summaryrefslogtreecommitdiffstats
path: root/clang/test/CodeGen/avx512bw-builtins.c
diff options
context:
space:
mode:
authorCraig Topper <craig.topper@intel.com>2019-04-15 17:17:35 +0000
committerCraig Topper <craig.topper@intel.com>2019-04-15 17:17:35 +0000
commit8e364c680fab1724bd0d4749b8d1177843164aa7 (patch)
treeb7b4d9d82e9afc1160fc1c2cf8c2688415162420 /clang/test/CodeGen/avx512bw-builtins.c
parent8d856488a897043877d9ed26861846cc80f4897e (diff)
downloadbcm5719-llvm-8e364c680fab1724bd0d4749b8d1177843164aa7.tar.gz
bcm5719-llvm-8e364c680fab1724bd0d4749b8d1177843164aa7.zip
[X86] Restore the pavg intrinsics.
The pattern we replaced these with may be too hard to match as demonstrated by PR41496 and PR41316. This patch restores the intrinsics and then we can start focusing on the optimizing the intrinsics. I've mostly reverted the original patch that removed them. Though I modified the avx512 intrinsics to not have masking built in. Differential Revision: https://reviews.llvm.org/D60674 llvm-svn: 358427
Diffstat (limited to 'clang/test/CodeGen/avx512bw-builtins.c')
-rw-r--r--clang/test/CodeGen/avx512bw-builtins.c50
1 files changed, 6 insertions, 44 deletions
diff --git a/clang/test/CodeGen/avx512bw-builtins.c b/clang/test/CodeGen/avx512bw-builtins.c
index 9e75baae77f..562bc312889 100644
--- a/clang/test/CodeGen/avx512bw-builtins.c
+++ b/clang/test/CodeGen/avx512bw-builtins.c
@@ -1066,73 +1066,35 @@ __m512i test_mm512_maskz_adds_epu16(__mmask32 __U, __m512i __A, __m512i __B) {
}
__m512i test_mm512_avg_epu8(__m512i __A, __m512i __B) {
// CHECK-LABEL: @test_mm512_avg_epu8
- // CHECK-NOT: @llvm.x86.avx512.mask.pavg.b.512
- // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
- // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
- // CHECK: add <64 x i16> %{{.*}}, %{{.*}}
- // CHECK: add <64 x i16> %{{.*}}, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
- // CHECK: lshr <64 x i16> %{{.*}}, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
- // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
+ // CHECK: @llvm.x86.avx512.pavg.b.512
return _mm512_avg_epu8(__A,__B);
}
__m512i test_mm512_mask_avg_epu8(__m512i __W, __mmask64 __U, __m512i __A, __m512i __B) {
// CHECK-LABEL: @test_mm512_mask_avg_epu8
- // CHECK-NOT: @llvm.x86.avx512.mask.pavg.b.512
- // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
- // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
- // CHECK: add <64 x i16> %{{.*}}, %{{.*}}
- // CHECK: add <64 x i16> %{{.*}}, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
- // CHECK: lshr <64 x i16> %{{.*}}, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
- // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
+ // CHECK: @llvm.x86.avx512.pavg.b.512
// CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
return _mm512_mask_avg_epu8(__W,__U,__A,__B);
}
__m512i test_mm512_maskz_avg_epu8(__mmask64 __U, __m512i __A, __m512i __B) {
// CHECK-LABEL: @test_mm512_maskz_avg_epu8
- // CHECK-NOT: @llvm.x86.avx512.mask.pavg.b.512
- // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
- // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
- // CHECK: add <64 x i16> %{{.*}}, %{{.*}}
- // CHECK: add <64 x i16> %{{.*}}, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
- // CHECK: lshr <64 x i16> %{{.*}}, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
- // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
- // CHECK: store <8 x i64> zeroinitializer
+ // CHECK: @llvm.x86.avx512.pavg.b.512
// CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
return _mm512_maskz_avg_epu8(__U,__A,__B);
}
__m512i test_mm512_avg_epu16(__m512i __A, __m512i __B) {
// CHECK-LABEL: @test_mm512_avg_epu16
- // CHECK-NOT: @llvm.x86.avx512.mask.pavg.w.512
- // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
- // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
- // CHECK: add <32 x i32> %{{.*}}, %{{.*}}
- // CHECK: add <32 x i32> %{{.*}}, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
- // CHECK: lshr <32 x i32> %{{.*}}, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
- // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
+ // CHECK: @llvm.x86.avx512.pavg.w.512
return _mm512_avg_epu16(__A,__B);
}
__m512i test_mm512_mask_avg_epu16(__m512i __W, __mmask32 __U, __m512i __A, __m512i __B) {
// CHECK-LABEL: @test_mm512_mask_avg_epu16
- // CHECK-NOT: @llvm.x86.avx512.mask.pavg.w.512
- // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
- // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
- // CHECK: add <32 x i32> %{{.*}}, %{{.*}}
- // CHECK: add <32 x i32> %{{.*}}, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
- // CHECK: lshr <32 x i32> %{{.*}}, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
- // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
+ // CHECK: @llvm.x86.avx512.pavg.w.512
// CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
return _mm512_mask_avg_epu16(__W,__U,__A,__B);
}
__m512i test_mm512_maskz_avg_epu16(__mmask32 __U, __m512i __A, __m512i __B) {
// CHECK-LABEL: @test_mm512_maskz_avg_epu16
- // CHECK-NOT: @llvm.x86.avx512.mask.pavg.w.512
- // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
- // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
- // CHECK: add <32 x i32> %{{.*}}, %{{.*}}
- // CHECK: add <32 x i32> %{{.*}}, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
- // CHECK: lshr <32 x i32> %{{.*}}, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
- // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
- // CHECK: store <8 x i64> zeroinitializer
+ // CHECK: @llvm.x86.avx512.pavg.w.512
// CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
return _mm512_maskz_avg_epu16(__U,__A,__B);
}
OpenPOWER on IntegriCloud