Fix implicit conversion rank ordering #106811
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
DXC prefers dimension-preserving conversions over precision-losing conversions. This means a double4 -> float4 conversion is preferred over a double4 -> double3 or double4 -> double conversion.
llvmbot
added
clang
|
@llvm/pr-subscribers-hlsl Author: Chris B (llvm-beanz) ChangesDXC prefers dimension-preserving conversions over precision-losing conversions. This means a double4 -> float4 conversion is preferred over a double4 -> double3 or double4 -> double conversion. Full diff: https://github.com/llvm/llvm-project/pull/106811.diff 2 Files Affected:
diff --git a/clang/include/clang/Sema/Overload.h b/clang/include/clang/Sema/Overload.h
index d6a6cee62a7528..f63f48e0fa3048 100644
--- a/clang/include/clang/Sema/Overload.h
+++ b/clang/include/clang/Sema/Overload.h
@@ -225,12 +225,12 @@ class Sema;
/// HLSL Scalar Widening with promotion
ICR_HLSL_Scalar_Widening_Promotion,
- /// HLSL Matching Dimension Reduction
- ICR_HLSL_Dimension_Reduction,
-
/// Conversion
ICR_Conversion,
+ /// HLSL Matching Dimension Reduction
+ ICR_HLSL_Dimension_Reduction,
+
/// OpenCL Scalar Widening
ICR_OCL_Scalar_Widening,
diff --git a/clang/test/SemaHLSL/TruncationOverloadResolution.hlsl b/clang/test/SemaHLSL/TruncationOverloadResolution.hlsl
index f8cfe22372e885..2bcb367c5669a3 100644
--- a/clang/test/SemaHLSL/TruncationOverloadResolution.hlsl
+++ b/clang/test/SemaHLSL/TruncationOverloadResolution.hlsl
@@ -1,24 +1,16 @@
-// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -fnative-half-type -finclude-default-header -fsyntax-only %s -DERROR=1 -verify
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -fnative-half-type -finclude-default-header -fsyntax-only -Wconversion %s -DERROR=1 -verify
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -fnative-half-type -finclude-default-header -ast-dump %s | FileCheck %s
-// Case 1: Prefer exact-match truncation over conversion.
-void Half4Float4Double2(double2 D);
-void Half4Float4Double2(float4 D);
-void Half4Float4Double2(half4 D);
+// Case 1: Prefer conversion over exact match truncation.
void Half4Float2(float2 D);
void Half4Float2(half4 D);
void Case1(float4 F, double4 D) {
// CHECK: CallExpr {{.*}} 'void'
- // CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(double2)' <FunctionToPointerDecay>
- // CHECK-NEXT: DeclRefExpr {{.*}} 'void (double2)' lvalue Function {{.*}} 'Half4Float4Double2' 'void (double2)'
- Half4Float4Double2(D); // expected-warning{{implicit conversion truncates vector: 'double4' (aka 'vector<double, 4>') to 'vector<double, 2>' (vector of 2 'double' values)}}
-
- // CHECK: CallExpr {{.*}} 'void'
- // CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float2)' <FunctionToPointerDecay>
- // CHECK-NEXT: DeclRefExpr {{.*}} 'void (float2)' lvalue Function {{.*}} 'Half4Float2' 'void (float2)'
- Half4Float2(F); // expected-warning{{implicit conversion truncates vector: 'float4' (aka 'vector<float, 4>') to 'vector<float, 2>' (vector of 2 'float' values)}}
+ // CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(half4)' <FunctionToPointerDecay>
+ // CHECK-NEXT: DeclRefExpr {{.*}} 'void (half4)' lvalue Function {{.*}} 'Half4Float2' 'void (half4)'
+ Half4Float2(F); // expected-warning{{implicit conversion loses floating-point precision: 'float4' (aka 'vector<float, 4>') to 'vector<half, 4>' (vector of 4 'half' values)}}
}
// Case 2: Prefer promotions over conversions when truncating.
@@ -46,7 +38,13 @@ void Half2Half3(half2 D); // expected-note{{candidate function}} expected-note{{
void Double2Double3(double3 D); // expected-note{{candidate function}} expected-note{{candidate function}} expected-note{{candidate function}}
void Double2Double3(double2 D); // expected-note{{candidate function}} expected-note{{candidate function}} expected-note{{candidate function}}
+void Half4Float4Double2(double2 D);
+void Half4Float4Double2(float4 D); // expected-note{{candidate function}}
+void Half4Float4Double2(half4 D); // expected-note{{candidate function}}
+
void Case1(half4 H, float4 F, double4 D) {
+ Half4Float4Double2(D); // expected-error {{call to 'Half4Float4Double2' is ambiguous}}
+
Float2Double2(H); // expected-error {{call to 'Float2Double2' is ambiguous}}
Half2Float2(D); // expected-error {{call to 'Half2Float2' is ambiguous}}
@@ -55,8 +53,8 @@ void Case1(half4 H, float4 F, double4 D) {
Half2Half3(F); // expected-error {{call to 'Half2Half3' is ambiguous}}
Half2Half3(D); // expected-error {{call to 'Half2Half3' is ambiguous}}
Half2Half3(H.xyz);
- Half2Half3(F.xyz);
- Half2Half3(D.xyz);
+ Half2Half3(F.xyz); // expected-warning {{implicit conversion loses floating-point precision: 'vector<float, 3>' (vector of 3 'float' values) to 'vector<half, 3>' (vector of 3 'half' values)}}
+ Half2Half3(D.xyz); // expected-warning {{implicit conversion loses floating-point precision: 'vector<double, 3>' (vector of 3 'double' values) to 'vector<half, 3>' (vector of 3 'half' values)}}
Double2Double3(H); // expected-error {{call to 'Double2Double3' is ambiguous}}
Double2Double3(F); // expected-error {{call to 'Double2Double3' is ambiguous}}
|
|
@llvm/pr-subscribers-clang Author: Chris B (llvm-beanz) ChangesDXC prefers dimension-preserving conversions over precision-losing conversions. This means a double4 -> float4 conversion is preferred over a double4 -> double3 or double4 -> double conversion. Full diff: https://github.com/llvm/llvm-project/pull/106811.diff 2 Files Affected:
diff --git a/clang/include/clang/Sema/Overload.h b/clang/include/clang/Sema/Overload.h
index d6a6cee62a7528..f63f48e0fa3048 100644
--- a/clang/include/clang/Sema/Overload.h
+++ b/clang/include/clang/Sema/Overload.h
@@ -225,12 +225,12 @@ class Sema;
/// HLSL Scalar Widening with promotion
ICR_HLSL_Scalar_Widening_Promotion,
- /// HLSL Matching Dimension Reduction
- ICR_HLSL_Dimension_Reduction,
-
/// Conversion
ICR_Conversion,
+ /// HLSL Matching Dimension Reduction
+ ICR_HLSL_Dimension_Reduction,
+
/// OpenCL Scalar Widening
ICR_OCL_Scalar_Widening,
diff --git a/clang/test/SemaHLSL/TruncationOverloadResolution.hlsl b/clang/test/SemaHLSL/TruncationOverloadResolution.hlsl
index f8cfe22372e885..2bcb367c5669a3 100644
--- a/clang/test/SemaHLSL/TruncationOverloadResolution.hlsl
+++ b/clang/test/SemaHLSL/TruncationOverloadResolution.hlsl
@@ -1,24 +1,16 @@
-// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -fnative-half-type -finclude-default-header -fsyntax-only %s -DERROR=1 -verify
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -fnative-half-type -finclude-default-header -fsyntax-only -Wconversion %s -DERROR=1 -verify
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -fnative-half-type -finclude-default-header -ast-dump %s | FileCheck %s
-// Case 1: Prefer exact-match truncation over conversion.
-void Half4Float4Double2(double2 D);
-void Half4Float4Double2(float4 D);
-void Half4Float4Double2(half4 D);
+// Case 1: Prefer conversion over exact match truncation.
void Half4Float2(float2 D);
void Half4Float2(half4 D);
void Case1(float4 F, double4 D) {
// CHECK: CallExpr {{.*}} 'void'
- // CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(double2)' <FunctionToPointerDecay>
- // CHECK-NEXT: DeclRefExpr {{.*}} 'void (double2)' lvalue Function {{.*}} 'Half4Float4Double2' 'void (double2)'
- Half4Float4Double2(D); // expected-warning{{implicit conversion truncates vector: 'double4' (aka 'vector<double, 4>') to 'vector<double, 2>' (vector of 2 'double' values)}}
-
- // CHECK: CallExpr {{.*}} 'void'
- // CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float2)' <FunctionToPointerDecay>
- // CHECK-NEXT: DeclRefExpr {{.*}} 'void (float2)' lvalue Function {{.*}} 'Half4Float2' 'void (float2)'
- Half4Float2(F); // expected-warning{{implicit conversion truncates vector: 'float4' (aka 'vector<float, 4>') to 'vector<float, 2>' (vector of 2 'float' values)}}
+ // CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(half4)' <FunctionToPointerDecay>
+ // CHECK-NEXT: DeclRefExpr {{.*}} 'void (half4)' lvalue Function {{.*}} 'Half4Float2' 'void (half4)'
+ Half4Float2(F); // expected-warning{{implicit conversion loses floating-point precision: 'float4' (aka 'vector<float, 4>') to 'vector<half, 4>' (vector of 4 'half' values)}}
}
// Case 2: Prefer promotions over conversions when truncating.
@@ -46,7 +38,13 @@ void Half2Half3(half2 D); // expected-note{{candidate function}} expected-note{{
void Double2Double3(double3 D); // expected-note{{candidate function}} expected-note{{candidate function}} expected-note{{candidate function}}
void Double2Double3(double2 D); // expected-note{{candidate function}} expected-note{{candidate function}} expected-note{{candidate function}}
+void Half4Float4Double2(double2 D);
+void Half4Float4Double2(float4 D); // expected-note{{candidate function}}
+void Half4Float4Double2(half4 D); // expected-note{{candidate function}}
+
void Case1(half4 H, float4 F, double4 D) {
+ Half4Float4Double2(D); // expected-error {{call to 'Half4Float4Double2' is ambiguous}}
+
Float2Double2(H); // expected-error {{call to 'Float2Double2' is ambiguous}}
Half2Float2(D); // expected-error {{call to 'Half2Float2' is ambiguous}}
@@ -55,8 +53,8 @@ void Case1(half4 H, float4 F, double4 D) {
Half2Half3(F); // expected-error {{call to 'Half2Half3' is ambiguous}}
Half2Half3(D); // expected-error {{call to 'Half2Half3' is ambiguous}}
Half2Half3(H.xyz);
- Half2Half3(F.xyz);
- Half2Half3(D.xyz);
+ Half2Half3(F.xyz); // expected-warning {{implicit conversion loses floating-point precision: 'vector<float, 3>' (vector of 3 'float' values) to 'vector<half, 3>' (vector of 3 'half' values)}}
+ Half2Half3(D.xyz); // expected-warning {{implicit conversion loses floating-point precision: 'vector<double, 3>' (vector of 3 'double' values) to 'vector<half, 3>' (vector of 3 'half' values)}}
Double2Double3(H); // expected-error {{call to 'Double2Double3' is ambiguous}}
Double2Double3(F); // expected-error {{call to 'Double2Double3' is ambiguous}}
|
damyanp
approved these changes
Aug 31, 2024
This is an NFC change. For a given argument in a call there cannot be a valid implicit conversion seqeunce of any of the scalar widening ranks and a conversion sequence for a different overload of dimension reduction because the widening ranks require the argument to be a scalar and the reduction ranks require the argument to be a vector. Changing the order in this change does not impact behavior in any way, but it does align the source with the spec's language and ordering which makes it cleaner to read.
bogner
approved these changes
Sep 6, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
DXC prefers dimension-preserving conversions over precision-losing conversions. This means a double4 -> float4 conversion is preferred over a double4 -> double3 or double4 -> double conversion.