Orlando Cazalet-Hyams 54d544b831
[KeyInstr][Clang] Ret atom (#134652)
This patch is part of a stack that teaches Clang to generate Key Instructions
metadata for C and C++.

When returning a value, stores to the `retval` allocas and branches to `return`
block are put in the same atom group. They are both rank 1, which could in
theory introduce an extra step in some optimized code. This low risk currently
feels an acceptable for keeping the code a bit simpler (as opposed to adding
scaffolding to make the store rank 2).

In the case of a single return (no control flow) the return instruction inherits
the atom group of the branch to the return block when the blocks get folded
togather.

RFC:
https://discourse.llvm.org/t/rfc-improving-is-stmt-placement-for-better-interactive-debugging/82668

The feature is only functional in LLVM if LLVM is built with CMake flag
LLVM_EXPERIMENTAL_KEY_INSTRUCTIONs. Eventually that flag will be removed.
2025-06-04 15:43:49 +01:00

101 lines
5.8 KiB
C

// RUN: %clang_cc1 -triple x86_64-linux-gnu -gkey-instructions -x c++ %s -debug-info-kind=line-tables-only -emit-llvm -o - \
// RUN: | FileCheck %s --implicit-check-not atomGroup --implicit-check-not atomRank --check-prefixes=CHECK,CHECK-CXX
// RUN: %clang_cc1 -triple x86_64-linux-gnu -gkey-instructions -x c %s -debug-info-kind=line-tables-only -emit-llvm -o - \
// RUN: | FileCheck %s --implicit-check-not atomGroup --implicit-check-not atomRank --check-prefixes=CHECK,CHECK-C
_Complex float ci;
float f;
void test() {
// CHECK: %ci.real = load float, ptr @ci{{.*}}, !dbg [[G1R2:!.*]]
// CHECK: %ci.imag = load float, ptr getelementptr inbounds nuw ({ float, float }, ptr @ci, i32 0, i32 1){{.*}}, !dbg [[G1R2]]
// CHECK: store float %ci.real, ptr %lc.realp{{.*}}, !dbg [[G1R1:!.*]]
// CHECK: store float %ci.imag, ptr %lc.imagp{{.*}}, !dbg [[G1R1]]
_Complex float lc = ci;
// CHECK: %ci.real1 = load float, ptr @ci{{.*}}, !dbg [[G2R2:!.*]]
// CHECK: %ci.imag2 = load float, ptr getelementptr inbounds nuw ({ float, float }, ptr @ci, i32 0, i32 1){{.*}}, !dbg [[G2R2]]
// CHECK: store float %ci.real1, ptr @ci{{.*}}, !dbg [[G2R1:!.*]]
// CHECK: store float %ci.imag2, ptr getelementptr inbounds nuw ({ float, float }, ptr @ci, i32 0, i32 1){{.*}}, !dbg [[G2R1]]
ci = ci;
// CHECK: %add.r = fadd float %ci.real5, %ci.real3, !dbg [[G3R2:!.*]]
// CHECK: %add.i = fadd float %ci.imag6, %ci.imag4, !dbg [[G3R2]]
// CHECK: store float %add.r, ptr @ci{{.*}}, !dbg [[G3R1:!.*]]
// CHECK: store float %add.i, ptr getelementptr inbounds nuw ({ float, float }, ptr @ci, i32 0, i32 1){{.*}}, !dbg [[G3R1]]
ci += ci;
// CHECK: %sub.r = fsub float %ci.real9, %ci.real7, !dbg [[G4R2:!.*]]
// CHECK: %sub.i = fsub float %ci.imag10, %ci.imag8, !dbg [[G4R2]]
// CHECK: store float %sub.r, ptr @ci, align 4, !dbg [[G4R1:!.*]]
// CHECK: store float %sub.i, ptr getelementptr inbounds nuw ({ float, float }, ptr @ci, i32 0, i32 1){{.*}}, !dbg [[G4R1]]
ci -= ci;
// There's control flow introduced here to skip around nan and lib calls
// (which is ignored in the test as none of those insts need be key). This does
// make PHIs "backup" instructions, which is... odd. FIXME: Do we want to make
// the instructions producing the values in the PHI backups instead/too?
// CHECK: %real_mul_phi = phi float [ %mul_r, %entry ], [ %mul_r, %complex_mul_imag_nan ], [ %coerce.real, %complex_mul_libcall ], !dbg [[G5R2:!.*]]
// CHECK: %imag_mul_phi = phi float [ %mul_i, %entry ], [ %mul_i, %complex_mul_imag_nan ], [ %coerce.imag, %complex_mul_libcall ], !dbg [[G5R2]]
// CHECK: store float %real_mul_phi, ptr @ci, align 4, !dbg [[G5R1:!.*]]
// CHECK: store float %imag_mul_phi, ptr getelementptr inbounds nuw ({ float, float }, ptr @ci, i32 0, i32 1), align 4, !dbg [[G5R1]]
ci *= ci;
// div goes straight to lib call, which gets saved into a temp.
// CHECK: %coerce21.real = load float, ptr %coerce21.realp, align 4, !dbg [[G6R2:!.*]]
// CHECK: %coerce21.imag = load float, ptr %coerce21.imagp, align 4, !dbg [[G6R2]]
// CHECK: store float %coerce21.real, ptr @ci, align 4, !dbg [[G6R1:!.*]]
// CHECK: store float %coerce21.imag, ptr getelementptr inbounds nuw ({ float, float }, ptr @ci, i32 0, i32 1), align 4, !dbg [[G6R1]]
ci /= ci;
// CHECK: %add = fadd float %0, %1, !dbg [[G7R2:!.*]]
// CHECK: store float %add, ptr getelementptr inbounds nuw ({ float, float }, ptr @ci, i32 0, i32 1){{.*}}, !dbg [[G7R1:!.*]]
__imag ci = __imag ci + __imag ci;
#ifndef __cplusplus
// CHECK-C: %2 = load float, ptr @f, align 4
// CHECK-C: %add.r24 = fadd float %2, %ci.real22, !dbg [[G8R2:!.*]]
// CHECK-C: store float %add.r24, ptr @f, align 4, !dbg [[G8R1:!.*]]
f += ci;
// CHECK-C: %3 = load float, ptr @f, align 4
// CHECK-C: %sub.r27 = fsub float %3, %ci.real25, !dbg [[G9R2:!.*]]
// CHECK-C: store float %sub.r27, ptr @f, align 4, !dbg [[G9R1:!.*]]
f -= ci;
// CHECK-C: %coerce32.real = load float, ptr %coerce32.realp, align 4, !dbg [[G10R2:!.*]]
// CHECK-C: store float %coerce32.real, ptr @f, align 4, !dbg [[G10R1:!.*]]
f /= ci;
// CHECK-C: %mul.rl = fmul float %5, %ci.real33, !dbg [[G11R2:!.*]]
// CHECK-C: store float %mul.rl, ptr @f, align 4, !dbg [[G11R1:!.*]]
f *= ci;
#endif
// CHECK: ret{{.*}}, !dbg [[RET:!.*]]
}
// CHECK: [[G1R2]] = !DILocation({{.*}}, atomGroup: 1, atomRank: 2)
// CHECK: [[G1R1]] = !DILocation({{.*}}, atomGroup: 1, atomRank: 1)
// CHECK: [[G2R2]] = !DILocation({{.*}}, atomGroup: 2, atomRank: 2)
// CHECK: [[G2R1]] = !DILocation({{.*}}, atomGroup: 2, atomRank: 1)
// CHECK: [[G3R2]] = !DILocation({{.*}}, atomGroup: 3, atomRank: 2)
// CHECK: [[G3R1]] = !DILocation({{.*}}, atomGroup: 3, atomRank: 1)
// CHECK: [[G4R2]] = !DILocation({{.*}}, atomGroup: 4, atomRank: 2)
// CHECK: [[G4R1]] = !DILocation({{.*}}, atomGroup: 4, atomRank: 1)
// CHECK: [[G5R2]] = !DILocation({{.*}}, atomGroup: 5, atomRank: 2)
// CHECK: [[G5R1]] = !DILocation({{.*}}, atomGroup: 5, atomRank: 1)
// CHECK: [[G6R2]] = !DILocation({{.*}}, atomGroup: 6, atomRank: 2)
// CHECK: [[G6R1]] = !DILocation({{.*}}, atomGroup: 6, atomRank: 1)
// CHECK: [[G7R2]] = !DILocation({{.*}}, atomGroup: 7, atomRank: 2)
// CHECK: [[G7R1]] = !DILocation({{.*}}, atomGroup: 7, atomRank: 1)
// CHECK-C: [[G8R2]] = !DILocation({{.*}}, atomGroup: 8, atomRank: 2)
// CHECK-C: [[G8R1]] = !DILocation({{.*}}, atomGroup: 8, atomRank: 1)
// CHECK-C: [[G9R2]] = !DILocation({{.*}}, atomGroup: 9, atomRank: 2)
// CHECK-C: [[G9R1]] = !DILocation({{.*}}, atomGroup: 9, atomRank: 1)
// CHECK-C: [[G10R2]] = !DILocation({{.*}}, atomGroup: 10, atomRank: 2)
// CHECK-C: [[G10R1]] = !DILocation({{.*}}, atomGroup: 10, atomRank: 1)
// CHECK-C: [[G11R2]] = !DILocation({{.*}}, atomGroup: 11, atomRank: 2)
// CHECK-C: [[G11R1]] = !DILocation({{.*}}, atomGroup: 11, atomRank: 1)
// CHECK-C: [[RET]] = !DILocation({{.*}}, atomGroup: 12, atomRank: 1)
// CHECK-CXX: [[RET]] = !DILocation({{.*}}, atomGroup: 8, atomRank: 1)