mirror of
https://github.com/llvm/llvm-project.git
synced 2025-04-27 18:16:06 +00:00
f3914b74c1
Previously we were just using extended vector operations in the header file. This unfortunately allowed non-constant indices to be used with the intrinsics. This is incompatible with gcc, icc, and MSVC. It also introduces a different performance characteristic because non-constant index gets lowered to a vector store and an element sized load. By adding the builtins we can check for the index to be a constant and ensure its in range of the vector element count. User code still has the option to use extended vector operations themselves if they need non-constant indexing. llvm-svn: 334057
83 lines
2.1 KiB
C
83 lines
2.1 KiB
C
// RUN: %clang_cc1 -ffreestanding -triple i386-apple-darwin9 -O1 -target-cpu corei7 -debug-info-kind=limited -emit-llvm %s -o - | FileCheck %s
|
|
typedef short __v4hi __attribute__ ((__vector_size__ (8)));
|
|
|
|
void test1() {
|
|
__v4hi A = (__v4hi)0LL;
|
|
}
|
|
|
|
__v4hi x = {1,2,3};
|
|
__v4hi y = {1,2,3,4};
|
|
|
|
typedef int vty __attribute((vector_size(16)));
|
|
int test2() { vty b; return b[2LL]; }
|
|
|
|
// PR4339
|
|
typedef float vec4 __attribute__((vector_size(16)));
|
|
|
|
void test3 ( vec4* a, char b, float c ) {
|
|
(*a)[b] = c;
|
|
}
|
|
|
|
|
|
|
|
|
|
#include <mmintrin.h>
|
|
|
|
int test4(int argc, char *argv[]) {
|
|
int array[16] = { 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15 };
|
|
__m64 *p = (__m64 *)array;
|
|
|
|
__m64 accum = _mm_setzero_si64();
|
|
|
|
for (int i=0; i<8; ++i)
|
|
accum = _mm_add_pi32(p[i], accum);
|
|
|
|
__m64 accum2 = _mm_unpackhi_pi32(accum, accum);
|
|
accum = _mm_add_pi32(accum, accum2);
|
|
|
|
int result = _mm_cvtsi64_si32(accum);
|
|
_mm_empty();
|
|
|
|
return result;
|
|
}
|
|
|
|
#include <smmintrin.h>
|
|
|
|
unsigned long test_epi8(__m128i x) { return _mm_extract_epi8(x, 4); }
|
|
// CHECK: @test_epi8
|
|
// CHECK: extractelement <16 x i8> {{.*}}, i32 4
|
|
// CHECK: zext i8 {{.*}} to i32
|
|
|
|
unsigned long test_epi16(__m128i x) { return _mm_extract_epi16(x, 3); }
|
|
|
|
// CHECK: @test_epi16
|
|
// CHECK: extractelement <8 x i16> {{.*}}, i32 3
|
|
// CHECK: zext i16 {{.*}} to i32
|
|
|
|
void extractinttypes() {
|
|
extern int check_extract_result_int;
|
|
extern __typeof(_mm_extract_epi8(_mm_setzero_si128(), 3)) check_result_int;
|
|
extern __typeof(_mm_extract_epi16(_mm_setzero_si128(), 3)) check_result_int;
|
|
extern __typeof(_mm_extract_epi32(_mm_setzero_si128(), 3)) check_result_int;
|
|
}
|
|
|
|
// Test some logic around our lax vector comparison rules with integers.
|
|
|
|
typedef int vec_int1 __attribute__((vector_size(4)));
|
|
vec_int1 lax_vector_compare1(int x, vec_int1 y) {
|
|
y = x == y;
|
|
return y;
|
|
}
|
|
|
|
// CHECK: define i32 @lax_vector_compare1(i32 {{.*}}, i32 {{.*}})
|
|
// CHECK: icmp eq <1 x i32>
|
|
|
|
typedef int vec_int2 __attribute__((vector_size(8)));
|
|
vec_int2 lax_vector_compare2(long long x, vec_int2 y) {
|
|
y = x == y;
|
|
return y;
|
|
}
|
|
|
|
// CHECK: define void @lax_vector_compare2(<2 x i32>* {{.*sret.*}}, i64 {{.*}}, i64 {{.*}})
|
|
// CHECK: icmp eq <2 x i32>
|