34 lines
1.1 KiB
Plaintext
34 lines
1.1 KiB
Plaintext
// RUN: %{ispc} --target=avx2-i32x8 --nowrap --x86-asm-syntax=intel --emit-asm -o - %s 2>&1 | FileCheck %s
|
|
// RUN: %{ispc} --target=avx512skx-x8 --nowrap --x86-asm-syntax=intel --emit-asm -o - %s 2>&1 | FileCheck %s
|
|
// RUN: %{ispc} --target=avx2-i32x16 --nowrap --x86-asm-syntax=intel --emit-asm -o - %s 2>&1 | FileCheck %s
|
|
// RUN: %{ispc} --target=avx512skx-x16 --nowrap --x86-asm-syntax=intel --emit-asm -o - %s 2>&1 | FileCheck %s
|
|
|
|
// REQUIRES: X86_ENABLED
|
|
// REQUIRES: LLVM_17_0+
|
|
|
|
struct vec4 {
|
|
double V[4];
|
|
};
|
|
|
|
uniform vec4 vmax(const uniform vec4 &V1, const uniform vec4 &V2) {
|
|
uniform vec4 Result;
|
|
|
|
foreach (i = 0 ... 4) {
|
|
Result.V[i] = max(V1.V[i], V2.V[i]);
|
|
}
|
|
|
|
return Result;
|
|
}
|
|
|
|
// CHECK-LABEL: foo:
|
|
// CHECK: vmovupd ymm{{.*}}, ymmword ptr [r{{.*}}]
|
|
// CHECK-NEXT: vmaxpd ymm{{.*}}, ymm{{.*}}, ymmword ptr [r{{.*}}]
|
|
// CHECK-NEXT: vmovupd ymmword ptr [r{{.*}}], ymm{{.*}}
|
|
export void foo(uniform double A[], uniform double B[]) {
|
|
|
|
uniform vec4 *uniform pA = (uniform vec4 * uniform) A;
|
|
uniform vec4 *uniform pB = (uniform vec4 * uniform) B;
|
|
|
|
*pA = vmax(*pA, *pB);
|
|
}
|