[llvm] [RISCV] Generate MIPS load/store pair instructions (PR #124717)
Min-Yih Hsu via llvm-commits
llvm-commits at lists.llvm.org
Tue Jan 28 09:30:22 PST 2025
================
@@ -0,0 +1,509 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
+; RUN: | FileCheck %s -check-prefix=RV32I
+; RUN: llc -mtriple=riscv32 -target-abi ilp32d -mattr=+d -verify-machineinstrs < %s \
+; RUN: | FileCheck %s -check-prefix=RV32D
+; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \
+; RUN: | FileCheck %s -check-prefix=RV64I
+; RUN: llc -mtriple=riscv64 -target-abi lp64d -mattr=+d -verify-machineinstrs < %s \
+; RUN: | FileCheck %s -check-prefix=RV64D
+; RUN: llc -mtriple=riscv32 -mattr=+Xmipslsp -mips-riscv-load-store-pairs=1 -verify-machineinstrs < %s \
+; RUN: | FileCheck %s -check-prefix=RV32I_PAIR
+; RUN: llc -mtriple=riscv32 -target-abi ilp32d -mattr=+d,+Xmipslsp -mips-riscv-load-store-pairs=1 -verify-machineinstrs < %s \
+; RUN: | FileCheck %s -check-prefix=RV32D_PAIR
+; RUN: llc -mtriple=riscv64 -mattr=+Xmipslsp -mips-riscv-load-store-pairs=1 -verify-machineinstrs < %s \
+; RUN: | FileCheck %s -check-prefix=RV64I_PAIR
+; RUN: llc -mtriple=riscv64 -mcpu mips-p8700 -mattr=+Xmipslsp -mips-riscv-load-store-pairs=1 -verify-machineinstrs < %s \
+; RUN: | FileCheck %s -check-prefix=RV64P_8700
+; RUN: llc -mtriple=riscv64 -target-abi lp64d -mattr=+d,+Xmipslsp -mips-riscv-load-store-pairs=1 -verify-machineinstrs < %s \
+; RUN: | FileCheck %s -check-prefix=RV64D_PAIR
+; RUN: llc -mtriple=riscv64 -target-abi lp64d -mattr=+d -verify-machineinstrs < %s \
+; RUN: | FileCheck %s -check-prefix=RV64D_8700
+
+define dso_local void @testi(i8** nocapture noundef readonly %a) local_unnamed_addr #0 {
+; RV32I-LABEL: testi:
+; RV32I: # %bb.0: # %entry
+; RV32I-NEXT: addi sp, sp, -16
+; RV32I-NEXT: .cfi_def_cfa_offset 16
+; RV32I-NEXT: sw s2, 12(sp) # 4-byte Folded Spill
+; RV32I-NEXT: sw s3, 8(sp) # 4-byte Folded Spill
+; RV32I-NEXT: sw s4, 4(sp) # 4-byte Folded Spill
+; RV32I-NEXT: sw s5, 0(sp) # 4-byte Folded Spill
+; RV32I-NEXT: .cfi_offset s2, -4
+; RV32I-NEXT: .cfi_offset s3, -8
+; RV32I-NEXT: .cfi_offset s4, -12
+; RV32I-NEXT: .cfi_offset s5, -16
+; RV32I-NEXT: lw s3, 0(a0)
+; RV32I-NEXT: lw s2, 4(a0)
+; RV32I-NEXT: lw s5, 8(a0)
+; RV32I-NEXT: lw s4, 12(a0)
+; RV32I-NEXT: #APP
+; RV32I-NEXT: #NO_APP
+; RV32I-NEXT: lw s2, 12(sp) # 4-byte Folded Reload
+; RV32I-NEXT: lw s3, 8(sp) # 4-byte Folded Reload
+; RV32I-NEXT: lw s4, 4(sp) # 4-byte Folded Reload
+; RV32I-NEXT: lw s5, 0(sp) # 4-byte Folded Reload
+; RV32I-NEXT: .cfi_restore s2
+; RV32I-NEXT: .cfi_restore s3
+; RV32I-NEXT: .cfi_restore s4
+; RV32I-NEXT: .cfi_restore s5
+; RV32I-NEXT: addi sp, sp, 16
+; RV32I-NEXT: .cfi_def_cfa_offset 0
+; RV32I-NEXT: ret
+;
+; RV32D-LABEL: testi:
+; RV32D: # %bb.0: # %entry
+; RV32D-NEXT: addi sp, sp, -16
+; RV32D-NEXT: .cfi_def_cfa_offset 16
+; RV32D-NEXT: sw s2, 12(sp) # 4-byte Folded Spill
+; RV32D-NEXT: sw s3, 8(sp) # 4-byte Folded Spill
+; RV32D-NEXT: sw s4, 4(sp) # 4-byte Folded Spill
+; RV32D-NEXT: sw s5, 0(sp) # 4-byte Folded Spill
+; RV32D-NEXT: .cfi_offset s2, -4
+; RV32D-NEXT: .cfi_offset s3, -8
+; RV32D-NEXT: .cfi_offset s4, -12
+; RV32D-NEXT: .cfi_offset s5, -16
+; RV32D-NEXT: lw s3, 0(a0)
+; RV32D-NEXT: lw s2, 4(a0)
+; RV32D-NEXT: lw s5, 8(a0)
+; RV32D-NEXT: lw s4, 12(a0)
+; RV32D-NEXT: #APP
+; RV32D-NEXT: #NO_APP
+; RV32D-NEXT: lw s2, 12(sp) # 4-byte Folded Reload
+; RV32D-NEXT: lw s3, 8(sp) # 4-byte Folded Reload
+; RV32D-NEXT: lw s4, 4(sp) # 4-byte Folded Reload
+; RV32D-NEXT: lw s5, 0(sp) # 4-byte Folded Reload
+; RV32D-NEXT: .cfi_restore s2
+; RV32D-NEXT: .cfi_restore s3
+; RV32D-NEXT: .cfi_restore s4
+; RV32D-NEXT: .cfi_restore s5
+; RV32D-NEXT: addi sp, sp, 16
+; RV32D-NEXT: .cfi_def_cfa_offset 0
+; RV32D-NEXT: ret
+;
+; RV64I-LABEL: testi:
+; RV64I: # %bb.0: # %entry
+; RV64I-NEXT: addi sp, sp, -32
+; RV64I-NEXT: .cfi_def_cfa_offset 32
+; RV64I-NEXT: sd s2, 24(sp) # 8-byte Folded Spill
+; RV64I-NEXT: sd s3, 16(sp) # 8-byte Folded Spill
+; RV64I-NEXT: sd s4, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT: sd s5, 0(sp) # 8-byte Folded Spill
+; RV64I-NEXT: .cfi_offset s2, -8
+; RV64I-NEXT: .cfi_offset s3, -16
+; RV64I-NEXT: .cfi_offset s4, -24
+; RV64I-NEXT: .cfi_offset s5, -32
+; RV64I-NEXT: ld s3, 0(a0)
+; RV64I-NEXT: ld s2, 8(a0)
+; RV64I-NEXT: ld s5, 16(a0)
+; RV64I-NEXT: ld s4, 24(a0)
+; RV64I-NEXT: #APP
+; RV64I-NEXT: #NO_APP
+; RV64I-NEXT: ld s2, 24(sp) # 8-byte Folded Reload
+; RV64I-NEXT: ld s3, 16(sp) # 8-byte Folded Reload
+; RV64I-NEXT: ld s4, 8(sp) # 8-byte Folded Reload
+; RV64I-NEXT: ld s5, 0(sp) # 8-byte Folded Reload
+; RV64I-NEXT: .cfi_restore s2
+; RV64I-NEXT: .cfi_restore s3
+; RV64I-NEXT: .cfi_restore s4
+; RV64I-NEXT: .cfi_restore s5
+; RV64I-NEXT: addi sp, sp, 32
+; RV64I-NEXT: .cfi_def_cfa_offset 0
+; RV64I-NEXT: ret
+;
+; RV64D-LABEL: testi:
+; RV64D: # %bb.0: # %entry
+; RV64D-NEXT: addi sp, sp, -32
+; RV64D-NEXT: .cfi_def_cfa_offset 32
+; RV64D-NEXT: sd s2, 24(sp) # 8-byte Folded Spill
+; RV64D-NEXT: sd s3, 16(sp) # 8-byte Folded Spill
+; RV64D-NEXT: sd s4, 8(sp) # 8-byte Folded Spill
+; RV64D-NEXT: sd s5, 0(sp) # 8-byte Folded Spill
+; RV64D-NEXT: .cfi_offset s2, -8
+; RV64D-NEXT: .cfi_offset s3, -16
+; RV64D-NEXT: .cfi_offset s4, -24
+; RV64D-NEXT: .cfi_offset s5, -32
+; RV64D-NEXT: ld s3, 0(a0)
+; RV64D-NEXT: ld s2, 8(a0)
+; RV64D-NEXT: ld s5, 16(a0)
+; RV64D-NEXT: ld s4, 24(a0)
+; RV64D-NEXT: #APP
+; RV64D-NEXT: #NO_APP
+; RV64D-NEXT: ld s2, 24(sp) # 8-byte Folded Reload
+; RV64D-NEXT: ld s3, 16(sp) # 8-byte Folded Reload
+; RV64D-NEXT: ld s4, 8(sp) # 8-byte Folded Reload
+; RV64D-NEXT: ld s5, 0(sp) # 8-byte Folded Reload
+; RV64D-NEXT: .cfi_restore s2
+; RV64D-NEXT: .cfi_restore s3
+; RV64D-NEXT: .cfi_restore s4
+; RV64D-NEXT: .cfi_restore s5
+; RV64D-NEXT: addi sp, sp, 32
+; RV64D-NEXT: .cfi_def_cfa_offset 0
+; RV64D-NEXT: ret
+;
+; RV32I_PAIR-LABEL: testi:
+; RV32I_PAIR: # %bb.0: # %entry
+; RV32I_PAIR-NEXT: addi sp, sp, -16
+; RV32I_PAIR-NEXT: .cfi_def_cfa_offset 16
+; RV32I_PAIR-NEXT: mips.swp s3, s2, 8(sp) # 8-byte Folded Spill
+; RV32I_PAIR-NEXT: mips.swp s5, s4, 0(sp) # 8-byte Folded Spill
+; RV32I_PAIR-NEXT: .cfi_offset s2, -4
+; RV32I_PAIR-NEXT: .cfi_offset s3, -8
+; RV32I_PAIR-NEXT: .cfi_offset s4, -12
+; RV32I_PAIR-NEXT: .cfi_offset s5, -16
+; RV32I_PAIR-NEXT: mips.lwp s3, s2, 0(a0)
+; RV32I_PAIR-NEXT: mips.lwp s5, s4, 8(a0)
+; RV32I_PAIR-NEXT: #APP
+; RV32I_PAIR-NEXT: #NO_APP
+; RV32I_PAIR-NEXT: mips.lwp s3, s2, 8(sp) # 8-byte Folded Reload
+; RV32I_PAIR-NEXT: mips.lwp s5, s4, 0(sp) # 8-byte Folded Reload
+; RV32I_PAIR-NEXT: .cfi_restore s2
+; RV32I_PAIR-NEXT: .cfi_restore s3
+; RV32I_PAIR-NEXT: .cfi_restore s4
+; RV32I_PAIR-NEXT: .cfi_restore s5
+; RV32I_PAIR-NEXT: addi sp, sp, 16
+; RV32I_PAIR-NEXT: .cfi_def_cfa_offset 0
+; RV32I_PAIR-NEXT: ret
+;
+; RV32D_PAIR-LABEL: testi:
+; RV32D_PAIR: # %bb.0: # %entry
+; RV32D_PAIR-NEXT: addi sp, sp, -16
+; RV32D_PAIR-NEXT: .cfi_def_cfa_offset 16
+; RV32D_PAIR-NEXT: mips.swp s3, s2, 8(sp) # 8-byte Folded Spill
+; RV32D_PAIR-NEXT: mips.swp s5, s4, 0(sp) # 8-byte Folded Spill
+; RV32D_PAIR-NEXT: .cfi_offset s2, -4
+; RV32D_PAIR-NEXT: .cfi_offset s3, -8
+; RV32D_PAIR-NEXT: .cfi_offset s4, -12
+; RV32D_PAIR-NEXT: .cfi_offset s5, -16
+; RV32D_PAIR-NEXT: mips.lwp s3, s2, 0(a0)
+; RV32D_PAIR-NEXT: mips.lwp s5, s4, 8(a0)
+; RV32D_PAIR-NEXT: #APP
+; RV32D_PAIR-NEXT: #NO_APP
+; RV32D_PAIR-NEXT: mips.lwp s3, s2, 8(sp) # 8-byte Folded Reload
+; RV32D_PAIR-NEXT: mips.lwp s5, s4, 0(sp) # 8-byte Folded Reload
+; RV32D_PAIR-NEXT: .cfi_restore s2
+; RV32D_PAIR-NEXT: .cfi_restore s3
+; RV32D_PAIR-NEXT: .cfi_restore s4
+; RV32D_PAIR-NEXT: .cfi_restore s5
+; RV32D_PAIR-NEXT: addi sp, sp, 16
+; RV32D_PAIR-NEXT: .cfi_def_cfa_offset 0
+; RV32D_PAIR-NEXT: ret
+;
+; RV64I_PAIR-LABEL: testi:
+; RV64I_PAIR: # %bb.0: # %entry
+; RV64I_PAIR-NEXT: addi sp, sp, -32
+; RV64I_PAIR-NEXT: .cfi_def_cfa_offset 32
+; RV64I_PAIR-NEXT: mips.sdp s3, s2, 16(sp) # 16-byte Folded Spill
+; RV64I_PAIR-NEXT: mips.sdp s5, s4, 0(sp) # 16-byte Folded Spill
+; RV64I_PAIR-NEXT: .cfi_offset s2, -8
+; RV64I_PAIR-NEXT: .cfi_offset s3, -16
+; RV64I_PAIR-NEXT: .cfi_offset s4, -24
+; RV64I_PAIR-NEXT: .cfi_offset s5, -32
+; RV64I_PAIR-NEXT: ld s3, 0(a0)
+; RV64I_PAIR-NEXT: ld s2, 8(a0)
+; RV64I_PAIR-NEXT: ld s5, 16(a0)
+; RV64I_PAIR-NEXT: ld s4, 24(a0)
+; RV64I_PAIR-NEXT: #APP
+; RV64I_PAIR-NEXT: #NO_APP
+; RV64I_PAIR-NEXT: mips.ldp s3, s2, 16(sp) # 16-byte Folded Reload
+; RV64I_PAIR-NEXT: mips.ldp s5, s4, 0(sp) # 16-byte Folded Reload
+; RV64I_PAIR-NEXT: .cfi_restore s2
+; RV64I_PAIR-NEXT: .cfi_restore s3
+; RV64I_PAIR-NEXT: .cfi_restore s4
+; RV64I_PAIR-NEXT: .cfi_restore s5
+; RV64I_PAIR-NEXT: addi sp, sp, 32
+; RV64I_PAIR-NEXT: .cfi_def_cfa_offset 0
+; RV64I_PAIR-NEXT: ret
+;
+; RV64P_8700-LABEL: testi:
+; RV64P_8700: # %bb.0: # %entry
+; RV64P_8700-NEXT: addi sp, sp, -32
+; RV64P_8700-NEXT: .cfi_def_cfa_offset 32
+; RV64P_8700-NEXT: mips.sdp s3, s2, 16(sp) # 16-byte Folded Spill
+; RV64P_8700-NEXT: mips.sdp s5, s4, 0(sp) # 16-byte Folded Spill
+; RV64P_8700-NEXT: .cfi_offset s2, -8
+; RV64P_8700-NEXT: .cfi_offset s3, -16
+; RV64P_8700-NEXT: .cfi_offset s4, -24
+; RV64P_8700-NEXT: .cfi_offset s5, -32
+; RV64P_8700-NEXT: ld s3, 0(a0)
+; RV64P_8700-NEXT: ld s2, 8(a0)
+; RV64P_8700-NEXT: ld s5, 16(a0)
+; RV64P_8700-NEXT: ld s4, 24(a0)
+; RV64P_8700-NEXT: #APP
+; RV64P_8700-NEXT: #NO_APP
+; RV64P_8700-NEXT: mips.ldp s3, s2, 16(sp) # 16-byte Folded Reload
+; RV64P_8700-NEXT: mips.ldp s5, s4, 0(sp) # 16-byte Folded Reload
+; RV64P_8700-NEXT: .cfi_restore s2
+; RV64P_8700-NEXT: .cfi_restore s3
+; RV64P_8700-NEXT: .cfi_restore s4
+; RV64P_8700-NEXT: .cfi_restore s5
+; RV64P_8700-NEXT: addi sp, sp, 32
+; RV64P_8700-NEXT: .cfi_def_cfa_offset 0
+; RV64P_8700-NEXT: ret
+;
+; RV64D_PAIR-LABEL: testi:
+; RV64D_PAIR: # %bb.0: # %entry
+; RV64D_PAIR-NEXT: addi sp, sp, -32
+; RV64D_PAIR-NEXT: .cfi_def_cfa_offset 32
+; RV64D_PAIR-NEXT: mips.sdp s3, s2, 16(sp) # 16-byte Folded Spill
+; RV64D_PAIR-NEXT: mips.sdp s5, s4, 0(sp) # 16-byte Folded Spill
+; RV64D_PAIR-NEXT: .cfi_offset s2, -8
+; RV64D_PAIR-NEXT: .cfi_offset s3, -16
+; RV64D_PAIR-NEXT: .cfi_offset s4, -24
+; RV64D_PAIR-NEXT: .cfi_offset s5, -32
+; RV64D_PAIR-NEXT: ld s3, 0(a0)
+; RV64D_PAIR-NEXT: ld s2, 8(a0)
+; RV64D_PAIR-NEXT: ld s5, 16(a0)
+; RV64D_PAIR-NEXT: ld s4, 24(a0)
+; RV64D_PAIR-NEXT: #APP
+; RV64D_PAIR-NEXT: #NO_APP
+; RV64D_PAIR-NEXT: mips.ldp s3, s2, 16(sp) # 16-byte Folded Reload
+; RV64D_PAIR-NEXT: mips.ldp s5, s4, 0(sp) # 16-byte Folded Reload
+; RV64D_PAIR-NEXT: .cfi_restore s2
+; RV64D_PAIR-NEXT: .cfi_restore s3
+; RV64D_PAIR-NEXT: .cfi_restore s4
+; RV64D_PAIR-NEXT: .cfi_restore s5
+; RV64D_PAIR-NEXT: addi sp, sp, 32
+; RV64D_PAIR-NEXT: .cfi_def_cfa_offset 0
+; RV64D_PAIR-NEXT: ret
+;
+; RV64D_8700-LABEL: testi:
+; RV64D_8700: # %bb.0: # %entry
+; RV64D_8700-NEXT: addi sp, sp, -32
+; RV64D_8700-NEXT: .cfi_def_cfa_offset 32
+; RV64D_8700-NEXT: sd s2, 24(sp) # 8-byte Folded Spill
+; RV64D_8700-NEXT: sd s3, 16(sp) # 8-byte Folded Spill
+; RV64D_8700-NEXT: sd s4, 8(sp) # 8-byte Folded Spill
+; RV64D_8700-NEXT: sd s5, 0(sp) # 8-byte Folded Spill
+; RV64D_8700-NEXT: .cfi_offset s2, -8
+; RV64D_8700-NEXT: .cfi_offset s3, -16
+; RV64D_8700-NEXT: .cfi_offset s4, -24
+; RV64D_8700-NEXT: .cfi_offset s5, -32
+; RV64D_8700-NEXT: ld s3, 0(a0)
+; RV64D_8700-NEXT: ld s2, 8(a0)
+; RV64D_8700-NEXT: ld s5, 16(a0)
+; RV64D_8700-NEXT: ld s4, 24(a0)
+; RV64D_8700-NEXT: #APP
+; RV64D_8700-NEXT: #NO_APP
+; RV64D_8700-NEXT: ld s2, 24(sp) # 8-byte Folded Reload
+; RV64D_8700-NEXT: ld s3, 16(sp) # 8-byte Folded Reload
+; RV64D_8700-NEXT: ld s4, 8(sp) # 8-byte Folded Reload
+; RV64D_8700-NEXT: ld s5, 0(sp) # 8-byte Folded Reload
+; RV64D_8700-NEXT: .cfi_restore s2
+; RV64D_8700-NEXT: .cfi_restore s3
+; RV64D_8700-NEXT: .cfi_restore s4
+; RV64D_8700-NEXT: .cfi_restore s5
+; RV64D_8700-NEXT: addi sp, sp, 32
+; RV64D_8700-NEXT: .cfi_def_cfa_offset 0
+; RV64D_8700-NEXT: ret
+entry:
+ %arrayidx = getelementptr inbounds i8*, i8** %a, i64 1
+ %0 = load i8*, i8** %arrayidx, align 8
+ %1 = load i8*, i8** %a, align 8
+ %arrayidx2 = getelementptr inbounds i8*, i8** %a, i64 3
+ %2 = load i8*, i8** %arrayidx2, align 8
+ %arrayidx3 = getelementptr inbounds i8*, i8** %a, i64 2
+ %3 = load i8*, i8** %arrayidx3, align 8
+ tail call void asm sideeffect "", "{x18},{x19},{x20},{x21}"(i8* %0, i8* %1, i8* %2, i8* %3)
+ ret void
+}
+
+
+define dso_local void @testf(float* nocapture noundef readonly %a) local_unnamed_addr #0 {
+; RV32I-LABEL: testf:
+; RV32I: # %bb.0: # %entry
+; RV32I-NEXT: lw a3, 0(a0)
+; RV32I-NEXT: lw a4, 4(a0)
+; RV32I-NEXT: lw a2, 8(a0)
+; RV32I-NEXT: lw a1, 12(a0)
+; RV32I-NEXT: mv a0, a4
+; RV32I-NEXT: tail sinkf
+;
+; RV32D-LABEL: testf:
+; RV32D: # %bb.0: # %entry
+; RV32D-NEXT: flw fa3, 0(a0)
+; RV32D-NEXT: flw fa0, 4(a0)
+; RV32D-NEXT: flw fa2, 8(a0)
+; RV32D-NEXT: flw fa1, 12(a0)
+; RV32D-NEXT: tail sinkf
+;
+; RV64I-LABEL: testf:
+; RV64I: # %bb.0: # %entry
+; RV64I-NEXT: lw a3, 0(a0)
+; RV64I-NEXT: lw a4, 4(a0)
+; RV64I-NEXT: lw a2, 8(a0)
+; RV64I-NEXT: lw a1, 12(a0)
+; RV64I-NEXT: mv a0, a4
+; RV64I-NEXT: tail sinkf
+;
+; RV64D-LABEL: testf:
+; RV64D: # %bb.0: # %entry
+; RV64D-NEXT: flw fa3, 0(a0)
+; RV64D-NEXT: flw fa0, 4(a0)
+; RV64D-NEXT: flw fa2, 8(a0)
+; RV64D-NEXT: flw fa1, 12(a0)
+; RV64D-NEXT: tail sinkf
+;
+; RV32I_PAIR-LABEL: testf:
+; RV32I_PAIR: # %bb.0: # %entry
+; RV32I_PAIR-NEXT: lw a3, 0(a0)
+; RV32I_PAIR-NEXT: lw a4, 4(a0)
+; RV32I_PAIR-NEXT: lw a2, 8(a0)
+; RV32I_PAIR-NEXT: lw a1, 12(a0)
+; RV32I_PAIR-NEXT: mv a0, a4
+; RV32I_PAIR-NEXT: tail sinkf
+;
+; RV32D_PAIR-LABEL: testf:
+; RV32D_PAIR: # %bb.0: # %entry
+; RV32D_PAIR-NEXT: flw fa3, 0(a0)
+; RV32D_PAIR-NEXT: flw fa0, 4(a0)
+; RV32D_PAIR-NEXT: flw fa2, 8(a0)
+; RV32D_PAIR-NEXT: flw fa1, 12(a0)
+; RV32D_PAIR-NEXT: tail sinkf
+;
+; RV64I_PAIR-LABEL: testf:
+; RV64I_PAIR: # %bb.0: # %entry
+; RV64I_PAIR-NEXT: lw a3, 0(a0)
+; RV64I_PAIR-NEXT: lw a4, 4(a0)
+; RV64I_PAIR-NEXT: lw a2, 8(a0)
+; RV64I_PAIR-NEXT: lw a1, 12(a0)
+; RV64I_PAIR-NEXT: mv a0, a4
+; RV64I_PAIR-NEXT: tail sinkf
+;
+; RV64P_8700-LABEL: testf:
+; RV64P_8700: # %bb.0: # %entry
+; RV64P_8700-NEXT: flw fa3, 0(a0)
+; RV64P_8700-NEXT: flw fa0, 4(a0)
+; RV64P_8700-NEXT: flw fa2, 8(a0)
+; RV64P_8700-NEXT: flw fa1, 12(a0)
+; RV64P_8700-NEXT: tail sinkf
+;
+; RV64D_PAIR-LABEL: testf:
+; RV64D_PAIR: # %bb.0: # %entry
+; RV64D_PAIR-NEXT: flw fa3, 0(a0)
+; RV64D_PAIR-NEXT: flw fa0, 4(a0)
+; RV64D_PAIR-NEXT: flw fa2, 8(a0)
+; RV64D_PAIR-NEXT: flw fa1, 12(a0)
+; RV64D_PAIR-NEXT: tail sinkf
+;
+; RV64D_8700-LABEL: testf:
+; RV64D_8700: # %bb.0: # %entry
+; RV64D_8700-NEXT: flw fa3, 0(a0)
+; RV64D_8700-NEXT: flw fa0, 4(a0)
+; RV64D_8700-NEXT: flw fa2, 8(a0)
+; RV64D_8700-NEXT: flw fa1, 12(a0)
+; RV64D_8700-NEXT: tail sinkf
+entry:
+ %arrayidx = getelementptr inbounds float, float* %a, i64 1
+ %0 = load float, float* %arrayidx, align 4
+ %arrayidx1 = getelementptr inbounds float, float* %a, i64 3
+ %1 = load float, float* %arrayidx1, align 4
+ %arrayidx2 = getelementptr inbounds float, float* %a, i64 2
+ %2 = load float, float* %arrayidx2, align 4
+ %3 = load float, float* %a, align 4
+ tail call void @sinkf(float noundef %0, float noundef %1, float noundef %2, float noundef %3)
+ ret void
+}
+
+declare dso_local void @sinkf(float noundef, float noundef, float noundef, float noundef) local_unnamed_addr
+
+define dso_local void @testd(double* nocapture noundef readonly %a) local_unnamed_addr #0 {
----------------
mshockwave wrote:
nit: `testf` and `testd` seems to be negative tests (please correct me if I'm wrong), could you either add comments or rename the function to make it more obvious?
https://github.com/llvm/llvm-project/pull/124717
More information about the llvm-commits
mailing list