[clang] 25713ed - [HLSL] Add additional overloads for min and max to allow for mixed scalar and vector arguments (#129334)
via cfe-commits
cfe-commits at lists.llvm.org
Mon Mar 3 11:30:02 PST 2025
Author: Sarah Spall
Date: 2025-03-03T11:29:58-08:00
New Revision: 25713ed85fd327f4733b2ac6083c23464aa9c646
URL: https://github.com/llvm/llvm-project/commit/25713ed85fd327f4733b2ac6083c23464aa9c646
DIFF: https://github.com/llvm/llvm-project/commit/25713ed85fd327f4733b2ac6083c23464aa9c646.diff
LOG: [HLSL] Add additional overloads for min and max to allow for mixed scalar and vector arguments (#129334)
Add additional overloads for min and max to support
min(vector<T,N>, T) and min(T, vector<T,N>)
max(vector<T,N>, T) and max(T, vector<T,N>)
Add tests
Closes #128231
Added:
Modified:
clang/lib/Headers/hlsl/hlsl_intrinsics.h
clang/test/CodeGenHLSL/builtins/max.hlsl
clang/test/CodeGenHLSL/builtins/min.hlsl
Removed:
################################################################################
diff --git a/clang/lib/Headers/hlsl/hlsl_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
index ed008eeb04ba8..734fc8d29b28f 100644
--- a/clang/lib/Headers/hlsl/hlsl_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
@@ -37,6 +37,26 @@ namespace hlsl {
#define _HLSL_16BIT_AVAILABILITY_STAGE(environment, version, stage)
#endif
+#define GEN_VEC_SCALAR_OVERLOADS(FUNC_NAME, BASE_TYPE, AVAIL) \
+ GEN_BOTH_OVERLOADS(FUNC_NAME, BASE_TYPE, BASE_TYPE##2, AVAIL) \
+ GEN_BOTH_OVERLOADS(FUNC_NAME, BASE_TYPE, BASE_TYPE##3, AVAIL) \
+ GEN_BOTH_OVERLOADS(FUNC_NAME, BASE_TYPE, BASE_TYPE##4, AVAIL)
+
+#define GEN_BOTH_OVERLOADS(FUNC_NAME, BASE_TYPE, VECTOR_TYPE, AVAIL) \
+ IF_TRUE_##AVAIL( \
+ _HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)) constexpr VECTOR_TYPE \
+ FUNC_NAME(VECTOR_TYPE p0, BASE_TYPE p1) { \
+ return __builtin_elementwise_##FUNC_NAME(p0, (VECTOR_TYPE)p1); \
+ } \
+ IF_TRUE_##AVAIL( \
+ _HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)) constexpr VECTOR_TYPE \
+ FUNC_NAME(BASE_TYPE p0, VECTOR_TYPE p1) { \
+ return __builtin_elementwise_##FUNC_NAME((VECTOR_TYPE)p0, p1); \
+ }
+
+#define IF_TRUE_0(EXPR)
+#define IF_TRUE_1(EXPR) EXPR
+
//===----------------------------------------------------------------------===//
// abs builtins
//===----------------------------------------------------------------------===//
@@ -1700,6 +1720,7 @@ half3 max(half3, half3);
_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
half4 max(half4, half4);
+GEN_VEC_SCALAR_OVERLOADS(max, half, 1)
#ifdef __HLSL_ENABLE_16_BIT
_HLSL_AVAILABILITY(shadermodel, 6.2)
@@ -1714,6 +1735,7 @@ int16_t3 max(int16_t3, int16_t3);
_HLSL_AVAILABILITY(shadermodel, 6.2)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
int16_t4 max(int16_t4, int16_t4);
+GEN_VEC_SCALAR_OVERLOADS(max, int16_t, 1)
_HLSL_AVAILABILITY(shadermodel, 6.2)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
@@ -1727,6 +1749,7 @@ uint16_t3 max(uint16_t3, uint16_t3);
_HLSL_AVAILABILITY(shadermodel, 6.2)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
uint16_t4 max(uint16_t4, uint16_t4);
+GEN_VEC_SCALAR_OVERLOADS(max, uint16_t, 1)
#endif
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
@@ -1737,6 +1760,7 @@ _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
int3 max(int3, int3);
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
int4 max(int4, int4);
+GEN_VEC_SCALAR_OVERLOADS(max, int, 0)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
uint max(uint, uint);
@@ -1746,6 +1770,7 @@ _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
uint3 max(uint3, uint3);
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
uint4 max(uint4, uint4);
+GEN_VEC_SCALAR_OVERLOADS(max, uint, 0)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
int64_t max(int64_t, int64_t);
@@ -1755,6 +1780,7 @@ _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
int64_t3 max(int64_t3, int64_t3);
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
int64_t4 max(int64_t4, int64_t4);
+GEN_VEC_SCALAR_OVERLOADS(max, int64_t, 0)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
uint64_t max(uint64_t, uint64_t);
@@ -1764,6 +1790,7 @@ _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
uint64_t3 max(uint64_t3, uint64_t3);
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
uint64_t4 max(uint64_t4, uint64_t4);
+GEN_VEC_SCALAR_OVERLOADS(max, uint64_t, 0)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
float max(float, float);
@@ -1773,6 +1800,7 @@ _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
float3 max(float3, float3);
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
float4 max(float4, float4);
+GEN_VEC_SCALAR_OVERLOADS(max, float, 0)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
double max(double, double);
@@ -1782,6 +1810,7 @@ _HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
double3 max(double3, double3);
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_max)
double4 max(double4, double4);
+GEN_VEC_SCALAR_OVERLOADS(max, double, 0)
//===----------------------------------------------------------------------===//
// min builtins
@@ -1804,6 +1833,7 @@ half3 min(half3, half3);
_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
half4 min(half4, half4);
+GEN_VEC_SCALAR_OVERLOADS(min, half, 1)
#ifdef __HLSL_ENABLE_16_BIT
_HLSL_AVAILABILITY(shadermodel, 6.2)
@@ -1818,6 +1848,7 @@ int16_t3 min(int16_t3, int16_t3);
_HLSL_AVAILABILITY(shadermodel, 6.2)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
int16_t4 min(int16_t4, int16_t4);
+GEN_VEC_SCALAR_OVERLOADS(min, int16_t, 1)
_HLSL_AVAILABILITY(shadermodel, 6.2)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
@@ -1831,6 +1862,7 @@ uint16_t3 min(uint16_t3, uint16_t3);
_HLSL_AVAILABILITY(shadermodel, 6.2)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
uint16_t4 min(uint16_t4, uint16_t4);
+GEN_VEC_SCALAR_OVERLOADS(min, uint16_t, 1)
#endif
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
@@ -1841,6 +1873,7 @@ _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
int3 min(int3, int3);
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
int4 min(int4, int4);
+GEN_VEC_SCALAR_OVERLOADS(min, int, 0)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
uint min(uint, uint);
@@ -1850,6 +1883,7 @@ _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
uint3 min(uint3, uint3);
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
uint4 min(uint4, uint4);
+GEN_VEC_SCALAR_OVERLOADS(min, uint, 0)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
float min(float, float);
@@ -1859,6 +1893,7 @@ _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
float3 min(float3, float3);
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
float4 min(float4, float4);
+GEN_VEC_SCALAR_OVERLOADS(min, float, 0)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
int64_t min(int64_t, int64_t);
@@ -1868,6 +1903,7 @@ _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
int64_t3 min(int64_t3, int64_t3);
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
int64_t4 min(int64_t4, int64_t4);
+GEN_VEC_SCALAR_OVERLOADS(min, int64_t, 0)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
uint64_t min(uint64_t, uint64_t);
@@ -1877,6 +1913,7 @@ _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
uint64_t3 min(uint64_t3, uint64_t3);
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
uint64_t4 min(uint64_t4, uint64_t4);
+GEN_VEC_SCALAR_OVERLOADS(min, uint64_t, 0)
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
double min(double, double);
@@ -1886,6 +1923,7 @@ _HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
double3 min(double3, double3);
_HLSL_BUILTIN_ALIAS(__builtin_elementwise_min)
double4 min(double4, double4);
+GEN_VEC_SCALAR_OVERLOADS(min, double, 0)
//===----------------------------------------------------------------------===//
// normalize builtins
diff --git a/clang/test/CodeGenHLSL/builtins/max.hlsl b/clang/test/CodeGenHLSL/builtins/max.hlsl
index 0b767335556ee..6b5fb6ae59534 100644
--- a/clang/test/CodeGenHLSL/builtins/max.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/max.hlsl
@@ -18,6 +18,9 @@ int16_t3 test_max_short3(int16_t3 p0, int16_t3 p1) { return max(p0, p1); }
// NATIVE_HALF-LABEL: define noundef <4 x i16> @_Z15test_max_short4
// NATIVE_HALF: call <4 x i16> @llvm.smax.v4i16
int16_t4 test_max_short4(int16_t4 p0, int16_t4 p1) { return max(p0, p1); }
+// NATIVE_HALF-LABEL: define noundef <4 x i16> {{.*}}test_max_short4_mismatch
+// NATIVE_HALF: call <4 x i16> @llvm.smax.v4i16
+int16_t4 test_max_short4_mismatch(int16_t4 p0, int16_t p1) { return max(p0, p1); }
// NATIVE_HALF-LABEL: define noundef i16 @_Z15test_max_ushort
// NATIVE_HALF: call i16 @llvm.umax.i16(
@@ -31,6 +34,9 @@ uint16_t3 test_max_ushort3(uint16_t3 p0, uint16_t3 p1) { return max(p0, p1); }
// NATIVE_HALF-LABEL: define noundef <4 x i16> @_Z16test_max_ushort4
// NATIVE_HALF: call <4 x i16> @llvm.umax.v4i16
uint16_t4 test_max_ushort4(uint16_t4 p0, uint16_t4 p1) { return max(p0, p1); }
+// NATIVE_HALF-LABEL: define noundef <4 x i16> {{.*}}test_max_ushort4_mismatch
+// NATIVE_HALF: call <4 x i16> @llvm.umax.v4i16
+uint16_t4 test_max_ushort4_mismatch(uint16_t4 p0, uint16_t p1) { return max(p0, p1); }
#endif
// CHECK-LABEL: define noundef i32 @_Z12test_max_int
@@ -45,6 +51,9 @@ int3 test_max_int3(int3 p0, int3 p1) { return max(p0, p1); }
// CHECK-LABEL: define noundef <4 x i32> @_Z13test_max_int4
// CHECK: call <4 x i32> @llvm.smax.v4i32
int4 test_max_int4(int4 p0, int4 p1) { return max(p0, p1); }
+// CHECK-LABEL: define noundef <4 x i32> {{.*}}test_max_int4_mismatch
+// CHECK: call <4 x i32> @llvm.smax.v4i32
+int4 test_max_int4_mismatch(int4 p0, int p1) { return max(p0, p1); }
// CHECK-LABEL: define noundef i32 @_Z13test_max_uint
// CHECK: call i32 @llvm.umax.i32(
@@ -58,6 +67,9 @@ uint3 test_max_uint3(uint3 p0, uint3 p1) { return max(p0, p1); }
// CHECK-LABEL: define noundef <4 x i32> @_Z14test_max_uint4
// CHECK: call <4 x i32> @llvm.umax.v4i32
uint4 test_max_uint4(uint4 p0, uint4 p1) { return max(p0, p1); }
+// CHECK-LABEL: define noundef <4 x i32> {{.*}}test_max_uint4_mismatch
+// CHECK: call <4 x i32> @llvm.umax.v4i32
+uint4 test_max_uint4_mismatch(uint4 p0, uint p1) { return max(p0, p1); }
// CHECK-LABEL: define noundef i64 @_Z13test_max_long
// CHECK: call i64 @llvm.smax.i64(
@@ -71,6 +83,9 @@ int64_t3 test_max_long3(int64_t3 p0, int64_t3 p1) { return max(p0, p1); }
// CHECK-LABEL: define noundef <4 x i64> @_Z14test_max_long4
// CHECK: call <4 x i64> @llvm.smax.v4i64
int64_t4 test_max_long4(int64_t4 p0, int64_t4 p1) { return max(p0, p1); }
+// CHECK-LABEL: define noundef <4 x i64> {{.*}}test_max_long4_mismatch
+// CHECK: call <4 x i64> @llvm.smax.v4i64
+int64_t4 test_max_long4_mismatch(int64_t4 p0, int64_t p1) { return max(p0, p1); }
// CHECK-LABEL: define noundef i64 @_Z14test_max_ulong
// CHECK: call i64 @llvm.umax.i64(
@@ -84,6 +99,9 @@ uint64_t3 test_max_ulong3(uint64_t3 p0, uint64_t3 p1) { return max(p0, p1); }
// CHECK-LABEL: define noundef <4 x i64> @_Z15test_max_ulong4
// CHECK: call <4 x i64> @llvm.umax.v4i64
uint64_t4 test_max_ulong4(uint64_t4 p0, uint64_t4 p1) { return max(p0, p1); }
+// CHECK-LABEL: define noundef <4 x i64> {{.*}}test_max_ulong4_mismatch
+// CHECK: call <4 x i64> @llvm.umax.v4i64
+uint64_t4 test_max_ulong4_mismatch(uint64_t4 p0, uint64_t p1) { return max(p0, p1); }
// NATIVE_HALF-LABEL: define noundef nofpclass(nan inf) half @_Z13test_max_half
// NATIVE_HALF: call reassoc nnan ninf nsz arcp afn half @llvm.maxnum.f16(
@@ -105,6 +123,11 @@ half3 test_max_half3(half3 p0, half3 p1) { return max(p0, p1); }
// NO_HALF-LABEL: define noundef nofpclass(nan inf) <4 x float> @_Z14test_max_half4
// NO_HALF: call reassoc nnan ninf nsz arcp afn <4 x float> @llvm.maxnum.v4f32(
half4 test_max_half4(half4 p0, half4 p1) { return max(p0, p1); }
+// NATIVE_HALF-LABEL: define noundef nofpclass(nan inf) <4 x half> {{.*}}test_max_half4_mismatch
+// NATIVE_HALF: call reassoc nnan ninf nsz arcp afn <4 x half> @llvm.maxnum.v4f16
+// NO_HALF-LABEL: define noundef nofpclass(nan inf) <4 x float> {{.*}}test_max_half4_mismatch
+// NO_HALF: call reassoc nnan ninf nsz arcp afn <4 x float> @llvm.maxnum.v4f32(
+half4 test_max_half4_mismatch(half4 p0, half p1) { return max(p0, p1); }
// CHECK-LABEL: define noundef nofpclass(nan inf) float @_Z14test_max_float
// CHECK: call reassoc nnan ninf nsz arcp afn float @llvm.maxnum.f32(
@@ -118,6 +141,9 @@ float3 test_max_float3(float3 p0, float3 p1) { return max(p0, p1); }
// CHECK-LABEL: define noundef nofpclass(nan inf) <4 x float> @_Z15test_max_float4
// CHECK: call reassoc nnan ninf nsz arcp afn <4 x float> @llvm.maxnum.v4f32
float4 test_max_float4(float4 p0, float4 p1) { return max(p0, p1); }
+// CHECK-LABEL: define noundef nofpclass(nan inf) <4 x float> {{.*}}test_max_float4_mismatch
+// CHECK: call reassoc nnan ninf nsz arcp afn <4 x float> @llvm.maxnum.v4f32
+float4 test_max_float4_mismatch(float4 p0, float p1) { return max(p0, p1); }
// CHECK-LABEL: define noundef nofpclass(nan inf) double @_Z15test_max_double
// CHECK: call reassoc nnan ninf nsz arcp afn double @llvm.maxnum.f64(
@@ -131,3 +157,9 @@ double3 test_max_double3(double3 p0, double3 p1) { return max(p0, p1); }
// CHECK-LABEL: define noundef nofpclass(nan inf) <4 x double> @_Z16test_max_double4
// CHECK: call reassoc nnan ninf nsz arcp afn <4 x double> @llvm.maxnum.v4f64
double4 test_max_double4(double4 p0, double4 p1) { return max(p0, p1); }
+// CHECK-LABEL: define noundef nofpclass(nan inf) <4 x double> {{.*}}test_max_double4_mismatch
+// CHECK: call reassoc nnan ninf nsz arcp afn <4 x double> @llvm.maxnum.v4f64
+double4 test_max_double4_mismatch(double4 p0, double p1) { return max(p0, p1); }
+// CHECK-LABEL: define noundef nofpclass(nan inf) <4 x double> {{.*}}test_max_double4_mismatch2
+// CHECK: call reassoc nnan ninf nsz arcp afn <4 x double> @llvm.maxnum.v4f64
+double4 test_max_double4_mismatch2(double4 p0, double p1) { return max(p1, p0); }
diff --git a/clang/test/CodeGenHLSL/builtins/min.hlsl b/clang/test/CodeGenHLSL/builtins/min.hlsl
index a352c72f29542..551db52878e37 100644
--- a/clang/test/CodeGenHLSL/builtins/min.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/min.hlsl
@@ -18,6 +18,9 @@ int16_t3 test_min_short3(int16_t3 p0, int16_t3 p1) { return min(p0, p1); }
// NATIVE_HALF-LABEL: define noundef <4 x i16> @_Z15test_min_short4
// NATIVE_HALF: call <4 x i16> @llvm.smin.v4i16
int16_t4 test_min_short4(int16_t4 p0, int16_t4 p1) { return min(p0, p1); }
+// NATIVE_HALF-LABEL: define noundef <4 x i16> {{.*}}test_min_short4_mismatch
+// NATIVE_HALF: call <4 x i16> @llvm.smin.v4i16
+int16_t4 test_min_short4_mismatch(int16_t4 p0, int16_t p1) { return min(p0, p1); }
// NATIVE_HALF-LABEL: define noundef i16 @_Z15test_min_ushort
// NATIVE_HALF: call i16 @llvm.umin.i16(
@@ -31,6 +34,9 @@ uint16_t3 test_min_ushort3(uint16_t3 p0, uint16_t3 p1) { return min(p0, p1); }
// NATIVE_HALF-LABEL: define noundef <4 x i16> @_Z16test_min_ushort4
// NATIVE_HALF: call <4 x i16> @llvm.umin.v4i16
uint16_t4 test_min_ushort4(uint16_t4 p0, uint16_t4 p1) { return min(p0, p1); }
+// NATIVE_HALF-LABEL: define noundef <4 x i16> {{.*}}test_min_ushort4_mismatch
+// NATIVE_HALF: call <4 x i16> @llvm.umin.v4i16
+uint16_t4 test_min_ushort4_mismatch(uint16_t4 p0, uint16_t p1) { return min(p0, p1); }
#endif
// CHECK-LABEL: define noundef i32 @_Z12test_min_int
@@ -45,6 +51,9 @@ int3 test_min_int3(int3 p0, int3 p1) { return min(p0, p1); }
// CHECK-LABEL: define noundef <4 x i32> @_Z13test_min_int4
// CHECK: call <4 x i32> @llvm.smin.v4i32
int4 test_min_int4(int4 p0, int4 p1) { return min(p0, p1); }
+// CHECK-LABEL: define noundef <4 x i32> {{.*}}test_min_int4_mismatch
+// CHECK: call <4 x i32> @llvm.smin.v4i32
+int4 test_min_int4_mismatch(int4 p0, int p1) { return min(p0, p1); }
// CHECK-LABEL: define noundef i32 @_Z13test_min_uint
// CHECK: call i32 @llvm.umin.i32(
@@ -58,6 +67,9 @@ uint3 test_min_uint3(uint3 p0, uint3 p1) { return min(p0, p1); }
// CHECK-LABEL: define noundef <4 x i32> @_Z14test_min_uint4
// CHECK: call <4 x i32> @llvm.umin.v4i32
uint4 test_min_uint4(uint4 p0, uint4 p1) { return min(p0, p1); }
+// CHECK-LABEL: define noundef <4 x i32> {{.*}}test_min_uint4_mismatch
+// CHECK: call <4 x i32> @llvm.umin.v4i32
+uint4 test_min_uint4_mismatch(uint4 p0, uint p1) { return min(p0, p1); }
// CHECK-LABEL: define noundef i64 @_Z13test_min_long
// CHECK: call i64 @llvm.smin.i64(
@@ -71,6 +83,9 @@ int64_t3 test_min_long3(int64_t3 p0, int64_t3 p1) { return min(p0, p1); }
// CHECK-LABEL: define noundef <4 x i64> @_Z14test_min_long4
// CHECK: call <4 x i64> @llvm.smin.v4i64
int64_t4 test_min_long4(int64_t4 p0, int64_t4 p1) { return min(p0, p1); }
+// CHECK-LABEL: define noundef <4 x i64> {{.*}}test_min_long4_mismatch
+// CHECK: call <4 x i64> @llvm.smin.v4i64
+int64_t4 test_min_long4_mismatch(int64_t4 p0, int64_t p1) { return min(p0, p1); }
// CHECK-LABEL: define noundef i64 @_Z14test_min_ulong
// CHECK: call i64 @llvm.umin.i64(
@@ -84,6 +99,9 @@ uint64_t3 test_min_ulong3(uint64_t3 p0, uint64_t3 p1) { return min(p0, p1); }
// CHECK-LABEL: define noundef <4 x i64> @_Z15test_min_ulong4
// CHECK: call <4 x i64> @llvm.umin.v4i64
uint64_t4 test_min_ulong4(uint64_t4 p0, uint64_t4 p1) { return min(p0, p1); }
+// CHECK-LABEL: define noundef <4 x i64> {{.*}}test_min_ulong4_mismatch
+// CHECK: call <4 x i64> @llvm.umin.v4i64
+uint64_t4 test_min_ulong4_mismatch(uint64_t4 p0, uint64_t p1) { return min(p0, p1); }
// NATIVE_HALF-LABEL: define noundef nofpclass(nan inf) half @_Z13test_min_half
// NATIVE_HALF: call reassoc nnan ninf nsz arcp afn half @llvm.minnum.f16(
@@ -105,6 +123,11 @@ half3 test_min_half3(half3 p0, half3 p1) { return min(p0, p1); }
// NO_HALF-LABEL: define noundef nofpclass(nan inf) <4 x float> @_Z14test_min_half4
// NO_HALF: call reassoc nnan ninf nsz arcp afn <4 x float> @llvm.minnum.v4f32(
half4 test_min_half4(half4 p0, half4 p1) { return min(p0, p1); }
+// NATIVE_HALF-LABEL: define noundef nofpclass(nan inf) <4 x half> {{.*}}test_min_half4_mismatch
+// NATIVE_HALF: call reassoc nnan ninf nsz arcp afn <4 x half> @llvm.minnum.v4f16
+// NO_HALF-LABEL: define noundef nofpclass(nan inf) <4 x float> {{.*}}test_min_half4_mismatch
+// NO_HALF: call reassoc nnan ninf nsz arcp afn <4 x float> @llvm.minnum.v4f32(
+half4 test_min_half4_mismatch(half4 p0, half p1) { return min(p0, p1); }
// CHECK-LABEL: define noundef nofpclass(nan inf) float @_Z14test_min_float
// CHECK: call reassoc nnan ninf nsz arcp afn float @llvm.minnum.f32(
@@ -118,6 +141,9 @@ float3 test_min_float3(float3 p0, float3 p1) { return min(p0, p1); }
// CHECK-LABEL: define noundef nofpclass(nan inf) <4 x float> @_Z15test_min_float4
// CHECK: call reassoc nnan ninf nsz arcp afn <4 x float> @llvm.minnum.v4f32
float4 test_min_float4(float4 p0, float4 p1) { return min(p0, p1); }
+// CHECK-LABEL: define noundef nofpclass(nan inf) <4 x float> {{.*}}test_min_float4_mismatch
+// CHECK: call reassoc nnan ninf nsz arcp afn <4 x float> @llvm.minnum.v4f32
+float4 test_min_float4_mismatch(float4 p0, float p1) { return min(p0, p1); }
// CHECK-LABEL: define noundef nofpclass(nan inf) double @_Z15test_min_double
// CHECK: call reassoc nnan ninf nsz arcp afn double @llvm.minnum.f64(
@@ -131,3 +157,9 @@ double3 test_min_double3(double3 p0, double3 p1) { return min(p0, p1); }
// CHECK-LABEL: define noundef nofpclass(nan inf) <4 x double> @_Z16test_min_double4
// CHECK: call reassoc nnan ninf nsz arcp afn <4 x double> @llvm.minnum.v4f64
double4 test_min_double4(double4 p0, double4 p1) { return min(p0, p1); }
+// CHECK-LABEL: define noundef nofpclass(nan inf) <4 x double> {{.*}}test_min_double4_mismatch
+// CHECK: call reassoc nnan ninf nsz arcp afn <4 x double> @llvm.minnum.v4f64
+double4 test_min_double4_mismatch(double4 p0, double p1) { return min(p0, p1); }
+// CHECK-LABEL: define noundef nofpclass(nan inf) <4 x double> {{.*}}test_min_double4_mismatch2
+// CHECK: call reassoc nnan ninf nsz arcp afn <4 x double> @llvm.minnum.v4f64
+double4 test_min_double4_mismatch2(double4 p0, double p1) { return min(p1, p0); }
More information about the cfe-commits
mailing list