Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[HLSL] Expose half types and intrinsics always #81782

Merged
merged 1 commit into from
Feb 15, 2024
Merged

[HLSL] Expose half types and intrinsics always #81782

merged 1 commit into from
Feb 15, 2024

Conversation

llvm-beanz
Copy link
Collaborator

We previously made an implmenetation error when adding half overloads for HLSL library functionalitly. The half type is always defined in HLSL and half intrinsics should not be conditionally included.

When native 16-bit types are disabled half is a unique 32-bit float type with lesser promotion rank than float.

Fixes #81049

@llvm-beanz
[HLSL] Expose half types and intrinsics always
e68a1db 
We previously made an implmenetation error when adding `half` overloads
for HLSL library functionalitly. The `half` type is always defined in
HLSL and `half` intrinsics should not be conditionally included.

When native 16-bit types are disabled `half` is a unique 32-bit float
type with lesser promotion rank than `float`.

Fixes llvm#81049
@llvmbot llvmbot added clang Clang issues not falling into any other category backend:X86 clang:headers Headers provided by Clang, e.g. for intrinsics HLSL HLSL Language Support labels Feb 14, 2024
@llvmbot
Copy link
Collaborator

llvmbot commented Feb 14, 2024 •
edited

@llvm/pr-subscribers-hlsl

@llvm/pr-subscribers-backend-x86

Author: Chris B (llvm-beanz)

Changes

We previously made an implmenetation error when adding half overloads for HLSL library functionalitly. The half type is always defined in HLSL and half intrinsics should not be conditionally included.

When native 16-bit types are disabled half is a unique 32-bit float type with lesser promotion rank than float.

Fixes #81049

Patch is 25.54 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/81782.diff

15 Files Affected:

  • (modified) clang/lib/Headers/hlsl/hlsl_basic_types.h (-2)
  • (modified) clang/lib/Headers/hlsl/hlsl_intrinsics.h (+59-72)
  • (modified) clang/test/CodeGenHLSL/builtins/abs.hlsl (+4-1)
  • (modified) clang/test/CodeGenHLSL/builtins/ceil.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/cos.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/floor.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/log.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/log10.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/log2.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/max.hlsl (+1-17)
  • (modified) clang/test/CodeGenHLSL/builtins/min.hlsl (+1-17)
  • (modified) clang/test/CodeGenHLSL/builtins/pow.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/reversebits.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/sin.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/trunc.hlsl (+1-1)
diff --git a/clang/lib/Headers/hlsl/hlsl_basic_types.h b/clang/lib/Headers/hlsl/hlsl_basic_types.h
index e96fa90b1ce469..3d0d296aadca3a 100644
--- a/clang/lib/Headers/hlsl/hlsl_basic_types.h
+++ b/clang/lib/Headers/hlsl/hlsl_basic_types.h
@@ -56,11 +56,9 @@ typedef vector<uint64_t, 2> uint64_t2;
 typedef vector<uint64_t, 3> uint64_t3;
 typedef vector<uint64_t, 4> uint64_t4;
 
-#ifdef __HLSL_ENABLE_16_BIT
 typedef vector<half, 2> half2;
 typedef vector<half, 3> half3;
 typedef vector<half, 4> half4;
-#endif
 
 typedef vector<float, 2> float2;
 typedef vector<float, 3> float3;
diff --git a/clang/lib/Headers/hlsl/hlsl_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
index a8b36d29c78607..f87ac977997962 100644
--- a/clang/lib/Headers/hlsl/hlsl_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
@@ -21,6 +21,13 @@ namespace hlsl {
 #define _HLSL_AVAILABILITY(environment, version)                               \
   __attribute__((availability(environment, introduced = version)))
 
+#ifdef __HLSL_ENABLE_16_BIT
+#define _HLSL_16BIT_AVAILABILITY(environment, version)                         \
+  __attribute__((availability(environment, introduced = version)))
+#else
+#define _HLSL_16BIT_AVAILABILITY(environment, version)
+#endif
+
 //===----------------------------------------------------------------------===//
 // abs builtins
 //===----------------------------------------------------------------------===//
@@ -42,20 +49,20 @@ int16_t3 abs(int16_t3);
 _HLSL_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_abs)
 int16_t4 abs(int16_t4);
+#endif
 
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_abs)
 half abs(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_abs)
 half2 abs(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_abs)
 half3 abs(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_abs)
 half4 abs(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_abs)
 int abs(int);
@@ -102,20 +109,18 @@ double4 abs(double4);
 /// the input value, \a Val.
 /// \param Val The input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_ceil)
 half ceil(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_ceil)
 half2 ceil(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_ceil)
 half3 ceil(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_ceil)
 half4 ceil(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_ceil)
 float ceil(float);
@@ -143,20 +148,18 @@ double4 ceil(double4);
 /// \brief Returns the cosine of the input value, \a Val.
 /// \param Val The input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_cos)
 half cos(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_cos)
 half2 cos(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_cos)
 half3 cos(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_cos)
 half4 cos(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_cos)
 float cos(float);
@@ -185,20 +188,18 @@ double4 cos(double4);
 /// value, \a Val.
 /// \param Val The input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_floor)
 half floor(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_floor)
 half2 floor(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_floor)
 half3 floor(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_floor)
 half4 floor(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_floor)
 float floor(float);
@@ -229,20 +230,18 @@ double4 floor(double4);
 /// If \a Val is negative, this result is undefined. If \a Val is 0, this
 /// function returns negative infinity.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log)
 half log(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log)
 half2 log(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log)
 half3 log(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log)
 half4 log(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log)
 float log(float);
@@ -273,20 +272,18 @@ double4 log(double4);
 /// If \a Val is negative, this result is undefined. If \a Val is 0, this
 /// function returns negative infinity.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log10)
 half log10(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log10)
 half2 log10(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log10)
 half3 log10(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log10)
 half4 log10(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log10)
 float log10(float);
@@ -317,20 +314,18 @@ double4 log10(double4);
 /// If \a Val is negative, this result is undefined. If \a Val is 0, this
 /// function returns negative infinity.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log2)
 half log2(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log2)
 half2 log2(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log2)
 half3 log2(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log2)
 half4 log2(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log2)
 float log2(float);
@@ -359,20 +354,20 @@ double4 log2(double4);
 /// \param X The X input value.
 /// \param Y The Y input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
 half max(half, half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
 half2 max(half2, half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
 half3 max(half3, half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
 half4 max(half4, half4);
 
+#ifdef __HLSL_ENABLE_16_BIT
 _HLSL_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
 int16_t max(int16_t, int16_t);
@@ -463,20 +458,20 @@ double4 max(double4, double4);
 /// \param X The X input value.
 /// \param Y The Y input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
 half min(half, half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
 half2 min(half2, half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
 half3 min(half3, half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
 half4 min(half4, half4);
 
+#ifdef __HLSL_ENABLE_16_BIT
 _HLSL_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
 int16_t min(int16_t, int16_t);
@@ -567,20 +562,18 @@ double4 min(double4, double4);
 /// \param Val The input value.
 /// \param Pow The specified power.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_pow)
 half pow(half, half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_pow)
 half2 pow(half2, half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_pow)
 half3 pow(half3, half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_pow)
 half4 pow(half4, half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_pow)
 float pow(float, float);
@@ -680,20 +673,18 @@ uint64_t4 reversebits(uint64_t4);
 /// \brief Returns the sine of the input value, \a Val.
 /// \param Val The input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_sin)
 half sin(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_sin)
 half2 sin(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_sin)
 half3 sin(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_sin)
 half4 sin(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_sin)
 float sin(float);
@@ -721,11 +712,9 @@ double4 sin(double4);
 /// \brief Returns the square root of the input value, \a Val.
 /// \param Val The input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_sqrtf16)
 half sqrt(half In);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_sqrtf)
 float sqrt(float In);
@@ -741,20 +730,18 @@ double sqrt(double In);
 /// \brief Returns the truncated integer value of the input value, \a Val.
 /// \param Val The input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_trunc)
 half trunc(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_trunc)
 half2 trunc(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_trunc)
 half3 trunc(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_trunc)
 half4 trunc(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_trunc)
 float trunc(float);
diff --git a/clang/test/CodeGenHLSL/builtins/abs.hlsl b/clang/test/CodeGenHLSL/builtins/abs.hlsl
index dc2187ae157f98..54c9d1a9dded45 100644
--- a/clang/test/CodeGenHLSL/builtins/abs.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/abs.hlsl
@@ -3,10 +3,11 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 using hlsl::abs;
 
+#ifdef __HLSL_ENABLE_16_BIT
 // CHECK: define noundef i16 @
 // CHECK: call i16 @llvm.abs.i16(
 int16_t test_abs_int16_t ( int16_t p0 ) {
@@ -27,6 +28,8 @@ int16_t3 test_abs_int16_t3 ( int16_t3 p0 ) {
 int16_t4 test_abs_int16_t4 ( int16_t4 p0 ) {
   return abs ( p0 );
 }
+#endif // __HLSL_ENABLE_16_BIT
+
 // CHECK: define noundef half @
 // CHECK: call half @llvm.fabs.f16(
 // NO_HALF: define noundef float @"?test_abs_half@@YA$halff@$halff@@Z"(
diff --git a/clang/test/CodeGenHLSL/builtins/ceil.hlsl b/clang/test/CodeGenHLSL/builtins/ceil.hlsl
index 5969028d56d1a0..f1672816e72bc2 100644
--- a/clang/test/CodeGenHLSL/builtins/ceil.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/ceil.hlsl
@@ -3,7 +3,7 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 using hlsl::ceil;
 
diff --git a/clang/test/CodeGenHLSL/builtins/cos.hlsl b/clang/test/CodeGenHLSL/builtins/cos.hlsl
index e05c0cfee34753..2fc1571949b2c5 100644
--- a/clang/test/CodeGenHLSL/builtins/cos.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/cos.hlsl
@@ -3,7 +3,7 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 // CHECK: define noundef half @
 // CHECK: call half @llvm.cos.f16(
diff --git a/clang/test/CodeGenHLSL/builtins/floor.hlsl b/clang/test/CodeGenHLSL/builtins/floor.hlsl
index 38834a146206fa..357661761b762a 100644
--- a/clang/test/CodeGenHLSL/builtins/floor.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/floor.hlsl
@@ -3,7 +3,7 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 using hlsl::floor;
 
diff --git a/clang/test/CodeGenHLSL/builtins/log.hlsl b/clang/test/CodeGenHLSL/builtins/log.hlsl
index 8a8a2de9f82e0c..6a8e4ac2e5f294 100644
--- a/clang/test/CodeGenHLSL/builtins/log.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/log.hlsl
@@ -3,7 +3,7 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 // CHECK: define noundef half @
 // CHECK: call half @llvm.log.f16(
diff --git a/clang/test/CodeGenHLSL/builtins/log10.hlsl b/clang/test/CodeGenHLSL/builtins/log10.hlsl
index 8b2bd26242cf72..8ce24fd530dd3c 100644
--- a/clang/test/CodeGenHLSL/builtins/log10.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/log10.hlsl
@@ -3,7 +3,7 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 // CHECK: define noundef half @
 // CHECK: call half @llvm.log10.f16(
diff --git a/clang/test/CodeGenHLSL/builtins/log2.hlsl b/clang/test/CodeGenHLSL/builtins/log2.hlsl
index 85060c559ddcc5..f0f0a6c7c50e81 100644
--- a/clang/test/CodeGenHLSL/builtins/log2.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/log2.hlsl
@@ -3,7 +3,7 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 // CHECK: define noundef half @
 // CHECK: call half @llvm.log2.f16(
diff --git a/clang/test/CodeGenHLSL/builtins/max.hlsl b/clang/test/CodeGenHLSL/builtins/max.hlsl
index d93db562c31405..d8879c3332fb88 100644
--- a/clang/test/CodeGenHLSL/builtins/max.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/max.hlsl
@@ -3,63 +3,47 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 #ifdef __HLSL_ENABLE_16_BIT
 // CHECK: define noundef i16 @
 // CHECK: call i16 @llvm.smax.i16(
-// NO_HALF: define noundef i16 @"?test_max_short@@YAFFF@Z"(
-// NO_HALF: call i16 @llvm.smax.i16(
 int16_t test_max_short ( int16_t p0, int16_t p1 ) {
   return max ( p0, p1 );
 }
 // CHECK: define noundef <2 x i16> @
 // CHECK: call <2 x i16> @llvm.smax.v2i16(
-// NO_HALF: define noundef <2 x i16> @"?test_max_short2@@YAT?$__vector@F$01@__clang@@T12@0@Z"(
-// NO_HALF: call <2 x i16> @llvm.smax.v2i16(
 int16_t2 test_max_short2 ( int16_t2 p0, int16_t2 p1 ) {
   return max ( p0, p1 );
 }
 // CHECK: define noundef <3 x i16> @
 // CHECK: call <3 x i16> @llvm.smax.v3i16
-// NO_HALF: define noundef <3 x i16> @"?test_max_short3@@YAT?$__vector@F$02@__clang@@T12@0@Z"(
-// NO_HALF: call <3 x i16> @llvm.smax.v3i16(
 int16_t3 test_max_short3 ( int16_t3 p0, int16_t3 p1 ) {
   return max ( p0, p1 );
 }
 // CHECK: define noundef <4 x i16> @
 // CHECK: call <4 x i16> @llvm.smax.v4i16
-// NO_HALF: define noundef <4 x i16> @"?test_max_short4@@YAT?$__vector@F$03@__clang@@T12@0@Z"(
-// NO_HALF: call <4 x i16> @llvm.smax.v4i16(
 int16_t4 test_max_short4 ( int16_t4 p0, int16_t4 p1 ) {
   return max ( p0, p1 );
 }
 
 // CHECK: define noundef i16 @
 // CHECK: call i16 @llvm.umax.i16(
-// NO_HALF: define noundef i16 @"?test_max_ushort@@YAGGG@Z"(
-// NO_HALF: call i16 @llvm.umax.i16(
 uint16_t test_max_ushort ( uint16_t p0, uint16_t p1 ) {
   return max ( p0, p1 );
 }
 // CHECK: define noundef <2 x i16> @
 // CHECK: call <2 x i16> @llvm.umax.v2i16
-// NO_HALF: define noundef <2 x i16> @"?test_max_ushort2@@YAT?$__vector@G$01@__clang@@T12@0@Z"(
-// NO_HALF: call <2 x i16> @llvm.umax.v2i16(
 uint16_t2 test_max_ushort2 ( uint16_t2 p0, uint16_t2 p1 ) {
   return max ( p0, p1 );
 }
 // CHECK: define noundef <3 x i16> @
 // CHECK: call <3 x i16> @llvm.umax.v3i16
-// NO_HALF: define noundef <3 x i16> @"?test_max_ushort3@@YAT?$__vector@G$02@__clang@@T12@0@Z"(
-// NO_HALF: call <3 x i16> @llvm.umax.v3i16(
 uint16_t3 test_max_ushort3 ( uint16_t3 p0, uint16_t3 p1 ) {
   return max ( p0, p1 );
 }
 // CHECK: define noundef <4 x i16> @
 // CHECK: call <4 x i16> @llvm.umax.v4i16
-// NO_HALF: define noundef <4 x i16> @"?test_max_ushort4@@YAT?$__vector@G$03@__clang@@T12@0@Z"(
-// NO_HALF: call <4 x i16> @llvm.umax.v4i16(
 uint16_t4 test_max_ushort4 ( uint16_t4 p0, uint16_t4 p1 ) {
   return max ( p0, p1 );
 }
diff --git a/clang/test/CodeGenHLSL/builtins/min.hlsl b/clang/test/CodeGenHLSL/builtins/min.hlsl
index 09066971b3f5d0..743053cbdd2620 10...
[truncated]

@llvmbot
Copy link
Collaborator

llvmbot commented Feb 14, 2024

@llvm/pr-subscribers-clang

Author: Chris B (llvm-beanz)

Changes

We previously made an implmenetation error when adding half overloads for HLSL library functionalitly. The half type is always defined in HLSL and half intrinsics should not be conditionally included.

When native 16-bit types are disabled half is a unique 32-bit float type with lesser promotion rank than float.

Fixes #81049

Patch is 25.54 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/81782.diff

15 Files Affected:

  • (modified) clang/lib/Headers/hlsl/hlsl_basic_types.h (-2)
  • (modified) clang/lib/Headers/hlsl/hlsl_intrinsics.h (+59-72)
  • (modified) clang/test/CodeGenHLSL/builtins/abs.hlsl (+4-1)
  • (modified) clang/test/CodeGenHLSL/builtins/ceil.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/cos.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/floor.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/log.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/log10.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/log2.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/max.hlsl (+1-17)
  • (modified) clang/test/CodeGenHLSL/builtins/min.hlsl (+1-17)
  • (modified) clang/test/CodeGenHLSL/builtins/pow.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/reversebits.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/sin.hlsl (+1-1)
  • (modified) clang/test/CodeGenHLSL/builtins/trunc.hlsl (+1-1)
diff --git a/clang/lib/Headers/hlsl/hlsl_basic_types.h b/clang/lib/Headers/hlsl/hlsl_basic_types.h
index e96fa90b1ce469..3d0d296aadca3a 100644
--- a/clang/lib/Headers/hlsl/hlsl_basic_types.h
+++ b/clang/lib/Headers/hlsl/hlsl_basic_types.h
@@ -56,11 +56,9 @@ typedef vector<uint64_t, 2> uint64_t2;
 typedef vector<uint64_t, 3> uint64_t3;
 typedef vector<uint64_t, 4> uint64_t4;
 
-#ifdef __HLSL_ENABLE_16_BIT
 typedef vector<half, 2> half2;
 typedef vector<half, 3> half3;
 typedef vector<half, 4> half4;
-#endif
 
 typedef vector<float, 2> float2;
 typedef vector<float, 3> float3;
diff --git a/clang/lib/Headers/hlsl/hlsl_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
index a8b36d29c78607..f87ac977997962 100644
--- a/clang/lib/Headers/hlsl/hlsl_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
@@ -21,6 +21,13 @@ namespace hlsl {
 #define _HLSL_AVAILABILITY(environment, version)                               \
   __attribute__((availability(environment, introduced = version)))
 
+#ifdef __HLSL_ENABLE_16_BIT
+#define _HLSL_16BIT_AVAILABILITY(environment, version)                         \
+  __attribute__((availability(environment, introduced = version)))
+#else
+#define _HLSL_16BIT_AVAILABILITY(environment, version)
+#endif
+
 //===----------------------------------------------------------------------===//
 // abs builtins
 //===----------------------------------------------------------------------===//
@@ -42,20 +49,20 @@ int16_t3 abs(int16_t3);
 _HLSL_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_abs)
 int16_t4 abs(int16_t4);
+#endif
 
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_abs)
 half abs(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_abs)
 half2 abs(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_abs)
 half3 abs(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_abs)
 half4 abs(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_abs)
 int abs(int);
@@ -102,20 +109,18 @@ double4 abs(double4);
 /// the input value, \a Val.
 /// \param Val The input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_ceil)
 half ceil(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_ceil)
 half2 ceil(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_ceil)
 half3 ceil(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_ceil)
 half4 ceil(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_ceil)
 float ceil(float);
@@ -143,20 +148,18 @@ double4 ceil(double4);
 /// \brief Returns the cosine of the input value, \a Val.
 /// \param Val The input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_cos)
 half cos(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_cos)
 half2 cos(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_cos)
 half3 cos(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_cos)
 half4 cos(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_cos)
 float cos(float);
@@ -185,20 +188,18 @@ double4 cos(double4);
 /// value, \a Val.
 /// \param Val The input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_floor)
 half floor(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_floor)
 half2 floor(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_floor)
 half3 floor(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_floor)
 half4 floor(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_floor)
 float floor(float);
@@ -229,20 +230,18 @@ double4 floor(double4);
 /// If \a Val is negative, this result is undefined. If \a Val is 0, this
 /// function returns negative infinity.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log)
 half log(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log)
 half2 log(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log)
 half3 log(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log)
 half4 log(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log)
 float log(float);
@@ -273,20 +272,18 @@ double4 log(double4);
 /// If \a Val is negative, this result is undefined. If \a Val is 0, this
 /// function returns negative infinity.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log10)
 half log10(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log10)
 half2 log10(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log10)
 half3 log10(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log10)
 half4 log10(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log10)
 float log10(float);
@@ -317,20 +314,18 @@ double4 log10(double4);
 /// If \a Val is negative, this result is undefined. If \a Val is 0, this
 /// function returns negative infinity.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log2)
 half log2(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log2)
 half2 log2(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log2)
 half3 log2(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log2)
 half4 log2(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_log2)
 float log2(float);
@@ -359,20 +354,20 @@ double4 log2(double4);
 /// \param X The X input value.
 /// \param Y The Y input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
 half max(half, half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
 half2 max(half2, half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
 half3 max(half3, half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
 half4 max(half4, half4);
 
+#ifdef __HLSL_ENABLE_16_BIT
 _HLSL_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
 int16_t max(int16_t, int16_t);
@@ -463,20 +458,20 @@ double4 max(double4, double4);
 /// \param X The X input value.
 /// \param Y The Y input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
 half min(half, half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
 half2 min(half2, half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
 half3 min(half3, half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
 half4 min(half4, half4);
 
+#ifdef __HLSL_ENABLE_16_BIT
 _HLSL_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
 int16_t min(int16_t, int16_t);
@@ -567,20 +562,18 @@ double4 min(double4, double4);
 /// \param Val The input value.
 /// \param Pow The specified power.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_pow)
 half pow(half, half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_pow)
 half2 pow(half2, half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_pow)
 half3 pow(half3, half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_pow)
 half4 pow(half4, half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_pow)
 float pow(float, float);
@@ -680,20 +673,18 @@ uint64_t4 reversebits(uint64_t4);
 /// \brief Returns the sine of the input value, \a Val.
 /// \param Val The input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_sin)
 half sin(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_sin)
 half2 sin(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_sin)
 half3 sin(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_sin)
 half4 sin(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_sin)
 float sin(float);
@@ -721,11 +712,9 @@ double4 sin(double4);
 /// \brief Returns the square root of the input value, \a Val.
 /// \param Val The input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_sqrtf16)
 half sqrt(half In);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_sqrtf)
 float sqrt(float In);
@@ -741,20 +730,18 @@ double sqrt(double In);
 /// \brief Returns the truncated integer value of the input value, \a Val.
 /// \param Val The input value.
 
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_trunc)
 half trunc(half);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_trunc)
 half2 trunc(half2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_trunc)
 half3 trunc(half3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_trunc)
 half4 trunc(half4);
-#endif
 
 _HLSL_BUILTIN_ALIAS(__builtin_elementwise_trunc)
 float trunc(float);
diff --git a/clang/test/CodeGenHLSL/builtins/abs.hlsl b/clang/test/CodeGenHLSL/builtins/abs.hlsl
index dc2187ae157f98..54c9d1a9dded45 100644
--- a/clang/test/CodeGenHLSL/builtins/abs.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/abs.hlsl
@@ -3,10 +3,11 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 using hlsl::abs;
 
+#ifdef __HLSL_ENABLE_16_BIT
 // CHECK: define noundef i16 @
 // CHECK: call i16 @llvm.abs.i16(
 int16_t test_abs_int16_t ( int16_t p0 ) {
@@ -27,6 +28,8 @@ int16_t3 test_abs_int16_t3 ( int16_t3 p0 ) {
 int16_t4 test_abs_int16_t4 ( int16_t4 p0 ) {
   return abs ( p0 );
 }
+#endif // __HLSL_ENABLE_16_BIT
+
 // CHECK: define noundef half @
 // CHECK: call half @llvm.fabs.f16(
 // NO_HALF: define noundef float @"?test_abs_half@@YA$halff@$halff@@Z"(
diff --git a/clang/test/CodeGenHLSL/builtins/ceil.hlsl b/clang/test/CodeGenHLSL/builtins/ceil.hlsl
index 5969028d56d1a0..f1672816e72bc2 100644
--- a/clang/test/CodeGenHLSL/builtins/ceil.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/ceil.hlsl
@@ -3,7 +3,7 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 using hlsl::ceil;
 
diff --git a/clang/test/CodeGenHLSL/builtins/cos.hlsl b/clang/test/CodeGenHLSL/builtins/cos.hlsl
index e05c0cfee34753..2fc1571949b2c5 100644
--- a/clang/test/CodeGenHLSL/builtins/cos.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/cos.hlsl
@@ -3,7 +3,7 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 // CHECK: define noundef half @
 // CHECK: call half @llvm.cos.f16(
diff --git a/clang/test/CodeGenHLSL/builtins/floor.hlsl b/clang/test/CodeGenHLSL/builtins/floor.hlsl
index 38834a146206fa..357661761b762a 100644
--- a/clang/test/CodeGenHLSL/builtins/floor.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/floor.hlsl
@@ -3,7 +3,7 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 using hlsl::floor;
 
diff --git a/clang/test/CodeGenHLSL/builtins/log.hlsl b/clang/test/CodeGenHLSL/builtins/log.hlsl
index 8a8a2de9f82e0c..6a8e4ac2e5f294 100644
--- a/clang/test/CodeGenHLSL/builtins/log.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/log.hlsl
@@ -3,7 +3,7 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 // CHECK: define noundef half @
 // CHECK: call half @llvm.log.f16(
diff --git a/clang/test/CodeGenHLSL/builtins/log10.hlsl b/clang/test/CodeGenHLSL/builtins/log10.hlsl
index 8b2bd26242cf72..8ce24fd530dd3c 100644
--- a/clang/test/CodeGenHLSL/builtins/log10.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/log10.hlsl
@@ -3,7 +3,7 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 // CHECK: define noundef half @
 // CHECK: call half @llvm.log10.f16(
diff --git a/clang/test/CodeGenHLSL/builtins/log2.hlsl b/clang/test/CodeGenHLSL/builtins/log2.hlsl
index 85060c559ddcc5..f0f0a6c7c50e81 100644
--- a/clang/test/CodeGenHLSL/builtins/log2.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/log2.hlsl
@@ -3,7 +3,7 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 // CHECK: define noundef half @
 // CHECK: call half @llvm.log2.f16(
diff --git a/clang/test/CodeGenHLSL/builtins/max.hlsl b/clang/test/CodeGenHLSL/builtins/max.hlsl
index d93db562c31405..d8879c3332fb88 100644
--- a/clang/test/CodeGenHLSL/builtins/max.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/max.hlsl
@@ -3,63 +3,47 @@
 // RUN:   -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
 // RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -D__HLSL_ENABLE_16_BIT -o - | FileCheck %s --check-prefix=NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefix=NO_HALF
 
 #ifdef __HLSL_ENABLE_16_BIT
 // CHECK: define noundef i16 @
 // CHECK: call i16 @llvm.smax.i16(
-// NO_HALF: define noundef i16 @"?test_max_short@@YAFFF@Z"(
-// NO_HALF: call i16 @llvm.smax.i16(
 int16_t test_max_short ( int16_t p0, int16_t p1 ) {
   return max ( p0, p1 );
 }
 // CHECK: define noundef <2 x i16> @
 // CHECK: call <2 x i16> @llvm.smax.v2i16(
-// NO_HALF: define noundef <2 x i16> @"?test_max_short2@@YAT?$__vector@F$01@__clang@@T12@0@Z"(
-// NO_HALF: call <2 x i16> @llvm.smax.v2i16(
 int16_t2 test_max_short2 ( int16_t2 p0, int16_t2 p1 ) {
   return max ( p0, p1 );
 }
 // CHECK: define noundef <3 x i16> @
 // CHECK: call <3 x i16> @llvm.smax.v3i16
-// NO_HALF: define noundef <3 x i16> @"?test_max_short3@@YAT?$__vector@F$02@__clang@@T12@0@Z"(
-// NO_HALF: call <3 x i16> @llvm.smax.v3i16(
 int16_t3 test_max_short3 ( int16_t3 p0, int16_t3 p1 ) {
   return max ( p0, p1 );
 }
 // CHECK: define noundef <4 x i16> @
 // CHECK: call <4 x i16> @llvm.smax.v4i16
-// NO_HALF: define noundef <4 x i16> @"?test_max_short4@@YAT?$__vector@F$03@__clang@@T12@0@Z"(
-// NO_HALF: call <4 x i16> @llvm.smax.v4i16(
 int16_t4 test_max_short4 ( int16_t4 p0, int16_t4 p1 ) {
   return max ( p0, p1 );
 }
 
 // CHECK: define noundef i16 @
 // CHECK: call i16 @llvm.umax.i16(
-// NO_HALF: define noundef i16 @"?test_max_ushort@@YAGGG@Z"(
-// NO_HALF: call i16 @llvm.umax.i16(
 uint16_t test_max_ushort ( uint16_t p0, uint16_t p1 ) {
   return max ( p0, p1 );
 }
 // CHECK: define noundef <2 x i16> @
 // CHECK: call <2 x i16> @llvm.umax.v2i16
-// NO_HALF: define noundef <2 x i16> @"?test_max_ushort2@@YAT?$__vector@G$01@__clang@@T12@0@Z"(
-// NO_HALF: call <2 x i16> @llvm.umax.v2i16(
 uint16_t2 test_max_ushort2 ( uint16_t2 p0, uint16_t2 p1 ) {
   return max ( p0, p1 );
 }
 // CHECK: define noundef <3 x i16> @
 // CHECK: call <3 x i16> @llvm.umax.v3i16
-// NO_HALF: define noundef <3 x i16> @"?test_max_ushort3@@YAT?$__vector@G$02@__clang@@T12@0@Z"(
-// NO_HALF: call <3 x i16> @llvm.umax.v3i16(
 uint16_t3 test_max_ushort3 ( uint16_t3 p0, uint16_t3 p1 ) {
   return max ( p0, p1 );
 }
 // CHECK: define noundef <4 x i16> @
 // CHECK: call <4 x i16> @llvm.umax.v4i16
-// NO_HALF: define noundef <4 x i16> @"?test_max_ushort4@@YAT?$__vector@G$03@__clang@@T12@0@Z"(
-// NO_HALF: call <4 x i16> @llvm.umax.v4i16(
 uint16_t4 test_max_ushort4 ( uint16_t4 p0, uint16_t4 p1 ) {
   return max ( p0, p1 );
 }
diff --git a/clang/test/CodeGenHLSL/builtins/min.hlsl b/clang/test/CodeGenHLSL/builtins/min.hlsl
index 09066971b3f5d0..743053cbdd2620 10...
[truncated]

dmpots
dmpots reviewed Feb 15, 2024
View reviewed changes
clang/lib/Headers/hlsl/hlsl_intrinsics.h

_HLSL_AVAILABILITY(shadermodel, 6.2)
_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Currently in dxc the16-bit types are only available when the -enable-16bit-types command line flag is provided. They are not available by default in SM 6.2+.

Is this availability macro making the 16-bit types always available when targeting SM 6.2+?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is complicated. half is a valid type regardless of whether or not 16bit types are enabled, and half is a distinctly different type when 16-bit types are enabled. What this does is put the availability annotation on the 16-bit half type's overloads, but not the 32-bit half type's overloads.

That way if you're writing a shader for SM 6.0, using 32-bit half you get don't get blocked by the SM 6.2 availability annotation.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah ok. So this isn't changing how the enabling of 16bit types are working. It just lets the 32-bit half overloads work in earlier shader models.

@llvm-beanz llvm-beanz merged commit e82659f into llvm:main Feb 15, 2024
9 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@bogner bogner bogner approved these changes

@dmpots dmpots dmpots approved these changes

@bob80905 bob80905 bob80905 approved these changes

Assignees
No one assigned
Labels
clang:headers Headers provided by Clang, e.g. for intrinsics clang Clang issues not falling into any other category HLSL HLSL Language Support
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

[HLSL] half overloads and typedefs are supported when 16-bit types are not enabled
5 participants
@llvm-beanz @llvmbot @bogner @dmpots @bob80905