[llvm-branch-commits] [llvm] 99001e3 - [RISCV] Correctly restore libcall CSRs with SCS
Douglas Yung via llvm-branch-commits
llvm-branch-commits at lists.llvm.org
Wed Oct 28 12:49:28 PDT 2026
Author: Sam Elliott
Date: 2026-10-28T19:49:07Z
New Revision: 99001e325442d2e36576200f54face3713fdf60f
URL: https://github.com/llvm/llvm-project/commit/99001e325442d2e36576200f54face3713fdf60f
DIFF: https://github.com/llvm/llvm-project/commit/99001e325442d2e36576200f54face3713fdf60f.diff
LOG: [RISCV] Correctly restore libcall CSRs with SCS
Fixes #233318
Before this change, we were restoring some callee-saved registers twice
when using save/restore libcalls with shadow call stack (SCS), and the
second restore was happening after the stack was deallocated using the
wrong register and offset calculation, which was causing lots of access
problems.
The problems were:
- We were using the entire list of Callee Saves rather than just the
subset that would have been restored by the libcall. This caused
duplicate restores because of how the CSI subsetting works.
- We were not using the correct offset calculation in
eliminateFrameIndex, which was causing us to calculate the wrong
address for the saves/restores.
Assisted-by: AI
Co-authored-by: Mitch Briles <mitchbriles at gmail.com>
Added:
llvm/test/CodeGen/RISCV/shadow-stack-frame-opts-v.ll
Modified:
llvm/lib/Target/RISCV/RISCVFrameLowering.cpp
llvm/test/CodeGen/RISCV/shadow-stack-frame-opts.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp b/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp
index 0cd2394229ff8c..2ccbceeda5907b 100644
--- a/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp
@@ -580,7 +580,10 @@ getUnmanagedCSI(const MachineFunction &MF,
for (auto &CS : CSI) {
int FI = CS.getFrameIdx();
- if (FI >= 0 && MFI.getStackID(FI) == TargetStackID::Default)
+ // Callee saves with negative frame indices have fixed positions on the
+ // stack, usually relating to libcalls or push/pop.
+ bool IsManagedFI = (FI < 0);
+ if (!IsManagedFI && MFI.getStackID(FI) == TargetStackID::Default)
NonLibcallCSI.push_back(CS);
}
@@ -1576,7 +1579,12 @@ RISCVFrameLowering::getFrameIndexReference(const MachineFunction &MF, int FI,
MaxCSFI = std::max(CSI.front().getFrameIdx(), CSI.back().getFrameIdx());
}
- if (FI >= MinCSFI && FI <= MaxCSFI) {
+ auto FIIsCalleeSave = [&](const CalleeSavedInfo &CS) -> bool {
+ return FI == CS.getFrameIdx();
+ };
+ bool IsInlineRestoredLibcallCSR = llvm::any_of(
+ getPushOrLibCallsSavedInfo(MF, MFI.getCalleeSavedInfo()), FIIsCalleeSave);
+ if ((FI >= MinCSFI && FI <= MaxCSFI) || IsInlineRestoredLibcallCSR) {
FrameReg = SPReg;
if (FirstSPAdjustAmount)
@@ -2552,10 +2560,13 @@ bool RISCVFrameLowering::restoreCalleeSavedRegisters(
PopBuilder.addDef(FixedCSRFIMap[i], RegState::ImplicitDefine);
}
} else if (const char *RestoreLibCall = getRestoreLibCallName(*MF, CSI)) {
+ assert(RVFI->useSaveRestoreLibCalls(*MF) &&
+ "Expected restore with libcall to use save restore libcalls");
+
// Restore is not compatible with shadow stacks, so do the restore manually.
// This ensures we still get the code size saving of `__riscv_save_<N>`.
if (RVFI->hasShadowStack(*MF)) {
- loadRegFromStackSlot(CSI);
+ loadRegFromStackSlot(getPushOrLibCallsSavedInfo(*MF, CSI));
return true;
}
diff --git a/llvm/test/CodeGen/RISCV/shadow-stack-frame-opts-v.ll b/llvm/test/CodeGen/RISCV/shadow-stack-frame-opts-v.ll
new file mode 100644
index 00000000000000..ab43a6b7007bd1
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/shadow-stack-frame-opts-v.ll
@@ -0,0 +1,136 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+
+; RUN: llc -mtriple=riscv32 -mattr=+v,+zimop,+save-restore %s -o - | FileCheck %s --check-prefixes=RV32V
+; RUN: llc -mtriple=riscv64 -mattr=+v,+zimop,+save-restore %s -o - | FileCheck %s --check-prefixes=RV64V
+
+
+declare void @g()
+
+define <vscale x 4 x i32> @w(<vscale x 4 x i32> %a) shadowcallstack {
+; RV32V-LABEL: w:
+; RV32V: # %bb.0:
+; RV32V-NEXT: addi gp, gp, 4
+; RV32V-NEXT: sw ra, -4(gp)
+; RV32V-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x7c #
+; RV32V-NEXT: call t0, __riscv_save_0
+; RV32V-NEXT: .cfi_def_cfa_offset 16
+; RV32V-NEXT: .cfi_offset ra, -4
+; RV32V-NEXT: addi sp, sp, -16
+; RV32V-NEXT: .cfi_def_cfa_offset 32
+; RV32V-NEXT: csrr a0, vlenb
+; RV32V-NEXT: slli a0, a0, 1
+; RV32V-NEXT: sub sp, sp, a0
+; RV32V-NEXT: .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x20, 0x22, 0x11, 0x02, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 32 + 2 * vlenb
+; RV32V-NEXT: addi a0, sp, 32
+; RV32V-NEXT: vs2r.v v8, (a0) # vscale x 16-byte Folded Spill
+; RV32V-NEXT: call g
+; RV32V-NEXT: addi a0, sp, 32
+; RV32V-NEXT: vl2r.v v8, (a0) # vscale x 16-byte Folded Reload
+; RV32V-NEXT: csrr a0, vlenb
+; RV32V-NEXT: slli a0, a0, 1
+; RV32V-NEXT: add sp, sp, a0
+; RV32V-NEXT: .cfi_def_cfa sp, 32
+; RV32V-NEXT: lw ra, 28(sp) # 4-byte Folded Reload
+; RV32V-NEXT: .cfi_restore ra
+; RV32V-NEXT: addi sp, sp, 32
+; RV32V-NEXT: .cfi_def_cfa_offset 0
+; RV32V-NEXT: lw ra, -4(gp)
+; RV32V-NEXT: addi gp, gp, -4
+; RV32V-NEXT: .cfi_restore gp
+; RV32V-NEXT: ret
+;
+; RV64V-LABEL: w:
+; RV64V: # %bb.0:
+; RV64V-NEXT: addi gp, gp, 8
+; RV64V-NEXT: sd ra, -8(gp)
+; RV64V-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x78 #
+; RV64V-NEXT: call t0, __riscv_save_0
+; RV64V-NEXT: .cfi_def_cfa_offset 16
+; RV64V-NEXT: .cfi_offset ra, -8
+; RV64V-NEXT: addi sp, sp, -16
+; RV64V-NEXT: .cfi_def_cfa_offset 32
+; RV64V-NEXT: csrr a0, vlenb
+; RV64V-NEXT: slli a0, a0, 1
+; RV64V-NEXT: sub sp, sp, a0
+; RV64V-NEXT: .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x20, 0x22, 0x11, 0x02, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 32 + 2 * vlenb
+; RV64V-NEXT: addi a0, sp, 32
+; RV64V-NEXT: vs2r.v v8, (a0) # vscale x 16-byte Folded Spill
+; RV64V-NEXT: call g
+; RV64V-NEXT: addi a0, sp, 32
+; RV64V-NEXT: vl2r.v v8, (a0) # vscale x 16-byte Folded Reload
+; RV64V-NEXT: csrr a0, vlenb
+; RV64V-NEXT: slli a0, a0, 1
+; RV64V-NEXT: add sp, sp, a0
+; RV64V-NEXT: .cfi_def_cfa sp, 32
+; RV64V-NEXT: ld ra, 24(sp) # 8-byte Folded Reload
+; RV64V-NEXT: .cfi_restore ra
+; RV64V-NEXT: addi sp, sp, 32
+; RV64V-NEXT: .cfi_def_cfa_offset 0
+; RV64V-NEXT: ld ra, -8(gp)
+; RV64V-NEXT: addi gp, gp, -8
+; RV64V-NEXT: .cfi_restore gp
+; RV64V-NEXT: ret
+ call void @g()
+ ret <vscale x 4 x i32> %a
+}
+
+define <vscale x 4 x i32> @z(<vscale x 4 x i32> %a) "hw-shadow-stack" {
+; RV32V-LABEL: z:
+; RV32V: # %bb.0:
+; RV32V-NEXT: sspush ra
+; RV32V-NEXT: call t0, __riscv_save_0
+; RV32V-NEXT: .cfi_def_cfa_offset 16
+; RV32V-NEXT: .cfi_offset ra, -4
+; RV32V-NEXT: addi sp, sp, -16
+; RV32V-NEXT: .cfi_def_cfa_offset 32
+; RV32V-NEXT: csrr a0, vlenb
+; RV32V-NEXT: slli a0, a0, 1
+; RV32V-NEXT: sub sp, sp, a0
+; RV32V-NEXT: .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x20, 0x22, 0x11, 0x02, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 32 + 2 * vlenb
+; RV32V-NEXT: addi a0, sp, 32
+; RV32V-NEXT: vs2r.v v8, (a0) # vscale x 16-byte Folded Spill
+; RV32V-NEXT: call g
+; RV32V-NEXT: addi a0, sp, 32
+; RV32V-NEXT: vl2r.v v8, (a0) # vscale x 16-byte Folded Reload
+; RV32V-NEXT: csrr a0, vlenb
+; RV32V-NEXT: slli a0, a0, 1
+; RV32V-NEXT: add sp, sp, a0
+; RV32V-NEXT: .cfi_def_cfa sp, 32
+; RV32V-NEXT: lw ra, 28(sp) # 4-byte Folded Reload
+; RV32V-NEXT: .cfi_restore ra
+; RV32V-NEXT: addi sp, sp, 32
+; RV32V-NEXT: .cfi_def_cfa_offset 0
+; RV32V-NEXT: sspopchk ra
+; RV32V-NEXT: ret
+;
+; RV64V-LABEL: z:
+; RV64V: # %bb.0:
+; RV64V-NEXT: sspush ra
+; RV64V-NEXT: call t0, __riscv_save_0
+; RV64V-NEXT: .cfi_def_cfa_offset 16
+; RV64V-NEXT: .cfi_offset ra, -8
+; RV64V-NEXT: addi sp, sp, -16
+; RV64V-NEXT: .cfi_def_cfa_offset 32
+; RV64V-NEXT: csrr a0, vlenb
+; RV64V-NEXT: slli a0, a0, 1
+; RV64V-NEXT: sub sp, sp, a0
+; RV64V-NEXT: .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x20, 0x22, 0x11, 0x02, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 32 + 2 * vlenb
+; RV64V-NEXT: addi a0, sp, 32
+; RV64V-NEXT: vs2r.v v8, (a0) # vscale x 16-byte Folded Spill
+; RV64V-NEXT: call g
+; RV64V-NEXT: addi a0, sp, 32
+; RV64V-NEXT: vl2r.v v8, (a0) # vscale x 16-byte Folded Reload
+; RV64V-NEXT: csrr a0, vlenb
+; RV64V-NEXT: slli a0, a0, 1
+; RV64V-NEXT: add sp, sp, a0
+; RV64V-NEXT: .cfi_def_cfa sp, 32
+; RV64V-NEXT: ld ra, 24(sp) # 8-byte Folded Reload
+; RV64V-NEXT: .cfi_restore ra
+; RV64V-NEXT: addi sp, sp, 32
+; RV64V-NEXT: .cfi_def_cfa_offset 0
+; RV64V-NEXT: sspopchk ra
+; RV64V-NEXT: ret
+ call void @g()
+ ret <vscale x 4 x i32> %a
+}
+
diff --git a/llvm/test/CodeGen/RISCV/shadow-stack-frame-opts.ll b/llvm/test/CodeGen/RISCV/shadow-stack-frame-opts.ll
index e73c461eaa81f0..8c872fdc6d5c23 100644
--- a/llvm/test/CodeGen/RISCV/shadow-stack-frame-opts.ll
+++ b/llvm/test/CodeGen/RISCV/shadow-stack-frame-opts.ll
@@ -5,6 +5,8 @@
; RUN: | FileCheck %s --check-prefix=RV32
; RUN: llc -mtriple=riscv32 -mattr=+zimop,+save-restore < %s \
; RUN: | FileCheck %s --check-prefix=RV32-SAVE-RESTORE
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zimop,+save-restore -target-abi=ilp32d < %s \
+; RUN: | FileCheck %s --check-prefix=RV32D-SAVE-RESTORE
; RUN: llc -mtriple=riscv32 -mattr=+zimop,+zcmp < %s \
; RUN: | FileCheck %s --check-prefix=RV32-ZCMP
; RUN: llc -mtriple=riscv32 -mattr=+zimop,+zcmp,+save-restore < %s \
@@ -16,6 +18,8 @@
; RUN: | FileCheck %s --check-prefix=RV64
; RUN: llc -mtriple=riscv64 -mattr=+zimop,+save-restore < %s \
; RUN: | FileCheck %s --check-prefix=RV64-SAVE-RESTORE
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zimop,+save-restore -target-abi=lp64d < %s \
+; RUN: | FileCheck %s --check-prefix=RV64D-SAVE-RESTORE
; RUN: llc -mtriple=riscv64 -mattr=+zimop,+zcmp < %s \
; RUN: | FileCheck %s --check-prefix=RV64-ZCMP
; RUN: llc -mtriple=riscv64 -mattr=+zimop,+zcmp,+save-restore < %s \
@@ -24,6 +28,7 @@
; RUN: | FileCheck %s --check-prefix=RV64-XQCCMP
declare i32 @callee(i32 %x)
+declare float @callee_float(float %x)
;; Software Shadow Call Stack
@@ -69,6 +74,26 @@ define i32 @caller_sw(i32 %x) shadowcallstack {
; RV32-SAVE-RESTORE-NEXT: .cfi_restore gp
; RV32-SAVE-RESTORE-NEXT: ret
;
+; RV32D-SAVE-RESTORE-LABEL: caller_sw:
+; RV32D-SAVE-RESTORE: # %bb.0:
+; RV32D-SAVE-RESTORE-NEXT: addi gp, gp, 4
+; RV32D-SAVE-RESTORE-NEXT: sw ra, -4(gp)
+; RV32D-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x7c #
+; RV32D-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset ra, -4
+; RV32D-SAVE-RESTORE-NEXT: neg a0, a0
+; RV32D-SAVE-RESTORE-NEXT: call callee
+; RV32D-SAVE-RESTORE-NEXT: addi a0, a0, 1
+; RV32D-SAVE-RESTORE-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV32D-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV32D-SAVE-RESTORE-NEXT: lw ra, -4(gp)
+; RV32D-SAVE-RESTORE-NEXT: addi gp, gp, -4
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV32D-SAVE-RESTORE-NEXT: ret
+;
; RV32-ZCMP-LABEL: caller_sw:
; RV32-ZCMP: # %bb.0:
; RV32-ZCMP-NEXT: addi gp, gp, 4
@@ -161,6 +186,26 @@ define i32 @caller_sw(i32 %x) shadowcallstack {
; RV64-SAVE-RESTORE-NEXT: .cfi_restore gp
; RV64-SAVE-RESTORE-NEXT: ret
;
+; RV64D-SAVE-RESTORE-LABEL: caller_sw:
+; RV64D-SAVE-RESTORE: # %bb.0:
+; RV64D-SAVE-RESTORE-NEXT: addi gp, gp, 8
+; RV64D-SAVE-RESTORE-NEXT: sd ra, -8(gp)
+; RV64D-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x78 #
+; RV64D-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset ra, -8
+; RV64D-SAVE-RESTORE-NEXT: negw a0, a0
+; RV64D-SAVE-RESTORE-NEXT: call callee
+; RV64D-SAVE-RESTORE-NEXT: addiw a0, a0, 1
+; RV64D-SAVE-RESTORE-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV64D-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV64D-SAVE-RESTORE-NEXT: ld ra, -8(gp)
+; RV64D-SAVE-RESTORE-NEXT: addi gp, gp, -8
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV64D-SAVE-RESTORE-NEXT: ret
+;
; RV64-ZCMP-LABEL: caller_sw:
; RV64-ZCMP: # %bb.0:
; RV64-ZCMP-NEXT: addi gp, gp, 8
@@ -228,6 +273,11 @@ define i32 @caller_sw_tail(i32 %x) shadowcallstack {
; RV32-SAVE-RESTORE-NEXT: neg a0, a0
; RV32-SAVE-RESTORE-NEXT: tail callee
;
+; RV32D-SAVE-RESTORE-LABEL: caller_sw_tail:
+; RV32D-SAVE-RESTORE: # %bb.0:
+; RV32D-SAVE-RESTORE-NEXT: neg a0, a0
+; RV32D-SAVE-RESTORE-NEXT: tail callee
+;
; RV32-ZCMP-LABEL: caller_sw_tail:
; RV32-ZCMP: # %bb.0:
; RV32-ZCMP-NEXT: neg a0, a0
@@ -253,6 +303,11 @@ define i32 @caller_sw_tail(i32 %x) shadowcallstack {
; RV64-SAVE-RESTORE-NEXT: negw a0, a0
; RV64-SAVE-RESTORE-NEXT: tail callee
;
+; RV64D-SAVE-RESTORE-LABEL: caller_sw_tail:
+; RV64D-SAVE-RESTORE: # %bb.0:
+; RV64D-SAVE-RESTORE-NEXT: negw a0, a0
+; RV64D-SAVE-RESTORE-NEXT: tail callee
+;
; RV64-ZCMP-LABEL: caller_sw_tail:
; RV64-ZCMP: # %bb.0:
; RV64-ZCMP-NEXT: negw a0, a0
@@ -327,6 +382,32 @@ define i32 @caller_sw_fp(i32 %x) shadowcallstack "frame-pointer"="all" {
; RV32-SAVE-RESTORE-NEXT: .cfi_restore gp
; RV32-SAVE-RESTORE-NEXT: ret
;
+; RV32D-SAVE-RESTORE-LABEL: caller_sw_fp:
+; RV32D-SAVE-RESTORE: # %bb.0:
+; RV32D-SAVE-RESTORE-NEXT: addi gp, gp, 4
+; RV32D-SAVE-RESTORE-NEXT: sw ra, -4(gp)
+; RV32D-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x7c #
+; RV32D-SAVE-RESTORE-NEXT: call t0, __riscv_save_1
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset ra, -4
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset s0, -8
+; RV32D-SAVE-RESTORE-NEXT: addi s0, sp, 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa s0, 0
+; RV32D-SAVE-RESTORE-NEXT: neg a0, a0
+; RV32D-SAVE-RESTORE-NEXT: call callee
+; RV32D-SAVE-RESTORE-NEXT: addi a0, a0, 1
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa sp, 16
+; RV32D-SAVE-RESTORE-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore s0
+; RV32D-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV32D-SAVE-RESTORE-NEXT: lw ra, -4(gp)
+; RV32D-SAVE-RESTORE-NEXT: addi gp, gp, -4
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV32D-SAVE-RESTORE-NEXT: ret
+;
; RV32-ZCMP-LABEL: caller_sw_fp:
; RV32-ZCMP: # %bb.0:
; RV32-ZCMP-NEXT: addi gp, gp, 4
@@ -455,6 +536,32 @@ define i32 @caller_sw_fp(i32 %x) shadowcallstack "frame-pointer"="all" {
; RV64-SAVE-RESTORE-NEXT: .cfi_restore gp
; RV64-SAVE-RESTORE-NEXT: ret
;
+; RV64D-SAVE-RESTORE-LABEL: caller_sw_fp:
+; RV64D-SAVE-RESTORE: # %bb.0:
+; RV64D-SAVE-RESTORE-NEXT: addi gp, gp, 8
+; RV64D-SAVE-RESTORE-NEXT: sd ra, -8(gp)
+; RV64D-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x78 #
+; RV64D-SAVE-RESTORE-NEXT: call t0, __riscv_save_1
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset ra, -8
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset s0, -16
+; RV64D-SAVE-RESTORE-NEXT: addi s0, sp, 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa s0, 0
+; RV64D-SAVE-RESTORE-NEXT: negw a0, a0
+; RV64D-SAVE-RESTORE-NEXT: call callee
+; RV64D-SAVE-RESTORE-NEXT: addiw a0, a0, 1
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa sp, 16
+; RV64D-SAVE-RESTORE-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: ld s0, 0(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore s0
+; RV64D-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV64D-SAVE-RESTORE-NEXT: ld ra, -8(gp)
+; RV64D-SAVE-RESTORE-NEXT: addi gp, gp, -8
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV64D-SAVE-RESTORE-NEXT: ret
+;
; RV64-ZCMP-LABEL: caller_sw_fp:
; RV64-ZCMP: # %bb.0:
; RV64-ZCMP-NEXT: addi gp, gp, 8
@@ -586,6 +693,31 @@ define i32 @caller_sw_cond(i32 %x) shadowcallstack {
; RV32-SAVE-RESTORE-NEXT: li a0, 1
; RV32-SAVE-RESTORE-NEXT: ret
;
+; RV32D-SAVE-RESTORE-LABEL: caller_sw_cond:
+; RV32D-SAVE-RESTORE: # %bb.0:
+; RV32D-SAVE-RESTORE-NEXT: beqz a0, .LBB3_2
+; RV32D-SAVE-RESTORE-NEXT: # %bb.1: # %if.end
+; RV32D-SAVE-RESTORE-NEXT: addi gp, gp, 4
+; RV32D-SAVE-RESTORE-NEXT: sw ra, -4(gp)
+; RV32D-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x7c #
+; RV32D-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset ra, -4
+; RV32D-SAVE-RESTORE-NEXT: neg a0, a0
+; RV32D-SAVE-RESTORE-NEXT: call callee
+; RV32D-SAVE-RESTORE-NEXT: addi a0, a0, 1
+; RV32D-SAVE-RESTORE-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV32D-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV32D-SAVE-RESTORE-NEXT: lw ra, -4(gp)
+; RV32D-SAVE-RESTORE-NEXT: addi gp, gp, -4
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV32D-SAVE-RESTORE-NEXT: ret
+; RV32D-SAVE-RESTORE-NEXT: .LBB3_2: # %return
+; RV32D-SAVE-RESTORE-NEXT: li a0, 1
+; RV32D-SAVE-RESTORE-NEXT: ret
+;
; RV32-ZCMP-LABEL: caller_sw_cond:
; RV32-ZCMP: # %bb.0:
; RV32-ZCMP-NEXT: beqz a0, .LBB3_2
@@ -705,6 +837,32 @@ define i32 @caller_sw_cond(i32 %x) shadowcallstack {
; RV64-SAVE-RESTORE-NEXT: li a0, 1
; RV64-SAVE-RESTORE-NEXT: ret
;
+; RV64D-SAVE-RESTORE-LABEL: caller_sw_cond:
+; RV64D-SAVE-RESTORE: # %bb.0:
+; RV64D-SAVE-RESTORE-NEXT: sext.w a1, a0
+; RV64D-SAVE-RESTORE-NEXT: beqz a1, .LBB3_2
+; RV64D-SAVE-RESTORE-NEXT: # %bb.1: # %if.end
+; RV64D-SAVE-RESTORE-NEXT: addi gp, gp, 8
+; RV64D-SAVE-RESTORE-NEXT: sd ra, -8(gp)
+; RV64D-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x78 #
+; RV64D-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset ra, -8
+; RV64D-SAVE-RESTORE-NEXT: negw a0, a0
+; RV64D-SAVE-RESTORE-NEXT: call callee
+; RV64D-SAVE-RESTORE-NEXT: addiw a0, a0, 1
+; RV64D-SAVE-RESTORE-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV64D-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV64D-SAVE-RESTORE-NEXT: ld ra, -8(gp)
+; RV64D-SAVE-RESTORE-NEXT: addi gp, gp, -8
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV64D-SAVE-RESTORE-NEXT: ret
+; RV64D-SAVE-RESTORE-NEXT: .LBB3_2: # %return
+; RV64D-SAVE-RESTORE-NEXT: li a0, 1
+; RV64D-SAVE-RESTORE-NEXT: ret
+;
; RV64-ZCMP-LABEL: caller_sw_cond:
; RV64-ZCMP: # %bb.0:
; RV64-ZCMP-NEXT: sext.w a1, a0
@@ -841,6 +999,32 @@ define i32 @caller_sw_clobber_s11(i32 %x) shadowcallstack {
; RV32-SAVE-RESTORE-NEXT: .cfi_restore gp
; RV32-SAVE-RESTORE-NEXT: ret
;
+; RV32D-SAVE-RESTORE-LABEL: caller_sw_clobber_s11:
+; RV32D-SAVE-RESTORE: # %bb.0:
+; RV32D-SAVE-RESTORE-NEXT: addi gp, gp, 4
+; RV32D-SAVE-RESTORE-NEXT: sw ra, -4(gp)
+; RV32D-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x7c #
+; RV32D-SAVE-RESTORE-NEXT: call t0, __riscv_save_12
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 64
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset ra, -4
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset s11, -52
+; RV32D-SAVE-RESTORE-NEXT: neg a0, a0
+; RV32D-SAVE-RESTORE-NEXT: call callee
+; RV32D-SAVE-RESTORE-NEXT: addi a0, a0, 1
+; RV32D-SAVE-RESTORE-NEXT: #APP
+; RV32D-SAVE-RESTORE-NEXT: nop
+; RV32D-SAVE-RESTORE-NEXT: #NO_APP
+; RV32D-SAVE-RESTORE-NEXT: lw ra, 60(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: lw s11, 12(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore s11
+; RV32D-SAVE-RESTORE-NEXT: addi sp, sp, 64
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV32D-SAVE-RESTORE-NEXT: lw ra, -4(gp)
+; RV32D-SAVE-RESTORE-NEXT: addi gp, gp, -4
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV32D-SAVE-RESTORE-NEXT: ret
+;
; RV32-ZCMP-LABEL: caller_sw_clobber_s11:
; RV32-ZCMP: # %bb.0:
; RV32-ZCMP-NEXT: addi gp, gp, 4
@@ -958,6 +1142,32 @@ define i32 @caller_sw_clobber_s11(i32 %x) shadowcallstack {
; RV64-SAVE-RESTORE-NEXT: .cfi_restore gp
; RV64-SAVE-RESTORE-NEXT: ret
;
+; RV64D-SAVE-RESTORE-LABEL: caller_sw_clobber_s11:
+; RV64D-SAVE-RESTORE: # %bb.0:
+; RV64D-SAVE-RESTORE-NEXT: addi gp, gp, 8
+; RV64D-SAVE-RESTORE-NEXT: sd ra, -8(gp)
+; RV64D-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x78 #
+; RV64D-SAVE-RESTORE-NEXT: call t0, __riscv_save_12
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 112
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset ra, -8
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset s11, -104
+; RV64D-SAVE-RESTORE-NEXT: negw a0, a0
+; RV64D-SAVE-RESTORE-NEXT: call callee
+; RV64D-SAVE-RESTORE-NEXT: addiw a0, a0, 1
+; RV64D-SAVE-RESTORE-NEXT: #APP
+; RV64D-SAVE-RESTORE-NEXT: nop
+; RV64D-SAVE-RESTORE-NEXT: #NO_APP
+; RV64D-SAVE-RESTORE-NEXT: ld ra, 104(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: ld s11, 8(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore s11
+; RV64D-SAVE-RESTORE-NEXT: addi sp, sp, 112
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV64D-SAVE-RESTORE-NEXT: ld ra, -8(gp)
+; RV64D-SAVE-RESTORE-NEXT: addi gp, gp, -8
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV64D-SAVE-RESTORE-NEXT: ret
+;
; RV64-ZCMP-LABEL: caller_sw_clobber_s11:
; RV64-ZCMP: # %bb.0:
; RV64-ZCMP-NEXT: addi gp, gp, 8
@@ -1027,6 +1237,322 @@ define i32 @caller_sw_clobber_s11(i32 %x) shadowcallstack {
ret i32 %add
}
+define float @caller_sw_float(float %x, float %y, float %z) shadowcallstack {
+; RV32-LABEL: caller_sw_float:
+; RV32: # %bb.0:
+; RV32-NEXT: addi gp, gp, 4
+; RV32-NEXT: sw ra, -4(gp)
+; RV32-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x7c #
+; RV32-NEXT: addi sp, sp, -16
+; RV32-NEXT: .cfi_def_cfa_offset 16
+; RV32-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
+; RV32-NEXT: sw s0, 8(sp) # 4-byte Folded Spill
+; RV32-NEXT: sw s1, 4(sp) # 4-byte Folded Spill
+; RV32-NEXT: .cfi_offset ra, -4
+; RV32-NEXT: .cfi_offset s0, -8
+; RV32-NEXT: .cfi_offset s1, -12
+; RV32-NEXT: mv s0, a2
+; RV32-NEXT: mv s1, a1
+; RV32-NEXT: call callee_float
+; RV32-NEXT: mv a0, s1
+; RV32-NEXT: mv a1, s0
+; RV32-NEXT: call __mulsf3
+; RV32-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
+; RV32-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
+; RV32-NEXT: lw s1, 4(sp) # 4-byte Folded Reload
+; RV32-NEXT: .cfi_restore ra
+; RV32-NEXT: .cfi_restore s0
+; RV32-NEXT: .cfi_restore s1
+; RV32-NEXT: addi sp, sp, 16
+; RV32-NEXT: .cfi_def_cfa_offset 0
+; RV32-NEXT: lw ra, -4(gp)
+; RV32-NEXT: addi gp, gp, -4
+; RV32-NEXT: .cfi_restore gp
+; RV32-NEXT: ret
+;
+; RV32-SAVE-RESTORE-LABEL: caller_sw_float:
+; RV32-SAVE-RESTORE: # %bb.0:
+; RV32-SAVE-RESTORE-NEXT: addi gp, gp, 4
+; RV32-SAVE-RESTORE-NEXT: sw ra, -4(gp)
+; RV32-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x7c #
+; RV32-SAVE-RESTORE-NEXT: call t0, __riscv_save_2
+; RV32-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV32-SAVE-RESTORE-NEXT: .cfi_offset ra, -4
+; RV32-SAVE-RESTORE-NEXT: .cfi_offset s0, -8
+; RV32-SAVE-RESTORE-NEXT: .cfi_offset s1, -12
+; RV32-SAVE-RESTORE-NEXT: mv s0, a2
+; RV32-SAVE-RESTORE-NEXT: mv s1, a1
+; RV32-SAVE-RESTORE-NEXT: call callee_float
+; RV32-SAVE-RESTORE-NEXT: mv a0, s1
+; RV32-SAVE-RESTORE-NEXT: mv a1, s0
+; RV32-SAVE-RESTORE-NEXT: call __mulsf3
+; RV32-SAVE-RESTORE-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
+; RV32-SAVE-RESTORE-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
+; RV32-SAVE-RESTORE-NEXT: lw s1, 4(sp) # 4-byte Folded Reload
+; RV32-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV32-SAVE-RESTORE-NEXT: .cfi_restore s0
+; RV32-SAVE-RESTORE-NEXT: .cfi_restore s1
+; RV32-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV32-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV32-SAVE-RESTORE-NEXT: lw ra, -4(gp)
+; RV32-SAVE-RESTORE-NEXT: addi gp, gp, -4
+; RV32-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV32-SAVE-RESTORE-NEXT: ret
+;
+; RV32D-SAVE-RESTORE-LABEL: caller_sw_float:
+; RV32D-SAVE-RESTORE: # %bb.0:
+; RV32D-SAVE-RESTORE-NEXT: addi gp, gp, 4
+; RV32D-SAVE-RESTORE-NEXT: sw ra, -4(gp)
+; RV32D-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x7c #
+; RV32D-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset ra, -4
+; RV32D-SAVE-RESTORE-NEXT: addi sp, sp, -16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 32
+; RV32D-SAVE-RESTORE-NEXT: fsd fs0, 8(sp) # 8-byte Folded Spill
+; RV32D-SAVE-RESTORE-NEXT: fsd fs1, 0(sp) # 8-byte Folded Spill
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset fs0, -24
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset fs1, -32
+; RV32D-SAVE-RESTORE-NEXT: fmv.s fs0, fa2
+; RV32D-SAVE-RESTORE-NEXT: fmv.s fs1, fa1
+; RV32D-SAVE-RESTORE-NEXT: call callee_float
+; RV32D-SAVE-RESTORE-NEXT: fmul.s fa0, fs1, fs0
+; RV32D-SAVE-RESTORE-NEXT: fld fs0, 8(sp) # 8-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: fld fs1, 0(sp) # 8-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: lw ra, 28(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore fs0
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore fs1
+; RV32D-SAVE-RESTORE-NEXT: addi sp, sp, 32
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV32D-SAVE-RESTORE-NEXT: lw ra, -4(gp)
+; RV32D-SAVE-RESTORE-NEXT: addi gp, gp, -4
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV32D-SAVE-RESTORE-NEXT: ret
+;
+; RV32-ZCMP-LABEL: caller_sw_float:
+; RV32-ZCMP: # %bb.0:
+; RV32-ZCMP-NEXT: addi gp, gp, 4
+; RV32-ZCMP-NEXT: sw ra, -4(gp)
+; RV32-ZCMP-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x7c #
+; RV32-ZCMP-NEXT: cm.push {ra, s0-s1}, -16
+; RV32-ZCMP-NEXT: .cfi_def_cfa_offset 16
+; RV32-ZCMP-NEXT: .cfi_offset ra, -12
+; RV32-ZCMP-NEXT: .cfi_offset s0, -8
+; RV32-ZCMP-NEXT: .cfi_offset s1, -4
+; RV32-ZCMP-NEXT: mv s0, a2
+; RV32-ZCMP-NEXT: mv s1, a1
+; RV32-ZCMP-NEXT: call callee_float
+; RV32-ZCMP-NEXT: cm.mva01s s1, s0
+; RV32-ZCMP-NEXT: call __mulsf3
+; RV32-ZCMP-NEXT: cm.pop {ra, s0-s1}, 16
+; RV32-ZCMP-NEXT: lw ra, -4(gp)
+; RV32-ZCMP-NEXT: addi gp, gp, -4
+; RV32-ZCMP-NEXT: .cfi_restore gp
+; RV32-ZCMP-NEXT: ret
+;
+; RV32-ZCMP-SAVE-RESTORE-LABEL: caller_sw_float:
+; RV32-ZCMP-SAVE-RESTORE: # %bb.0:
+; RV32-ZCMP-SAVE-RESTORE-NEXT: addi gp, gp, 4
+; RV32-ZCMP-SAVE-RESTORE-NEXT: sw ra, -4(gp)
+; RV32-ZCMP-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x7c #
+; RV32-ZCMP-SAVE-RESTORE-NEXT: cm.push {ra, s0-s1}, -16
+; RV32-ZCMP-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV32-ZCMP-SAVE-RESTORE-NEXT: .cfi_offset ra, -12
+; RV32-ZCMP-SAVE-RESTORE-NEXT: .cfi_offset s0, -8
+; RV32-ZCMP-SAVE-RESTORE-NEXT: .cfi_offset s1, -4
+; RV32-ZCMP-SAVE-RESTORE-NEXT: mv s0, a2
+; RV32-ZCMP-SAVE-RESTORE-NEXT: mv s1, a1
+; RV32-ZCMP-SAVE-RESTORE-NEXT: call callee_float
+; RV32-ZCMP-SAVE-RESTORE-NEXT: cm.mva01s s1, s0
+; RV32-ZCMP-SAVE-RESTORE-NEXT: call __mulsf3
+; RV32-ZCMP-SAVE-RESTORE-NEXT: cm.pop {ra, s0-s1}, 16
+; RV32-ZCMP-SAVE-RESTORE-NEXT: lw ra, -4(gp)
+; RV32-ZCMP-SAVE-RESTORE-NEXT: addi gp, gp, -4
+; RV32-ZCMP-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV32-ZCMP-SAVE-RESTORE-NEXT: ret
+;
+; RV32-XQCCMP-LABEL: caller_sw_float:
+; RV32-XQCCMP: # %bb.0:
+; RV32-XQCCMP-NEXT: addi gp, gp, 4
+; RV32-XQCCMP-NEXT: sw ra, -4(gp)
+; RV32-XQCCMP-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x7c #
+; RV32-XQCCMP-NEXT: qc.cm.push {ra, s0-s1}, -16
+; RV32-XQCCMP-NEXT: .cfi_def_cfa_offset 16
+; RV32-XQCCMP-NEXT: .cfi_offset ra, -4
+; RV32-XQCCMP-NEXT: .cfi_offset s0, -8
+; RV32-XQCCMP-NEXT: .cfi_offset s1, -12
+; RV32-XQCCMP-NEXT: mv s0, a2
+; RV32-XQCCMP-NEXT: mv s1, a1
+; RV32-XQCCMP-NEXT: call callee_float
+; RV32-XQCCMP-NEXT: qc.cm.mva01s s1, s0
+; RV32-XQCCMP-NEXT: call __mulsf3
+; RV32-XQCCMP-NEXT: qc.cm.pop {ra, s0-s1}, 16
+; RV32-XQCCMP-NEXT: lw ra, -4(gp)
+; RV32-XQCCMP-NEXT: addi gp, gp, -4
+; RV32-XQCCMP-NEXT: .cfi_restore gp
+; RV32-XQCCMP-NEXT: ret
+;
+; RV64-LABEL: caller_sw_float:
+; RV64: # %bb.0:
+; RV64-NEXT: addi gp, gp, 8
+; RV64-NEXT: sd ra, -8(gp)
+; RV64-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x78 #
+; RV64-NEXT: addi sp, sp, -32
+; RV64-NEXT: .cfi_def_cfa_offset 32
+; RV64-NEXT: sd ra, 24(sp) # 8-byte Folded Spill
+; RV64-NEXT: sd s0, 16(sp) # 8-byte Folded Spill
+; RV64-NEXT: sd s1, 8(sp) # 8-byte Folded Spill
+; RV64-NEXT: .cfi_offset ra, -8
+; RV64-NEXT: .cfi_offset s0, -16
+; RV64-NEXT: .cfi_offset s1, -24
+; RV64-NEXT: mv s0, a2
+; RV64-NEXT: mv s1, a1
+; RV64-NEXT: call callee_float
+; RV64-NEXT: mv a0, s1
+; RV64-NEXT: mv a1, s0
+; RV64-NEXT: call __mulsf3
+; RV64-NEXT: ld ra, 24(sp) # 8-byte Folded Reload
+; RV64-NEXT: ld s0, 16(sp) # 8-byte Folded Reload
+; RV64-NEXT: ld s1, 8(sp) # 8-byte Folded Reload
+; RV64-NEXT: .cfi_restore ra
+; RV64-NEXT: .cfi_restore s0
+; RV64-NEXT: .cfi_restore s1
+; RV64-NEXT: addi sp, sp, 32
+; RV64-NEXT: .cfi_def_cfa_offset 0
+; RV64-NEXT: ld ra, -8(gp)
+; RV64-NEXT: addi gp, gp, -8
+; RV64-NEXT: .cfi_restore gp
+; RV64-NEXT: ret
+;
+; RV64-SAVE-RESTORE-LABEL: caller_sw_float:
+; RV64-SAVE-RESTORE: # %bb.0:
+; RV64-SAVE-RESTORE-NEXT: addi gp, gp, 8
+; RV64-SAVE-RESTORE-NEXT: sd ra, -8(gp)
+; RV64-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x78 #
+; RV64-SAVE-RESTORE-NEXT: call t0, __riscv_save_2
+; RV64-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 32
+; RV64-SAVE-RESTORE-NEXT: .cfi_offset ra, -8
+; RV64-SAVE-RESTORE-NEXT: .cfi_offset s0, -16
+; RV64-SAVE-RESTORE-NEXT: .cfi_offset s1, -24
+; RV64-SAVE-RESTORE-NEXT: mv s0, a2
+; RV64-SAVE-RESTORE-NEXT: mv s1, a1
+; RV64-SAVE-RESTORE-NEXT: call callee_float
+; RV64-SAVE-RESTORE-NEXT: mv a0, s1
+; RV64-SAVE-RESTORE-NEXT: mv a1, s0
+; RV64-SAVE-RESTORE-NEXT: call __mulsf3
+; RV64-SAVE-RESTORE-NEXT: ld ra, 24(sp) # 8-byte Folded Reload
+; RV64-SAVE-RESTORE-NEXT: ld s0, 16(sp) # 8-byte Folded Reload
+; RV64-SAVE-RESTORE-NEXT: ld s1, 8(sp) # 8-byte Folded Reload
+; RV64-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV64-SAVE-RESTORE-NEXT: .cfi_restore s0
+; RV64-SAVE-RESTORE-NEXT: .cfi_restore s1
+; RV64-SAVE-RESTORE-NEXT: addi sp, sp, 32
+; RV64-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV64-SAVE-RESTORE-NEXT: ld ra, -8(gp)
+; RV64-SAVE-RESTORE-NEXT: addi gp, gp, -8
+; RV64-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV64-SAVE-RESTORE-NEXT: ret
+;
+; RV64D-SAVE-RESTORE-LABEL: caller_sw_float:
+; RV64D-SAVE-RESTORE: # %bb.0:
+; RV64D-SAVE-RESTORE-NEXT: addi gp, gp, 8
+; RV64D-SAVE-RESTORE-NEXT: sd ra, -8(gp)
+; RV64D-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x78 #
+; RV64D-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset ra, -8
+; RV64D-SAVE-RESTORE-NEXT: addi sp, sp, -16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 32
+; RV64D-SAVE-RESTORE-NEXT: fsd fs0, 8(sp) # 8-byte Folded Spill
+; RV64D-SAVE-RESTORE-NEXT: fsd fs1, 0(sp) # 8-byte Folded Spill
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset fs0, -24
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset fs1, -32
+; RV64D-SAVE-RESTORE-NEXT: fmv.s fs0, fa2
+; RV64D-SAVE-RESTORE-NEXT: fmv.s fs1, fa1
+; RV64D-SAVE-RESTORE-NEXT: call callee_float
+; RV64D-SAVE-RESTORE-NEXT: fmul.s fa0, fs1, fs0
+; RV64D-SAVE-RESTORE-NEXT: fld fs0, 8(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: fld fs1, 0(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: ld ra, 24(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore fs0
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore fs1
+; RV64D-SAVE-RESTORE-NEXT: addi sp, sp, 32
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV64D-SAVE-RESTORE-NEXT: ld ra, -8(gp)
+; RV64D-SAVE-RESTORE-NEXT: addi gp, gp, -8
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV64D-SAVE-RESTORE-NEXT: ret
+;
+; RV64-ZCMP-LABEL: caller_sw_float:
+; RV64-ZCMP: # %bb.0:
+; RV64-ZCMP-NEXT: addi gp, gp, 8
+; RV64-ZCMP-NEXT: sd ra, -8(gp)
+; RV64-ZCMP-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x78 #
+; RV64-ZCMP-NEXT: cm.push {ra, s0-s1}, -32
+; RV64-ZCMP-NEXT: .cfi_def_cfa_offset 32
+; RV64-ZCMP-NEXT: .cfi_offset ra, -24
+; RV64-ZCMP-NEXT: .cfi_offset s0, -16
+; RV64-ZCMP-NEXT: .cfi_offset s1, -8
+; RV64-ZCMP-NEXT: mv s0, a2
+; RV64-ZCMP-NEXT: mv s1, a1
+; RV64-ZCMP-NEXT: call callee_float
+; RV64-ZCMP-NEXT: cm.mva01s s1, s0
+; RV64-ZCMP-NEXT: call __mulsf3
+; RV64-ZCMP-NEXT: cm.pop {ra, s0-s1}, 32
+; RV64-ZCMP-NEXT: ld ra, -8(gp)
+; RV64-ZCMP-NEXT: addi gp, gp, -8
+; RV64-ZCMP-NEXT: .cfi_restore gp
+; RV64-ZCMP-NEXT: ret
+;
+; RV64-ZCMP-SAVE-RESTORE-LABEL: caller_sw_float:
+; RV64-ZCMP-SAVE-RESTORE: # %bb.0:
+; RV64-ZCMP-SAVE-RESTORE-NEXT: addi gp, gp, 8
+; RV64-ZCMP-SAVE-RESTORE-NEXT: sd ra, -8(gp)
+; RV64-ZCMP-SAVE-RESTORE-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x78 #
+; RV64-ZCMP-SAVE-RESTORE-NEXT: cm.push {ra, s0-s1}, -32
+; RV64-ZCMP-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 32
+; RV64-ZCMP-SAVE-RESTORE-NEXT: .cfi_offset ra, -24
+; RV64-ZCMP-SAVE-RESTORE-NEXT: .cfi_offset s0, -16
+; RV64-ZCMP-SAVE-RESTORE-NEXT: .cfi_offset s1, -8
+; RV64-ZCMP-SAVE-RESTORE-NEXT: mv s0, a2
+; RV64-ZCMP-SAVE-RESTORE-NEXT: mv s1, a1
+; RV64-ZCMP-SAVE-RESTORE-NEXT: call callee_float
+; RV64-ZCMP-SAVE-RESTORE-NEXT: cm.mva01s s1, s0
+; RV64-ZCMP-SAVE-RESTORE-NEXT: call __mulsf3
+; RV64-ZCMP-SAVE-RESTORE-NEXT: cm.pop {ra, s0-s1}, 32
+; RV64-ZCMP-SAVE-RESTORE-NEXT: ld ra, -8(gp)
+; RV64-ZCMP-SAVE-RESTORE-NEXT: addi gp, gp, -8
+; RV64-ZCMP-SAVE-RESTORE-NEXT: .cfi_restore gp
+; RV64-ZCMP-SAVE-RESTORE-NEXT: ret
+;
+; RV64-XQCCMP-LABEL: caller_sw_float:
+; RV64-XQCCMP: # %bb.0:
+; RV64-XQCCMP-NEXT: addi gp, gp, 8
+; RV64-XQCCMP-NEXT: sd ra, -8(gp)
+; RV64-XQCCMP-NEXT: .cfi_escape 0x16, 0x03, 0x02, 0x73, 0x78 #
+; RV64-XQCCMP-NEXT: qc.cm.push {ra, s0-s1}, -32
+; RV64-XQCCMP-NEXT: .cfi_def_cfa_offset 32
+; RV64-XQCCMP-NEXT: .cfi_offset ra, -8
+; RV64-XQCCMP-NEXT: .cfi_offset s0, -16
+; RV64-XQCCMP-NEXT: .cfi_offset s1, -24
+; RV64-XQCCMP-NEXT: mv s0, a2
+; RV64-XQCCMP-NEXT: mv s1, a1
+; RV64-XQCCMP-NEXT: call callee_float
+; RV64-XQCCMP-NEXT: qc.cm.mva01s s1, s0
+; RV64-XQCCMP-NEXT: call __mulsf3
+; RV64-XQCCMP-NEXT: qc.cm.pop {ra, s0-s1}, 32
+; RV64-XQCCMP-NEXT: ld ra, -8(gp)
+; RV64-XQCCMP-NEXT: addi gp, gp, -8
+; RV64-XQCCMP-NEXT: .cfi_restore gp
+; RV64-XQCCMP-NEXT: ret
+ %a = call float @callee_float(float %x)
+ %b = fmul float %y, %z
+ ret float %b
+}
+
+
;; Hardware Shadow Call Stack
define i32 @caller_hw(i32 %x) "hw-shadow-stack" {
@@ -1063,6 +1589,22 @@ define i32 @caller_hw(i32 %x) "hw-shadow-stack" {
; RV32-SAVE-RESTORE-NEXT: sspopchk ra
; RV32-SAVE-RESTORE-NEXT: ret
;
+; RV32D-SAVE-RESTORE-LABEL: caller_hw:
+; RV32D-SAVE-RESTORE: # %bb.0:
+; RV32D-SAVE-RESTORE-NEXT: sspush ra
+; RV32D-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset ra, -4
+; RV32D-SAVE-RESTORE-NEXT: neg a0, a0
+; RV32D-SAVE-RESTORE-NEXT: call callee
+; RV32D-SAVE-RESTORE-NEXT: addi a0, a0, 1
+; RV32D-SAVE-RESTORE-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV32D-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV32D-SAVE-RESTORE-NEXT: sspopchk ra
+; RV32D-SAVE-RESTORE-NEXT: ret
+;
; RV32-ZCMP-LABEL: caller_hw:
; RV32-ZCMP: # %bb.0:
; RV32-ZCMP-NEXT: sspush ra
@@ -1135,6 +1677,22 @@ define i32 @caller_hw(i32 %x) "hw-shadow-stack" {
; RV64-SAVE-RESTORE-NEXT: sspopchk ra
; RV64-SAVE-RESTORE-NEXT: ret
;
+; RV64D-SAVE-RESTORE-LABEL: caller_hw:
+; RV64D-SAVE-RESTORE: # %bb.0:
+; RV64D-SAVE-RESTORE-NEXT: sspush ra
+; RV64D-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset ra, -8
+; RV64D-SAVE-RESTORE-NEXT: negw a0, a0
+; RV64D-SAVE-RESTORE-NEXT: call callee
+; RV64D-SAVE-RESTORE-NEXT: addiw a0, a0, 1
+; RV64D-SAVE-RESTORE-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV64D-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV64D-SAVE-RESTORE-NEXT: sspopchk ra
+; RV64D-SAVE-RESTORE-NEXT: ret
+;
; RV64-ZCMP-LABEL: caller_hw:
; RV64-ZCMP: # %bb.0:
; RV64-ZCMP-NEXT: sspush ra
@@ -1190,6 +1748,11 @@ define i32 @caller_hw_tail(i32 %x) "hw-shadow-stack" {
; RV32-SAVE-RESTORE-NEXT: neg a0, a0
; RV32-SAVE-RESTORE-NEXT: tail callee
;
+; RV32D-SAVE-RESTORE-LABEL: caller_hw_tail:
+; RV32D-SAVE-RESTORE: # %bb.0:
+; RV32D-SAVE-RESTORE-NEXT: neg a0, a0
+; RV32D-SAVE-RESTORE-NEXT: tail callee
+;
; RV32-ZCMP-LABEL: caller_hw_tail:
; RV32-ZCMP: # %bb.0:
; RV32-ZCMP-NEXT: neg a0, a0
@@ -1215,6 +1778,11 @@ define i32 @caller_hw_tail(i32 %x) "hw-shadow-stack" {
; RV64-SAVE-RESTORE-NEXT: negw a0, a0
; RV64-SAVE-RESTORE-NEXT: tail callee
;
+; RV64D-SAVE-RESTORE-LABEL: caller_hw_tail:
+; RV64D-SAVE-RESTORE: # %bb.0:
+; RV64D-SAVE-RESTORE-NEXT: negw a0, a0
+; RV64D-SAVE-RESTORE-NEXT: tail callee
+;
; RV64-ZCMP-LABEL: caller_hw_tail:
; RV64-ZCMP: # %bb.0:
; RV64-ZCMP-NEXT: negw a0, a0
@@ -1281,6 +1849,28 @@ define i32 @caller_hw_fp(i32 %x) "hw-shadow-stack" "frame-pointer"="all" {
; RV32-SAVE-RESTORE-NEXT: sspopchk ra
; RV32-SAVE-RESTORE-NEXT: ret
;
+; RV32D-SAVE-RESTORE-LABEL: caller_hw_fp:
+; RV32D-SAVE-RESTORE: # %bb.0:
+; RV32D-SAVE-RESTORE-NEXT: sspush ra
+; RV32D-SAVE-RESTORE-NEXT: call t0, __riscv_save_1
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset ra, -4
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset s0, -8
+; RV32D-SAVE-RESTORE-NEXT: addi s0, sp, 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa s0, 0
+; RV32D-SAVE-RESTORE-NEXT: neg a0, a0
+; RV32D-SAVE-RESTORE-NEXT: call callee
+; RV32D-SAVE-RESTORE-NEXT: addi a0, a0, 1
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa sp, 16
+; RV32D-SAVE-RESTORE-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore s0
+; RV32D-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV32D-SAVE-RESTORE-NEXT: sspopchk ra
+; RV32D-SAVE-RESTORE-NEXT: ret
+;
; RV32-ZCMP-LABEL: caller_hw_fp:
; RV32-ZCMP: # %bb.0:
; RV32-ZCMP-NEXT: sspush ra
@@ -1389,6 +1979,28 @@ define i32 @caller_hw_fp(i32 %x) "hw-shadow-stack" "frame-pointer"="all" {
; RV64-SAVE-RESTORE-NEXT: sspopchk ra
; RV64-SAVE-RESTORE-NEXT: ret
;
+; RV64D-SAVE-RESTORE-LABEL: caller_hw_fp:
+; RV64D-SAVE-RESTORE: # %bb.0:
+; RV64D-SAVE-RESTORE-NEXT: sspush ra
+; RV64D-SAVE-RESTORE-NEXT: call t0, __riscv_save_1
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset ra, -8
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset s0, -16
+; RV64D-SAVE-RESTORE-NEXT: addi s0, sp, 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa s0, 0
+; RV64D-SAVE-RESTORE-NEXT: negw a0, a0
+; RV64D-SAVE-RESTORE-NEXT: call callee
+; RV64D-SAVE-RESTORE-NEXT: addiw a0, a0, 1
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa sp, 16
+; RV64D-SAVE-RESTORE-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: ld s0, 0(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore s0
+; RV64D-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV64D-SAVE-RESTORE-NEXT: sspopchk ra
+; RV64D-SAVE-RESTORE-NEXT: ret
+;
; RV64-ZCMP-LABEL: caller_hw_fp:
; RV64-ZCMP: # %bb.0:
; RV64-ZCMP-NEXT: sspush ra
@@ -1460,7 +2072,7 @@ define i32 @caller_hw_fp(i32 %x) "hw-shadow-stack" "frame-pointer"="all" {
define i32 @caller_hw_cond(i32 %x) "hw-shadow-stack" {
; RV32-LABEL: caller_hw_cond:
; RV32: # %bb.0:
-; RV32-NEXT: beqz a0, .LBB8_2
+; RV32-NEXT: beqz a0, .LBB9_2
; RV32-NEXT: # %bb.1: # %if.end
; RV32-NEXT: sspush ra
; RV32-NEXT: addi sp, sp, -16
@@ -1476,13 +2088,13 @@ define i32 @caller_hw_cond(i32 %x) "hw-shadow-stack" {
; RV32-NEXT: .cfi_def_cfa_offset 0
; RV32-NEXT: sspopchk ra
; RV32-NEXT: ret
-; RV32-NEXT: .LBB8_2: # %return
+; RV32-NEXT: .LBB9_2: # %return
; RV32-NEXT: li a0, 1
; RV32-NEXT: ret
;
; RV32-SAVE-RESTORE-LABEL: caller_hw_cond:
; RV32-SAVE-RESTORE: # %bb.0:
-; RV32-SAVE-RESTORE-NEXT: beqz a0, .LBB8_2
+; RV32-SAVE-RESTORE-NEXT: beqz a0, .LBB9_2
; RV32-SAVE-RESTORE-NEXT: # %bb.1: # %if.end
; RV32-SAVE-RESTORE-NEXT: sspush ra
; RV32-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
@@ -1497,13 +2109,34 @@ define i32 @caller_hw_cond(i32 %x) "hw-shadow-stack" {
; RV32-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
; RV32-SAVE-RESTORE-NEXT: sspopchk ra
; RV32-SAVE-RESTORE-NEXT: ret
-; RV32-SAVE-RESTORE-NEXT: .LBB8_2: # %return
+; RV32-SAVE-RESTORE-NEXT: .LBB9_2: # %return
; RV32-SAVE-RESTORE-NEXT: li a0, 1
; RV32-SAVE-RESTORE-NEXT: ret
;
+; RV32D-SAVE-RESTORE-LABEL: caller_hw_cond:
+; RV32D-SAVE-RESTORE: # %bb.0:
+; RV32D-SAVE-RESTORE-NEXT: beqz a0, .LBB9_2
+; RV32D-SAVE-RESTORE-NEXT: # %bb.1: # %if.end
+; RV32D-SAVE-RESTORE-NEXT: sspush ra
+; RV32D-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset ra, -4
+; RV32D-SAVE-RESTORE-NEXT: neg a0, a0
+; RV32D-SAVE-RESTORE-NEXT: call callee
+; RV32D-SAVE-RESTORE-NEXT: addi a0, a0, 1
+; RV32D-SAVE-RESTORE-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV32D-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV32D-SAVE-RESTORE-NEXT: sspopchk ra
+; RV32D-SAVE-RESTORE-NEXT: ret
+; RV32D-SAVE-RESTORE-NEXT: .LBB9_2: # %return
+; RV32D-SAVE-RESTORE-NEXT: li a0, 1
+; RV32D-SAVE-RESTORE-NEXT: ret
+;
; RV32-ZCMP-LABEL: caller_hw_cond:
; RV32-ZCMP: # %bb.0:
-; RV32-ZCMP-NEXT: beqz a0, .LBB8_2
+; RV32-ZCMP-NEXT: beqz a0, .LBB9_2
; RV32-ZCMP-NEXT: # %bb.1: # %if.end
; RV32-ZCMP-NEXT: sspush ra
; RV32-ZCMP-NEXT: cm.push {ra}, -16
@@ -1515,13 +2148,13 @@ define i32 @caller_hw_cond(i32 %x) "hw-shadow-stack" {
; RV32-ZCMP-NEXT: cm.pop {ra}, 16
; RV32-ZCMP-NEXT: sspopchk ra
; RV32-ZCMP-NEXT: ret
-; RV32-ZCMP-NEXT: .LBB8_2: # %return
+; RV32-ZCMP-NEXT: .LBB9_2: # %return
; RV32-ZCMP-NEXT: li a0, 1
; RV32-ZCMP-NEXT: ret
;
; RV32-ZCMP-SAVE-RESTORE-LABEL: caller_hw_cond:
; RV32-ZCMP-SAVE-RESTORE: # %bb.0:
-; RV32-ZCMP-SAVE-RESTORE-NEXT: beqz a0, .LBB8_2
+; RV32-ZCMP-SAVE-RESTORE-NEXT: beqz a0, .LBB9_2
; RV32-ZCMP-SAVE-RESTORE-NEXT: # %bb.1: # %if.end
; RV32-ZCMP-SAVE-RESTORE-NEXT: sspush ra
; RV32-ZCMP-SAVE-RESTORE-NEXT: cm.push {ra}, -16
@@ -1533,13 +2166,13 @@ define i32 @caller_hw_cond(i32 %x) "hw-shadow-stack" {
; RV32-ZCMP-SAVE-RESTORE-NEXT: cm.pop {ra}, 16
; RV32-ZCMP-SAVE-RESTORE-NEXT: sspopchk ra
; RV32-ZCMP-SAVE-RESTORE-NEXT: ret
-; RV32-ZCMP-SAVE-RESTORE-NEXT: .LBB8_2: # %return
+; RV32-ZCMP-SAVE-RESTORE-NEXT: .LBB9_2: # %return
; RV32-ZCMP-SAVE-RESTORE-NEXT: li a0, 1
; RV32-ZCMP-SAVE-RESTORE-NEXT: ret
;
; RV32-XQCCMP-LABEL: caller_hw_cond:
; RV32-XQCCMP: # %bb.0:
-; RV32-XQCCMP-NEXT: beqz a0, .LBB8_2
+; RV32-XQCCMP-NEXT: beqz a0, .LBB9_2
; RV32-XQCCMP-NEXT: # %bb.1: # %if.end
; RV32-XQCCMP-NEXT: sspush ra
; RV32-XQCCMP-NEXT: qc.cm.push {ra}, -16
@@ -1551,14 +2184,14 @@ define i32 @caller_hw_cond(i32 %x) "hw-shadow-stack" {
; RV32-XQCCMP-NEXT: qc.cm.pop {ra}, 16
; RV32-XQCCMP-NEXT: sspopchk ra
; RV32-XQCCMP-NEXT: ret
-; RV32-XQCCMP-NEXT: .LBB8_2: # %return
+; RV32-XQCCMP-NEXT: .LBB9_2: # %return
; RV32-XQCCMP-NEXT: li a0, 1
; RV32-XQCCMP-NEXT: ret
;
; RV64-LABEL: caller_hw_cond:
; RV64: # %bb.0:
; RV64-NEXT: sext.w a1, a0
-; RV64-NEXT: beqz a1, .LBB8_2
+; RV64-NEXT: beqz a1, .LBB9_2
; RV64-NEXT: # %bb.1: # %if.end
; RV64-NEXT: sspush ra
; RV64-NEXT: addi sp, sp, -16
@@ -1574,14 +2207,14 @@ define i32 @caller_hw_cond(i32 %x) "hw-shadow-stack" {
; RV64-NEXT: .cfi_def_cfa_offset 0
; RV64-NEXT: sspopchk ra
; RV64-NEXT: ret
-; RV64-NEXT: .LBB8_2: # %return
+; RV64-NEXT: .LBB9_2: # %return
; RV64-NEXT: li a0, 1
; RV64-NEXT: ret
;
; RV64-SAVE-RESTORE-LABEL: caller_hw_cond:
; RV64-SAVE-RESTORE: # %bb.0:
; RV64-SAVE-RESTORE-NEXT: sext.w a1, a0
-; RV64-SAVE-RESTORE-NEXT: beqz a1, .LBB8_2
+; RV64-SAVE-RESTORE-NEXT: beqz a1, .LBB9_2
; RV64-SAVE-RESTORE-NEXT: # %bb.1: # %if.end
; RV64-SAVE-RESTORE-NEXT: sspush ra
; RV64-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
@@ -1596,14 +2229,36 @@ define i32 @caller_hw_cond(i32 %x) "hw-shadow-stack" {
; RV64-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
; RV64-SAVE-RESTORE-NEXT: sspopchk ra
; RV64-SAVE-RESTORE-NEXT: ret
-; RV64-SAVE-RESTORE-NEXT: .LBB8_2: # %return
+; RV64-SAVE-RESTORE-NEXT: .LBB9_2: # %return
; RV64-SAVE-RESTORE-NEXT: li a0, 1
; RV64-SAVE-RESTORE-NEXT: ret
;
+; RV64D-SAVE-RESTORE-LABEL: caller_hw_cond:
+; RV64D-SAVE-RESTORE: # %bb.0:
+; RV64D-SAVE-RESTORE-NEXT: sext.w a1, a0
+; RV64D-SAVE-RESTORE-NEXT: beqz a1, .LBB9_2
+; RV64D-SAVE-RESTORE-NEXT: # %bb.1: # %if.end
+; RV64D-SAVE-RESTORE-NEXT: sspush ra
+; RV64D-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset ra, -8
+; RV64D-SAVE-RESTORE-NEXT: negw a0, a0
+; RV64D-SAVE-RESTORE-NEXT: call callee
+; RV64D-SAVE-RESTORE-NEXT: addiw a0, a0, 1
+; RV64D-SAVE-RESTORE-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV64D-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV64D-SAVE-RESTORE-NEXT: sspopchk ra
+; RV64D-SAVE-RESTORE-NEXT: ret
+; RV64D-SAVE-RESTORE-NEXT: .LBB9_2: # %return
+; RV64D-SAVE-RESTORE-NEXT: li a0, 1
+; RV64D-SAVE-RESTORE-NEXT: ret
+;
; RV64-ZCMP-LABEL: caller_hw_cond:
; RV64-ZCMP: # %bb.0:
; RV64-ZCMP-NEXT: sext.w a1, a0
-; RV64-ZCMP-NEXT: beqz a1, .LBB8_2
+; RV64-ZCMP-NEXT: beqz a1, .LBB9_2
; RV64-ZCMP-NEXT: # %bb.1: # %if.end
; RV64-ZCMP-NEXT: sspush ra
; RV64-ZCMP-NEXT: cm.push {ra}, -16
@@ -1615,14 +2270,14 @@ define i32 @caller_hw_cond(i32 %x) "hw-shadow-stack" {
; RV64-ZCMP-NEXT: cm.pop {ra}, 16
; RV64-ZCMP-NEXT: sspopchk ra
; RV64-ZCMP-NEXT: ret
-; RV64-ZCMP-NEXT: .LBB8_2: # %return
+; RV64-ZCMP-NEXT: .LBB9_2: # %return
; RV64-ZCMP-NEXT: li a0, 1
; RV64-ZCMP-NEXT: ret
;
; RV64-ZCMP-SAVE-RESTORE-LABEL: caller_hw_cond:
; RV64-ZCMP-SAVE-RESTORE: # %bb.0:
; RV64-ZCMP-SAVE-RESTORE-NEXT: sext.w a1, a0
-; RV64-ZCMP-SAVE-RESTORE-NEXT: beqz a1, .LBB8_2
+; RV64-ZCMP-SAVE-RESTORE-NEXT: beqz a1, .LBB9_2
; RV64-ZCMP-SAVE-RESTORE-NEXT: # %bb.1: # %if.end
; RV64-ZCMP-SAVE-RESTORE-NEXT: sspush ra
; RV64-ZCMP-SAVE-RESTORE-NEXT: cm.push {ra}, -16
@@ -1634,14 +2289,14 @@ define i32 @caller_hw_cond(i32 %x) "hw-shadow-stack" {
; RV64-ZCMP-SAVE-RESTORE-NEXT: cm.pop {ra}, 16
; RV64-ZCMP-SAVE-RESTORE-NEXT: sspopchk ra
; RV64-ZCMP-SAVE-RESTORE-NEXT: ret
-; RV64-ZCMP-SAVE-RESTORE-NEXT: .LBB8_2: # %return
+; RV64-ZCMP-SAVE-RESTORE-NEXT: .LBB9_2: # %return
; RV64-ZCMP-SAVE-RESTORE-NEXT: li a0, 1
; RV64-ZCMP-SAVE-RESTORE-NEXT: ret
;
; RV64-XQCCMP-LABEL: caller_hw_cond:
; RV64-XQCCMP: # %bb.0:
; RV64-XQCCMP-NEXT: sext.w a1, a0
-; RV64-XQCCMP-NEXT: beqz a1, .LBB8_2
+; RV64-XQCCMP-NEXT: beqz a1, .LBB9_2
; RV64-XQCCMP-NEXT: # %bb.1: # %if.end
; RV64-XQCCMP-NEXT: sspush ra
; RV64-XQCCMP-NEXT: qc.cm.push {ra}, -16
@@ -1653,7 +2308,7 @@ define i32 @caller_hw_cond(i32 %x) "hw-shadow-stack" {
; RV64-XQCCMP-NEXT: qc.cm.pop {ra}, 16
; RV64-XQCCMP-NEXT: sspopchk ra
; RV64-XQCCMP-NEXT: ret
-; RV64-XQCCMP-NEXT: .LBB8_2: # %return
+; RV64-XQCCMP-NEXT: .LBB9_2: # %return
; RV64-XQCCMP-NEXT: li a0, 1
; RV64-XQCCMP-NEXT: ret
%cmp = icmp eq i32 %x, 0
@@ -1716,6 +2371,28 @@ define i32 @caller_hw_clobber_s11(i32 %x) "hw-shadow-stack" {
; RV32-SAVE-RESTORE-NEXT: sspopchk ra
; RV32-SAVE-RESTORE-NEXT: ret
;
+; RV32D-SAVE-RESTORE-LABEL: caller_hw_clobber_s11:
+; RV32D-SAVE-RESTORE: # %bb.0:
+; RV32D-SAVE-RESTORE-NEXT: sspush ra
+; RV32D-SAVE-RESTORE-NEXT: call t0, __riscv_save_12
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 64
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset ra, -4
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset s11, -52
+; RV32D-SAVE-RESTORE-NEXT: neg a0, a0
+; RV32D-SAVE-RESTORE-NEXT: call callee
+; RV32D-SAVE-RESTORE-NEXT: addi a0, a0, 1
+; RV32D-SAVE-RESTORE-NEXT: #APP
+; RV32D-SAVE-RESTORE-NEXT: nop
+; RV32D-SAVE-RESTORE-NEXT: #NO_APP
+; RV32D-SAVE-RESTORE-NEXT: lw ra, 60(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: lw s11, 12(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore s11
+; RV32D-SAVE-RESTORE-NEXT: addi sp, sp, 64
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV32D-SAVE-RESTORE-NEXT: sspopchk ra
+; RV32D-SAVE-RESTORE-NEXT: ret
+;
; RV32-ZCMP-LABEL: caller_hw_clobber_s11:
; RV32-ZCMP: # %bb.0:
; RV32-ZCMP-NEXT: sspush ra
@@ -1813,6 +2490,28 @@ define i32 @caller_hw_clobber_s11(i32 %x) "hw-shadow-stack" {
; RV64-SAVE-RESTORE-NEXT: sspopchk ra
; RV64-SAVE-RESTORE-NEXT: ret
;
+; RV64D-SAVE-RESTORE-LABEL: caller_hw_clobber_s11:
+; RV64D-SAVE-RESTORE: # %bb.0:
+; RV64D-SAVE-RESTORE-NEXT: sspush ra
+; RV64D-SAVE-RESTORE-NEXT: call t0, __riscv_save_12
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 112
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset ra, -8
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset s11, -104
+; RV64D-SAVE-RESTORE-NEXT: negw a0, a0
+; RV64D-SAVE-RESTORE-NEXT: call callee
+; RV64D-SAVE-RESTORE-NEXT: addiw a0, a0, 1
+; RV64D-SAVE-RESTORE-NEXT: #APP
+; RV64D-SAVE-RESTORE-NEXT: nop
+; RV64D-SAVE-RESTORE-NEXT: #NO_APP
+; RV64D-SAVE-RESTORE-NEXT: ld ra, 104(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: ld s11, 8(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore s11
+; RV64D-SAVE-RESTORE-NEXT: addi sp, sp, 112
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV64D-SAVE-RESTORE-NEXT: sspopchk ra
+; RV64D-SAVE-RESTORE-NEXT: ret
+;
; RV64-ZCMP-LABEL: caller_hw_clobber_s11:
; RV64-ZCMP: # %bb.0:
; RV64-ZCMP-NEXT: sspush ra
@@ -1869,3 +2568,270 @@ define i32 @caller_hw_clobber_s11(i32 %x) "hw-shadow-stack" {
call void asm sideeffect "nop", "~{x27}"()
ret i32 %add
}
+
+define float @caller_hw_float(float %x, float %y, float %z) "hw-shadow-stack" {
+; RV32-LABEL: caller_hw_float:
+; RV32: # %bb.0:
+; RV32-NEXT: sspush ra
+; RV32-NEXT: addi sp, sp, -16
+; RV32-NEXT: .cfi_def_cfa_offset 16
+; RV32-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
+; RV32-NEXT: sw s0, 8(sp) # 4-byte Folded Spill
+; RV32-NEXT: sw s1, 4(sp) # 4-byte Folded Spill
+; RV32-NEXT: .cfi_offset ra, -4
+; RV32-NEXT: .cfi_offset s0, -8
+; RV32-NEXT: .cfi_offset s1, -12
+; RV32-NEXT: mv s0, a2
+; RV32-NEXT: mv s1, a1
+; RV32-NEXT: call callee_float
+; RV32-NEXT: mv a0, s1
+; RV32-NEXT: mv a1, s0
+; RV32-NEXT: call __mulsf3
+; RV32-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
+; RV32-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
+; RV32-NEXT: lw s1, 4(sp) # 4-byte Folded Reload
+; RV32-NEXT: .cfi_restore ra
+; RV32-NEXT: .cfi_restore s0
+; RV32-NEXT: .cfi_restore s1
+; RV32-NEXT: addi sp, sp, 16
+; RV32-NEXT: .cfi_def_cfa_offset 0
+; RV32-NEXT: sspopchk ra
+; RV32-NEXT: ret
+;
+; RV32-SAVE-RESTORE-LABEL: caller_hw_float:
+; RV32-SAVE-RESTORE: # %bb.0:
+; RV32-SAVE-RESTORE-NEXT: sspush ra
+; RV32-SAVE-RESTORE-NEXT: call t0, __riscv_save_2
+; RV32-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV32-SAVE-RESTORE-NEXT: .cfi_offset ra, -4
+; RV32-SAVE-RESTORE-NEXT: .cfi_offset s0, -8
+; RV32-SAVE-RESTORE-NEXT: .cfi_offset s1, -12
+; RV32-SAVE-RESTORE-NEXT: mv s0, a2
+; RV32-SAVE-RESTORE-NEXT: mv s1, a1
+; RV32-SAVE-RESTORE-NEXT: call callee_float
+; RV32-SAVE-RESTORE-NEXT: mv a0, s1
+; RV32-SAVE-RESTORE-NEXT: mv a1, s0
+; RV32-SAVE-RESTORE-NEXT: call __mulsf3
+; RV32-SAVE-RESTORE-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
+; RV32-SAVE-RESTORE-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
+; RV32-SAVE-RESTORE-NEXT: lw s1, 4(sp) # 4-byte Folded Reload
+; RV32-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV32-SAVE-RESTORE-NEXT: .cfi_restore s0
+; RV32-SAVE-RESTORE-NEXT: .cfi_restore s1
+; RV32-SAVE-RESTORE-NEXT: addi sp, sp, 16
+; RV32-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV32-SAVE-RESTORE-NEXT: sspopchk ra
+; RV32-SAVE-RESTORE-NEXT: ret
+;
+; RV32D-SAVE-RESTORE-LABEL: caller_hw_float:
+; RV32D-SAVE-RESTORE: # %bb.0:
+; RV32D-SAVE-RESTORE-NEXT: sspush ra
+; RV32D-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset ra, -4
+; RV32D-SAVE-RESTORE-NEXT: addi sp, sp, -16
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 32
+; RV32D-SAVE-RESTORE-NEXT: fsd fs0, 8(sp) # 8-byte Folded Spill
+; RV32D-SAVE-RESTORE-NEXT: fsd fs1, 0(sp) # 8-byte Folded Spill
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset fs0, -24
+; RV32D-SAVE-RESTORE-NEXT: .cfi_offset fs1, -32
+; RV32D-SAVE-RESTORE-NEXT: fmv.s fs0, fa2
+; RV32D-SAVE-RESTORE-NEXT: fmv.s fs1, fa1
+; RV32D-SAVE-RESTORE-NEXT: call callee_float
+; RV32D-SAVE-RESTORE-NEXT: fmul.s fa0, fs1, fs0
+; RV32D-SAVE-RESTORE-NEXT: fld fs0, 8(sp) # 8-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: fld fs1, 0(sp) # 8-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: lw ra, 28(sp) # 4-byte Folded Reload
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore fs0
+; RV32D-SAVE-RESTORE-NEXT: .cfi_restore fs1
+; RV32D-SAVE-RESTORE-NEXT: addi sp, sp, 32
+; RV32D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV32D-SAVE-RESTORE-NEXT: sspopchk ra
+; RV32D-SAVE-RESTORE-NEXT: ret
+;
+; RV32-ZCMP-LABEL: caller_hw_float:
+; RV32-ZCMP: # %bb.0:
+; RV32-ZCMP-NEXT: sspush ra
+; RV32-ZCMP-NEXT: cm.push {ra, s0-s1}, -16
+; RV32-ZCMP-NEXT: .cfi_def_cfa_offset 16
+; RV32-ZCMP-NEXT: .cfi_offset ra, -12
+; RV32-ZCMP-NEXT: .cfi_offset s0, -8
+; RV32-ZCMP-NEXT: .cfi_offset s1, -4
+; RV32-ZCMP-NEXT: mv s0, a2
+; RV32-ZCMP-NEXT: mv s1, a1
+; RV32-ZCMP-NEXT: call callee_float
+; RV32-ZCMP-NEXT: cm.mva01s s1, s0
+; RV32-ZCMP-NEXT: call __mulsf3
+; RV32-ZCMP-NEXT: cm.pop {ra, s0-s1}, 16
+; RV32-ZCMP-NEXT: sspopchk ra
+; RV32-ZCMP-NEXT: ret
+;
+; RV32-ZCMP-SAVE-RESTORE-LABEL: caller_hw_float:
+; RV32-ZCMP-SAVE-RESTORE: # %bb.0:
+; RV32-ZCMP-SAVE-RESTORE-NEXT: sspush ra
+; RV32-ZCMP-SAVE-RESTORE-NEXT: cm.push {ra, s0-s1}, -16
+; RV32-ZCMP-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV32-ZCMP-SAVE-RESTORE-NEXT: .cfi_offset ra, -12
+; RV32-ZCMP-SAVE-RESTORE-NEXT: .cfi_offset s0, -8
+; RV32-ZCMP-SAVE-RESTORE-NEXT: .cfi_offset s1, -4
+; RV32-ZCMP-SAVE-RESTORE-NEXT: mv s0, a2
+; RV32-ZCMP-SAVE-RESTORE-NEXT: mv s1, a1
+; RV32-ZCMP-SAVE-RESTORE-NEXT: call callee_float
+; RV32-ZCMP-SAVE-RESTORE-NEXT: cm.mva01s s1, s0
+; RV32-ZCMP-SAVE-RESTORE-NEXT: call __mulsf3
+; RV32-ZCMP-SAVE-RESTORE-NEXT: cm.pop {ra, s0-s1}, 16
+; RV32-ZCMP-SAVE-RESTORE-NEXT: sspopchk ra
+; RV32-ZCMP-SAVE-RESTORE-NEXT: ret
+;
+; RV32-XQCCMP-LABEL: caller_hw_float:
+; RV32-XQCCMP: # %bb.0:
+; RV32-XQCCMP-NEXT: sspush ra
+; RV32-XQCCMP-NEXT: qc.cm.push {ra, s0-s1}, -16
+; RV32-XQCCMP-NEXT: .cfi_def_cfa_offset 16
+; RV32-XQCCMP-NEXT: .cfi_offset ra, -4
+; RV32-XQCCMP-NEXT: .cfi_offset s0, -8
+; RV32-XQCCMP-NEXT: .cfi_offset s1, -12
+; RV32-XQCCMP-NEXT: mv s0, a2
+; RV32-XQCCMP-NEXT: mv s1, a1
+; RV32-XQCCMP-NEXT: call callee_float
+; RV32-XQCCMP-NEXT: qc.cm.mva01s s1, s0
+; RV32-XQCCMP-NEXT: call __mulsf3
+; RV32-XQCCMP-NEXT: qc.cm.pop {ra, s0-s1}, 16
+; RV32-XQCCMP-NEXT: sspopchk ra
+; RV32-XQCCMP-NEXT: ret
+;
+; RV64-LABEL: caller_hw_float:
+; RV64: # %bb.0:
+; RV64-NEXT: sspush ra
+; RV64-NEXT: addi sp, sp, -32
+; RV64-NEXT: .cfi_def_cfa_offset 32
+; RV64-NEXT: sd ra, 24(sp) # 8-byte Folded Spill
+; RV64-NEXT: sd s0, 16(sp) # 8-byte Folded Spill
+; RV64-NEXT: sd s1, 8(sp) # 8-byte Folded Spill
+; RV64-NEXT: .cfi_offset ra, -8
+; RV64-NEXT: .cfi_offset s0, -16
+; RV64-NEXT: .cfi_offset s1, -24
+; RV64-NEXT: mv s0, a2
+; RV64-NEXT: mv s1, a1
+; RV64-NEXT: call callee_float
+; RV64-NEXT: mv a0, s1
+; RV64-NEXT: mv a1, s0
+; RV64-NEXT: call __mulsf3
+; RV64-NEXT: ld ra, 24(sp) # 8-byte Folded Reload
+; RV64-NEXT: ld s0, 16(sp) # 8-byte Folded Reload
+; RV64-NEXT: ld s1, 8(sp) # 8-byte Folded Reload
+; RV64-NEXT: .cfi_restore ra
+; RV64-NEXT: .cfi_restore s0
+; RV64-NEXT: .cfi_restore s1
+; RV64-NEXT: addi sp, sp, 32
+; RV64-NEXT: .cfi_def_cfa_offset 0
+; RV64-NEXT: sspopchk ra
+; RV64-NEXT: ret
+;
+; RV64-SAVE-RESTORE-LABEL: caller_hw_float:
+; RV64-SAVE-RESTORE: # %bb.0:
+; RV64-SAVE-RESTORE-NEXT: sspush ra
+; RV64-SAVE-RESTORE-NEXT: call t0, __riscv_save_2
+; RV64-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 32
+; RV64-SAVE-RESTORE-NEXT: .cfi_offset ra, -8
+; RV64-SAVE-RESTORE-NEXT: .cfi_offset s0, -16
+; RV64-SAVE-RESTORE-NEXT: .cfi_offset s1, -24
+; RV64-SAVE-RESTORE-NEXT: mv s0, a2
+; RV64-SAVE-RESTORE-NEXT: mv s1, a1
+; RV64-SAVE-RESTORE-NEXT: call callee_float
+; RV64-SAVE-RESTORE-NEXT: mv a0, s1
+; RV64-SAVE-RESTORE-NEXT: mv a1, s0
+; RV64-SAVE-RESTORE-NEXT: call __mulsf3
+; RV64-SAVE-RESTORE-NEXT: ld ra, 24(sp) # 8-byte Folded Reload
+; RV64-SAVE-RESTORE-NEXT: ld s0, 16(sp) # 8-byte Folded Reload
+; RV64-SAVE-RESTORE-NEXT: ld s1, 8(sp) # 8-byte Folded Reload
+; RV64-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV64-SAVE-RESTORE-NEXT: .cfi_restore s0
+; RV64-SAVE-RESTORE-NEXT: .cfi_restore s1
+; RV64-SAVE-RESTORE-NEXT: addi sp, sp, 32
+; RV64-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV64-SAVE-RESTORE-NEXT: sspopchk ra
+; RV64-SAVE-RESTORE-NEXT: ret
+;
+; RV64D-SAVE-RESTORE-LABEL: caller_hw_float:
+; RV64D-SAVE-RESTORE: # %bb.0:
+; RV64D-SAVE-RESTORE-NEXT: sspush ra
+; RV64D-SAVE-RESTORE-NEXT: call t0, __riscv_save_0
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset ra, -8
+; RV64D-SAVE-RESTORE-NEXT: addi sp, sp, -16
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 32
+; RV64D-SAVE-RESTORE-NEXT: fsd fs0, 8(sp) # 8-byte Folded Spill
+; RV64D-SAVE-RESTORE-NEXT: fsd fs1, 0(sp) # 8-byte Folded Spill
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset fs0, -24
+; RV64D-SAVE-RESTORE-NEXT: .cfi_offset fs1, -32
+; RV64D-SAVE-RESTORE-NEXT: fmv.s fs0, fa2
+; RV64D-SAVE-RESTORE-NEXT: fmv.s fs1, fa1
+; RV64D-SAVE-RESTORE-NEXT: call callee_float
+; RV64D-SAVE-RESTORE-NEXT: fmul.s fa0, fs1, fs0
+; RV64D-SAVE-RESTORE-NEXT: fld fs0, 8(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: fld fs1, 0(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: ld ra, 24(sp) # 8-byte Folded Reload
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore ra
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore fs0
+; RV64D-SAVE-RESTORE-NEXT: .cfi_restore fs1
+; RV64D-SAVE-RESTORE-NEXT: addi sp, sp, 32
+; RV64D-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 0
+; RV64D-SAVE-RESTORE-NEXT: sspopchk ra
+; RV64D-SAVE-RESTORE-NEXT: ret
+;
+; RV64-ZCMP-LABEL: caller_hw_float:
+; RV64-ZCMP: # %bb.0:
+; RV64-ZCMP-NEXT: sspush ra
+; RV64-ZCMP-NEXT: cm.push {ra, s0-s1}, -32
+; RV64-ZCMP-NEXT: .cfi_def_cfa_offset 32
+; RV64-ZCMP-NEXT: .cfi_offset ra, -24
+; RV64-ZCMP-NEXT: .cfi_offset s0, -16
+; RV64-ZCMP-NEXT: .cfi_offset s1, -8
+; RV64-ZCMP-NEXT: mv s0, a2
+; RV64-ZCMP-NEXT: mv s1, a1
+; RV64-ZCMP-NEXT: call callee_float
+; RV64-ZCMP-NEXT: cm.mva01s s1, s0
+; RV64-ZCMP-NEXT: call __mulsf3
+; RV64-ZCMP-NEXT: cm.pop {ra, s0-s1}, 32
+; RV64-ZCMP-NEXT: sspopchk ra
+; RV64-ZCMP-NEXT: ret
+;
+; RV64-ZCMP-SAVE-RESTORE-LABEL: caller_hw_float:
+; RV64-ZCMP-SAVE-RESTORE: # %bb.0:
+; RV64-ZCMP-SAVE-RESTORE-NEXT: sspush ra
+; RV64-ZCMP-SAVE-RESTORE-NEXT: cm.push {ra, s0-s1}, -32
+; RV64-ZCMP-SAVE-RESTORE-NEXT: .cfi_def_cfa_offset 32
+; RV64-ZCMP-SAVE-RESTORE-NEXT: .cfi_offset ra, -24
+; RV64-ZCMP-SAVE-RESTORE-NEXT: .cfi_offset s0, -16
+; RV64-ZCMP-SAVE-RESTORE-NEXT: .cfi_offset s1, -8
+; RV64-ZCMP-SAVE-RESTORE-NEXT: mv s0, a2
+; RV64-ZCMP-SAVE-RESTORE-NEXT: mv s1, a1
+; RV64-ZCMP-SAVE-RESTORE-NEXT: call callee_float
+; RV64-ZCMP-SAVE-RESTORE-NEXT: cm.mva01s s1, s0
+; RV64-ZCMP-SAVE-RESTORE-NEXT: call __mulsf3
+; RV64-ZCMP-SAVE-RESTORE-NEXT: cm.pop {ra, s0-s1}, 32
+; RV64-ZCMP-SAVE-RESTORE-NEXT: sspopchk ra
+; RV64-ZCMP-SAVE-RESTORE-NEXT: ret
+;
+; RV64-XQCCMP-LABEL: caller_hw_float:
+; RV64-XQCCMP: # %bb.0:
+; RV64-XQCCMP-NEXT: sspush ra
+; RV64-XQCCMP-NEXT: qc.cm.push {ra, s0-s1}, -32
+; RV64-XQCCMP-NEXT: .cfi_def_cfa_offset 32
+; RV64-XQCCMP-NEXT: .cfi_offset ra, -8
+; RV64-XQCCMP-NEXT: .cfi_offset s0, -16
+; RV64-XQCCMP-NEXT: .cfi_offset s1, -24
+; RV64-XQCCMP-NEXT: mv s0, a2
+; RV64-XQCCMP-NEXT: mv s1, a1
+; RV64-XQCCMP-NEXT: call callee_float
+; RV64-XQCCMP-NEXT: qc.cm.mva01s s1, s0
+; RV64-XQCCMP-NEXT: call __mulsf3
+; RV64-XQCCMP-NEXT: qc.cm.pop {ra, s0-s1}, 32
+; RV64-XQCCMP-NEXT: sspopchk ra
+; RV64-XQCCMP-NEXT: ret
+ %a = call float @callee_float(float %x)
+ %b = fmul float %y, %z
+ ret float %b
+}
More information about the llvm-branch-commits
mailing list