXT: .LBB108_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB108_2 Depth=1
> -; RV32I-NEXT: sh a0, 6(sp)
> -; RV32I-NEXT: mv a0, s4
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sh a1, 6(sp)
> ; RV32I-NEXT: addi a3, zero, 4
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s4
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 6(sp)
> -; RV32I-NEXT: bnez a1, .LBB108_4
> +; RV32I-NEXT: lh a1, 6(sp)
> +; RV32I-NEXT: bnez a0, .LBB108_4
> ; RV32I-NEXT: .LBB108_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: and a1, a0, s0
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bgeu s1, a1, .LBB108_1
> +; RV32I-NEXT: and a0, a1, s0
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bgeu s1, a0, .LBB108_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB108_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB108_1
> ; RV32I-NEXT: .LBB108_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s4, 8(sp)
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> @@ -9968,28 +9968,28 @@ define i16 @atomicrmw_umin_i16_acq_rel(i
> ;
> ; RV32IA-LABEL: atomicrmw_umin_i16_acq_rel:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a6, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB108_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aq a4, (a0)
> -; RV32IA-NEXT: and a2, a4, a6
> -; RV32IA-NEXT: mv a5, a4
> +; RV32IA-NEXT: lr.w.aq a3, (a6)
> +; RV32IA-NEXT: and a2, a3, a4
> +; RV32IA-NEXT: mv a5, a3
> ; RV32IA-NEXT: bgeu a1, a2, .LBB108_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB108_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> -; RV32IA-NEXT: xor a5, a4, a5
> +; RV32IA-NEXT: xor a5, a3, a1
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> ; RV32IA-NEXT: .LBB108_3: # in Loop: Header=BB108_1 Depth=1
> -; RV32IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB108_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umin_i16_acq_rel:
> @@ -10003,33 +10003,33 @@ define i16 @atomicrmw_umin_i16_acq_rel(i
> ; RV64I-NEXT: sd s4, 16(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s4, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: lui a1, 16
> -; RV64I-NEXT: addiw s0, a1, -1
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: lui a0, 16
> +; RV64I-NEXT: addiw s0, a0, -1
> ; RV64I-NEXT: and s1, s2, s0
> ; RV64I-NEXT: addi s3, sp, 14
> ; RV64I-NEXT: j .LBB108_2
> ; RV64I-NEXT: .LBB108_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB108_2 Depth=1
> -; RV64I-NEXT: sh a0, 14(sp)
> -; RV64I-NEXT: mv a0, s4
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sh a1, 14(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s4
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 14(sp)
> -; RV64I-NEXT: bnez a1, .LBB108_4
> +; RV64I-NEXT: lh a1, 14(sp)
> +; RV64I-NEXT: bnez a0, .LBB108_4
> ; RV64I-NEXT: .LBB108_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: and a1, a0, s0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s1, a1, .LBB108_1
> +; RV64I-NEXT: and a0, a1, s0
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bgeu s1, a0, .LBB108_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB108_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB108_1
> ; RV64I-NEXT: .LBB108_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s4, 16(sp)
> ; RV64I-NEXT: ld s3, 24(sp)
> ; RV64I-NEXT: ld s2, 32(sp)
> @@ -10041,28 +10041,28 @@ define i16 @atomicrmw_umin_i16_acq_rel(i
> ;
> ; RV64IA-LABEL: atomicrmw_umin_i16_acq_rel:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a6, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB108_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aq a4, (a0)
> -; RV64IA-NEXT: and a2, a4, a6
> -; RV64IA-NEXT: mv a5, a4
> +; RV64IA-NEXT: lr.w.aq a3, (a6)
> +; RV64IA-NEXT: and a2, a3, a4
> +; RV64IA-NEXT: mv a5, a3
> ; RV64IA-NEXT: bgeu a1, a2, .LBB108_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB108_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> -; RV64IA-NEXT: xor a5, a4, a5
> +; RV64IA-NEXT: xor a5, a3, a1
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> ; RV64IA-NEXT: .LBB108_3: # in Loop: Header=BB108_1 Depth=1
> -; RV64IA-NEXT: sc.w.rl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.rl a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB108_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umin i16* %a, i16 %b acq_rel
> ret i16 %1
> @@ -10080,33 +10080,33 @@ define i16 @atomicrmw_umin_i16_seq_cst(i
> ; RV32I-NEXT: sw s4, 8(sp)
> ; RV32I-NEXT: mv s2, a1
> ; RV32I-NEXT: mv s4, a0
> -; RV32I-NEXT: lhu a0, 0(a0)
> -; RV32I-NEXT: lui a1, 16
> -; RV32I-NEXT: addi s0, a1, -1
> +; RV32I-NEXT: lhu a1, 0(a0)
> +; RV32I-NEXT: lui a0, 16
> +; RV32I-NEXT: addi s0, a0, -1
> ; RV32I-NEXT: and s1, s2, s0
> ; RV32I-NEXT: addi s3, sp, 6
> ; RV32I-NEXT: j .LBB109_2
> ; RV32I-NEXT: .LBB109_1: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB109_2 Depth=1
> -; RV32I-NEXT: sh a0, 6(sp)
> -; RV32I-NEXT: mv a0, s4
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sh a1, 6(sp)
> ; RV32I-NEXT: addi a3, zero, 5
> ; RV32I-NEXT: addi a4, zero, 5
> +; RV32I-NEXT: mv a0, s4
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_2
> -; RV32I-NEXT: mv a1, a0
> -; RV32I-NEXT: lh a0, 6(sp)
> -; RV32I-NEXT: bnez a1, .LBB109_4
> +; RV32I-NEXT: lh a1, 6(sp)
> +; RV32I-NEXT: bnez a0, .LBB109_4
> ; RV32I-NEXT: .LBB109_2: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: and a1, a0, s0
> -; RV32I-NEXT: mv a2, a0
> -; RV32I-NEXT: bgeu s1, a1, .LBB109_1
> +; RV32I-NEXT: and a0, a1, s0
> +; RV32I-NEXT: mv a2, a1
> +; RV32I-NEXT: bgeu s1, a0, .LBB109_1
> ; RV32I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB109_2 Depth=1
> ; RV32I-NEXT: mv a2, s2
> ; RV32I-NEXT: j .LBB109_1
> ; RV32I-NEXT: .LBB109_4: # %atomicrmw.end
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: lw s4, 8(sp)
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> @@ -10118,28 +10118,28 @@ define i16 @atomicrmw_umin_i16_seq_cst(i
> ;
> ; RV32IA-LABEL: atomicrmw_umin_i16_seq_cst:
> ; RV32IA: # %bb.0:
> -; RV32IA-NEXT: lui a2, 16
> -; RV32IA-NEXT: addi a2, a2, -1
> -; RV32IA-NEXT: and a1, a1, a2
> -; RV32IA-NEXT: slli a3, a0, 3
> -; RV32IA-NEXT: andi a3, a3, 24
> -; RV32IA-NEXT: sll a6, a2, a3
> -; RV32IA-NEXT: sll a1, a1, a3
> -; RV32IA-NEXT: andi a0, a0, -4
> +; RV32IA-NEXT: andi a6, a0, -4
> +; RV32IA-NEXT: slli a0, a0, 3
> +; RV32IA-NEXT: andi a0, a0, 24
> +; RV32IA-NEXT: lui a3, 16
> +; RV32IA-NEXT: addi a3, a3, -1
> +; RV32IA-NEXT: sll a4, a3, a0
> +; RV32IA-NEXT: and a1, a1, a3
> +; RV32IA-NEXT: sll a1, a1, a0
> ; RV32IA-NEXT: .LBB109_1: # =>This Inner Loop Header: Depth=1
> -; RV32IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV32IA-NEXT: and a2, a4, a6
> -; RV32IA-NEXT: mv a5, a4
> +; RV32IA-NEXT: lr.w.aqrl a3, (a6)
> +; RV32IA-NEXT: and a2, a3, a4
> +; RV32IA-NEXT: mv a5, a3
> ; RV32IA-NEXT: bgeu a1, a2, .LBB109_3
> ; RV32IA-NEXT: # %bb.2: # in Loop: Header=BB109_1 Depth=1
> -; RV32IA-NEXT: xor a5, a4, a1
> -; RV32IA-NEXT: and a5, a5, a6
> -; RV32IA-NEXT: xor a5, a4, a5
> +; RV32IA-NEXT: xor a5, a3, a1
> +; RV32IA-NEXT: and a5, a5, a4
> +; RV32IA-NEXT: xor a5, a3, a5
> ; RV32IA-NEXT: .LBB109_3: # in Loop: Header=BB109_1 Depth=1
> -; RV32IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV32IA-NEXT: sc.w.aqrl a5, a5, (a6)
> ; RV32IA-NEXT: bnez a5, .LBB109_1
> ; RV32IA-NEXT: # %bb.4:
> -; RV32IA-NEXT: srl a0, a4, a3
> +; RV32IA-NEXT: srl a0, a3, a0
> ; RV32IA-NEXT: ret
> ;
> ; RV64I-LABEL: atomicrmw_umin_i16_seq_cst:
> @@ -10153,33 +10153,33 @@ define i16 @atomicrmw_umin_i16_seq_cst(i
> ; RV64I-NEXT: sd s4, 16(sp)
> ; RV64I-NEXT: mv s2, a1
> ; RV64I-NEXT: mv s4, a0
> -; RV64I-NEXT: lhu a0, 0(a0)
> -; RV64I-NEXT: lui a1, 16
> -; RV64I-NEXT: addiw s0, a1, -1
> +; RV64I-NEXT: lhu a1, 0(a0)
> +; RV64I-NEXT: lui a0, 16
> +; RV64I-NEXT: addiw s0, a0, -1
> ; RV64I-NEXT: and s1, s2, s0
> ; RV64I-NEXT: addi s3, sp, 14
> ; RV64I-NEXT: j .LBB109_2
> ; RV64I-NEXT: .LBB109_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB109_2 Depth=1
> -; RV64I-NEXT: sh a0, 14(sp)
> -; RV64I-NEXT: mv a0, s4
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sh a1, 14(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s4
> +; RV64I-NEXT: mv a1, s3
> ; RV64I-NEXT: call __atomic_compare_exchange_2
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lh a0, 14(sp)
> -; RV64I-NEXT: bnez a1, .LBB109_4
> +; RV64I-NEXT: lh a1, 14(sp)
> +; RV64I-NEXT: bnez a0, .LBB109_4
> ; RV64I-NEXT: .LBB109_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: and a1, a0, s0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s1, a1, .LBB109_1
> +; RV64I-NEXT: and a0, a1, s0
> +; RV64I-NEXT: mv a2, a1
> +; RV64I-NEXT: bgeu s1, a0, .LBB109_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB109_2 Depth=1
> ; RV64I-NEXT: mv a2, s2
> ; RV64I-NEXT: j .LBB109_1
> ; RV64I-NEXT: .LBB109_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ld s4, 16(sp)
> ; RV64I-NEXT: ld s3, 24(sp)
> ; RV64I-NEXT: ld s2, 32(sp)
> @@ -10191,28 +10191,28 @@ define i16 @atomicrmw_umin_i16_seq_cst(i
> ;
> ; RV64IA-LABEL: atomicrmw_umin_i16_seq_cst:
> ; RV64IA: # %bb.0:
> -; RV64IA-NEXT: lui a2, 16
> -; RV64IA-NEXT: addiw a2, a2, -1
> -; RV64IA-NEXT: and a1, a1, a2
> -; RV64IA-NEXT: slli a3, a0, 3
> -; RV64IA-NEXT: andi a3, a3, 24
> -; RV64IA-NEXT: sllw a6, a2, a3
> -; RV64IA-NEXT: sllw a1, a1, a3
> -; RV64IA-NEXT: andi a0, a0, -4
> +; RV64IA-NEXT: andi a6, a0, -4
> +; RV64IA-NEXT: slli a0, a0, 3
> +; RV64IA-NEXT: andi a0, a0, 24
> +; RV64IA-NEXT: lui a3, 16
> +; RV64IA-NEXT: addiw a3, a3, -1
> +; RV64IA-NEXT: sllw a4, a3, a0
> +; RV64IA-NEXT: and a1, a1, a3
> +; RV64IA-NEXT: sllw a1, a1, a0
> ; RV64IA-NEXT: .LBB109_1: # =>This Inner Loop Header: Depth=1
> -; RV64IA-NEXT: lr.w.aqrl a4, (a0)
> -; RV64IA-NEXT: and a2, a4, a6
> -; RV64IA-NEXT: mv a5, a4
> +; RV64IA-NEXT: lr.w.aqrl a3, (a6)
> +; RV64IA-NEXT: and a2, a3, a4
> +; RV64IA-NEXT: mv a5, a3
> ; RV64IA-NEXT: bgeu a1, a2, .LBB109_3
> ; RV64IA-NEXT: # %bb.2: # in Loop: Header=BB109_1 Depth=1
> -; RV64IA-NEXT: xor a5, a4, a1
> -; RV64IA-NEXT: and a5, a5, a6
> -; RV64IA-NEXT: xor a5, a4, a5
> +; RV64IA-NEXT: xor a5, a3, a1
> +; RV64IA-NEXT: and a5, a5, a4
> +; RV64IA-NEXT: xor a5, a3, a5
> ; RV64IA-NEXT: .LBB109_3: # in Loop: Header=BB109_1 Depth=1
> -; RV64IA-NEXT: sc.w.aqrl a5, a5, (a0)
> +; RV64IA-NEXT: sc.w.aqrl a5, a5, (a6)
> ; RV64IA-NEXT: bnez a5, .LBB109_1
> ; RV64IA-NEXT: # %bb.4:
> -; RV64IA-NEXT: srlw a0, a4, a3
> +; RV64IA-NEXT: srlw a0, a3, a0
> ; RV64IA-NEXT: ret
> %1 = atomicrmw umin i16* %a, i16 %b seq_cst
> ret i16 %1
> @@ -11496,30 +11496,31 @@ define i32 @atomicrmw_max_i32_monotonic(
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bge s0, a2, .LBB145_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bge s1, a3, .LBB145_3
> ; RV32I-NEXT: .LBB145_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: sw a3, 12(sp)
> +; RV32I-NEXT: mv a0, s0
> ; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: mv a3, zero
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB145_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB145_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: blt s0, a2, .LBB145_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: blt s1, a3, .LBB145_1
> ; RV32I-NEXT: .LBB145_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB145_1
> ; RV32I-NEXT: .LBB145_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -11540,33 +11541,33 @@ define i32 @atomicrmw_max_i32_monotonic(
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB145_2
> ; RV64I-NEXT: .LBB145_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB145_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB145_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB145_4
> ; RV64I-NEXT: .LBB145_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB145_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a0, .LBB145_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB145_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB145_1
> ; RV64I-NEXT: .LBB145_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -11591,30 +11592,31 @@ define i32 @atomicrmw_max_i32_acquire(i3
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bge s0, a2, .LBB146_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bge s1, a3, .LBB146_3
> ; RV32I-NEXT: .LBB146_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 2
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB146_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB146_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: blt s0, a2, .LBB146_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: blt s1, a3, .LBB146_1
> ; RV32I-NEXT: .LBB146_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB146_1
> ; RV32I-NEXT: .LBB146_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -11635,33 +11637,33 @@ define i32 @atomicrmw_max_i32_acquire(i3
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB146_2
> ; RV64I-NEXT: .LBB146_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB146_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB146_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB146_4
> ; RV64I-NEXT: .LBB146_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB146_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a0, .LBB146_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB146_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB146_1
> ; RV64I-NEXT: .LBB146_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -11686,30 +11688,31 @@ define i32 @atomicrmw_max_i32_release(i3
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bge s0, a2, .LBB147_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bge s1, a3, .LBB147_3
> ; RV32I-NEXT: .LBB147_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 3
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB147_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB147_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: blt s0, a2, .LBB147_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: blt s1, a3, .LBB147_1
> ; RV32I-NEXT: .LBB147_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB147_1
> ; RV32I-NEXT: .LBB147_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -11730,33 +11733,33 @@ define i32 @atomicrmw_max_i32_release(i3
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB147_2
> ; RV64I-NEXT: .LBB147_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB147_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 3
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB147_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB147_4
> ; RV64I-NEXT: .LBB147_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB147_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a0, .LBB147_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB147_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB147_1
> ; RV64I-NEXT: .LBB147_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -11781,30 +11784,31 @@ define i32 @atomicrmw_max_i32_acq_rel(i3
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bge s0, a2, .LBB148_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bge s1, a3, .LBB148_3
> ; RV32I-NEXT: .LBB148_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 4
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB148_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB148_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: blt s0, a2, .LBB148_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: blt s1, a3, .LBB148_1
> ; RV32I-NEXT: .LBB148_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB148_1
> ; RV32I-NEXT: .LBB148_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -11825,33 +11829,33 @@ define i32 @atomicrmw_max_i32_acq_rel(i3
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB148_2
> ; RV64I-NEXT: .LBB148_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB148_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB148_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB148_4
> ; RV64I-NEXT: .LBB148_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB148_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a0, .LBB148_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB148_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB148_1
> ; RV64I-NEXT: .LBB148_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -11876,30 +11880,31 @@ define i32 @atomicrmw_max_i32_seq_cst(i3
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bge s0, a2, .LBB149_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bge s1, a3, .LBB149_3
> ; RV32I-NEXT: .LBB149_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 5
> ; RV32I-NEXT: addi a4, zero, 5
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB149_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB149_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: blt s0, a2, .LBB149_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: blt s1, a3, .LBB149_1
> ; RV32I-NEXT: .LBB149_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB149_1
> ; RV32I-NEXT: .LBB149_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -11920,33 +11925,33 @@ define i32 @atomicrmw_max_i32_seq_cst(i3
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB149_2
> ; RV64I-NEXT: .LBB149_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB149_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB149_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB149_4
> ; RV64I-NEXT: .LBB149_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: blt s0, a1, .LBB149_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a0, .LBB149_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB149_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB149_1
> ; RV64I-NEXT: .LBB149_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -11971,30 +11976,31 @@ define i32 @atomicrmw_min_i32_monotonic(
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: blt s0, a2, .LBB150_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: blt s1, a3, .LBB150_3
> ; RV32I-NEXT: .LBB150_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: sw a3, 12(sp)
> +; RV32I-NEXT: mv a0, s0
> ; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: mv a3, zero
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB150_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB150_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bge s0, a2, .LBB150_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bge s1, a3, .LBB150_1
> ; RV32I-NEXT: .LBB150_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB150_1
> ; RV32I-NEXT: .LBB150_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -12015,33 +12021,33 @@ define i32 @atomicrmw_min_i32_monotonic(
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB150_2
> ; RV64I-NEXT: .LBB150_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB150_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB150_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB150_4
> ; RV64I-NEXT: .LBB150_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB150_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a0, .LBB150_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB150_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB150_1
> ; RV64I-NEXT: .LBB150_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -12066,30 +12072,31 @@ define i32 @atomicrmw_min_i32_acquire(i3
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: blt s0, a2, .LBB151_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: blt s1, a3, .LBB151_3
> ; RV32I-NEXT: .LBB151_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 2
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB151_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB151_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bge s0, a2, .LBB151_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bge s1, a3, .LBB151_1
> ; RV32I-NEXT: .LBB151_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB151_1
> ; RV32I-NEXT: .LBB151_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -12110,33 +12117,33 @@ define i32 @atomicrmw_min_i32_acquire(i3
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB151_2
> ; RV64I-NEXT: .LBB151_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB151_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB151_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB151_4
> ; RV64I-NEXT: .LBB151_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB151_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a0, .LBB151_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB151_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB151_1
> ; RV64I-NEXT: .LBB151_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -12161,30 +12168,31 @@ define i32 @atomicrmw_min_i32_release(i3
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: blt s0, a2, .LBB152_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: blt s1, a3, .LBB152_3
> ; RV32I-NEXT: .LBB152_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 3
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB152_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB152_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bge s0, a2, .LBB152_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bge s1, a3, .LBB152_1
> ; RV32I-NEXT: .LBB152_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB152_1
> ; RV32I-NEXT: .LBB152_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -12205,33 +12213,33 @@ define i32 @atomicrmw_min_i32_release(i3
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB152_2
> ; RV64I-NEXT: .LBB152_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB152_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 3
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB152_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB152_4
> ; RV64I-NEXT: .LBB152_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB152_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a0, .LBB152_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB152_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB152_1
> ; RV64I-NEXT: .LBB152_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -12256,30 +12264,31 @@ define i32 @atomicrmw_min_i32_acq_rel(i3
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: blt s0, a2, .LBB153_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: blt s1, a3, .LBB153_3
> ; RV32I-NEXT: .LBB153_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 4
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB153_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB153_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bge s0, a2, .LBB153_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bge s1, a3, .LBB153_1
> ; RV32I-NEXT: .LBB153_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB153_1
> ; RV32I-NEXT: .LBB153_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -12300,33 +12309,33 @@ define i32 @atomicrmw_min_i32_acq_rel(i3
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB153_2
> ; RV64I-NEXT: .LBB153_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB153_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB153_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB153_4
> ; RV64I-NEXT: .LBB153_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB153_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a0, .LBB153_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB153_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB153_1
> ; RV64I-NEXT: .LBB153_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -12351,30 +12360,31 @@ define i32 @atomicrmw_min_i32_seq_cst(i3
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: blt s0, a2, .LBB154_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: blt s1, a3, .LBB154_3
> ; RV32I-NEXT: .LBB154_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 5
> ; RV32I-NEXT: addi a4, zero, 5
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB154_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB154_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bge s0, a2, .LBB154_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bge s1, a3, .LBB154_1
> ; RV32I-NEXT: .LBB154_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB154_1
> ; RV32I-NEXT: .LBB154_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -12395,33 +12405,33 @@ define i32 @atomicrmw_min_i32_seq_cst(i3
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB154_2
> ; RV64I-NEXT: .LBB154_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB154_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB154_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB154_4
> ; RV64I-NEXT: .LBB154_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bge s0, a1, .LBB154_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a0, .LBB154_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB154_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB154_1
> ; RV64I-NEXT: .LBB154_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -12446,30 +12456,31 @@ define i32 @atomicrmw_umax_i32_monotonic
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bgeu s0, a2, .LBB155_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a3, .LBB155_3
> ; RV32I-NEXT: .LBB155_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: sw a3, 12(sp)
> +; RV32I-NEXT: mv a0, s0
> ; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: mv a3, zero
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB155_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB155_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bltu s0, a2, .LBB155_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a3, .LBB155_1
> ; RV32I-NEXT: .LBB155_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB155_1
> ; RV32I-NEXT: .LBB155_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -12490,33 +12501,33 @@ define i32 @atomicrmw_umax_i32_monotonic
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB155_2
> ; RV64I-NEXT: .LBB155_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB155_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB155_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB155_4
> ; RV64I-NEXT: .LBB155_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s0, a1, .LBB155_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a0, .LBB155_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB155_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB155_1
> ; RV64I-NEXT: .LBB155_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -12541,30 +12552,31 @@ define i32 @atomicrmw_umax_i32_acquire(i
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bgeu s0, a2, .LBB156_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a3, .LBB156_3
> ; RV32I-NEXT: .LBB156_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 2
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB156_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB156_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bltu s0, a2, .LBB156_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a3, .LBB156_1
> ; RV32I-NEXT: .LBB156_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB156_1
> ; RV32I-NEXT: .LBB156_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -12585,33 +12597,33 @@ define i32 @atomicrmw_umax_i32_acquire(i
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB156_2
> ; RV64I-NEXT: .LBB156_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB156_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB156_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB156_4
> ; RV64I-NEXT: .LBB156_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s0, a1, .LBB156_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a0, .LBB156_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB156_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB156_1
> ; RV64I-NEXT: .LBB156_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -12636,30 +12648,31 @@ define i32 @atomicrmw_umax_i32_release(i
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bgeu s0, a2, .LBB157_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a3, .LBB157_3
> ; RV32I-NEXT: .LBB157_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 3
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB157_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB157_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bltu s0, a2, .LBB157_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a3, .LBB157_1
> ; RV32I-NEXT: .LBB157_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB157_1
> ; RV32I-NEXT: .LBB157_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -12680,33 +12693,33 @@ define i32 @atomicrmw_umax_i32_release(i
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB157_2
> ; RV64I-NEXT: .LBB157_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB157_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 3
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB157_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB157_4
> ; RV64I-NEXT: .LBB157_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s0, a1, .LBB157_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a0, .LBB157_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB157_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB157_1
> ; RV64I-NEXT: .LBB157_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -12731,30 +12744,31 @@ define i32 @atomicrmw_umax_i32_acq_rel(i
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bgeu s0, a2, .LBB158_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a3, .LBB158_3
> ; RV32I-NEXT: .LBB158_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 4
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB158_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB158_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bltu s0, a2, .LBB158_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a3, .LBB158_1
> ; RV32I-NEXT: .LBB158_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB158_1
> ; RV32I-NEXT: .LBB158_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -12775,33 +12789,33 @@ define i32 @atomicrmw_umax_i32_acq_rel(i
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB158_2
> ; RV64I-NEXT: .LBB158_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB158_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB158_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB158_4
> ; RV64I-NEXT: .LBB158_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s0, a1, .LBB158_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a0, .LBB158_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB158_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB158_1
> ; RV64I-NEXT: .LBB158_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -12826,30 +12840,31 @@ define i32 @atomicrmw_umax_i32_seq_cst(i
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bgeu s0, a2, .LBB159_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a3, .LBB159_3
> ; RV32I-NEXT: .LBB159_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 5
> ; RV32I-NEXT: addi a4, zero, 5
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB159_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB159_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bltu s0, a2, .LBB159_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a3, .LBB159_1
> ; RV32I-NEXT: .LBB159_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB159_1
> ; RV32I-NEXT: .LBB159_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -12870,33 +12885,33 @@ define i32 @atomicrmw_umax_i32_seq_cst(i
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB159_2
> ; RV64I-NEXT: .LBB159_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB159_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB159_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB159_4
> ; RV64I-NEXT: .LBB159_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bltu s0, a1, .LBB159_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a0, .LBB159_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB159_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB159_1
> ; RV64I-NEXT: .LBB159_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -12921,30 +12936,31 @@ define i32 @atomicrmw_umin_i32_monotonic
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bltu s0, a2, .LBB160_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a3, .LBB160_3
> ; RV32I-NEXT: .LBB160_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: sw a3, 12(sp)
> +; RV32I-NEXT: mv a0, s0
> ; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: mv a3, zero
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB160_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB160_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bgeu s0, a2, .LBB160_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a3, .LBB160_1
> ; RV32I-NEXT: .LBB160_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB160_1
> ; RV32I-NEXT: .LBB160_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -12965,33 +12981,33 @@ define i32 @atomicrmw_umin_i32_monotonic
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB160_2
> ; RV64I-NEXT: .LBB160_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB160_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB160_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB160_4
> ; RV64I-NEXT: .LBB160_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s0, a1, .LBB160_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a0, .LBB160_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB160_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB160_1
> ; RV64I-NEXT: .LBB160_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -13016,30 +13032,31 @@ define i32 @atomicrmw_umin_i32_acquire(i
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bltu s0, a2, .LBB161_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a3, .LBB161_3
> ; RV32I-NEXT: .LBB161_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 2
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB161_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB161_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bgeu s0, a2, .LBB161_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a3, .LBB161_1
> ; RV32I-NEXT: .LBB161_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB161_1
> ; RV32I-NEXT: .LBB161_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -13060,33 +13077,33 @@ define i32 @atomicrmw_umin_i32_acquire(i
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB161_2
> ; RV64I-NEXT: .LBB161_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB161_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB161_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB161_4
> ; RV64I-NEXT: .LBB161_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s0, a1, .LBB161_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a0, .LBB161_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB161_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB161_1
> ; RV64I-NEXT: .LBB161_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -13111,30 +13128,31 @@ define i32 @atomicrmw_umin_i32_release(i
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bltu s0, a2, .LBB162_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a3, .LBB162_3
> ; RV32I-NEXT: .LBB162_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 3
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB162_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB162_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bgeu s0, a2, .LBB162_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a3, .LBB162_1
> ; RV32I-NEXT: .LBB162_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB162_1
> ; RV32I-NEXT: .LBB162_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -13155,33 +13173,33 @@ define i32 @atomicrmw_umin_i32_release(i
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB162_2
> ; RV64I-NEXT: .LBB162_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB162_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 3
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB162_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB162_4
> ; RV64I-NEXT: .LBB162_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s0, a1, .LBB162_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a0, .LBB162_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB162_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB162_1
> ; RV64I-NEXT: .LBB162_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -13206,30 +13224,31 @@ define i32 @atomicrmw_umin_i32_acq_rel(i
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bltu s0, a2, .LBB163_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a3, .LBB163_3
> ; RV32I-NEXT: .LBB163_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 4
> ; RV32I-NEXT: addi a4, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB163_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB163_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bgeu s0, a2, .LBB163_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a3, .LBB163_1
> ; RV32I-NEXT: .LBB163_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB163_1
> ; RV32I-NEXT: .LBB163_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -13250,33 +13269,33 @@ define i32 @atomicrmw_umin_i32_acq_rel(i
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB163_2
> ; RV64I-NEXT: .LBB163_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB163_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB163_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB163_4
> ; RV64I-NEXT: .LBB163_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s0, a1, .LBB163_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a0, .LBB163_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB163_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB163_1
> ; RV64I-NEXT: .LBB163_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -13301,30 +13320,31 @@ define i32 @atomicrmw_umin_i32_seq_cst(i
> ; RV32I-NEXT: sw s0, 24(sp)
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> -; RV32I-NEXT: mv s0, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a2, 0(a0)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a3, 0(a0)
> +; RV32I-NEXT: mv s1, a1
> ; RV32I-NEXT: addi s2, sp, 12
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bltu s0, a2, .LBB164_3
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bltu s1, a3, .LBB164_3
> ; RV32I-NEXT: .LBB164_1: # %atomicrmw.start
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s2
> +; RV32I-NEXT: sw a3, 12(sp)
> ; RV32I-NEXT: addi a3, zero, 5
> ; RV32I-NEXT: addi a4, zero, 5
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s2
> ; RV32I-NEXT: call __atomic_compare_exchange_4
> -; RV32I-NEXT: lw a2, 12(sp)
> +; RV32I-NEXT: lw a3, 12(sp)
> ; RV32I-NEXT: bnez a0, .LBB164_4
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV32I-NEXT: # in Loop: Header=BB164_1 Depth=1
> -; RV32I-NEXT: sw a2, 12(sp)
> -; RV32I-NEXT: bgeu s0, a2, .LBB164_1
> +; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bgeu s1, a3, .LBB164_1
> ; RV32I-NEXT: .LBB164_3: # %atomicrmw.start
> -; RV32I-NEXT: mv a2, s0
> +; RV32I-NEXT: mv a2, s1
> ; RV32I-NEXT: j .LBB164_1
> ; RV32I-NEXT: .LBB164_4: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> ; RV32I-NEXT: lw s0, 24(sp)
> @@ -13345,33 +13365,33 @@ define i32 @atomicrmw_umin_i32_seq_cst(i
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> ; RV64I-NEXT: sd s3, 8(sp)
> -; RV64I-NEXT: mv s2, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: lwu a0, 0(a0)
> -; RV64I-NEXT: sext.w s0, a1
> -; RV64I-NEXT: addi s3, sp, 4
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: lwu a3, 0(a0)
> +; RV64I-NEXT: mv s3, a1
> +; RV64I-NEXT: sext.w s1, a1
> +; RV64I-NEXT: addi s2, sp, 4
> ; RV64I-NEXT: j .LBB164_2
> ; RV64I-NEXT: .LBB164_1: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB164_2 Depth=1
> -; RV64I-NEXT: sw a0, 4(sp)
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s3
> +; RV64I-NEXT: sw a3, 4(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_4
> -; RV64I-NEXT: mv a1, a0
> -; RV64I-NEXT: lw a0, 4(sp)
> -; RV64I-NEXT: bnez a1, .LBB164_4
> +; RV64I-NEXT: lw a3, 4(sp)
> +; RV64I-NEXT: bnez a0, .LBB164_4
> ; RV64I-NEXT: .LBB164_2: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sext.w a1, a0
> -; RV64I-NEXT: mv a2, a0
> -; RV64I-NEXT: bgeu s0, a1, .LBB164_1
> +; RV64I-NEXT: sext.w a0, a3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a0, .LBB164_1
> ; RV64I-NEXT: # %bb.3: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB164_2 Depth=1
> -; RV64I-NEXT: mv a2, s2
> +; RV64I-NEXT: mv a2, s3
> ; RV64I-NEXT: j .LBB164_1
> ; RV64I-NEXT: .LBB164_4: # %atomicrmw.end
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s3, 8(sp)
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> @@ -14802,41 +14822,43 @@ define i64 @atomicrmw_max_i64_monotonic(
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB200_3
> +; RV32I-NEXT: bne a5, s1, .LBB200_3
> ; RV32I-NEXT: j .LBB200_4
> ; RV32I-NEXT: .LBB200_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> +; RV32I-NEXT: mv a0, s0
> ; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: mv a5, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB200_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB200_4
> +; RV32I-NEXT: beq a5, s1, .LBB200_4
> ; RV32I-NEXT: .LBB200_3: # %atomicrmw.start
> -; RV32I-NEXT: slt a0, s0, a1
> +; RV32I-NEXT: slt a0, s1, a5
> ; RV32I-NEXT: j .LBB200_5
> ; RV32I-NEXT: .LBB200_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB200_5: # %atomicrmw.start
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB200_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB200_1
> ; RV32I-NEXT: .LBB200_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -14853,41 +14875,43 @@ define i64 @atomicrmw_max_i64_monotonic(
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB200_3
> +; RV32IA-NEXT: bne a5, s1, .LBB200_3
> ; RV32IA-NEXT: j .LBB200_4
> ; RV32IA-NEXT: .LBB200_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> +; RV32IA-NEXT: mv a0, s0
> ; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: mv a4, zero
> ; RV32IA-NEXT: mv a5, zero
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB200_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB200_4
> +; RV32IA-NEXT: beq a5, s1, .LBB200_4
> ; RV32IA-NEXT: .LBB200_3: # %atomicrmw.start
> -; RV32IA-NEXT: slt a0, s0, a1
> +; RV32IA-NEXT: slt a0, s1, a5
> ; RV32IA-NEXT: j .LBB200_5
> ; RV32IA-NEXT: .LBB200_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB200_5: # %atomicrmw.start
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB200_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB200_1
> ; RV32IA-NEXT: .LBB200_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -14903,30 +14927,31 @@ define i64 @atomicrmw_max_i64_monotonic(
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bge s0, a2, .LBB200_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a3, .LBB200_3
> ; RV64I-NEXT: .LBB200_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: sd a3, 8(sp)
> +; RV64I-NEXT: mv a0, s0
> ; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB200_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB200_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: blt s0, a2, .LBB200_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a3, .LBB200_1
> ; RV64I-NEXT: .LBB200_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB200_1
> ; RV64I-NEXT: .LBB200_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -14951,41 +14976,43 @@ define i64 @atomicrmw_max_i64_acquire(i6
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB201_3
> +; RV32I-NEXT: bne a5, s1, .LBB201_3
> ; RV32I-NEXT: j .LBB201_4
> ; RV32I-NEXT: .LBB201_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 2
> ; RV32I-NEXT: addi a5, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB201_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB201_4
> +; RV32I-NEXT: beq a5, s1, .LBB201_4
> ; RV32I-NEXT: .LBB201_3: # %atomicrmw.start
> -; RV32I-NEXT: slt a0, s0, a1
> +; RV32I-NEXT: slt a0, s1, a5
> ; RV32I-NEXT: j .LBB201_5
> ; RV32I-NEXT: .LBB201_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB201_5: # %atomicrmw.start
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB201_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB201_1
> ; RV32I-NEXT: .LBB201_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -15002,41 +15029,43 @@ define i64 @atomicrmw_max_i64_acquire(i6
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB201_3
> +; RV32IA-NEXT: bne a5, s1, .LBB201_3
> ; RV32IA-NEXT: j .LBB201_4
> ; RV32IA-NEXT: .LBB201_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 2
> ; RV32IA-NEXT: addi a5, zero, 2
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB201_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB201_4
> +; RV32IA-NEXT: beq a5, s1, .LBB201_4
> ; RV32IA-NEXT: .LBB201_3: # %atomicrmw.start
> -; RV32IA-NEXT: slt a0, s0, a1
> +; RV32IA-NEXT: slt a0, s1, a5
> ; RV32IA-NEXT: j .LBB201_5
> ; RV32IA-NEXT: .LBB201_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB201_5: # %atomicrmw.start
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB201_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB201_1
> ; RV32IA-NEXT: .LBB201_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -15052,30 +15081,31 @@ define i64 @atomicrmw_max_i64_acquire(i6
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bge s0, a2, .LBB201_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a3, .LBB201_3
> ; RV64I-NEXT: .LBB201_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB201_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB201_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: blt s0, a2, .LBB201_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a3, .LBB201_1
> ; RV64I-NEXT: .LBB201_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB201_1
> ; RV64I-NEXT: .LBB201_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -15100,41 +15130,43 @@ define i64 @atomicrmw_max_i64_release(i6
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB202_3
> +; RV32I-NEXT: bne a5, s1, .LBB202_3
> ; RV32I-NEXT: j .LBB202_4
> ; RV32I-NEXT: .LBB202_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 3
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a5, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB202_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB202_4
> +; RV32I-NEXT: beq a5, s1, .LBB202_4
> ; RV32I-NEXT: .LBB202_3: # %atomicrmw.start
> -; RV32I-NEXT: slt a0, s0, a1
> +; RV32I-NEXT: slt a0, s1, a5
> ; RV32I-NEXT: j .LBB202_5
> ; RV32I-NEXT: .LBB202_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB202_5: # %atomicrmw.start
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB202_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB202_1
> ; RV32I-NEXT: .LBB202_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -15151,41 +15183,43 @@ define i64 @atomicrmw_max_i64_release(i6
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB202_3
> +; RV32IA-NEXT: bne a5, s1, .LBB202_3
> ; RV32IA-NEXT: j .LBB202_4
> ; RV32IA-NEXT: .LBB202_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 3
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: mv a5, zero
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB202_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB202_4
> +; RV32IA-NEXT: beq a5, s1, .LBB202_4
> ; RV32IA-NEXT: .LBB202_3: # %atomicrmw.start
> -; RV32IA-NEXT: slt a0, s0, a1
> +; RV32IA-NEXT: slt a0, s1, a5
> ; RV32IA-NEXT: j .LBB202_5
> ; RV32IA-NEXT: .LBB202_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB202_5: # %atomicrmw.start
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB202_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB202_1
> ; RV32IA-NEXT: .LBB202_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -15201,30 +15235,31 @@ define i64 @atomicrmw_max_i64_release(i6
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bge s0, a2, .LBB202_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a3, .LBB202_3
> ; RV64I-NEXT: .LBB202_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 3
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB202_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB202_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: blt s0, a2, .LBB202_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a3, .LBB202_1
> ; RV64I-NEXT: .LBB202_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB202_1
> ; RV64I-NEXT: .LBB202_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -15249,41 +15284,43 @@ define i64 @atomicrmw_max_i64_acq_rel(i6
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB203_3
> +; RV32I-NEXT: bne a5, s1, .LBB203_3
> ; RV32I-NEXT: j .LBB203_4
> ; RV32I-NEXT: .LBB203_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 4
> ; RV32I-NEXT: addi a5, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB203_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB203_4
> +; RV32I-NEXT: beq a5, s1, .LBB203_4
> ; RV32I-NEXT: .LBB203_3: # %atomicrmw.start
> -; RV32I-NEXT: slt a0, s0, a1
> +; RV32I-NEXT: slt a0, s1, a5
> ; RV32I-NEXT: j .LBB203_5
> ; RV32I-NEXT: .LBB203_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB203_5: # %atomicrmw.start
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB203_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB203_1
> ; RV32I-NEXT: .LBB203_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -15300,41 +15337,43 @@ define i64 @atomicrmw_max_i64_acq_rel(i6
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB203_3
> +; RV32IA-NEXT: bne a5, s1, .LBB203_3
> ; RV32IA-NEXT: j .LBB203_4
> ; RV32IA-NEXT: .LBB203_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 4
> ; RV32IA-NEXT: addi a5, zero, 2
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB203_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB203_4
> +; RV32IA-NEXT: beq a5, s1, .LBB203_4
> ; RV32IA-NEXT: .LBB203_3: # %atomicrmw.start
> -; RV32IA-NEXT: slt a0, s0, a1
> +; RV32IA-NEXT: slt a0, s1, a5
> ; RV32IA-NEXT: j .LBB203_5
> ; RV32IA-NEXT: .LBB203_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB203_5: # %atomicrmw.start
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB203_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB203_1
> ; RV32IA-NEXT: .LBB203_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -15350,30 +15389,31 @@ define i64 @atomicrmw_max_i64_acq_rel(i6
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bge s0, a2, .LBB203_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a3, .LBB203_3
> ; RV64I-NEXT: .LBB203_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB203_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB203_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: blt s0, a2, .LBB203_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a3, .LBB203_1
> ; RV64I-NEXT: .LBB203_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB203_1
> ; RV64I-NEXT: .LBB203_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -15398,41 +15438,43 @@ define i64 @atomicrmw_max_i64_seq_cst(i6
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB204_3
> +; RV32I-NEXT: bne a5, s1, .LBB204_3
> ; RV32I-NEXT: j .LBB204_4
> ; RV32I-NEXT: .LBB204_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 5
> ; RV32I-NEXT: addi a5, zero, 5
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB204_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB204_4
> +; RV32I-NEXT: beq a5, s1, .LBB204_4
> ; RV32I-NEXT: .LBB204_3: # %atomicrmw.start
> -; RV32I-NEXT: slt a0, s0, a1
> +; RV32I-NEXT: slt a0, s1, a5
> ; RV32I-NEXT: j .LBB204_5
> ; RV32I-NEXT: .LBB204_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB204_5: # %atomicrmw.start
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB204_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB204_1
> ; RV32I-NEXT: .LBB204_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -15449,41 +15491,43 @@ define i64 @atomicrmw_max_i64_seq_cst(i6
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB204_3
> +; RV32IA-NEXT: bne a5, s1, .LBB204_3
> ; RV32IA-NEXT: j .LBB204_4
> ; RV32IA-NEXT: .LBB204_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 5
> ; RV32IA-NEXT: addi a5, zero, 5
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB204_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB204_4
> +; RV32IA-NEXT: beq a5, s1, .LBB204_4
> ; RV32IA-NEXT: .LBB204_3: # %atomicrmw.start
> -; RV32IA-NEXT: slt a0, s0, a1
> +; RV32IA-NEXT: slt a0, s1, a5
> ; RV32IA-NEXT: j .LBB204_5
> ; RV32IA-NEXT: .LBB204_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB204_5: # %atomicrmw.start
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB204_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB204_1
> ; RV32IA-NEXT: .LBB204_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -15499,30 +15543,31 @@ define i64 @atomicrmw_max_i64_seq_cst(i6
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bge s0, a2, .LBB204_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a3, .LBB204_3
> ; RV64I-NEXT: .LBB204_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB204_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB204_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: blt s0, a2, .LBB204_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a3, .LBB204_1
> ; RV64I-NEXT: .LBB204_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB204_1
> ; RV64I-NEXT: .LBB204_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -15547,42 +15592,44 @@ define i64 @atomicrmw_min_i64_monotonic(
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB205_3
> +; RV32I-NEXT: bne a5, s1, .LBB205_3
> ; RV32I-NEXT: j .LBB205_4
> ; RV32I-NEXT: .LBB205_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> +; RV32I-NEXT: mv a0, s0
> ; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: mv a5, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB205_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB205_4
> +; RV32I-NEXT: beq a5, s1, .LBB205_4
> ; RV32I-NEXT: .LBB205_3: # %atomicrmw.start
> -; RV32I-NEXT: slt a0, s0, a1
> +; RV32I-NEXT: slt a0, s1, a5
> ; RV32I-NEXT: j .LBB205_5
> ; RV32I-NEXT: .LBB205_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB205_5: # %atomicrmw.start
> ; RV32I-NEXT: xori a0, a0, 1
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB205_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB205_1
> ; RV32I-NEXT: .LBB205_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -15599,42 +15646,44 @@ define i64 @atomicrmw_min_i64_monotonic(
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB205_3
> +; RV32IA-NEXT: bne a5, s1, .LBB205_3
> ; RV32IA-NEXT: j .LBB205_4
> ; RV32IA-NEXT: .LBB205_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> +; RV32IA-NEXT: mv a0, s0
> ; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: mv a4, zero
> ; RV32IA-NEXT: mv a5, zero
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB205_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB205_4
> +; RV32IA-NEXT: beq a5, s1, .LBB205_4
> ; RV32IA-NEXT: .LBB205_3: # %atomicrmw.start
> -; RV32IA-NEXT: slt a0, s0, a1
> +; RV32IA-NEXT: slt a0, s1, a5
> ; RV32IA-NEXT: j .LBB205_5
> ; RV32IA-NEXT: .LBB205_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB205_5: # %atomicrmw.start
> ; RV32IA-NEXT: xori a0, a0, 1
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB205_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB205_1
> ; RV32IA-NEXT: .LBB205_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -15650,30 +15699,31 @@ define i64 @atomicrmw_min_i64_monotonic(
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: blt s0, a2, .LBB205_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a3, .LBB205_3
> ; RV64I-NEXT: .LBB205_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: sd a3, 8(sp)
> +; RV64I-NEXT: mv a0, s0
> ; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB205_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB205_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bge s0, a2, .LBB205_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a3, .LBB205_1
> ; RV64I-NEXT: .LBB205_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB205_1
> ; RV64I-NEXT: .LBB205_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -15698,42 +15748,44 @@ define i64 @atomicrmw_min_i64_acquire(i6
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB206_3
> +; RV32I-NEXT: bne a5, s1, .LBB206_3
> ; RV32I-NEXT: j .LBB206_4
> ; RV32I-NEXT: .LBB206_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 2
> ; RV32I-NEXT: addi a5, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB206_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB206_4
> +; RV32I-NEXT: beq a5, s1, .LBB206_4
> ; RV32I-NEXT: .LBB206_3: # %atomicrmw.start
> -; RV32I-NEXT: slt a0, s0, a1
> +; RV32I-NEXT: slt a0, s1, a5
> ; RV32I-NEXT: j .LBB206_5
> ; RV32I-NEXT: .LBB206_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB206_5: # %atomicrmw.start
> ; RV32I-NEXT: xori a0, a0, 1
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB206_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB206_1
> ; RV32I-NEXT: .LBB206_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -15750,42 +15802,44 @@ define i64 @atomicrmw_min_i64_acquire(i6
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB206_3
> +; RV32IA-NEXT: bne a5, s1, .LBB206_3
> ; RV32IA-NEXT: j .LBB206_4
> ; RV32IA-NEXT: .LBB206_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 2
> ; RV32IA-NEXT: addi a5, zero, 2
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB206_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB206_4
> +; RV32IA-NEXT: beq a5, s1, .LBB206_4
> ; RV32IA-NEXT: .LBB206_3: # %atomicrmw.start
> -; RV32IA-NEXT: slt a0, s0, a1
> +; RV32IA-NEXT: slt a0, s1, a5
> ; RV32IA-NEXT: j .LBB206_5
> ; RV32IA-NEXT: .LBB206_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB206_5: # %atomicrmw.start
> ; RV32IA-NEXT: xori a0, a0, 1
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB206_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB206_1
> ; RV32IA-NEXT: .LBB206_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -15801,30 +15855,31 @@ define i64 @atomicrmw_min_i64_acquire(i6
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: blt s0, a2, .LBB206_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a3, .LBB206_3
> ; RV64I-NEXT: .LBB206_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB206_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB206_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bge s0, a2, .LBB206_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a3, .LBB206_1
> ; RV64I-NEXT: .LBB206_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB206_1
> ; RV64I-NEXT: .LBB206_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -15849,42 +15904,44 @@ define i64 @atomicrmw_min_i64_release(i6
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB207_3
> +; RV32I-NEXT: bne a5, s1, .LBB207_3
> ; RV32I-NEXT: j .LBB207_4
> ; RV32I-NEXT: .LBB207_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 3
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a5, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB207_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB207_4
> +; RV32I-NEXT: beq a5, s1, .LBB207_4
> ; RV32I-NEXT: .LBB207_3: # %atomicrmw.start
> -; RV32I-NEXT: slt a0, s0, a1
> +; RV32I-NEXT: slt a0, s1, a5
> ; RV32I-NEXT: j .LBB207_5
> ; RV32I-NEXT: .LBB207_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB207_5: # %atomicrmw.start
> ; RV32I-NEXT: xori a0, a0, 1
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB207_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB207_1
> ; RV32I-NEXT: .LBB207_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -15901,42 +15958,44 @@ define i64 @atomicrmw_min_i64_release(i6
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB207_3
> +; RV32IA-NEXT: bne a5, s1, .LBB207_3
> ; RV32IA-NEXT: j .LBB207_4
> ; RV32IA-NEXT: .LBB207_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 3
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: mv a5, zero
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB207_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB207_4
> +; RV32IA-NEXT: beq a5, s1, .LBB207_4
> ; RV32IA-NEXT: .LBB207_3: # %atomicrmw.start
> -; RV32IA-NEXT: slt a0, s0, a1
> +; RV32IA-NEXT: slt a0, s1, a5
> ; RV32IA-NEXT: j .LBB207_5
> ; RV32IA-NEXT: .LBB207_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB207_5: # %atomicrmw.start
> ; RV32IA-NEXT: xori a0, a0, 1
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB207_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB207_1
> ; RV32IA-NEXT: .LBB207_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -15952,30 +16011,31 @@ define i64 @atomicrmw_min_i64_release(i6
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: blt s0, a2, .LBB207_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a3, .LBB207_3
> ; RV64I-NEXT: .LBB207_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 3
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB207_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB207_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bge s0, a2, .LBB207_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a3, .LBB207_1
> ; RV64I-NEXT: .LBB207_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB207_1
> ; RV64I-NEXT: .LBB207_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -16000,42 +16060,44 @@ define i64 @atomicrmw_min_i64_acq_rel(i6
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB208_3
> +; RV32I-NEXT: bne a5, s1, .LBB208_3
> ; RV32I-NEXT: j .LBB208_4
> ; RV32I-NEXT: .LBB208_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 4
> ; RV32I-NEXT: addi a5, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB208_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB208_4
> +; RV32I-NEXT: beq a5, s1, .LBB208_4
> ; RV32I-NEXT: .LBB208_3: # %atomicrmw.start
> -; RV32I-NEXT: slt a0, s0, a1
> +; RV32I-NEXT: slt a0, s1, a5
> ; RV32I-NEXT: j .LBB208_5
> ; RV32I-NEXT: .LBB208_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB208_5: # %atomicrmw.start
> ; RV32I-NEXT: xori a0, a0, 1
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB208_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB208_1
> ; RV32I-NEXT: .LBB208_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -16052,42 +16114,44 @@ define i64 @atomicrmw_min_i64_acq_rel(i6
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB208_3
> +; RV32IA-NEXT: bne a5, s1, .LBB208_3
> ; RV32IA-NEXT: j .LBB208_4
> ; RV32IA-NEXT: .LBB208_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 4
> ; RV32IA-NEXT: addi a5, zero, 2
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB208_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB208_4
> +; RV32IA-NEXT: beq a5, s1, .LBB208_4
> ; RV32IA-NEXT: .LBB208_3: # %atomicrmw.start
> -; RV32IA-NEXT: slt a0, s0, a1
> +; RV32IA-NEXT: slt a0, s1, a5
> ; RV32IA-NEXT: j .LBB208_5
> ; RV32IA-NEXT: .LBB208_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB208_5: # %atomicrmw.start
> ; RV32IA-NEXT: xori a0, a0, 1
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB208_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB208_1
> ; RV32IA-NEXT: .LBB208_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -16103,30 +16167,31 @@ define i64 @atomicrmw_min_i64_acq_rel(i6
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: blt s0, a2, .LBB208_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a3, .LBB208_3
> ; RV64I-NEXT: .LBB208_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB208_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB208_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bge s0, a2, .LBB208_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a3, .LBB208_1
> ; RV64I-NEXT: .LBB208_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB208_1
> ; RV64I-NEXT: .LBB208_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -16151,42 +16216,44 @@ define i64 @atomicrmw_min_i64_seq_cst(i6
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB209_3
> +; RV32I-NEXT: bne a5, s1, .LBB209_3
> ; RV32I-NEXT: j .LBB209_4
> ; RV32I-NEXT: .LBB209_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 5
> ; RV32I-NEXT: addi a5, zero, 5
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB209_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB209_4
> +; RV32I-NEXT: beq a5, s1, .LBB209_4
> ; RV32I-NEXT: .LBB209_3: # %atomicrmw.start
> -; RV32I-NEXT: slt a0, s0, a1
> +; RV32I-NEXT: slt a0, s1, a5
> ; RV32I-NEXT: j .LBB209_5
> ; RV32I-NEXT: .LBB209_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB209_5: # %atomicrmw.start
> ; RV32I-NEXT: xori a0, a0, 1
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB209_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB209_1
> ; RV32I-NEXT: .LBB209_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -16203,42 +16270,44 @@ define i64 @atomicrmw_min_i64_seq_cst(i6
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB209_3
> +; RV32IA-NEXT: bne a5, s1, .LBB209_3
> ; RV32IA-NEXT: j .LBB209_4
> ; RV32IA-NEXT: .LBB209_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 5
> ; RV32IA-NEXT: addi a5, zero, 5
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB209_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB209_4
> +; RV32IA-NEXT: beq a5, s1, .LBB209_4
> ; RV32IA-NEXT: .LBB209_3: # %atomicrmw.start
> -; RV32IA-NEXT: slt a0, s0, a1
> +; RV32IA-NEXT: slt a0, s1, a5
> ; RV32IA-NEXT: j .LBB209_5
> ; RV32IA-NEXT: .LBB209_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB209_5: # %atomicrmw.start
> ; RV32IA-NEXT: xori a0, a0, 1
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB209_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB209_1
> ; RV32IA-NEXT: .LBB209_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -16254,30 +16323,31 @@ define i64 @atomicrmw_min_i64_seq_cst(i6
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: blt s0, a2, .LBB209_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: blt s1, a3, .LBB209_3
> ; RV64I-NEXT: .LBB209_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB209_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB209_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bge s0, a2, .LBB209_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bge s1, a3, .LBB209_1
> ; RV64I-NEXT: .LBB209_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB209_1
> ; RV64I-NEXT: .LBB209_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -16302,41 +16372,43 @@ define i64 @atomicrmw_umax_i64_monotonic
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB210_3
> +; RV32I-NEXT: bne a5, s1, .LBB210_3
> ; RV32I-NEXT: j .LBB210_4
> ; RV32I-NEXT: .LBB210_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> +; RV32I-NEXT: mv a0, s0
> ; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: mv a5, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB210_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB210_4
> +; RV32I-NEXT: beq a5, s1, .LBB210_4
> ; RV32I-NEXT: .LBB210_3: # %atomicrmw.start
> -; RV32I-NEXT: sltu a0, s0, a1
> +; RV32I-NEXT: sltu a0, s1, a5
> ; RV32I-NEXT: j .LBB210_5
> ; RV32I-NEXT: .LBB210_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB210_5: # %atomicrmw.start
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB210_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB210_1
> ; RV32I-NEXT: .LBB210_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -16353,41 +16425,43 @@ define i64 @atomicrmw_umax_i64_monotonic
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB210_3
> +; RV32IA-NEXT: bne a5, s1, .LBB210_3
> ; RV32IA-NEXT: j .LBB210_4
> ; RV32IA-NEXT: .LBB210_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> +; RV32IA-NEXT: mv a0, s0
> ; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: mv a4, zero
> ; RV32IA-NEXT: mv a5, zero
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB210_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB210_4
> +; RV32IA-NEXT: beq a5, s1, .LBB210_4
> ; RV32IA-NEXT: .LBB210_3: # %atomicrmw.start
> -; RV32IA-NEXT: sltu a0, s0, a1
> +; RV32IA-NEXT: sltu a0, s1, a5
> ; RV32IA-NEXT: j .LBB210_5
> ; RV32IA-NEXT: .LBB210_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB210_5: # %atomicrmw.start
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB210_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB210_1
> ; RV32IA-NEXT: .LBB210_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -16403,30 +16477,31 @@ define i64 @atomicrmw_umax_i64_monotonic
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bgeu s0, a2, .LBB210_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a3, .LBB210_3
> ; RV64I-NEXT: .LBB210_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: sd a3, 8(sp)
> +; RV64I-NEXT: mv a0, s0
> ; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB210_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB210_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bltu s0, a2, .LBB210_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a3, .LBB210_1
> ; RV64I-NEXT: .LBB210_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB210_1
> ; RV64I-NEXT: .LBB210_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -16451,41 +16526,43 @@ define i64 @atomicrmw_umax_i64_acquire(i
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB211_3
> +; RV32I-NEXT: bne a5, s1, .LBB211_3
> ; RV32I-NEXT: j .LBB211_4
> ; RV32I-NEXT: .LBB211_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 2
> ; RV32I-NEXT: addi a5, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB211_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB211_4
> +; RV32I-NEXT: beq a5, s1, .LBB211_4
> ; RV32I-NEXT: .LBB211_3: # %atomicrmw.start
> -; RV32I-NEXT: sltu a0, s0, a1
> +; RV32I-NEXT: sltu a0, s1, a5
> ; RV32I-NEXT: j .LBB211_5
> ; RV32I-NEXT: .LBB211_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB211_5: # %atomicrmw.start
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB211_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB211_1
> ; RV32I-NEXT: .LBB211_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -16502,41 +16579,43 @@ define i64 @atomicrmw_umax_i64_acquire(i
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB211_3
> +; RV32IA-NEXT: bne a5, s1, .LBB211_3
> ; RV32IA-NEXT: j .LBB211_4
> ; RV32IA-NEXT: .LBB211_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 2
> ; RV32IA-NEXT: addi a5, zero, 2
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB211_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB211_4
> +; RV32IA-NEXT: beq a5, s1, .LBB211_4
> ; RV32IA-NEXT: .LBB211_3: # %atomicrmw.start
> -; RV32IA-NEXT: sltu a0, s0, a1
> +; RV32IA-NEXT: sltu a0, s1, a5
> ; RV32IA-NEXT: j .LBB211_5
> ; RV32IA-NEXT: .LBB211_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB211_5: # %atomicrmw.start
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB211_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB211_1
> ; RV32IA-NEXT: .LBB211_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -16552,30 +16631,31 @@ define i64 @atomicrmw_umax_i64_acquire(i
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bgeu s0, a2, .LBB211_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a3, .LBB211_3
> ; RV64I-NEXT: .LBB211_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB211_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB211_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bltu s0, a2, .LBB211_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a3, .LBB211_1
> ; RV64I-NEXT: .LBB211_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB211_1
> ; RV64I-NEXT: .LBB211_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -16600,41 +16680,43 @@ define i64 @atomicrmw_umax_i64_release(i
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB212_3
> +; RV32I-NEXT: bne a5, s1, .LBB212_3
> ; RV32I-NEXT: j .LBB212_4
> ; RV32I-NEXT: .LBB212_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 3
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a5, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB212_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB212_4
> +; RV32I-NEXT: beq a5, s1, .LBB212_4
> ; RV32I-NEXT: .LBB212_3: # %atomicrmw.start
> -; RV32I-NEXT: sltu a0, s0, a1
> +; RV32I-NEXT: sltu a0, s1, a5
> ; RV32I-NEXT: j .LBB212_5
> ; RV32I-NEXT: .LBB212_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB212_5: # %atomicrmw.start
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB212_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB212_1
> ; RV32I-NEXT: .LBB212_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -16651,41 +16733,43 @@ define i64 @atomicrmw_umax_i64_release(i
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB212_3
> +; RV32IA-NEXT: bne a5, s1, .LBB212_3
> ; RV32IA-NEXT: j .LBB212_4
> ; RV32IA-NEXT: .LBB212_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 3
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: mv a5, zero
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB212_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB212_4
> +; RV32IA-NEXT: beq a5, s1, .LBB212_4
> ; RV32IA-NEXT: .LBB212_3: # %atomicrmw.start
> -; RV32IA-NEXT: sltu a0, s0, a1
> +; RV32IA-NEXT: sltu a0, s1, a5
> ; RV32IA-NEXT: j .LBB212_5
> ; RV32IA-NEXT: .LBB212_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB212_5: # %atomicrmw.start
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB212_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB212_1
> ; RV32IA-NEXT: .LBB212_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -16701,30 +16785,31 @@ define i64 @atomicrmw_umax_i64_release(i
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bgeu s0, a2, .LBB212_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a3, .LBB212_3
> ; RV64I-NEXT: .LBB212_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 3
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB212_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB212_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bltu s0, a2, .LBB212_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a3, .LBB212_1
> ; RV64I-NEXT: .LBB212_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB212_1
> ; RV64I-NEXT: .LBB212_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -16749,41 +16834,43 @@ define i64 @atomicrmw_umax_i64_acq_rel(i
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB213_3
> +; RV32I-NEXT: bne a5, s1, .LBB213_3
> ; RV32I-NEXT: j .LBB213_4
> ; RV32I-NEXT: .LBB213_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 4
> ; RV32I-NEXT: addi a5, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB213_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB213_4
> +; RV32I-NEXT: beq a5, s1, .LBB213_4
> ; RV32I-NEXT: .LBB213_3: # %atomicrmw.start
> -; RV32I-NEXT: sltu a0, s0, a1
> +; RV32I-NEXT: sltu a0, s1, a5
> ; RV32I-NEXT: j .LBB213_5
> ; RV32I-NEXT: .LBB213_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB213_5: # %atomicrmw.start
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB213_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB213_1
> ; RV32I-NEXT: .LBB213_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -16800,41 +16887,43 @@ define i64 @atomicrmw_umax_i64_acq_rel(i
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB213_3
> +; RV32IA-NEXT: bne a5, s1, .LBB213_3
> ; RV32IA-NEXT: j .LBB213_4
> ; RV32IA-NEXT: .LBB213_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 4
> ; RV32IA-NEXT: addi a5, zero, 2
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB213_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB213_4
> +; RV32IA-NEXT: beq a5, s1, .LBB213_4
> ; RV32IA-NEXT: .LBB213_3: # %atomicrmw.start
> -; RV32IA-NEXT: sltu a0, s0, a1
> +; RV32IA-NEXT: sltu a0, s1, a5
> ; RV32IA-NEXT: j .LBB213_5
> ; RV32IA-NEXT: .LBB213_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB213_5: # %atomicrmw.start
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB213_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB213_1
> ; RV32IA-NEXT: .LBB213_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -16850,30 +16939,31 @@ define i64 @atomicrmw_umax_i64_acq_rel(i
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bgeu s0, a2, .LBB213_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a3, .LBB213_3
> ; RV64I-NEXT: .LBB213_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB213_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB213_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bltu s0, a2, .LBB213_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a3, .LBB213_1
> ; RV64I-NEXT: .LBB213_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB213_1
> ; RV64I-NEXT: .LBB213_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -16898,41 +16988,43 @@ define i64 @atomicrmw_umax_i64_seq_cst(i
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB214_3
> +; RV32I-NEXT: bne a5, s1, .LBB214_3
> ; RV32I-NEXT: j .LBB214_4
> ; RV32I-NEXT: .LBB214_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 5
> ; RV32I-NEXT: addi a5, zero, 5
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB214_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB214_4
> +; RV32I-NEXT: beq a5, s1, .LBB214_4
> ; RV32I-NEXT: .LBB214_3: # %atomicrmw.start
> -; RV32I-NEXT: sltu a0, s0, a1
> +; RV32I-NEXT: sltu a0, s1, a5
> ; RV32I-NEXT: j .LBB214_5
> ; RV32I-NEXT: .LBB214_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB214_5: # %atomicrmw.start
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB214_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB214_1
> ; RV32I-NEXT: .LBB214_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -16949,41 +17041,43 @@ define i64 @atomicrmw_umax_i64_seq_cst(i
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB214_3
> +; RV32IA-NEXT: bne a5, s1, .LBB214_3
> ; RV32IA-NEXT: j .LBB214_4
> ; RV32IA-NEXT: .LBB214_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 5
> ; RV32IA-NEXT: addi a5, zero, 5
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB214_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB214_4
> +; RV32IA-NEXT: beq a5, s1, .LBB214_4
> ; RV32IA-NEXT: .LBB214_3: # %atomicrmw.start
> -; RV32IA-NEXT: sltu a0, s0, a1
> +; RV32IA-NEXT: sltu a0, s1, a5
> ; RV32IA-NEXT: j .LBB214_5
> ; RV32IA-NEXT: .LBB214_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB214_5: # %atomicrmw.start
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB214_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB214_1
> ; RV32IA-NEXT: .LBB214_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -16999,30 +17093,31 @@ define i64 @atomicrmw_umax_i64_seq_cst(i
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bgeu s0, a2, .LBB214_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a3, .LBB214_3
> ; RV64I-NEXT: .LBB214_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB214_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB214_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bltu s0, a2, .LBB214_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a3, .LBB214_1
> ; RV64I-NEXT: .LBB214_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB214_1
> ; RV64I-NEXT: .LBB214_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -17047,42 +17142,44 @@ define i64 @atomicrmw_umin_i64_monotonic
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB215_3
> +; RV32I-NEXT: bne a5, s1, .LBB215_3
> ; RV32I-NEXT: j .LBB215_4
> ; RV32I-NEXT: .LBB215_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> +; RV32I-NEXT: mv a0, s0
> ; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a4, zero
> ; RV32I-NEXT: mv a5, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB215_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB215_4
> +; RV32I-NEXT: beq a5, s1, .LBB215_4
> ; RV32I-NEXT: .LBB215_3: # %atomicrmw.start
> -; RV32I-NEXT: sltu a0, s0, a1
> +; RV32I-NEXT: sltu a0, s1, a5
> ; RV32I-NEXT: j .LBB215_5
> ; RV32I-NEXT: .LBB215_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB215_5: # %atomicrmw.start
> ; RV32I-NEXT: xori a0, a0, 1
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB215_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB215_1
> ; RV32I-NEXT: .LBB215_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -17099,42 +17196,44 @@ define i64 @atomicrmw_umin_i64_monotonic
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB215_3
> +; RV32IA-NEXT: bne a5, s1, .LBB215_3
> ; RV32IA-NEXT: j .LBB215_4
> ; RV32IA-NEXT: .LBB215_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> +; RV32IA-NEXT: mv a0, s0
> ; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: mv a4, zero
> ; RV32IA-NEXT: mv a5, zero
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB215_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB215_4
> +; RV32IA-NEXT: beq a5, s1, .LBB215_4
> ; RV32IA-NEXT: .LBB215_3: # %atomicrmw.start
> -; RV32IA-NEXT: sltu a0, s0, a1
> +; RV32IA-NEXT: sltu a0, s1, a5
> ; RV32IA-NEXT: j .LBB215_5
> ; RV32IA-NEXT: .LBB215_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB215_5: # %atomicrmw.start
> ; RV32IA-NEXT: xori a0, a0, 1
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB215_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB215_1
> ; RV32IA-NEXT: .LBB215_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -17150,30 +17249,31 @@ define i64 @atomicrmw_umin_i64_monotonic
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bltu s0, a2, .LBB215_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a3, .LBB215_3
> ; RV64I-NEXT: .LBB215_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> +; RV64I-NEXT: sd a3, 8(sp)
> +; RV64I-NEXT: mv a0, s0
> ; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a3, zero
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB215_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB215_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bgeu s0, a2, .LBB215_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a3, .LBB215_1
> ; RV64I-NEXT: .LBB215_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB215_1
> ; RV64I-NEXT: .LBB215_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -17198,42 +17298,44 @@ define i64 @atomicrmw_umin_i64_acquire(i
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB216_3
> +; RV32I-NEXT: bne a5, s1, .LBB216_3
> ; RV32I-NEXT: j .LBB216_4
> ; RV32I-NEXT: .LBB216_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 2
> ; RV32I-NEXT: addi a5, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB216_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB216_4
> +; RV32I-NEXT: beq a5, s1, .LBB216_4
> ; RV32I-NEXT: .LBB216_3: # %atomicrmw.start
> -; RV32I-NEXT: sltu a0, s0, a1
> +; RV32I-NEXT: sltu a0, s1, a5
> ; RV32I-NEXT: j .LBB216_5
> ; RV32I-NEXT: .LBB216_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB216_5: # %atomicrmw.start
> ; RV32I-NEXT: xori a0, a0, 1
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB216_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB216_1
> ; RV32I-NEXT: .LBB216_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -17250,42 +17352,44 @@ define i64 @atomicrmw_umin_i64_acquire(i
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB216_3
> +; RV32IA-NEXT: bne a5, s1, .LBB216_3
> ; RV32IA-NEXT: j .LBB216_4
> ; RV32IA-NEXT: .LBB216_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 2
> ; RV32IA-NEXT: addi a5, zero, 2
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB216_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB216_4
> +; RV32IA-NEXT: beq a5, s1, .LBB216_4
> ; RV32IA-NEXT: .LBB216_3: # %atomicrmw.start
> -; RV32IA-NEXT: sltu a0, s0, a1
> +; RV32IA-NEXT: sltu a0, s1, a5
> ; RV32IA-NEXT: j .LBB216_5
> ; RV32IA-NEXT: .LBB216_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB216_5: # %atomicrmw.start
> ; RV32IA-NEXT: xori a0, a0, 1
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB216_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB216_1
> ; RV32IA-NEXT: .LBB216_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -17301,30 +17405,31 @@ define i64 @atomicrmw_umin_i64_acquire(i
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bltu s0, a2, .LBB216_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a3, .LBB216_3
> ; RV64I-NEXT: .LBB216_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 2
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB216_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB216_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bgeu s0, a2, .LBB216_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a3, .LBB216_1
> ; RV64I-NEXT: .LBB216_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB216_1
> ; RV64I-NEXT: .LBB216_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -17349,42 +17454,44 @@ define i64 @atomicrmw_umin_i64_release(i
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB217_3
> +; RV32I-NEXT: bne a5, s1, .LBB217_3
> ; RV32I-NEXT: j .LBB217_4
> ; RV32I-NEXT: .LBB217_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 3
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: mv a5, zero
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB217_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB217_4
> +; RV32I-NEXT: beq a5, s1, .LBB217_4
> ; RV32I-NEXT: .LBB217_3: # %atomicrmw.start
> -; RV32I-NEXT: sltu a0, s0, a1
> +; RV32I-NEXT: sltu a0, s1, a5
> ; RV32I-NEXT: j .LBB217_5
> ; RV32I-NEXT: .LBB217_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB217_5: # %atomicrmw.start
> ; RV32I-NEXT: xori a0, a0, 1
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB217_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB217_1
> ; RV32I-NEXT: .LBB217_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -17401,42 +17508,44 @@ define i64 @atomicrmw_umin_i64_release(i
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB217_3
> +; RV32IA-NEXT: bne a5, s1, .LBB217_3
> ; RV32IA-NEXT: j .LBB217_4
> ; RV32IA-NEXT: .LBB217_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 3
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: mv a5, zero
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB217_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB217_4
> +; RV32IA-NEXT: beq a5, s1, .LBB217_4
> ; RV32IA-NEXT: .LBB217_3: # %atomicrmw.start
> -; RV32IA-NEXT: sltu a0, s0, a1
> +; RV32IA-NEXT: sltu a0, s1, a5
> ; RV32IA-NEXT: j .LBB217_5
> ; RV32IA-NEXT: .LBB217_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB217_5: # %atomicrmw.start
> ; RV32IA-NEXT: xori a0, a0, 1
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB217_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB217_1
> ; RV32IA-NEXT: .LBB217_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -17452,30 +17561,31 @@ define i64 @atomicrmw_umin_i64_release(i
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bltu s0, a2, .LBB217_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a3, .LBB217_3
> ; RV64I-NEXT: .LBB217_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 3
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB217_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB217_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bgeu s0, a2, .LBB217_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a3, .LBB217_1
> ; RV64I-NEXT: .LBB217_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB217_1
> ; RV64I-NEXT: .LBB217_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -17500,42 +17610,44 @@ define i64 @atomicrmw_umin_i64_acq_rel(i
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB218_3
> +; RV32I-NEXT: bne a5, s1, .LBB218_3
> ; RV32I-NEXT: j .LBB218_4
> ; RV32I-NEXT: .LBB218_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 4
> ; RV32I-NEXT: addi a5, zero, 2
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB218_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB218_4
> +; RV32I-NEXT: beq a5, s1, .LBB218_4
> ; RV32I-NEXT: .LBB218_3: # %atomicrmw.start
> -; RV32I-NEXT: sltu a0, s0, a1
> +; RV32I-NEXT: sltu a0, s1, a5
> ; RV32I-NEXT: j .LBB218_5
> ; RV32I-NEXT: .LBB218_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB218_5: # %atomicrmw.start
> ; RV32I-NEXT: xori a0, a0, 1
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB218_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB218_1
> ; RV32I-NEXT: .LBB218_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -17552,42 +17664,44 @@ define i64 @atomicrmw_umin_i64_acq_rel(i
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB218_3
> +; RV32IA-NEXT: bne a5, s1, .LBB218_3
> ; RV32IA-NEXT: j .LBB218_4
> ; RV32IA-NEXT: .LBB218_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 4
> ; RV32IA-NEXT: addi a5, zero, 2
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB218_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB218_4
> +; RV32IA-NEXT: beq a5, s1, .LBB218_4
> ; RV32IA-NEXT: .LBB218_3: # %atomicrmw.start
> -; RV32IA-NEXT: sltu a0, s0, a1
> +; RV32IA-NEXT: sltu a0, s1, a5
> ; RV32IA-NEXT: j .LBB218_5
> ; RV32IA-NEXT: .LBB218_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB218_5: # %atomicrmw.start
> ; RV32IA-NEXT: xori a0, a0, 1
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB218_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB218_1
> ; RV32IA-NEXT: .LBB218_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -17603,30 +17717,31 @@ define i64 @atomicrmw_umin_i64_acq_rel(i
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bltu s0, a2, .LBB218_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a3, .LBB218_3
> ; RV64I-NEXT: .LBB218_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 2
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB218_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB218_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bgeu s0, a2, .LBB218_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a3, .LBB218_1
> ; RV64I-NEXT: .LBB218_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB218_1
> ; RV64I-NEXT: .LBB218_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
> @@ -17651,42 +17766,44 @@ define i64 @atomicrmw_umin_i64_seq_cst(i
> ; RV32I-NEXT: sw s1, 20(sp)
> ; RV32I-NEXT: sw s2, 16(sp)
> ; RV32I-NEXT: sw s3, 12(sp)
> -; RV32I-NEXT: mv s0, a2
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: lw a5, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: mv s1, a2
> ; RV32I-NEXT: mv s2, a1
> -; RV32I-NEXT: mv s1, a0
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: lw a2, 0(a0)
> ; RV32I-NEXT: mv s3, sp
> -; RV32I-NEXT: bne a1, s0, .LBB219_3
> +; RV32I-NEXT: bne a5, s1, .LBB219_3
> ; RV32I-NEXT: j .LBB219_4
> ; RV32I-NEXT: .LBB219_1: # %atomicrmw.start
> -; RV32I-NEXT: sw a1, 4(sp)
> -; RV32I-NEXT: mv a0, s1
> -; RV32I-NEXT: mv a1, s3
> +; RV32I-NEXT: sw a4, 0(sp)
> +; RV32I-NEXT: sw a5, 4(sp)
> ; RV32I-NEXT: addi a4, zero, 5
> ; RV32I-NEXT: addi a5, zero, 5
> +; RV32I-NEXT: mv a0, s0
> +; RV32I-NEXT: mv a1, s3
> ; RV32I-NEXT: call __atomic_compare_exchange_8
> -; RV32I-NEXT: lw a1, 4(sp)
> -; RV32I-NEXT: lw a2, 0(sp)
> +; RV32I-NEXT: lw a5, 4(sp)
> +; RV32I-NEXT: lw a4, 0(sp)
> ; RV32I-NEXT: bnez a0, .LBB219_7
> ; RV32I-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32I-NEXT: beq a1, s0, .LBB219_4
> +; RV32I-NEXT: beq a5, s1, .LBB219_4
> ; RV32I-NEXT: .LBB219_3: # %atomicrmw.start
> -; RV32I-NEXT: sltu a0, s0, a1
> +; RV32I-NEXT: sltu a0, s1, a5
> ; RV32I-NEXT: j .LBB219_5
> ; RV32I-NEXT: .LBB219_4:
> -; RV32I-NEXT: sltu a0, s2, a2
> +; RV32I-NEXT: sltu a0, s2, a4
> ; RV32I-NEXT: .LBB219_5: # %atomicrmw.start
> ; RV32I-NEXT: xori a0, a0, 1
> -; RV32I-NEXT: sw a2, 0(sp)
> -; RV32I-NEXT: mv a3, a1
> +; RV32I-NEXT: mv a2, a4
> +; RV32I-NEXT: mv a3, a5
> ; RV32I-NEXT: bnez a0, .LBB219_1
> ; RV32I-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32I-NEXT: mv a2, s2
> -; RV32I-NEXT: mv a3, s0
> +; RV32I-NEXT: mv a3, s1
> ; RV32I-NEXT: j .LBB219_1
> ; RV32I-NEXT: .LBB219_7: # %atomicrmw.end
> -; RV32I-NEXT: mv a0, a2
> +; RV32I-NEXT: mv a0, a4
> +; RV32I-NEXT: mv a1, a5
> ; RV32I-NEXT: lw s3, 12(sp)
> ; RV32I-NEXT: lw s2, 16(sp)
> ; RV32I-NEXT: lw s1, 20(sp)
> @@ -17703,42 +17820,44 @@ define i64 @atomicrmw_umin_i64_seq_cst(i
> ; RV32IA-NEXT: sw s1, 20(sp)
> ; RV32IA-NEXT: sw s2, 16(sp)
> ; RV32IA-NEXT: sw s3, 12(sp)
> -; RV32IA-NEXT: mv s0, a2
> +; RV32IA-NEXT: mv s0, a0
> +; RV32IA-NEXT: lw a5, 4(a0)
> +; RV32IA-NEXT: lw a4, 0(a0)
> +; RV32IA-NEXT: mv s1, a2
> ; RV32IA-NEXT: mv s2, a1
> -; RV32IA-NEXT: mv s1, a0
> -; RV32IA-NEXT: lw a1, 4(a0)
> -; RV32IA-NEXT: lw a2, 0(a0)
> ; RV32IA-NEXT: mv s3, sp
> -; RV32IA-NEXT: bne a1, s0, .LBB219_3
> +; RV32IA-NEXT: bne a5, s1, .LBB219_3
> ; RV32IA-NEXT: j .LBB219_4
> ; RV32IA-NEXT: .LBB219_1: # %atomicrmw.start
> -; RV32IA-NEXT: sw a1, 4(sp)
> -; RV32IA-NEXT: mv a0, s1
> -; RV32IA-NEXT: mv a1, s3
> +; RV32IA-NEXT: sw a4, 0(sp)
> +; RV32IA-NEXT: sw a5, 4(sp)
> ; RV32IA-NEXT: addi a4, zero, 5
> ; RV32IA-NEXT: addi a5, zero, 5
> +; RV32IA-NEXT: mv a0, s0
> +; RV32IA-NEXT: mv a1, s3
> ; RV32IA-NEXT: call __atomic_compare_exchange_8
> -; RV32IA-NEXT: lw a1, 4(sp)
> -; RV32IA-NEXT: lw a2, 0(sp)
> +; RV32IA-NEXT: lw a5, 4(sp)
> +; RV32IA-NEXT: lw a4, 0(sp)
> ; RV32IA-NEXT: bnez a0, .LBB219_7
> ; RV32IA-NEXT: # %bb.2: # %atomicrmw.start
> -; RV32IA-NEXT: beq a1, s0, .LBB219_4
> +; RV32IA-NEXT: beq a5, s1, .LBB219_4
> ; RV32IA-NEXT: .LBB219_3: # %atomicrmw.start
> -; RV32IA-NEXT: sltu a0, s0, a1
> +; RV32IA-NEXT: sltu a0, s1, a5
> ; RV32IA-NEXT: j .LBB219_5
> ; RV32IA-NEXT: .LBB219_4:
> -; RV32IA-NEXT: sltu a0, s2, a2
> +; RV32IA-NEXT: sltu a0, s2, a4
> ; RV32IA-NEXT: .LBB219_5: # %atomicrmw.start
> ; RV32IA-NEXT: xori a0, a0, 1
> -; RV32IA-NEXT: sw a2, 0(sp)
> -; RV32IA-NEXT: mv a3, a1
> +; RV32IA-NEXT: mv a2, a4
> +; RV32IA-NEXT: mv a3, a5
> ; RV32IA-NEXT: bnez a0, .LBB219_1
> ; RV32IA-NEXT: # %bb.6: # %atomicrmw.start
> ; RV32IA-NEXT: mv a2, s2
> -; RV32IA-NEXT: mv a3, s0
> +; RV32IA-NEXT: mv a3, s1
> ; RV32IA-NEXT: j .LBB219_1
> ; RV32IA-NEXT: .LBB219_7: # %atomicrmw.end
> -; RV32IA-NEXT: mv a0, a2
> +; RV32IA-NEXT: mv a0, a4
> +; RV32IA-NEXT: mv a1, a5
> ; RV32IA-NEXT: lw s3, 12(sp)
> ; RV32IA-NEXT: lw s2, 16(sp)
> ; RV32IA-NEXT: lw s1, 20(sp)
> @@ -17754,30 +17873,31 @@ define i64 @atomicrmw_umin_i64_seq_cst(i
> ; RV64I-NEXT: sd s0, 32(sp)
> ; RV64I-NEXT: sd s1, 24(sp)
> ; RV64I-NEXT: sd s2, 16(sp)
> -; RV64I-NEXT: mv s0, a1
> -; RV64I-NEXT: mv s1, a0
> -; RV64I-NEXT: ld a2, 0(a0)
> +; RV64I-NEXT: mv s0, a0
> +; RV64I-NEXT: ld a3, 0(a0)
> +; RV64I-NEXT: mv s1, a1
> ; RV64I-NEXT: addi s2, sp, 8
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bltu s0, a2, .LBB219_3
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bltu s1, a3, .LBB219_3
> ; RV64I-NEXT: .LBB219_1: # %atomicrmw.start
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: mv a0, s1
> -; RV64I-NEXT: mv a1, s2
> +; RV64I-NEXT: sd a3, 8(sp)
> ; RV64I-NEXT: addi a3, zero, 5
> ; RV64I-NEXT: addi a4, zero, 5
> +; RV64I-NEXT: mv a0, s0
> +; RV64I-NEXT: mv a1, s2
> ; RV64I-NEXT: call __atomic_compare_exchange_8
> -; RV64I-NEXT: ld a2, 8(sp)
> +; RV64I-NEXT: ld a3, 8(sp)
> ; RV64I-NEXT: bnez a0, .LBB219_4
> ; RV64I-NEXT: # %bb.2: # %atomicrmw.start
> ; RV64I-NEXT: # in Loop: Header=BB219_1 Depth=1
> -; RV64I-NEXT: sd a2, 8(sp)
> -; RV64I-NEXT: bgeu s0, a2, .LBB219_1
> +; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bgeu s1, a3, .LBB219_1
> ; RV64I-NEXT: .LBB219_3: # %atomicrmw.start
> -; RV64I-NEXT: mv a2, s0
> +; RV64I-NEXT: mv a2, s1
> ; RV64I-NEXT: j .LBB219_1
> ; RV64I-NEXT: .LBB219_4: # %atomicrmw.end
> -; RV64I-NEXT: mv a0, a2
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: ld s2, 16(sp)
> ; RV64I-NEXT: ld s1, 24(sp)
> ; RV64I-NEXT: ld s0, 32(sp)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/bare-select.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/bare-select.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/bare-select.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/bare-select.ll Tue Sep 17 02:43:08 2019
> @@ -5,12 +5,12 @@
> define i32 @bare_select(i1 %a, i32 %b, i32 %c) nounwind {
> ; RV32I-LABEL: bare_select:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: andi a0, a0, 1
> -; RV32I-NEXT: bnez a0, .LBB0_2
> +; RV32I-NEXT: andi a3, a0, 1
> +; RV32I-NEXT: mv a0, a1
> +; RV32I-NEXT: bnez a3, .LBB0_2
> ; RV32I-NEXT: # %bb.1:
> -; RV32I-NEXT: mv a1, a2
> +; RV32I-NEXT: mv a0, a2
> ; RV32I-NEXT: .LBB0_2:
> -; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: ret
> %1 = select i1 %a, i32 %b, i32 %c
> ret i32 %1
> @@ -19,12 +19,12 @@ define i32 @bare_select(i1 %a, i32 %b, i
> define float @bare_select_float(i1 %a, float %b, float %c) nounwind {
> ; RV32I-LABEL: bare_select_float:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: andi a0, a0, 1
> -; RV32I-NEXT: bnez a0, .LBB1_2
> +; RV32I-NEXT: andi a3, a0, 1
> +; RV32I-NEXT: mv a0, a1
> +; RV32I-NEXT: bnez a3, .LBB1_2
> ; RV32I-NEXT: # %bb.1:
> -; RV32I-NEXT: mv a1, a2
> +; RV32I-NEXT: mv a0, a2
> ; RV32I-NEXT: .LBB1_2:
> -; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: ret
> %1 = select i1 %a, float %b, float %c
> ret float %1
>
> Modified: llvm/trunk/test/CodeGen/RISCV/blockaddress.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/blockaddress.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/blockaddress.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/blockaddress.ll Tue Sep 17 02:43:08 2019
> @@ -9,11 +9,11 @@ define void @test_blockaddress() nounwin
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi sp, sp, -16
> ; RV32I-NEXT: sw ra, 12(sp)
> -; RV32I-NEXT: lui a0, %hi(.Ltmp0)
> -; RV32I-NEXT: addi a0, a0, %lo(.Ltmp0)
> -; RV32I-NEXT: lui a1, %hi(addr)
> -; RV32I-NEXT: sw a0, %lo(addr)(a1)
> -; RV32I-NEXT: lw a0, %lo(addr)(a1)
> +; RV32I-NEXT: lui a0, %hi(addr)
> +; RV32I-NEXT: lui a1, %hi(.Ltmp0)
> +; RV32I-NEXT: addi a1, a1, %lo(.Ltmp0)
> +; RV32I-NEXT: sw a1, %lo(addr)(a0)
> +; RV32I-NEXT: lw a0, %lo(addr)(a0)
> ; RV32I-NEXT: jr a0
> ; RV32I-NEXT: .Ltmp0: # Block address taken
> ; RV32I-NEXT: .LBB0_1: # %block
>
> Modified: llvm/trunk/test/CodeGen/RISCV/bswap-ctlz-cttz-ctpop.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/bswap-ctlz-cttz-ctpop.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/bswap-ctlz-cttz-ctpop.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/bswap-ctlz-cttz-ctpop.ll Tue Sep 17 02:43:08 2019
> @@ -29,10 +29,10 @@ define i16 @test_bswap_i16(i16 %a) nounw
> define i32 @test_bswap_i32(i32 %a) nounwind {
> ; RV32I-LABEL: test_bswap_i32:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: lui a1, 16
> -; RV32I-NEXT: addi a1, a1, -256
> -; RV32I-NEXT: srli a2, a0, 8
> -; RV32I-NEXT: and a1, a2, a1
> +; RV32I-NEXT: srli a1, a0, 8
> +; RV32I-NEXT: lui a2, 16
> +; RV32I-NEXT: addi a2, a2, -256
> +; RV32I-NEXT: and a1, a1, a2
> ; RV32I-NEXT: srli a2, a0, 24
> ; RV32I-NEXT: or a1, a1, a2
> ; RV32I-NEXT: slli a2, a0, 8
> @@ -49,9 +49,9 @@ define i32 @test_bswap_i32(i32 %a) nounw
> define i64 @test_bswap_i64(i64 %a) nounwind {
> ; RV32I-LABEL: test_bswap_i64:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: lui a2, 16
> -; RV32I-NEXT: addi a3, a2, -256
> ; RV32I-NEXT: srli a2, a1, 8
> +; RV32I-NEXT: lui a3, 16
> +; RV32I-NEXT: addi a3, a3, -256
> ; RV32I-NEXT: and a2, a2, a3
> ; RV32I-NEXT: srli a4, a1, 24
> ; RV32I-NEXT: or a2, a2, a4
> @@ -87,10 +87,10 @@ define i8 @test_cttz_i8(i8 %a) nounwind
> ; RV32I-NEXT: addi a1, a0, -1
> ; RV32I-NEXT: not a0, a0
> ; RV32I-NEXT: and a0, a0, a1
> -; RV32I-NEXT: lui a1, 349525
> -; RV32I-NEXT: addi a1, a1, 1365
> -; RV32I-NEXT: srli a2, a0, 1
> -; RV32I-NEXT: and a1, a2, a1
> +; RV32I-NEXT: srli a1, a0, 1
> +; RV32I-NEXT: lui a2, 349525
> +; RV32I-NEXT: addi a2, a2, 1365
> +; RV32I-NEXT: and a1, a1, a2
> ; RV32I-NEXT: sub a0, a0, a1
> ; RV32I-NEXT: lui a1, 209715
> ; RV32I-NEXT: addi a1, a1, 819
> @@ -131,10 +131,10 @@ define i16 @test_cttz_i16(i16 %a) nounwi
> ; RV32I-NEXT: addi a1, a0, -1
> ; RV32I-NEXT: not a0, a0
> ; RV32I-NEXT: and a0, a0, a1
> -; RV32I-NEXT: lui a1, 349525
> -; RV32I-NEXT: addi a1, a1, 1365
> -; RV32I-NEXT: srli a2, a0, 1
> -; RV32I-NEXT: and a1, a2, a1
> +; RV32I-NEXT: srli a1, a0, 1
> +; RV32I-NEXT: lui a2, 349525
> +; RV32I-NEXT: addi a2, a2, 1365
> +; RV32I-NEXT: and a1, a1, a2
> ; RV32I-NEXT: sub a0, a0, a1
> ; RV32I-NEXT: lui a1, 209715
> ; RV32I-NEXT: addi a1, a1, 819
> @@ -172,10 +172,10 @@ define i32 @test_cttz_i32(i32 %a) nounwi
> ; RV32I-NEXT: addi a1, a0, -1
> ; RV32I-NEXT: not a0, a0
> ; RV32I-NEXT: and a0, a0, a1
> -; RV32I-NEXT: lui a1, 349525
> -; RV32I-NEXT: addi a1, a1, 1365
> -; RV32I-NEXT: srli a2, a0, 1
> -; RV32I-NEXT: and a1, a2, a1
> +; RV32I-NEXT: srli a1, a0, 1
> +; RV32I-NEXT: lui a2, 349525
> +; RV32I-NEXT: addi a2, a2, 1365
> +; RV32I-NEXT: and a1, a1, a2
> ; RV32I-NEXT: sub a0, a0, a1
> ; RV32I-NEXT: lui a1, 209715
> ; RV32I-NEXT: addi a1, a1, 819
> @@ -220,11 +220,11 @@ define i32 @test_ctlz_i32(i32 %a) nounwi
> ; RV32I-NEXT: or a0, a0, a1
> ; RV32I-NEXT: srli a1, a0, 16
> ; RV32I-NEXT: or a0, a0, a1
> -; RV32I-NEXT: lui a1, 349525
> -; RV32I-NEXT: addi a1, a1, 1365
> ; RV32I-NEXT: not a0, a0
> -; RV32I-NEXT: srli a2, a0, 1
> -; RV32I-NEXT: and a1, a2, a1
> +; RV32I-NEXT: srli a1, a0, 1
> +; RV32I-NEXT: lui a2, 349525
> +; RV32I-NEXT: addi a2, a2, 1365
> +; RV32I-NEXT: and a1, a1, a2
> ; RV32I-NEXT: sub a0, a0, a1
> ; RV32I-NEXT: lui a1, 209715
> ; RV32I-NEXT: addi a1, a1, 819
> @@ -269,9 +269,9 @@ define i64 @test_cttz_i64(i64 %a) nounwi
> ; RV32I-NEXT: addi a0, a0, -1
> ; RV32I-NEXT: not a1, s4
> ; RV32I-NEXT: and a0, a1, a0
> -; RV32I-NEXT: lui a1, 349525
> -; RV32I-NEXT: addi s5, a1, 1365
> ; RV32I-NEXT: srli a1, a0, 1
> +; RV32I-NEXT: lui a2, 349525
> +; RV32I-NEXT: addi s5, a2, 1365
> ; RV32I-NEXT: and a1, a1, s5
> ; RV32I-NEXT: sub a0, a0, a1
> ; RV32I-NEXT: lui a1, 209715
> @@ -282,12 +282,12 @@ define i64 @test_cttz_i64(i64 %a) nounwi
> ; RV32I-NEXT: add a0, a1, a0
> ; RV32I-NEXT: srli a1, a0, 4
> ; RV32I-NEXT: add a0, a0, a1
> -; RV32I-NEXT: lui a1, 4112
> -; RV32I-NEXT: addi s6, a1, 257
> ; RV32I-NEXT: lui a1, 61681
> -; RV32I-NEXT: addi s1, a1, -241
> -; RV32I-NEXT: and a0, a0, s1
> -; RV32I-NEXT: mv a1, s6
> +; RV32I-NEXT: addi s6, a1, -241
> +; RV32I-NEXT: and a0, a0, s6
> +; RV32I-NEXT: lui a1, 4112
> +; RV32I-NEXT: addi s1, a1, 257
> +; RV32I-NEXT: mv a1, s1
> ; RV32I-NEXT: call __mulsi3
> ; RV32I-NEXT: mv s2, a0
> ; RV32I-NEXT: addi a0, s3, -1
> @@ -302,8 +302,8 @@ define i64 @test_cttz_i64(i64 %a) nounwi
> ; RV32I-NEXT: add a0, a1, a0
> ; RV32I-NEXT: srli a1, a0, 4
> ; RV32I-NEXT: add a0, a0, a1
> -; RV32I-NEXT: and a0, a0, s1
> -; RV32I-NEXT: mv a1, s6
> +; RV32I-NEXT: and a0, a0, s6
> +; RV32I-NEXT: mv a1, s1
> ; RV32I-NEXT: call __mulsi3
> ; RV32I-NEXT: bnez s4, .LBB7_2
> ; RV32I-NEXT: # %bb.1:
> @@ -336,10 +336,10 @@ define i8 @test_cttz_i8_zero_undef(i8 %a
> ; RV32I-NEXT: addi a1, a0, -1
> ; RV32I-NEXT: not a0, a0
> ; RV32I-NEXT: and a0, a0, a1
> -; RV32I-NEXT: lui a1, 349525
> -; RV32I-NEXT: addi a1, a1, 1365
> -; RV32I-NEXT: srli a2, a0, 1
> -; RV32I-NEXT: and a1, a2, a1
> +; RV32I-NEXT: srli a1, a0, 1
> +; RV32I-NEXT: lui a2, 349525
> +; RV32I-NEXT: addi a2, a2, 1365
> +; RV32I-NEXT: and a1, a1, a2
> ; RV32I-NEXT: sub a0, a0, a1
> ; RV32I-NEXT: lui a1, 209715
> ; RV32I-NEXT: addi a1, a1, 819
> @@ -371,10 +371,10 @@ define i16 @test_cttz_i16_zero_undef(i16
> ; RV32I-NEXT: addi a1, a0, -1
> ; RV32I-NEXT: not a0, a0
> ; RV32I-NEXT: and a0, a0, a1
> -; RV32I-NEXT: lui a1, 349525
> -; RV32I-NEXT: addi a1, a1, 1365
> -; RV32I-NEXT: srli a2, a0, 1
> -; RV32I-NEXT: and a1, a2, a1
> +; RV32I-NEXT: srli a1, a0, 1
> +; RV32I-NEXT: lui a2, 349525
> +; RV32I-NEXT: addi a2, a2, 1365
> +; RV32I-NEXT: and a1, a1, a2
> ; RV32I-NEXT: sub a0, a0, a1
> ; RV32I-NEXT: lui a1, 209715
> ; RV32I-NEXT: addi a1, a1, 819
> @@ -406,10 +406,10 @@ define i32 @test_cttz_i32_zero_undef(i32
> ; RV32I-NEXT: addi a1, a0, -1
> ; RV32I-NEXT: not a0, a0
> ; RV32I-NEXT: and a0, a0, a1
> -; RV32I-NEXT: lui a1, 349525
> -; RV32I-NEXT: addi a1, a1, 1365
> -; RV32I-NEXT: srli a2, a0, 1
> -; RV32I-NEXT: and a1, a2, a1
> +; RV32I-NEXT: srli a1, a0, 1
> +; RV32I-NEXT: lui a2, 349525
> +; RV32I-NEXT: addi a2, a2, 1365
> +; RV32I-NEXT: and a1, a1, a2
> ; RV32I-NEXT: sub a0, a0, a1
> ; RV32I-NEXT: lui a1, 209715
> ; RV32I-NEXT: addi a1, a1, 819
> @@ -450,9 +450,9 @@ define i64 @test_cttz_i64_zero_undef(i64
> ; RV32I-NEXT: addi a0, a0, -1
> ; RV32I-NEXT: not a1, s4
> ; RV32I-NEXT: and a0, a1, a0
> -; RV32I-NEXT: lui a1, 349525
> -; RV32I-NEXT: addi s5, a1, 1365
> ; RV32I-NEXT: srli a1, a0, 1
> +; RV32I-NEXT: lui a2, 349525
> +; RV32I-NEXT: addi s5, a2, 1365
> ; RV32I-NEXT: and a1, a1, s5
> ; RV32I-NEXT: sub a0, a0, a1
> ; RV32I-NEXT: lui a1, 209715
> @@ -463,12 +463,12 @@ define i64 @test_cttz_i64_zero_undef(i64
> ; RV32I-NEXT: add a0, a1, a0
> ; RV32I-NEXT: srli a1, a0, 4
> ; RV32I-NEXT: add a0, a0, a1
> -; RV32I-NEXT: lui a1, 4112
> -; RV32I-NEXT: addi s6, a1, 257
> ; RV32I-NEXT: lui a1, 61681
> -; RV32I-NEXT: addi s1, a1, -241
> -; RV32I-NEXT: and a0, a0, s1
> -; RV32I-NEXT: mv a1, s6
> +; RV32I-NEXT: addi s6, a1, -241
> +; RV32I-NEXT: and a0, a0, s6
> +; RV32I-NEXT: lui a1, 4112
> +; RV32I-NEXT: addi s1, a1, 257
> +; RV32I-NEXT: mv a1, s1
> ; RV32I-NEXT: call __mulsi3
> ; RV32I-NEXT: mv s2, a0
> ; RV32I-NEXT: addi a0, s3, -1
> @@ -483,8 +483,8 @@ define i64 @test_cttz_i64_zero_undef(i64
> ; RV32I-NEXT: add a0, a1, a0
> ; RV32I-NEXT: srli a1, a0, 4
> ; RV32I-NEXT: add a0, a0, a1
> -; RV32I-NEXT: and a0, a0, s1
> -; RV32I-NEXT: mv a1, s6
> +; RV32I-NEXT: and a0, a0, s6
> +; RV32I-NEXT: mv a1, s1
> ; RV32I-NEXT: call __mulsi3
> ; RV32I-NEXT: bnez s4, .LBB11_2
> ; RV32I-NEXT: # %bb.1:
> @@ -514,10 +514,10 @@ define i32 @test_ctpop_i32(i32 %a) nounw
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi sp, sp, -16
> ; RV32I-NEXT: sw ra, 12(sp)
> -; RV32I-NEXT: lui a1, 349525
> -; RV32I-NEXT: addi a1, a1, 1365
> -; RV32I-NEXT: srli a2, a0, 1
> -; RV32I-NEXT: and a1, a2, a1
> +; RV32I-NEXT: srli a1, a0, 1
> +; RV32I-NEXT: lui a2, 349525
> +; RV32I-NEXT: addi a2, a2, 1365
> +; RV32I-NEXT: and a1, a1, a2
> ; RV32I-NEXT: sub a0, a0, a1
> ; RV32I-NEXT: lui a1, 209715
> ; RV32I-NEXT: addi a1, a1, 819
>
> Modified: llvm/trunk/test/CodeGen/RISCV/callee-saved-fpr32s.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/callee-saved-fpr32s.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/callee-saved-fpr32s.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/callee-saved-fpr32s.ll Tue Sep 17 02:43:08 2019
> @@ -23,8 +23,8 @@ define void @callee() nounwind {
> ; ILP32-LP64-LABEL: callee:
> ; ILP32-LP64: # %bb.0:
> ; ILP32-LP64-NEXT: lui a0, %hi(var)
> -; ILP32-LP64-NEXT: addi a1, a0, %lo(var)
> ; ILP32-LP64-NEXT: flw ft0, %lo(var)(a0)
> +; ILP32-LP64-NEXT: addi a1, a0, %lo(var)
> ; ILP32-LP64-NEXT: flw ft1, 4(a1)
> ; ILP32-LP64-NEXT: flw ft2, 8(a1)
> ; ILP32-LP64-NEXT: flw ft3, 12(a1)
> @@ -52,14 +52,14 @@ define void @callee() nounwind {
> ; ILP32-LP64-NEXT: flw fs5, 100(a1)
> ; ILP32-LP64-NEXT: flw fs6, 104(a1)
> ; ILP32-LP64-NEXT: flw fs7, 108(a1)
> -; ILP32-LP64-NEXT: flw fs8, 112(a1)
> -; ILP32-LP64-NEXT: flw fs9, 116(a1)
> -; ILP32-LP64-NEXT: flw fs10, 120(a1)
> -; ILP32-LP64-NEXT: flw fs11, 124(a1)
> -; ILP32-LP64-NEXT: fsw fs11, 124(a1)
> -; ILP32-LP64-NEXT: fsw fs10, 120(a1)
> -; ILP32-LP64-NEXT: fsw fs9, 116(a1)
> -; ILP32-LP64-NEXT: fsw fs8, 112(a1)
> +; ILP32-LP64-NEXT: flw fs8, 124(a1)
> +; ILP32-LP64-NEXT: flw fs9, 120(a1)
> +; ILP32-LP64-NEXT: flw fs10, 116(a1)
> +; ILP32-LP64-NEXT: flw fs11, 112(a1)
> +; ILP32-LP64-NEXT: fsw fs8, 124(a1)
> +; ILP32-LP64-NEXT: fsw fs9, 120(a1)
> +; ILP32-LP64-NEXT: fsw fs10, 116(a1)
> +; ILP32-LP64-NEXT: fsw fs11, 112(a1)
> ; ILP32-LP64-NEXT: fsw fs7, 108(a1)
> ; ILP32-LP64-NEXT: fsw fs6, 104(a1)
> ; ILP32-LP64-NEXT: fsw fs5, 100(a1)
> @@ -106,6 +106,7 @@ define void @callee() nounwind {
> ; ILP32F-LP64F-NEXT: fsw fs10, 4(sp)
> ; ILP32F-LP64F-NEXT: fsw fs11, 0(sp)
> ; ILP32F-LP64F-NEXT: lui a0, %hi(var)
> +; ILP32F-LP64F-NEXT: flw ft0, %lo(var)(a0)
> ; ILP32F-LP64F-NEXT: addi a1, a0, %lo(var)
> ;
> ; ILP32D-LP64D-LABEL: callee:
> @@ -124,6 +125,7 @@ define void @callee() nounwind {
> ; ILP32D-LP64D-NEXT: fsd fs10, 8(sp)
> ; ILP32D-LP64D-NEXT: fsd fs11, 0(sp)
> ; ILP32D-LP64D-NEXT: lui a0, %hi(var)
> +; ILP32D-LP64D-NEXT: flw ft0, %lo(var)(a0)
> ; ILP32D-LP64D-NEXT: addi a1, a0, %lo(var)
> %val = load [32 x float], [32 x float]* @var
> store volatile [32 x float] %val, [32 x float]* @var
>
> Modified: llvm/trunk/test/CodeGen/RISCV/callee-saved-fpr64s.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/callee-saved-fpr64s.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/callee-saved-fpr64s.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/callee-saved-fpr64s.ll Tue Sep 17 02:43:08 2019
> @@ -19,8 +19,8 @@ define void @callee() nounwind {
> ; ILP32-LP64-LABEL: callee:
> ; ILP32-LP64: # %bb.0:
> ; ILP32-LP64-NEXT: lui a0, %hi(var)
> -; ILP32-LP64-NEXT: addi a1, a0, %lo(var)
> ; ILP32-LP64-NEXT: fld ft0, %lo(var)(a0)
> +; ILP32-LP64-NEXT: addi a1, a0, %lo(var)
> ; ILP32-LP64-NEXT: fld ft1, 8(a1)
> ; ILP32-LP64-NEXT: fld ft2, 16(a1)
> ; ILP32-LP64-NEXT: fld ft3, 24(a1)
> @@ -48,14 +48,14 @@ define void @callee() nounwind {
> ; ILP32-LP64-NEXT: fld fs5, 200(a1)
> ; ILP32-LP64-NEXT: fld fs6, 208(a1)
> ; ILP32-LP64-NEXT: fld fs7, 216(a1)
> -; ILP32-LP64-NEXT: fld fs8, 224(a1)
> -; ILP32-LP64-NEXT: fld fs9, 232(a1)
> -; ILP32-LP64-NEXT: fld fs10, 240(a1)
> -; ILP32-LP64-NEXT: fld fs11, 248(a1)
> -; ILP32-LP64-NEXT: fsd fs11, 248(a1)
> -; ILP32-LP64-NEXT: fsd fs10, 240(a1)
> -; ILP32-LP64-NEXT: fsd fs9, 232(a1)
> -; ILP32-LP64-NEXT: fsd fs8, 224(a1)
> +; ILP32-LP64-NEXT: fld fs8, 248(a1)
> +; ILP32-LP64-NEXT: fld fs9, 240(a1)
> +; ILP32-LP64-NEXT: fld fs10, 232(a1)
> +; ILP32-LP64-NEXT: fld fs11, 224(a1)
> +; ILP32-LP64-NEXT: fsd fs8, 248(a1)
> +; ILP32-LP64-NEXT: fsd fs9, 240(a1)
> +; ILP32-LP64-NEXT: fsd fs10, 232(a1)
> +; ILP32-LP64-NEXT: fsd fs11, 224(a1)
> ; ILP32-LP64-NEXT: fsd fs7, 216(a1)
> ; ILP32-LP64-NEXT: fsd fs6, 208(a1)
> ; ILP32-LP64-NEXT: fsd fs5, 200(a1)
> @@ -102,6 +102,7 @@ define void @callee() nounwind {
> ; ILP32D-LP64D-NEXT: fsd fs10, 8(sp)
> ; ILP32D-LP64D-NEXT: fsd fs11, 0(sp)
> ; ILP32D-LP64D-NEXT: lui a0, %hi(var)
> +; ILP32D-LP64D-NEXT: fld ft0, %lo(var)(a0)
> ; ILP32D-LP64D-NEXT: addi a1, a0, %lo(var)
> %val = load [32 x double], [32 x double]* @var
> store volatile [32 x double] %val, [32 x double]* @var
>
> Modified: llvm/trunk/test/CodeGen/RISCV/callee-saved-gprs.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/callee-saved-gprs.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/callee-saved-gprs.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/callee-saved-gprs.ll Tue Sep 17 02:43:08 2019
> @@ -41,7 +41,9 @@ define void @callee() nounwind {
> ; RV32I-NEXT: sw s10, 36(sp)
> ; RV32I-NEXT: sw s11, 32(sp)
> ; RV32I-NEXT: lui a0, %hi(var)
> -; RV32I-NEXT: addi a1, a0, %lo(var)
> +; RV32I-NEXT: lw a1, %lo(var)(a0)
> +; RV32I-NEXT: sw a1, 28(sp)
> +; RV32I-NEXT: addi a2, a0, %lo(var)
> ;
> ; RV32I-WITH-FP-LABEL: callee:
> ; RV32I-WITH-FP: # %bb.0:
> @@ -61,7 +63,9 @@ define void @callee() nounwind {
> ; RV32I-WITH-FP-NEXT: sw s11, 28(sp)
> ; RV32I-WITH-FP-NEXT: addi s0, sp, 80
> ; RV32I-WITH-FP-NEXT: lui a0, %hi(var)
> -; RV32I-WITH-FP-NEXT: addi a1, a0, %lo(var)
> +; RV32I-WITH-FP-NEXT: lw a1, %lo(var)(a0)
> +; RV32I-WITH-FP-NEXT: sw a1, -56(s0)
> +; RV32I-WITH-FP-NEXT: addi a2, a0, %lo(var)
> ;
> ; RV64I-LABEL: callee:
> ; RV64I: # %bb.0:
> @@ -79,7 +83,9 @@ define void @callee() nounwind {
> ; RV64I-NEXT: sd s10, 56(sp)
> ; RV64I-NEXT: sd s11, 48(sp)
> ; RV64I-NEXT: lui a0, %hi(var)
> -; RV64I-NEXT: addi a1, a0, %lo(var)
> +; RV64I-NEXT: lw a1, %lo(var)(a0)
> +; RV64I-NEXT: sd a1, 40(sp)
> +; RV64I-NEXT: addi a2, a0, %lo(var)
> ;
> ; RV64I-WITH-FP-LABEL: callee:
> ; RV64I-WITH-FP: # %bb.0:
> @@ -99,7 +105,9 @@ define void @callee() nounwind {
> ; RV64I-WITH-FP-NEXT: sd s11, 56(sp)
> ; RV64I-WITH-FP-NEXT: addi s0, sp, 160
> ; RV64I-WITH-FP-NEXT: lui a0, %hi(var)
> -; RV64I-WITH-FP-NEXT: addi a1, a0, %lo(var)
> +; RV64I-WITH-FP-NEXT: lw a1, %lo(var)(a0)
> +; RV64I-WITH-FP-NEXT: sd a1, -112(s0)
> +; RV64I-WITH-FP-NEXT: addi a2, a0, %lo(var)
> %val = load [32 x i32], [32 x i32]* @var
> store volatile [32 x i32] %val, [32 x i32]* @var
> ret void
> @@ -111,36 +119,41 @@ define void @callee() nounwind {
> define void @caller() nounwind {
> ; RV32I-LABEL: caller:
> ; RV32I: lui a0, %hi(var)
> -; RV32I-NEXT: addi s1, a0, %lo(var)
> +; RV32I-NEXT: lw a1, %lo(var)(a0)
> +; RV32I-NEXT: sw a1, 88(sp)
> +; RV32I-NEXT: addi s0, a0, %lo(var)
> +
> ; RV32I: sw a0, 8(sp)
> -; RV32I-NEXT: lw s2, 84(s1)
> -; RV32I-NEXT: lw s3, 88(s1)
> -; RV32I-NEXT: lw s4, 92(s1)
> -; RV32I-NEXT: lw s5, 96(s1)
> -; RV32I-NEXT: lw s6, 100(s1)
> -; RV32I-NEXT: lw s7, 104(s1)
> -; RV32I-NEXT: lw s8, 108(s1)
> -; RV32I-NEXT: lw s9, 112(s1)
> -; RV32I-NEXT: lw s10, 116(s1)
> -; RV32I-NEXT: lw s11, 120(s1)
> -; RV32I-NEXT: lw s0, 124(s1)
> +; RV32I-NEXT: lw s2, 84(s0)
> +; RV32I-NEXT: lw s3, 88(s0)
> +; RV32I-NEXT: lw s4, 92(s0)
> +; RV32I-NEXT: lw s5, 96(s0)
> +; RV32I-NEXT: lw s6, 100(s0)
> +; RV32I-NEXT: lw s7, 104(s0)
> +; RV32I-NEXT: lw s8, 108(s0)
> +; RV32I-NEXT: lw s9, 112(s0)
> +; RV32I-NEXT: lw s10, 116(s0)
> +; RV32I-NEXT: lw s11, 120(s0)
> +; RV32I-NEXT: lw s1, 124(s0)
> ; RV32I-NEXT: call callee
> -; RV32I-NEXT: sw s0, 124(s1)
> -; RV32I-NEXT: sw s11, 120(s1)
> -; RV32I-NEXT: sw s10, 116(s1)
> -; RV32I-NEXT: sw s9, 112(s1)
> -; RV32I-NEXT: sw s8, 108(s1)
> -; RV32I-NEXT: sw s7, 104(s1)
> -; RV32I-NEXT: sw s6, 100(s1)
> -; RV32I-NEXT: sw s5, 96(s1)
> -; RV32I-NEXT: sw s4, 92(s1)
> -; RV32I-NEXT: sw s3, 88(s1)
> -; RV32I-NEXT: sw s2, 84(s1)
> +; RV32I-NEXT: sw s1, 124(s0)
> +; RV32I-NEXT: sw s11, 120(s0)
> +; RV32I-NEXT: sw s10, 116(s0)
> +; RV32I-NEXT: sw s9, 112(s0)
> +; RV32I-NEXT: sw s8, 108(s0)
> +; RV32I-NEXT: sw s7, 104(s0)
> +; RV32I-NEXT: sw s6, 100(s0)
> +; RV32I-NEXT: sw s5, 96(s0)
> +; RV32I-NEXT: sw s4, 92(s0)
> +; RV32I-NEXT: sw s3, 88(s0)
> +; RV32I-NEXT: sw s2, 84(s0)
> ; RV32I-NEXT: lw a0, 8(sp)
> ;
> ; RV32I-WITH-FP-LABEL: caller:
> ; RV32I-WITH-FP: addi s0, sp, 144
> ; RV32I-WITH-FP-NEXT: lui a0, %hi(var)
> +; RV32I-WITH-FP-NEXT: lw a1, %lo(var)(a0)
> +; RV32I-WITH-FP-NEXT: sw a1, -56(s0)
> ; RV32I-WITH-FP-NEXT: addi s1, a0, %lo(var)
> ; RV32I-WITH-FP: sw a0, -140(s0)
> ; RV32I-WITH-FP-NEXT: lw s5, 88(s1)
> @@ -168,36 +181,40 @@ define void @caller() nounwind {
> ;
> ; RV64I-LABEL: caller:
> ; RV64I: lui a0, %hi(var)
> -; RV64I-NEXT: addi s1, a0, %lo(var)
> +; RV64I-NEXT: lw a1, %lo(var)(a0)
> +; RV64I-NEXT: sd a1, 160(sp)
> +; RV64I-NEXT: addi s0, a0, %lo(var)
> ; RV64I: sd a0, 0(sp)
> -; RV64I-NEXT: lw s2, 84(s1)
> -; RV64I-NEXT: lw s3, 88(s1)
> -; RV64I-NEXT: lw s4, 92(s1)
> -; RV64I-NEXT: lw s5, 96(s1)
> -; RV64I-NEXT: lw s6, 100(s1)
> -; RV64I-NEXT: lw s7, 104(s1)
> -; RV64I-NEXT: lw s8, 108(s1)
> -; RV64I-NEXT: lw s9, 112(s1)
> -; RV64I-NEXT: lw s10, 116(s1)
> -; RV64I-NEXT: lw s11, 120(s1)
> -; RV64I-NEXT: lw s0, 124(s1)
> +; RV64I-NEXT: lw s2, 84(s0)
> +; RV64I-NEXT: lw s3, 88(s0)
> +; RV64I-NEXT: lw s4, 92(s0)
> +; RV64I-NEXT: lw s5, 96(s0)
> +; RV64I-NEXT: lw s6, 100(s0)
> +; RV64I-NEXT: lw s7, 104(s0)
> +; RV64I-NEXT: lw s8, 108(s0)
> +; RV64I-NEXT: lw s9, 112(s0)
> +; RV64I-NEXT: lw s10, 116(s0)
> +; RV64I-NEXT: lw s11, 120(s0)
> +; RV64I-NEXT: lw s1, 124(s0)
> ; RV64I-NEXT: call callee
> -; RV64I-NEXT: sw s0, 124(s1)
> -; RV64I-NEXT: sw s11, 120(s1)
> -; RV64I-NEXT: sw s10, 116(s1)
> -; RV64I-NEXT: sw s9, 112(s1)
> -; RV64I-NEXT: sw s8, 108(s1)
> -; RV64I-NEXT: sw s7, 104(s1)
> -; RV64I-NEXT: sw s6, 100(s1)
> -; RV64I-NEXT: sw s5, 96(s1)
> -; RV64I-NEXT: sw s4, 92(s1)
> -; RV64I-NEXT: sw s3, 88(s1)
> -; RV64I-NEXT: sw s2, 84(s1)
> +; RV64I-NEXT: sw s1, 124(s0)
> +; RV64I-NEXT: sw s11, 120(s0)
> +; RV64I-NEXT: sw s10, 116(s0)
> +; RV64I-NEXT: sw s9, 112(s0)
> +; RV64I-NEXT: sw s8, 108(s0)
> +; RV64I-NEXT: sw s7, 104(s0)
> +; RV64I-NEXT: sw s6, 100(s0)
> +; RV64I-NEXT: sw s5, 96(s0)
> +; RV64I-NEXT: sw s4, 92(s0)
> +; RV64I-NEXT: sw s3, 88(s0)
> +; RV64I-NEXT: sw s2, 84(s0)
> ; RV64I-NEXT: ld a0, 0(sp)
> ;
> ; RV64I-WITH-FP-LABEL: caller:
> ; RV64I-WITH-FP: addi s0, sp, 288
> ; RV64I-WITH-FP-NEXT: lui a0, %hi(var)
> +; RV64I-WITH-FP-NEXT: lw a1, %lo(var)(a0)
> +; RV64I-WITH-FP-NEXT: sd a1, -112(s0)
> ; RV64I-WITH-FP-NEXT: addi s1, a0, %lo(var)
> ; RV64I-WITH-FP: sd a0, -280(s0)
> ; RV64I-WITH-FP-NEXT: lw s5, 88(s1)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32-ilp32f-common.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32-ilp32f-common.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32-ilp32f-common.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32-ilp32f-common.ll Tue Sep 17 02:43:08 2019
> @@ -61,8 +61,8 @@ define i32 @caller_double_in_regs() noun
> ; RV32I-FPELIM-NEXT: addi sp, sp, -16
> ; RV32I-FPELIM-NEXT: sw ra, 12(sp)
> ; RV32I-FPELIM-NEXT: addi a0, zero, 1
> -; RV32I-FPELIM-NEXT: mv a1, zero
> ; RV32I-FPELIM-NEXT: lui a2, 262144
> +; RV32I-FPELIM-NEXT: mv a1, zero
> ; RV32I-FPELIM-NEXT: call callee_double_in_regs
> ; RV32I-FPELIM-NEXT: lw ra, 12(sp)
> ; RV32I-FPELIM-NEXT: addi sp, sp, 16
> @@ -75,8 +75,8 @@ define i32 @caller_double_in_regs() noun
> ; RV32I-WITHFP-NEXT: sw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: addi s0, sp, 16
> ; RV32I-WITHFP-NEXT: addi a0, zero, 1
> -; RV32I-WITHFP-NEXT: mv a1, zero
> ; RV32I-WITHFP-NEXT: lui a2, 262144
> +; RV32I-WITHFP-NEXT: mv a1, zero
> ; RV32I-WITHFP-NEXT: call callee_double_in_regs
> ; RV32I-WITHFP-NEXT: lw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 12(sp)
> @@ -94,14 +94,14 @@ define i32 @callee_aligned_stack(i32 %a,
> ; RV32I-FPELIM-LABEL: callee_aligned_stack:
> ; RV32I-FPELIM: # %bb.0:
> ; RV32I-FPELIM-NEXT: lw a0, 0(a2)
> -; RV32I-FPELIM-NEXT: add a0, a0, a7
> -; RV32I-FPELIM-NEXT: lw a1, 0(sp)
> -; RV32I-FPELIM-NEXT: add a0, a0, a1
> -; RV32I-FPELIM-NEXT: lw a1, 8(sp)
> -; RV32I-FPELIM-NEXT: add a0, a0, a1
> -; RV32I-FPELIM-NEXT: lw a1, 16(sp)
> -; RV32I-FPELIM-NEXT: add a0, a0, a1
> ; RV32I-FPELIM-NEXT: lw a1, 20(sp)
> +; RV32I-FPELIM-NEXT: lw a2, 0(sp)
> +; RV32I-FPELIM-NEXT: lw a3, 8(sp)
> +; RV32I-FPELIM-NEXT: lw a4, 16(sp)
> +; RV32I-FPELIM-NEXT: add a0, a0, a7
> +; RV32I-FPELIM-NEXT: add a0, a0, a2
> +; RV32I-FPELIM-NEXT: add a0, a0, a3
> +; RV32I-FPELIM-NEXT: add a0, a0, a4
> ; RV32I-FPELIM-NEXT: add a0, a0, a1
> ; RV32I-FPELIM-NEXT: ret
> ;
> @@ -112,14 +112,14 @@ define i32 @callee_aligned_stack(i32 %a,
> ; RV32I-WITHFP-NEXT: sw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: addi s0, sp, 16
> ; RV32I-WITHFP-NEXT: lw a0, 0(a2)
> -; RV32I-WITHFP-NEXT: add a0, a0, a7
> -; RV32I-WITHFP-NEXT: lw a1, 0(s0)
> -; RV32I-WITHFP-NEXT: add a0, a0, a1
> -; RV32I-WITHFP-NEXT: lw a1, 8(s0)
> -; RV32I-WITHFP-NEXT: add a0, a0, a1
> -; RV32I-WITHFP-NEXT: lw a1, 16(s0)
> -; RV32I-WITHFP-NEXT: add a0, a0, a1
> ; RV32I-WITHFP-NEXT: lw a1, 20(s0)
> +; RV32I-WITHFP-NEXT: lw a2, 0(s0)
> +; RV32I-WITHFP-NEXT: lw a3, 8(s0)
> +; RV32I-WITHFP-NEXT: lw a4, 16(s0)
> +; RV32I-WITHFP-NEXT: add a0, a0, a7
> +; RV32I-WITHFP-NEXT: add a0, a0, a2
> +; RV32I-WITHFP-NEXT: add a0, a0, a3
> +; RV32I-WITHFP-NEXT: add a0, a0, a4
> ; RV32I-WITHFP-NEXT: add a0, a0, a1
> ; RV32I-WITHFP-NEXT: lw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 12(sp)
> @@ -169,8 +169,7 @@ define void @caller_aligned_stack() noun
> ; RV32I-FPELIM-NEXT: addi a0, a0, -328
> ; RV32I-FPELIM-NEXT: sw a0, 36(sp)
> ; RV32I-FPELIM-NEXT: lui a0, 335544
> -; RV32I-FPELIM-NEXT: addi a0, a0, 1311
> -; RV32I-FPELIM-NEXT: sw a0, 32(sp)
> +; RV32I-FPELIM-NEXT: addi t0, a0, 1311
> ; RV32I-FPELIM-NEXT: lui a0, 688509
> ; RV32I-FPELIM-NEXT: addi a5, a0, -2048
> ; RV32I-FPELIM-NEXT: addi a2, sp, 32
> @@ -180,6 +179,7 @@ define void @caller_aligned_stack() noun
> ; RV32I-FPELIM-NEXT: addi a4, zero, 13
> ; RV32I-FPELIM-NEXT: addi a6, zero, 4
> ; RV32I-FPELIM-NEXT: addi a7, zero, 14
> +; RV32I-FPELIM-NEXT: sw t0, 32(sp)
> ; RV32I-FPELIM-NEXT: call callee_aligned_stack
> ; RV32I-FPELIM-NEXT: lw ra, 60(sp)
> ; RV32I-FPELIM-NEXT: addi sp, sp, 64
> @@ -215,8 +215,7 @@ define void @caller_aligned_stack() noun
> ; RV32I-WITHFP-NEXT: addi a0, a0, -328
> ; RV32I-WITHFP-NEXT: sw a0, -28(s0)
> ; RV32I-WITHFP-NEXT: lui a0, 335544
> -; RV32I-WITHFP-NEXT: addi a0, a0, 1311
> -; RV32I-WITHFP-NEXT: sw a0, -32(s0)
> +; RV32I-WITHFP-NEXT: addi t0, a0, 1311
> ; RV32I-WITHFP-NEXT: lui a0, 688509
> ; RV32I-WITHFP-NEXT: addi a5, a0, -2048
> ; RV32I-WITHFP-NEXT: addi a2, s0, -32
> @@ -226,6 +225,7 @@ define void @caller_aligned_stack() noun
> ; RV32I-WITHFP-NEXT: addi a4, zero, 13
> ; RV32I-WITHFP-NEXT: addi a6, zero, 4
> ; RV32I-WITHFP-NEXT: addi a7, zero, 14
> +; RV32I-WITHFP-NEXT: sw t0, -32(s0)
> ; RV32I-WITHFP-NEXT: call callee_aligned_stack
> ; RV32I-WITHFP-NEXT: lw s0, 56(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 60(sp)
> @@ -241,8 +241,8 @@ define void @caller_aligned_stack() noun
> define double @callee_small_scalar_ret() nounwind {
> ; RV32I-FPELIM-LABEL: callee_small_scalar_ret:
> ; RV32I-FPELIM: # %bb.0:
> -; RV32I-FPELIM-NEXT: mv a0, zero
> ; RV32I-FPELIM-NEXT: lui a1, 261888
> +; RV32I-FPELIM-NEXT: mv a0, zero
> ; RV32I-FPELIM-NEXT: ret
> ;
> ; RV32I-WITHFP-LABEL: callee_small_scalar_ret:
> @@ -251,8 +251,8 @@ define double @callee_small_scalar_ret()
> ; RV32I-WITHFP-NEXT: sw ra, 12(sp)
> ; RV32I-WITHFP-NEXT: sw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: addi s0, sp, 16
> -; RV32I-WITHFP-NEXT: mv a0, zero
> ; RV32I-WITHFP-NEXT: lui a1, 261888
> +; RV32I-WITHFP-NEXT: mv a0, zero
> ; RV32I-WITHFP-NEXT: lw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 12(sp)
> ; RV32I-WITHFP-NEXT: addi sp, sp, 16
>
> Modified: llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32-ilp32f-ilp32d-common.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32-ilp32f-ilp32d-common.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32-ilp32f-ilp32d-common.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32-ilp32f-ilp32d-common.ll Tue Sep 17 02:43:08 2019
> @@ -82,22 +82,22 @@ define i32 @caller_i64_in_regs() nounwin
> define i32 @callee_many_scalars(i8 %a, i16 %b, i32 %c, i64 %d, i32 %e, i32 %f, i64 %g, i32 %h) nounwind {
> ; RV32I-FPELIM-LABEL: callee_many_scalars:
> ; RV32I-FPELIM: # %bb.0:
> -; RV32I-FPELIM-NEXT: lw t0, 0(sp)
> -; RV32I-FPELIM-NEXT: xor a4, a4, t0
> -; RV32I-FPELIM-NEXT: xor a3, a3, a7
> -; RV32I-FPELIM-NEXT: or a3, a3, a4
> -; RV32I-FPELIM-NEXT: lui a4, 16
> -; RV32I-FPELIM-NEXT: addi a4, a4, -1
> -; RV32I-FPELIM-NEXT: and a1, a1, a4
> -; RV32I-FPELIM-NEXT: andi a0, a0, 255
> -; RV32I-FPELIM-NEXT: add a0, a0, a1
> +; RV32I-FPELIM-NEXT: lw t0, 4(sp)
> +; RV32I-FPELIM-NEXT: lw t1, 0(sp)
> +; RV32I-FPELIM-NEXT: andi t2, a0, 255
> +; RV32I-FPELIM-NEXT: lui a0, 16
> +; RV32I-FPELIM-NEXT: addi a0, a0, -1
> +; RV32I-FPELIM-NEXT: and a0, a1, a0
> +; RV32I-FPELIM-NEXT: add a0, t2, a0
> ; RV32I-FPELIM-NEXT: add a0, a0, a2
> -; RV32I-FPELIM-NEXT: seqz a1, a3
> +; RV32I-FPELIM-NEXT: xor a1, a4, t1
> +; RV32I-FPELIM-NEXT: xor a2, a3, a7
> +; RV32I-FPELIM-NEXT: or a1, a2, a1
> +; RV32I-FPELIM-NEXT: seqz a1, a1
> ; RV32I-FPELIM-NEXT: add a0, a1, a0
> ; RV32I-FPELIM-NEXT: add a0, a0, a5
> ; RV32I-FPELIM-NEXT: add a0, a0, a6
> -; RV32I-FPELIM-NEXT: lw a1, 4(sp)
> -; RV32I-FPELIM-NEXT: add a0, a0, a1
> +; RV32I-FPELIM-NEXT: add a0, a0, t0
> ; RV32I-FPELIM-NEXT: ret
> ;
> ; RV32I-WITHFP-LABEL: callee_many_scalars:
> @@ -106,22 +106,22 @@ define i32 @callee_many_scalars(i8 %a, i
> ; RV32I-WITHFP-NEXT: sw ra, 12(sp)
> ; RV32I-WITHFP-NEXT: sw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: addi s0, sp, 16
> -; RV32I-WITHFP-NEXT: lw t0, 0(s0)
> -; RV32I-WITHFP-NEXT: xor a4, a4, t0
> -; RV32I-WITHFP-NEXT: xor a3, a3, a7
> -; RV32I-WITHFP-NEXT: or a3, a3, a4
> -; RV32I-WITHFP-NEXT: lui a4, 16
> -; RV32I-WITHFP-NEXT: addi a4, a4, -1
> -; RV32I-WITHFP-NEXT: and a1, a1, a4
> -; RV32I-WITHFP-NEXT: andi a0, a0, 255
> -; RV32I-WITHFP-NEXT: add a0, a0, a1
> +; RV32I-WITHFP-NEXT: lw t0, 4(s0)
> +; RV32I-WITHFP-NEXT: lw t1, 0(s0)
> +; RV32I-WITHFP-NEXT: andi t2, a0, 255
> +; RV32I-WITHFP-NEXT: lui a0, 16
> +; RV32I-WITHFP-NEXT: addi a0, a0, -1
> +; RV32I-WITHFP-NEXT: and a0, a1, a0
> +; RV32I-WITHFP-NEXT: add a0, t2, a0
> ; RV32I-WITHFP-NEXT: add a0, a0, a2
> -; RV32I-WITHFP-NEXT: seqz a1, a3
> +; RV32I-WITHFP-NEXT: xor a1, a4, t1
> +; RV32I-WITHFP-NEXT: xor a2, a3, a7
> +; RV32I-WITHFP-NEXT: or a1, a2, a1
> +; RV32I-WITHFP-NEXT: seqz a1, a1
> ; RV32I-WITHFP-NEXT: add a0, a1, a0
> ; RV32I-WITHFP-NEXT: add a0, a0, a5
> ; RV32I-WITHFP-NEXT: add a0, a0, a6
> -; RV32I-WITHFP-NEXT: lw a1, 4(s0)
> -; RV32I-WITHFP-NEXT: add a0, a0, a1
> +; RV32I-WITHFP-NEXT: add a0, a0, t0
> ; RV32I-WITHFP-NEXT: lw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 12(sp)
> ; RV32I-WITHFP-NEXT: addi sp, sp, 16
> @@ -146,15 +146,15 @@ define i32 @caller_many_scalars() nounwi
> ; RV32I-FPELIM-NEXT: sw ra, 12(sp)
> ; RV32I-FPELIM-NEXT: addi a0, zero, 8
> ; RV32I-FPELIM-NEXT: sw a0, 4(sp)
> -; RV32I-FPELIM-NEXT: sw zero, 0(sp)
> ; RV32I-FPELIM-NEXT: addi a0, zero, 1
> ; RV32I-FPELIM-NEXT: addi a1, zero, 2
> ; RV32I-FPELIM-NEXT: addi a2, zero, 3
> ; RV32I-FPELIM-NEXT: addi a3, zero, 4
> -; RV32I-FPELIM-NEXT: mv a4, zero
> ; RV32I-FPELIM-NEXT: addi a5, zero, 5
> ; RV32I-FPELIM-NEXT: addi a6, zero, 6
> ; RV32I-FPELIM-NEXT: addi a7, zero, 7
> +; RV32I-FPELIM-NEXT: sw zero, 0(sp)
> +; RV32I-FPELIM-NEXT: mv a4, zero
> ; RV32I-FPELIM-NEXT: call callee_many_scalars
> ; RV32I-FPELIM-NEXT: lw ra, 12(sp)
> ; RV32I-FPELIM-NEXT: addi sp, sp, 16
> @@ -168,15 +168,15 @@ define i32 @caller_many_scalars() nounwi
> ; RV32I-WITHFP-NEXT: addi s0, sp, 16
> ; RV32I-WITHFP-NEXT: addi a0, zero, 8
> ; RV32I-WITHFP-NEXT: sw a0, 4(sp)
> -; RV32I-WITHFP-NEXT: sw zero, 0(sp)
> ; RV32I-WITHFP-NEXT: addi a0, zero, 1
> ; RV32I-WITHFP-NEXT: addi a1, zero, 2
> ; RV32I-WITHFP-NEXT: addi a2, zero, 3
> ; RV32I-WITHFP-NEXT: addi a3, zero, 4
> -; RV32I-WITHFP-NEXT: mv a4, zero
> ; RV32I-WITHFP-NEXT: addi a5, zero, 5
> ; RV32I-WITHFP-NEXT: addi a6, zero, 6
> ; RV32I-WITHFP-NEXT: addi a7, zero, 7
> +; RV32I-WITHFP-NEXT: sw zero, 0(sp)
> +; RV32I-WITHFP-NEXT: mv a4, zero
> ; RV32I-WITHFP-NEXT: call callee_many_scalars
> ; RV32I-WITHFP-NEXT: lw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 12(sp)
> @@ -192,20 +192,20 @@ define i32 @caller_many_scalars() nounwi
> define i32 @callee_large_scalars(i128 %a, fp128 %b) nounwind {
> ; RV32I-FPELIM-LABEL: callee_large_scalars:
> ; RV32I-FPELIM: # %bb.0:
> -; RV32I-FPELIM-NEXT: lw a2, 12(a1)
> -; RV32I-FPELIM-NEXT: lw a3, 12(a0)
> -; RV32I-FPELIM-NEXT: xor a2, a3, a2
> -; RV32I-FPELIM-NEXT: lw a3, 4(a1)
> -; RV32I-FPELIM-NEXT: lw a4, 4(a0)
> -; RV32I-FPELIM-NEXT: xor a3, a4, a3
> +; RV32I-FPELIM-NEXT: lw a6, 0(a1)
> +; RV32I-FPELIM-NEXT: lw a7, 0(a0)
> +; RV32I-FPELIM-NEXT: lw a4, 4(a1)
> +; RV32I-FPELIM-NEXT: lw a5, 12(a1)
> +; RV32I-FPELIM-NEXT: lw a2, 12(a0)
> +; RV32I-FPELIM-NEXT: lw a3, 4(a0)
> +; RV32I-FPELIM-NEXT: lw a1, 8(a1)
> +; RV32I-FPELIM-NEXT: lw a0, 8(a0)
> +; RV32I-FPELIM-NEXT: xor a2, a2, a5
> +; RV32I-FPELIM-NEXT: xor a3, a3, a4
> ; RV32I-FPELIM-NEXT: or a2, a3, a2
> -; RV32I-FPELIM-NEXT: lw a3, 8(a1)
> -; RV32I-FPELIM-NEXT: lw a4, 8(a0)
> -; RV32I-FPELIM-NEXT: xor a3, a4, a3
> -; RV32I-FPELIM-NEXT: lw a1, 0(a1)
> -; RV32I-FPELIM-NEXT: lw a0, 0(a0)
> ; RV32I-FPELIM-NEXT: xor a0, a0, a1
> -; RV32I-FPELIM-NEXT: or a0, a0, a3
> +; RV32I-FPELIM-NEXT: xor a1, a7, a6
> +; RV32I-FPELIM-NEXT: or a0, a1, a0
> ; RV32I-FPELIM-NEXT: or a0, a0, a2
> ; RV32I-FPELIM-NEXT: seqz a0, a0
> ; RV32I-FPELIM-NEXT: ret
> @@ -216,20 +216,20 @@ define i32 @callee_large_scalars(i128 %a
> ; RV32I-WITHFP-NEXT: sw ra, 12(sp)
> ; RV32I-WITHFP-NEXT: sw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: addi s0, sp, 16
> -; RV32I-WITHFP-NEXT: lw a2, 12(a1)
> -; RV32I-WITHFP-NEXT: lw a3, 12(a0)
> -; RV32I-WITHFP-NEXT: xor a2, a3, a2
> -; RV32I-WITHFP-NEXT: lw a3, 4(a1)
> -; RV32I-WITHFP-NEXT: lw a4, 4(a0)
> -; RV32I-WITHFP-NEXT: xor a3, a4, a3
> +; RV32I-WITHFP-NEXT: lw a6, 0(a1)
> +; RV32I-WITHFP-NEXT: lw a7, 0(a0)
> +; RV32I-WITHFP-NEXT: lw a4, 4(a1)
> +; RV32I-WITHFP-NEXT: lw a5, 12(a1)
> +; RV32I-WITHFP-NEXT: lw a2, 12(a0)
> +; RV32I-WITHFP-NEXT: lw a3, 4(a0)
> +; RV32I-WITHFP-NEXT: lw a1, 8(a1)
> +; RV32I-WITHFP-NEXT: lw a0, 8(a0)
> +; RV32I-WITHFP-NEXT: xor a2, a2, a5
> +; RV32I-WITHFP-NEXT: xor a3, a3, a4
> ; RV32I-WITHFP-NEXT: or a2, a3, a2
> -; RV32I-WITHFP-NEXT: lw a3, 8(a1)
> -; RV32I-WITHFP-NEXT: lw a4, 8(a0)
> -; RV32I-WITHFP-NEXT: xor a3, a4, a3
> -; RV32I-WITHFP-NEXT: lw a1, 0(a1)
> -; RV32I-WITHFP-NEXT: lw a0, 0(a0)
> ; RV32I-WITHFP-NEXT: xor a0, a0, a1
> -; RV32I-WITHFP-NEXT: or a0, a0, a3
> +; RV32I-WITHFP-NEXT: xor a1, a7, a6
> +; RV32I-WITHFP-NEXT: or a0, a1, a0
> ; RV32I-WITHFP-NEXT: or a0, a0, a2
> ; RV32I-WITHFP-NEXT: seqz a0, a0
> ; RV32I-WITHFP-NEXT: lw s0, 8(sp)
> @@ -255,10 +255,10 @@ define i32 @caller_large_scalars() nounw
> ; RV32I-FPELIM-NEXT: sw zero, 36(sp)
> ; RV32I-FPELIM-NEXT: sw zero, 32(sp)
> ; RV32I-FPELIM-NEXT: sw zero, 28(sp)
> -; RV32I-FPELIM-NEXT: addi a0, zero, 1
> -; RV32I-FPELIM-NEXT: sw a0, 24(sp)
> +; RV32I-FPELIM-NEXT: addi a2, zero, 1
> ; RV32I-FPELIM-NEXT: addi a0, sp, 24
> ; RV32I-FPELIM-NEXT: mv a1, sp
> +; RV32I-FPELIM-NEXT: sw a2, 24(sp)
> ; RV32I-FPELIM-NEXT: call callee_large_scalars
> ; RV32I-FPELIM-NEXT: lw ra, 44(sp)
> ; RV32I-FPELIM-NEXT: addi sp, sp, 48
> @@ -278,10 +278,10 @@ define i32 @caller_large_scalars() nounw
> ; RV32I-WITHFP-NEXT: sw zero, -12(s0)
> ; RV32I-WITHFP-NEXT: sw zero, -16(s0)
> ; RV32I-WITHFP-NEXT: sw zero, -20(s0)
> -; RV32I-WITHFP-NEXT: addi a0, zero, 1
> -; RV32I-WITHFP-NEXT: sw a0, -24(s0)
> +; RV32I-WITHFP-NEXT: addi a2, zero, 1
> ; RV32I-WITHFP-NEXT: addi a0, s0, -24
> ; RV32I-WITHFP-NEXT: addi a1, s0, -48
> +; RV32I-WITHFP-NEXT: sw a2, -24(s0)
> ; RV32I-WITHFP-NEXT: call callee_large_scalars
> ; RV32I-WITHFP-NEXT: lw s0, 40(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 44(sp)
> @@ -299,20 +299,20 @@ define i32 @callee_large_scalars_exhaust
> ; RV32I-FPELIM-LABEL: callee_large_scalars_exhausted_regs:
> ; RV32I-FPELIM: # %bb.0:
> ; RV32I-FPELIM-NEXT: lw a0, 4(sp)
> -; RV32I-FPELIM-NEXT: lw a1, 12(a0)
> -; RV32I-FPELIM-NEXT: lw a2, 12(a7)
> -; RV32I-FPELIM-NEXT: xor a1, a2, a1
> -; RV32I-FPELIM-NEXT: lw a2, 4(a0)
> -; RV32I-FPELIM-NEXT: lw a3, 4(a7)
> -; RV32I-FPELIM-NEXT: xor a2, a3, a2
> -; RV32I-FPELIM-NEXT: or a1, a2, a1
> -; RV32I-FPELIM-NEXT: lw a2, 8(a0)
> -; RV32I-FPELIM-NEXT: lw a3, 8(a7)
> -; RV32I-FPELIM-NEXT: xor a2, a3, a2
> -; RV32I-FPELIM-NEXT: lw a0, 0(a0)
> -; RV32I-FPELIM-NEXT: lw a3, 0(a7)
> -; RV32I-FPELIM-NEXT: xor a0, a3, a0
> -; RV32I-FPELIM-NEXT: or a0, a0, a2
> +; RV32I-FPELIM-NEXT: lw a6, 0(a0)
> +; RV32I-FPELIM-NEXT: lw t0, 0(a7)
> +; RV32I-FPELIM-NEXT: lw a3, 4(a0)
> +; RV32I-FPELIM-NEXT: lw a4, 12(a0)
> +; RV32I-FPELIM-NEXT: lw a5, 12(a7)
> +; RV32I-FPELIM-NEXT: lw a1, 4(a7)
> +; RV32I-FPELIM-NEXT: lw a0, 8(a0)
> +; RV32I-FPELIM-NEXT: lw a2, 8(a7)
> +; RV32I-FPELIM-NEXT: xor a4, a5, a4
> +; RV32I-FPELIM-NEXT: xor a1, a1, a3
> +; RV32I-FPELIM-NEXT: or a1, a1, a4
> +; RV32I-FPELIM-NEXT: xor a0, a2, a0
> +; RV32I-FPELIM-NEXT: xor a2, t0, a6
> +; RV32I-FPELIM-NEXT: or a0, a2, a0
> ; RV32I-FPELIM-NEXT: or a0, a0, a1
> ; RV32I-FPELIM-NEXT: seqz a0, a0
> ; RV32I-FPELIM-NEXT: ret
> @@ -324,20 +324,20 @@ define i32 @callee_large_scalars_exhaust
> ; RV32I-WITHFP-NEXT: sw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: addi s0, sp, 16
> ; RV32I-WITHFP-NEXT: lw a0, 4(s0)
> -; RV32I-WITHFP-NEXT: lw a1, 12(a0)
> -; RV32I-WITHFP-NEXT: lw a2, 12(a7)
> -; RV32I-WITHFP-NEXT: xor a1, a2, a1
> -; RV32I-WITHFP-NEXT: lw a2, 4(a0)
> -; RV32I-WITHFP-NEXT: lw a3, 4(a7)
> -; RV32I-WITHFP-NEXT: xor a2, a3, a2
> -; RV32I-WITHFP-NEXT: or a1, a2, a1
> -; RV32I-WITHFP-NEXT: lw a2, 8(a0)
> -; RV32I-WITHFP-NEXT: lw a3, 8(a7)
> -; RV32I-WITHFP-NEXT: xor a2, a3, a2
> -; RV32I-WITHFP-NEXT: lw a0, 0(a0)
> -; RV32I-WITHFP-NEXT: lw a3, 0(a7)
> -; RV32I-WITHFP-NEXT: xor a0, a3, a0
> -; RV32I-WITHFP-NEXT: or a0, a0, a2
> +; RV32I-WITHFP-NEXT: lw a6, 0(a0)
> +; RV32I-WITHFP-NEXT: lw t0, 0(a7)
> +; RV32I-WITHFP-NEXT: lw a3, 4(a0)
> +; RV32I-WITHFP-NEXT: lw a4, 12(a0)
> +; RV32I-WITHFP-NEXT: lw a5, 12(a7)
> +; RV32I-WITHFP-NEXT: lw a1, 4(a7)
> +; RV32I-WITHFP-NEXT: lw a0, 8(a0)
> +; RV32I-WITHFP-NEXT: lw a2, 8(a7)
> +; RV32I-WITHFP-NEXT: xor a4, a5, a4
> +; RV32I-WITHFP-NEXT: xor a1, a1, a3
> +; RV32I-WITHFP-NEXT: or a1, a1, a4
> +; RV32I-WITHFP-NEXT: xor a0, a2, a0
> +; RV32I-WITHFP-NEXT: xor a2, t0, a6
> +; RV32I-WITHFP-NEXT: or a0, a2, a0
> ; RV32I-WITHFP-NEXT: or a0, a0, a1
> ; RV32I-WITHFP-NEXT: seqz a0, a0
> ; RV32I-WITHFP-NEXT: lw s0, 8(sp)
> @@ -367,8 +367,7 @@ define i32 @caller_large_scalars_exhaust
> ; RV32I-FPELIM-NEXT: sw zero, 52(sp)
> ; RV32I-FPELIM-NEXT: sw zero, 48(sp)
> ; RV32I-FPELIM-NEXT: sw zero, 44(sp)
> -; RV32I-FPELIM-NEXT: addi a0, zero, 8
> -; RV32I-FPELIM-NEXT: sw a0, 40(sp)
> +; RV32I-FPELIM-NEXT: addi t0, zero, 8
> ; RV32I-FPELIM-NEXT: addi a7, sp, 40
> ; RV32I-FPELIM-NEXT: addi a0, zero, 1
> ; RV32I-FPELIM-NEXT: addi a1, zero, 2
> @@ -377,6 +376,7 @@ define i32 @caller_large_scalars_exhaust
> ; RV32I-FPELIM-NEXT: addi a4, zero, 5
> ; RV32I-FPELIM-NEXT: addi a5, zero, 6
> ; RV32I-FPELIM-NEXT: addi a6, zero, 7
> +; RV32I-FPELIM-NEXT: sw t0, 40(sp)
> ; RV32I-FPELIM-NEXT: call callee_large_scalars_exhausted_regs
> ; RV32I-FPELIM-NEXT: lw ra, 60(sp)
> ; RV32I-FPELIM-NEXT: addi sp, sp, 64
> @@ -400,8 +400,7 @@ define i32 @caller_large_scalars_exhaust
> ; RV32I-WITHFP-NEXT: sw zero, -12(s0)
> ; RV32I-WITHFP-NEXT: sw zero, -16(s0)
> ; RV32I-WITHFP-NEXT: sw zero, -20(s0)
> -; RV32I-WITHFP-NEXT: addi a0, zero, 8
> -; RV32I-WITHFP-NEXT: sw a0, -24(s0)
> +; RV32I-WITHFP-NEXT: addi t0, zero, 8
> ; RV32I-WITHFP-NEXT: addi a7, s0, -24
> ; RV32I-WITHFP-NEXT: addi a0, zero, 1
> ; RV32I-WITHFP-NEXT: addi a1, zero, 2
> @@ -410,6 +409,7 @@ define i32 @caller_large_scalars_exhaust
> ; RV32I-WITHFP-NEXT: addi a4, zero, 5
> ; RV32I-WITHFP-NEXT: addi a5, zero, 6
> ; RV32I-WITHFP-NEXT: addi a6, zero, 7
> +; RV32I-WITHFP-NEXT: sw t0, -24(s0)
> ; RV32I-WITHFP-NEXT: call callee_large_scalars_exhausted_regs
> ; RV32I-WITHFP-NEXT: lw s0, 56(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 60(sp)
> @@ -524,9 +524,9 @@ define i32 @caller_small_coerced_struct(
> define i32 @callee_large_struct(%struct.large* byval align 4 %a) nounwind {
> ; RV32I-FPELIM-LABEL: callee_large_struct:
> ; RV32I-FPELIM: # %bb.0:
> -; RV32I-FPELIM-NEXT: lw a1, 12(a0)
> -; RV32I-FPELIM-NEXT: lw a0, 0(a0)
> -; RV32I-FPELIM-NEXT: add a0, a0, a1
> +; RV32I-FPELIM-NEXT: lw a1, 0(a0)
> +; RV32I-FPELIM-NEXT: lw a0, 12(a0)
> +; RV32I-FPELIM-NEXT: add a0, a1, a0
> ; RV32I-FPELIM-NEXT: ret
> ;
> ; RV32I-WITHFP-LABEL: callee_large_struct:
> @@ -535,9 +535,9 @@ define i32 @callee_large_struct(%struct.
> ; RV32I-WITHFP-NEXT: sw ra, 12(sp)
> ; RV32I-WITHFP-NEXT: sw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: addi s0, sp, 16
> -; RV32I-WITHFP-NEXT: lw a1, 12(a0)
> -; RV32I-WITHFP-NEXT: lw a0, 0(a0)
> -; RV32I-WITHFP-NEXT: add a0, a0, a1
> +; RV32I-WITHFP-NEXT: lw a1, 0(a0)
> +; RV32I-WITHFP-NEXT: lw a0, 12(a0)
> +; RV32I-WITHFP-NEXT: add a0, a1, a0
> ; RV32I-WITHFP-NEXT: lw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 12(sp)
> ; RV32I-WITHFP-NEXT: addi sp, sp, 16
> @@ -557,16 +557,16 @@ define i32 @caller_large_struct() nounwi
> ; RV32I-FPELIM-NEXT: sw ra, 44(sp)
> ; RV32I-FPELIM-NEXT: addi a0, zero, 1
> ; RV32I-FPELIM-NEXT: sw a0, 24(sp)
> +; RV32I-FPELIM-NEXT: addi a1, zero, 2
> +; RV32I-FPELIM-NEXT: sw a1, 28(sp)
> +; RV32I-FPELIM-NEXT: addi a2, zero, 3
> +; RV32I-FPELIM-NEXT: sw a2, 32(sp)
> +; RV32I-FPELIM-NEXT: addi a3, zero, 4
> +; RV32I-FPELIM-NEXT: sw a3, 36(sp)
> ; RV32I-FPELIM-NEXT: sw a0, 8(sp)
> -; RV32I-FPELIM-NEXT: addi a0, zero, 2
> -; RV32I-FPELIM-NEXT: sw a0, 28(sp)
> -; RV32I-FPELIM-NEXT: sw a0, 12(sp)
> -; RV32I-FPELIM-NEXT: addi a0, zero, 3
> -; RV32I-FPELIM-NEXT: sw a0, 32(sp)
> -; RV32I-FPELIM-NEXT: sw a0, 16(sp)
> -; RV32I-FPELIM-NEXT: addi a0, zero, 4
> -; RV32I-FPELIM-NEXT: sw a0, 36(sp)
> -; RV32I-FPELIM-NEXT: sw a0, 20(sp)
> +; RV32I-FPELIM-NEXT: sw a1, 12(sp)
> +; RV32I-FPELIM-NEXT: sw a2, 16(sp)
> +; RV32I-FPELIM-NEXT: sw a3, 20(sp)
> ; RV32I-FPELIM-NEXT: addi a0, sp, 8
> ; RV32I-FPELIM-NEXT: call callee_large_struct
> ; RV32I-FPELIM-NEXT: lw ra, 44(sp)
> @@ -581,16 +581,16 @@ define i32 @caller_large_struct() nounwi
> ; RV32I-WITHFP-NEXT: addi s0, sp, 48
> ; RV32I-WITHFP-NEXT: addi a0, zero, 1
> ; RV32I-WITHFP-NEXT: sw a0, -24(s0)
> +; RV32I-WITHFP-NEXT: addi a1, zero, 2
> +; RV32I-WITHFP-NEXT: sw a1, -20(s0)
> +; RV32I-WITHFP-NEXT: addi a2, zero, 3
> +; RV32I-WITHFP-NEXT: sw a2, -16(s0)
> +; RV32I-WITHFP-NEXT: addi a3, zero, 4
> +; RV32I-WITHFP-NEXT: sw a3, -12(s0)
> ; RV32I-WITHFP-NEXT: sw a0, -40(s0)
> -; RV32I-WITHFP-NEXT: addi a0, zero, 2
> -; RV32I-WITHFP-NEXT: sw a0, -20(s0)
> -; RV32I-WITHFP-NEXT: sw a0, -36(s0)
> -; RV32I-WITHFP-NEXT: addi a0, zero, 3
> -; RV32I-WITHFP-NEXT: sw a0, -16(s0)
> -; RV32I-WITHFP-NEXT: sw a0, -32(s0)
> -; RV32I-WITHFP-NEXT: addi a0, zero, 4
> -; RV32I-WITHFP-NEXT: sw a0, -12(s0)
> -; RV32I-WITHFP-NEXT: sw a0, -28(s0)
> +; RV32I-WITHFP-NEXT: sw a1, -36(s0)
> +; RV32I-WITHFP-NEXT: sw a2, -32(s0)
> +; RV32I-WITHFP-NEXT: sw a3, -28(s0)
> ; RV32I-WITHFP-NEXT: addi a0, s0, -40
> ; RV32I-WITHFP-NEXT: call callee_large_struct
> ; RV32I-WITHFP-NEXT: lw s0, 40(sp)
> @@ -619,14 +619,14 @@ define i32 @callee_aligned_stack(i32 %a,
> ; RV32I-FPELIM-LABEL: callee_aligned_stack:
> ; RV32I-FPELIM: # %bb.0:
> ; RV32I-FPELIM-NEXT: lw a0, 0(a2)
> -; RV32I-FPELIM-NEXT: add a0, a0, a7
> -; RV32I-FPELIM-NEXT: lw a1, 0(sp)
> -; RV32I-FPELIM-NEXT: add a0, a0, a1
> -; RV32I-FPELIM-NEXT: lw a1, 8(sp)
> -; RV32I-FPELIM-NEXT: add a0, a0, a1
> -; RV32I-FPELIM-NEXT: lw a1, 16(sp)
> -; RV32I-FPELIM-NEXT: add a0, a0, a1
> ; RV32I-FPELIM-NEXT: lw a1, 20(sp)
> +; RV32I-FPELIM-NEXT: lw a2, 0(sp)
> +; RV32I-FPELIM-NEXT: lw a3, 8(sp)
> +; RV32I-FPELIM-NEXT: lw a4, 16(sp)
> +; RV32I-FPELIM-NEXT: add a0, a0, a7
> +; RV32I-FPELIM-NEXT: add a0, a0, a2
> +; RV32I-FPELIM-NEXT: add a0, a0, a3
> +; RV32I-FPELIM-NEXT: add a0, a0, a4
> ; RV32I-FPELIM-NEXT: add a0, a0, a1
> ; RV32I-FPELIM-NEXT: ret
> ;
> @@ -637,14 +637,14 @@ define i32 @callee_aligned_stack(i32 %a,
> ; RV32I-WITHFP-NEXT: sw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: addi s0, sp, 16
> ; RV32I-WITHFP-NEXT: lw a0, 0(a2)
> -; RV32I-WITHFP-NEXT: add a0, a0, a7
> -; RV32I-WITHFP-NEXT: lw a1, 0(s0)
> -; RV32I-WITHFP-NEXT: add a0, a0, a1
> -; RV32I-WITHFP-NEXT: lw a1, 8(s0)
> -; RV32I-WITHFP-NEXT: add a0, a0, a1
> -; RV32I-WITHFP-NEXT: lw a1, 16(s0)
> -; RV32I-WITHFP-NEXT: add a0, a0, a1
> ; RV32I-WITHFP-NEXT: lw a1, 20(s0)
> +; RV32I-WITHFP-NEXT: lw a2, 0(s0)
> +; RV32I-WITHFP-NEXT: lw a3, 8(s0)
> +; RV32I-WITHFP-NEXT: lw a4, 16(s0)
> +; RV32I-WITHFP-NEXT: add a0, a0, a7
> +; RV32I-WITHFP-NEXT: add a0, a0, a2
> +; RV32I-WITHFP-NEXT: add a0, a0, a3
> +; RV32I-WITHFP-NEXT: add a0, a0, a4
> ; RV32I-WITHFP-NEXT: add a0, a0, a1
> ; RV32I-WITHFP-NEXT: lw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 12(sp)
> @@ -690,8 +690,7 @@ define void @caller_aligned_stack() noun
> ; RV32I-FPELIM-NEXT: addi a0, a0, -328
> ; RV32I-FPELIM-NEXT: sw a0, 36(sp)
> ; RV32I-FPELIM-NEXT: lui a0, 335544
> -; RV32I-FPELIM-NEXT: addi a0, a0, 1311
> -; RV32I-FPELIM-NEXT: sw a0, 32(sp)
> +; RV32I-FPELIM-NEXT: addi t0, a0, 1311
> ; RV32I-FPELIM-NEXT: lui a0, 688509
> ; RV32I-FPELIM-NEXT: addi a5, a0, -2048
> ; RV32I-FPELIM-NEXT: addi a2, sp, 32
> @@ -701,6 +700,7 @@ define void @caller_aligned_stack() noun
> ; RV32I-FPELIM-NEXT: addi a4, zero, 13
> ; RV32I-FPELIM-NEXT: addi a6, zero, 4
> ; RV32I-FPELIM-NEXT: addi a7, zero, 14
> +; RV32I-FPELIM-NEXT: sw t0, 32(sp)
> ; RV32I-FPELIM-NEXT: call callee_aligned_stack
> ; RV32I-FPELIM-NEXT: lw ra, 60(sp)
> ; RV32I-FPELIM-NEXT: addi sp, sp, 64
> @@ -733,8 +733,7 @@ define void @caller_aligned_stack() noun
> ; RV32I-WITHFP-NEXT: addi a0, a0, -328
> ; RV32I-WITHFP-NEXT: sw a0, -28(s0)
> ; RV32I-WITHFP-NEXT: lui a0, 335544
> -; RV32I-WITHFP-NEXT: addi a0, a0, 1311
> -; RV32I-WITHFP-NEXT: sw a0, -32(s0)
> +; RV32I-WITHFP-NEXT: addi t0, a0, 1311
> ; RV32I-WITHFP-NEXT: lui a0, 688509
> ; RV32I-WITHFP-NEXT: addi a5, a0, -2048
> ; RV32I-WITHFP-NEXT: addi a2, s0, -32
> @@ -744,6 +743,7 @@ define void @caller_aligned_stack() noun
> ; RV32I-WITHFP-NEXT: addi a4, zero, 13
> ; RV32I-WITHFP-NEXT: addi a6, zero, 4
> ; RV32I-WITHFP-NEXT: addi a7, zero, 14
> +; RV32I-WITHFP-NEXT: sw t0, -32(s0)
> ; RV32I-WITHFP-NEXT: call callee_aligned_stack
> ; RV32I-WITHFP-NEXT: lw s0, 56(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 60(sp)
> @@ -787,17 +787,15 @@ define i32 @caller_small_scalar_ret() no
> ; RV32I-FPELIM: # %bb.0:
> ; RV32I-FPELIM-NEXT: addi sp, sp, -16
> ; RV32I-FPELIM-NEXT: sw ra, 12(sp)
> -; RV32I-FPELIM-NEXT: sw s0, 8(sp)
> -; RV32I-FPELIM-NEXT: lui a0, 56
> -; RV32I-FPELIM-NEXT: addi s0, a0, 580
> ; RV32I-FPELIM-NEXT: call callee_small_scalar_ret
> -; RV32I-FPELIM-NEXT: xor a1, a1, s0
> +; RV32I-FPELIM-NEXT: lui a2, 56
> +; RV32I-FPELIM-NEXT: addi a2, a2, 580
> +; RV32I-FPELIM-NEXT: xor a1, a1, a2
> ; RV32I-FPELIM-NEXT: lui a2, 200614
> ; RV32I-FPELIM-NEXT: addi a2, a2, 647
> ; RV32I-FPELIM-NEXT: xor a0, a0, a2
> ; RV32I-FPELIM-NEXT: or a0, a0, a1
> ; RV32I-FPELIM-NEXT: seqz a0, a0
> -; RV32I-FPELIM-NEXT: lw s0, 8(sp)
> ; RV32I-FPELIM-NEXT: lw ra, 12(sp)
> ; RV32I-FPELIM-NEXT: addi sp, sp, 16
> ; RV32I-FPELIM-NEXT: ret
> @@ -807,18 +805,16 @@ define i32 @caller_small_scalar_ret() no
> ; RV32I-WITHFP-NEXT: addi sp, sp, -16
> ; RV32I-WITHFP-NEXT: sw ra, 12(sp)
> ; RV32I-WITHFP-NEXT: sw s0, 8(sp)
> -; RV32I-WITHFP-NEXT: sw s1, 4(sp)
> ; RV32I-WITHFP-NEXT: addi s0, sp, 16
> -; RV32I-WITHFP-NEXT: lui a0, 56
> -; RV32I-WITHFP-NEXT: addi s1, a0, 580
> ; RV32I-WITHFP-NEXT: call callee_small_scalar_ret
> -; RV32I-WITHFP-NEXT: xor a1, a1, s1
> +; RV32I-WITHFP-NEXT: lui a2, 56
> +; RV32I-WITHFP-NEXT: addi a2, a2, 580
> +; RV32I-WITHFP-NEXT: xor a1, a1, a2
> ; RV32I-WITHFP-NEXT: lui a2, 200614
> ; RV32I-WITHFP-NEXT: addi a2, a2, 647
> ; RV32I-WITHFP-NEXT: xor a0, a0, a2
> ; RV32I-WITHFP-NEXT: or a0, a0, a1
> ; RV32I-WITHFP-NEXT: seqz a0, a0
> -; RV32I-WITHFP-NEXT: lw s1, 4(sp)
> ; RV32I-WITHFP-NEXT: lw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 12(sp)
> ; RV32I-WITHFP-NEXT: addi sp, sp, 16
> @@ -946,14 +942,14 @@ define void @caller_large_scalar_ret() n
> define void @callee_large_struct_ret(%struct.large* noalias sret %agg.result) nounwind {
> ; RV32I-FPELIM-LABEL: callee_large_struct_ret:
> ; RV32I-FPELIM: # %bb.0:
> -; RV32I-FPELIM-NEXT: addi a1, zero, 4
> -; RV32I-FPELIM-NEXT: sw a1, 12(a0)
> -; RV32I-FPELIM-NEXT: addi a1, zero, 3
> -; RV32I-FPELIM-NEXT: sw a1, 8(a0)
> -; RV32I-FPELIM-NEXT: addi a1, zero, 2
> -; RV32I-FPELIM-NEXT: sw a1, 4(a0)
> ; RV32I-FPELIM-NEXT: addi a1, zero, 1
> ; RV32I-FPELIM-NEXT: sw a1, 0(a0)
> +; RV32I-FPELIM-NEXT: addi a1, zero, 2
> +; RV32I-FPELIM-NEXT: sw a1, 4(a0)
> +; RV32I-FPELIM-NEXT: addi a1, zero, 3
> +; RV32I-FPELIM-NEXT: sw a1, 8(a0)
> +; RV32I-FPELIM-NEXT: addi a1, zero, 4
> +; RV32I-FPELIM-NEXT: sw a1, 12(a0)
> ; RV32I-FPELIM-NEXT: ret
> ;
> ; RV32I-WITHFP-LABEL: callee_large_struct_ret:
> @@ -962,14 +958,14 @@ define void @callee_large_struct_ret(%st
> ; RV32I-WITHFP-NEXT: sw ra, 12(sp)
> ; RV32I-WITHFP-NEXT: sw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: addi s0, sp, 16
> -; RV32I-WITHFP-NEXT: addi a1, zero, 4
> -; RV32I-WITHFP-NEXT: sw a1, 12(a0)
> -; RV32I-WITHFP-NEXT: addi a1, zero, 3
> -; RV32I-WITHFP-NEXT: sw a1, 8(a0)
> -; RV32I-WITHFP-NEXT: addi a1, zero, 2
> -; RV32I-WITHFP-NEXT: sw a1, 4(a0)
> ; RV32I-WITHFP-NEXT: addi a1, zero, 1
> ; RV32I-WITHFP-NEXT: sw a1, 0(a0)
> +; RV32I-WITHFP-NEXT: addi a1, zero, 2
> +; RV32I-WITHFP-NEXT: sw a1, 4(a0)
> +; RV32I-WITHFP-NEXT: addi a1, zero, 3
> +; RV32I-WITHFP-NEXT: sw a1, 8(a0)
> +; RV32I-WITHFP-NEXT: addi a1, zero, 4
> +; RV32I-WITHFP-NEXT: sw a1, 12(a0)
> ; RV32I-WITHFP-NEXT: lw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 12(sp)
> ; RV32I-WITHFP-NEXT: addi sp, sp, 16
> @@ -992,9 +988,9 @@ define i32 @caller_large_struct_ret() no
> ; RV32I-FPELIM-NEXT: sw ra, 28(sp)
> ; RV32I-FPELIM-NEXT: addi a0, sp, 8
> ; RV32I-FPELIM-NEXT: call callee_large_struct_ret
> -; RV32I-FPELIM-NEXT: lw a0, 20(sp)
> -; RV32I-FPELIM-NEXT: lw a1, 8(sp)
> -; RV32I-FPELIM-NEXT: add a0, a1, a0
> +; RV32I-FPELIM-NEXT: lw a0, 8(sp)
> +; RV32I-FPELIM-NEXT: lw a1, 20(sp)
> +; RV32I-FPELIM-NEXT: add a0, a0, a1
> ; RV32I-FPELIM-NEXT: lw ra, 28(sp)
> ; RV32I-FPELIM-NEXT: addi sp, sp, 32
> ; RV32I-FPELIM-NEXT: ret
> @@ -1007,9 +1003,9 @@ define i32 @caller_large_struct_ret() no
> ; RV32I-WITHFP-NEXT: addi s0, sp, 32
> ; RV32I-WITHFP-NEXT: addi a0, s0, -24
> ; RV32I-WITHFP-NEXT: call callee_large_struct_ret
> -; RV32I-WITHFP-NEXT: lw a0, -12(s0)
> -; RV32I-WITHFP-NEXT: lw a1, -24(s0)
> -; RV32I-WITHFP-NEXT: add a0, a1, a0
> +; RV32I-WITHFP-NEXT: lw a0, -24(s0)
> +; RV32I-WITHFP-NEXT: lw a1, -12(s0)
> +; RV32I-WITHFP-NEXT: add a0, a0, a1
> ; RV32I-WITHFP-NEXT: lw s0, 24(sp)
> ; RV32I-WITHFP-NEXT: lw ra, 28(sp)
> ; RV32I-WITHFP-NEXT: addi sp, sp, 32
>
> Modified: llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32.ll Tue Sep 17 02:43:08 2019
> @@ -107,15 +107,15 @@ define i32 @caller_float_on_stack() noun
> ; RV32I-FPELIM: # %bb.0:
> ; RV32I-FPELIM-NEXT: addi sp, sp, -16
> ; RV32I-FPELIM-NEXT: sw ra, 12(sp)
> -; RV32I-FPELIM-NEXT: lui a0, 264704
> -; RV32I-FPELIM-NEXT: sw a0, 0(sp)
> +; RV32I-FPELIM-NEXT: lui a1, 264704
> ; RV32I-FPELIM-NEXT: addi a0, zero, 1
> -; RV32I-FPELIM-NEXT: mv a1, zero
> ; RV32I-FPELIM-NEXT: addi a2, zero, 2
> -; RV32I-FPELIM-NEXT: mv a3, zero
> ; RV32I-FPELIM-NEXT: addi a4, zero, 3
> -; RV32I-FPELIM-NEXT: mv a5, zero
> ; RV32I-FPELIM-NEXT: addi a6, zero, 4
> +; RV32I-FPELIM-NEXT: sw a1, 0(sp)
> +; RV32I-FPELIM-NEXT: mv a1, zero
> +; RV32I-FPELIM-NEXT: mv a3, zero
> +; RV32I-FPELIM-NEXT: mv a5, zero
> ; RV32I-FPELIM-NEXT: mv a7, zero
> ; RV32I-FPELIM-NEXT: call callee_float_on_stack
> ; RV32I-FPELIM-NEXT: lw ra, 12(sp)
> @@ -128,15 +128,15 @@ define i32 @caller_float_on_stack() noun
> ; RV32I-WITHFP-NEXT: sw ra, 12(sp)
> ; RV32I-WITHFP-NEXT: sw s0, 8(sp)
> ; RV32I-WITHFP-NEXT: addi s0, sp, 16
> -; RV32I-WITHFP-NEXT: lui a0, 264704
> -; RV32I-WITHFP-NEXT: sw a0, 0(sp)
> +; RV32I-WITHFP-NEXT: lui a1, 264704
> ; RV32I-WITHFP-NEXT: addi a0, zero, 1
> -; RV32I-WITHFP-NEXT: mv a1, zero
> ; RV32I-WITHFP-NEXT: addi a2, zero, 2
> -; RV32I-WITHFP-NEXT: mv a3, zero
> ; RV32I-WITHFP-NEXT: addi a4, zero, 3
> -; RV32I-WITHFP-NEXT: mv a5, zero
> ; RV32I-WITHFP-NEXT: addi a6, zero, 4
> +; RV32I-WITHFP-NEXT: sw a1, 0(sp)
> +; RV32I-WITHFP-NEXT: mv a1, zero
> +; RV32I-WITHFP-NEXT: mv a3, zero
> +; RV32I-WITHFP-NEXT: mv a5, zero
> ; RV32I-WITHFP-NEXT: mv a7, zero
> ; RV32I-WITHFP-NEXT: call callee_float_on_stack
> ; RV32I-WITHFP-NEXT: lw s0, 8(sp)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32d.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32d.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32d.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32d.ll Tue Sep 17 02:43:08 2019
> @@ -38,9 +38,9 @@ define i32 @caller_double_in_fpr() nounw
> define i32 @callee_double_in_fpr_exhausted_gprs(i64 %a, i64 %b, i64 %c, i64 %d, i32 %e, double %f) nounwind {
> ; RV32-ILP32D-LABEL: callee_double_in_fpr_exhausted_gprs:
> ; RV32-ILP32D: # %bb.0:
> -; RV32-ILP32D-NEXT: fcvt.w.d a0, fa0, rtz
> -; RV32-ILP32D-NEXT: lw a1, 0(sp)
> -; RV32-ILP32D-NEXT: add a0, a1, a0
> +; RV32-ILP32D-NEXT: lw a0, 0(sp)
> +; RV32-ILP32D-NEXT: fcvt.w.d a1, fa0, rtz
> +; RV32-ILP32D-NEXT: add a0, a0, a1
> ; RV32-ILP32D-NEXT: ret
> %f_fptosi = fptosi double %f to i32
> %1 = add i32 %e, %f_fptosi
> @@ -52,18 +52,18 @@ define i32 @caller_double_in_fpr_exhaust
> ; RV32-ILP32D: # %bb.0:
> ; RV32-ILP32D-NEXT: addi sp, sp, -16
> ; RV32-ILP32D-NEXT: sw ra, 12(sp)
> -; RV32-ILP32D-NEXT: addi a0, zero, 5
> -; RV32-ILP32D-NEXT: sw a0, 0(sp)
> +; RV32-ILP32D-NEXT: addi a1, zero, 5
> ; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI3_0)
> ; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI3_0)
> ; RV32-ILP32D-NEXT: fld fa0, 0(a0)
> ; RV32-ILP32D-NEXT: addi a0, zero, 1
> -; RV32-ILP32D-NEXT: mv a1, zero
> ; RV32-ILP32D-NEXT: addi a2, zero, 2
> -; RV32-ILP32D-NEXT: mv a3, zero
> ; RV32-ILP32D-NEXT: addi a4, zero, 3
> -; RV32-ILP32D-NEXT: mv a5, zero
> ; RV32-ILP32D-NEXT: addi a6, zero, 4
> +; RV32-ILP32D-NEXT: sw a1, 0(sp)
> +; RV32-ILP32D-NEXT: mv a1, zero
> +; RV32-ILP32D-NEXT: mv a3, zero
> +; RV32-ILP32D-NEXT: mv a5, zero
> ; RV32-ILP32D-NEXT: mv a7, zero
> ; RV32-ILP32D-NEXT: call callee_double_in_fpr_exhausted_gprs
> ; RV32-ILP32D-NEXT: lw ra, 12(sp)
> @@ -82,9 +82,9 @@ define i32 @callee_double_in_gpr_exhaust
> ; RV32-ILP32D-NEXT: sw a0, 8(sp)
> ; RV32-ILP32D-NEXT: sw a1, 12(sp)
> ; RV32-ILP32D-NEXT: fld ft0, 8(sp)
> -; RV32-ILP32D-NEXT: fcvt.w.d a0, ft0, rtz
> -; RV32-ILP32D-NEXT: fcvt.w.d a1, fa7, rtz
> -; RV32-ILP32D-NEXT: add a0, a1, a0
> +; RV32-ILP32D-NEXT: fcvt.w.d a0, fa7, rtz
> +; RV32-ILP32D-NEXT: fcvt.w.d a1, ft0, rtz
> +; RV32-ILP32D-NEXT: add a0, a0, a1
> ; RV32-ILP32D-NEXT: addi sp, sp, 16
> ; RV32-ILP32D-NEXT: ret
> %h_fptosi = fptosi double %h to i32
> @@ -100,21 +100,21 @@ define i32 @caller_double_in_gpr_exhaust
> ; RV32-ILP32D-NEXT: sw ra, 12(sp)
> ; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_0)
> ; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_0)
> -; RV32-ILP32D-NEXT: lui a1, %hi(.LCPI5_1)
> -; RV32-ILP32D-NEXT: addi a1, a1, %lo(.LCPI5_1)
> -; RV32-ILP32D-NEXT: lui a2, %hi(.LCPI5_2)
> -; RV32-ILP32D-NEXT: addi a2, a2, %lo(.LCPI5_2)
> -; RV32-ILP32D-NEXT: lui a3, %hi(.LCPI5_3)
> -; RV32-ILP32D-NEXT: addi a3, a3, %lo(.LCPI5_3)
> -; RV32-ILP32D-NEXT: lui a4, %hi(.LCPI5_4)
> -; RV32-ILP32D-NEXT: addi a4, a4, %lo(.LCPI5_4)
> -; RV32-ILP32D-NEXT: lui a5, %hi(.LCPI5_5)
> -; RV32-ILP32D-NEXT: addi a5, a5, %lo(.LCPI5_5)
> -; RV32-ILP32D-NEXT: fld fa0, 0(a5)
> -; RV32-ILP32D-NEXT: fld fa1, 0(a4)
> -; RV32-ILP32D-NEXT: fld fa2, 0(a3)
> -; RV32-ILP32D-NEXT: fld fa3, 0(a2)
> -; RV32-ILP32D-NEXT: fld fa4, 0(a1)
> +; RV32-ILP32D-NEXT: fld fa0, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_1)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_1)
> +; RV32-ILP32D-NEXT: fld fa1, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_2)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_2)
> +; RV32-ILP32D-NEXT: fld fa2, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_3)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_3)
> +; RV32-ILP32D-NEXT: fld fa3, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_4)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_4)
> +; RV32-ILP32D-NEXT: fld fa4, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_5)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_5)
> ; RV32-ILP32D-NEXT: fld fa5, 0(a0)
> ; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_6)
> ; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_6)
> @@ -122,8 +122,8 @@ define i32 @caller_double_in_gpr_exhaust
> ; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_7)
> ; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_7)
> ; RV32-ILP32D-NEXT: fld fa7, 0(a0)
> -; RV32-ILP32D-NEXT: mv a0, zero
> ; RV32-ILP32D-NEXT: lui a1, 262688
> +; RV32-ILP32D-NEXT: mv a0, zero
> ; RV32-ILP32D-NEXT: call callee_double_in_gpr_exhausted_fprs
> ; RV32-ILP32D-NEXT: lw ra, 12(sp)
> ; RV32-ILP32D-NEXT: addi sp, sp, 16
> @@ -157,39 +157,39 @@ define i32 @caller_double_in_gpr_and_sta
> ; RV32-ILP32D: # %bb.0:
> ; RV32-ILP32D-NEXT: addi sp, sp, -16
> ; RV32-ILP32D-NEXT: sw ra, 12(sp)
> -; RV32-ILP32D-NEXT: lui a0, 262816
> -; RV32-ILP32D-NEXT: sw a0, 0(sp)
> +; RV32-ILP32D-NEXT: lui a1, 262816
> ; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_0)
> -; RV32-ILP32D-NEXT: addi a6, a0, %lo(.LCPI7_0)
> -; RV32-ILP32D-NEXT: lui a1, %hi(.LCPI7_1)
> -; RV32-ILP32D-NEXT: addi a1, a1, %lo(.LCPI7_1)
> -; RV32-ILP32D-NEXT: lui a2, %hi(.LCPI7_2)
> -; RV32-ILP32D-NEXT: addi a2, a2, %lo(.LCPI7_2)
> -; RV32-ILP32D-NEXT: lui a3, %hi(.LCPI7_3)
> -; RV32-ILP32D-NEXT: addi a3, a3, %lo(.LCPI7_3)
> -; RV32-ILP32D-NEXT: lui a4, %hi(.LCPI7_4)
> -; RV32-ILP32D-NEXT: addi a4, a4, %lo(.LCPI7_4)
> -; RV32-ILP32D-NEXT: lui a5, %hi(.LCPI7_5)
> -; RV32-ILP32D-NEXT: addi a5, a5, %lo(.LCPI7_5)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_0)
> +; RV32-ILP32D-NEXT: fld fa0, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_1)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_1)
> +; RV32-ILP32D-NEXT: fld fa1, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_2)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_2)
> +; RV32-ILP32D-NEXT: fld fa2, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_3)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_3)
> +; RV32-ILP32D-NEXT: fld fa3, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_4)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_4)
> +; RV32-ILP32D-NEXT: fld fa4, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_5)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_5)
> +; RV32-ILP32D-NEXT: fld fa5, 0(a0)
> ; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_6)
> ; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_6)
> -; RV32-ILP32D-NEXT: fld fa0, 0(a0)
> -; RV32-ILP32D-NEXT: fld fa1, 0(a5)
> -; RV32-ILP32D-NEXT: fld fa2, 0(a4)
> -; RV32-ILP32D-NEXT: fld fa3, 0(a3)
> -; RV32-ILP32D-NEXT: fld fa4, 0(a2)
> -; RV32-ILP32D-NEXT: fld fa5, 0(a1)
> -; RV32-ILP32D-NEXT: fld fa6, 0(a6)
> +; RV32-ILP32D-NEXT: fld fa6, 0(a0)
> ; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_7)
> ; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_7)
> ; RV32-ILP32D-NEXT: fld fa7, 0(a0)
> ; RV32-ILP32D-NEXT: addi a0, zero, 1
> -; RV32-ILP32D-NEXT: mv a1, zero
> ; RV32-ILP32D-NEXT: addi a2, zero, 3
> -; RV32-ILP32D-NEXT: mv a3, zero
> ; RV32-ILP32D-NEXT: addi a4, zero, 5
> -; RV32-ILP32D-NEXT: mv a5, zero
> ; RV32-ILP32D-NEXT: addi a6, zero, 7
> +; RV32-ILP32D-NEXT: sw a1, 0(sp)
> +; RV32-ILP32D-NEXT: mv a1, zero
> +; RV32-ILP32D-NEXT: mv a3, zero
> +; RV32-ILP32D-NEXT: mv a5, zero
> ; RV32-ILP32D-NEXT: mv a7, zero
> ; RV32-ILP32D-NEXT: call callee_double_in_gpr_and_stack_almost_exhausted_gprs_fprs
> ; RV32-ILP32D-NEXT: lw ra, 12(sp)
> @@ -223,38 +223,38 @@ define i32 @caller_double_on_stack_exhau
> ; RV32-ILP32D-NEXT: sw ra, 12(sp)
> ; RV32-ILP32D-NEXT: lui a0, 262816
> ; RV32-ILP32D-NEXT: sw a0, 4(sp)
> -; RV32-ILP32D-NEXT: sw zero, 0(sp)
> ; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_0)
> -; RV32-ILP32D-NEXT: addi a6, a0, %lo(.LCPI9_0)
> -; RV32-ILP32D-NEXT: lui a1, %hi(.LCPI9_1)
> -; RV32-ILP32D-NEXT: addi a1, a1, %lo(.LCPI9_1)
> -; RV32-ILP32D-NEXT: lui a2, %hi(.LCPI9_2)
> -; RV32-ILP32D-NEXT: addi a2, a2, %lo(.LCPI9_2)
> -; RV32-ILP32D-NEXT: lui a3, %hi(.LCPI9_3)
> -; RV32-ILP32D-NEXT: addi a3, a3, %lo(.LCPI9_3)
> -; RV32-ILP32D-NEXT: lui a4, %hi(.LCPI9_4)
> -; RV32-ILP32D-NEXT: addi a4, a4, %lo(.LCPI9_4)
> -; RV32-ILP32D-NEXT: lui a5, %hi(.LCPI9_5)
> -; RV32-ILP32D-NEXT: addi a5, a5, %lo(.LCPI9_5)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_0)
> +; RV32-ILP32D-NEXT: fld fa0, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_1)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_1)
> +; RV32-ILP32D-NEXT: fld fa1, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_2)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_2)
> +; RV32-ILP32D-NEXT: fld fa2, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_3)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_3)
> +; RV32-ILP32D-NEXT: fld fa3, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_4)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_4)
> +; RV32-ILP32D-NEXT: fld fa4, 0(a0)
> +; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_5)
> +; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_5)
> +; RV32-ILP32D-NEXT: fld fa5, 0(a0)
> ; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_6)
> ; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_6)
> -; RV32-ILP32D-NEXT: fld fa0, 0(a0)
> -; RV32-ILP32D-NEXT: fld fa1, 0(a5)
> -; RV32-ILP32D-NEXT: fld fa2, 0(a4)
> -; RV32-ILP32D-NEXT: fld fa3, 0(a3)
> -; RV32-ILP32D-NEXT: fld fa4, 0(a2)
> -; RV32-ILP32D-NEXT: fld fa5, 0(a1)
> -; RV32-ILP32D-NEXT: fld fa6, 0(a6)
> +; RV32-ILP32D-NEXT: fld fa6, 0(a0)
> ; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_7)
> ; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_7)
> ; RV32-ILP32D-NEXT: fld fa7, 0(a0)
> ; RV32-ILP32D-NEXT: addi a0, zero, 1
> -; RV32-ILP32D-NEXT: mv a1, zero
> ; RV32-ILP32D-NEXT: addi a2, zero, 3
> -; RV32-ILP32D-NEXT: mv a3, zero
> ; RV32-ILP32D-NEXT: addi a4, zero, 5
> -; RV32-ILP32D-NEXT: mv a5, zero
> ; RV32-ILP32D-NEXT: addi a6, zero, 7
> +; RV32-ILP32D-NEXT: sw zero, 0(sp)
> +; RV32-ILP32D-NEXT: mv a1, zero
> +; RV32-ILP32D-NEXT: mv a3, zero
> +; RV32-ILP32D-NEXT: mv a5, zero
> ; RV32-ILP32D-NEXT: mv a7, zero
> ; RV32-ILP32D-NEXT: call callee_double_on_stack_exhausted_gprs_fprs
> ; RV32-ILP32D-NEXT: lw ra, 12(sp)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32f-ilp32d-common.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32f-ilp32d-common.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32f-ilp32d-common.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/calling-conv-ilp32f-ilp32d-common.ll Tue Sep 17 02:43:08 2019
> @@ -41,9 +41,9 @@ define i32 @caller_float_in_fpr() nounwi
> define i32 @callee_float_in_fpr_exhausted_gprs(i64 %a, i64 %b, i64 %c, i64 %d, i32 %e, float %f) nounwind {
> ; RV32-ILP32FD-LABEL: callee_float_in_fpr_exhausted_gprs:
> ; RV32-ILP32FD: # %bb.0:
> -; RV32-ILP32FD-NEXT: fcvt.w.s a0, fa0, rtz
> -; RV32-ILP32FD-NEXT: lw a1, 0(sp)
> -; RV32-ILP32FD-NEXT: add a0, a1, a0
> +; RV32-ILP32FD-NEXT: lw a0, 0(sp)
> +; RV32-ILP32FD-NEXT: fcvt.w.s a1, fa0, rtz
> +; RV32-ILP32FD-NEXT: add a0, a0, a1
> ; RV32-ILP32FD-NEXT: ret
> %f_fptosi = fptosi float %f to i32
> %1 = add i32 %e, %f_fptosi
> @@ -55,18 +55,18 @@ define i32 @caller_float_in_fpr_exhauste
> ; RV32-ILP32FD: # %bb.0:
> ; RV32-ILP32FD-NEXT: addi sp, sp, -16
> ; RV32-ILP32FD-NEXT: sw ra, 12(sp)
> -; RV32-ILP32FD-NEXT: addi a0, zero, 5
> -; RV32-ILP32FD-NEXT: sw a0, 0(sp)
> +; RV32-ILP32FD-NEXT: addi a1, zero, 5
> ; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI3_0)
> ; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI3_0)
> ; RV32-ILP32FD-NEXT: flw fa0, 0(a0)
> ; RV32-ILP32FD-NEXT: addi a0, zero, 1
> -; RV32-ILP32FD-NEXT: mv a1, zero
> ; RV32-ILP32FD-NEXT: addi a2, zero, 2
> -; RV32-ILP32FD-NEXT: mv a3, zero
> ; RV32-ILP32FD-NEXT: addi a4, zero, 3
> -; RV32-ILP32FD-NEXT: mv a5, zero
> ; RV32-ILP32FD-NEXT: addi a6, zero, 4
> +; RV32-ILP32FD-NEXT: sw a1, 0(sp)
> +; RV32-ILP32FD-NEXT: mv a1, zero
> +; RV32-ILP32FD-NEXT: mv a3, zero
> +; RV32-ILP32FD-NEXT: mv a5, zero
> ; RV32-ILP32FD-NEXT: mv a7, zero
> ; RV32-ILP32FD-NEXT: call callee_float_in_fpr_exhausted_gprs
> ; RV32-ILP32FD-NEXT: lw ra, 12(sp)
> @@ -81,10 +81,10 @@ define i32 @caller_float_in_fpr_exhauste
> define i32 @callee_float_in_gpr_exhausted_fprs(float %a, float %b, float %c, float %d, float %e, float %f, float %g, float %h, float %i) nounwind {
> ; RV32-ILP32FD-LABEL: callee_float_in_gpr_exhausted_fprs:
> ; RV32-ILP32FD: # %bb.0:
> -; RV32-ILP32FD-NEXT: fcvt.w.s a1, fa7, rtz
> ; RV32-ILP32FD-NEXT: fmv.w.x ft0, a0
> -; RV32-ILP32FD-NEXT: fcvt.w.s a0, ft0, rtz
> -; RV32-ILP32FD-NEXT: add a0, a1, a0
> +; RV32-ILP32FD-NEXT: fcvt.w.s a0, fa7, rtz
> +; RV32-ILP32FD-NEXT: fcvt.w.s a1, ft0, rtz
> +; RV32-ILP32FD-NEXT: add a0, a0, a1
> ; RV32-ILP32FD-NEXT: ret
> %h_fptosi = fptosi float %h to i32
> %i_fptosi = fptosi float %i to i32
> @@ -99,21 +99,21 @@ define i32 @caller_float_in_gpr_exhauste
> ; RV32-ILP32FD-NEXT: sw ra, 12(sp)
> ; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_0)
> ; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_0)
> -; RV32-ILP32FD-NEXT: lui a1, %hi(.LCPI5_1)
> -; RV32-ILP32FD-NEXT: addi a1, a1, %lo(.LCPI5_1)
> -; RV32-ILP32FD-NEXT: lui a2, %hi(.LCPI5_2)
> -; RV32-ILP32FD-NEXT: addi a2, a2, %lo(.LCPI5_2)
> -; RV32-ILP32FD-NEXT: lui a3, %hi(.LCPI5_3)
> -; RV32-ILP32FD-NEXT: addi a3, a3, %lo(.LCPI5_3)
> -; RV32-ILP32FD-NEXT: lui a4, %hi(.LCPI5_4)
> -; RV32-ILP32FD-NEXT: addi a4, a4, %lo(.LCPI5_4)
> -; RV32-ILP32FD-NEXT: lui a5, %hi(.LCPI5_5)
> -; RV32-ILP32FD-NEXT: addi a5, a5, %lo(.LCPI5_5)
> -; RV32-ILP32FD-NEXT: flw fa0, 0(a5)
> -; RV32-ILP32FD-NEXT: flw fa1, 0(a4)
> -; RV32-ILP32FD-NEXT: flw fa2, 0(a3)
> -; RV32-ILP32FD-NEXT: flw fa3, 0(a2)
> -; RV32-ILP32FD-NEXT: flw fa4, 0(a1)
> +; RV32-ILP32FD-NEXT: flw fa0, 0(a0)
> +; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_1)
> +; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_1)
> +; RV32-ILP32FD-NEXT: flw fa1, 0(a0)
> +; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_2)
> +; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_2)
> +; RV32-ILP32FD-NEXT: flw fa2, 0(a0)
> +; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_3)
> +; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_3)
> +; RV32-ILP32FD-NEXT: flw fa3, 0(a0)
> +; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_4)
> +; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_4)
> +; RV32-ILP32FD-NEXT: flw fa4, 0(a0)
> +; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_5)
> +; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_5)
> ; RV32-ILP32FD-NEXT: flw fa5, 0(a0)
> ; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_6)
> ; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_6)
> @@ -151,39 +151,39 @@ define i32 @caller_float_on_stack_exhaus
> ; RV32-ILP32FD: # %bb.0:
> ; RV32-ILP32FD-NEXT: addi sp, sp, -16
> ; RV32-ILP32FD-NEXT: sw ra, 12(sp)
> -; RV32-ILP32FD-NEXT: lui a0, 267520
> -; RV32-ILP32FD-NEXT: sw a0, 0(sp)
> +; RV32-ILP32FD-NEXT: lui a1, 267520
> ; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_0)
> -; RV32-ILP32FD-NEXT: addi a6, a0, %lo(.LCPI7_0)
> -; RV32-ILP32FD-NEXT: lui a1, %hi(.LCPI7_1)
> -; RV32-ILP32FD-NEXT: addi a1, a1, %lo(.LCPI7_1)
> -; RV32-ILP32FD-NEXT: lui a2, %hi(.LCPI7_2)
> -; RV32-ILP32FD-NEXT: addi a2, a2, %lo(.LCPI7_2)
> -; RV32-ILP32FD-NEXT: lui a3, %hi(.LCPI7_3)
> -; RV32-ILP32FD-NEXT: addi a3, a3, %lo(.LCPI7_3)
> -; RV32-ILP32FD-NEXT: lui a4, %hi(.LCPI7_4)
> -; RV32-ILP32FD-NEXT: addi a4, a4, %lo(.LCPI7_4)
> -; RV32-ILP32FD-NEXT: lui a5, %hi(.LCPI7_5)
> -; RV32-ILP32FD-NEXT: addi a5, a5, %lo(.LCPI7_5)
> +; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_0)
> +; RV32-ILP32FD-NEXT: flw fa0, 0(a0)
> +; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_1)
> +; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_1)
> +; RV32-ILP32FD-NEXT: flw fa1, 0(a0)
> +; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_2)
> +; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_2)
> +; RV32-ILP32FD-NEXT: flw fa2, 0(a0)
> +; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_3)
> +; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_3)
> +; RV32-ILP32FD-NEXT: flw fa3, 0(a0)
> +; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_4)
> +; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_4)
> +; RV32-ILP32FD-NEXT: flw fa4, 0(a0)
> +; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_5)
> +; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_5)
> +; RV32-ILP32FD-NEXT: flw fa5, 0(a0)
> ; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_6)
> ; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_6)
> -; RV32-ILP32FD-NEXT: flw fa0, 0(a0)
> -; RV32-ILP32FD-NEXT: flw fa1, 0(a5)
> -; RV32-ILP32FD-NEXT: flw fa2, 0(a4)
> -; RV32-ILP32FD-NEXT: flw fa3, 0(a3)
> -; RV32-ILP32FD-NEXT: flw fa4, 0(a2)
> -; RV32-ILP32FD-NEXT: flw fa5, 0(a1)
> -; RV32-ILP32FD-NEXT: flw fa6, 0(a6)
> +; RV32-ILP32FD-NEXT: flw fa6, 0(a0)
> ; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_7)
> ; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_7)
> ; RV32-ILP32FD-NEXT: flw fa7, 0(a0)
> ; RV32-ILP32FD-NEXT: addi a0, zero, 1
> -; RV32-ILP32FD-NEXT: mv a1, zero
> ; RV32-ILP32FD-NEXT: addi a2, zero, 3
> -; RV32-ILP32FD-NEXT: mv a3, zero
> ; RV32-ILP32FD-NEXT: addi a4, zero, 5
> -; RV32-ILP32FD-NEXT: mv a5, zero
> ; RV32-ILP32FD-NEXT: addi a6, zero, 7
> +; RV32-ILP32FD-NEXT: sw a1, 0(sp)
> +; RV32-ILP32FD-NEXT: mv a1, zero
> +; RV32-ILP32FD-NEXT: mv a3, zero
> +; RV32-ILP32FD-NEXT: mv a5, zero
> ; RV32-ILP32FD-NEXT: mv a7, zero
> ; RV32-ILP32FD-NEXT: call callee_float_on_stack_exhausted_gprs_fprs
> ; RV32-ILP32FD-NEXT: lw ra, 12(sp)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/calling-conv-lp64-lp64f-lp64d-common.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/calling-conv-lp64-lp64f-lp64d-common.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/calling-conv-lp64-lp64f-lp64d-common.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/calling-conv-lp64-lp64f-lp64d-common.ll Tue Sep 17 02:43:08 2019
> @@ -48,22 +48,22 @@ define i64 @caller_i128_in_regs() nounwi
> define i32 @callee_many_scalars(i8 %a, i16 %b, i32 %c, i128 %d, i32 %e, i32 %f, i128 %g, i32 %h) nounwind {
> ; RV64I-LABEL: callee_many_scalars:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: ld t0, 0(sp)
> -; RV64I-NEXT: xor a4, a4, t0
> -; RV64I-NEXT: xor a3, a3, a7
> -; RV64I-NEXT: or a3, a3, a4
> -; RV64I-NEXT: lui a4, 16
> -; RV64I-NEXT: addiw a4, a4, -1
> -; RV64I-NEXT: and a1, a1, a4
> -; RV64I-NEXT: andi a0, a0, 255
> -; RV64I-NEXT: add a0, a0, a1
> +; RV64I-NEXT: lw t0, 8(sp)
> +; RV64I-NEXT: ld t1, 0(sp)
> +; RV64I-NEXT: andi t2, a0, 255
> +; RV64I-NEXT: lui a0, 16
> +; RV64I-NEXT: addiw a0, a0, -1
> +; RV64I-NEXT: and a0, a1, a0
> +; RV64I-NEXT: add a0, t2, a0
> ; RV64I-NEXT: add a0, a0, a2
> -; RV64I-NEXT: seqz a1, a3
> +; RV64I-NEXT: xor a1, a4, t1
> +; RV64I-NEXT: xor a2, a3, a7
> +; RV64I-NEXT: or a1, a2, a1
> +; RV64I-NEXT: seqz a1, a1
> ; RV64I-NEXT: add a0, a1, a0
> ; RV64I-NEXT: add a0, a0, a5
> ; RV64I-NEXT: add a0, a0, a6
> -; RV64I-NEXT: lw a1, 8(sp)
> -; RV64I-NEXT: addw a0, a0, a1
> +; RV64I-NEXT: addw a0, a0, t0
> ; RV64I-NEXT: ret
> %a_ext = zext i8 %a to i32
> %b_ext = zext i16 %b to i32
> @@ -85,15 +85,15 @@ define i32 @caller_many_scalars() nounwi
> ; RV64I-NEXT: sd ra, 24(sp)
> ; RV64I-NEXT: addi a0, zero, 8
> ; RV64I-NEXT: sd a0, 8(sp)
> -; RV64I-NEXT: sd zero, 0(sp)
> ; RV64I-NEXT: addi a0, zero, 1
> ; RV64I-NEXT: addi a1, zero, 2
> ; RV64I-NEXT: addi a2, zero, 3
> ; RV64I-NEXT: addi a3, zero, 4
> -; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: addi a5, zero, 5
> ; RV64I-NEXT: addi a6, zero, 6
> ; RV64I-NEXT: addi a7, zero, 7
> +; RV64I-NEXT: sd zero, 0(sp)
> +; RV64I-NEXT: mv a4, zero
> ; RV64I-NEXT: call callee_many_scalars
> ; RV64I-NEXT: ld ra, 24(sp)
> ; RV64I-NEXT: addi sp, sp, 32
> @@ -107,20 +107,20 @@ define i32 @caller_many_scalars() nounwi
> define i64 @callee_large_scalars(i256 %a, i256 %b) nounwind {
> ; RV64I-LABEL: callee_large_scalars:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: ld a2, 24(a1)
> -; RV64I-NEXT: ld a3, 24(a0)
> -; RV64I-NEXT: xor a2, a3, a2
> -; RV64I-NEXT: ld a3, 8(a1)
> -; RV64I-NEXT: ld a4, 8(a0)
> -; RV64I-NEXT: xor a3, a4, a3
> +; RV64I-NEXT: ld a6, 0(a1)
> +; RV64I-NEXT: ld a7, 0(a0)
> +; RV64I-NEXT: ld a4, 8(a1)
> +; RV64I-NEXT: ld a5, 24(a1)
> +; RV64I-NEXT: ld a2, 24(a0)
> +; RV64I-NEXT: ld a3, 8(a0)
> +; RV64I-NEXT: ld a1, 16(a1)
> +; RV64I-NEXT: ld a0, 16(a0)
> +; RV64I-NEXT: xor a2, a2, a5
> +; RV64I-NEXT: xor a3, a3, a4
> ; RV64I-NEXT: or a2, a3, a2
> -; RV64I-NEXT: ld a3, 16(a1)
> -; RV64I-NEXT: ld a4, 16(a0)
> -; RV64I-NEXT: xor a3, a4, a3
> -; RV64I-NEXT: ld a1, 0(a1)
> -; RV64I-NEXT: ld a0, 0(a0)
> ; RV64I-NEXT: xor a0, a0, a1
> -; RV64I-NEXT: or a0, a0, a3
> +; RV64I-NEXT: xor a1, a7, a6
> +; RV64I-NEXT: or a0, a1, a0
> ; RV64I-NEXT: or a0, a0, a2
> ; RV64I-NEXT: seqz a0, a0
> ; RV64I-NEXT: ret
> @@ -134,18 +134,18 @@ define i64 @caller_large_scalars() nounw
> ; RV64I: # %bb.0:
> ; RV64I-NEXT: addi sp, sp, -80
> ; RV64I-NEXT: sd ra, 72(sp)
> -; RV64I-NEXT: addi a0, zero, 2
> -; RV64I-NEXT: sd a0, 0(sp)
> ; RV64I-NEXT: sd zero, 24(sp)
> ; RV64I-NEXT: sd zero, 16(sp)
> ; RV64I-NEXT: sd zero, 8(sp)
> +; RV64I-NEXT: addi a0, zero, 2
> +; RV64I-NEXT: sd a0, 0(sp)
> ; RV64I-NEXT: sd zero, 56(sp)
> ; RV64I-NEXT: sd zero, 48(sp)
> ; RV64I-NEXT: sd zero, 40(sp)
> -; RV64I-NEXT: addi a0, zero, 1
> -; RV64I-NEXT: sd a0, 32(sp)
> +; RV64I-NEXT: addi a2, zero, 1
> ; RV64I-NEXT: addi a0, sp, 32
> ; RV64I-NEXT: mv a1, sp
> +; RV64I-NEXT: sd a2, 32(sp)
> ; RV64I-NEXT: call callee_large_scalars
> ; RV64I-NEXT: ld ra, 72(sp)
> ; RV64I-NEXT: addi sp, sp, 80
> @@ -162,20 +162,20 @@ define i64 @callee_large_scalars_exhaust
> ; RV64I-LABEL: callee_large_scalars_exhausted_regs:
> ; RV64I: # %bb.0:
> ; RV64I-NEXT: ld a0, 8(sp)
> -; RV64I-NEXT: ld a1, 24(a0)
> -; RV64I-NEXT: ld a2, 24(a7)
> -; RV64I-NEXT: xor a1, a2, a1
> -; RV64I-NEXT: ld a2, 8(a0)
> -; RV64I-NEXT: ld a3, 8(a7)
> -; RV64I-NEXT: xor a2, a3, a2
> -; RV64I-NEXT: or a1, a2, a1
> -; RV64I-NEXT: ld a2, 16(a0)
> -; RV64I-NEXT: ld a3, 16(a7)
> -; RV64I-NEXT: xor a2, a3, a2
> -; RV64I-NEXT: ld a0, 0(a0)
> -; RV64I-NEXT: ld a3, 0(a7)
> -; RV64I-NEXT: xor a0, a3, a0
> -; RV64I-NEXT: or a0, a0, a2
> +; RV64I-NEXT: ld a6, 0(a0)
> +; RV64I-NEXT: ld t0, 0(a7)
> +; RV64I-NEXT: ld a3, 8(a0)
> +; RV64I-NEXT: ld a4, 24(a0)
> +; RV64I-NEXT: ld a5, 24(a7)
> +; RV64I-NEXT: ld a1, 8(a7)
> +; RV64I-NEXT: ld a0, 16(a0)
> +; RV64I-NEXT: ld a2, 16(a7)
> +; RV64I-NEXT: xor a4, a5, a4
> +; RV64I-NEXT: xor a1, a1, a3
> +; RV64I-NEXT: or a1, a1, a4
> +; RV64I-NEXT: xor a0, a2, a0
> +; RV64I-NEXT: xor a2, t0, a6
> +; RV64I-NEXT: or a0, a2, a0
> ; RV64I-NEXT: or a0, a0, a1
> ; RV64I-NEXT: seqz a0, a0
> ; RV64I-NEXT: ret
> @@ -193,16 +193,15 @@ define i64 @caller_large_scalars_exhaust
> ; RV64I-NEXT: sd a0, 8(sp)
> ; RV64I-NEXT: addi a0, zero, 9
> ; RV64I-NEXT: sd a0, 0(sp)
> -; RV64I-NEXT: addi a0, zero, 10
> -; RV64I-NEXT: sd a0, 16(sp)
> ; RV64I-NEXT: sd zero, 40(sp)
> ; RV64I-NEXT: sd zero, 32(sp)
> ; RV64I-NEXT: sd zero, 24(sp)
> +; RV64I-NEXT: addi a0, zero, 10
> +; RV64I-NEXT: sd a0, 16(sp)
> ; RV64I-NEXT: sd zero, 72(sp)
> ; RV64I-NEXT: sd zero, 64(sp)
> ; RV64I-NEXT: sd zero, 56(sp)
> -; RV64I-NEXT: addi a0, zero, 8
> -; RV64I-NEXT: sd a0, 48(sp)
> +; RV64I-NEXT: addi t0, zero, 8
> ; RV64I-NEXT: addi a7, sp, 48
> ; RV64I-NEXT: addi a0, zero, 1
> ; RV64I-NEXT: addi a1, zero, 2
> @@ -211,6 +210,7 @@ define i64 @caller_large_scalars_exhaust
> ; RV64I-NEXT: addi a4, zero, 5
> ; RV64I-NEXT: addi a5, zero, 6
> ; RV64I-NEXT: addi a6, zero, 7
> +; RV64I-NEXT: sd t0, 48(sp)
> ; RV64I-NEXT: call callee_large_scalars_exhausted_regs
> ; RV64I-NEXT: ld ra, 88(sp)
> ; RV64I-NEXT: addi sp, sp, 96
> @@ -277,9 +277,9 @@ define i64 @caller_small_coerced_struct(
> define i64 @callee_large_struct(%struct.large* byval align 8 %a) nounwind {
> ; RV64I-LABEL: callee_large_struct:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: ld a1, 24(a0)
> -; RV64I-NEXT: ld a0, 0(a0)
> -; RV64I-NEXT: add a0, a0, a1
> +; RV64I-NEXT: ld a1, 0(a0)
> +; RV64I-NEXT: ld a0, 24(a0)
> +; RV64I-NEXT: add a0, a1, a0
> ; RV64I-NEXT: ret
> %1 = getelementptr inbounds %struct.large, %struct.large* %a, i64 0, i32 0
> %2 = getelementptr inbounds %struct.large, %struct.large* %a, i64 0, i32 3
> @@ -296,16 +296,16 @@ define i64 @caller_large_struct() nounwi
> ; RV64I-NEXT: sd ra, 72(sp)
> ; RV64I-NEXT: addi a0, zero, 1
> ; RV64I-NEXT: sd a0, 40(sp)
> +; RV64I-NEXT: addi a1, zero, 2
> +; RV64I-NEXT: sd a1, 48(sp)
> +; RV64I-NEXT: addi a2, zero, 3
> +; RV64I-NEXT: sd a2, 56(sp)
> +; RV64I-NEXT: addi a3, zero, 4
> +; RV64I-NEXT: sd a3, 64(sp)
> ; RV64I-NEXT: sd a0, 8(sp)
> -; RV64I-NEXT: addi a0, zero, 2
> -; RV64I-NEXT: sd a0, 48(sp)
> -; RV64I-NEXT: sd a0, 16(sp)
> -; RV64I-NEXT: addi a0, zero, 3
> -; RV64I-NEXT: sd a0, 56(sp)
> -; RV64I-NEXT: sd a0, 24(sp)
> -; RV64I-NEXT: addi a0, zero, 4
> -; RV64I-NEXT: sd a0, 64(sp)
> -; RV64I-NEXT: sd a0, 32(sp)
> +; RV64I-NEXT: sd a1, 16(sp)
> +; RV64I-NEXT: sd a2, 24(sp)
> +; RV64I-NEXT: sd a3, 32(sp)
> ; RV64I-NEXT: addi a0, sp, 8
> ; RV64I-NEXT: call callee_large_struct
> ; RV64I-NEXT: ld ra, 72(sp)
> @@ -332,15 +332,15 @@ define i64 @callee_aligned_stack(i64 %a,
> ; should only be 8-byte aligned
> ; RV64I-LABEL: callee_aligned_stack:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: add a0, a5, a7
> +; RV64I-NEXT: ld a0, 40(sp)
> ; RV64I-NEXT: ld a1, 0(sp)
> -; RV64I-NEXT: add a0, a0, a1
> -; RV64I-NEXT: ld a1, 16(sp)
> -; RV64I-NEXT: add a0, a0, a1
> -; RV64I-NEXT: ld a1, 32(sp)
> -; RV64I-NEXT: add a0, a0, a1
> -; RV64I-NEXT: ld a1, 40(sp)
> -; RV64I-NEXT: add a0, a0, a1
> +; RV64I-NEXT: ld a2, 16(sp)
> +; RV64I-NEXT: ld a3, 32(sp)
> +; RV64I-NEXT: add a4, a5, a7
> +; RV64I-NEXT: add a1, a4, a1
> +; RV64I-NEXT: add a1, a1, a2
> +; RV64I-NEXT: add a1, a1, a3
> +; RV64I-NEXT: add a0, a1, a0
> ; RV64I-NEXT: ret
> %f_trunc = trunc i128 %f to i64
> %1 = add i64 %f_trunc, %g
> @@ -366,19 +366,19 @@ define void @caller_aligned_stack() noun
> ; RV64I-NEXT: sd a0, 40(sp)
> ; RV64I-NEXT: addi a0, zero, 10
> ; RV64I-NEXT: sd a0, 32(sp)
> +; RV64I-NEXT: sd zero, 24(sp)
> ; RV64I-NEXT: addi a0, zero, 9
> ; RV64I-NEXT: sd a0, 16(sp)
> -; RV64I-NEXT: addi a0, zero, 8
> -; RV64I-NEXT: sd a0, 0(sp)
> -; RV64I-NEXT: sd zero, 24(sp)
> +; RV64I-NEXT: addi a6, zero, 8
> ; RV64I-NEXT: addi a0, zero, 1
> ; RV64I-NEXT: addi a1, zero, 2
> ; RV64I-NEXT: addi a2, zero, 3
> ; RV64I-NEXT: addi a3, zero, 4
> ; RV64I-NEXT: addi a4, zero, 5
> ; RV64I-NEXT: addi a5, zero, 6
> -; RV64I-NEXT: mv a6, zero
> ; RV64I-NEXT: addi a7, zero, 7
> +; RV64I-NEXT: sd a6, 0(sp)
> +; RV64I-NEXT: mv a6, zero
> ; RV64I-NEXT: call callee_aligned_stack
> ; RV64I-NEXT: ld ra, 56(sp)
> ; RV64I-NEXT: addi sp, sp, 64
> @@ -482,18 +482,18 @@ define void @caller_large_scalar_ret() n
> define void @callee_large_struct_ret(%struct.large* noalias sret %agg.result) nounwind {
> ; RV64I-LABEL: callee_large_struct_ret:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: addi a1, zero, 4
> -; RV64I-NEXT: sw a1, 24(a0)
> -; RV64I-NEXT: addi a1, zero, 3
> -; RV64I-NEXT: sw a1, 16(a0)
> -; RV64I-NEXT: addi a1, zero, 2
> -; RV64I-NEXT: sw a1, 8(a0)
> -; RV64I-NEXT: sw zero, 28(a0)
> -; RV64I-NEXT: sw zero, 20(a0)
> -; RV64I-NEXT: sw zero, 12(a0)
> ; RV64I-NEXT: sw zero, 4(a0)
> ; RV64I-NEXT: addi a1, zero, 1
> ; RV64I-NEXT: sw a1, 0(a0)
> +; RV64I-NEXT: sw zero, 12(a0)
> +; RV64I-NEXT: addi a1, zero, 2
> +; RV64I-NEXT: sw a1, 8(a0)
> +; RV64I-NEXT: sw zero, 20(a0)
> +; RV64I-NEXT: addi a1, zero, 3
> +; RV64I-NEXT: sw a1, 16(a0)
> +; RV64I-NEXT: sw zero, 28(a0)
> +; RV64I-NEXT: addi a1, zero, 4
> +; RV64I-NEXT: sw a1, 24(a0)
> ; RV64I-NEXT: ret
> %a = getelementptr inbounds %struct.large, %struct.large* %agg.result, i64 0, i32 0
> store i64 1, i64* %a, align 4
> @@ -513,9 +513,9 @@ define i64 @caller_large_struct_ret() no
> ; RV64I-NEXT: sd ra, 40(sp)
> ; RV64I-NEXT: addi a0, sp, 8
> ; RV64I-NEXT: call callee_large_struct_ret
> -; RV64I-NEXT: ld a0, 32(sp)
> -; RV64I-NEXT: ld a1, 8(sp)
> -; RV64I-NEXT: add a0, a1, a0
> +; RV64I-NEXT: ld a0, 8(sp)
> +; RV64I-NEXT: ld a1, 32(sp)
> +; RV64I-NEXT: add a0, a0, a1
> ; RV64I-NEXT: ld ra, 40(sp)
> ; RV64I-NEXT: addi sp, sp, 48
> ; RV64I-NEXT: ret
>
> Modified: llvm/trunk/test/CodeGen/RISCV/calling-conv-lp64.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/calling-conv-lp64.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/calling-conv-lp64.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/calling-conv-lp64.ll Tue Sep 17 02:43:08 2019
> @@ -108,15 +108,15 @@ define i64 @caller_float_on_stack() noun
> ; RV64I-FPELIM: # %bb.0:
> ; RV64I-FPELIM-NEXT: addi sp, sp, -16
> ; RV64I-FPELIM-NEXT: sd ra, 8(sp)
> -; RV64I-FPELIM-NEXT: lui a0, 264704
> -; RV64I-FPELIM-NEXT: sd a0, 0(sp)
> +; RV64I-FPELIM-NEXT: lui a1, 264704
> ; RV64I-FPELIM-NEXT: addi a0, zero, 1
> -; RV64I-FPELIM-NEXT: mv a1, zero
> ; RV64I-FPELIM-NEXT: addi a2, zero, 2
> -; RV64I-FPELIM-NEXT: mv a3, zero
> ; RV64I-FPELIM-NEXT: addi a4, zero, 3
> -; RV64I-FPELIM-NEXT: mv a5, zero
> ; RV64I-FPELIM-NEXT: addi a6, zero, 4
> +; RV64I-FPELIM-NEXT: sd a1, 0(sp)
> +; RV64I-FPELIM-NEXT: mv a1, zero
> +; RV64I-FPELIM-NEXT: mv a3, zero
> +; RV64I-FPELIM-NEXT: mv a5, zero
> ; RV64I-FPELIM-NEXT: mv a7, zero
> ; RV64I-FPELIM-NEXT: call callee_float_on_stack
> ; RV64I-FPELIM-NEXT: ld ra, 8(sp)
> @@ -129,15 +129,15 @@ define i64 @caller_float_on_stack() noun
> ; RV64I-WITHFP-NEXT: sd ra, 24(sp)
> ; RV64I-WITHFP-NEXT: sd s0, 16(sp)
> ; RV64I-WITHFP-NEXT: addi s0, sp, 32
> -; RV64I-WITHFP-NEXT: lui a0, 264704
> -; RV64I-WITHFP-NEXT: sd a0, 0(sp)
> +; RV64I-WITHFP-NEXT: lui a1, 264704
> ; RV64I-WITHFP-NEXT: addi a0, zero, 1
> -; RV64I-WITHFP-NEXT: mv a1, zero
> ; RV64I-WITHFP-NEXT: addi a2, zero, 2
> -; RV64I-WITHFP-NEXT: mv a3, zero
> ; RV64I-WITHFP-NEXT: addi a4, zero, 3
> -; RV64I-WITHFP-NEXT: mv a5, zero
> ; RV64I-WITHFP-NEXT: addi a6, zero, 4
> +; RV64I-WITHFP-NEXT: sd a1, 0(sp)
> +; RV64I-WITHFP-NEXT: mv a1, zero
> +; RV64I-WITHFP-NEXT: mv a3, zero
> +; RV64I-WITHFP-NEXT: mv a5, zero
> ; RV64I-WITHFP-NEXT: mv a7, zero
> ; RV64I-WITHFP-NEXT: call callee_float_on_stack
> ; RV64I-WITHFP-NEXT: ld s0, 16(sp)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/calling-conv-rv32f-ilp32.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/calling-conv-rv32f-ilp32.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/calling-conv-rv32f-ilp32.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/calling-conv-rv32f-ilp32.ll Tue Sep 17 02:43:08 2019
> @@ -33,15 +33,15 @@ define float @caller_onstack_f32_noop(fl
> ; RV32IF-NEXT: addi sp, sp, -16
> ; RV32IF-NEXT: sw ra, 12(sp)
> ; RV32IF-NEXT: sw a0, 4(sp)
> -; RV32IF-NEXT: lui a0, 264704
> -; RV32IF-NEXT: sw a0, 0(sp)
> +; RV32IF-NEXT: lui a1, 264704
> ; RV32IF-NEXT: addi a0, zero, 1
> -; RV32IF-NEXT: mv a1, zero
> ; RV32IF-NEXT: addi a2, zero, 2
> -; RV32IF-NEXT: mv a3, zero
> ; RV32IF-NEXT: addi a4, zero, 3
> -; RV32IF-NEXT: mv a5, zero
> ; RV32IF-NEXT: addi a6, zero, 4
> +; RV32IF-NEXT: sw a1, 0(sp)
> +; RV32IF-NEXT: mv a1, zero
> +; RV32IF-NEXT: mv a3, zero
> +; RV32IF-NEXT: mv a5, zero
> ; RV32IF-NEXT: mv a7, zero
> ; RV32IF-NEXT: call onstack_f32_noop
> ; RV32IF-NEXT: lw ra, 12(sp)
> @@ -56,19 +56,19 @@ define float @caller_onstack_f32_fadd(fl
> ; RV32IF: # %bb.0:
> ; RV32IF-NEXT: addi sp, sp, -16
> ; RV32IF-NEXT: sw ra, 12(sp)
> -; RV32IF-NEXT: fmv.w.x ft0, a0
> -; RV32IF-NEXT: fmv.w.x ft1, a1
> -; RV32IF-NEXT: fsub.s ft2, ft1, ft0
> -; RV32IF-NEXT: fsw ft2, 4(sp)
> -; RV32IF-NEXT: fadd.s ft0, ft0, ft1
> -; RV32IF-NEXT: fsw ft0, 0(sp)
> +; RV32IF-NEXT: fmv.w.x ft0, a1
> +; RV32IF-NEXT: fmv.w.x ft1, a0
> +; RV32IF-NEXT: fadd.s ft2, ft1, ft0
> +; RV32IF-NEXT: fsub.s ft0, ft0, ft1
> +; RV32IF-NEXT: fsw ft0, 4(sp)
> ; RV32IF-NEXT: addi a0, zero, 1
> -; RV32IF-NEXT: mv a1, zero
> ; RV32IF-NEXT: addi a2, zero, 2
> -; RV32IF-NEXT: mv a3, zero
> ; RV32IF-NEXT: addi a4, zero, 3
> -; RV32IF-NEXT: mv a5, zero
> ; RV32IF-NEXT: addi a6, zero, 4
> +; RV32IF-NEXT: fsw ft2, 0(sp)
> +; RV32IF-NEXT: mv a1, zero
> +; RV32IF-NEXT: mv a3, zero
> +; RV32IF-NEXT: mv a5, zero
> ; RV32IF-NEXT: mv a7, zero
> ; RV32IF-NEXT: call onstack_f32_noop
> ; RV32IF-NEXT: lw ra, 12(sp)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/codemodel-lowering.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/codemodel-lowering.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/codemodel-lowering.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/codemodel-lowering.ll Tue Sep 17 02:43:08 2019
> @@ -61,9 +61,9 @@ define signext i32 @lower_blockaddress_d
> ; RV32I-SMALL-NEXT: sw ra, 12(sp)
> ; RV32I-SMALL-NEXT: lui a1, %hi(.Ltmp0)
> ; RV32I-SMALL-NEXT: addi a1, a1, %lo(.Ltmp0)
> +; RV32I-SMALL-NEXT: addi a2, zero, 101
> ; RV32I-SMALL-NEXT: sw a1, 8(sp)
> -; RV32I-SMALL-NEXT: addi a1, zero, 101
> -; RV32I-SMALL-NEXT: blt a0, a1, .LBB2_3
> +; RV32I-SMALL-NEXT: blt a0, a2, .LBB2_3
> ; RV32I-SMALL-NEXT: # %bb.1: # %if.then
> ; RV32I-SMALL-NEXT: lw a0, 8(sp)
> ; RV32I-SMALL-NEXT: jr a0
> @@ -86,9 +86,9 @@ define signext i32 @lower_blockaddress_d
> ; RV32I-MEDIUM-NEXT: # Label of block must be emitted
> ; RV32I-MEDIUM-NEXT: auipc a1, %pcrel_hi(.Ltmp0)
> ; RV32I-MEDIUM-NEXT: addi a1, a1, %pcrel_lo(.LBB2_5)
> +; RV32I-MEDIUM-NEXT: addi a2, zero, 101
> ; RV32I-MEDIUM-NEXT: sw a1, 8(sp)
> -; RV32I-MEDIUM-NEXT: addi a1, zero, 101
> -; RV32I-MEDIUM-NEXT: blt a0, a1, .LBB2_3
> +; RV32I-MEDIUM-NEXT: blt a0, a2, .LBB2_3
> ; RV32I-MEDIUM-NEXT: # %bb.1: # %if.then
> ; RV32I-MEDIUM-NEXT: lw a0, 8(sp)
> ; RV32I-MEDIUM-NEXT: jr a0
> @@ -131,11 +131,11 @@ indirectgoto:
> define float @lower_constantpool(float %a) nounwind {
> ; RV32I-SMALL-LABEL: lower_constantpool:
> ; RV32I-SMALL: # %bb.0:
> -; RV32I-SMALL-NEXT: fmv.w.x ft0, a0
> -; RV32I-SMALL-NEXT: lui a0, %hi(.LCPI3_0)
> -; RV32I-SMALL-NEXT: addi a0, a0, %lo(.LCPI3_0)
> -; RV32I-SMALL-NEXT: flw ft1, 0(a0)
> -; RV32I-SMALL-NEXT: fadd.s ft0, ft0, ft1
> +; RV32I-SMALL-NEXT: lui a1, %hi(.LCPI3_0)
> +; RV32I-SMALL-NEXT: addi a1, a1, %lo(.LCPI3_0)
> +; RV32I-SMALL-NEXT: flw ft0, 0(a1)
> +; RV32I-SMALL-NEXT: fmv.w.x ft1, a0
> +; RV32I-SMALL-NEXT: fadd.s ft0, ft1, ft0
> ; RV32I-SMALL-NEXT: fmv.x.w a0, ft0
> ; RV32I-SMALL-NEXT: ret
> ;
>
> Modified: llvm/trunk/test/CodeGen/RISCV/compress.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/compress.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/compress.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/compress.ll Tue Sep 17 02:43:08 2019
> @@ -19,10 +19,10 @@
>
> define i32 @simple_arith(i32 %a, i32 %b) nounwind {
> ; RV32IC-LABEL: simple_arith:
> -; RV32IC: c.srai a1, 9
> -; RV32IC-NEXT: addi a2, a0, 1
> +; RV32IC: addi a2, a0, 1
> ; RV32IC-NEXT: c.andi a2, 11
> ; RV32IC-NEXT: c.slli a2, 7
> +; RV32IC-NEXT: c.srai a1, 9
> ; RV32IC-NEXT: c.add a1, a2
> ; RV32IC-NEXT: sub a0, a1, a0
> ; RV32IC-NEXT: c.jr ra
>
> Modified: llvm/trunk/test/CodeGen/RISCV/div.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/div.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/div.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/div.ll Tue Sep 17 02:43:08 2019
> @@ -457,8 +457,8 @@ define i64 @sdiv64_sext_operands(i32 %a,
> ;
> ; RV64IM-LABEL: sdiv64_sext_operands:
> ; RV64IM: # %bb.0:
> -; RV64IM-NEXT: sext.w a1, a1
> ; RV64IM-NEXT: sext.w a0, a0
> +; RV64IM-NEXT: sext.w a1, a1
> ; RV64IM-NEXT: div a0, a0, a1
> ; RV64IM-NEXT: ret
> %1 = sext i32 %a to i64
>
> Modified: llvm/trunk/test/CodeGen/RISCV/double-arith.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/double-arith.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/double-arith.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/double-arith.ll Tue Sep 17 02:43:08 2019
> @@ -473,13 +473,13 @@ define double @fmsub_d(double %a, double
> ;
> ; RV64IFD-LABEL: fmsub_d:
> ; RV64IFD: # %bb.0:
> -; RV64IFD-NEXT: fmv.d.x ft0, a2
> -; RV64IFD-NEXT: lui a2, %hi(.LCPI15_0)
> -; RV64IFD-NEXT: addi a2, a2, %lo(.LCPI15_0)
> -; RV64IFD-NEXT: fld ft1, 0(a2)
> -; RV64IFD-NEXT: fadd.d ft0, ft0, ft1
> +; RV64IFD-NEXT: lui a3, %hi(.LCPI15_0)
> +; RV64IFD-NEXT: addi a3, a3, %lo(.LCPI15_0)
> +; RV64IFD-NEXT: fld ft0, 0(a3)
> ; RV64IFD-NEXT: fmv.d.x ft1, a1
> ; RV64IFD-NEXT: fmv.d.x ft2, a0
> +; RV64IFD-NEXT: fmv.d.x ft3, a2
> +; RV64IFD-NEXT: fadd.d ft0, ft3, ft0
> ; RV64IFD-NEXT: fmsub.d ft0, ft2, ft1, ft0
> ; RV64IFD-NEXT: fmv.x.d a0, ft0
> ; RV64IFD-NEXT: ret
> @@ -496,18 +496,18 @@ define double @fnmadd_d(double %a, doubl
> ; RV32IFD-NEXT: sw a2, 8(sp)
> ; RV32IFD-NEXT: sw a3, 12(sp)
> ; RV32IFD-NEXT: fld ft0, 8(sp)
> -; RV32IFD-NEXT: sw a0, 8(sp)
> -; RV32IFD-NEXT: sw a1, 12(sp)
> -; RV32IFD-NEXT: fld ft1, 8(sp)
> ; RV32IFD-NEXT: sw a4, 8(sp)
> ; RV32IFD-NEXT: sw a5, 12(sp)
> +; RV32IFD-NEXT: fld ft1, 8(sp)
> +; RV32IFD-NEXT: sw a0, 8(sp)
> +; RV32IFD-NEXT: sw a1, 12(sp)
> ; RV32IFD-NEXT: fld ft2, 8(sp)
> ; RV32IFD-NEXT: lui a0, %hi(.LCPI16_0)
> ; RV32IFD-NEXT: addi a0, a0, %lo(.LCPI16_0)
> ; RV32IFD-NEXT: fld ft3, 0(a0)
> ; RV32IFD-NEXT: fadd.d ft2, ft2, ft3
> ; RV32IFD-NEXT: fadd.d ft1, ft1, ft3
> -; RV32IFD-NEXT: fnmadd.d ft0, ft1, ft0, ft2
> +; RV32IFD-NEXT: fnmadd.d ft0, ft2, ft0, ft1
> ; RV32IFD-NEXT: fsd ft0, 8(sp)
> ; RV32IFD-NEXT: lw a0, 8(sp)
> ; RV32IFD-NEXT: lw a1, 12(sp)
> @@ -516,15 +516,15 @@ define double @fnmadd_d(double %a, doubl
> ;
> ; RV64IFD-LABEL: fnmadd_d:
> ; RV64IFD: # %bb.0:
> -; RV64IFD-NEXT: fmv.d.x ft0, a2
> -; RV64IFD-NEXT: lui a2, %hi(.LCPI16_0)
> -; RV64IFD-NEXT: addi a2, a2, %lo(.LCPI16_0)
> -; RV64IFD-NEXT: fld ft1, 0(a2)
> -; RV64IFD-NEXT: fadd.d ft0, ft0, ft1
> -; RV64IFD-NEXT: fmv.d.x ft2, a0
> -; RV64IFD-NEXT: fadd.d ft1, ft2, ft1
> -; RV64IFD-NEXT: fmv.d.x ft2, a1
> -; RV64IFD-NEXT: fnmadd.d ft0, ft1, ft2, ft0
> +; RV64IFD-NEXT: lui a3, %hi(.LCPI16_0)
> +; RV64IFD-NEXT: addi a3, a3, %lo(.LCPI16_0)
> +; RV64IFD-NEXT: fld ft0, 0(a3)
> +; RV64IFD-NEXT: fmv.d.x ft1, a1
> +; RV64IFD-NEXT: fmv.d.x ft2, a2
> +; RV64IFD-NEXT: fmv.d.x ft3, a0
> +; RV64IFD-NEXT: fadd.d ft3, ft3, ft0
> +; RV64IFD-NEXT: fadd.d ft0, ft2, ft0
> +; RV64IFD-NEXT: fnmadd.d ft0, ft3, ft1, ft0
> ; RV64IFD-NEXT: fmv.x.d a0, ft0
> ; RV64IFD-NEXT: ret
> %a_ = fadd double 0.0, %a
> @@ -561,13 +561,13 @@ define double @fnmsub_d(double %a, doubl
> ;
> ; RV64IFD-LABEL: fnmsub_d:
> ; RV64IFD: # %bb.0:
> -; RV64IFD-NEXT: fmv.d.x ft0, a0
> -; RV64IFD-NEXT: lui a0, %hi(.LCPI17_0)
> -; RV64IFD-NEXT: addi a0, a0, %lo(.LCPI17_0)
> -; RV64IFD-NEXT: fld ft1, 0(a0)
> -; RV64IFD-NEXT: fadd.d ft0, ft0, ft1
> +; RV64IFD-NEXT: lui a3, %hi(.LCPI17_0)
> +; RV64IFD-NEXT: addi a3, a3, %lo(.LCPI17_0)
> +; RV64IFD-NEXT: fld ft0, 0(a3)
> ; RV64IFD-NEXT: fmv.d.x ft1, a2
> ; RV64IFD-NEXT: fmv.d.x ft2, a1
> +; RV64IFD-NEXT: fmv.d.x ft3, a0
> +; RV64IFD-NEXT: fadd.d ft0, ft3, ft0
> ; RV64IFD-NEXT: fnmsub.d ft0, ft0, ft2, ft1
> ; RV64IFD-NEXT: fmv.x.d a0, ft0
> ; RV64IFD-NEXT: ret
>
> Modified: llvm/trunk/test/CodeGen/RISCV/double-bitmanip-dagcombines.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/double-bitmanip-dagcombines.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/double-bitmanip-dagcombines.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/double-bitmanip-dagcombines.ll Tue Sep 17 02:43:08 2019
> @@ -118,9 +118,9 @@ define double @fcopysign_fneg(double %a,
> ;
> ; RV64I-LABEL: fcopysign_fneg:
> ; RV64I: # %bb.0:
> +; RV64I-NEXT: not a1, a1
> ; RV64I-NEXT: addi a2, zero, -1
> ; RV64I-NEXT: slli a2, a2, 63
> -; RV64I-NEXT: not a1, a1
> ; RV64I-NEXT: and a1, a1, a2
> ; RV64I-NEXT: addi a2, a2, -1
> ; RV64I-NEXT: and a0, a0, a2
>
> Modified: llvm/trunk/test/CodeGen/RISCV/double-br-fcmp.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/double-br-fcmp.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/double-br-fcmp.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/double-br-fcmp.ll Tue Sep 17 02:43:08 2019
> @@ -385,11 +385,11 @@ define void @br_fcmp_ord(double %a, doub
> ; RV64IFD: # %bb.0:
> ; RV64IFD-NEXT: addi sp, sp, -16
> ; RV64IFD-NEXT: sd ra, 8(sp)
> -; RV64IFD-NEXT: fmv.d.x ft0, a1
> -; RV64IFD-NEXT: feq.d a1, ft0, ft0
> ; RV64IFD-NEXT: fmv.d.x ft0, a0
> -; RV64IFD-NEXT: feq.d a0, ft0, ft0
> -; RV64IFD-NEXT: and a0, a0, a1
> +; RV64IFD-NEXT: fmv.d.x ft1, a1
> +; RV64IFD-NEXT: feq.d a0, ft1, ft1
> +; RV64IFD-NEXT: feq.d a1, ft0, ft0
> +; RV64IFD-NEXT: and a0, a1, a0
> ; RV64IFD-NEXT: bnez a0, .LBB8_2
> ; RV64IFD-NEXT: # %bb.1: # %if.else
> ; RV64IFD-NEXT: ld ra, 8(sp)
> @@ -712,11 +712,11 @@ define void @br_fcmp_uno(double %a, doub
> ; RV64IFD: # %bb.0:
> ; RV64IFD-NEXT: addi sp, sp, -16
> ; RV64IFD-NEXT: sd ra, 8(sp)
> -; RV64IFD-NEXT: fmv.d.x ft0, a1
> -; RV64IFD-NEXT: feq.d a1, ft0, ft0
> ; RV64IFD-NEXT: fmv.d.x ft0, a0
> -; RV64IFD-NEXT: feq.d a0, ft0, ft0
> -; RV64IFD-NEXT: and a0, a0, a1
> +; RV64IFD-NEXT: fmv.d.x ft1, a1
> +; RV64IFD-NEXT: feq.d a0, ft1, ft1
> +; RV64IFD-NEXT: feq.d a1, ft0, ft0
> +; RV64IFD-NEXT: and a0, a1, a0
> ; RV64IFD-NEXT: seqz a0, a0
> ; RV64IFD-NEXT: bnez a0, .LBB15_2
> ; RV64IFD-NEXT: # %bb.1: # %if.else
>
> Modified: llvm/trunk/test/CodeGen/RISCV/double-calling-conv.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/double-calling-conv.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/double-calling-conv.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/double-calling-conv.ll Tue Sep 17 02:43:08 2019
> @@ -76,16 +76,16 @@ define double @caller_double_split_reg_s
> ; RV32IFD-NEXT: addi sp, sp, -16
> ; RV32IFD-NEXT: sw ra, 12(sp)
> ; RV32IFD-NEXT: lui a0, 262510
> -; RV32IFD-NEXT: addi a0, a0, 327
> -; RV32IFD-NEXT: sw a0, 0(sp)
> +; RV32IFD-NEXT: addi a2, a0, 327
> ; RV32IFD-NEXT: lui a0, 262446
> ; RV32IFD-NEXT: addi a6, a0, 327
> ; RV32IFD-NEXT: lui a0, 713032
> ; RV32IFD-NEXT: addi a5, a0, -1311
> ; RV32IFD-NEXT: addi a0, zero, 1
> ; RV32IFD-NEXT: addi a1, zero, 2
> -; RV32IFD-NEXT: mv a2, zero
> ; RV32IFD-NEXT: addi a3, zero, 3
> +; RV32IFD-NEXT: sw a2, 0(sp)
> +; RV32IFD-NEXT: mv a2, zero
> ; RV32IFD-NEXT: mv a4, zero
> ; RV32IFD-NEXT: mv a7, a5
> ; RV32IFD-NEXT: call callee_double_split_reg_stack
> @@ -120,20 +120,20 @@ define double @caller_double_stack() nou
> ; RV32IFD-NEXT: lui a0, 262510
> ; RV32IFD-NEXT: addi a0, a0, 327
> ; RV32IFD-NEXT: sw a0, 4(sp)
> +; RV32IFD-NEXT: lui a0, 713032
> +; RV32IFD-NEXT: addi a1, a0, -1311
> +; RV32IFD-NEXT: sw a1, 0(sp)
> ; RV32IFD-NEXT: lui a0, 262574
> ; RV32IFD-NEXT: addi a0, a0, 327
> ; RV32IFD-NEXT: sw a0, 12(sp)
> -; RV32IFD-NEXT: lui a0, 713032
> -; RV32IFD-NEXT: addi a0, a0, -1311
> -; RV32IFD-NEXT: sw a0, 0(sp)
> -; RV32IFD-NEXT: sw a0, 8(sp)
> ; RV32IFD-NEXT: addi a0, zero, 1
> -; RV32IFD-NEXT: mv a1, zero
> ; RV32IFD-NEXT: addi a2, zero, 2
> -; RV32IFD-NEXT: mv a3, zero
> ; RV32IFD-NEXT: addi a4, zero, 3
> -; RV32IFD-NEXT: mv a5, zero
> ; RV32IFD-NEXT: addi a6, zero, 4
> +; RV32IFD-NEXT: sw a1, 8(sp)
> +; RV32IFD-NEXT: mv a1, zero
> +; RV32IFD-NEXT: mv a3, zero
> +; RV32IFD-NEXT: mv a5, zero
> ; RV32IFD-NEXT: mv a7, zero
> ; RV32IFD-NEXT: call callee_double_stack
> ; RV32IFD-NEXT: lw ra, 28(sp)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/double-convert.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/double-convert.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/double-convert.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/double-convert.ll Tue Sep 17 02:43:08 2019
> @@ -257,9 +257,9 @@ define double @fmv_d_x(i64 %a, i64 %b) n
> ;
> ; RV64IFD-LABEL: fmv_d_x:
> ; RV64IFD: # %bb.0:
> -; RV64IFD-NEXT: fmv.d.x ft0, a1
> -; RV64IFD-NEXT: fmv.d.x ft1, a0
> -; RV64IFD-NEXT: fadd.d ft0, ft1, ft0
> +; RV64IFD-NEXT: fmv.d.x ft0, a0
> +; RV64IFD-NEXT: fmv.d.x ft1, a1
> +; RV64IFD-NEXT: fadd.d ft0, ft0, ft1
> ; RV64IFD-NEXT: fmv.x.d a0, ft0
> ; RV64IFD-NEXT: ret
> %1 = bitcast i64 %a to double
>
> Modified: llvm/trunk/test/CodeGen/RISCV/double-fcmp.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/double-fcmp.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/double-fcmp.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/double-fcmp.ll Tue Sep 17 02:43:08 2019
> @@ -197,11 +197,11 @@ define i32 @fcmp_ord(double %a, double %
> ;
> ; RV64IFD-LABEL: fcmp_ord:
> ; RV64IFD: # %bb.0:
> -; RV64IFD-NEXT: fmv.d.x ft0, a1
> -; RV64IFD-NEXT: feq.d a1, ft0, ft0
> ; RV64IFD-NEXT: fmv.d.x ft0, a0
> -; RV64IFD-NEXT: feq.d a0, ft0, ft0
> -; RV64IFD-NEXT: and a0, a0, a1
> +; RV64IFD-NEXT: fmv.d.x ft1, a1
> +; RV64IFD-NEXT: feq.d a0, ft1, ft1
> +; RV64IFD-NEXT: feq.d a1, ft0, ft0
> +; RV64IFD-NEXT: and a0, a1, a0
> ; RV64IFD-NEXT: ret
> %1 = fcmp ord double %a, %b
> %2 = zext i1 %1 to i32
> @@ -397,11 +397,11 @@ define i32 @fcmp_uno(double %a, double %
> ;
> ; RV64IFD-LABEL: fcmp_uno:
> ; RV64IFD: # %bb.0:
> -; RV64IFD-NEXT: fmv.d.x ft0, a1
> -; RV64IFD-NEXT: feq.d a1, ft0, ft0
> ; RV64IFD-NEXT: fmv.d.x ft0, a0
> -; RV64IFD-NEXT: feq.d a0, ft0, ft0
> -; RV64IFD-NEXT: and a0, a0, a1
> +; RV64IFD-NEXT: fmv.d.x ft1, a1
> +; RV64IFD-NEXT: feq.d a0, ft1, ft1
> +; RV64IFD-NEXT: feq.d a1, ft0, ft0
> +; RV64IFD-NEXT: and a0, a1, a0
> ; RV64IFD-NEXT: seqz a0, a0
> ; RV64IFD-NEXT: ret
> %1 = fcmp uno double %a, %b
>
> Modified: llvm/trunk/test/CodeGen/RISCV/double-imm.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/double-imm.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/double-imm.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/double-imm.ll Tue Sep 17 02:43:08 2019
> @@ -49,11 +49,11 @@ define double @double_imm_op(double %a)
> ;
> ; RV64IFD-LABEL: double_imm_op:
> ; RV64IFD: # %bb.0:
> -; RV64IFD-NEXT: fmv.d.x ft0, a0
> -; RV64IFD-NEXT: lui a0, %hi(.LCPI1_0)
> -; RV64IFD-NEXT: addi a0, a0, %lo(.LCPI1_0)
> -; RV64IFD-NEXT: fld ft1, 0(a0)
> -; RV64IFD-NEXT: fadd.d ft0, ft0, ft1
> +; RV64IFD-NEXT: lui a1, %hi(.LCPI1_0)
> +; RV64IFD-NEXT: addi a1, a1, %lo(.LCPI1_0)
> +; RV64IFD-NEXT: fld ft0, 0(a1)
> +; RV64IFD-NEXT: fmv.d.x ft1, a0
> +; RV64IFD-NEXT: fadd.d ft0, ft1, ft0
> ; RV64IFD-NEXT: fmv.x.d a0, ft0
> ; RV64IFD-NEXT: ret
> %1 = fadd double %a, 1.0
>
> Modified: llvm/trunk/test/CodeGen/RISCV/double-intrinsics.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/double-intrinsics.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/double-intrinsics.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/double-intrinsics.ll Tue Sep 17 02:43:08 2019
> @@ -111,28 +111,24 @@ define double @sincos_f64(double %a) nou
> ; RV32IFD-NEXT: sw ra, 28(sp)
> ; RV32IFD-NEXT: sw s0, 24(sp)
> ; RV32IFD-NEXT: sw s1, 20(sp)
> -; RV32IFD-NEXT: sw s2, 16(sp)
> -; RV32IFD-NEXT: sw s3, 12(sp)
> ; RV32IFD-NEXT: mv s0, a1
> ; RV32IFD-NEXT: mv s1, a0
> ; RV32IFD-NEXT: call sin
> -; RV32IFD-NEXT: mv s2, a0
> -; RV32IFD-NEXT: mv s3, a1
> +; RV32IFD-NEXT: sw a0, 8(sp)
> +; RV32IFD-NEXT: sw a1, 12(sp)
> +; RV32IFD-NEXT: fld ft0, 8(sp)
> +; RV32IFD-NEXT: fsd ft0, 0(sp)
> ; RV32IFD-NEXT: mv a0, s1
> ; RV32IFD-NEXT: mv a1, s0
> ; RV32IFD-NEXT: call cos
> -; RV32IFD-NEXT: sw a0, 0(sp)
> -; RV32IFD-NEXT: sw a1, 4(sp)
> -; RV32IFD-NEXT: fld ft0, 0(sp)
> -; RV32IFD-NEXT: sw s2, 0(sp)
> -; RV32IFD-NEXT: sw s3, 4(sp)
> +; RV32IFD-NEXT: sw a0, 8(sp)
> +; RV32IFD-NEXT: sw a1, 12(sp)
> +; RV32IFD-NEXT: fld ft0, 8(sp)
> ; RV32IFD-NEXT: fld ft1, 0(sp)
> ; RV32IFD-NEXT: fadd.d ft0, ft1, ft0
> -; RV32IFD-NEXT: fsd ft0, 0(sp)
> -; RV32IFD-NEXT: lw a0, 0(sp)
> -; RV32IFD-NEXT: lw a1, 4(sp)
> -; RV32IFD-NEXT: lw s3, 12(sp)
> -; RV32IFD-NEXT: lw s2, 16(sp)
> +; RV32IFD-NEXT: fsd ft0, 8(sp)
> +; RV32IFD-NEXT: lw a0, 8(sp)
> +; RV32IFD-NEXT: lw a1, 12(sp)
> ; RV32IFD-NEXT: lw s1, 20(sp)
> ; RV32IFD-NEXT: lw s0, 24(sp)
> ; RV32IFD-NEXT: lw ra, 28(sp)
> @@ -144,17 +140,16 @@ define double @sincos_f64(double %a) nou
> ; RV64IFD-NEXT: addi sp, sp, -32
> ; RV64IFD-NEXT: sd ra, 24(sp)
> ; RV64IFD-NEXT: sd s0, 16(sp)
> -; RV64IFD-NEXT: sd s1, 8(sp)
> ; RV64IFD-NEXT: mv s0, a0
> ; RV64IFD-NEXT: call sin
> -; RV64IFD-NEXT: mv s1, a0
> +; RV64IFD-NEXT: fmv.d.x ft0, a0
> +; RV64IFD-NEXT: fsd ft0, 8(sp)
> ; RV64IFD-NEXT: mv a0, s0
> ; RV64IFD-NEXT: call cos
> ; RV64IFD-NEXT: fmv.d.x ft0, a0
> -; RV64IFD-NEXT: fmv.d.x ft1, s1
> +; RV64IFD-NEXT: fld ft1, 8(sp)
> ; RV64IFD-NEXT: fadd.d ft0, ft1, ft0
> ; RV64IFD-NEXT: fmv.x.d a0, ft0
> -; RV64IFD-NEXT: ld s1, 8(sp)
> ; RV64IFD-NEXT: ld s0, 16(sp)
> ; RV64IFD-NEXT: ld ra, 24(sp)
> ; RV64IFD-NEXT: addi sp, sp, 32
> @@ -350,17 +345,17 @@ define double @fmuladd_f64(double %a, do
> ; RV32IFD-LABEL: fmuladd_f64:
> ; RV32IFD: # %bb.0:
> ; RV32IFD-NEXT: addi sp, sp, -16
> +; RV32IFD-NEXT: sw a4, 8(sp)
> +; RV32IFD-NEXT: sw a5, 12(sp)
> +; RV32IFD-NEXT: fld ft0, 8(sp)
> ; RV32IFD-NEXT: sw a2, 8(sp)
> ; RV32IFD-NEXT: sw a3, 12(sp)
> -; RV32IFD-NEXT: fld ft0, 8(sp)
> +; RV32IFD-NEXT: fld ft1, 8(sp)
> ; RV32IFD-NEXT: sw a0, 8(sp)
> ; RV32IFD-NEXT: sw a1, 12(sp)
> -; RV32IFD-NEXT: fld ft1, 8(sp)
> -; RV32IFD-NEXT: fmul.d ft0, ft1, ft0
> -; RV32IFD-NEXT: sw a4, 8(sp)
> -; RV32IFD-NEXT: sw a5, 12(sp)
> -; RV32IFD-NEXT: fld ft1, 8(sp)
> -; RV32IFD-NEXT: fadd.d ft0, ft0, ft1
> +; RV32IFD-NEXT: fld ft2, 8(sp)
> +; RV32IFD-NEXT: fmul.d ft1, ft2, ft1
> +; RV32IFD-NEXT: fadd.d ft0, ft1, ft0
> ; RV32IFD-NEXT: fsd ft0, 8(sp)
> ; RV32IFD-NEXT: lw a0, 8(sp)
> ; RV32IFD-NEXT: lw a1, 12(sp)
> @@ -369,11 +364,11 @@ define double @fmuladd_f64(double %a, do
> ;
> ; RV64IFD-LABEL: fmuladd_f64:
> ; RV64IFD: # %bb.0:
> -; RV64IFD-NEXT: fmv.d.x ft0, a1
> -; RV64IFD-NEXT: fmv.d.x ft1, a0
> -; RV64IFD-NEXT: fmul.d ft0, ft1, ft0
> -; RV64IFD-NEXT: fmv.d.x ft1, a2
> -; RV64IFD-NEXT: fadd.d ft0, ft0, ft1
> +; RV64IFD-NEXT: fmv.d.x ft0, a2
> +; RV64IFD-NEXT: fmv.d.x ft1, a1
> +; RV64IFD-NEXT: fmv.d.x ft2, a0
> +; RV64IFD-NEXT: fmul.d ft1, ft2, ft1
> +; RV64IFD-NEXT: fadd.d ft0, ft1, ft0
> ; RV64IFD-NEXT: fmv.x.d a0, ft0
> ; RV64IFD-NEXT: ret
> %1 = call double @llvm.fmuladd.f64(double %a, double %b, double %c)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/double-mem.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/double-mem.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/double-mem.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/double-mem.ll Tue Sep 17 02:43:08 2019
> @@ -8,9 +8,9 @@ define double @fld(double *%a) nounwind
> ; RV32IFD-LABEL: fld:
> ; RV32IFD: # %bb.0:
> ; RV32IFD-NEXT: addi sp, sp, -16
> -; RV32IFD-NEXT: fld ft0, 24(a0)
> -; RV32IFD-NEXT: fld ft1, 0(a0)
> -; RV32IFD-NEXT: fadd.d ft0, ft1, ft0
> +; RV32IFD-NEXT: fld ft0, 0(a0)
> +; RV32IFD-NEXT: fld ft1, 24(a0)
> +; RV32IFD-NEXT: fadd.d ft0, ft0, ft1
> ; RV32IFD-NEXT: fsd ft0, 8(sp)
> ; RV32IFD-NEXT: lw a0, 8(sp)
> ; RV32IFD-NEXT: lw a1, 12(sp)
> @@ -19,9 +19,9 @@ define double @fld(double *%a) nounwind
> ;
> ; RV64IFD-LABEL: fld:
> ; RV64IFD: # %bb.0:
> -; RV64IFD-NEXT: fld ft0, 24(a0)
> -; RV64IFD-NEXT: fld ft1, 0(a0)
> -; RV64IFD-NEXT: fadd.d ft0, ft1, ft0
> +; RV64IFD-NEXT: fld ft0, 0(a0)
> +; RV64IFD-NEXT: fld ft1, 24(a0)
> +; RV64IFD-NEXT: fadd.d ft0, ft0, ft1
> ; RV64IFD-NEXT: fmv.x.d a0, ft0
> ; RV64IFD-NEXT: ret
> %1 = load double, double* %a
> @@ -44,8 +44,8 @@ define void @fsd(double *%a, double %b,
> ; RV32IFD-NEXT: sw a2, 12(sp)
> ; RV32IFD-NEXT: fld ft1, 8(sp)
> ; RV32IFD-NEXT: fadd.d ft0, ft1, ft0
> -; RV32IFD-NEXT: fsd ft0, 64(a0)
> ; RV32IFD-NEXT: fsd ft0, 0(a0)
> +; RV32IFD-NEXT: fsd ft0, 64(a0)
> ; RV32IFD-NEXT: addi sp, sp, 16
> ; RV32IFD-NEXT: ret
> ;
> @@ -54,8 +54,8 @@ define void @fsd(double *%a, double %b,
> ; RV64IFD-NEXT: fmv.d.x ft0, a2
> ; RV64IFD-NEXT: fmv.d.x ft1, a1
> ; RV64IFD-NEXT: fadd.d ft0, ft1, ft0
> -; RV64IFD-NEXT: fsd ft0, 64(a0)
> ; RV64IFD-NEXT: fsd ft0, 0(a0)
> +; RV64IFD-NEXT: fsd ft0, 64(a0)
> ; RV64IFD-NEXT: ret
> ; Use %b and %c in an FP op to ensure floating point registers are used, even
> ; for the soft float ABI
> @@ -100,10 +100,10 @@ define double @fld_fsd_global(double %a,
> ; RV64IFD-NEXT: lui a0, %hi(G)
> ; RV64IFD-NEXT: fld ft1, %lo(G)(a0)
> ; RV64IFD-NEXT: fsd ft0, %lo(G)(a0)
> -; RV64IFD-NEXT: addi a0, a0, %lo(G)
> -; RV64IFD-NEXT: fld ft1, 72(a0)
> -; RV64IFD-NEXT: fsd ft0, 72(a0)
> +; RV64IFD-NEXT: addi a1, a0, %lo(G)
> +; RV64IFD-NEXT: fld ft1, 72(a1)
> ; RV64IFD-NEXT: fmv.x.d a0, ft0
> +; RV64IFD-NEXT: fsd ft0, 72(a1)
> ; RV64IFD-NEXT: ret
> ; Use %a and %b in an FP op to ensure floating point registers are used, even
> ; for the soft float ABI
> @@ -136,14 +136,14 @@ define double @fld_fsd_constant(double %
> ;
> ; RV64IFD-LABEL: fld_fsd_constant:
> ; RV64IFD: # %bb.0:
> -; RV64IFD-NEXT: fmv.d.x ft0, a0
> -; RV64IFD-NEXT: lui a0, 56
> -; RV64IFD-NEXT: addiw a0, a0, -1353
> -; RV64IFD-NEXT: slli a0, a0, 14
> -; RV64IFD-NEXT: fld ft1, -273(a0)
> -; RV64IFD-NEXT: fadd.d ft0, ft0, ft1
> -; RV64IFD-NEXT: fsd ft0, -273(a0)
> +; RV64IFD-NEXT: lui a1, 56
> +; RV64IFD-NEXT: addiw a1, a1, -1353
> +; RV64IFD-NEXT: slli a1, a1, 14
> +; RV64IFD-NEXT: fld ft0, -273(a1)
> +; RV64IFD-NEXT: fmv.d.x ft1, a0
> +; RV64IFD-NEXT: fadd.d ft0, ft1, ft0
> ; RV64IFD-NEXT: fmv.x.d a0, ft0
> +; RV64IFD-NEXT: fsd ft0, -273(a1)
> ; RV64IFD-NEXT: ret
> %1 = inttoptr i32 3735928559 to double*
> %2 = load volatile double, double* %1
> @@ -159,22 +159,18 @@ define double @fld_stack(double %a) noun
> ; RV32IFD: # %bb.0:
> ; RV32IFD-NEXT: addi sp, sp, -32
> ; RV32IFD-NEXT: sw ra, 28(sp)
> -; RV32IFD-NEXT: sw s0, 24(sp)
> -; RV32IFD-NEXT: sw s1, 20(sp)
> -; RV32IFD-NEXT: mv s0, a1
> -; RV32IFD-NEXT: mv s1, a0
> -; RV32IFD-NEXT: addi a0, sp, 8
> -; RV32IFD-NEXT: call notdead
> -; RV32IFD-NEXT: sw s1, 0(sp)
> -; RV32IFD-NEXT: sw s0, 4(sp)
> -; RV32IFD-NEXT: fld ft0, 0(sp)
> -; RV32IFD-NEXT: fld ft1, 8(sp)
> -; RV32IFD-NEXT: fadd.d ft0, ft1, ft0
> +; RV32IFD-NEXT: sw a0, 8(sp)
> +; RV32IFD-NEXT: sw a1, 12(sp)
> +; RV32IFD-NEXT: fld ft0, 8(sp)
> ; RV32IFD-NEXT: fsd ft0, 0(sp)
> -; RV32IFD-NEXT: lw a0, 0(sp)
> -; RV32IFD-NEXT: lw a1, 4(sp)
> -; RV32IFD-NEXT: lw s1, 20(sp)
> -; RV32IFD-NEXT: lw s0, 24(sp)
> +; RV32IFD-NEXT: addi a0, sp, 16
> +; RV32IFD-NEXT: call notdead
> +; RV32IFD-NEXT: fld ft0, 16(sp)
> +; RV32IFD-NEXT: fld ft1, 0(sp)
> +; RV32IFD-NEXT: fadd.d ft0, ft0, ft1
> +; RV32IFD-NEXT: fsd ft0, 8(sp)
> +; RV32IFD-NEXT: lw a0, 8(sp)
> +; RV32IFD-NEXT: lw a1, 12(sp)
> ; RV32IFD-NEXT: lw ra, 28(sp)
> ; RV32IFD-NEXT: addi sp, sp, 32
> ; RV32IFD-NEXT: ret
> @@ -183,15 +179,14 @@ define double @fld_stack(double %a) noun
> ; RV64IFD: # %bb.0:
> ; RV64IFD-NEXT: addi sp, sp, -32
> ; RV64IFD-NEXT: sd ra, 24(sp)
> -; RV64IFD-NEXT: sd s0, 16(sp)
> -; RV64IFD-NEXT: mv s0, a0
> -; RV64IFD-NEXT: addi a0, sp, 8
> +; RV64IFD-NEXT: fmv.d.x ft0, a0
> +; RV64IFD-NEXT: fsd ft0, 8(sp)
> +; RV64IFD-NEXT: addi a0, sp, 16
> ; RV64IFD-NEXT: call notdead
> -; RV64IFD-NEXT: fmv.d.x ft0, s0
> +; RV64IFD-NEXT: fld ft0, 16(sp)
> ; RV64IFD-NEXT: fld ft1, 8(sp)
> -; RV64IFD-NEXT: fadd.d ft0, ft1, ft0
> +; RV64IFD-NEXT: fadd.d ft0, ft0, ft1
> ; RV64IFD-NEXT: fmv.x.d a0, ft0
> -; RV64IFD-NEXT: ld s0, 16(sp)
> ; RV64IFD-NEXT: ld ra, 24(sp)
> ; RV64IFD-NEXT: addi sp, sp, 32
> ; RV64IFD-NEXT: ret
>
> Modified: llvm/trunk/test/CodeGen/RISCV/double-previous-failure.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/double-previous-failure.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/double-previous-failure.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/double-previous-failure.ll Tue Sep 17 02:43:08 2019
> @@ -17,8 +17,8 @@ define i32 @main() nounwind {
> ; RV32IFD: # %bb.0: # %entry
> ; RV32IFD-NEXT: addi sp, sp, -16
> ; RV32IFD-NEXT: sw ra, 12(sp)
> -; RV32IFD-NEXT: mv a0, zero
> ; RV32IFD-NEXT: lui a1, 262144
> +; RV32IFD-NEXT: mv a0, zero
> ; RV32IFD-NEXT: call test
> ; RV32IFD-NEXT: sw a0, 0(sp)
> ; RV32IFD-NEXT: sw a1, 4(sp)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/double-select-fcmp.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/double-select-fcmp.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/double-select-fcmp.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/double-select-fcmp.ll Tue Sep 17 02:43:08 2019
> @@ -298,23 +298,23 @@ define double @select_fcmp_ueq(double %a
> ; RV32IFD-LABEL: select_fcmp_ueq:
> ; RV32IFD: # %bb.0:
> ; RV32IFD-NEXT: addi sp, sp, -16
> -; RV32IFD-NEXT: sw a0, 8(sp)
> -; RV32IFD-NEXT: sw a1, 12(sp)
> -; RV32IFD-NEXT: fld ft0, 8(sp)
> ; RV32IFD-NEXT: sw a2, 8(sp)
> ; RV32IFD-NEXT: sw a3, 12(sp)
> +; RV32IFD-NEXT: fld ft0, 8(sp)
> +; RV32IFD-NEXT: sw a0, 8(sp)
> +; RV32IFD-NEXT: sw a1, 12(sp)
> ; RV32IFD-NEXT: fld ft1, 8(sp)
> -; RV32IFD-NEXT: feq.d a0, ft1, ft1
> +; RV32IFD-NEXT: feq.d a0, ft1, ft0
> ; RV32IFD-NEXT: feq.d a1, ft0, ft0
> -; RV32IFD-NEXT: and a0, a1, a0
> -; RV32IFD-NEXT: seqz a0, a0
> -; RV32IFD-NEXT: feq.d a1, ft0, ft1
> -; RV32IFD-NEXT: or a0, a1, a0
> +; RV32IFD-NEXT: feq.d a2, ft1, ft1
> +; RV32IFD-NEXT: and a1, a2, a1
> +; RV32IFD-NEXT: seqz a1, a1
> +; RV32IFD-NEXT: or a0, a0, a1
> ; RV32IFD-NEXT: bnez a0, .LBB8_2
> ; RV32IFD-NEXT: # %bb.1:
> -; RV32IFD-NEXT: fmv.d ft0, ft1
> +; RV32IFD-NEXT: fmv.d ft1, ft0
> ; RV32IFD-NEXT: .LBB8_2:
> -; RV32IFD-NEXT: fsd ft0, 8(sp)
> +; RV32IFD-NEXT: fsd ft1, 8(sp)
> ; RV32IFD-NEXT: lw a0, 8(sp)
> ; RV32IFD-NEXT: lw a1, 12(sp)
> ; RV32IFD-NEXT: addi sp, sp, 16
> @@ -322,14 +322,14 @@ define double @select_fcmp_ueq(double %a
> ;
> ; RV64IFD-LABEL: select_fcmp_ueq:
> ; RV64IFD: # %bb.0:
> -; RV64IFD-NEXT: fmv.d.x ft0, a0
> ; RV64IFD-NEXT: fmv.d.x ft1, a1
> -; RV64IFD-NEXT: feq.d a0, ft1, ft1
> -; RV64IFD-NEXT: feq.d a1, ft0, ft0
> -; RV64IFD-NEXT: and a0, a1, a0
> -; RV64IFD-NEXT: seqz a0, a0
> -; RV64IFD-NEXT: feq.d a1, ft0, ft1
> -; RV64IFD-NEXT: or a0, a1, a0
> +; RV64IFD-NEXT: fmv.d.x ft0, a0
> +; RV64IFD-NEXT: feq.d a0, ft0, ft1
> +; RV64IFD-NEXT: feq.d a1, ft1, ft1
> +; RV64IFD-NEXT: feq.d a2, ft0, ft0
> +; RV64IFD-NEXT: and a1, a2, a1
> +; RV64IFD-NEXT: seqz a1, a1
> +; RV64IFD-NEXT: or a0, a0, a1
> ; RV64IFD-NEXT: bnez a0, .LBB8_2
> ; RV64IFD-NEXT: # %bb.1:
> ; RV64IFD-NEXT: fmv.d ft0, ft1
> @@ -604,12 +604,12 @@ define i32 @i32_select_fcmp_oeq(double %
> ; RV32IFD-NEXT: sw a0, 8(sp)
> ; RV32IFD-NEXT: sw a1, 12(sp)
> ; RV32IFD-NEXT: fld ft1, 8(sp)
> -; RV32IFD-NEXT: feq.d a0, ft1, ft0
> -; RV32IFD-NEXT: bnez a0, .LBB16_2
> +; RV32IFD-NEXT: feq.d a1, ft1, ft0
> +; RV32IFD-NEXT: mv a0, a4
> +; RV32IFD-NEXT: bnez a1, .LBB16_2
> ; RV32IFD-NEXT: # %bb.1:
> -; RV32IFD-NEXT: mv a4, a5
> +; RV32IFD-NEXT: mv a0, a5
> ; RV32IFD-NEXT: .LBB16_2:
> -; RV32IFD-NEXT: mv a0, a4
> ; RV32IFD-NEXT: addi sp, sp, 16
> ; RV32IFD-NEXT: ret
> ;
> @@ -617,12 +617,12 @@ define i32 @i32_select_fcmp_oeq(double %
> ; RV64IFD: # %bb.0:
> ; RV64IFD-NEXT: fmv.d.x ft0, a1
> ; RV64IFD-NEXT: fmv.d.x ft1, a0
> -; RV64IFD-NEXT: feq.d a0, ft1, ft0
> -; RV64IFD-NEXT: bnez a0, .LBB16_2
> +; RV64IFD-NEXT: feq.d a1, ft1, ft0
> +; RV64IFD-NEXT: mv a0, a2
> +; RV64IFD-NEXT: bnez a1, .LBB16_2
> ; RV64IFD-NEXT: # %bb.1:
> -; RV64IFD-NEXT: mv a2, a3
> +; RV64IFD-NEXT: mv a0, a3
> ; RV64IFD-NEXT: .LBB16_2:
> -; RV64IFD-NEXT: mv a0, a2
> ; RV64IFD-NEXT: ret
> %1 = fcmp oeq double %a, %b
> %2 = select i1 %1, i32 %c, i32 %d
>
> Modified: llvm/trunk/test/CodeGen/RISCV/double-stack-spill-restore.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/double-stack-spill-restore.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/double-stack-spill-restore.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/double-stack-spill-restore.ll Tue Sep 17 02:43:08 2019
> @@ -37,10 +37,10 @@ define double @func(double %d, i32 %n) n
> ; RV64IFD: # %bb.0: # %entry
> ; RV64IFD-NEXT: addi sp, sp, -16
> ; RV64IFD-NEXT: sd ra, 8(sp)
> +; RV64IFD-NEXT: slli a2, a1, 32
> +; RV64IFD-NEXT: srli a2, a2, 32
> ; RV64IFD-NEXT: fmv.d.x ft0, a0
> -; RV64IFD-NEXT: slli a0, a1, 32
> -; RV64IFD-NEXT: srli a0, a0, 32
> -; RV64IFD-NEXT: beqz a0, .LBB0_2
> +; RV64IFD-NEXT: beqz a2, .LBB0_2
> ; RV64IFD-NEXT: # %bb.1: # %if.else
> ; RV64IFD-NEXT: addi a1, a1, -1
> ; RV64IFD-NEXT: fmv.x.d a0, ft0
>
> Modified: llvm/trunk/test/CodeGen/RISCV/float-arith.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/float-arith.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/float-arith.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/float-arith.ll Tue Sep 17 02:43:08 2019
> @@ -339,26 +339,26 @@ define float @fmadd_s(float %a, float %b
> define float @fmsub_s(float %a, float %b, float %c) nounwind {
> ; RV32IF-LABEL: fmsub_s:
> ; RV32IF: # %bb.0:
> -; RV32IF-NEXT: fmv.w.x ft0, a2
> -; RV32IF-NEXT: lui a2, %hi(.LCPI15_0)
> -; RV32IF-NEXT: addi a2, a2, %lo(.LCPI15_0)
> -; RV32IF-NEXT: flw ft1, 0(a2)
> -; RV32IF-NEXT: fadd.s ft0, ft0, ft1
> +; RV32IF-NEXT: lui a3, %hi(.LCPI15_0)
> +; RV32IF-NEXT: addi a3, a3, %lo(.LCPI15_0)
> +; RV32IF-NEXT: flw ft0, 0(a3)
> ; RV32IF-NEXT: fmv.w.x ft1, a1
> ; RV32IF-NEXT: fmv.w.x ft2, a0
> +; RV32IF-NEXT: fmv.w.x ft3, a2
> +; RV32IF-NEXT: fadd.s ft0, ft3, ft0
> ; RV32IF-NEXT: fmsub.s ft0, ft2, ft1, ft0
> ; RV32IF-NEXT: fmv.x.w a0, ft0
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: fmsub_s:
> ; RV64IF: # %bb.0:
> -; RV64IF-NEXT: fmv.w.x ft0, a2
> -; RV64IF-NEXT: lui a2, %hi(.LCPI15_0)
> -; RV64IF-NEXT: addi a2, a2, %lo(.LCPI15_0)
> -; RV64IF-NEXT: flw ft1, 0(a2)
> -; RV64IF-NEXT: fadd.s ft0, ft0, ft1
> +; RV64IF-NEXT: lui a3, %hi(.LCPI15_0)
> +; RV64IF-NEXT: addi a3, a3, %lo(.LCPI15_0)
> +; RV64IF-NEXT: flw ft0, 0(a3)
> ; RV64IF-NEXT: fmv.w.x ft1, a1
> ; RV64IF-NEXT: fmv.w.x ft2, a0
> +; RV64IF-NEXT: fmv.w.x ft3, a2
> +; RV64IF-NEXT: fadd.s ft0, ft3, ft0
> ; RV64IF-NEXT: fmsub.s ft0, ft2, ft1, ft0
> ; RV64IF-NEXT: fmv.x.w a0, ft0
> ; RV64IF-NEXT: ret
> @@ -371,29 +371,29 @@ define float @fmsub_s(float %a, float %b
> define float @fnmadd_s(float %a, float %b, float %c) nounwind {
> ; RV32IF-LABEL: fnmadd_s:
> ; RV32IF: # %bb.0:
> -; RV32IF-NEXT: fmv.w.x ft0, a2
> -; RV32IF-NEXT: lui a2, %hi(.LCPI16_0)
> -; RV32IF-NEXT: addi a2, a2, %lo(.LCPI16_0)
> -; RV32IF-NEXT: flw ft1, 0(a2)
> -; RV32IF-NEXT: fadd.s ft0, ft0, ft1
> -; RV32IF-NEXT: fmv.w.x ft2, a0
> -; RV32IF-NEXT: fadd.s ft1, ft2, ft1
> -; RV32IF-NEXT: fmv.w.x ft2, a1
> -; RV32IF-NEXT: fnmadd.s ft0, ft1, ft2, ft0
> +; RV32IF-NEXT: lui a3, %hi(.LCPI16_0)
> +; RV32IF-NEXT: addi a3, a3, %lo(.LCPI16_0)
> +; RV32IF-NEXT: flw ft0, 0(a3)
> +; RV32IF-NEXT: fmv.w.x ft1, a1
> +; RV32IF-NEXT: fmv.w.x ft2, a2
> +; RV32IF-NEXT: fmv.w.x ft3, a0
> +; RV32IF-NEXT: fadd.s ft3, ft3, ft0
> +; RV32IF-NEXT: fadd.s ft0, ft2, ft0
> +; RV32IF-NEXT: fnmadd.s ft0, ft3, ft1, ft0
> ; RV32IF-NEXT: fmv.x.w a0, ft0
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: fnmadd_s:
> ; RV64IF: # %bb.0:
> -; RV64IF-NEXT: fmv.w.x ft0, a2
> -; RV64IF-NEXT: lui a2, %hi(.LCPI16_0)
> -; RV64IF-NEXT: addi a2, a2, %lo(.LCPI16_0)
> -; RV64IF-NEXT: flw ft1, 0(a2)
> -; RV64IF-NEXT: fadd.s ft0, ft0, ft1
> -; RV64IF-NEXT: fmv.w.x ft2, a0
> -; RV64IF-NEXT: fadd.s ft1, ft2, ft1
> -; RV64IF-NEXT: fmv.w.x ft2, a1
> -; RV64IF-NEXT: fnmadd.s ft0, ft1, ft2, ft0
> +; RV64IF-NEXT: lui a3, %hi(.LCPI16_0)
> +; RV64IF-NEXT: addi a3, a3, %lo(.LCPI16_0)
> +; RV64IF-NEXT: flw ft0, 0(a3)
> +; RV64IF-NEXT: fmv.w.x ft1, a1
> +; RV64IF-NEXT: fmv.w.x ft2, a2
> +; RV64IF-NEXT: fmv.w.x ft3, a0
> +; RV64IF-NEXT: fadd.s ft3, ft3, ft0
> +; RV64IF-NEXT: fadd.s ft0, ft2, ft0
> +; RV64IF-NEXT: fnmadd.s ft0, ft3, ft1, ft0
> ; RV64IF-NEXT: fmv.x.w a0, ft0
> ; RV64IF-NEXT: ret
> %a_ = fadd float 0.0, %a
> @@ -407,26 +407,26 @@ define float @fnmadd_s(float %a, float %
> define float @fnmsub_s(float %a, float %b, float %c) nounwind {
> ; RV32IF-LABEL: fnmsub_s:
> ; RV32IF: # %bb.0:
> -; RV32IF-NEXT: fmv.w.x ft0, a0
> -; RV32IF-NEXT: lui a0, %hi(.LCPI17_0)
> -; RV32IF-NEXT: addi a0, a0, %lo(.LCPI17_0)
> -; RV32IF-NEXT: flw ft1, 0(a0)
> -; RV32IF-NEXT: fadd.s ft0, ft0, ft1
> +; RV32IF-NEXT: lui a3, %hi(.LCPI17_0)
> +; RV32IF-NEXT: addi a3, a3, %lo(.LCPI17_0)
> +; RV32IF-NEXT: flw ft0, 0(a3)
> ; RV32IF-NEXT: fmv.w.x ft1, a2
> ; RV32IF-NEXT: fmv.w.x ft2, a1
> +; RV32IF-NEXT: fmv.w.x ft3, a0
> +; RV32IF-NEXT: fadd.s ft0, ft3, ft0
> ; RV32IF-NEXT: fnmsub.s ft0, ft0, ft2, ft1
> ; RV32IF-NEXT: fmv.x.w a0, ft0
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: fnmsub_s:
> ; RV64IF: # %bb.0:
> -; RV64IF-NEXT: fmv.w.x ft0, a0
> -; RV64IF-NEXT: lui a0, %hi(.LCPI17_0)
> -; RV64IF-NEXT: addi a0, a0, %lo(.LCPI17_0)
> -; RV64IF-NEXT: flw ft1, 0(a0)
> -; RV64IF-NEXT: fadd.s ft0, ft0, ft1
> +; RV64IF-NEXT: lui a3, %hi(.LCPI17_0)
> +; RV64IF-NEXT: addi a3, a3, %lo(.LCPI17_0)
> +; RV64IF-NEXT: flw ft0, 0(a3)
> ; RV64IF-NEXT: fmv.w.x ft1, a2
> ; RV64IF-NEXT: fmv.w.x ft2, a1
> +; RV64IF-NEXT: fmv.w.x ft3, a0
> +; RV64IF-NEXT: fadd.s ft0, ft3, ft0
> ; RV64IF-NEXT: fnmsub.s ft0, ft0, ft2, ft1
> ; RV64IF-NEXT: fmv.x.w a0, ft0
> ; RV64IF-NEXT: ret
>
> Modified: llvm/trunk/test/CodeGen/RISCV/float-br-fcmp.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/float-br-fcmp.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/float-br-fcmp.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/float-br-fcmp.ll Tue Sep 17 02:43:08 2019
> @@ -337,11 +337,11 @@ define void @br_fcmp_ord(float %a, float
> ; RV32IF: # %bb.0:
> ; RV32IF-NEXT: addi sp, sp, -16
> ; RV32IF-NEXT: sw ra, 12(sp)
> -; RV32IF-NEXT: fmv.w.x ft0, a1
> -; RV32IF-NEXT: feq.s a1, ft0, ft0
> ; RV32IF-NEXT: fmv.w.x ft0, a0
> -; RV32IF-NEXT: feq.s a0, ft0, ft0
> -; RV32IF-NEXT: and a0, a0, a1
> +; RV32IF-NEXT: fmv.w.x ft1, a1
> +; RV32IF-NEXT: feq.s a0, ft1, ft1
> +; RV32IF-NEXT: feq.s a1, ft0, ft0
> +; RV32IF-NEXT: and a0, a1, a0
> ; RV32IF-NEXT: bnez a0, .LBB8_2
> ; RV32IF-NEXT: # %bb.1: # %if.else
> ; RV32IF-NEXT: lw ra, 12(sp)
> @@ -354,11 +354,11 @@ define void @br_fcmp_ord(float %a, float
> ; RV64IF: # %bb.0:
> ; RV64IF-NEXT: addi sp, sp, -16
> ; RV64IF-NEXT: sd ra, 8(sp)
> -; RV64IF-NEXT: fmv.w.x ft0, a1
> -; RV64IF-NEXT: feq.s a1, ft0, ft0
> ; RV64IF-NEXT: fmv.w.x ft0, a0
> -; RV64IF-NEXT: feq.s a0, ft0, ft0
> -; RV64IF-NEXT: and a0, a0, a1
> +; RV64IF-NEXT: fmv.w.x ft1, a1
> +; RV64IF-NEXT: feq.s a0, ft1, ft1
> +; RV64IF-NEXT: feq.s a1, ft0, ft0
> +; RV64IF-NEXT: and a0, a1, a0
> ; RV64IF-NEXT: bnez a0, .LBB8_2
> ; RV64IF-NEXT: # %bb.1: # %if.else
> ; RV64IF-NEXT: ld ra, 8(sp)
> @@ -635,11 +635,11 @@ define void @br_fcmp_uno(float %a, float
> ; RV32IF: # %bb.0:
> ; RV32IF-NEXT: addi sp, sp, -16
> ; RV32IF-NEXT: sw ra, 12(sp)
> -; RV32IF-NEXT: fmv.w.x ft0, a1
> -; RV32IF-NEXT: feq.s a1, ft0, ft0
> ; RV32IF-NEXT: fmv.w.x ft0, a0
> -; RV32IF-NEXT: feq.s a0, ft0, ft0
> -; RV32IF-NEXT: and a0, a0, a1
> +; RV32IF-NEXT: fmv.w.x ft1, a1
> +; RV32IF-NEXT: feq.s a0, ft1, ft1
> +; RV32IF-NEXT: feq.s a1, ft0, ft0
> +; RV32IF-NEXT: and a0, a1, a0
> ; RV32IF-NEXT: seqz a0, a0
> ; RV32IF-NEXT: bnez a0, .LBB15_2
> ; RV32IF-NEXT: # %bb.1: # %if.else
> @@ -653,11 +653,11 @@ define void @br_fcmp_uno(float %a, float
> ; RV64IF: # %bb.0:
> ; RV64IF-NEXT: addi sp, sp, -16
> ; RV64IF-NEXT: sd ra, 8(sp)
> -; RV64IF-NEXT: fmv.w.x ft0, a1
> -; RV64IF-NEXT: feq.s a1, ft0, ft0
> ; RV64IF-NEXT: fmv.w.x ft0, a0
> -; RV64IF-NEXT: feq.s a0, ft0, ft0
> -; RV64IF-NEXT: and a0, a0, a1
> +; RV64IF-NEXT: fmv.w.x ft1, a1
> +; RV64IF-NEXT: feq.s a0, ft1, ft1
> +; RV64IF-NEXT: feq.s a1, ft0, ft0
> +; RV64IF-NEXT: and a0, a1, a0
> ; RV64IF-NEXT: seqz a0, a0
> ; RV64IF-NEXT: bnez a0, .LBB15_2
> ; RV64IF-NEXT: # %bb.1: # %if.else
>
> Modified: llvm/trunk/test/CodeGen/RISCV/float-convert.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/float-convert.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/float-convert.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/float-convert.ll Tue Sep 17 02:43:08 2019
> @@ -97,17 +97,17 @@ define float @fcvt_s_wu(i32 %a) nounwind
> define float @fmv_w_x(i32 %a, i32 %b) nounwind {
> ; RV32IF-LABEL: fmv_w_x:
> ; RV32IF: # %bb.0:
> -; RV32IF-NEXT: fmv.w.x ft0, a1
> -; RV32IF-NEXT: fmv.w.x ft1, a0
> -; RV32IF-NEXT: fadd.s ft0, ft1, ft0
> +; RV32IF-NEXT: fmv.w.x ft0, a0
> +; RV32IF-NEXT: fmv.w.x ft1, a1
> +; RV32IF-NEXT: fadd.s ft0, ft0, ft1
> ; RV32IF-NEXT: fmv.x.w a0, ft0
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: fmv_w_x:
> ; RV64IF: # %bb.0:
> -; RV64IF-NEXT: fmv.w.x ft0, a1
> -; RV64IF-NEXT: fmv.w.x ft1, a0
> -; RV64IF-NEXT: fadd.s ft0, ft1, ft0
> +; RV64IF-NEXT: fmv.w.x ft0, a0
> +; RV64IF-NEXT: fmv.w.x ft1, a1
> +; RV64IF-NEXT: fadd.s ft0, ft0, ft1
> ; RV64IF-NEXT: fmv.x.w a0, ft0
> ; RV64IF-NEXT: ret
> ; Ensure fmv.w.x is generated even for a soft float calling convention
>
> Modified: llvm/trunk/test/CodeGen/RISCV/float-fcmp.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/float-fcmp.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/float-fcmp.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/float-fcmp.ll Tue Sep 17 02:43:08 2019
> @@ -146,20 +146,20 @@ define i32 @fcmp_one(float %a, float %b)
> define i32 @fcmp_ord(float %a, float %b) nounwind {
> ; RV32IF-LABEL: fcmp_ord:
> ; RV32IF: # %bb.0:
> -; RV32IF-NEXT: fmv.w.x ft0, a1
> -; RV32IF-NEXT: feq.s a1, ft0, ft0
> ; RV32IF-NEXT: fmv.w.x ft0, a0
> -; RV32IF-NEXT: feq.s a0, ft0, ft0
> -; RV32IF-NEXT: and a0, a0, a1
> +; RV32IF-NEXT: fmv.w.x ft1, a1
> +; RV32IF-NEXT: feq.s a0, ft1, ft1
> +; RV32IF-NEXT: feq.s a1, ft0, ft0
> +; RV32IF-NEXT: and a0, a1, a0
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: fcmp_ord:
> ; RV64IF: # %bb.0:
> -; RV64IF-NEXT: fmv.w.x ft0, a1
> -; RV64IF-NEXT: feq.s a1, ft0, ft0
> ; RV64IF-NEXT: fmv.w.x ft0, a0
> -; RV64IF-NEXT: feq.s a0, ft0, ft0
> -; RV64IF-NEXT: and a0, a0, a1
> +; RV64IF-NEXT: fmv.w.x ft1, a1
> +; RV64IF-NEXT: feq.s a0, ft1, ft1
> +; RV64IF-NEXT: feq.s a1, ft0, ft0
> +; RV64IF-NEXT: and a0, a1, a0
> ; RV64IF-NEXT: ret
> %1 = fcmp ord float %a, %b
> %2 = zext i1 %1 to i32
> @@ -303,21 +303,21 @@ define i32 @fcmp_une(float %a, float %b)
> define i32 @fcmp_uno(float %a, float %b) nounwind {
> ; RV32IF-LABEL: fcmp_uno:
> ; RV32IF: # %bb.0:
> -; RV32IF-NEXT: fmv.w.x ft0, a1
> -; RV32IF-NEXT: feq.s a1, ft0, ft0
> ; RV32IF-NEXT: fmv.w.x ft0, a0
> -; RV32IF-NEXT: feq.s a0, ft0, ft0
> -; RV32IF-NEXT: and a0, a0, a1
> +; RV32IF-NEXT: fmv.w.x ft1, a1
> +; RV32IF-NEXT: feq.s a0, ft1, ft1
> +; RV32IF-NEXT: feq.s a1, ft0, ft0
> +; RV32IF-NEXT: and a0, a1, a0
> ; RV32IF-NEXT: seqz a0, a0
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: fcmp_uno:
> ; RV64IF: # %bb.0:
> -; RV64IF-NEXT: fmv.w.x ft0, a1
> -; RV64IF-NEXT: feq.s a1, ft0, ft0
> ; RV64IF-NEXT: fmv.w.x ft0, a0
> -; RV64IF-NEXT: feq.s a0, ft0, ft0
> -; RV64IF-NEXT: and a0, a0, a1
> +; RV64IF-NEXT: fmv.w.x ft1, a1
> +; RV64IF-NEXT: feq.s a0, ft1, ft1
> +; RV64IF-NEXT: feq.s a1, ft0, ft0
> +; RV64IF-NEXT: and a0, a1, a0
> ; RV64IF-NEXT: seqz a0, a0
> ; RV64IF-NEXT: ret
> %1 = fcmp uno float %a, %b
>
> Modified: llvm/trunk/test/CodeGen/RISCV/float-imm.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/float-imm.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/float-imm.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/float-imm.ll Tue Sep 17 02:43:08 2019
> @@ -26,21 +26,21 @@ define float @float_imm_op(float %a) nou
> ; TODO: addi should be folded in to the flw
> ; RV32IF-LABEL: float_imm_op:
> ; RV32IF: # %bb.0:
> -; RV32IF-NEXT: fmv.w.x ft0, a0
> -; RV32IF-NEXT: lui a0, %hi(.LCPI1_0)
> -; RV32IF-NEXT: addi a0, a0, %lo(.LCPI1_0)
> -; RV32IF-NEXT: flw ft1, 0(a0)
> -; RV32IF-NEXT: fadd.s ft0, ft0, ft1
> +; RV32IF-NEXT: lui a1, %hi(.LCPI1_0)
> +; RV32IF-NEXT: addi a1, a1, %lo(.LCPI1_0)
> +; RV32IF-NEXT: flw ft0, 0(a1)
> +; RV32IF-NEXT: fmv.w.x ft1, a0
> +; RV32IF-NEXT: fadd.s ft0, ft1, ft0
> ; RV32IF-NEXT: fmv.x.w a0, ft0
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: float_imm_op:
> ; RV64IF: # %bb.0:
> -; RV64IF-NEXT: fmv.w.x ft0, a0
> -; RV64IF-NEXT: lui a0, %hi(.LCPI1_0)
> -; RV64IF-NEXT: addi a0, a0, %lo(.LCPI1_0)
> -; RV64IF-NEXT: flw ft1, 0(a0)
> -; RV64IF-NEXT: fadd.s ft0, ft0, ft1
> +; RV64IF-NEXT: lui a1, %hi(.LCPI1_0)
> +; RV64IF-NEXT: addi a1, a1, %lo(.LCPI1_0)
> +; RV64IF-NEXT: flw ft0, 0(a1)
> +; RV64IF-NEXT: fmv.w.x ft1, a0
> +; RV64IF-NEXT: fadd.s ft0, ft1, ft0
> ; RV64IF-NEXT: fmv.x.w a0, ft0
> ; RV64IF-NEXT: ret
> %1 = fadd float %a, 1.0
>
> Modified: llvm/trunk/test/CodeGen/RISCV/float-intrinsics.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/float-intrinsics.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/float-intrinsics.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/float-intrinsics.ll Tue Sep 17 02:43:08 2019
> @@ -108,17 +108,16 @@ define float @sincos_f32(float %a) nounw
> ; RV32IF-NEXT: addi sp, sp, -16
> ; RV32IF-NEXT: sw ra, 12(sp)
> ; RV32IF-NEXT: sw s0, 8(sp)
> -; RV32IF-NEXT: sw s1, 4(sp)
> ; RV32IF-NEXT: mv s0, a0
> ; RV32IF-NEXT: call sinf
> -; RV32IF-NEXT: mv s1, a0
> +; RV32IF-NEXT: fmv.w.x ft0, a0
> +; RV32IF-NEXT: fsw ft0, 4(sp)
> ; RV32IF-NEXT: mv a0, s0
> ; RV32IF-NEXT: call cosf
> ; RV32IF-NEXT: fmv.w.x ft0, a0
> -; RV32IF-NEXT: fmv.w.x ft1, s1
> +; RV32IF-NEXT: flw ft1, 4(sp)
> ; RV32IF-NEXT: fadd.s ft0, ft1, ft0
> ; RV32IF-NEXT: fmv.x.w a0, ft0
> -; RV32IF-NEXT: lw s1, 4(sp)
> ; RV32IF-NEXT: lw s0, 8(sp)
> ; RV32IF-NEXT: lw ra, 12(sp)
> ; RV32IF-NEXT: addi sp, sp, 16
> @@ -129,17 +128,16 @@ define float @sincos_f32(float %a) nounw
> ; RV64IF-NEXT: addi sp, sp, -32
> ; RV64IF-NEXT: sd ra, 24(sp)
> ; RV64IF-NEXT: sd s0, 16(sp)
> -; RV64IF-NEXT: sd s1, 8(sp)
> ; RV64IF-NEXT: mv s0, a0
> ; RV64IF-NEXT: call sinf
> -; RV64IF-NEXT: mv s1, a0
> +; RV64IF-NEXT: fmv.w.x ft0, a0
> +; RV64IF-NEXT: fsw ft0, 12(sp)
> ; RV64IF-NEXT: mv a0, s0
> ; RV64IF-NEXT: call cosf
> ; RV64IF-NEXT: fmv.w.x ft0, a0
> -; RV64IF-NEXT: fmv.w.x ft1, s1
> +; RV64IF-NEXT: flw ft1, 12(sp)
> ; RV64IF-NEXT: fadd.s ft0, ft1, ft0
> ; RV64IF-NEXT: fmv.x.w a0, ft0
> -; RV64IF-NEXT: ld s1, 8(sp)
> ; RV64IF-NEXT: ld s0, 16(sp)
> ; RV64IF-NEXT: ld ra, 24(sp)
> ; RV64IF-NEXT: addi sp, sp, 32
> @@ -324,21 +322,21 @@ define float @fmuladd_f32(float %a, floa
> ; Use of fmadd depends on TargetLowering::isFMAFasterthanFMulAndFAdd
> ; RV32IF-LABEL: fmuladd_f32:
> ; RV32IF: # %bb.0:
> -; RV32IF-NEXT: fmv.w.x ft0, a1
> -; RV32IF-NEXT: fmv.w.x ft1, a0
> -; RV32IF-NEXT: fmul.s ft0, ft1, ft0
> -; RV32IF-NEXT: fmv.w.x ft1, a2
> -; RV32IF-NEXT: fadd.s ft0, ft0, ft1
> +; RV32IF-NEXT: fmv.w.x ft0, a2
> +; RV32IF-NEXT: fmv.w.x ft1, a1
> +; RV32IF-NEXT: fmv.w.x ft2, a0
> +; RV32IF-NEXT: fmul.s ft1, ft2, ft1
> +; RV32IF-NEXT: fadd.s ft0, ft1, ft0
> ; RV32IF-NEXT: fmv.x.w a0, ft0
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: fmuladd_f32:
> ; RV64IF: # %bb.0:
> -; RV64IF-NEXT: fmv.w.x ft0, a1
> -; RV64IF-NEXT: fmv.w.x ft1, a0
> -; RV64IF-NEXT: fmul.s ft0, ft1, ft0
> -; RV64IF-NEXT: fmv.w.x ft1, a2
> -; RV64IF-NEXT: fadd.s ft0, ft0, ft1
> +; RV64IF-NEXT: fmv.w.x ft0, a2
> +; RV64IF-NEXT: fmv.w.x ft1, a1
> +; RV64IF-NEXT: fmv.w.x ft2, a0
> +; RV64IF-NEXT: fmul.s ft1, ft2, ft1
> +; RV64IF-NEXT: fadd.s ft0, ft1, ft0
> ; RV64IF-NEXT: fmv.x.w a0, ft0
> ; RV64IF-NEXT: ret
> %1 = call float @llvm.fmuladd.f32(float %a, float %b, float %c)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/float-mem.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/float-mem.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/float-mem.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/float-mem.ll Tue Sep 17 02:43:08 2019
> @@ -7,17 +7,17 @@
> define float @flw(float *%a) nounwind {
> ; RV32IF-LABEL: flw:
> ; RV32IF: # %bb.0:
> -; RV32IF-NEXT: flw ft0, 12(a0)
> -; RV32IF-NEXT: flw ft1, 0(a0)
> -; RV32IF-NEXT: fadd.s ft0, ft1, ft0
> +; RV32IF-NEXT: flw ft0, 0(a0)
> +; RV32IF-NEXT: flw ft1, 12(a0)
> +; RV32IF-NEXT: fadd.s ft0, ft0, ft1
> ; RV32IF-NEXT: fmv.x.w a0, ft0
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: flw:
> ; RV64IF: # %bb.0:
> -; RV64IF-NEXT: flw ft0, 12(a0)
> -; RV64IF-NEXT: flw ft1, 0(a0)
> -; RV64IF-NEXT: fadd.s ft0, ft1, ft0
> +; RV64IF-NEXT: flw ft0, 0(a0)
> +; RV64IF-NEXT: flw ft1, 12(a0)
> +; RV64IF-NEXT: fadd.s ft0, ft0, ft1
> ; RV64IF-NEXT: fmv.x.w a0, ft0
> ; RV64IF-NEXT: ret
> %1 = load float, float* %a
> @@ -37,8 +37,8 @@ define void @fsw(float *%a, float %b, fl
> ; RV32IF-NEXT: fmv.w.x ft0, a2
> ; RV32IF-NEXT: fmv.w.x ft1, a1
> ; RV32IF-NEXT: fadd.s ft0, ft1, ft0
> -; RV32IF-NEXT: fsw ft0, 32(a0)
> ; RV32IF-NEXT: fsw ft0, 0(a0)
> +; RV32IF-NEXT: fsw ft0, 32(a0)
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: fsw:
> @@ -46,8 +46,8 @@ define void @fsw(float *%a, float %b, fl
> ; RV64IF-NEXT: fmv.w.x ft0, a2
> ; RV64IF-NEXT: fmv.w.x ft1, a1
> ; RV64IF-NEXT: fadd.s ft0, ft1, ft0
> -; RV64IF-NEXT: fsw ft0, 32(a0)
> ; RV64IF-NEXT: fsw ft0, 0(a0)
> +; RV64IF-NEXT: fsw ft0, 32(a0)
> ; RV64IF-NEXT: ret
> %1 = fadd float %b, %c
> store float %1, float* %a
> @@ -70,10 +70,10 @@ define float @flw_fsw_global(float %a, f
> ; RV32IF-NEXT: lui a0, %hi(G)
> ; RV32IF-NEXT: flw ft1, %lo(G)(a0)
> ; RV32IF-NEXT: fsw ft0, %lo(G)(a0)
> -; RV32IF-NEXT: addi a0, a0, %lo(G)
> -; RV32IF-NEXT: flw ft1, 36(a0)
> -; RV32IF-NEXT: fsw ft0, 36(a0)
> +; RV32IF-NEXT: addi a1, a0, %lo(G)
> +; RV32IF-NEXT: flw ft1, 36(a1)
> ; RV32IF-NEXT: fmv.x.w a0, ft0
> +; RV32IF-NEXT: fsw ft0, 36(a1)
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: flw_fsw_global:
> @@ -84,10 +84,10 @@ define float @flw_fsw_global(float %a, f
> ; RV64IF-NEXT: lui a0, %hi(G)
> ; RV64IF-NEXT: flw ft1, %lo(G)(a0)
> ; RV64IF-NEXT: fsw ft0, %lo(G)(a0)
> -; RV64IF-NEXT: addi a0, a0, %lo(G)
> -; RV64IF-NEXT: flw ft1, 36(a0)
> -; RV64IF-NEXT: fsw ft0, 36(a0)
> +; RV64IF-NEXT: addi a1, a0, %lo(G)
> +; RV64IF-NEXT: flw ft1, 36(a1)
> ; RV64IF-NEXT: fmv.x.w a0, ft0
> +; RV64IF-NEXT: fsw ft0, 36(a1)
> ; RV64IF-NEXT: ret
> %1 = fadd float %a, %b
> %2 = load volatile float, float* @G
> @@ -102,24 +102,24 @@ define float @flw_fsw_global(float %a, f
> define float @flw_fsw_constant(float %a) nounwind {
> ; RV32IF-LABEL: flw_fsw_constant:
> ; RV32IF: # %bb.0:
> -; RV32IF-NEXT: fmv.w.x ft0, a0
> -; RV32IF-NEXT: lui a0, 912092
> -; RV32IF-NEXT: flw ft1, -273(a0)
> -; RV32IF-NEXT: fadd.s ft0, ft0, ft1
> -; RV32IF-NEXT: fsw ft0, -273(a0)
> +; RV32IF-NEXT: lui a1, 912092
> +; RV32IF-NEXT: flw ft0, -273(a1)
> +; RV32IF-NEXT: fmv.w.x ft1, a0
> +; RV32IF-NEXT: fadd.s ft0, ft1, ft0
> ; RV32IF-NEXT: fmv.x.w a0, ft0
> +; RV32IF-NEXT: fsw ft0, -273(a1)
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: flw_fsw_constant:
> ; RV64IF: # %bb.0:
> -; RV64IF-NEXT: fmv.w.x ft0, a0
> -; RV64IF-NEXT: lui a0, 56
> -; RV64IF-NEXT: addiw a0, a0, -1353
> -; RV64IF-NEXT: slli a0, a0, 14
> -; RV64IF-NEXT: flw ft1, -273(a0)
> -; RV64IF-NEXT: fadd.s ft0, ft0, ft1
> -; RV64IF-NEXT: fsw ft0, -273(a0)
> +; RV64IF-NEXT: lui a1, 56
> +; RV64IF-NEXT: addiw a1, a1, -1353
> +; RV64IF-NEXT: slli a1, a1, 14
> +; RV64IF-NEXT: flw ft0, -273(a1)
> +; RV64IF-NEXT: fmv.w.x ft1, a0
> +; RV64IF-NEXT: fadd.s ft0, ft1, ft0
> ; RV64IF-NEXT: fmv.x.w a0, ft0
> +; RV64IF-NEXT: fsw ft0, -273(a1)
> ; RV64IF-NEXT: ret
> %1 = inttoptr i32 3735928559 to float*
> %2 = load volatile float, float* %1
> @@ -135,34 +135,32 @@ define float @flw_stack(float %a) nounwi
> ; RV32IF: # %bb.0:
> ; RV32IF-NEXT: addi sp, sp, -16
> ; RV32IF-NEXT: sw ra, 12(sp)
> -; RV32IF-NEXT: sw s0, 8(sp)
> -; RV32IF-NEXT: mv s0, a0
> -; RV32IF-NEXT: addi a0, sp, 4
> +; RV32IF-NEXT: fmv.w.x ft0, a0
> +; RV32IF-NEXT: fsw ft0, 4(sp)
> +; RV32IF-NEXT: addi a0, sp, 8
> ; RV32IF-NEXT: call notdead
> -; RV32IF-NEXT: fmv.w.x ft0, s0
> +; RV32IF-NEXT: flw ft0, 8(sp)
> ; RV32IF-NEXT: flw ft1, 4(sp)
> -; RV32IF-NEXT: fadd.s ft0, ft1, ft0
> +; RV32IF-NEXT: fadd.s ft0, ft0, ft1
> ; RV32IF-NEXT: fmv.x.w a0, ft0
> -; RV32IF-NEXT: lw s0, 8(sp)
> ; RV32IF-NEXT: lw ra, 12(sp)
> ; RV32IF-NEXT: addi sp, sp, 16
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: flw_stack:
> ; RV64IF: # %bb.0:
> -; RV64IF-NEXT: addi sp, sp, -32
> -; RV64IF-NEXT: sd ra, 24(sp)
> -; RV64IF-NEXT: sd s0, 16(sp)
> -; RV64IF-NEXT: mv s0, a0
> -; RV64IF-NEXT: addi a0, sp, 12
> +; RV64IF-NEXT: addi sp, sp, -16
> +; RV64IF-NEXT: sd ra, 8(sp)
> +; RV64IF-NEXT: fmv.w.x ft0, a0
> +; RV64IF-NEXT: fsw ft0, 0(sp)
> +; RV64IF-NEXT: addi a0, sp, 4
> ; RV64IF-NEXT: call notdead
> -; RV64IF-NEXT: fmv.w.x ft0, s0
> -; RV64IF-NEXT: flw ft1, 12(sp)
> -; RV64IF-NEXT: fadd.s ft0, ft1, ft0
> +; RV64IF-NEXT: flw ft0, 4(sp)
> +; RV64IF-NEXT: flw ft1, 0(sp)
> +; RV64IF-NEXT: fadd.s ft0, ft0, ft1
> ; RV64IF-NEXT: fmv.x.w a0, ft0
> -; RV64IF-NEXT: ld s0, 16(sp)
> -; RV64IF-NEXT: ld ra, 24(sp)
> -; RV64IF-NEXT: addi sp, sp, 32
> +; RV64IF-NEXT: ld ra, 8(sp)
> +; RV64IF-NEXT: addi sp, sp, 16
> ; RV64IF-NEXT: ret
> %1 = alloca float, align 4
> %2 = bitcast float* %1 to i8*
>
> Modified: llvm/trunk/test/CodeGen/RISCV/float-select-fcmp.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/float-select-fcmp.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/float-select-fcmp.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/float-select-fcmp.ll Tue Sep 17 02:43:08 2019
> @@ -240,14 +240,14 @@ define float @select_fcmp_ord(float %a,
> define float @select_fcmp_ueq(float %a, float %b) nounwind {
> ; RV32IF-LABEL: select_fcmp_ueq:
> ; RV32IF: # %bb.0:
> -; RV32IF-NEXT: fmv.w.x ft0, a0
> ; RV32IF-NEXT: fmv.w.x ft1, a1
> -; RV32IF-NEXT: feq.s a0, ft1, ft1
> -; RV32IF-NEXT: feq.s a1, ft0, ft0
> -; RV32IF-NEXT: and a0, a1, a0
> -; RV32IF-NEXT: seqz a0, a0
> -; RV32IF-NEXT: feq.s a1, ft0, ft1
> -; RV32IF-NEXT: or a0, a1, a0
> +; RV32IF-NEXT: fmv.w.x ft0, a0
> +; RV32IF-NEXT: feq.s a0, ft0, ft1
> +; RV32IF-NEXT: feq.s a1, ft1, ft1
> +; RV32IF-NEXT: feq.s a2, ft0, ft0
> +; RV32IF-NEXT: and a1, a2, a1
> +; RV32IF-NEXT: seqz a1, a1
> +; RV32IF-NEXT: or a0, a0, a1
> ; RV32IF-NEXT: bnez a0, .LBB8_2
> ; RV32IF-NEXT: # %bb.1:
> ; RV32IF-NEXT: fmv.s ft0, ft1
> @@ -257,14 +257,14 @@ define float @select_fcmp_ueq(float %a,
> ;
> ; RV64IF-LABEL: select_fcmp_ueq:
> ; RV64IF: # %bb.0:
> -; RV64IF-NEXT: fmv.w.x ft0, a0
> ; RV64IF-NEXT: fmv.w.x ft1, a1
> -; RV64IF-NEXT: feq.s a0, ft1, ft1
> -; RV64IF-NEXT: feq.s a1, ft0, ft0
> -; RV64IF-NEXT: and a0, a1, a0
> -; RV64IF-NEXT: seqz a0, a0
> -; RV64IF-NEXT: feq.s a1, ft0, ft1
> -; RV64IF-NEXT: or a0, a1, a0
> +; RV64IF-NEXT: fmv.w.x ft0, a0
> +; RV64IF-NEXT: feq.s a0, ft0, ft1
> +; RV64IF-NEXT: feq.s a1, ft1, ft1
> +; RV64IF-NEXT: feq.s a2, ft0, ft0
> +; RV64IF-NEXT: and a1, a2, a1
> +; RV64IF-NEXT: seqz a1, a1
> +; RV64IF-NEXT: or a0, a0, a1
> ; RV64IF-NEXT: bnez a0, .LBB8_2
> ; RV64IF-NEXT: # %bb.1:
> ; RV64IF-NEXT: fmv.s ft0, ft1
> @@ -486,24 +486,24 @@ define i32 @i32_select_fcmp_oeq(float %a
> ; RV32IF: # %bb.0:
> ; RV32IF-NEXT: fmv.w.x ft0, a1
> ; RV32IF-NEXT: fmv.w.x ft1, a0
> -; RV32IF-NEXT: feq.s a0, ft1, ft0
> -; RV32IF-NEXT: bnez a0, .LBB16_2
> +; RV32IF-NEXT: feq.s a1, ft1, ft0
> +; RV32IF-NEXT: mv a0, a2
> +; RV32IF-NEXT: bnez a1, .LBB16_2
> ; RV32IF-NEXT: # %bb.1:
> -; RV32IF-NEXT: mv a2, a3
> +; RV32IF-NEXT: mv a0, a3
> ; RV32IF-NEXT: .LBB16_2:
> -; RV32IF-NEXT: mv a0, a2
> ; RV32IF-NEXT: ret
> ;
> ; RV64IF-LABEL: i32_select_fcmp_oeq:
> ; RV64IF: # %bb.0:
> ; RV64IF-NEXT: fmv.w.x ft0, a1
> ; RV64IF-NEXT: fmv.w.x ft1, a0
> -; RV64IF-NEXT: feq.s a0, ft1, ft0
> -; RV64IF-NEXT: bnez a0, .LBB16_2
> +; RV64IF-NEXT: feq.s a1, ft1, ft0
> +; RV64IF-NEXT: mv a0, a2
> +; RV64IF-NEXT: bnez a1, .LBB16_2
> ; RV64IF-NEXT: # %bb.1:
> -; RV64IF-NEXT: mv a2, a3
> +; RV64IF-NEXT: mv a0, a3
> ; RV64IF-NEXT: .LBB16_2:
> -; RV64IF-NEXT: mv a0, a2
> ; RV64IF-NEXT: ret
> %1 = fcmp oeq float %a, %b
> %2 = select i1 %1, i32 %c, i32 %d
>
> Modified: llvm/trunk/test/CodeGen/RISCV/fp128.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/fp128.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/fp128.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/fp128.ll Tue Sep 17 02:43:08 2019
> @@ -13,28 +13,28 @@ define i32 @test_load_and_cmp() nounwind
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi sp, sp, -48
> ; RV32I-NEXT: sw ra, 44(sp)
> -; RV32I-NEXT: lui a0, %hi(y)
> -; RV32I-NEXT: lw a1, %lo(y)(a0)
> -; RV32I-NEXT: sw a1, 8(sp)
> -; RV32I-NEXT: lui a1, %hi(x)
> -; RV32I-NEXT: lw a2, %lo(x)(a1)
> -; RV32I-NEXT: sw a2, 24(sp)
> -; RV32I-NEXT: addi a0, a0, %lo(y)
> -; RV32I-NEXT: lw a2, 12(a0)
> -; RV32I-NEXT: sw a2, 20(sp)
> -; RV32I-NEXT: lw a2, 8(a0)
> -; RV32I-NEXT: sw a2, 16(sp)
> -; RV32I-NEXT: lw a0, 4(a0)
> -; RV32I-NEXT: sw a0, 12(sp)
> -; RV32I-NEXT: addi a0, a1, %lo(x)
> -; RV32I-NEXT: lw a1, 12(a0)
> +; RV32I-NEXT: lui a0, %hi(x)
> +; RV32I-NEXT: addi a1, a0, %lo(x)
> +; RV32I-NEXT: lw a6, 4(a1)
> +; RV32I-NEXT: lw a7, 8(a1)
> +; RV32I-NEXT: lw a1, 12(a1)
> +; RV32I-NEXT: lw a0, %lo(x)(a0)
> +; RV32I-NEXT: lui a4, %hi(y)
> +; RV32I-NEXT: addi a5, a4, %lo(y)
> +; RV32I-NEXT: lw a2, 4(a5)
> +; RV32I-NEXT: lw a3, 8(a5)
> +; RV32I-NEXT: lw a5, 12(a5)
> +; RV32I-NEXT: lw a4, %lo(y)(a4)
> +; RV32I-NEXT: sw a4, 8(sp)
> +; RV32I-NEXT: sw a0, 24(sp)
> +; RV32I-NEXT: sw a5, 20(sp)
> +; RV32I-NEXT: sw a3, 16(sp)
> +; RV32I-NEXT: sw a2, 12(sp)
> ; RV32I-NEXT: sw a1, 36(sp)
> -; RV32I-NEXT: lw a1, 8(a0)
> -; RV32I-NEXT: sw a1, 32(sp)
> -; RV32I-NEXT: lw a0, 4(a0)
> -; RV32I-NEXT: sw a0, 28(sp)
> +; RV32I-NEXT: sw a7, 32(sp)
> ; RV32I-NEXT: addi a0, sp, 24
> ; RV32I-NEXT: addi a1, sp, 8
> +; RV32I-NEXT: sw a6, 28(sp)
> ; RV32I-NEXT: call __netf2
> ; RV32I-NEXT: snez a0, a0
> ; RV32I-NEXT: lw ra, 44(sp)
> @@ -52,39 +52,39 @@ define i32 @test_add_and_fptosi() nounwi
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi sp, sp, -80
> ; RV32I-NEXT: sw ra, 76(sp)
> -; RV32I-NEXT: lui a0, %hi(y)
> -; RV32I-NEXT: lw a1, %lo(y)(a0)
> -; RV32I-NEXT: sw a1, 24(sp)
> -; RV32I-NEXT: lui a1, %hi(x)
> -; RV32I-NEXT: lw a2, %lo(x)(a1)
> -; RV32I-NEXT: sw a2, 40(sp)
> -; RV32I-NEXT: addi a0, a0, %lo(y)
> -; RV32I-NEXT: lw a2, 12(a0)
> -; RV32I-NEXT: sw a2, 36(sp)
> -; RV32I-NEXT: lw a2, 8(a0)
> +; RV32I-NEXT: lui a0, %hi(x)
> +; RV32I-NEXT: addi a1, a0, %lo(x)
> +; RV32I-NEXT: lw a6, 4(a1)
> +; RV32I-NEXT: lw a7, 8(a1)
> +; RV32I-NEXT: lw a1, 12(a1)
> +; RV32I-NEXT: lw a0, %lo(x)(a0)
> +; RV32I-NEXT: lui a4, %hi(y)
> +; RV32I-NEXT: addi a5, a4, %lo(y)
> +; RV32I-NEXT: lw a3, 4(a5)
> +; RV32I-NEXT: lw a2, 8(a5)
> +; RV32I-NEXT: lw a5, 12(a5)
> +; RV32I-NEXT: lw a4, %lo(y)(a4)
> +; RV32I-NEXT: sw a4, 24(sp)
> +; RV32I-NEXT: sw a0, 40(sp)
> +; RV32I-NEXT: sw a5, 36(sp)
> ; RV32I-NEXT: sw a2, 32(sp)
> -; RV32I-NEXT: lw a0, 4(a0)
> -; RV32I-NEXT: sw a0, 28(sp)
> -; RV32I-NEXT: addi a0, a1, %lo(x)
> -; RV32I-NEXT: lw a1, 12(a0)
> +; RV32I-NEXT: sw a3, 28(sp)
> ; RV32I-NEXT: sw a1, 52(sp)
> -; RV32I-NEXT: lw a1, 8(a0)
> -; RV32I-NEXT: sw a1, 48(sp)
> -; RV32I-NEXT: lw a0, 4(a0)
> -; RV32I-NEXT: sw a0, 44(sp)
> +; RV32I-NEXT: sw a7, 48(sp)
> ; RV32I-NEXT: addi a0, sp, 56
> ; RV32I-NEXT: addi a1, sp, 40
> ; RV32I-NEXT: addi a2, sp, 24
> +; RV32I-NEXT: sw a6, 44(sp)
> ; RV32I-NEXT: call __addtf3
> -; RV32I-NEXT: lw a0, 68(sp)
> -; RV32I-NEXT: sw a0, 20(sp)
> -; RV32I-NEXT: lw a0, 64(sp)
> -; RV32I-NEXT: sw a0, 16(sp)
> +; RV32I-NEXT: lw a1, 56(sp)
> ; RV32I-NEXT: lw a0, 60(sp)
> +; RV32I-NEXT: lw a2, 64(sp)
> +; RV32I-NEXT: lw a3, 68(sp)
> +; RV32I-NEXT: sw a3, 20(sp)
> +; RV32I-NEXT: sw a2, 16(sp)
> ; RV32I-NEXT: sw a0, 12(sp)
> -; RV32I-NEXT: lw a0, 56(sp)
> -; RV32I-NEXT: sw a0, 8(sp)
> ; RV32I-NEXT: addi a0, sp, 8
> +; RV32I-NEXT: sw a1, 8(sp)
> ; RV32I-NEXT: call __fixtfsi
> ; RV32I-NEXT: lw ra, 76(sp)
> ; RV32I-NEXT: addi sp, sp, 80
>
> Modified: llvm/trunk/test/CodeGen/RISCV/frame-info.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/frame-info.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/frame-info.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/frame-info.ll Tue Sep 17 02:43:08 2019
> @@ -38,12 +38,12 @@ define void @foo(i32 signext %size) {
> ; RV64-NEXT: .cfi_offset s0, -16
> ; RV64-NEXT: addi s0, sp, 16
> ; RV64-NEXT: .cfi_def_cfa s0, 0
> -; RV64-NEXT: addi a1, zero, 1
> -; RV64-NEXT: slli a1, a1, 33
> -; RV64-NEXT: addi a1, a1, -16
> ; RV64-NEXT: slli a0, a0, 32
> ; RV64-NEXT: srli a0, a0, 32
> ; RV64-NEXT: addi a0, a0, 15
> +; RV64-NEXT: addi a1, zero, 1
> +; RV64-NEXT: slli a1, a1, 33
> +; RV64-NEXT: addi a1, a1, -16
> ; RV64-NEXT: and a0, a0, a1
> ; RV64-NEXT: sub a0, sp, a0
> ; RV64-NEXT: mv sp, a0
>
> Modified: llvm/trunk/test/CodeGen/RISCV/get-setcc-result-type.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/get-setcc-result-type.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/get-setcc-result-type.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/get-setcc-result-type.ll Tue Sep 17 02:43:08 2019
> @@ -6,21 +6,21 @@ define void @getSetCCResultType(<4 x i32
> ; RV32I-LABEL: getSetCCResultType:
> ; RV32I: # %bb.0: # %entry
> ; RV32I-NEXT: lw a1, 12(a0)
> +; RV32I-NEXT: lw a2, 8(a0)
> +; RV32I-NEXT: lw a3, 4(a0)
> +; RV32I-NEXT: lw a4, 0(a0)
> ; RV32I-NEXT: seqz a1, a1
> +; RV32I-NEXT: seqz a2, a2
> +; RV32I-NEXT: seqz a3, a3
> +; RV32I-NEXT: seqz a4, a4
> +; RV32I-NEXT: neg a4, a4
> +; RV32I-NEXT: neg a3, a3
> +; RV32I-NEXT: neg a2, a2
> ; RV32I-NEXT: neg a1, a1
> ; RV32I-NEXT: sw a1, 12(a0)
> -; RV32I-NEXT: lw a1, 8(a0)
> -; RV32I-NEXT: seqz a1, a1
> -; RV32I-NEXT: neg a1, a1
> -; RV32I-NEXT: sw a1, 8(a0)
> -; RV32I-NEXT: lw a1, 4(a0)
> -; RV32I-NEXT: seqz a1, a1
> -; RV32I-NEXT: neg a1, a1
> -; RV32I-NEXT: sw a1, 4(a0)
> -; RV32I-NEXT: lw a1, 0(a0)
> -; RV32I-NEXT: seqz a1, a1
> -; RV32I-NEXT: neg a1, a1
> -; RV32I-NEXT: sw a1, 0(a0)
> +; RV32I-NEXT: sw a2, 8(a0)
> +; RV32I-NEXT: sw a3, 4(a0)
> +; RV32I-NEXT: sw a4, 0(a0)
> ; RV32I-NEXT: ret
> entry:
> %0 = load <4 x i32>, <4 x i32>* %p, align 16
>
> Modified: llvm/trunk/test/CodeGen/RISCV/hoist-global-addr-base.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/hoist-global-addr-base.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/hoist-global-addr-base.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/hoist-global-addr-base.ll Tue Sep 17 02:43:08 2019
> @@ -12,10 +12,10 @@ define dso_local void @multiple_stores()
> ; CHECK: # %bb.0: # %entry
> ; CHECK-NEXT: lui a0, %hi(s)
> ; CHECK-NEXT: addi a0, a0, %lo(s)
> -; CHECK-NEXT: addi a1, zero, 20
> -; CHECK-NEXT: sw a1, 164(a0)
> ; CHECK-NEXT: addi a1, zero, 10
> ; CHECK-NEXT: sw a1, 160(a0)
> +; CHECK-NEXT: addi a1, zero, 20
> +; CHECK-NEXT: sw a1, 164(a0)
> ; CHECK-NEXT: ret
> entry:
> store i32 10, i32* getelementptr inbounds (%struct.S, %struct.S* @s, i32 0, i32 1), align 4
>
> Modified: llvm/trunk/test/CodeGen/RISCV/imm-cse.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/imm-cse.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/imm-cse.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/imm-cse.ll Tue Sep 17 02:43:08 2019
> @@ -10,19 +10,19 @@
> define void @imm32_cse() nounwind {
> ; RV32I-LABEL: imm32_cse:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: lui a0, 1
> -; RV32I-NEXT: addi a0, a0, 1
> -; RV32I-NEXT: lui a1, %hi(src)
> -; RV32I-NEXT: lw a2, %lo(src)(a1)
> -; RV32I-NEXT: add a2, a2, a0
> -; RV32I-NEXT: lui a3, %hi(dst)
> -; RV32I-NEXT: sw a2, %lo(dst)(a3)
> -; RV32I-NEXT: lw a2, %lo(src)(a1)
> -; RV32I-NEXT: add a2, a2, a0
> +; RV32I-NEXT: lui a0, %hi(src)
> +; RV32I-NEXT: lw a1, %lo(src)(a0)
> +; RV32I-NEXT: lui a2, 1
> ; RV32I-NEXT: addi a2, a2, 1
> -; RV32I-NEXT: sw a2, %lo(dst)(a3)
> -; RV32I-NEXT: lw a1, %lo(src)(a1)
> -; RV32I-NEXT: add a0, a1, a0
> +; RV32I-NEXT: add a1, a1, a2
> +; RV32I-NEXT: lui a3, %hi(dst)
> +; RV32I-NEXT: sw a1, %lo(dst)(a3)
> +; RV32I-NEXT: lw a1, %lo(src)(a0)
> +; RV32I-NEXT: add a1, a1, a2
> +; RV32I-NEXT: addi a1, a1, 1
> +; RV32I-NEXT: sw a1, %lo(dst)(a3)
> +; RV32I-NEXT: lw a0, %lo(src)(a0)
> +; RV32I-NEXT: add a0, a0, a2
> ; RV32I-NEXT: addi a0, a0, 2
> ; RV32I-NEXT: sw a0, %lo(dst)(a3)
> ; RV32I-NEXT: ret
>
> Modified: llvm/trunk/test/CodeGen/RISCV/imm.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/imm.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/imm.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/imm.ll Tue Sep 17 02:43:08 2019
> @@ -142,8 +142,8 @@ define i64 @imm64_2() nounwind {
> define i64 @imm64_3() nounwind {
> ; RV32I-LABEL: imm64_3:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: mv a0, zero
> ; RV32I-NEXT: addi a1, zero, 1
> +; RV32I-NEXT: mv a0, zero
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: imm64_3:
> @@ -157,8 +157,8 @@ define i64 @imm64_3() nounwind {
> define i64 @imm64_4() nounwind {
> ; RV32I-LABEL: imm64_4:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: mv a0, zero
> ; RV32I-NEXT: lui a1, 524288
> +; RV32I-NEXT: mv a0, zero
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: imm64_4:
> @@ -172,8 +172,8 @@ define i64 @imm64_4() nounwind {
> define i64 @imm64_5() nounwind {
> ; RV32I-LABEL: imm64_5:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: mv a0, zero
> ; RV32I-NEXT: lui a1, 524288
> +; RV32I-NEXT: mv a0, zero
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: imm64_5:
> @@ -249,7 +249,7 @@ define i64 @imm64_9() nounwind {
> ; RV32I-LABEL: imm64_9:
> ; RV32I: # %bb.0:
> ; RV32I-NEXT: addi a0, zero, -1
> -; RV32I-NEXT: addi a1, zero, -1
> +; RV32I-NEXT: addi a1, zero, -1
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: imm64_9:
>
> Modified: llvm/trunk/test/CodeGen/RISCV/indirectbr.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/indirectbr.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/indirectbr.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/indirectbr.ll Tue Sep 17 02:43:08 2019
> @@ -8,7 +8,7 @@ define i32 @indirectbr(i8* %target) noun
> ; RV32I-NEXT: addi sp, sp, -16
> ; RV32I-NEXT: sw ra, 12(sp)
> ; RV32I-NEXT: jr a0
> -; RV32I-NEXT: .LBB0_1:
> +; RV32I-NEXT: .LBB0_1: # %test_label
> ; RV32I-NEXT: mv a0, zero
> ; RV32I-NEXT: lw ra, 12(sp)
> ; RV32I-NEXT: addi sp, sp, 16
> @@ -26,7 +26,7 @@ define i32 @indirectbr_with_offset(i8* %
> ; RV32I-NEXT: addi sp, sp, -16
> ; RV32I-NEXT: sw ra, 12(sp)
> ; RV32I-NEXT: jr 1380(a0)
> -; RV32I-NEXT: .LBB1_1:
> +; RV32I-NEXT: .LBB1_1: # %test_label
> ; RV32I-NEXT: mv a0, zero
> ; RV32I-NEXT: lw ra, 12(sp)
> ; RV32I-NEXT: addi sp, sp, 16
>
> Modified: llvm/trunk/test/CodeGen/RISCV/inline-asm-d-constraint-f.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/inline-asm-d-constraint-f.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/inline-asm-d-constraint-f.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/inline-asm-d-constraint-f.ll Tue Sep 17 02:43:08 2019
> @@ -26,11 +26,11 @@ define double @constraint_f_double(doubl
> ;
> ; RV64F-LABEL: constraint_f_double:
> ; RV64F: # %bb.0:
> -; RV64F-NEXT: fmv.d.x ft0, a0
> -; RV64F-NEXT: lui a0, %hi(gd)
> -; RV64F-NEXT: fld ft1, %lo(gd)(a0)
> +; RV64F-NEXT: lui a1, %hi(gd)
> +; RV64F-NEXT: fld ft0, %lo(gd)(a1)
> +; RV64F-NEXT: fmv.d.x ft1, a0
> ; RV64F-NEXT: #APP
> -; RV64F-NEXT: fadd.d ft0, ft0, ft1
> +; RV64F-NEXT: fadd.d ft0, ft1, ft0
> ; RV64F-NEXT: #NO_APP
> ; RV64F-NEXT: fmv.x.d a0, ft0
> ; RV64F-NEXT: ret
> @@ -59,9 +59,9 @@ define double @constraint_f_double_abi_n
> ;
> ; RV64F-LABEL: constraint_f_double_abi_name:
> ; RV64F: # %bb.0:
> +; RV64F-NEXT: lui a1, %hi(gd)
> +; RV64F-NEXT: fld fs0, %lo(gd)(a1)
> ; RV64F-NEXT: fmv.d.x fa1, a0
> -; RV64F-NEXT: lui a0, %hi(gd)
> -; RV64F-NEXT: fld fs0, %lo(gd)(a0)
> ; RV64F-NEXT: #APP
> ; RV64F-NEXT: fadd.d ft0, fa1, fs0
> ; RV64F-NEXT: #NO_APP
>
> Modified: llvm/trunk/test/CodeGen/RISCV/inline-asm-f-constraint-f.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/inline-asm-f-constraint-f.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/inline-asm-f-constraint-f.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/inline-asm-f-constraint-f.ll Tue Sep 17 02:43:08 2019
> @@ -9,22 +9,22 @@
> define float @constraint_f_float(float %a) nounwind {
> ; RV32F-LABEL: constraint_f_float:
> ; RV32F: # %bb.0:
> -; RV32F-NEXT: fmv.w.x ft0, a0
> -; RV32F-NEXT: lui a0, %hi(gf)
> -; RV32F-NEXT: flw ft1, %lo(gf)(a0)
> +; RV32F-NEXT: lui a1, %hi(gf)
> +; RV32F-NEXT: flw ft0, %lo(gf)(a1)
> +; RV32F-NEXT: fmv.w.x ft1, a0
> ; RV32F-NEXT: #APP
> -; RV32F-NEXT: fadd.s ft0, ft0, ft1
> +; RV32F-NEXT: fadd.s ft0, ft1, ft0
> ; RV32F-NEXT: #NO_APP
> ; RV32F-NEXT: fmv.x.w a0, ft0
> ; RV32F-NEXT: ret
> ;
> ; RV64F-LABEL: constraint_f_float:
> ; RV64F: # %bb.0:
> -; RV64F-NEXT: fmv.w.x ft0, a0
> -; RV64F-NEXT: lui a0, %hi(gf)
> -; RV64F-NEXT: flw ft1, %lo(gf)(a0)
> +; RV64F-NEXT: lui a1, %hi(gf)
> +; RV64F-NEXT: flw ft0, %lo(gf)(a1)
> +; RV64F-NEXT: fmv.w.x ft1, a0
> ; RV64F-NEXT: #APP
> -; RV64F-NEXT: fadd.s ft0, ft0, ft1
> +; RV64F-NEXT: fadd.s ft0, ft1, ft0
> ; RV64F-NEXT: #NO_APP
> ; RV64F-NEXT: fmv.x.w a0, ft0
> ; RV64F-NEXT: ret
> @@ -36,9 +36,9 @@ define float @constraint_f_float(float %
> define float @constraint_f_float_abi_name(float %a) nounwind {
> ; RV32F-LABEL: constraint_f_float_abi_name:
> ; RV32F: # %bb.0:
> +; RV32F-NEXT: lui a1, %hi(gf)
> +; RV32F-NEXT: flw fs0, %lo(gf)(a1)
> ; RV32F-NEXT: fmv.w.x fa0, a0
> -; RV32F-NEXT: lui a0, %hi(gf)
> -; RV32F-NEXT: flw fs0, %lo(gf)(a0)
> ; RV32F-NEXT: #APP
> ; RV32F-NEXT: fadd.s ft0, fa0, fs0
> ; RV32F-NEXT: #NO_APP
> @@ -47,9 +47,9 @@ define float @constraint_f_float_abi_nam
> ;
> ; RV64F-LABEL: constraint_f_float_abi_name:
> ; RV64F: # %bb.0:
> +; RV64F-NEXT: lui a1, %hi(gf)
> +; RV64F-NEXT: flw fs0, %lo(gf)(a1)
> ; RV64F-NEXT: fmv.w.x fa0, a0
> -; RV64F-NEXT: lui a0, %hi(gf)
> -; RV64F-NEXT: flw fs0, %lo(gf)(a0)
> ; RV64F-NEXT: #APP
> ; RV64F-NEXT: fadd.s ft0, fa0, fs0
> ; RV64F-NEXT: #NO_APP
>
> Modified: llvm/trunk/test/CodeGen/RISCV/interrupt-attr-nocall.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/interrupt-attr-nocall.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/interrupt-attr-nocall.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/interrupt-attr-nocall.ll Tue Sep 17 02:43:08 2019
> @@ -213,11 +213,11 @@ define void @foo_float() nounwind #0 {
> ; CHECK-RV32IF-NEXT: sw a0, 12(sp)
> ; CHECK-RV32IF-NEXT: fsw ft0, 8(sp)
> ; CHECK-RV32IF-NEXT: fsw ft1, 4(sp)
> -; CHECK-RV32IF-NEXT: lui a0, %hi(f)
> -; CHECK-RV32IF-NEXT: flw ft0, %lo(f)(a0)
> ; CHECK-RV32IF-NEXT: lui a0, %hi(e)
> -; CHECK-RV32IF-NEXT: flw ft1, %lo(e)(a0)
> -; CHECK-RV32IF-NEXT: fadd.s ft0, ft1, ft0
> +; CHECK-RV32IF-NEXT: flw ft0, %lo(e)(a0)
> +; CHECK-RV32IF-NEXT: lui a0, %hi(f)
> +; CHECK-RV32IF-NEXT: flw ft1, %lo(f)(a0)
> +; CHECK-RV32IF-NEXT: fadd.s ft0, ft0, ft1
> ; CHECK-RV32IF-NEXT: lui a0, %hi(d)
> ; CHECK-RV32IF-NEXT: fsw ft0, %lo(d)(a0)
> ; CHECK-RV32IF-NEXT: flw ft1, 4(sp)
> @@ -232,11 +232,11 @@ define void @foo_float() nounwind #0 {
> ; CHECK-RV32IFD-NEXT: sw a0, 28(sp)
> ; CHECK-RV32IFD-NEXT: fsd ft0, 16(sp)
> ; CHECK-RV32IFD-NEXT: fsd ft1, 8(sp)
> -; CHECK-RV32IFD-NEXT: lui a0, %hi(f)
> -; CHECK-RV32IFD-NEXT: flw ft0, %lo(f)(a0)
> ; CHECK-RV32IFD-NEXT: lui a0, %hi(e)
> -; CHECK-RV32IFD-NEXT: flw ft1, %lo(e)(a0)
> -; CHECK-RV32IFD-NEXT: fadd.s ft0, ft1, ft0
> +; CHECK-RV32IFD-NEXT: flw ft0, %lo(e)(a0)
> +; CHECK-RV32IFD-NEXT: lui a0, %hi(f)
> +; CHECK-RV32IFD-NEXT: flw ft1, %lo(f)(a0)
> +; CHECK-RV32IFD-NEXT: fadd.s ft0, ft0, ft1
> ; CHECK-RV32IFD-NEXT: lui a0, %hi(d)
> ; CHECK-RV32IFD-NEXT: fsw ft0, %lo(d)(a0)
> ; CHECK-RV32IFD-NEXT: fld ft1, 8(sp)
> @@ -312,11 +312,11 @@ define void @foo_fp_float() nounwind #1
> ; CHECK-RV32IF-NEXT: fsw ft0, 16(sp)
> ; CHECK-RV32IF-NEXT: fsw ft1, 12(sp)
> ; CHECK-RV32IF-NEXT: addi s0, sp, 32
> -; CHECK-RV32IF-NEXT: lui a0, %hi(f)
> -; CHECK-RV32IF-NEXT: flw ft0, %lo(f)(a0)
> ; CHECK-RV32IF-NEXT: lui a0, %hi(e)
> -; CHECK-RV32IF-NEXT: flw ft1, %lo(e)(a0)
> -; CHECK-RV32IF-NEXT: fadd.s ft0, ft1, ft0
> +; CHECK-RV32IF-NEXT: flw ft0, %lo(e)(a0)
> +; CHECK-RV32IF-NEXT: lui a0, %hi(f)
> +; CHECK-RV32IF-NEXT: flw ft1, %lo(f)(a0)
> +; CHECK-RV32IF-NEXT: fadd.s ft0, ft0, ft1
> ; CHECK-RV32IF-NEXT: lui a0, %hi(d)
> ; CHECK-RV32IF-NEXT: fsw ft0, %lo(d)(a0)
> ; CHECK-RV32IF-NEXT: flw ft1, 12(sp)
> @@ -336,11 +336,11 @@ define void @foo_fp_float() nounwind #1
> ; CHECK-RV32IFD-NEXT: fsd ft0, 8(sp)
> ; CHECK-RV32IFD-NEXT: fsd ft1, 0(sp)
> ; CHECK-RV32IFD-NEXT: addi s0, sp, 32
> -; CHECK-RV32IFD-NEXT: lui a0, %hi(f)
> -; CHECK-RV32IFD-NEXT: flw ft0, %lo(f)(a0)
> ; CHECK-RV32IFD-NEXT: lui a0, %hi(e)
> -; CHECK-RV32IFD-NEXT: flw ft1, %lo(e)(a0)
> -; CHECK-RV32IFD-NEXT: fadd.s ft0, ft1, ft0
> +; CHECK-RV32IFD-NEXT: flw ft0, %lo(e)(a0)
> +; CHECK-RV32IFD-NEXT: lui a0, %hi(f)
> +; CHECK-RV32IFD-NEXT: flw ft1, %lo(f)(a0)
> +; CHECK-RV32IFD-NEXT: fadd.s ft0, ft0, ft1
> ; CHECK-RV32IFD-NEXT: lui a0, %hi(d)
> ; CHECK-RV32IFD-NEXT: fsw ft0, %lo(d)(a0)
> ; CHECK-RV32IFD-NEXT: fld ft1, 0(sp)
> @@ -534,11 +534,11 @@ define void @foo_double() nounwind #0 {
> ; CHECK-RV32IFD-NEXT: sw a0, 28(sp)
> ; CHECK-RV32IFD-NEXT: fsd ft0, 16(sp)
> ; CHECK-RV32IFD-NEXT: fsd ft1, 8(sp)
> -; CHECK-RV32IFD-NEXT: lui a0, %hi(i)
> -; CHECK-RV32IFD-NEXT: fld ft0, %lo(i)(a0)
> ; CHECK-RV32IFD-NEXT: lui a0, %hi(h)
> -; CHECK-RV32IFD-NEXT: fld ft1, %lo(h)(a0)
> -; CHECK-RV32IFD-NEXT: fadd.d ft0, ft1, ft0
> +; CHECK-RV32IFD-NEXT: fld ft0, %lo(h)(a0)
> +; CHECK-RV32IFD-NEXT: lui a0, %hi(i)
> +; CHECK-RV32IFD-NEXT: fld ft1, %lo(i)(a0)
> +; CHECK-RV32IFD-NEXT: fadd.d ft0, ft0, ft1
> ; CHECK-RV32IFD-NEXT: lui a0, %hi(g)
> ; CHECK-RV32IFD-NEXT: fsd ft0, %lo(g)(a0)
> ; CHECK-RV32IFD-NEXT: fld ft1, 8(sp)
> @@ -738,11 +738,11 @@ define void @foo_fp_double() nounwind #1
> ; CHECK-RV32IFD-NEXT: fsd ft0, 8(sp)
> ; CHECK-RV32IFD-NEXT: fsd ft1, 0(sp)
> ; CHECK-RV32IFD-NEXT: addi s0, sp, 32
> -; CHECK-RV32IFD-NEXT: lui a0, %hi(i)
> -; CHECK-RV32IFD-NEXT: fld ft0, %lo(i)(a0)
> ; CHECK-RV32IFD-NEXT: lui a0, %hi(h)
> -; CHECK-RV32IFD-NEXT: fld ft1, %lo(h)(a0)
> -; CHECK-RV32IFD-NEXT: fadd.d ft0, ft1, ft0
> +; CHECK-RV32IFD-NEXT: fld ft0, %lo(h)(a0)
> +; CHECK-RV32IFD-NEXT: lui a0, %hi(i)
> +; CHECK-RV32IFD-NEXT: fld ft1, %lo(i)(a0)
> +; CHECK-RV32IFD-NEXT: fadd.d ft0, ft0, ft1
> ; CHECK-RV32IFD-NEXT: lui a0, %hi(g)
> ; CHECK-RV32IFD-NEXT: fsd ft0, %lo(g)(a0)
> ; CHECK-RV32IFD-NEXT: fld ft1, 0(sp)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/legalize-fneg.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/legalize-fneg.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/legalize-fneg.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/legalize-fneg.ll Tue Sep 17 02:43:08 2019
> @@ -15,9 +15,9 @@ define void @test1(float* %a, float* %b)
> ;
> ; RV64-LABEL: test1:
> ; RV64: # %bb.0: # %entry
> +; RV64-NEXT: lw a1, 0(a1)
> ; RV64-NEXT: addi a2, zero, 1
> ; RV64-NEXT: slli a2, a2, 31
> -; RV64-NEXT: lw a1, 0(a1)
> ; RV64-NEXT: xor a1, a1, a2
> ; RV64-NEXT: sw a1, 0(a0)
> ; RV64-NEXT: ret
> @@ -33,17 +33,17 @@ define void @test2(double* %a, double* %
> ; RV32: # %bb.0: # %entry
> ; RV32-NEXT: lw a2, 4(a1)
> ; RV32-NEXT: lw a1, 0(a1)
> +; RV32-NEXT: lui a3, 524288
> +; RV32-NEXT: xor a2, a2, a3
> ; RV32-NEXT: sw a1, 0(a0)
> -; RV32-NEXT: lui a1, 524288
> -; RV32-NEXT: xor a1, a2, a1
> -; RV32-NEXT: sw a1, 4(a0)
> +; RV32-NEXT: sw a2, 4(a0)
> ; RV32-NEXT: ret
> ;
> ; RV64-LABEL: test2:
> ; RV64: # %bb.0: # %entry
> +; RV64-NEXT: ld a1, 0(a1)
> ; RV64-NEXT: addi a2, zero, -1
> ; RV64-NEXT: slli a2, a2, 63
> -; RV64-NEXT: ld a1, 0(a1)
> ; RV64-NEXT: xor a1, a1, a2
> ; RV64-NEXT: sd a1, 0(a0)
> ; RV64-NEXT: ret
> @@ -57,27 +57,27 @@ entry:
> define void @test3(fp128* %a, fp128* %b) nounwind {
> ; RV32-LABEL: test3:
> ; RV32: # %bb.0: # %entry
> -; RV32-NEXT: lw a2, 12(a1)
> -; RV32-NEXT: lw a3, 4(a1)
> -; RV32-NEXT: lw a4, 0(a1)
> -; RV32-NEXT: lw a1, 8(a1)
> -; RV32-NEXT: sw a1, 8(a0)
> -; RV32-NEXT: sw a4, 0(a0)
> -; RV32-NEXT: sw a3, 4(a0)
> -; RV32-NEXT: lui a1, 524288
> -; RV32-NEXT: xor a1, a2, a1
> -; RV32-NEXT: sw a1, 12(a0)
> +; RV32-NEXT: lw a2, 4(a1)
> +; RV32-NEXT: lw a3, 12(a1)
> +; RV32-NEXT: lw a4, 8(a1)
> +; RV32-NEXT: lw a1, 0(a1)
> +; RV32-NEXT: lui a5, 524288
> +; RV32-NEXT: xor a3, a3, a5
> +; RV32-NEXT: sw a4, 8(a0)
> +; RV32-NEXT: sw a1, 0(a0)
> +; RV32-NEXT: sw a2, 4(a0)
> +; RV32-NEXT: sw a3, 12(a0)
> ; RV32-NEXT: ret
> ;
> ; RV64-LABEL: test3:
> ; RV64: # %bb.0: # %entry
> ; RV64-NEXT: ld a2, 8(a1)
> ; RV64-NEXT: ld a1, 0(a1)
> +; RV64-NEXT: addi a3, zero, -1
> +; RV64-NEXT: slli a3, a3, 63
> +; RV64-NEXT: xor a2, a2, a3
> ; RV64-NEXT: sd a1, 0(a0)
> -; RV64-NEXT: addi a1, zero, -1
> -; RV64-NEXT: slli a1, a1, 63
> -; RV64-NEXT: xor a1, a2, a1
> -; RV64-NEXT: sd a1, 8(a0)
> +; RV64-NEXT: sd a2, 8(a0)
> ; RV64-NEXT: ret
> entry:
> %0 = load fp128, fp128* %b
>
> Modified: llvm/trunk/test/CodeGen/RISCV/lsr-legaladdimm.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/lsr-legaladdimm.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/lsr-legaladdimm.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/lsr-legaladdimm.ll Tue Sep 17 02:43:08 2019
> @@ -11,21 +11,21 @@
> define i32 @main() nounwind {
> ; RV32I-LABEL: main:
> ; RV32I: # %bb.0: # %entry
> -; RV32I-NEXT: lui a0, %hi(b)
> -; RV32I-NEXT: addi a0, a0, %lo(b)
> -; RV32I-NEXT: lui a1, %hi(a)
> -; RV32I-NEXT: addi a1, a1, %lo(a)
> -; RV32I-NEXT: lui a2, 1
> -; RV32I-NEXT: mv a3, zero
> +; RV32I-NEXT: mv a0, zero
> +; RV32I-NEXT: lui a1, %hi(b)
> +; RV32I-NEXT: addi a1, a1, %lo(b)
> +; RV32I-NEXT: lui a2, %hi(a)
> +; RV32I-NEXT: addi a2, a2, %lo(a)
> +; RV32I-NEXT: lui a3, 1
> ; RV32I-NEXT: .LBB0_1: # %for.body
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: addi a4, a3, -2048
> -; RV32I-NEXT: sw a4, 0(a1)
> +; RV32I-NEXT: addi a4, a0, -2048
> +; RV32I-NEXT: sw a4, 0(a2)
> +; RV32I-NEXT: sw a0, 0(a1)
> +; RV32I-NEXT: addi a0, a0, 1
> ; RV32I-NEXT: addi a1, a1, 4
> -; RV32I-NEXT: sw a3, 0(a0)
> -; RV32I-NEXT: addi a0, a0, 4
> -; RV32I-NEXT: addi a3, a3, 1
> -; RV32I-NEXT: bne a3, a2, .LBB0_1
> +; RV32I-NEXT: addi a2, a2, 4
> +; RV32I-NEXT: bne a0, a3, .LBB0_1
> ; RV32I-NEXT: # %bb.2: # %for.end
> ; RV32I-NEXT: mv a0, zero
> ; RV32I-NEXT: ret
>
> Modified: llvm/trunk/test/CodeGen/RISCV/mem.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/mem.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/mem.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/mem.ll Tue Sep 17 02:43:08 2019
> @@ -7,8 +7,9 @@
> define i32 @lb(i8 *%a) nounwind {
> ; RV32I-LABEL: lb:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: lb a1, 0(a0)
> -; RV32I-NEXT: lb a0, 1(a0)
> +; RV32I-NEXT: lb a1, 1(a0)
> +; RV32I-NEXT: lb a0, 0(a0)
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: ret
> %1 = getelementptr i8, i8* %a, i32 1
> %2 = load i8, i8* %1
> @@ -21,8 +22,9 @@ define i32 @lb(i8 *%a) nounwind {
> define i32 @lh(i16 *%a) nounwind {
> ; RV32I-LABEL: lh:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: lh a1, 0(a0)
> -; RV32I-NEXT: lh a0, 4(a0)
> +; RV32I-NEXT: lh a1, 4(a0)
> +; RV32I-NEXT: lh a0, 0(a0)
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: ret
> %1 = getelementptr i16, i16* %a, i32 2
> %2 = load i16, i16* %1
> @@ -35,8 +37,9 @@ define i32 @lh(i16 *%a) nounwind {
> define i32 @lw(i32 *%a) nounwind {
> ; RV32I-LABEL: lw:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: lw a1, 0(a0)
> -; RV32I-NEXT: lw a0, 12(a0)
> +; RV32I-NEXT: lw a1, 12(a0)
> +; RV32I-NEXT: lw a0, 0(a0)
> +; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: ret
> %1 = getelementptr i32, i32* %a, i32 3
> %2 = load i32, i32* %1
> @@ -47,9 +50,9 @@ define i32 @lw(i32 *%a) nounwind {
> define i32 @lbu(i8 *%a) nounwind {
> ; RV32I-LABEL: lbu:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: lbu a1, 0(a0)
> -; RV32I-NEXT: lbu a0, 4(a0)
> -; RV32I-NEXT: add a0, a0, a1
> +; RV32I-NEXT: lbu a1, 4(a0)
> +; RV32I-NEXT: lbu a0, 0(a0)
> +; RV32I-NEXT: add a0, a1, a0
> ; RV32I-NEXT: ret
> %1 = getelementptr i8, i8* %a, i32 4
> %2 = load i8, i8* %1
> @@ -63,9 +66,9 @@ define i32 @lbu(i8 *%a) nounwind {
> define i32 @lhu(i16 *%a) nounwind {
> ; RV32I-LABEL: lhu:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: lhu a1, 0(a0)
> -; RV32I-NEXT: lhu a0, 10(a0)
> -; RV32I-NEXT: add a0, a0, a1
> +; RV32I-NEXT: lhu a1, 10(a0)
> +; RV32I-NEXT: lhu a0, 0(a0)
> +; RV32I-NEXT: add a0, a1, a0
> ; RV32I-NEXT: ret
> %1 = getelementptr i16, i16* %a, i32 5
> %2 = load i16, i16* %1
> @@ -81,8 +84,8 @@ define i32 @lhu(i16 *%a) nounwind {
> define void @sb(i8 *%a, i8 %b) nounwind {
> ; RV32I-LABEL: sb:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: sb a1, 6(a0)
> ; RV32I-NEXT: sb a1, 0(a0)
> +; RV32I-NEXT: sb a1, 6(a0)
> ; RV32I-NEXT: ret
> store i8 %b, i8* %a
> %1 = getelementptr i8, i8* %a, i32 6
> @@ -93,8 +96,8 @@ define void @sb(i8 *%a, i8 %b) nounwind
> define void @sh(i16 *%a, i16 %b) nounwind {
> ; RV32I-LABEL: sh:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: sh a1, 14(a0)
> ; RV32I-NEXT: sh a1, 0(a0)
> +; RV32I-NEXT: sh a1, 14(a0)
> ; RV32I-NEXT: ret
> store i16 %b, i16* %a
> %1 = getelementptr i16, i16* %a, i32 7
> @@ -105,8 +108,8 @@ define void @sh(i16 *%a, i16 %b) nounwin
> define void @sw(i32 *%a, i32 %b) nounwind {
> ; RV32I-LABEL: sw:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: sw a1, 32(a0)
> ; RV32I-NEXT: sw a1, 0(a0)
> +; RV32I-NEXT: sw a1, 32(a0)
> ; RV32I-NEXT: ret
> store i32 %b, i32* %a
> %1 = getelementptr i32, i32* %a, i32 8
> @@ -118,10 +121,10 @@ define void @sw(i32 *%a, i32 %b) nounwin
> define i32 @load_sext_zext_anyext_i1(i1 *%a) nounwind {
> ; RV32I-LABEL: load_sext_zext_anyext_i1:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: lb a1, 0(a0)
> ; RV32I-NEXT: lbu a1, 1(a0)
> -; RV32I-NEXT: lbu a0, 2(a0)
> -; RV32I-NEXT: sub a0, a0, a1
> +; RV32I-NEXT: lbu a2, 2(a0)
> +; RV32I-NEXT: lb a0, 0(a0)
> +; RV32I-NEXT: sub a0, a2, a1
> ; RV32I-NEXT: ret
> ; sextload i1
> %1 = getelementptr i1, i1* %a, i32 1
> @@ -140,10 +143,10 @@ define i32 @load_sext_zext_anyext_i1(i1
> define i16 @load_sext_zext_anyext_i1_i16(i1 *%a) nounwind {
> ; RV32I-LABEL: load_sext_zext_anyext_i1_i16:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: lb a1, 0(a0)
> ; RV32I-NEXT: lbu a1, 1(a0)
> -; RV32I-NEXT: lbu a0, 2(a0)
> -; RV32I-NEXT: sub a0, a0, a1
> +; RV32I-NEXT: lbu a2, 2(a0)
> +; RV32I-NEXT: lb a0, 0(a0)
> +; RV32I-NEXT: sub a0, a2, a1
> ; RV32I-NEXT: ret
> ; sextload i1
> %1 = getelementptr i1, i1* %a, i32 1
>
> Modified: llvm/trunk/test/CodeGen/RISCV/mem64.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/mem64.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/mem64.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/mem64.ll Tue Sep 17 02:43:08 2019
> @@ -7,8 +7,9 @@
> define i64 @lb(i8 *%a) nounwind {
> ; RV64I-LABEL: lb:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: lb a1, 0(a0)
> -; RV64I-NEXT: lb a0, 1(a0)
> +; RV64I-NEXT: lb a1, 1(a0)
> +; RV64I-NEXT: lb a0, 0(a0)
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ret
> %1 = getelementptr i8, i8* %a, i32 1
> %2 = load i8, i8* %1
> @@ -21,8 +22,9 @@ define i64 @lb(i8 *%a) nounwind {
> define i64 @lh(i16 *%a) nounwind {
> ; RV64I-LABEL: lh:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: lh a1, 0(a0)
> -; RV64I-NEXT: lh a0, 4(a0)
> +; RV64I-NEXT: lh a1, 4(a0)
> +; RV64I-NEXT: lh a0, 0(a0)
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ret
> %1 = getelementptr i16, i16* %a, i32 2
> %2 = load i16, i16* %1
> @@ -35,8 +37,9 @@ define i64 @lh(i16 *%a) nounwind {
> define i64 @lw(i32 *%a) nounwind {
> ; RV64I-LABEL: lw:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: lw a1, 0(a0)
> -; RV64I-NEXT: lw a0, 12(a0)
> +; RV64I-NEXT: lw a1, 12(a0)
> +; RV64I-NEXT: lw a0, 0(a0)
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ret
> %1 = getelementptr i32, i32* %a, i32 3
> %2 = load i32, i32* %1
> @@ -49,9 +52,9 @@ define i64 @lw(i32 *%a) nounwind {
> define i64 @lbu(i8 *%a) nounwind {
> ; RV64I-LABEL: lbu:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: lbu a1, 0(a0)
> -; RV64I-NEXT: lbu a0, 4(a0)
> -; RV64I-NEXT: add a0, a0, a1
> +; RV64I-NEXT: lbu a1, 4(a0)
> +; RV64I-NEXT: lbu a0, 0(a0)
> +; RV64I-NEXT: add a0, a1, a0
> ; RV64I-NEXT: ret
> %1 = getelementptr i8, i8* %a, i32 4
> %2 = load i8, i8* %1
> @@ -65,9 +68,9 @@ define i64 @lbu(i8 *%a) nounwind {
> define i64 @lhu(i16 *%a) nounwind {
> ; RV64I-LABEL: lhu:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: lhu a1, 0(a0)
> -; RV64I-NEXT: lhu a0, 10(a0)
> -; RV64I-NEXT: add a0, a0, a1
> +; RV64I-NEXT: lhu a1, 10(a0)
> +; RV64I-NEXT: lhu a0, 0(a0)
> +; RV64I-NEXT: add a0, a1, a0
> ; RV64I-NEXT: ret
> %1 = getelementptr i16, i16* %a, i32 5
> %2 = load i16, i16* %1
> @@ -81,9 +84,9 @@ define i64 @lhu(i16 *%a) nounwind {
> define i64 @lwu(i32 *%a) nounwind {
> ; RV64I-LABEL: lwu:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: lwu a1, 0(a0)
> -; RV64I-NEXT: lwu a0, 24(a0)
> -; RV64I-NEXT: add a0, a0, a1
> +; RV64I-NEXT: lwu a1, 24(a0)
> +; RV64I-NEXT: lwu a0, 0(a0)
> +; RV64I-NEXT: add a0, a1, a0
> ; RV64I-NEXT: ret
> %1 = getelementptr i32, i32* %a, i32 6
> %2 = load i32, i32* %1
> @@ -99,8 +102,8 @@ define i64 @lwu(i32 *%a) nounwind {
> define void @sb(i8 *%a, i8 %b) nounwind {
> ; RV64I-LABEL: sb:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: sb a1, 7(a0)
> ; RV64I-NEXT: sb a1, 0(a0)
> +; RV64I-NEXT: sb a1, 7(a0)
> ; RV64I-NEXT: ret
> store i8 %b, i8* %a
> %1 = getelementptr i8, i8* %a, i32 7
> @@ -111,8 +114,8 @@ define void @sb(i8 *%a, i8 %b) nounwind
> define void @sh(i16 *%a, i16 %b) nounwind {
> ; RV64I-LABEL: sh:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: sh a1, 16(a0)
> ; RV64I-NEXT: sh a1, 0(a0)
> +; RV64I-NEXT: sh a1, 16(a0)
> ; RV64I-NEXT: ret
> store i16 %b, i16* %a
> %1 = getelementptr i16, i16* %a, i32 8
> @@ -123,8 +126,8 @@ define void @sh(i16 *%a, i16 %b) nounwin
> define void @sw(i32 *%a, i32 %b) nounwind {
> ; RV64I-LABEL: sw:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: sw a1, 36(a0)
> ; RV64I-NEXT: sw a1, 0(a0)
> +; RV64I-NEXT: sw a1, 36(a0)
> ; RV64I-NEXT: ret
> store i32 %b, i32* %a
> %1 = getelementptr i32, i32* %a, i32 9
> @@ -137,8 +140,9 @@ define void @sw(i32 *%a, i32 %b) nounwin
> define i64 @ld(i64 *%a) nounwind {
> ; RV64I-LABEL: ld:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: ld a1, 0(a0)
> -; RV64I-NEXT: ld a0, 80(a0)
> +; RV64I-NEXT: ld a1, 80(a0)
> +; RV64I-NEXT: ld a0, 0(a0)
> +; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ret
> %1 = getelementptr i64, i64* %a, i32 10
> %2 = load i64, i64* %1
> @@ -149,8 +153,8 @@ define i64 @ld(i64 *%a) nounwind {
> define void @sd(i64 *%a, i64 %b) nounwind {
> ; RV64I-LABEL: sd:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: sd a1, 88(a0)
> ; RV64I-NEXT: sd a1, 0(a0)
> +; RV64I-NEXT: sd a1, 88(a0)
> ; RV64I-NEXT: ret
> store i64 %b, i64* %a
> %1 = getelementptr i64, i64* %a, i32 11
> @@ -162,10 +166,10 @@ define void @sd(i64 *%a, i64 %b) nounwin
> define i64 @load_sext_zext_anyext_i1(i1 *%a) nounwind {
> ; RV64I-LABEL: load_sext_zext_anyext_i1:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: lb a1, 0(a0)
> ; RV64I-NEXT: lbu a1, 1(a0)
> -; RV64I-NEXT: lbu a0, 2(a0)
> -; RV64I-NEXT: sub a0, a0, a1
> +; RV64I-NEXT: lbu a2, 2(a0)
> +; RV64I-NEXT: lb a0, 0(a0)
> +; RV64I-NEXT: sub a0, a2, a1
> ; RV64I-NEXT: ret
> ; sextload i1
> %1 = getelementptr i1, i1* %a, i32 1
> @@ -184,10 +188,10 @@ define i64 @load_sext_zext_anyext_i1(i1
> define i16 @load_sext_zext_anyext_i1_i16(i1 *%a) nounwind {
> ; RV64I-LABEL: load_sext_zext_anyext_i1_i16:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: lb a1, 0(a0)
> ; RV64I-NEXT: lbu a1, 1(a0)
> -; RV64I-NEXT: lbu a0, 2(a0)
> -; RV64I-NEXT: sub a0, a0, a1
> +; RV64I-NEXT: lbu a2, 2(a0)
> +; RV64I-NEXT: lb a0, 0(a0)
> +; RV64I-NEXT: sub a0, a2, a1
> ; RV64I-NEXT: ret
> ; sextload i1
> %1 = getelementptr i1, i1* %a, i32 1
>
> Modified: llvm/trunk/test/CodeGen/RISCV/mul.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/mul.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/mul.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/mul.ll Tue Sep 17 02:43:08 2019
> @@ -247,8 +247,8 @@ define i32 @mulhs(i32 %a, i32 %b) nounwi
> ;
> ; RV64IM-LABEL: mulhs:
> ; RV64IM: # %bb.0:
> -; RV64IM-NEXT: sext.w a1, a1
> ; RV64IM-NEXT: sext.w a0, a0
> +; RV64IM-NEXT: sext.w a1, a1
> ; RV64IM-NEXT: mul a0, a0, a1
> ; RV64IM-NEXT: srli a0, a0, 32
> ; RV64IM-NEXT: ret
>
> Modified: llvm/trunk/test/CodeGen/RISCV/remat.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/remat.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/remat.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/remat.ll Tue Sep 17 02:43:08 2019
> @@ -37,19 +37,19 @@ define i32 @test() nounwind {
> ; RV32I-NEXT: sw s9, 20(sp)
> ; RV32I-NEXT: sw s10, 16(sp)
> ; RV32I-NEXT: sw s11, 12(sp)
> -; RV32I-NEXT: lui s9, %hi(a)
> -; RV32I-NEXT: lw a0, %lo(a)(s9)
> +; RV32I-NEXT: lui s6, %hi(a)
> +; RV32I-NEXT: lw a0, %lo(a)(s6)
> ; RV32I-NEXT: beqz a0, .LBB0_11
> ; RV32I-NEXT: # %bb.1: # %for.body.preheader
> ; RV32I-NEXT: lui s2, %hi(l)
> ; RV32I-NEXT: lui s3, %hi(k)
> ; RV32I-NEXT: lui s4, %hi(j)
> -; RV32I-NEXT: lui s6, %hi(i)
> -; RV32I-NEXT: lui s5, %hi(h)
> -; RV32I-NEXT: lui s7, %hi(g)
> -; RV32I-NEXT: lui s8, %hi(f)
> -; RV32I-NEXT: lui s1, %hi(e)
> -; RV32I-NEXT: lui s0, %hi(d)
> +; RV32I-NEXT: lui s5, %hi(i)
> +; RV32I-NEXT: lui s1, %hi(d)
> +; RV32I-NEXT: lui s0, %hi(e)
> +; RV32I-NEXT: lui s7, %hi(f)
> +; RV32I-NEXT: lui s8, %hi(g)
> +; RV32I-NEXT: lui s9, %hi(h)
> ; RV32I-NEXT: lui s10, %hi(c)
> ; RV32I-NEXT: lui s11, %hi(b)
> ; RV32I-NEXT: lw a1, %lo(l)(s2)
> @@ -57,19 +57,19 @@ define i32 @test() nounwind {
> ; RV32I-NEXT: j .LBB0_5
> ; RV32I-NEXT: .LBB0_2: # %for.inc
> ; RV32I-NEXT: # in Loop: Header=BB0_5 Depth=1
> -; RV32I-NEXT: lw a0, %lo(a)(s9)
> +; RV32I-NEXT: lw a0, %lo(a)(s6)
> ; RV32I-NEXT: addi a0, a0, -1
> -; RV32I-NEXT: sw a0, %lo(a)(s9)
> +; RV32I-NEXT: sw a0, %lo(a)(s6)
> ; RV32I-NEXT: beqz a0, .LBB0_11
> ; RV32I-NEXT: # %bb.3: # %for.body
> ; RV32I-NEXT: # in Loop: Header=BB0_5 Depth=1
> ; RV32I-NEXT: lw a1, %lo(l)(s2)
> ; RV32I-NEXT: beqz a1, .LBB0_5
> ; RV32I-NEXT: .LBB0_4: # %if.then
> -; RV32I-NEXT: lw a4, %lo(e)(s1)
> -; RV32I-NEXT: lw a3, %lo(d)(s0)
> -; RV32I-NEXT: lw a2, %lo(c)(s10)
> ; RV32I-NEXT: lw a1, %lo(b)(s11)
> +; RV32I-NEXT: lw a2, %lo(c)(s10)
> +; RV32I-NEXT: lw a3, %lo(d)(s1)
> +; RV32I-NEXT: lw a4, %lo(e)(s0)
> ; RV32I-NEXT: addi a5, zero, 32
> ; RV32I-NEXT: call foo
> ; RV32I-NEXT: .LBB0_5: # %if.end
> @@ -78,11 +78,11 @@ define i32 @test() nounwind {
> ; RV32I-NEXT: beqz a0, .LBB0_7
> ; RV32I-NEXT: # %bb.6: # %if.then3
> ; RV32I-NEXT: # in Loop: Header=BB0_5 Depth=1
> -; RV32I-NEXT: lw a4, %lo(f)(s8)
> -; RV32I-NEXT: lw a3, %lo(e)(s1)
> -; RV32I-NEXT: lw a2, %lo(d)(s0)
> -; RV32I-NEXT: lw a1, %lo(c)(s10)
> ; RV32I-NEXT: lw a0, %lo(b)(s11)
> +; RV32I-NEXT: lw a1, %lo(c)(s10)
> +; RV32I-NEXT: lw a2, %lo(d)(s1)
> +; RV32I-NEXT: lw a3, %lo(e)(s0)
> +; RV32I-NEXT: lw a4, %lo(f)(s7)
> ; RV32I-NEXT: addi a5, zero, 64
> ; RV32I-NEXT: call foo
> ; RV32I-NEXT: .LBB0_7: # %if.end5
> @@ -91,24 +91,24 @@ define i32 @test() nounwind {
> ; RV32I-NEXT: beqz a0, .LBB0_9
> ; RV32I-NEXT: # %bb.8: # %if.then7
> ; RV32I-NEXT: # in Loop: Header=BB0_5 Depth=1
> -; RV32I-NEXT: lw a4, %lo(g)(s7)
> -; RV32I-NEXT: lw a3, %lo(f)(s8)
> -; RV32I-NEXT: lw a2, %lo(e)(s1)
> -; RV32I-NEXT: lw a1, %lo(d)(s0)
> ; RV32I-NEXT: lw a0, %lo(c)(s10)
> +; RV32I-NEXT: lw a1, %lo(d)(s1)
> +; RV32I-NEXT: lw a2, %lo(e)(s0)
> +; RV32I-NEXT: lw a3, %lo(f)(s7)
> +; RV32I-NEXT: lw a4, %lo(g)(s8)
> ; RV32I-NEXT: addi a5, zero, 32
> ; RV32I-NEXT: call foo
> ; RV32I-NEXT: .LBB0_9: # %if.end9
> ; RV32I-NEXT: # in Loop: Header=BB0_5 Depth=1
> -; RV32I-NEXT: lw a0, %lo(i)(s6)
> +; RV32I-NEXT: lw a0, %lo(i)(s5)
> ; RV32I-NEXT: beqz a0, .LBB0_2
> ; RV32I-NEXT: # %bb.10: # %if.then11
> ; RV32I-NEXT: # in Loop: Header=BB0_5 Depth=1
> -; RV32I-NEXT: lw a4, %lo(h)(s5)
> -; RV32I-NEXT: lw a3, %lo(g)(s7)
> -; RV32I-NEXT: lw a2, %lo(f)(s8)
> -; RV32I-NEXT: lw a1, %lo(e)(s1)
> -; RV32I-NEXT: lw a0, %lo(d)(s0)
> +; RV32I-NEXT: lw a0, %lo(d)(s1)
> +; RV32I-NEXT: lw a1, %lo(e)(s0)
> +; RV32I-NEXT: lw a2, %lo(f)(s7)
> +; RV32I-NEXT: lw a3, %lo(g)(s8)
> +; RV32I-NEXT: lw a4, %lo(h)(s9)
> ; RV32I-NEXT: addi a5, zero, 32
> ; RV32I-NEXT: call foo
> ; RV32I-NEXT: j .LBB0_2
>
> Modified: llvm/trunk/test/CodeGen/RISCV/rv64f-float-convert.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/rv64f-float-convert.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/rv64f-float-convert.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/rv64f-float-convert.ll Tue Sep 17 02:43:08 2019
> @@ -113,9 +113,9 @@ define zeroext i32 @bcvt_f32_to_zext_i32
> define float @bcvt_i64_to_f32_via_i32(i64 %a, i64 %b) nounwind {
> ; RV64IF-LABEL: bcvt_i64_to_f32_via_i32:
> ; RV64IF: # %bb.0:
> -; RV64IF-NEXT: fmv.w.x ft0, a1
> -; RV64IF-NEXT: fmv.w.x ft1, a0
> -; RV64IF-NEXT: fadd.s ft0, ft1, ft0
> +; RV64IF-NEXT: fmv.w.x ft0, a0
> +; RV64IF-NEXT: fmv.w.x ft1, a1
> +; RV64IF-NEXT: fadd.s ft0, ft0, ft1
> ; RV64IF-NEXT: fmv.x.w a0, ft0
> ; RV64IF-NEXT: ret
> %1 = trunc i64 %a to i32
>
> Modified: llvm/trunk/test/CodeGen/RISCV/rv64i-complex-float.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/rv64i-complex-float.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/rv64i-complex-float.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/rv64i-complex-float.ll Tue Sep 17 02:43:08 2019
> @@ -13,16 +13,16 @@ define i64 @complex_float_add(i64 %a.coe
> ; CHECK-NEXT: sd s0, 16(sp)
> ; CHECK-NEXT: sd s1, 8(sp)
> ; CHECK-NEXT: sd s2, 0(sp)
> -; CHECK-NEXT: mv s0, a1
> -; CHECK-NEXT: mv s1, a0
> +; CHECK-NEXT: srli s2, a0, 32
> +; CHECK-NEXT: srli s1, a1, 32
> ; CHECK-NEXT: call __addsf3
> -; CHECK-NEXT: mv s2, a0
> -; CHECK-NEXT: srli a0, s1, 32
> -; CHECK-NEXT: srli a1, s0, 32
> +; CHECK-NEXT: mv s0, a0
> +; CHECK-NEXT: mv a0, s2
> +; CHECK-NEXT: mv a1, s1
> ; CHECK-NEXT: call __addsf3
> -; CHECK-NEXT: slli a1, s2, 32
> -; CHECK-NEXT: srli a1, a1, 32
> ; CHECK-NEXT: slli a0, a0, 32
> +; CHECK-NEXT: slli a1, s0, 32
> +; CHECK-NEXT: srli a1, a1, 32
> ; CHECK-NEXT: or a0, a0, a1
> ; CHECK-NEXT: ld s2, 0(sp)
> ; CHECK-NEXT: ld s1, 8(sp)
>
> Modified: llvm/trunk/test/CodeGen/RISCV/rv64i-w-insts-legalization.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/rv64i-w-insts-legalization.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/rv64i-w-insts-legalization.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/rv64i-w-insts-legalization.ll Tue Sep 17 02:43:08 2019
> @@ -8,17 +8,17 @@ define signext i32 @addw(i32 signext %s,
> ; CHECK-NEXT: # %bb.1: # %for.body.preheader
> ; CHECK-NEXT: not a2, a0
> ; CHECK-NEXT: add a2, a2, a1
> +; CHECK-NEXT: addi a3, a0, 1
> +; CHECK-NEXT: mul a3, a2, a3
> +; CHECK-NEXT: slli a2, a2, 32
> +; CHECK-NEXT: srli a2, a2, 32
> ; CHECK-NEXT: sub a1, a1, a0
> ; CHECK-NEXT: addi a1, a1, -2
> ; CHECK-NEXT: slli a1, a1, 32
> ; CHECK-NEXT: srli a1, a1, 32
> -; CHECK-NEXT: slli a3, a2, 32
> -; CHECK-NEXT: srli a3, a3, 32
> -; CHECK-NEXT: mul a1, a3, a1
> -; CHECK-NEXT: addi a3, a0, 1
> -; CHECK-NEXT: mul a2, a2, a3
> -; CHECK-NEXT: add a0, a2, a0
> +; CHECK-NEXT: mul a1, a2, a1
> ; CHECK-NEXT: srli a1, a1, 1
> +; CHECK-NEXT: add a0, a3, a0
> ; CHECK-NEXT: addw a0, a0, a1
> ; CHECK-NEXT: ret
> ; CHECK-NEXT: .LBB0_2:
> @@ -54,18 +54,18 @@ define signext i32 @subw(i32 signext %s,
> ; CHECK: # %bb.0: # %entry
> ; CHECK-NEXT: bge a0, a1, .LBB1_2
> ; CHECK-NEXT: # %bb.1: # %for.body.preheader
> -; CHECK-NEXT: sub a2, a1, a0
> -; CHECK-NEXT: addi a2, a2, -2
> -; CHECK-NEXT: slli a2, a2, 32
> -; CHECK-NEXT: srli a2, a2, 32
> -; CHECK-NEXT: not a3, a0
> -; CHECK-NEXT: add a1, a3, a1
> -; CHECK-NEXT: slli a4, a1, 32
> -; CHECK-NEXT: srli a4, a4, 32
> -; CHECK-NEXT: mul a2, a4, a2
> -; CHECK-NEXT: mul a1, a1, a3
> -; CHECK-NEXT: sub a0, a1, a0
> -; CHECK-NEXT: srli a1, a2, 1
> +; CHECK-NEXT: not a2, a0
> +; CHECK-NEXT: add a3, a2, a1
> +; CHECK-NEXT: mul a2, a3, a2
> +; CHECK-NEXT: slli a3, a3, 32
> +; CHECK-NEXT: srli a3, a3, 32
> +; CHECK-NEXT: sub a1, a1, a0
> +; CHECK-NEXT: addi a1, a1, -2
> +; CHECK-NEXT: slli a1, a1, 32
> +; CHECK-NEXT: srli a1, a1, 32
> +; CHECK-NEXT: mul a1, a3, a1
> +; CHECK-NEXT: srli a1, a1, 1
> +; CHECK-NEXT: sub a0, a2, a0
> ; CHECK-NEXT: subw a0, a0, a1
> ; CHECK-NEXT: ret
> ; CHECK-NEXT: .LBB1_2:
>
> Modified: llvm/trunk/test/CodeGen/RISCV/select-optimize-multiple.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/select-optimize-multiple.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/select-optimize-multiple.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/select-optimize-multiple.ll Tue Sep 17 02:43:08 2019
> @@ -40,27 +40,41 @@ define i128 @cmovcc128(i64 signext %a, i
> ; RV32I-LABEL: cmovcc128:
> ; RV32I: # %bb.0: # %entry
> ; RV32I-NEXT: xori a1, a1, 123
> -; RV32I-NEXT: or a1, a1, a2
> -; RV32I-NEXT: beqz a1, .LBB1_2
> +; RV32I-NEXT: or a2, a1, a2
> +; RV32I-NEXT: mv a1, a3
> +; RV32I-NEXT: beqz a2, .LBB1_2
> ; RV32I-NEXT: # %bb.1: # %entry
> +; RV32I-NEXT: mv a1, a4
> +; RV32I-NEXT: .LBB1_2: # %entry
> +; RV32I-NEXT: lw a6, 0(a1)
> +; RV32I-NEXT: beqz a2, .LBB1_6
> +; RV32I-NEXT: # %bb.3: # %entry
> ; RV32I-NEXT: addi a1, a4, 4
> -; RV32I-NEXT: addi a2, a4, 8
> -; RV32I-NEXT: addi a5, a4, 12
> -; RV32I-NEXT: mv a3, a4
> -; RV32I-NEXT: j .LBB1_3
> -; RV32I-NEXT: .LBB1_2:
> +; RV32I-NEXT: lw a5, 0(a1)
> +; RV32I-NEXT: bnez a2, .LBB1_7
> +; RV32I-NEXT: .LBB1_4:
> +; RV32I-NEXT: addi a1, a3, 8
> +; RV32I-NEXT: lw a1, 0(a1)
> +; RV32I-NEXT: bnez a2, .LBB1_8
> +; RV32I-NEXT: .LBB1_5:
> +; RV32I-NEXT: addi a2, a3, 12
> +; RV32I-NEXT: j .LBB1_9
> +; RV32I-NEXT: .LBB1_6:
> ; RV32I-NEXT: addi a1, a3, 4
> -; RV32I-NEXT: addi a2, a3, 8
> -; RV32I-NEXT: addi a5, a3, 12
> -; RV32I-NEXT: .LBB1_3: # %entry
> -; RV32I-NEXT: lw a4, 0(a5)
> -; RV32I-NEXT: sw a4, 12(a0)
> -; RV32I-NEXT: lw a2, 0(a2)
> -; RV32I-NEXT: sw a2, 8(a0)
> +; RV32I-NEXT: lw a5, 0(a1)
> +; RV32I-NEXT: beqz a2, .LBB1_4
> +; RV32I-NEXT: .LBB1_7: # %entry
> +; RV32I-NEXT: addi a1, a4, 8
> ; RV32I-NEXT: lw a1, 0(a1)
> -; RV32I-NEXT: sw a1, 4(a0)
> -; RV32I-NEXT: lw a1, 0(a3)
> -; RV32I-NEXT: sw a1, 0(a0)
> +; RV32I-NEXT: beqz a2, .LBB1_5
> +; RV32I-NEXT: .LBB1_8: # %entry
> +; RV32I-NEXT: addi a2, a4, 12
> +; RV32I-NEXT: .LBB1_9: # %entry
> +; RV32I-NEXT: lw a2, 0(a2)
> +; RV32I-NEXT: sw a2, 12(a0)
> +; RV32I-NEXT: sw a1, 8(a0)
> +; RV32I-NEXT: sw a5, 4(a0)
> +; RV32I-NEXT: sw a6, 0(a0)
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: cmovcc128:
> @@ -83,24 +97,24 @@ entry:
> define i64 @cmov64(i1 %a, i64 %b, i64 %c) nounwind {
> ; RV32I-LABEL: cmov64:
> ; RV32I: # %bb.0: # %entry
> -; RV32I-NEXT: andi a0, a0, 1
> -; RV32I-NEXT: bnez a0, .LBB2_2
> +; RV32I-NEXT: andi a5, a0, 1
> +; RV32I-NEXT: mv a0, a1
> +; RV32I-NEXT: bnez a5, .LBB2_2
> ; RV32I-NEXT: # %bb.1: # %entry
> -; RV32I-NEXT: mv a1, a3
> +; RV32I-NEXT: mv a0, a3
> ; RV32I-NEXT: mv a2, a4
> ; RV32I-NEXT: .LBB2_2: # %entry
> -; RV32I-NEXT: mv a0, a1
> ; RV32I-NEXT: mv a1, a2
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: cmov64:
> ; RV64I: # %bb.0: # %entry
> -; RV64I-NEXT: andi a0, a0, 1
> -; RV64I-NEXT: bnez a0, .LBB2_2
> +; RV64I-NEXT: andi a3, a0, 1
> +; RV64I-NEXT: mv a0, a1
> +; RV64I-NEXT: bnez a3, .LBB2_2
> ; RV64I-NEXT: # %bb.1: # %entry
> -; RV64I-NEXT: mv a1, a2
> +; RV64I-NEXT: mv a0, a2
> ; RV64I-NEXT: .LBB2_2: # %entry
> -; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: ret
> entry:
> %cond = select i1 %a, i64 %b, i64 %c
> @@ -110,38 +124,52 @@ entry:
> define i128 @cmov128(i1 %a, i128 %b, i128 %c) nounwind {
> ; RV32I-LABEL: cmov128:
> ; RV32I: # %bb.0: # %entry
> -; RV32I-NEXT: andi a1, a1, 1
> -; RV32I-NEXT: bnez a1, .LBB3_2
> +; RV32I-NEXT: andi a4, a1, 1
> +; RV32I-NEXT: mv a1, a2
> +; RV32I-NEXT: bnez a4, .LBB3_2
> ; RV32I-NEXT: # %bb.1: # %entry
> +; RV32I-NEXT: mv a1, a3
> +; RV32I-NEXT: .LBB3_2: # %entry
> +; RV32I-NEXT: lw a6, 0(a1)
> +; RV32I-NEXT: bnez a4, .LBB3_6
> +; RV32I-NEXT: # %bb.3: # %entry
> ; RV32I-NEXT: addi a1, a3, 4
> -; RV32I-NEXT: addi a4, a3, 8
> -; RV32I-NEXT: addi a5, a3, 12
> -; RV32I-NEXT: mv a2, a3
> -; RV32I-NEXT: j .LBB3_3
> -; RV32I-NEXT: .LBB3_2:
> +; RV32I-NEXT: lw a5, 0(a1)
> +; RV32I-NEXT: beqz a4, .LBB3_7
> +; RV32I-NEXT: .LBB3_4:
> +; RV32I-NEXT: addi a1, a2, 8
> +; RV32I-NEXT: lw a1, 0(a1)
> +; RV32I-NEXT: beqz a4, .LBB3_8
> +; RV32I-NEXT: .LBB3_5:
> +; RV32I-NEXT: addi a2, a2, 12
> +; RV32I-NEXT: j .LBB3_9
> +; RV32I-NEXT: .LBB3_6:
> ; RV32I-NEXT: addi a1, a2, 4
> -; RV32I-NEXT: addi a4, a2, 8
> -; RV32I-NEXT: addi a5, a2, 12
> -; RV32I-NEXT: .LBB3_3: # %entry
> -; RV32I-NEXT: lw a3, 0(a5)
> -; RV32I-NEXT: sw a3, 12(a0)
> -; RV32I-NEXT: lw a3, 0(a4)
> -; RV32I-NEXT: sw a3, 8(a0)
> +; RV32I-NEXT: lw a5, 0(a1)
> +; RV32I-NEXT: bnez a4, .LBB3_4
> +; RV32I-NEXT: .LBB3_7: # %entry
> +; RV32I-NEXT: addi a1, a3, 8
> ; RV32I-NEXT: lw a1, 0(a1)
> -; RV32I-NEXT: sw a1, 4(a0)
> -; RV32I-NEXT: lw a1, 0(a2)
> -; RV32I-NEXT: sw a1, 0(a0)
> +; RV32I-NEXT: bnez a4, .LBB3_5
> +; RV32I-NEXT: .LBB3_8: # %entry
> +; RV32I-NEXT: addi a2, a3, 12
> +; RV32I-NEXT: .LBB3_9: # %entry
> +; RV32I-NEXT: lw a2, 0(a2)
> +; RV32I-NEXT: sw a2, 12(a0)
> +; RV32I-NEXT: sw a1, 8(a0)
> +; RV32I-NEXT: sw a5, 4(a0)
> +; RV32I-NEXT: sw a6, 0(a0)
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: cmov128:
> ; RV64I: # %bb.0: # %entry
> -; RV64I-NEXT: andi a0, a0, 1
> -; RV64I-NEXT: bnez a0, .LBB3_2
> +; RV64I-NEXT: andi a5, a0, 1
> +; RV64I-NEXT: mv a0, a1
> +; RV64I-NEXT: bnez a5, .LBB3_2
> ; RV64I-NEXT: # %bb.1: # %entry
> -; RV64I-NEXT: mv a1, a3
> +; RV64I-NEXT: mv a0, a3
> ; RV64I-NEXT: mv a2, a4
> ; RV64I-NEXT: .LBB3_2: # %entry
> -; RV64I-NEXT: mv a0, a1
> ; RV64I-NEXT: mv a1, a2
> ; RV64I-NEXT: ret
> entry:
> @@ -280,39 +308,37 @@ entry:
> define i32 @cmovdiffcc(i1 %a, i1 %b, i32 %c, i32 %d, i32 %e, i32 %f) nounwind {
> ; RV32I-LABEL: cmovdiffcc:
> ; RV32I: # %bb.0: # %entry
> +; RV32I-NEXT: andi a0, a0, 1
> ; RV32I-NEXT: andi a1, a1, 1
> -; RV32I-NEXT: beqz a1, .LBB7_3
> +; RV32I-NEXT: beqz a0, .LBB7_3
> ; RV32I-NEXT: # %bb.1: # %entry
> -; RV32I-NEXT: andi a0, a0, 1
> -; RV32I-NEXT: beqz a0, .LBB7_4
> +; RV32I-NEXT: beqz a1, .LBB7_4
> ; RV32I-NEXT: .LBB7_2: # %entry
> ; RV32I-NEXT: add a0, a2, a4
> ; RV32I-NEXT: ret
> ; RV32I-NEXT: .LBB7_3: # %entry
> -; RV32I-NEXT: mv a4, a5
> -; RV32I-NEXT: andi a0, a0, 1
> -; RV32I-NEXT: bnez a0, .LBB7_2
> -; RV32I-NEXT: .LBB7_4: # %entry
> ; RV32I-NEXT: mv a2, a3
> +; RV32I-NEXT: bnez a1, .LBB7_2
> +; RV32I-NEXT: .LBB7_4: # %entry
> +; RV32I-NEXT: mv a4, a5
> ; RV32I-NEXT: add a0, a2, a4
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: cmovdiffcc:
> ; RV64I: # %bb.0: # %entry
> +; RV64I-NEXT: andi a0, a0, 1
> ; RV64I-NEXT: andi a1, a1, 1
> -; RV64I-NEXT: beqz a1, .LBB7_3
> +; RV64I-NEXT: beqz a0, .LBB7_3
> ; RV64I-NEXT: # %bb.1: # %entry
> -; RV64I-NEXT: andi a0, a0, 1
> -; RV64I-NEXT: beqz a0, .LBB7_4
> +; RV64I-NEXT: beqz a1, .LBB7_4
> ; RV64I-NEXT: .LBB7_2: # %entry
> ; RV64I-NEXT: addw a0, a2, a4
> ; RV64I-NEXT: ret
> ; RV64I-NEXT: .LBB7_3: # %entry
> -; RV64I-NEXT: mv a4, a5
> -; RV64I-NEXT: andi a0, a0, 1
> -; RV64I-NEXT: bnez a0, .LBB7_2
> -; RV64I-NEXT: .LBB7_4: # %entry
> ; RV64I-NEXT: mv a2, a3
> +; RV64I-NEXT: bnez a1, .LBB7_2
> +; RV64I-NEXT: .LBB7_4: # %entry
> +; RV64I-NEXT: mv a4, a5
> ; RV64I-NEXT: addw a0, a2, a4
> ; RV64I-NEXT: ret
> entry:
>
> Modified: llvm/trunk/test/CodeGen/RISCV/setcc-logic.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/setcc-logic.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/setcc-logic.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/setcc-logic.ll Tue Sep 17 02:43:08 2019
> @@ -7,17 +7,17 @@
> define i1 @and_icmp_eq(i32 %a, i32 %b, i32 %c, i32 %d) nounwind {
> ; RV32I-LABEL: and_icmp_eq:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: xor a2, a2, a3
> ; RV32I-NEXT: xor a0, a0, a1
> -; RV32I-NEXT: or a0, a0, a2
> +; RV32I-NEXT: xor a1, a2, a3
> +; RV32I-NEXT: or a0, a0, a1
> ; RV32I-NEXT: seqz a0, a0
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: and_icmp_eq:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: xor a2, a2, a3
> ; RV64I-NEXT: xor a0, a0, a1
> -; RV64I-NEXT: or a0, a0, a2
> +; RV64I-NEXT: xor a1, a2, a3
> +; RV64I-NEXT: or a0, a0, a1
> ; RV64I-NEXT: slli a0, a0, 32
> ; RV64I-NEXT: srli a0, a0, 32
> ; RV64I-NEXT: seqz a0, a0
> @@ -31,17 +31,17 @@ define i1 @and_icmp_eq(i32 %a, i32 %b, i
> define i1 @or_icmp_ne(i32 %a, i32 %b, i32 %c, i32 %d) nounwind {
> ; RV32I-LABEL: or_icmp_ne:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: xor a2, a2, a3
> ; RV32I-NEXT: xor a0, a0, a1
> -; RV32I-NEXT: or a0, a0, a2
> +; RV32I-NEXT: xor a1, a2, a3
> +; RV32I-NEXT: or a0, a0, a1
> ; RV32I-NEXT: snez a0, a0
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: or_icmp_ne:
> ; RV64I: # %bb.0:
> -; RV64I-NEXT: xor a2, a2, a3
> ; RV64I-NEXT: xor a0, a0, a1
> -; RV64I-NEXT: or a0, a0, a2
> +; RV64I-NEXT: xor a1, a2, a3
> +; RV64I-NEXT: or a0, a0, a1
> ; RV64I-NEXT: slli a0, a0, 32
> ; RV64I-NEXT: srli a0, a0, 32
> ; RV64I-NEXT: snez a0, a0
> @@ -102,22 +102,22 @@ define i1 @and_icmps_const_1bit_diff(i32
> define i1 @and_icmps_const_not1bit_diff(i32 %x) nounwind {
> ; RV32I-LABEL: and_icmps_const_not1bit_diff:
> ; RV32I: # %bb.0:
> -; RV32I-NEXT: xori a1, a0, 92
> +; RV32I-NEXT: xori a1, a0, 44
> ; RV32I-NEXT: snez a1, a1
> -; RV32I-NEXT: xori a0, a0, 44
> +; RV32I-NEXT: xori a0, a0, 92
> ; RV32I-NEXT: snez a0, a0
> -; RV32I-NEXT: and a0, a0, a1
> +; RV32I-NEXT: and a0, a1, a0
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: and_icmps_const_not1bit_diff:
> ; RV64I: # %bb.0:
> ; RV64I-NEXT: slli a0, a0, 32
> ; RV64I-NEXT: srli a0, a0, 32
> -; RV64I-NEXT: xori a1, a0, 92
> +; RV64I-NEXT: xori a1, a0, 44
> ; RV64I-NEXT: snez a1, a1
> -; RV64I-NEXT: xori a0, a0, 44
> +; RV64I-NEXT: xori a0, a0, 92
> ; RV64I-NEXT: snez a0, a0
> -; RV64I-NEXT: and a0, a0, a1
> +; RV64I-NEXT: and a0, a1, a0
> ; RV64I-NEXT: ret
> %a = icmp ne i32 %x, 44
> %b = icmp ne i32 %x, 92
>
> Modified: llvm/trunk/test/CodeGen/RISCV/shifts.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/shifts.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/shifts.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/shifts.ll Tue Sep 17 02:43:08 2019
> @@ -17,13 +17,14 @@ define i64 @lshr64(i64 %a, i64 %b) nounw
> ; RV32I-NEXT: mv a1, zero
> ; RV32I-NEXT: ret
> ; RV32I-NEXT: .LBB0_2:
> +; RV32I-NEXT: srl a0, a0, a2
> ; RV32I-NEXT: addi a3, zero, 31
> ; RV32I-NEXT: sub a3, a3, a2
> ; RV32I-NEXT: slli a4, a1, 1
> ; RV32I-NEXT: sll a3, a4, a3
> -; RV32I-NEXT: srl a0, a0, a2
> ; RV32I-NEXT: or a0, a0, a3
> -; RV32I-NEXT: srl a1, a1, a2
> +; RV32I-NEXT: srl a2, a1, a2
> +; RV32I-NEXT: mv a1, a2
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: lshr64:
> @@ -62,11 +63,11 @@ define i64 @ashr64(i64 %a, i64 %b) nounw
> ; RV32I-NEXT: srai a1, a1, 31
> ; RV32I-NEXT: ret
> ; RV32I-NEXT: .LBB2_2:
> +; RV32I-NEXT: srl a0, a0, a2
> ; RV32I-NEXT: addi a3, zero, 31
> ; RV32I-NEXT: sub a3, a3, a2
> ; RV32I-NEXT: slli a4, a1, 1
> ; RV32I-NEXT: sll a3, a4, a3
> -; RV32I-NEXT: srl a0, a0, a2
> ; RV32I-NEXT: or a0, a0, a3
> ; RV32I-NEXT: sra a1, a1, a2
> ; RV32I-NEXT: ret
> @@ -107,13 +108,14 @@ define i64 @shl64(i64 %a, i64 %b) nounwi
> ; RV32I-NEXT: mv a0, zero
> ; RV32I-NEXT: ret
> ; RV32I-NEXT: .LBB4_2:
> +; RV32I-NEXT: sll a1, a1, a2
> ; RV32I-NEXT: addi a3, zero, 31
> ; RV32I-NEXT: sub a3, a3, a2
> ; RV32I-NEXT: srli a4, a0, 1
> ; RV32I-NEXT: srl a3, a4, a3
> -; RV32I-NEXT: sll a1, a1, a2
> ; RV32I-NEXT: or a1, a1, a3
> -; RV32I-NEXT: sll a0, a0, a2
> +; RV32I-NEXT: sll a2, a0, a2
> +; RV32I-NEXT: mv a0, a2
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: shl64:
> @@ -148,27 +150,27 @@ define i128 @lshr128(i128 %a, i128 %b) n
> ; RV32I-NEXT: addi sp, sp, -48
> ; RV32I-NEXT: sw ra, 44(sp)
> ; RV32I-NEXT: sw s0, 40(sp)
> -; RV32I-NEXT: mv s0, a0
> -; RV32I-NEXT: lw a0, 12(a1)
> -; RV32I-NEXT: sw a0, 20(sp)
> -; RV32I-NEXT: lw a0, 8(a1)
> -; RV32I-NEXT: sw a0, 16(sp)
> -; RV32I-NEXT: lw a0, 4(a1)
> -; RV32I-NEXT: sw a0, 12(sp)
> -; RV32I-NEXT: lw a0, 0(a1)
> -; RV32I-NEXT: sw a0, 8(sp)
> ; RV32I-NEXT: lw a2, 0(a2)
> +; RV32I-NEXT: lw a3, 0(a1)
> +; RV32I-NEXT: lw a4, 4(a1)
> +; RV32I-NEXT: lw a5, 8(a1)
> +; RV32I-NEXT: lw a1, 12(a1)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: sw a1, 20(sp)
> +; RV32I-NEXT: sw a5, 16(sp)
> +; RV32I-NEXT: sw a4, 12(sp)
> ; RV32I-NEXT: addi a0, sp, 24
> ; RV32I-NEXT: addi a1, sp, 8
> +; RV32I-NEXT: sw a3, 8(sp)
> ; RV32I-NEXT: call __lshrti3
> ; RV32I-NEXT: lw a0, 36(sp)
> +; RV32I-NEXT: lw a1, 32(sp)
> +; RV32I-NEXT: lw a2, 28(sp)
> +; RV32I-NEXT: lw a3, 24(sp)
> ; RV32I-NEXT: sw a0, 12(s0)
> -; RV32I-NEXT: lw a0, 32(sp)
> -; RV32I-NEXT: sw a0, 8(s0)
> -; RV32I-NEXT: lw a0, 28(sp)
> -; RV32I-NEXT: sw a0, 4(s0)
> -; RV32I-NEXT: lw a0, 24(sp)
> -; RV32I-NEXT: sw a0, 0(s0)
> +; RV32I-NEXT: sw a1, 8(s0)
> +; RV32I-NEXT: sw a2, 4(s0)
> +; RV32I-NEXT: sw a3, 0(s0)
> ; RV32I-NEXT: lw s0, 40(sp)
> ; RV32I-NEXT: lw ra, 44(sp)
> ; RV32I-NEXT: addi sp, sp, 48
> @@ -183,13 +185,14 @@ define i128 @lshr128(i128 %a, i128 %b) n
> ; RV64I-NEXT: mv a1, zero
> ; RV64I-NEXT: ret
> ; RV64I-NEXT: .LBB6_2:
> +; RV64I-NEXT: srl a0, a0, a2
> ; RV64I-NEXT: addi a3, zero, 63
> ; RV64I-NEXT: sub a3, a3, a2
> ; RV64I-NEXT: slli a4, a1, 1
> ; RV64I-NEXT: sll a3, a4, a3
> -; RV64I-NEXT: srl a0, a0, a2
> ; RV64I-NEXT: or a0, a0, a3
> -; RV64I-NEXT: srl a1, a1, a2
> +; RV64I-NEXT: srl a2, a1, a2
> +; RV64I-NEXT: mv a1, a2
> ; RV64I-NEXT: ret
> %1 = lshr i128 %a, %b
> ret i128 %1
> @@ -201,27 +204,27 @@ define i128 @ashr128(i128 %a, i128 %b) n
> ; RV32I-NEXT: addi sp, sp, -48
> ; RV32I-NEXT: sw ra, 44(sp)
> ; RV32I-NEXT: sw s0, 40(sp)
> -; RV32I-NEXT: mv s0, a0
> -; RV32I-NEXT: lw a0, 12(a1)
> -; RV32I-NEXT: sw a0, 20(sp)
> -; RV32I-NEXT: lw a0, 8(a1)
> -; RV32I-NEXT: sw a0, 16(sp)
> -; RV32I-NEXT: lw a0, 4(a1)
> -; RV32I-NEXT: sw a0, 12(sp)
> -; RV32I-NEXT: lw a0, 0(a1)
> -; RV32I-NEXT: sw a0, 8(sp)
> ; RV32I-NEXT: lw a2, 0(a2)
> +; RV32I-NEXT: lw a3, 0(a1)
> +; RV32I-NEXT: lw a4, 4(a1)
> +; RV32I-NEXT: lw a5, 8(a1)
> +; RV32I-NEXT: lw a1, 12(a1)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: sw a1, 20(sp)
> +; RV32I-NEXT: sw a5, 16(sp)
> +; RV32I-NEXT: sw a4, 12(sp)
> ; RV32I-NEXT: addi a0, sp, 24
> ; RV32I-NEXT: addi a1, sp, 8
> +; RV32I-NEXT: sw a3, 8(sp)
> ; RV32I-NEXT: call __ashrti3
> ; RV32I-NEXT: lw a0, 36(sp)
> +; RV32I-NEXT: lw a1, 32(sp)
> +; RV32I-NEXT: lw a2, 28(sp)
> +; RV32I-NEXT: lw a3, 24(sp)
> ; RV32I-NEXT: sw a0, 12(s0)
> -; RV32I-NEXT: lw a0, 32(sp)
> -; RV32I-NEXT: sw a0, 8(s0)
> -; RV32I-NEXT: lw a0, 28(sp)
> -; RV32I-NEXT: sw a0, 4(s0)
> -; RV32I-NEXT: lw a0, 24(sp)
> -; RV32I-NEXT: sw a0, 0(s0)
> +; RV32I-NEXT: sw a1, 8(s0)
> +; RV32I-NEXT: sw a2, 4(s0)
> +; RV32I-NEXT: sw a3, 0(s0)
> ; RV32I-NEXT: lw s0, 40(sp)
> ; RV32I-NEXT: lw ra, 44(sp)
> ; RV32I-NEXT: addi sp, sp, 48
> @@ -236,11 +239,11 @@ define i128 @ashr128(i128 %a, i128 %b) n
> ; RV64I-NEXT: srai a1, a1, 63
> ; RV64I-NEXT: ret
> ; RV64I-NEXT: .LBB7_2:
> +; RV64I-NEXT: srl a0, a0, a2
> ; RV64I-NEXT: addi a3, zero, 63
> ; RV64I-NEXT: sub a3, a3, a2
> ; RV64I-NEXT: slli a4, a1, 1
> ; RV64I-NEXT: sll a3, a4, a3
> -; RV64I-NEXT: srl a0, a0, a2
> ; RV64I-NEXT: or a0, a0, a3
> ; RV64I-NEXT: sra a1, a1, a2
> ; RV64I-NEXT: ret
> @@ -254,27 +257,27 @@ define i128 @shl128(i128 %a, i128 %b) no
> ; RV32I-NEXT: addi sp, sp, -48
> ; RV32I-NEXT: sw ra, 44(sp)
> ; RV32I-NEXT: sw s0, 40(sp)
> -; RV32I-NEXT: mv s0, a0
> -; RV32I-NEXT: lw a0, 12(a1)
> -; RV32I-NEXT: sw a0, 20(sp)
> -; RV32I-NEXT: lw a0, 8(a1)
> -; RV32I-NEXT: sw a0, 16(sp)
> -; RV32I-NEXT: lw a0, 4(a1)
> -; RV32I-NEXT: sw a0, 12(sp)
> -; RV32I-NEXT: lw a0, 0(a1)
> -; RV32I-NEXT: sw a0, 8(sp)
> ; RV32I-NEXT: lw a2, 0(a2)
> +; RV32I-NEXT: lw a3, 0(a1)
> +; RV32I-NEXT: lw a4, 4(a1)
> +; RV32I-NEXT: lw a5, 8(a1)
> +; RV32I-NEXT: lw a1, 12(a1)
> +; RV32I-NEXT: mv s0, a0
> +; RV32I-NEXT: sw a1, 20(sp)
> +; RV32I-NEXT: sw a5, 16(sp)
> +; RV32I-NEXT: sw a4, 12(sp)
> ; RV32I-NEXT: addi a0, sp, 24
> ; RV32I-NEXT: addi a1, sp, 8
> +; RV32I-NEXT: sw a3, 8(sp)
> ; RV32I-NEXT: call __ashlti3
> ; RV32I-NEXT: lw a0, 36(sp)
> +; RV32I-NEXT: lw a1, 32(sp)
> +; RV32I-NEXT: lw a2, 28(sp)
> +; RV32I-NEXT: lw a3, 24(sp)
> ; RV32I-NEXT: sw a0, 12(s0)
> -; RV32I-NEXT: lw a0, 32(sp)
> -; RV32I-NEXT: sw a0, 8(s0)
> -; RV32I-NEXT: lw a0, 28(sp)
> -; RV32I-NEXT: sw a0, 4(s0)
> -; RV32I-NEXT: lw a0, 24(sp)
> -; RV32I-NEXT: sw a0, 0(s0)
> +; RV32I-NEXT: sw a1, 8(s0)
> +; RV32I-NEXT: sw a2, 4(s0)
> +; RV32I-NEXT: sw a3, 0(s0)
> ; RV32I-NEXT: lw s0, 40(sp)
> ; RV32I-NEXT: lw ra, 44(sp)
> ; RV32I-NEXT: addi sp, sp, 48
> @@ -289,13 +292,14 @@ define i128 @shl128(i128 %a, i128 %b) no
> ; RV64I-NEXT: mv a0, zero
> ; RV64I-NEXT: ret
> ; RV64I-NEXT: .LBB8_2:
> +; RV64I-NEXT: sll a1, a1, a2
> ; RV64I-NEXT: addi a3, zero, 63
> ; RV64I-NEXT: sub a3, a3, a2
> ; RV64I-NEXT: srli a4, a0, 1
> ; RV64I-NEXT: srl a3, a4, a3
> -; RV64I-NEXT: sll a1, a1, a2
> ; RV64I-NEXT: or a1, a1, a3
> -; RV64I-NEXT: sll a0, a0, a2
> +; RV64I-NEXT: sll a2, a0, a2
> +; RV64I-NEXT: mv a0, a2
> ; RV64I-NEXT: ret
> %1 = shl i128 %a, %b
> ret i128 %1
>
> Modified: llvm/trunk/test/CodeGen/RISCV/split-offsets.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/split-offsets.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/split-offsets.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/split-offsets.ll Tue Sep 17 02:43:08 2019
> @@ -11,33 +11,33 @@
> define void @test1([65536 x i32]** %sp, [65536 x i32]* %t, i32 %n) {
> ; RV32I-LABEL: test1:
> ; RV32I: # %bb.0: # %entry
> +; RV32I-NEXT: lw a0, 0(a0)
> ; RV32I-NEXT: lui a2, 20
> ; RV32I-NEXT: addi a2, a2, -1920
> -; RV32I-NEXT: lw a0, 0(a0)
> +; RV32I-NEXT: add a1, a1, a2
> ; RV32I-NEXT: add a0, a0, a2
> +; RV32I-NEXT: addi a2, zero, 2
> +; RV32I-NEXT: sw a2, 0(a0)
> ; RV32I-NEXT: addi a3, zero, 1
> ; RV32I-NEXT: sw a3, 4(a0)
> -; RV32I-NEXT: addi a4, zero, 2
> -; RV32I-NEXT: sw a4, 0(a0)
> -; RV32I-NEXT: add a0, a1, a2
> -; RV32I-NEXT: sw a4, 4(a0)
> -; RV32I-NEXT: sw a3, 0(a0)
> +; RV32I-NEXT: sw a3, 0(a1)
> +; RV32I-NEXT: sw a2, 4(a1)
> ; RV32I-NEXT: .cfi_def_cfa_offset 0
> ; RV32I-NEXT: ret
> ;
> ; RV64I-LABEL: test1:
> ; RV64I: # %bb.0: # %entry
> +; RV64I-NEXT: ld a0, 0(a0)
> ; RV64I-NEXT: lui a2, 20
> ; RV64I-NEXT: addiw a2, a2, -1920
> -; RV64I-NEXT: ld a0, 0(a0)
> +; RV64I-NEXT: add a1, a1, a2
> ; RV64I-NEXT: add a0, a0, a2
> +; RV64I-NEXT: addi a2, zero, 2
> +; RV64I-NEXT: sw a2, 0(a0)
> ; RV64I-NEXT: addi a3, zero, 1
> ; RV64I-NEXT: sw a3, 4(a0)
> -; RV64I-NEXT: addi a4, zero, 2
> -; RV64I-NEXT: sw a4, 0(a0)
> -; RV64I-NEXT: add a0, a1, a2
> -; RV64I-NEXT: sw a4, 4(a0)
> -; RV64I-NEXT: sw a3, 0(a0)
> +; RV64I-NEXT: sw a3, 0(a1)
> +; RV64I-NEXT: sw a2, 4(a1)
> ; RV64I-NEXT: .cfi_def_cfa_offset 0
> ; RV64I-NEXT: ret
> entry:
> @@ -57,20 +57,20 @@ entry:
> define void @test2([65536 x i32]** %sp, [65536 x i32]* %t, i32 %n) {
> ; RV32I-LABEL: test2:
> ; RV32I: # %bb.0: # %entry
> -; RV32I-NEXT: lui a3, 20
> -; RV32I-NEXT: addi a3, a3, -1920
> -; RV32I-NEXT: lw a0, 0(a0)
> -; RV32I-NEXT: add a0, a0, a3
> -; RV32I-NEXT: add a1, a1, a3
> ; RV32I-NEXT: mv a3, zero
> +; RV32I-NEXT: lw a4, 0(a0)
> +; RV32I-NEXT: lui a0, 20
> +; RV32I-NEXT: addi a5, a0, -1920
> +; RV32I-NEXT: add a0, a1, a5
> +; RV32I-NEXT: add a1, a4, a5
> ; RV32I-NEXT: bge a3, a2, .LBB1_2
> ; RV32I-NEXT: .LBB1_1: # %while_body
> ; RV32I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV32I-NEXT: sw a3, 4(a0)
> ; RV32I-NEXT: addi a4, a3, 1
> -; RV32I-NEXT: sw a4, 0(a0)
> -; RV32I-NEXT: sw a3, 4(a1)
> ; RV32I-NEXT: sw a4, 0(a1)
> +; RV32I-NEXT: sw a3, 4(a1)
> +; RV32I-NEXT: sw a4, 0(a0)
> +; RV32I-NEXT: sw a3, 4(a0)
> ; RV32I-NEXT: mv a3, a4
> ; RV32I-NEXT: blt a3, a2, .LBB1_1
> ; RV32I-NEXT: .LBB1_2: # %while_end
> @@ -79,22 +79,22 @@ define void @test2([65536 x i32]** %sp,
> ;
> ; RV64I-LABEL: test2:
> ; RV64I: # %bb.0: # %entry
> -; RV64I-NEXT: lui a3, 20
> -; RV64I-NEXT: addiw a3, a3, -1920
> -; RV64I-NEXT: ld a0, 0(a0)
> -; RV64I-NEXT: add a0, a0, a3
> -; RV64I-NEXT: add a1, a1, a3
> -; RV64I-NEXT: sext.w a2, a2
> ; RV64I-NEXT: mv a3, zero
> +; RV64I-NEXT: ld a4, 0(a0)
> +; RV64I-NEXT: lui a0, 20
> +; RV64I-NEXT: addiw a5, a0, -1920
> +; RV64I-NEXT: add a0, a1, a5
> +; RV64I-NEXT: add a1, a4, a5
> +; RV64I-NEXT: sext.w a2, a2
> ; RV64I-NEXT: sext.w a4, a3
> ; RV64I-NEXT: bge a4, a2, .LBB1_2
> ; RV64I-NEXT: .LBB1_1: # %while_body
> ; RV64I-NEXT: # =>This Inner Loop Header: Depth=1
> -; RV64I-NEXT: sw a3, 4(a0)
> ; RV64I-NEXT: addi a4, a3, 1
> -; RV64I-NEXT: sw a4, 0(a0)
> -; RV64I-NEXT: sw a3, 4(a1)
> ; RV64I-NEXT: sw a4, 0(a1)
> +; RV64I-NEXT: sw a3, 4(a1)
> +; RV64I-NEXT: sw a4, 0(a0)
> +; RV64I-NEXT: sw a3, 4(a0)
> ; RV64I-NEXT: mv a3, a4
> ; RV64I-NEXT: sext.w a4, a3
> ; RV64I-NEXT: blt a4, a2, .LBB1_1
>
> Modified: llvm/trunk/test/CodeGen/RISCV/umulo-128-legalisation-lowering.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/umulo-128-legalisation-lowering.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/umulo-128-legalisation-lowering.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/umulo-128-legalisation-lowering.ll Tue Sep 17 02:43:08 2019
> @@ -4,111 +4,113 @@
> define { i128, i8 } @muloti_test(i128 %l, i128 %r) #0 {
> ; RISCV32-LABEL: muloti_test:
> ; RISCV32: # %bb.0: # %start
> -; RISCV32-NEXT: addi sp, sp, -80
> -; RISCV32-NEXT: sw ra, 76(sp)
> -; RISCV32-NEXT: sw s0, 72(sp)
> -; RISCV32-NEXT: sw s1, 68(sp)
> -; RISCV32-NEXT: sw s2, 64(sp)
> -; RISCV32-NEXT: sw s3, 60(sp)
> -; RISCV32-NEXT: sw s4, 56(sp)
> -; RISCV32-NEXT: sw s5, 52(sp)
> -; RISCV32-NEXT: sw s6, 48(sp)
> -; RISCV32-NEXT: mv s1, a2
> -; RISCV32-NEXT: mv s0, a1
> -; RISCV32-NEXT: mv s2, a0
> -; RISCV32-NEXT: sw zero, 12(sp)
> -; RISCV32-NEXT: sw zero, 8(sp)
> -; RISCV32-NEXT: sw zero, 28(sp)
> -; RISCV32-NEXT: sw zero, 24(sp)
> -; RISCV32-NEXT: lw s3, 4(a2)
> -; RISCV32-NEXT: sw s3, 4(sp)
> -; RISCV32-NEXT: lw s5, 0(a2)
> -; RISCV32-NEXT: sw s5, 0(sp)
> -; RISCV32-NEXT: lw s4, 4(a1)
> -; RISCV32-NEXT: sw s4, 20(sp)
> -; RISCV32-NEXT: lw s6, 0(a1)
> -; RISCV32-NEXT: sw s6, 16(sp)
> -; RISCV32-NEXT: addi a0, sp, 32
> -; RISCV32-NEXT: addi a1, sp, 16
> -; RISCV32-NEXT: mv a2, sp
> +; RISCV32-NEXT: addi sp, sp, -96
> +; RISCV32-NEXT: sw ra, 92(sp)
> +; RISCV32-NEXT: sw s0, 88(sp)
> +; RISCV32-NEXT: sw s1, 84(sp)
> +; RISCV32-NEXT: sw s2, 80(sp)
> +; RISCV32-NEXT: sw s3, 76(sp)
> +; RISCV32-NEXT: sw s4, 72(sp)
> +; RISCV32-NEXT: sw s5, 68(sp)
> +; RISCV32-NEXT: sw s6, 64(sp)
> +; RISCV32-NEXT: sw s7, 60(sp)
> +; RISCV32-NEXT: sw s8, 56(sp)
> +; RISCV32-NEXT: lw s2, 12(a1)
> +; RISCV32-NEXT: lw s6, 8(a1)
> +; RISCV32-NEXT: lw s3, 12(a2)
> +; RISCV32-NEXT: lw s7, 8(a2)
> +; RISCV32-NEXT: lw s0, 0(a1)
> +; RISCV32-NEXT: lw s8, 4(a1)
> +; RISCV32-NEXT: lw s1, 0(a2)
> +; RISCV32-NEXT: lw s5, 4(a2)
> +; RISCV32-NEXT: mv s4, a0
> +; RISCV32-NEXT: sw zero, 20(sp)
> +; RISCV32-NEXT: sw zero, 16(sp)
> +; RISCV32-NEXT: sw zero, 36(sp)
> +; RISCV32-NEXT: sw zero, 32(sp)
> +; RISCV32-NEXT: sw s5, 12(sp)
> +; RISCV32-NEXT: sw s1, 8(sp)
> +; RISCV32-NEXT: sw s8, 28(sp)
> +; RISCV32-NEXT: addi a0, sp, 40
> +; RISCV32-NEXT: addi a1, sp, 24
> +; RISCV32-NEXT: addi a2, sp, 8
> +; RISCV32-NEXT: sw s0, 24(sp)
> ; RISCV32-NEXT: call __multi3
> -; RISCV32-NEXT: lw a0, 12(s0)
> -; RISCV32-NEXT: lw a1, 8(s0)
> -; RISCV32-NEXT: mul a2, s3, a1
> -; RISCV32-NEXT: mul a3, a0, s5
> -; RISCV32-NEXT: add a4, a3, a2
> -; RISCV32-NEXT: lw a2, 12(s1)
> -; RISCV32-NEXT: lw a3, 8(s1)
> -; RISCV32-NEXT: mul a5, s4, a3
> -; RISCV32-NEXT: mul s1, a2, s6
> -; RISCV32-NEXT: add a5, s1, a5
> -; RISCV32-NEXT: mul s1, a3, s6
> -; RISCV32-NEXT: mul s0, a1, s5
> -; RISCV32-NEXT: add s1, s0, s1
> -; RISCV32-NEXT: sltu s0, s1, s0
> -; RISCV32-NEXT: mulhu a6, a3, s6
> -; RISCV32-NEXT: add t1, a6, a5
> -; RISCV32-NEXT: mulhu t2, a1, s5
> -; RISCV32-NEXT: add t3, t2, a4
> -; RISCV32-NEXT: add a5, t3, t1
> -; RISCV32-NEXT: add a5, a5, s0
> -; RISCV32-NEXT: lw s0, 44(sp)
> -; RISCV32-NEXT: add a5, s0, a5
> -; RISCV32-NEXT: lw a4, 40(sp)
> -; RISCV32-NEXT: add a7, a4, s1
> -; RISCV32-NEXT: sltu t0, a7, a4
> -; RISCV32-NEXT: add a5, a5, t0
> -; RISCV32-NEXT: beq a5, s0, .LBB0_2
> +; RISCV32-NEXT: mul a0, s8, s7
> +; RISCV32-NEXT: mul a1, s3, s0
> +; RISCV32-NEXT: add a0, a1, a0
> +; RISCV32-NEXT: mulhu a5, s7, s0
> +; RISCV32-NEXT: add a0, a5, a0
> +; RISCV32-NEXT: mul a1, s5, s6
> +; RISCV32-NEXT: mul a2, s2, s1
> +; RISCV32-NEXT: add a1, a2, a1
> +; RISCV32-NEXT: mulhu t0, s6, s1
> +; RISCV32-NEXT: add t1, t0, a1
> +; RISCV32-NEXT: add a6, t1, a0
> +; RISCV32-NEXT: mul a1, s7, s0
> +; RISCV32-NEXT: mul a3, s6, s1
> +; RISCV32-NEXT: add a4, a3, a1
> +; RISCV32-NEXT: lw a1, 52(sp)
> +; RISCV32-NEXT: lw a2, 48(sp)
> +; RISCV32-NEXT: sltu a3, a4, a3
> +; RISCV32-NEXT: add a3, a6, a3
> +; RISCV32-NEXT: add a3, a1, a3
> +; RISCV32-NEXT: add a6, a2, a4
> +; RISCV32-NEXT: sltu a2, a6, a2
> +; RISCV32-NEXT: add a7, a3, a2
> +; RISCV32-NEXT: beq a7, a1, .LBB0_2
> ; RISCV32-NEXT: # %bb.1: # %start
> -; RISCV32-NEXT: sltu t0, a5, s0
> +; RISCV32-NEXT: sltu a2, a7, a1
> ; RISCV32-NEXT: .LBB0_2: # %start
> -; RISCV32-NEXT: snez a4, s3
> -; RISCV32-NEXT: snez s1, a0
> -; RISCV32-NEXT: and a4, s1, a4
> -; RISCV32-NEXT: snez s1, s4
> -; RISCV32-NEXT: snez s0, a2
> -; RISCV32-NEXT: and s1, s0, s1
> -; RISCV32-NEXT: mulhu s0, a2, s6
> -; RISCV32-NEXT: snez s0, s0
> -; RISCV32-NEXT: or s1, s1, s0
> -; RISCV32-NEXT: mulhu s0, a0, s5
> -; RISCV32-NEXT: snez s0, s0
> -; RISCV32-NEXT: or a4, a4, s0
> -; RISCV32-NEXT: sltu t2, t3, t2
> -; RISCV32-NEXT: mulhu s0, s3, a1
> -; RISCV32-NEXT: snez s0, s0
> -; RISCV32-NEXT: or t3, a4, s0
> -; RISCV32-NEXT: sltu s0, t1, a6
> -; RISCV32-NEXT: mulhu a4, s4, a3
> +; RISCV32-NEXT: sltu a0, a0, a5
> +; RISCV32-NEXT: snez a1, s8
> +; RISCV32-NEXT: snez a3, s3
> +; RISCV32-NEXT: and a1, a3, a1
> +; RISCV32-NEXT: mulhu a3, s3, s0
> +; RISCV32-NEXT: snez a3, a3
> +; RISCV32-NEXT: or a1, a1, a3
> +; RISCV32-NEXT: mulhu a3, s8, s7
> +; RISCV32-NEXT: snez a3, a3
> +; RISCV32-NEXT: or a1, a1, a3
> +; RISCV32-NEXT: or a0, a1, a0
> +; RISCV32-NEXT: sltu a1, t1, t0
> +; RISCV32-NEXT: snez a3, s5
> +; RISCV32-NEXT: snez a4, s2
> +; RISCV32-NEXT: and a3, a4, a3
> +; RISCV32-NEXT: mulhu a4, s2, s1
> +; RISCV32-NEXT: snez a4, a4
> +; RISCV32-NEXT: or a3, a3, a4
> +; RISCV32-NEXT: mulhu a4, s5, s6
> +; RISCV32-NEXT: snez a4, a4
> +; RISCV32-NEXT: or a3, a3, a4
> +; RISCV32-NEXT: or a1, a3, a1
> +; RISCV32-NEXT: or a3, s7, s3
> +; RISCV32-NEXT: snez a3, a3
> +; RISCV32-NEXT: or a4, s6, s2
> ; RISCV32-NEXT: snez a4, a4
> -; RISCV32-NEXT: or a4, s1, a4
> -; RISCV32-NEXT: lw s1, 36(sp)
> -; RISCV32-NEXT: sw s1, 4(s2)
> -; RISCV32-NEXT: lw s1, 32(sp)
> -; RISCV32-NEXT: sw s1, 0(s2)
> -; RISCV32-NEXT: sw a7, 8(s2)
> -; RISCV32-NEXT: sw a5, 12(s2)
> -; RISCV32-NEXT: or a4, a4, s0
> -; RISCV32-NEXT: or a5, t3, t2
> +; RISCV32-NEXT: and a3, a4, a3
> +; RISCV32-NEXT: or a1, a3, a1
> ; RISCV32-NEXT: or a0, a1, a0
> -; RISCV32-NEXT: or a1, a3, a2
> -; RISCV32-NEXT: snez a1, a1
> -; RISCV32-NEXT: snez a0, a0
> -; RISCV32-NEXT: and a0, a0, a1
> -; RISCV32-NEXT: or a0, a0, a5
> -; RISCV32-NEXT: or a0, a0, a4
> -; RISCV32-NEXT: or a0, a0, t0
> +; RISCV32-NEXT: lw a1, 44(sp)
> +; RISCV32-NEXT: lw a3, 40(sp)
> +; RISCV32-NEXT: or a0, a0, a2
> ; RISCV32-NEXT: andi a0, a0, 1
> -; RISCV32-NEXT: sb a0, 16(s2)
> -; RISCV32-NEXT: lw s6, 48(sp)
> -; RISCV32-NEXT: lw s5, 52(sp)
> -; RISCV32-NEXT: lw s4, 56(sp)
> -; RISCV32-NEXT: lw s3, 60(sp)
> -; RISCV32-NEXT: lw s2, 64(sp)
> -; RISCV32-NEXT: lw s1, 68(sp)
> -; RISCV32-NEXT: lw s0, 72(sp)
> -; RISCV32-NEXT: lw ra, 76(sp)
> -; RISCV32-NEXT: addi sp, sp, 80
> +; RISCV32-NEXT: sw a1, 4(s4)
> +; RISCV32-NEXT: sw a3, 0(s4)
> +; RISCV32-NEXT: sw a6, 8(s4)
> +; RISCV32-NEXT: sw a7, 12(s4)
> +; RISCV32-NEXT: sb a0, 16(s4)
> +; RISCV32-NEXT: lw s8, 56(sp)
> +; RISCV32-NEXT: lw s7, 60(sp)
> +; RISCV32-NEXT: lw s6, 64(sp)
> +; RISCV32-NEXT: lw s5, 68(sp)
> +; RISCV32-NEXT: lw s4, 72(sp)
> +; RISCV32-NEXT: lw s3, 76(sp)
> +; RISCV32-NEXT: lw s2, 80(sp)
> +; RISCV32-NEXT: lw s1, 84(sp)
> +; RISCV32-NEXT: lw s0, 88(sp)
> +; RISCV32-NEXT: lw ra, 92(sp)
> +; RISCV32-NEXT: addi sp, sp, 96
> ; RISCV32-NEXT: ret
> start:
> %0 = tail call { i128, i1 } @llvm.umul.with.overflow.i128(i128 %l, i128 %r) #2
>
> Modified: llvm/trunk/test/CodeGen/RISCV/vararg.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/vararg.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/vararg.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/vararg.ll Tue Sep 17 02:43:08 2019
> @@ -46,9 +46,9 @@ define i32 @va1(i8* %fmt, ...) nounwind
> ; ILP32-ILP32F-FPELIM-NEXT: sw a4, 32(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: sw a3, 28(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: sw a2, 24(sp)
> +; ILP32-ILP32F-FPELIM-NEXT: sw a1, 20(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: addi a1, sp, 24
> ; ILP32-ILP32F-FPELIM-NEXT: sw a1, 12(sp)
> -; ILP32-ILP32F-FPELIM-NEXT: sw a0, 20(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: addi sp, sp, 48
> ; ILP32-ILP32F-FPELIM-NEXT: ret
> ;
> @@ -65,9 +65,9 @@ define i32 @va1(i8* %fmt, ...) nounwind
> ; ILP32-ILP32F-WITHFP-NEXT: sw a4, 16(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: sw a3, 12(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: sw a2, 8(s0)
> +; ILP32-ILP32F-WITHFP-NEXT: sw a1, 4(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: addi a1, s0, 8
> ; ILP32-ILP32F-WITHFP-NEXT: sw a1, -12(s0)
> -; ILP32-ILP32F-WITHFP-NEXT: sw a0, 4(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: lw s0, 8(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: lw ra, 12(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: addi sp, sp, 48
> @@ -83,9 +83,9 @@ define i32 @va1(i8* %fmt, ...) nounwind
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a4, 32(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a3, 28(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a2, 24(sp)
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a1, 20(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a1, sp, 24
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a1, 12(sp)
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a0, 20(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi sp, sp, 48
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: ret
> ;
> @@ -150,9 +150,9 @@ define i32 @va1_va_arg(i8* %fmt, ...) no
> ; ILP32-ILP32F-FPELIM-NEXT: sw a4, 32(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: sw a3, 28(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: sw a2, 24(sp)
> +; ILP32-ILP32F-FPELIM-NEXT: sw a1, 20(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: addi a1, sp, 24
> ; ILP32-ILP32F-FPELIM-NEXT: sw a1, 12(sp)
> -; ILP32-ILP32F-FPELIM-NEXT: sw a0, 20(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: addi sp, sp, 48
> ; ILP32-ILP32F-FPELIM-NEXT: ret
> ;
> @@ -169,9 +169,9 @@ define i32 @va1_va_arg(i8* %fmt, ...) no
> ; ILP32-ILP32F-WITHFP-NEXT: sw a4, 16(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: sw a3, 12(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: sw a2, 8(s0)
> +; ILP32-ILP32F-WITHFP-NEXT: sw a1, 4(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: addi a1, s0, 8
> ; ILP32-ILP32F-WITHFP-NEXT: sw a1, -12(s0)
> -; ILP32-ILP32F-WITHFP-NEXT: sw a0, 4(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: lw s0, 8(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: lw ra, 12(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: addi sp, sp, 48
> @@ -187,9 +187,9 @@ define i32 @va1_va_arg(i8* %fmt, ...) no
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a4, 32(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a3, 28(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a2, 24(sp)
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a1, 20(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a1, sp, 24
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a1, 12(sp)
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a0, 20(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi sp, sp, 48
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: ret
> ;
> @@ -203,10 +203,10 @@ define i32 @va1_va_arg(i8* %fmt, ...) no
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a4, 48(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a3, 40(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a2, 32(sp)
> +; LP64-LP64F-LP64D-FPELIM-NEXT: sd a1, 24(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a1, sp, 24
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a1, a1, 8
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a1, 8(sp)
> -; LP64-LP64F-LP64D-FPELIM-NEXT: sd a0, 24(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi sp, sp, 80
> ; LP64-LP64F-LP64D-FPELIM-NEXT: ret
> ;
> @@ -223,10 +223,10 @@ define i32 @va1_va_arg(i8* %fmt, ...) no
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a4, 32(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a3, 24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a2, 16(s0)
> +; LP64-LP64F-LP64D-WITHFP-NEXT: sd a1, 8(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a1, s0, 8
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a1, a1, 8
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a1, -24(s0)
> -; LP64-LP64F-LP64D-WITHFP-NEXT: sd a0, 8(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: ld s0, 16(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: ld ra, 24(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi sp, sp, 96
> @@ -256,9 +256,9 @@ define i32 @va1_va_arg_alloca(i8* %fmt,
> ; ILP32-ILP32F-FPELIM-NEXT: sw a4, 16(s0)
> ; ILP32-ILP32F-FPELIM-NEXT: sw a3, 12(s0)
> ; ILP32-ILP32F-FPELIM-NEXT: sw a2, 8(s0)
> +; ILP32-ILP32F-FPELIM-NEXT: sw a1, 4(s0)
> ; ILP32-ILP32F-FPELIM-NEXT: addi a0, s0, 8
> ; ILP32-ILP32F-FPELIM-NEXT: sw a0, -16(s0)
> -; ILP32-ILP32F-FPELIM-NEXT: sw a1, 4(s0)
> ; ILP32-ILP32F-FPELIM-NEXT: addi a0, a1, 15
> ; ILP32-ILP32F-FPELIM-NEXT: andi a0, a0, -16
> ; ILP32-ILP32F-FPELIM-NEXT: sub a0, sp, a0
> @@ -286,9 +286,9 @@ define i32 @va1_va_arg_alloca(i8* %fmt,
> ; ILP32-ILP32F-WITHFP-NEXT: sw a4, 16(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: sw a3, 12(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: sw a2, 8(s0)
> +; ILP32-ILP32F-WITHFP-NEXT: sw a1, 4(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: addi a0, s0, 8
> ; ILP32-ILP32F-WITHFP-NEXT: sw a0, -16(s0)
> -; ILP32-ILP32F-WITHFP-NEXT: sw a1, 4(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: addi a0, a1, 15
> ; ILP32-ILP32F-WITHFP-NEXT: andi a0, a0, -16
> ; ILP32-ILP32F-WITHFP-NEXT: sub a0, sp, a0
> @@ -316,9 +316,9 @@ define i32 @va1_va_arg_alloca(i8* %fmt,
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a4, 16(s0)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a3, 12(s0)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a2, 8(s0)
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a1, 4(s0)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a0, s0, 8
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a0, -16(s0)
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a1, 4(s0)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a0, a1, 15
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: andi a0, a0, -16
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sub a0, sp, a0
> @@ -346,17 +346,17 @@ define i32 @va1_va_arg_alloca(i8* %fmt,
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a4, 32(s0)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a3, 24(s0)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a2, 16(s0)
> +; LP64-LP64F-LP64D-FPELIM-NEXT: sd a1, 8(s0)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, s0, 8
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, a0, 8
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a0, -32(s0)
> -; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, zero, 1
> -; LP64-LP64F-LP64D-FPELIM-NEXT: slli a0, a0, 33
> -; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, a0, -16
> -; LP64-LP64F-LP64D-FPELIM-NEXT: slli a1, a1, 32
> -; LP64-LP64F-LP64D-FPELIM-NEXT: srli a1, a1, 32
> -; LP64-LP64F-LP64D-FPELIM-NEXT: addi a1, a1, 15
> -; LP64-LP64F-LP64D-FPELIM-NEXT: and a0, a1, a0
> -; LP64-LP64F-LP64D-FPELIM-NEXT: sd s1, 8(s0)
> +; LP64-LP64F-LP64D-FPELIM-NEXT: slli a0, a1, 32
> +; LP64-LP64F-LP64D-FPELIM-NEXT: srli a0, a0, 32
> +; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, a0, 15
> +; LP64-LP64F-LP64D-FPELIM-NEXT: addi a1, zero, 1
> +; LP64-LP64F-LP64D-FPELIM-NEXT: slli a1, a1, 33
> +; LP64-LP64F-LP64D-FPELIM-NEXT: addi a1, a1, -16
> +; LP64-LP64F-LP64D-FPELIM-NEXT: and a0, a0, a1
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sub a0, sp, a0
> ; LP64-LP64F-LP64D-FPELIM-NEXT: mv sp, a0
> ; LP64-LP64F-LP64D-FPELIM-NEXT: call notdead
> @@ -382,17 +382,17 @@ define i32 @va1_va_arg_alloca(i8* %fmt,
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a4, 32(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a3, 24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a2, 16(s0)
> +; LP64-LP64F-LP64D-WITHFP-NEXT: sd a1, 8(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, s0, 8
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, a0, 8
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a0, -32(s0)
> -; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, zero, 1
> -; LP64-LP64F-LP64D-WITHFP-NEXT: slli a0, a0, 33
> -; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, a0, -16
> -; LP64-LP64F-LP64D-WITHFP-NEXT: slli a1, a1, 32
> -; LP64-LP64F-LP64D-WITHFP-NEXT: srli a1, a1, 32
> -; LP64-LP64F-LP64D-WITHFP-NEXT: addi a1, a1, 15
> -; LP64-LP64F-LP64D-WITHFP-NEXT: and a0, a1, a0
> -; LP64-LP64F-LP64D-WITHFP-NEXT: sd s1, 8(s0)
> +; LP64-LP64F-LP64D-WITHFP-NEXT: slli a0, a1, 32
> +; LP64-LP64F-LP64D-WITHFP-NEXT: srli a0, a0, 32
> +; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, a0, 15
> +; LP64-LP64F-LP64D-WITHFP-NEXT: addi a1, zero, 1
> +; LP64-LP64F-LP64D-WITHFP-NEXT: slli a1, a1, 33
> +; LP64-LP64F-LP64D-WITHFP-NEXT: addi a1, a1, -16
> +; LP64-LP64F-LP64D-WITHFP-NEXT: and a0, a0, a1
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sub a0, sp, a0
> ; LP64-LP64F-LP64D-WITHFP-NEXT: mv sp, a0
> ; LP64-LP64F-LP64D-WITHFP-NEXT: call notdead
> @@ -419,9 +419,9 @@ define void @va1_caller() nounwind {
> ; ILP32-ILP32F-FPELIM: # %bb.0:
> ; ILP32-ILP32F-FPELIM-NEXT: addi sp, sp, -16
> ; ILP32-ILP32F-FPELIM-NEXT: sw ra, 12(sp)
> -; ILP32-ILP32F-FPELIM-NEXT: mv a2, zero
> ; ILP32-ILP32F-FPELIM-NEXT: lui a3, 261888
> ; ILP32-ILP32F-FPELIM-NEXT: addi a4, zero, 2
> +; ILP32-ILP32F-FPELIM-NEXT: mv a2, zero
> ; ILP32-ILP32F-FPELIM-NEXT: call va1
> ; ILP32-ILP32F-FPELIM-NEXT: lw ra, 12(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: addi sp, sp, 16
> @@ -433,9 +433,9 @@ define void @va1_caller() nounwind {
> ; ILP32-ILP32F-WITHFP-NEXT: sw ra, 12(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: sw s0, 8(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: addi s0, sp, 16
> -; ILP32-ILP32F-WITHFP-NEXT: mv a2, zero
> ; ILP32-ILP32F-WITHFP-NEXT: lui a3, 261888
> ; ILP32-ILP32F-WITHFP-NEXT: addi a4, zero, 2
> +; ILP32-ILP32F-WITHFP-NEXT: mv a2, zero
> ; ILP32-ILP32F-WITHFP-NEXT: call va1
> ; ILP32-ILP32F-WITHFP-NEXT: lw s0, 8(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: lw ra, 12(sp)
> @@ -446,9 +446,9 @@ define void @va1_caller() nounwind {
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM: # %bb.0:
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi sp, sp, -16
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw ra, 12(sp)
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: mv a2, zero
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lui a3, 261888
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a4, zero, 2
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: mv a2, zero
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: call va1
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw ra, 12(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi sp, sp, 16
> @@ -498,10 +498,10 @@ define i64 @va2(i8 *%fmt, ...) nounwind
> ; ILP32-ILP32F-FPELIM-NEXT: sw a3, 28(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: sw a2, 24(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: sw a1, 20(sp)
> -; ILP32-ILP32F-FPELIM-NEXT: addi a0, sp, 35
> -; ILP32-ILP32F-FPELIM-NEXT: sw a0, 12(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: addi a0, sp, 27
> ; ILP32-ILP32F-FPELIM-NEXT: andi a1, a0, -8
> +; ILP32-ILP32F-FPELIM-NEXT: addi a0, sp, 35
> +; ILP32-ILP32F-FPELIM-NEXT: sw a0, 12(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: lw a0, 0(a1)
> ; ILP32-ILP32F-FPELIM-NEXT: ori a1, a1, 4
> ; ILP32-ILP32F-FPELIM-NEXT: lw a1, 0(a1)
> @@ -521,10 +521,10 @@ define i64 @va2(i8 *%fmt, ...) nounwind
> ; ILP32-ILP32F-WITHFP-NEXT: sw a3, 12(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: sw a2, 8(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: sw a1, 4(s0)
> -; ILP32-ILP32F-WITHFP-NEXT: addi a0, s0, 19
> -; ILP32-ILP32F-WITHFP-NEXT: sw a0, -12(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: addi a0, s0, 11
> ; ILP32-ILP32F-WITHFP-NEXT: andi a1, a0, -8
> +; ILP32-ILP32F-WITHFP-NEXT: addi a0, s0, 19
> +; ILP32-ILP32F-WITHFP-NEXT: sw a0, -12(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: lw a0, 0(a1)
> ; ILP32-ILP32F-WITHFP-NEXT: ori a1, a1, 4
> ; ILP32-ILP32F-WITHFP-NEXT: lw a1, 0(a1)
> @@ -543,10 +543,10 @@ define i64 @va2(i8 *%fmt, ...) nounwind
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a3, 28(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a2, 24(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a1, 20(sp)
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a0, sp, 35
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a0, 12(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a0, sp, 27
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: andi a1, a0, -8
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a0, sp, 35
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a0, 12(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw a0, 0(a1)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: ori a1, a1, 4
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw a1, 0(a1)
> @@ -556,8 +556,6 @@ define i64 @va2(i8 *%fmt, ...) nounwind
> ; LP64-LP64F-LP64D-FPELIM-LABEL: va2:
> ; LP64-LP64F-LP64D-FPELIM: # %bb.0:
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi sp, sp, -80
> -; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, sp, 24
> -; LP64-LP64F-LP64D-FPELIM-NEXT: sd a0, 8(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a7, 72(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a6, 64(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a5, 56(sp)
> @@ -565,6 +563,8 @@ define i64 @va2(i8 *%fmt, ...) nounwind
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a3, 40(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a2, 32(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a1, 24(sp)
> +; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, sp, 24
> +; LP64-LP64F-LP64D-FPELIM-NEXT: sd a0, 8(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: lw a0, 8(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, a0, 7
> ; LP64-LP64F-LP64D-FPELIM-NEXT: slli a1, a0, 32
> @@ -585,8 +585,6 @@ define i64 @va2(i8 *%fmt, ...) nounwind
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd ra, 24(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd s0, 16(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi s0, sp, 32
> -; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, s0, 8
> -; LP64-LP64F-LP64D-WITHFP-NEXT: sd a0, -24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a7, 56(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a6, 48(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a5, 40(s0)
> @@ -594,6 +592,8 @@ define i64 @va2(i8 *%fmt, ...) nounwind
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a3, 24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a2, 16(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a1, 8(s0)
> +; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, s0, 8
> +; LP64-LP64F-LP64D-WITHFP-NEXT: sd a0, -24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: lw a0, -24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, a0, 7
> ; LP64-LP64F-LP64D-WITHFP-NEXT: slli a1, a0, 32
> @@ -705,10 +705,10 @@ define i64 @va2_va_arg(i8 *%fmt, ...) no
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a4, 48(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a3, 40(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a2, 32(sp)
> +; LP64-LP64F-LP64D-FPELIM-NEXT: sd a1, 24(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a1, sp, 24
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a1, a1, 8
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a1, 8(sp)
> -; LP64-LP64F-LP64D-FPELIM-NEXT: sd a0, 24(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi sp, sp, 80
> ; LP64-LP64F-LP64D-FPELIM-NEXT: ret
> ;
> @@ -725,10 +725,10 @@ define i64 @va2_va_arg(i8 *%fmt, ...) no
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a4, 32(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a3, 24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a2, 16(s0)
> +; LP64-LP64F-LP64D-WITHFP-NEXT: sd a1, 8(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a1, s0, 8
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a1, a1, 8
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a1, -24(s0)
> -; LP64-LP64F-LP64D-WITHFP-NEXT: sd a0, 8(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: ld s0, 16(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: ld ra, 24(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi sp, sp, 96
> @@ -747,8 +747,8 @@ define void @va2_caller() nounwind {
> ; ILP32-ILP32F-FPELIM: # %bb.0:
> ; ILP32-ILP32F-FPELIM-NEXT: addi sp, sp, -16
> ; ILP32-ILP32F-FPELIM-NEXT: sw ra, 12(sp)
> -; ILP32-ILP32F-FPELIM-NEXT: mv a2, zero
> ; ILP32-ILP32F-FPELIM-NEXT: lui a3, 261888
> +; ILP32-ILP32F-FPELIM-NEXT: mv a2, zero
> ; ILP32-ILP32F-FPELIM-NEXT: call va2
> ; ILP32-ILP32F-FPELIM-NEXT: lw ra, 12(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: addi sp, sp, 16
> @@ -760,8 +760,8 @@ define void @va2_caller() nounwind {
> ; ILP32-ILP32F-WITHFP-NEXT: sw ra, 12(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: sw s0, 8(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: addi s0, sp, 16
> -; ILP32-ILP32F-WITHFP-NEXT: mv a2, zero
> ; ILP32-ILP32F-WITHFP-NEXT: lui a3, 261888
> +; ILP32-ILP32F-WITHFP-NEXT: mv a2, zero
> ; ILP32-ILP32F-WITHFP-NEXT: call va2
> ; ILP32-ILP32F-WITHFP-NEXT: lw s0, 8(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: lw ra, 12(sp)
> @@ -772,8 +772,8 @@ define void @va2_caller() nounwind {
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM: # %bb.0:
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi sp, sp, -16
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw ra, 12(sp)
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: mv a2, zero
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lui a3, 261888
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: mv a2, zero
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: call va2
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw ra, 12(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi sp, sp, 16
> @@ -819,16 +819,16 @@ define i64 @va3(i32 %a, i64 %b, ...) nou
> ; ILP32-ILP32F-FPELIM-NEXT: sw a5, 20(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: sw a4, 16(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: sw a3, 12(sp)
> -; ILP32-ILP32F-FPELIM-NEXT: addi a0, sp, 27
> -; ILP32-ILP32F-FPELIM-NEXT: sw a0, 4(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: addi a0, sp, 19
> ; ILP32-ILP32F-FPELIM-NEXT: andi a0, a0, -8
> -; ILP32-ILP32F-FPELIM-NEXT: ori a3, a0, 4
> -; ILP32-ILP32F-FPELIM-NEXT: lw a3, 0(a3)
> -; ILP32-ILP32F-FPELIM-NEXT: add a2, a2, a3
> -; ILP32-ILP32F-FPELIM-NEXT: lw a0, 0(a0)
> -; ILP32-ILP32F-FPELIM-NEXT: add a0, a1, a0
> +; ILP32-ILP32F-FPELIM-NEXT: addi a3, sp, 27
> +; ILP32-ILP32F-FPELIM-NEXT: sw a3, 4(sp)
> +; ILP32-ILP32F-FPELIM-NEXT: lw a3, 0(a0)
> +; ILP32-ILP32F-FPELIM-NEXT: ori a0, a0, 4
> +; ILP32-ILP32F-FPELIM-NEXT: lw a4, 0(a0)
> +; ILP32-ILP32F-FPELIM-NEXT: add a0, a1, a3
> ; ILP32-ILP32F-FPELIM-NEXT: sltu a1, a0, a1
> +; ILP32-ILP32F-FPELIM-NEXT: add a2, a2, a4
> ; ILP32-ILP32F-FPELIM-NEXT: add a1, a2, a1
> ; ILP32-ILP32F-FPELIM-NEXT: addi sp, sp, 32
> ; ILP32-ILP32F-FPELIM-NEXT: ret
> @@ -844,16 +844,16 @@ define i64 @va3(i32 %a, i64 %b, ...) nou
> ; ILP32-ILP32F-WITHFP-NEXT: sw a5, 12(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: sw a4, 8(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: sw a3, 4(s0)
> -; ILP32-ILP32F-WITHFP-NEXT: addi a0, s0, 19
> -; ILP32-ILP32F-WITHFP-NEXT: sw a0, -12(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: addi a0, s0, 11
> ; ILP32-ILP32F-WITHFP-NEXT: andi a0, a0, -8
> -; ILP32-ILP32F-WITHFP-NEXT: ori a3, a0, 4
> -; ILP32-ILP32F-WITHFP-NEXT: lw a3, 0(a3)
> -; ILP32-ILP32F-WITHFP-NEXT: add a2, a2, a3
> -; ILP32-ILP32F-WITHFP-NEXT: lw a0, 0(a0)
> -; ILP32-ILP32F-WITHFP-NEXT: add a0, a1, a0
> +; ILP32-ILP32F-WITHFP-NEXT: addi a3, s0, 19
> +; ILP32-ILP32F-WITHFP-NEXT: sw a3, -12(s0)
> +; ILP32-ILP32F-WITHFP-NEXT: lw a3, 0(a0)
> +; ILP32-ILP32F-WITHFP-NEXT: ori a0, a0, 4
> +; ILP32-ILP32F-WITHFP-NEXT: lw a4, 0(a0)
> +; ILP32-ILP32F-WITHFP-NEXT: add a0, a1, a3
> ; ILP32-ILP32F-WITHFP-NEXT: sltu a1, a0, a1
> +; ILP32-ILP32F-WITHFP-NEXT: add a2, a2, a4
> ; ILP32-ILP32F-WITHFP-NEXT: add a1, a2, a1
> ; ILP32-ILP32F-WITHFP-NEXT: lw s0, 16(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: lw ra, 20(sp)
> @@ -868,16 +868,16 @@ define i64 @va3(i32 %a, i64 %b, ...) nou
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a5, 20(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a4, 16(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a3, 12(sp)
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a0, sp, 27
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a0, 4(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a0, sp, 19
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: andi a0, a0, -8
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: ori a3, a0, 4
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw a3, 0(a3)
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: add a2, a2, a3
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw a0, 0(a0)
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: add a0, a1, a0
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a3, sp, 27
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a3, 4(sp)
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw a3, 0(a0)
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: ori a0, a0, 4
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw a4, 0(a0)
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: add a0, a1, a3
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sltu a1, a0, a1
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: add a2, a2, a4
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: add a1, a2, a1
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi sp, sp, 32
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: ret
> @@ -885,15 +885,15 @@ define i64 @va3(i32 %a, i64 %b, ...) nou
> ; LP64-LP64F-LP64D-FPELIM-LABEL: va3:
> ; LP64-LP64F-LP64D-FPELIM: # %bb.0:
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi sp, sp, -64
> -; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, sp, 16
> -; LP64-LP64F-LP64D-FPELIM-NEXT: sd a0, 8(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a7, 56(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a6, 48(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a5, 40(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a4, 32(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a3, 24(sp)
> -; LP64-LP64F-LP64D-FPELIM-NEXT: sd a2, 16(sp)
> +; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, sp, 16
> +; LP64-LP64F-LP64D-FPELIM-NEXT: sd a0, 8(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: lw a0, 8(sp)
> +; LP64-LP64F-LP64D-FPELIM-NEXT: sd a2, 16(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, a0, 7
> ; LP64-LP64F-LP64D-FPELIM-NEXT: slli a2, a0, 32
> ; LP64-LP64F-LP64D-FPELIM-NEXT: srli a2, a2, 32
> @@ -914,15 +914,15 @@ define i64 @va3(i32 %a, i64 %b, ...) nou
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd ra, 24(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd s0, 16(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi s0, sp, 32
> -; LP64-LP64F-LP64D-WITHFP-NEXT: mv a0, s0
> -; LP64-LP64F-LP64D-WITHFP-NEXT: sd a0, -24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a7, 40(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a6, 32(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a5, 24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a4, 16(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a3, 8(s0)
> -; LP64-LP64F-LP64D-WITHFP-NEXT: sd a2, 0(s0)
> +; LP64-LP64F-LP64D-WITHFP-NEXT: mv a0, s0
> +; LP64-LP64F-LP64D-WITHFP-NEXT: sd a0, -24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: lw a0, -24(s0)
> +; LP64-LP64F-LP64D-WITHFP-NEXT: sd a2, 0(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, a0, 7
> ; LP64-LP64F-LP64D-WITHFP-NEXT: slli a2, a0, 32
> ; LP64-LP64F-LP64D-WITHFP-NEXT: srli a2, a2, 32
> @@ -973,9 +973,9 @@ define i64 @va3_va_arg(i32 %a, i64 %b, .
> ; ILP32-ILP32F-FPELIM-NEXT: addi a4, a3, 4
> ; ILP32-ILP32F-FPELIM-NEXT: sw a4, 4(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: lw a3, 0(a3)
> -; ILP32-ILP32F-FPELIM-NEXT: add a2, a2, a3
> ; ILP32-ILP32F-FPELIM-NEXT: add a0, a1, a0
> ; ILP32-ILP32F-FPELIM-NEXT: sltu a1, a0, a1
> +; ILP32-ILP32F-FPELIM-NEXT: add a2, a2, a3
> ; ILP32-ILP32F-FPELIM-NEXT: add a1, a2, a1
> ; ILP32-ILP32F-FPELIM-NEXT: addi sp, sp, 32
> ; ILP32-ILP32F-FPELIM-NEXT: ret
> @@ -999,9 +999,9 @@ define i64 @va3_va_arg(i32 %a, i64 %b, .
> ; ILP32-ILP32F-WITHFP-NEXT: addi a4, a3, 4
> ; ILP32-ILP32F-WITHFP-NEXT: sw a4, -12(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: lw a3, 0(a3)
> -; ILP32-ILP32F-WITHFP-NEXT: add a2, a2, a3
> ; ILP32-ILP32F-WITHFP-NEXT: add a0, a1, a0
> ; ILP32-ILP32F-WITHFP-NEXT: sltu a1, a0, a1
> +; ILP32-ILP32F-WITHFP-NEXT: add a2, a2, a3
> ; ILP32-ILP32F-WITHFP-NEXT: add a1, a2, a1
> ; ILP32-ILP32F-WITHFP-NEXT: lw s0, 16(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: lw ra, 20(sp)
> @@ -1023,9 +1023,9 @@ define i64 @va3_va_arg(i32 %a, i64 %b, .
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: fld ft0, 0(a0)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: fsd ft0, 8(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw a0, 12(sp)
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw a3, 8(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: add a2, a2, a0
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw a0, 8(sp)
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: add a0, a1, a0
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: add a0, a1, a3
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sltu a1, a0, a1
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: add a1, a2, a1
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi sp, sp, 48
> @@ -1039,11 +1039,11 @@ define i64 @va3_va_arg(i32 %a, i64 %b, .
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a5, 40(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a4, 32(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a3, 24(sp)
> -; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, sp, 16
> -; LP64-LP64F-LP64D-FPELIM-NEXT: ori a0, a0, 8
> -; LP64-LP64F-LP64D-FPELIM-NEXT: sd a0, 8(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a2, 16(sp)
> +; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, sp, 16
> +; LP64-LP64F-LP64D-FPELIM-NEXT: ori a3, a0, 8
> ; LP64-LP64F-LP64D-FPELIM-NEXT: add a0, a1, a2
> +; LP64-LP64F-LP64D-FPELIM-NEXT: sd a3, 8(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi sp, sp, 64
> ; LP64-LP64F-LP64D-FPELIM-NEXT: ret
> ;
> @@ -1058,11 +1058,11 @@ define i64 @va3_va_arg(i32 %a, i64 %b, .
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a5, 24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a4, 16(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a3, 8(s0)
> -; LP64-LP64F-LP64D-WITHFP-NEXT: mv a0, s0
> -; LP64-LP64F-LP64D-WITHFP-NEXT: ori a0, a0, 8
> -; LP64-LP64F-LP64D-WITHFP-NEXT: sd a0, -24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a2, 0(s0)
> +; LP64-LP64F-LP64D-WITHFP-NEXT: mv a0, s0
> +; LP64-LP64F-LP64D-WITHFP-NEXT: ori a3, a0, 8
> ; LP64-LP64F-LP64D-WITHFP-NEXT: add a0, a1, a2
> +; LP64-LP64F-LP64D-WITHFP-NEXT: sd a3, -24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: ld s0, 16(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: ld ra, 24(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi sp, sp, 80
> @@ -1084,9 +1084,9 @@ define void @va3_caller() nounwind {
> ; ILP32-ILP32F-FPELIM-NEXT: sw ra, 12(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: addi a0, zero, 2
> ; ILP32-ILP32F-FPELIM-NEXT: addi a1, zero, 1111
> +; ILP32-ILP32F-FPELIM-NEXT: lui a5, 262144
> ; ILP32-ILP32F-FPELIM-NEXT: mv a2, zero
> ; ILP32-ILP32F-FPELIM-NEXT: mv a4, zero
> -; ILP32-ILP32F-FPELIM-NEXT: lui a5, 262144
> ; ILP32-ILP32F-FPELIM-NEXT: call va3
> ; ILP32-ILP32F-FPELIM-NEXT: lw ra, 12(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: addi sp, sp, 16
> @@ -1100,9 +1100,9 @@ define void @va3_caller() nounwind {
> ; ILP32-ILP32F-WITHFP-NEXT: addi s0, sp, 16
> ; ILP32-ILP32F-WITHFP-NEXT: addi a0, zero, 2
> ; ILP32-ILP32F-WITHFP-NEXT: addi a1, zero, 1111
> +; ILP32-ILP32F-WITHFP-NEXT: lui a5, 262144
> ; ILP32-ILP32F-WITHFP-NEXT: mv a2, zero
> ; ILP32-ILP32F-WITHFP-NEXT: mv a4, zero
> -; ILP32-ILP32F-WITHFP-NEXT: lui a5, 262144
> ; ILP32-ILP32F-WITHFP-NEXT: call va3
> ; ILP32-ILP32F-WITHFP-NEXT: lw s0, 8(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: lw ra, 12(sp)
> @@ -1115,9 +1115,9 @@ define void @va3_caller() nounwind {
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw ra, 12(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a0, zero, 2
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a1, zero, 1111
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lui a5, 262144
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: mv a2, zero
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: mv a4, zero
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lui a5, 262144
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: call va3
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw ra, 12(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi sp, sp, 16
> @@ -1190,9 +1190,9 @@ define i32 @va4_va_copy(i32 %argno, ...)
> ; ILP32-ILP32F-FPELIM-NEXT: andi a0, a0, -4
> ; ILP32-ILP32F-FPELIM-NEXT: addi a3, a0, 4
> ; ILP32-ILP32F-FPELIM-NEXT: sw a3, 4(sp)
> +; ILP32-ILP32F-FPELIM-NEXT: lw a0, 0(a0)
> ; ILP32-ILP32F-FPELIM-NEXT: add a1, a1, s0
> ; ILP32-ILP32F-FPELIM-NEXT: add a1, a1, a2
> -; ILP32-ILP32F-FPELIM-NEXT: lw a0, 0(a0)
> ; ILP32-ILP32F-FPELIM-NEXT: add a0, a1, a0
> ; ILP32-ILP32F-FPELIM-NEXT: lw s0, 8(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: lw ra, 12(sp)
> @@ -1233,9 +1233,9 @@ define i32 @va4_va_copy(i32 %argno, ...)
> ; ILP32-ILP32F-WITHFP-NEXT: andi a0, a0, -4
> ; ILP32-ILP32F-WITHFP-NEXT: addi a3, a0, 4
> ; ILP32-ILP32F-WITHFP-NEXT: sw a3, -16(s0)
> +; ILP32-ILP32F-WITHFP-NEXT: lw a0, 0(a0)
> ; ILP32-ILP32F-WITHFP-NEXT: add a1, a1, s1
> ; ILP32-ILP32F-WITHFP-NEXT: add a1, a1, a2
> -; ILP32-ILP32F-WITHFP-NEXT: lw a0, 0(a0)
> ; ILP32-ILP32F-WITHFP-NEXT: add a0, a1, a0
> ; ILP32-ILP32F-WITHFP-NEXT: lw s1, 20(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: lw s0, 24(sp)
> @@ -1275,9 +1275,9 @@ define i32 @va4_va_copy(i32 %argno, ...)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: andi a0, a0, -4
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a3, a0, 4
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a3, 4(sp)
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw a0, 0(a0)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: add a1, a1, s0
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: add a1, a1, a2
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw a0, 0(a0)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: add a0, a1, a0
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw s0, 8(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw ra, 12(sp)
> @@ -1317,9 +1317,9 @@ define i32 @va4_va_copy(i32 %argno, ...)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: andi a0, a0, -4
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a3, a0, 8
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a3, 8(sp)
> +; LP64-LP64F-LP64D-FPELIM-NEXT: ld a0, 0(a0)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: add a1, a1, s0
> ; LP64-LP64F-LP64D-FPELIM-NEXT: add a1, a1, a2
> -; LP64-LP64F-LP64D-FPELIM-NEXT: ld a0, 0(a0)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addw a0, a1, a0
> ; LP64-LP64F-LP64D-FPELIM-NEXT: ld s0, 16(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: ld ra, 24(sp)
> @@ -1361,9 +1361,9 @@ define i32 @va4_va_copy(i32 %argno, ...)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: andi a0, a0, -4
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a3, a0, 8
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a3, -32(s0)
> +; LP64-LP64F-LP64D-WITHFP-NEXT: ld a0, 0(a0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: add a1, a1, s1
> ; LP64-LP64F-LP64D-WITHFP-NEXT: add a1, a1, a2
> -; LP64-LP64F-LP64D-WITHFP-NEXT: ld a0, 0(a0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addw a0, a1, a0
> ; LP64-LP64F-LP64D-WITHFP-NEXT: ld s1, 24(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: ld s0, 32(sp)
> @@ -1425,8 +1425,7 @@ define void @va5_aligned_stack_caller()
> ; ILP32-ILP32F-FPELIM-NEXT: addi a0, a0, -328
> ; ILP32-ILP32F-FPELIM-NEXT: sw a0, 36(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: lui a0, 335544
> -; ILP32-ILP32F-FPELIM-NEXT: addi a0, a0, 1311
> -; ILP32-ILP32F-FPELIM-NEXT: sw a0, 32(sp)
> +; ILP32-ILP32F-FPELIM-NEXT: addi a5, a0, 1311
> ; ILP32-ILP32F-FPELIM-NEXT: lui a0, 688509
> ; ILP32-ILP32F-FPELIM-NEXT: addi a6, a0, -2048
> ; ILP32-ILP32F-FPELIM-NEXT: addi a2, sp, 32
> @@ -1435,6 +1434,7 @@ define void @va5_aligned_stack_caller()
> ; ILP32-ILP32F-FPELIM-NEXT: addi a3, zero, 12
> ; ILP32-ILP32F-FPELIM-NEXT: addi a4, zero, 13
> ; ILP32-ILP32F-FPELIM-NEXT: addi a7, zero, 4
> +; ILP32-ILP32F-FPELIM-NEXT: sw a5, 32(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: call va5_aligned_stack_callee
> ; ILP32-ILP32F-FPELIM-NEXT: lw ra, 60(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: addi sp, sp, 64
> @@ -1470,8 +1470,7 @@ define void @va5_aligned_stack_caller()
> ; ILP32-ILP32F-WITHFP-NEXT: addi a0, a0, -328
> ; ILP32-ILP32F-WITHFP-NEXT: sw a0, -28(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: lui a0, 335544
> -; ILP32-ILP32F-WITHFP-NEXT: addi a0, a0, 1311
> -; ILP32-ILP32F-WITHFP-NEXT: sw a0, -32(s0)
> +; ILP32-ILP32F-WITHFP-NEXT: addi a5, a0, 1311
> ; ILP32-ILP32F-WITHFP-NEXT: lui a0, 688509
> ; ILP32-ILP32F-WITHFP-NEXT: addi a6, a0, -2048
> ; ILP32-ILP32F-WITHFP-NEXT: addi a2, s0, -32
> @@ -1480,6 +1479,7 @@ define void @va5_aligned_stack_caller()
> ; ILP32-ILP32F-WITHFP-NEXT: addi a3, zero, 12
> ; ILP32-ILP32F-WITHFP-NEXT: addi a4, zero, 13
> ; ILP32-ILP32F-WITHFP-NEXT: addi a7, zero, 4
> +; ILP32-ILP32F-WITHFP-NEXT: sw a5, -32(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: call va5_aligned_stack_callee
> ; ILP32-ILP32F-WITHFP-NEXT: lw s0, 56(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: lw ra, 60(sp)
> @@ -1514,8 +1514,7 @@ define void @va5_aligned_stack_caller()
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a0, a0, -328
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a0, 36(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lui a0, 335544
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a0, a0, 1311
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a0, 32(sp)
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a5, a0, 1311
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lui a0, 688509
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a6, a0, -2048
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a2, sp, 32
> @@ -1524,6 +1523,7 @@ define void @va5_aligned_stack_caller()
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a3, zero, 12
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a4, zero, 13
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a7, zero, 4
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a5, 32(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: call va5_aligned_stack_callee
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: lw ra, 60(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi sp, sp, 64
> @@ -1546,33 +1546,33 @@ define void @va5_aligned_stack_caller()
> ; LP64-LP64F-LP64D-FPELIM-NEXT: slli a0, a0, 14
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, a0, 655
> ; LP64-LP64F-LP64D-FPELIM-NEXT: slli a0, a0, 12
> -; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, a0, 1475
> -; LP64-LP64F-LP64D-FPELIM-NEXT: sd a0, 0(sp)
> +; LP64-LP64F-LP64D-FPELIM-NEXT: addi t0, a0, 1475
> +; LP64-LP64F-LP64D-FPELIM-NEXT: lui a0, 1192
> +; LP64-LP64F-LP64D-FPELIM-NEXT: addiw a0, a0, 381
> +; LP64-LP64F-LP64D-FPELIM-NEXT: slli a0, a0, 12
> +; LP64-LP64F-LP64D-FPELIM-NEXT: addi a6, a0, -2048
> ; LP64-LP64F-LP64D-FPELIM-NEXT: lui a0, 1048248
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addiw a0, a0, 1311
> ; LP64-LP64F-LP64D-FPELIM-NEXT: slli a0, a0, 12
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, a0, -1147
> ; LP64-LP64F-LP64D-FPELIM-NEXT: slli a0, a0, 13
> -; LP64-LP64F-LP64D-FPELIM-NEXT: lui a1, 512
> -; LP64-LP64F-LP64D-FPELIM-NEXT: addiw a1, a1, 73
> -; LP64-LP64F-LP64D-FPELIM-NEXT: slli a1, a1, 15
> -; LP64-LP64F-LP64D-FPELIM-NEXT: addi a1, a1, -1311
> -; LP64-LP64F-LP64D-FPELIM-NEXT: slli a1, a1, 12
> -; LP64-LP64F-LP64D-FPELIM-NEXT: addi a1, a1, 1147
> -; LP64-LP64F-LP64D-FPELIM-NEXT: slli a1, a1, 14
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, a0, 983
> ; LP64-LP64F-LP64D-FPELIM-NEXT: slli a0, a0, 15
> -; LP64-LP64F-LP64D-FPELIM-NEXT: lui a2, 1192
> -; LP64-LP64F-LP64D-FPELIM-NEXT: addiw a2, a2, 381
> -; LP64-LP64F-LP64D-FPELIM-NEXT: slli a2, a2, 12
> -; LP64-LP64F-LP64D-FPELIM-NEXT: addi a6, a2, -2048
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a2, a0, 1311
> -; LP64-LP64F-LP64D-FPELIM-NEXT: addi a3, a1, -1967
> +; LP64-LP64F-LP64D-FPELIM-NEXT: lui a0, 512
> +; LP64-LP64F-LP64D-FPELIM-NEXT: addiw a0, a0, 73
> +; LP64-LP64F-LP64D-FPELIM-NEXT: slli a0, a0, 15
> +; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, a0, -1311
> +; LP64-LP64F-LP64D-FPELIM-NEXT: slli a0, a0, 12
> +; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, a0, 1147
> +; LP64-LP64F-LP64D-FPELIM-NEXT: slli a0, a0, 14
> +; LP64-LP64F-LP64D-FPELIM-NEXT: addi a3, a0, -1967
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a0, zero, 1
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a1, zero, 11
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a4, zero, 12
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a5, zero, 13
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a7, zero, 14
> +; LP64-LP64F-LP64D-FPELIM-NEXT: sd t0, 0(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: call va5_aligned_stack_callee
> ; LP64-LP64F-LP64D-FPELIM-NEXT: ld ra, 40(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi sp, sp, 48
> @@ -1597,33 +1597,33 @@ define void @va5_aligned_stack_caller()
> ; LP64-LP64F-LP64D-WITHFP-NEXT: slli a0, a0, 14
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, a0, 655
> ; LP64-LP64F-LP64D-WITHFP-NEXT: slli a0, a0, 12
> -; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, a0, 1475
> -; LP64-LP64F-LP64D-WITHFP-NEXT: sd a0, 0(sp)
> +; LP64-LP64F-LP64D-WITHFP-NEXT: addi t0, a0, 1475
> +; LP64-LP64F-LP64D-WITHFP-NEXT: lui a0, 1192
> +; LP64-LP64F-LP64D-WITHFP-NEXT: addiw a0, a0, 381
> +; LP64-LP64F-LP64D-WITHFP-NEXT: slli a0, a0, 12
> +; LP64-LP64F-LP64D-WITHFP-NEXT: addi a6, a0, -2048
> ; LP64-LP64F-LP64D-WITHFP-NEXT: lui a0, 1048248
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addiw a0, a0, 1311
> ; LP64-LP64F-LP64D-WITHFP-NEXT: slli a0, a0, 12
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, a0, -1147
> ; LP64-LP64F-LP64D-WITHFP-NEXT: slli a0, a0, 13
> -; LP64-LP64F-LP64D-WITHFP-NEXT: lui a1, 512
> -; LP64-LP64F-LP64D-WITHFP-NEXT: addiw a1, a1, 73
> -; LP64-LP64F-LP64D-WITHFP-NEXT: slli a1, a1, 15
> -; LP64-LP64F-LP64D-WITHFP-NEXT: addi a1, a1, -1311
> -; LP64-LP64F-LP64D-WITHFP-NEXT: slli a1, a1, 12
> -; LP64-LP64F-LP64D-WITHFP-NEXT: addi a1, a1, 1147
> -; LP64-LP64F-LP64D-WITHFP-NEXT: slli a1, a1, 14
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, a0, 983
> ; LP64-LP64F-LP64D-WITHFP-NEXT: slli a0, a0, 15
> -; LP64-LP64F-LP64D-WITHFP-NEXT: lui a2, 1192
> -; LP64-LP64F-LP64D-WITHFP-NEXT: addiw a2, a2, 381
> -; LP64-LP64F-LP64D-WITHFP-NEXT: slli a2, a2, 12
> -; LP64-LP64F-LP64D-WITHFP-NEXT: addi a6, a2, -2048
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a2, a0, 1311
> -; LP64-LP64F-LP64D-WITHFP-NEXT: addi a3, a1, -1967
> +; LP64-LP64F-LP64D-WITHFP-NEXT: lui a0, 512
> +; LP64-LP64F-LP64D-WITHFP-NEXT: addiw a0, a0, 73
> +; LP64-LP64F-LP64D-WITHFP-NEXT: slli a0, a0, 15
> +; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, a0, -1311
> +; LP64-LP64F-LP64D-WITHFP-NEXT: slli a0, a0, 12
> +; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, a0, 1147
> +; LP64-LP64F-LP64D-WITHFP-NEXT: slli a0, a0, 14
> +; LP64-LP64F-LP64D-WITHFP-NEXT: addi a3, a0, -1967
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a0, zero, 1
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a1, zero, 11
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a4, zero, 12
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a5, zero, 13
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi a7, zero, 14
> +; LP64-LP64F-LP64D-WITHFP-NEXT: sd t0, 0(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: call va5_aligned_stack_callee
> ; LP64-LP64F-LP64D-WITHFP-NEXT: ld s0, 32(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: ld ra, 40(sp)
> @@ -1650,9 +1650,9 @@ define i32 @va6_no_fixed_args(...) nounw
> ; ILP32-ILP32F-FPELIM-NEXT: sw a3, 28(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: sw a2, 24(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: sw a1, 20(sp)
> +; ILP32-ILP32F-FPELIM-NEXT: sw a0, 16(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: addi a1, sp, 20
> ; ILP32-ILP32F-FPELIM-NEXT: sw a1, 12(sp)
> -; ILP32-ILP32F-FPELIM-NEXT: sw a0, 16(sp)
> ; ILP32-ILP32F-FPELIM-NEXT: addi sp, sp, 48
> ; ILP32-ILP32F-FPELIM-NEXT: ret
> ;
> @@ -1669,9 +1669,9 @@ define i32 @va6_no_fixed_args(...) nounw
> ; ILP32-ILP32F-WITHFP-NEXT: sw a3, 12(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: sw a2, 8(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: sw a1, 4(s0)
> +; ILP32-ILP32F-WITHFP-NEXT: sw a0, 0(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: addi a1, s0, 4
> ; ILP32-ILP32F-WITHFP-NEXT: sw a1, -12(s0)
> -; ILP32-ILP32F-WITHFP-NEXT: sw a0, 0(s0)
> ; ILP32-ILP32F-WITHFP-NEXT: lw s0, 8(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: lw ra, 12(sp)
> ; ILP32-ILP32F-WITHFP-NEXT: addi sp, sp, 48
> @@ -1687,9 +1687,9 @@ define i32 @va6_no_fixed_args(...) nounw
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a3, 28(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a2, 24(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a1, 20(sp)
> +; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a0, 16(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi a1, sp, 20
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a1, 12(sp)
> -; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: sw a0, 16(sp)
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: addi sp, sp, 48
> ; RV32D-ILP32-ILP32F-ILP32D-FPELIM-NEXT: ret
> ;
> @@ -1703,10 +1703,10 @@ define i32 @va6_no_fixed_args(...) nounw
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a3, 40(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a2, 32(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a1, 24(sp)
> +; LP64-LP64F-LP64D-FPELIM-NEXT: sd a0, 16(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi a1, sp, 16
> ; LP64-LP64F-LP64D-FPELIM-NEXT: ori a1, a1, 8
> ; LP64-LP64F-LP64D-FPELIM-NEXT: sd a1, 8(sp)
> -; LP64-LP64F-LP64D-FPELIM-NEXT: sd a0, 16(sp)
> ; LP64-LP64F-LP64D-FPELIM-NEXT: addi sp, sp, 80
> ; LP64-LP64F-LP64D-FPELIM-NEXT: ret
> ;
> @@ -1723,10 +1723,10 @@ define i32 @va6_no_fixed_args(...) nounw
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a3, 24(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a2, 16(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a1, 8(s0)
> +; LP64-LP64F-LP64D-WITHFP-NEXT: sd a0, 0(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: mv a1, s0
> ; LP64-LP64F-LP64D-WITHFP-NEXT: ori a1, a1, 8
> ; LP64-LP64F-LP64D-WITHFP-NEXT: sd a1, -24(s0)
> -; LP64-LP64F-LP64D-WITHFP-NEXT: sd a0, 0(s0)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: ld s0, 16(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: ld ra, 24(sp)
> ; LP64-LP64F-LP64D-WITHFP-NEXT: addi sp, sp, 96
>
> Modified: llvm/trunk/test/CodeGen/RISCV/zext-with-load-is-free.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/RISCV/zext-with-load-is-free.ll?rev=372092&r1=372091&r2=372092&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/RISCV/zext-with-load-is-free.ll (original)
> +++ llvm/trunk/test/CodeGen/RISCV/zext-with-load-is-free.ll Tue Sep 17 02:43:08 2019
> @@ -45,10 +45,10 @@ define i32 @test_zext_i16() nounwind {
> ; RV32I-LABEL: test_zext_i16:
> ; RV32I: # %bb.0: # %entry
> ; RV32I-NEXT: lui a0, %hi(shorts)
> -; RV32I-NEXT: lui a1, 16
> -; RV32I-NEXT: addi a1, a1, -120
> -; RV32I-NEXT: lhu a2, %lo(shorts)(a0)
> -; RV32I-NEXT: bne a2, a1, .LBB1_3
> +; RV32I-NEXT: lhu a1, %lo(shorts)(a0)
> +; RV32I-NEXT: lui a2, 16
> +; RV32I-NEXT: addi a2, a2, -120
> +; RV32I-NEXT: bne a1, a2, .LBB1_3
> ; RV32I-NEXT: # %bb.1: # %entry
> ; RV32I-NEXT: addi a0, a0, %lo(shorts)
> ; RV32I-NEXT: lhu a0, 2(a0)
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
More information about the llvm-commits
mailing list