[HLSL] Preserve vec3 for HLSL.

[python3kgae]

Summary

Preserve vec3 for HLSL by set -fpreserve-vec3-type.

[beanz]

clang/lib/Driver/ToolChains/Clang.cpp

3524 ↗

(On Diff #456559)

Preserving vec3 is required for HLSL correctness, this shouldn't be tied to the DXC driver mode. Instead, the option should be implied by the HLSL language mode. You should be able to add the following to the TableGen definition for the preserve-vec3-type flag:

ImpliedByAnyOf<[hlsl.KeyPath]>;

[python3kgae]

Switch to ImpliedByAnyOf.

[beanz]

LGTM

[chapuni]

The test fails for -Asserts.

clang/test/CodeGenHLSL/float3.hlsl

4

Also %a may be unnamed, like %0

[python3kgae]

clang/test/CodeGenHLSL/float3.hlsl

4

Thanks for pointing this out.
I'll fix it.

[python3kgae]

In D132913#3781494, @chapuni wrote:

The test fails for -Asserts.

Fixed the %0 issue in https://github.com/llvm/llvm-project/commit/71fae33f5e7c1b11e32db695fd24dd50aead737c
It passed locally on build with -DLLVM_ENABLE_ASSERTIONS=ON

Is -DLLVM_ENABLE_ASSERTIONS=ON what you mean by -Asserts?

[chapuni]

The fix looks good. Thanks.

Is -DLLVM_ENABLE_ASSERTIONS=ON what you mean by -Asserts?

Oh excuse me, "-Asserts" was an ancient term, "w/o assertsions" in the autoconf age.
An antonym was "+Asserts". it means LLVM_ENABLE_ASSERTIONS=ON

[python3kgae]

In D132913#3781591, @chapuni wrote:

The fix looks good. Thanks.

Is -DLLVM_ENABLE_ASSERTIONS=ON what you mean by -Asserts?

Oh excuse me, "-Asserts" was an ancient term, "w/o assertsions" in the autoconf age.
An antonym was "+Asserts". it means LLVM_ENABLE_ASSERTIONS=ON

I see. I'll test LLVM_ENABLE_ASSERTIONS=OFF this time.

[python3kgae]

In D132913#3781611, @python3kgae wrote:

In D132913#3781591, @chapuni wrote:

The fix looks good. Thanks.

Is -DLLVM_ENABLE_ASSERTIONS=ON what you mean by -Asserts?

Oh excuse me, "-Asserts" was an ancient term, "w/o assertsions" in the autoconf age.
An antonym was "+Asserts". it means LLVM_ENABLE_ASSERTIONS=ON

I see. I'll test LLVM_ENABLE_ASSERTIONS=OFF this time.

Confirmed pass with LLVM_ENABLE_ASSERTIONS=OFF

[alexgatea]

This test fails because the actual output has <3 x float>* rather than ptr. Could you please fix this test case?

[python3kgae]

In D132913#3786432, @alexgatea wrote:

This test fails because the actual output has <3 x float>* rather than ptr. Could you please fix this test case?

Thanks for reporting the issue.
But I cannot repro the fail.
Do you mind sharing your cmake command?

[alexgatea]

Thanks for reporting the issue.
But I cannot repro the fail.
Do you mind sharing your cmake command?

Hmm, I'm just running the command in the test case:
clang --driver-mode=dxc -Tlib_6_7 -fcgl -Fo - clang/test/CodeGenHLSL/float3.hlsl | FileCheck clang/test/CodeGenHLSL/float3.hlsl

[python3kgae]

In D132913#3786679, @alexgatea wrote:

Thanks for reporting the issue.
But I cannot repro the fail.
Do you mind sharing your cmake command?

Hmm, I'm just running the command in the test case:
clang --driver-mode=dxc -Tlib_6_7 -fcgl -Fo - clang/test/CodeGenHLSL/float3.hlsl | FileCheck clang/test/CodeGenHLSL/float3.hlsl

This is the output I got with clang --driver-mode=dxc -Tlib_6_7 -fcgl -Fo - clang/test/CodeGenHLSL/float3.hlsl

target datalayout = "e-m:e-p:32:32-i1:32-i8:8-i16:16-i32:32-i64:64-f16:16-f32:32-f64:64-n8:16:32:64"
target triple = "dxil-unknown-shadermodel6.7-library"

; Function Attrs: noinline nounwind optnone
define noundef <3 x float> @"?foo@@YAT?$__vector@M$02@__clang@@T12@@Z"(<3 x float> noundef %a) #0 {
entry:
  %a.addr = alloca <3 x float>, align 16
  store <3 x float> %a, ptr %a.addr, align 16
  %0 = load <3 x float>, ptr %a.addr, align 16
  ret <3 x float> %0
}

attributes #0 = { noinline nounwind optnone "frame-pointer"="all" "min-legal-vector-width"="96" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }

!llvm.module.flags = !{!0, !1}
!dx.valver = !{!2}
!llvm.ident = !{!3}

!0 = !{i32 1, !"wchar_size", i32 4}
!1 = !{i32 7, !"frame-pointer", i32 2}
!2 = !{i32 1, i32 7}
!3 = !{!"clang version 16.0.0 (https://github.com/llvm/llvm-project c9d2b6b92d6c29d00f6adc0527cf2331dbaae31a)"}

Maybe you build clang with a different setting? What are the CMake options you're using?

[alexgatea]

In D132913#3786734, @python3kgae wrote:

In D132913#3786679, @alexgatea wrote:

Thanks for reporting the issue.
But I cannot repro the fail.
Do you mind sharing your cmake command?

Hmm, I'm just running the command in the test case:
clang --driver-mode=dxc -Tlib_6_7 -fcgl -Fo - clang/test/CodeGenHLSL/float3.hlsl | FileCheck clang/test/CodeGenHLSL/float3.hlsl

This is the output I got with clang --driver-mode=dxc -Tlib_6_7 -fcgl -Fo - clang/test/CodeGenHLSL/float3.hlsl

target datalayout = "e-m:e-p:32:32-i1:32-i8:8-i16:16-i32:32-i64:64-f16:16-f32:32-f64:64-n8:16:32:64"
target triple = "dxil-unknown-shadermodel6.7-library"

; Function Attrs: noinline nounwind optnone
define noundef <3 x float> @"?foo@@YAT?$__vector@M$02@__clang@@T12@@Z"(<3 x float> noundef %a) #0 {
entry:
  %a.addr = alloca <3 x float>, align 16
  store <3 x float> %a, ptr %a.addr, align 16
  %0 = load <3 x float>, ptr %a.addr, align 16
  ret <3 x float> %0
}

attributes #0 = { noinline nounwind optnone "frame-pointer"="all" "min-legal-vector-width"="96" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }

!llvm.module.flags = !{!0, !1}
!dx.valver = !{!2}
!llvm.ident = !{!3}

!0 = !{i32 1, !"wchar_size", i32 4}
!1 = !{i32 7, !"frame-pointer", i32 2}
!2 = !{i32 1, i32 7}
!3 = !{!"clang version 16.0.0 (https://github.com/llvm/llvm-project c9d2b6b92d6c29d00f6adc0527cf2331dbaae31a)"}

Maybe you build clang with a different setting? What are the CMake options you're using?

Sorry for the delay and thank you for verifying. It turns out this is not due to a problem with the test itself, but rather a local problem with my compiler. No revision is necessary here, my apologies.

[tahonermann]

clang/test/CodeGenHLSL/float3.hlsl

1

Does this need to use driver mode? That is odd for a code gen test.

[python3kgae]

clang/test/CodeGenHLSL/float3.hlsl

1

Nice catch.
Fixed with https://github.com/llvm/llvm-project/commit/8b2f8b309885e872ad98af381ee5459716955877

Thanks.

clang/include/clang/Driver/Options.td

 // These options cannot be marshalled, because they are used to set up the LangOptions defaults.
 def finclude_default_header : Flag<["-"], "finclude-default-header">,
   HelpText<"Include default header file for OpenCL and HLSL">;
 def fdeclare_opencl_builtins : Flag<["-"], "fdeclare-opencl-builtins">,
   HelpText<"Add OpenCL builtin function declarations (experimental)">;
 
 def fpreserve_vec3_type : Flag<["-"], "fpreserve-vec3-type">,
   HelpText<"Preserve 3-component vector type">,
-  MarshallingInfoFlag<CodeGenOpts<"PreserveVec3Type">>;
+  MarshallingInfoFlag<CodeGenOpts<"PreserveVec3Type">>,
+  ImpliedByAnyOf<[hlsl.KeyPath]>;
 def fwchar_type_EQ : Joined<["-"], "fwchar-type=">,
   HelpText<"Select underlying type for wchar_t">,
   Values<"char,short,int">,
   NormalizedValues<["1", "2", "4"]>,
   MarshallingInfoEnum<LangOpts<"WCharSize">, "0">;
 defm signed_wchar : BoolOption<"f", "signed-wchar",
   LangOpts<"WCharIsSigned">, DefaultTrue,
   NegFlag<SetFalse, [CC1Option], "Use an unsigned">, PosFlag<SetTrue, [], "Use a signed">,

clang/test/CodeGenHLSL/float3.hlsl

+// RUN: %clang --driver-mode=dxc -Tlib_6_7 -fcgl  -Fo - %s | FileCheck %s
+
+// Make sure float3 is not changed into float4.
+// CHECK:<3 x float> @"?foo@@YAT?$__vector@M$02@__clang@@T12@@Z"(<3 x float> noundef %a)
+// CHECK:%[[A_ADDR:.+]] = alloca <3 x float>, align 16
+// CHECK-NEXT:store <3 x float> %a, ptr %[[A_ADDR]], align 16
+// CHECK-NEXT:%[[V:[0-9]+]] = load <3 x float>, ptr %[[A_ADDR]], align 16
+// CHECK-NEXT:ret <3 x float> %[[V]]
+float3 foo(float3 a) {
+  return a;
+}