[ARM] Pre-Commit Tests for PR60510

[lenary]

llvm/test/CodeGen/ARM/fp16-return-pr60510.ll

6

I need to go back and re-enable these tests, or remove these run lines. Oops.

[lenary]

Most recent change adds tests with multiple uses of the returned value.

llvm/test/CodeGen/ARM/fp16-return-pr60510.ll

6

Done.

[dmgreen]

It's a bit odd to read code with the filter-out push/pop, but LGTM.

[lenary]

In D143712#4165622, @dmgreen wrote:

It's a bit odd to read code with the filter-out push/pop, but LGTM.

I figured most of that was prolog/epilog that wasn't where the bug was, and the issue i was running into was the arm/thumb prologs/epilogs were different, but the function bodies were not. This still shows the movs to set the registers used for return values.

llvm/test/CodeGen/ARM/fp16-return-pr60510.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --filter-out "(\.v?save|push|pop)"
+
+; No FP16/BF16
+; RUN: llc -mtriple=arm-none-eabi -float-abi=soft -mattr=+armv8.2-a,+fp-armv8,-fullfp16,-bf16,-neon %s -o - | FileCheck %s --check-prefixes=NO-FP16-SOFT
+; RUN: llc -mtriple=thumb-none-eabi -float-abi=soft -mattr=+armv8.2-a,+fp-armv8,-fullfp16,-bf16,-neon %s -o - | FileCheck %s --check-prefixes=NO-FP16-SOFT
+; RUN: llc -mtriple=arm-none-eabi -mattr=+armv8.2-a,+fp-armv8,-fullfp16,-bf16,-neon %s -o - | FileCheck %s --check-prefixes=NO-FP16-SOFT
+; RUN: llc -mtriple=thumb-none-eabi -mattr=+armv8.2-a,+fp-armv8,-fullfp16,-bf16,-neon %s -o - | FileCheck %s --check-prefixes=NO-FP16-SOFT
+; RUN: llc -mtriple=arm-none-eabihf -mattr=+armv8.2-a,+fp-armv8,-fullfp16,-bf16,-neon %s -o - | FileCheck %s --check-prefixes=NO-FP16-HARD
+; RUN: llc -mtriple=thumb-none-eabihf -mattr=+armv8.2-a,+fp-armv8,-fullfp16,-bf16,-neon %s -o - | FileCheck %s --check-prefixes=NO-FP16-HARD
+
+; FP16/BF16
+; RUN: llc -mtriple=arm-none-eabi -float-abi=soft -mattr=+armv8.2-a,+fp-armv8,+fullfp16,+bf16,-neon %s -o - | FileCheck %s --check-prefixes=FP16-SOFT
+; RUN: llc -mtriple=thumb-none-eabi -float-abi=soft -mattr=+armv8.2-a,+fp-armv8,+fullfp16,+bf16,-neon %s -o - | FileCheck %s --check-prefixes=FP16-SOFT
+; RUN: llc -mtriple=arm-none-eabi -mattr=+armv8.2-a,+fp-armv8,+fullfp16,+bf16,-neon %s -o - | FileCheck %s --check-prefixes=FP16-SOFT
+; RUN: llc -mtriple=thumb-none-eabi -mattr=+armv8.2-a,+fp-armv8,+fullfp16,+bf16,-neon %s -o - | FileCheck %s --check-prefixes=FP16-SOFT
+; RUN: llc -mtriple=arm-none-eabihf -mattr=+armv8.2-a,+fp-armv8,+fullfp16,+bf16,-neon %s -o - | FileCheck %s --check-prefixes=FP16-HARD
+; RUN: llc -mtriple=thumb-none-eabihf -mattr=+armv8.2-a,+fp-armv8,+fullfp16,+bf16,-neon %s -o - | FileCheck %s --check-prefixes=FP16-HARD
+
+; PR60510 showed a bug where the return from `*_inner` was getting "lost" by an
+; optimisation, and a garbage value was being left in `s0`.
+
+declare dso_local float @other(float) nounwind
+declare dso_local void @fp16_sink(half) nounwind
+declare dso_local void @bf16_sink(bfloat) nounwind
+
+declare dso_local bfloat @fp16_inner() nounwind
+declare dso_local bfloat @bf_inner() nounwind
+
+define half @fp16_out_call_oneuse(float %arg) nounwind {
+; NO-FP16-SOFT-LABEL: fp16_out_call_oneuse:
+; NO-FP16-SOFT:  @ %bb.0:
+; NO-FP16-SOFT:    mov r4, r0
+; NO-FP16-SOFT:    bl fp16_inner
+; NO-FP16-SOFT:    mov r5, r0
+; NO-FP16-SOFT:    mov r0, r4
+; NO-FP16-SOFT:    bl other
+; NO-FP16-SOFT:    mov r0, r5
+;
+; NO-FP16-HARD-LABEL: fp16_out_call_oneuse:
+; NO-FP16-HARD:  @ %bb.0:
+; NO-FP16-HARD:    vmov.f32 s16, s0
+; NO-FP16-HARD:    bl fp16_inner
+; NO-FP16-HARD:    vmov.f32 s18, s0
+; NO-FP16-HARD:    vmov.f32 s0, s16
+; NO-FP16-HARD:    bl other
+; NO-FP16-HARD:    vmov.f32 s0, s18
+;
+; FP16-SOFT-LABEL: fp16_out_call_oneuse:
+; FP16-SOFT:  @ %bb.0:
+; FP16-SOFT:    mov r4, r0
+; FP16-SOFT:    bl fp16_inner
+; FP16-SOFT:    mov r5, r0
+; FP16-SOFT:    mov r0, r4
+; FP16-SOFT:    bl other
+; FP16-SOFT:    vmov.f16 s0, r5
+; FP16-SOFT:    vmov r0, s0
+;
+; FP16-HARD-LABEL: fp16_out_call_oneuse:
+; FP16-HARD:  @ %bb.0:
+; FP16-HARD:    vmov.f32 s16, s0
+; FP16-HARD:    bl fp16_inner
+; FP16-HARD:    vmov.f32 s0, s16
+; FP16-HARD:    bl other
+  %call = call half @fp16_inner()
+  %call1 = call float @other(float %arg)
+  ret half %call
+}
+
+define half @fp16_out_call_multiuse(float %arg) nounwind {
+; NO-FP16-SOFT-LABEL: fp16_out_call_multiuse:
+; NO-FP16-SOFT:  @ %bb.0:
+; NO-FP16-SOFT:    mov r4, r0
+; NO-FP16-SOFT:    bl fp16_inner
+; NO-FP16-SOFT:    mov r5, r0
+; NO-FP16-SOFT:    mov r0, r4
+; NO-FP16-SOFT:    bl other
+; NO-FP16-SOFT:    mov r0, r5
+; NO-FP16-SOFT:    bl fp16_sink
+; NO-FP16-SOFT:    mov r0, r5
+;
+; NO-FP16-HARD-LABEL: fp16_out_call_multiuse:
+; NO-FP16-HARD:  @ %bb.0:
+; NO-FP16-HARD:    vmov.f32 s16, s0
+; NO-FP16-HARD:    bl fp16_inner
+; NO-FP16-HARD:    vmov.f32 s18, s0
+; NO-FP16-HARD:    vmov.f32 s0, s16
+; NO-FP16-HARD:    bl other
+; NO-FP16-HARD:    vmov.f32 s0, s18
+; NO-FP16-HARD:    bl fp16_sink
+; NO-FP16-HARD:    vmov.f32 s0, s18
+;
+; FP16-SOFT-LABEL: fp16_out_call_multiuse:
+; FP16-SOFT:  @ %bb.0:
+; FP16-SOFT:    mov r4, r0
+; FP16-SOFT:    bl fp16_inner
+; FP16-SOFT:    mov r5, r0
+; FP16-SOFT:    mov r0, r4
+; FP16-SOFT:    bl other
+; FP16-SOFT:    vmov.f16 s16, r5
+; FP16-SOFT:    vmov.f16 r0, s16
+; FP16-SOFT:    bl fp16_sink
+; FP16-SOFT:    vmov r0, s16
+;
+; FP16-HARD-LABEL: fp16_out_call_multiuse:
+; FP16-HARD:  @ %bb.0:
+; FP16-HARD:    vmov.f32 s16, s0
+; FP16-HARD:    bl fp16_inner
+; FP16-HARD:    vmov.f32 s0, s16
+; FP16-HARD:    bl other
+; FP16-HARD:    vmov.f16 r0, s0
+; FP16-HARD:    vmov.f32 s16, s0
+; FP16-HARD:    vmov s0, r0
+; FP16-HARD:    bl fp16_sink
+; FP16-HARD:    vmov.f32 s0, s16
+  %call = call half @fp16_inner()
+  %call1 = call float @other(float %arg)
+  call void @fp16_sink(half %call)
+  ret half %call
+}
+
+define bfloat @bf_out_call_oneuse(float %arg) nounwind {
+; NO-FP16-SOFT-LABEL: bf_out_call_oneuse:
+; NO-FP16-SOFT:  @ %bb.0:
+; NO-FP16-SOFT:    mov r4, r0
+; NO-FP16-SOFT:    bl bf_inner
+; NO-FP16-SOFT:    mov r5, r0
+; NO-FP16-SOFT:    mov r0, r4
+; NO-FP16-SOFT:    bl other
+; NO-FP16-SOFT:    mov r0, r5
+;
+; NO-FP16-HARD-LABEL: bf_out_call_oneuse:
+; NO-FP16-HARD:  @ %bb.0:
+; NO-FP16-HARD:    vmov.f32 s16, s0
+; NO-FP16-HARD:    bl bf_inner
+; NO-FP16-HARD:    vmov.f32 s18, s0
+; NO-FP16-HARD:    vmov.f32 s0, s16
+; NO-FP16-HARD:    bl other
+; NO-FP16-HARD:    vmov.f32 s0, s18
+;
+; FP16-SOFT-LABEL: bf_out_call_oneuse:
+; FP16-SOFT:  @ %bb.0:
+; FP16-SOFT:    mov r4, r0
+; FP16-SOFT:    bl bf_inner
+; FP16-SOFT:    mov r5, r0
+; FP16-SOFT:    mov r0, r4
+; FP16-SOFT:    bl other
+; FP16-SOFT:    mov r0, r5
+;
+; FP16-HARD-LABEL: bf_out_call_oneuse:
+; FP16-HARD:  @ %bb.0:
+; FP16-HARD:    vmov.f32 s16, s0
+; FP16-HARD:    bl bf_inner
+; FP16-HARD:    vmov.f32 s18, s0
+; FP16-HARD:    vmov.f32 s0, s16
+; FP16-HARD:    bl other
+; FP16-HARD:    vmov.f32 s0, s18
+  %call = call bfloat @bf_inner()
+  %call1 = call float @other(float %arg)
+  ret bfloat %call
+}
+
+define bfloat @bf_out_call_multiuse(float %arg) nounwind {
+; NO-FP16-SOFT-LABEL: bf_out_call_multiuse:
+; NO-FP16-SOFT:  @ %bb.0:
+; NO-FP16-SOFT:    mov r4, r0
+; NO-FP16-SOFT:    bl bf_inner
+; NO-FP16-SOFT:    mov r5, r0
+; NO-FP16-SOFT:    mov r0, r4
+; NO-FP16-SOFT:    bl other
+; NO-FP16-SOFT:    mov r0, r5
+; NO-FP16-SOFT:    bl bf16_sink
+; NO-FP16-SOFT:    mov r0, r5
+;
+; NO-FP16-HARD-LABEL: bf_out_call_multiuse:
+; NO-FP16-HARD:  @ %bb.0:
+; NO-FP16-HARD:    vmov.f32 s16, s0
+; NO-FP16-HARD:    bl bf_inner
+; NO-FP16-HARD:    vmov.f32 s18, s0
+; NO-FP16-HARD:    vmov.f32 s0, s16
+; NO-FP16-HARD:    bl other
+; NO-FP16-HARD:    vmov.f32 s0, s18
+; NO-FP16-HARD:    bl bf16_sink
+; NO-FP16-HARD:    vmov.f32 s0, s18
+;
+; FP16-SOFT-LABEL: bf_out_call_multiuse:
+; FP16-SOFT:  @ %bb.0:
+; FP16-SOFT:    mov r4, r0
+; FP16-SOFT:    bl bf_inner
+; FP16-SOFT:    mov r5, r0
+; FP16-SOFT:    mov r0, r4
+; FP16-SOFT:    bl other
+; FP16-SOFT:    mov r0, r5
+; FP16-SOFT:    bl bf16_sink
+; FP16-SOFT:    mov r0, r5
+;
+; FP16-HARD-LABEL: bf_out_call_multiuse:
+; FP16-HARD:  @ %bb.0:
+; FP16-HARD:    vmov.f32 s16, s0
+; FP16-HARD:    bl bf_inner
+; FP16-HARD:    vmov.f32 s18, s0
+; FP16-HARD:    vmov.f32 s0, s16
+; FP16-HARD:    bl other
+; FP16-HARD:    vmov.f32 s0, s18
+; FP16-HARD:    bl bf16_sink
+; FP16-HARD:    vmov.f32 s0, s18
+  %call = call bfloat @bf_inner()
+  %call1 = call float @other(float %arg)
+  call void @bf16_sink(bfloat %call)
+  ret bfloat %call
+}