Author: ctopper
Date: Tue Apr 10 21:55:10 2018
New Revision: 329775
URL: http://llvm.org/viewvc/llvm-project?rev=329775&view=rev
Log:
[X86] Replace 512-bit masked pmaddubsw and pmaddwd intrinsic with unmasked
intrinsic and a select.
This makes it consistent with the 128/256-bit functions.
Someday maybe we'll have all the masking moved to selects.
Modified:
cfe/trunk/include/clang/Basic/BuiltinsX86.def
cfe/trunk/lib/Headers/avx512bwintrin.h
cfe/trunk/test/CodeGen/avx512bw-builtins.c
Modified: cfe/trunk/include/clang/Basic/BuiltinsX86.def
URL:
http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/BuiltinsX86.def?rev=329775&r1=329774&r2=329775&view=diff
==============================================================================
--- cfe/trunk/include/clang/Basic/BuiltinsX86.def (original)
+++ cfe/trunk/include/clang/Basic/BuiltinsX86.def Tue Apr 10 21:55:10 2018
@@ -1151,8 +1151,8 @@ TARGET_BUILTIN(__builtin_ia32_mulps512_m
TARGET_BUILTIN(__builtin_ia32_subpd512_mask, "V8dV8dV8dV8dUcIi", "", "avx512f")
TARGET_BUILTIN(__builtin_ia32_subps512_mask, "V16fV16fV16fV16fUsIi", "",
"avx512f")
-TARGET_BUILTIN(__builtin_ia32_pmaddubsw512_mask, "V32sV64cV64cV32sUi", "",
"avx512bw")
-TARGET_BUILTIN(__builtin_ia32_pmaddwd512_mask, "V16iV32sV32sV16iUs", "",
"avx512bw")
+TARGET_BUILTIN(__builtin_ia32_pmaddubsw512, "V32sV64cV64c", "", "avx512bw")
+TARGET_BUILTIN(__builtin_ia32_pmaddwd512, "V16iV32sV32s", "", "avx512bw")
TARGET_BUILTIN(__builtin_ia32_addss_round_mask, "V4fV4fV4fV4fUcIi", "",
"avx512f")
TARGET_BUILTIN(__builtin_ia32_divss_round_mask, "V4fV4fV4fV4fUcIi", "",
"avx512f")
Modified: cfe/trunk/lib/Headers/avx512bwintrin.h
URL:
http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Headers/avx512bwintrin.h?rev=329775&r1=329774&r2=329775&view=diff
==============================================================================
--- cfe/trunk/lib/Headers/avx512bwintrin.h (original)
+++ cfe/trunk/lib/Headers/avx512bwintrin.h Tue Apr 10 21:55:10 2018
@@ -1075,53 +1075,42 @@ _mm512_maskz_mulhi_epu16 (__mmask32 __U,
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_maddubs_epi16 (__m512i __X, __m512i __Y) {
- return (__m512i) __builtin_ia32_pmaddubsw512_mask ((__v64qi) __X,
- (__v64qi) __Y,
- (__v32hi) _mm512_setzero_hi(),
- (__mmask32) -1);
+_mm512_maddubs_epi16(__m512i __X, __m512i __Y) {
+ return (__m512i)__builtin_ia32_pmaddubsw512((__v64qi)__X, (__v64qi)__Y);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_mask_maddubs_epi16 (__m512i __W, __mmask32 __U, __m512i __X,
- __m512i __Y) {
- return (__m512i) __builtin_ia32_pmaddubsw512_mask ((__v64qi) __X,
- (__v64qi) __Y,
- (__v32hi) __W,
- (__mmask32) __U);
+_mm512_mask_maddubs_epi16(__m512i __W, __mmask32 __U, __m512i __X,
+ __m512i __Y) {
+ return (__m512i)__builtin_ia32_selectw_512((__mmask32) __U,
+ (__v32hi)_mm512_maddubs_epi16(__X,
__Y),
+ (__v32hi)__W);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_maskz_maddubs_epi16 (__mmask32 __U, __m512i __X, __m512i __Y) {
- return (__m512i) __builtin_ia32_pmaddubsw512_mask ((__v64qi) __X,
- (__v64qi) __Y,
- (__v32hi) _mm512_setzero_hi(),
- (__mmask32) __U);
+_mm512_maskz_maddubs_epi16(__mmask32 __U, __m512i __X, __m512i __Y) {
+ return (__m512i)__builtin_ia32_selectw_512((__mmask32) __U,
+ (__v32hi)_mm512_maddubs_epi16(__X,
__Y),
+ (__v32hi)_mm512_setzero_hi());
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_madd_epi16 (__m512i __A, __m512i __B) {
- return (__m512i) __builtin_ia32_pmaddwd512_mask ((__v32hi) __A,
- (__v32hi) __B,
- (__v16si) _mm512_setzero_si512(),
- (__mmask16) -1);
+_mm512_madd_epi16(__m512i __A, __m512i __B) {
+ return (__m512i)__builtin_ia32_pmaddwd512((__v32hi)__A, (__v32hi)__B);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_mask_madd_epi16 (__m512i __W, __mmask16 __U, __m512i __A,
- __m512i __B) {
- return (__m512i) __builtin_ia32_pmaddwd512_mask ((__v32hi) __A,
- (__v32hi) __B,
- (__v16si) __W,
- (__mmask16) __U);
+_mm512_mask_madd_epi16(__m512i __W, __mmask16 __U, __m512i __A, __m512i __B) {
+ return (__m512i)__builtin_ia32_selectd_512((__mask16)__U,
+ (__v16si)_mm512_madd_epi16(__A,
__B),
+ (__v16si)__W);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
-_mm512_maskz_madd_epi16 (__mmask16 __U, __m512i __A, __m512i __B) {
- return (__m512i) __builtin_ia32_pmaddwd512_mask ((__v32hi) __A,
- (__v32hi) __B,
- (__v16si) _mm512_setzero_si512(),
- (__mmask16) __U);
+_mm512_maskz_madd_epi16(__mmask16 __U, __m512i __A, __m512i __B) {
+ return (__m512i)__builtin_ia32_selectd_512((__mask16)__U,
+ (__v16si)_mm512_madd_epi16(__A,
__B),
+ (__v16si)_mm512_setzero_si512());
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
Modified: cfe/trunk/test/CodeGen/avx512bw-builtins.c
URL:
http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGen/avx512bw-builtins.c?rev=329775&r1=329774&r2=329775&view=diff
==============================================================================
--- cfe/trunk/test/CodeGen/avx512bw-builtins.c (original)
+++ cfe/trunk/test/CodeGen/avx512bw-builtins.c Tue Apr 10 21:55:10 2018
@@ -1036,32 +1036,36 @@ __m512i test_mm512_maskz_mulhi_epu16(__m
__m512i test_mm512_maddubs_epi16(__m512i __X, __m512i __Y) {
// CHECK-LABEL: @test_mm512_maddubs_epi16
- // CHECK: @llvm.x86.avx512.mask.pmaddubs.w.512
+ // CHECK: @llvm.x86.avx512.pmaddubs.w.512
return _mm512_maddubs_epi16(__X,__Y);
}
__m512i test_mm512_mask_maddubs_epi16(__m512i __W, __mmask32 __U, __m512i __X,
__m512i __Y) {
// CHECK-LABEL: @test_mm512_mask_maddubs_epi16
- // CHECK: @llvm.x86.avx512.mask.pmaddubs.w.512
+ // CHECK: @llvm.x86.avx512.pmaddubs.w.512
+ // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
return _mm512_mask_maddubs_epi16(__W,__U,__X,__Y);
}
__m512i test_mm512_maskz_maddubs_epi16(__mmask32 __U, __m512i __X, __m512i
__Y) {
// CHECK-LABEL: @test_mm512_maskz_maddubs_epi16
- // CHECK: @llvm.x86.avx512.mask.pmaddubs.w.512
+ // CHECK: @llvm.x86.avx512.pmaddubs.w.512
+ // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
return _mm512_maskz_maddubs_epi16(__U,__X,__Y);
}
__m512i test_mm512_madd_epi16(__m512i __A, __m512i __B) {
// CHECK-LABEL: @test_mm512_madd_epi16
- // CHECK: @llvm.x86.avx512.mask.pmaddw.d.512
+ // CHECK: @llvm.x86.avx512.pmaddw.d.512
return _mm512_madd_epi16(__A,__B);
}
__m512i test_mm512_mask_madd_epi16(__m512i __W, __mmask16 __U, __m512i __A,
__m512i __B) {
// CHECK-LABEL: @test_mm512_mask_madd_epi16
- // CHECK: @llvm.x86.avx512.mask.pmaddw.d.512
+ // CHECK: @llvm.x86.avx512.pmaddw.d.512
+ // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> %{{.*}}
return _mm512_mask_madd_epi16(__W,__U,__A,__B);
}
__m512i test_mm512_maskz_madd_epi16(__mmask16 __U, __m512i __A, __m512i __B) {
// CHECK-LABEL: @test_mm512_maskz_madd_epi16
- // CHECK: @llvm.x86.avx512.mask.pmaddw.d.512
+ // CHECK: @llvm.x86.avx512.pmaddw.d.512
+ // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> %{{.*}}
return _mm512_maskz_madd_epi16(__U,__A,__B);
}
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
