mirror of
https://github.com/llvm/llvm-project.git
synced 2025-04-17 09:56:40 +00:00
d91ff3f240
This PR reworks HLSL's implicit conversion sequences. Initially I was seeking to match DXC's behavior more closely, but that was leading to a pile of special case rules to tie-break ambiguous cases that should really be left as ambiguous. We've decided that we're going to break compatibility with DXC here, and we may port this new behavior over to DXC instead. This change is a bit closer to C++'s overload resolution rules, but it does have a bit of nuance around how dimension adjustment conversions are ranked. Conversion sequence ranks for HLSL are: * Exact match * Scalar Widening (i.e. splat) * Promotion * Scalar Widening with Promotion * Conversion * Scalar Widening with Conversion * Dimension Reduction (i.e. truncation) * Dimension Reduction with Promotion * Dimension Reduction with Conversion In this implementation I've folded the disambiguation into the conversion sequence ranks which does add some complexity as compared to C++, however this avoids needing to add special casing in `CompareStandardConversionSequences`. I believe the added conversion rank values provide a simpler approach, but feedback is appreciated. The HLSL language spec updates are in the PR here: https://github.com/microsoft/hlsl-specs/pull/261
167 lines
7.7 KiB
HLSL
167 lines
7.7 KiB
HLSL
// 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 -ast-dump %s | FileCheck %s
|
|
|
|
|
|
// Case 1: Prioritize splat without conversion over conversion. In this case the
|
|
// called functions have valid overloads for each type, however one of the
|
|
// overloads is a vector rather than scalar. Each call should resolve to the
|
|
// same type, and the vector should splat.
|
|
void HalfFloatDoubleV(double2 D);
|
|
void HalfFloatDoubleV(float F);
|
|
void HalfFloatDoubleV(half H);
|
|
|
|
void HalfFloatVDouble(double D);
|
|
void HalfFloatVDouble(float2 F);
|
|
void HalfFloatVDouble(half H);
|
|
|
|
void HalfVFloatDouble(double D);
|
|
void HalfVFloatDouble(float F);
|
|
void HalfVFloatDouble(half2 H);
|
|
|
|
|
|
// CHECK-LABEL: FunctionDecl {{.*}} Case1 'void (half, float, double)'
|
|
void Case1(half H, float F, double D) {
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(half)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (half)' lvalue Function {{.*}} 'HalfFloatDoubleV' 'void (half)'
|
|
HalfFloatDoubleV(H);
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (float)' lvalue Function {{.*}} 'HalfFloatDoubleV' 'void (float)'
|
|
HalfFloatDoubleV(F);
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(double2)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (double2)' lvalue Function {{.*}} 'HalfFloatDoubleV' 'void (double2)'
|
|
HalfFloatDoubleV(D);
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(half)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (half)' lvalue Function {{.*}} 'HalfFloatVDouble' 'void (half)'
|
|
HalfFloatVDouble(H);
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float2)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (float2)' lvalue Function {{.*}} 'HalfFloatVDouble' 'void (float2)'
|
|
HalfFloatVDouble(F);
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(double)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (double)' lvalue Function {{.*}} 'HalfFloatVDouble' 'void (double)'
|
|
HalfFloatVDouble(D);
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(half2)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (half2)' lvalue Function {{.*}} 'HalfVFloatDouble' 'void (half2)'
|
|
HalfVFloatDouble(H);
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (float)' lvalue Function {{.*}} 'HalfVFloatDouble' 'void (float)'
|
|
HalfVFloatDouble(F);
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(double)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (double)' lvalue Function {{.*}} 'HalfVFloatDouble' 'void (double)'
|
|
HalfVFloatDouble(D);
|
|
}
|
|
|
|
// Case 2: Prefer splat+promotion over conversion. In this case the overloads
|
|
// require a splat+promotion or a conversion. The call will resolve to the
|
|
// splat+promotion.
|
|
void HalfDoubleV(double2 D);
|
|
void HalfDoubleV(half H);
|
|
|
|
// CHECK-LABEL: FunctionDecl {{.*}} Case2 'void (float)'
|
|
void Case2(float F) {
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(double2)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (double2)' lvalue Function {{.*}} 'HalfDoubleV' 'void (double2)'
|
|
HalfDoubleV(F);
|
|
}
|
|
|
|
// Case 3: Prefer promotion or conversion without splat over the splat. In this
|
|
// case the scalar value will overload to the scalar function.
|
|
void DoubleV(double D);
|
|
void DoubleV(double2 V);
|
|
|
|
void HalfV(half D);
|
|
void HalfV(half2 V);
|
|
|
|
// CHECK-LABEL: FunctionDecl {{.*}} Case3 'void (float)'
|
|
void Case3(float F) {
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(double)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (double)' lvalue Function {{.*}} 'DoubleV' 'void (double)'
|
|
DoubleV(F);
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(half)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (half)' lvalue Function {{.*}} 'HalfV' 'void (half)'
|
|
HalfV(F);
|
|
}
|
|
|
|
#if ERROR
|
|
// Case 4: It is ambiguous to resolve two splat+conversion or splat+promotion
|
|
// functions. In all the calls below an error occurs.
|
|
void FloatVDoubleV(float2 F); // expected-note {{candidate function}}
|
|
void FloatVDoubleV(double2 D); // expected-note {{candidate function}}
|
|
|
|
void HalfVFloatV(half2 H); // expected-note {{candidate function}}
|
|
void HalfVFloatV(float2 F); // expected-note {{candidate function}}
|
|
|
|
void Case4(half H, double D) {
|
|
FloatVDoubleV(H); // expected-error {{call to 'FloatVDoubleV' is ambiguous}}
|
|
|
|
HalfVFloatV(D); // expected-error {{call to 'HalfVFloatV' is ambiguous}}
|
|
}
|
|
|
|
// Case 5: It is ambiguous to resolve two splats of different lengths.
|
|
void FloatV(float2 V); // expected-note {{candidate function}} expected-note {{candidate function}} expected-note {{candidate function}}
|
|
void FloatV(float4 V); // expected-note {{candidate function}} expected-note {{candidate function}} expected-note {{candidate function}}
|
|
|
|
void Case5(half H, float F, double D) {
|
|
FloatV(H); // expected-error {{call to 'FloatV' is ambiguous}}
|
|
FloatV(F); // expected-error {{call to 'FloatV' is ambiguous}}
|
|
FloatV(D); // expected-error {{call to 'FloatV' is ambiguous}}
|
|
}
|
|
#endif
|
|
|
|
// Case 5: Vectors truncate or match, but don't extend.
|
|
void FloatV24(float2 V);
|
|
void FloatV24(float4 V);
|
|
|
|
// CHECK-LABEL: FunctionDecl {{.*}} Case5 'void (half3, float3, double3, half4, float4, double4)'
|
|
void Case5(half3 H3, float3 F3, double3 D3, half4 H4, float4 F4, double4 D4) {
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float2)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (float2)' lvalue Function {{.*}} 'FloatV24' 'void (float2)'
|
|
FloatV24(H3); // expected-warning{{implicit conversion truncates vector: 'half3' (aka 'vector<half, 3>') to 'vector<float, 2>' (vector of 2 'float' values)}}
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float2)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (float2)' lvalue Function {{.*}} 'FloatV24' 'void (float2)'
|
|
FloatV24(F3); // expected-warning{{implicit conversion truncates vector: 'float3' (aka 'vector<float, 3>') to 'vector<float, 2>' (vector of 2 'float' values)}}
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float2)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (float2)' lvalue Function {{.*}} 'FloatV24' 'void (float2)'
|
|
FloatV24(D3); // expected-warning{{implicit conversion truncates vector: 'double3' (aka 'vector<double, 3>') to 'vector<float, 2>' (vector of 2 'float' values)}}
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float4)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (float4)' lvalue Function {{.*}} 'FloatV24' 'void (float4)'
|
|
FloatV24(H4);
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float4)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (float4)' lvalue Function {{.*}} 'FloatV24' 'void (float4)'
|
|
FloatV24(F4);
|
|
|
|
// CHECK: CallExpr {{.*}} 'void'
|
|
// CHECK-NEXT: ImplicitCastExpr {{.*}} 'void (*)(float4)' <FunctionToPointerDecay>
|
|
// CHECK-NEXT: DeclRefExpr {{.*}} 'void (float4)' lvalue Function {{.*}} 'FloatV24' 'void (float4)'
|
|
FloatV24(D4);
|
|
}
|