[Clang][Intrinsics][avx512] Adding round roundscale to clang

Differential Revision: http://reviews.llvm.org/D20815

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@271368 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Headers/avx512fintrin.h b/lib/Headers/avx512fintrin.h
index 4d13fbfa586f102b5342407aa08ac4c90fbcffb1..368db3dbfd755dd9938714228cea65570d464907 100644 (file)
--- a/lib/Headers/avx512fintrin.h
+++ b/lib/Headers/avx512fintrin.h
@@ -2382,6 +2382,24 @@ _mm512_maskz_div_ps(__mmask16 __U, __m512 __A, __m512 __B) {
                                          (__mmask16)(A), \
                                          _MM_FROUND_CUR_DIRECTION); })
 
+#define _mm512_mask_roundscale_round_ps( __A, __B, __C, __imm, __R) __extension__ ({ \
+  (__m512)__builtin_ia32_rndscaleps_mask ((__v16sf)( __C), (int)__imm,\
+                                          (__v16sf)( __A),\
+                                          (__mmask16)( __B),(int) __R);\
+})
+
+#define _mm512_maskz_roundscale_round_ps( __A, __B, __imm,__R) __extension__ ({ \
+  (__m512)__builtin_ia32_rndscaleps_mask ((__v16sf)( __B), (int)__imm,\
+                                          (__v16sf)_mm512_setzero_ps (),\
+                                          (__mmask16)( __A),(int) __R);\
+})
+
+#define _mm512_roundscale_round_ps( __A, __imm, __R) __extension__ ({ \
+  (__m512)__builtin_ia32_rndscaleps_mask ((__v16sf)( __A),(int) __imm,\
+                                          (__v16sf) _mm512_undefined_ps (),\
+                                          (__mmask16) -1,(int) __R);\
+})
+
 #define _mm512_roundscale_pd(A, B) __extension__ ({ \
   (__m512d)__builtin_ia32_rndscalepd_mask((__v8df)(__m512d)(A), (int)(B), \
                                           (__v8df)(__m512d)(A), (__mmask8)-1, \
@@ -2398,6 +2416,24 @@ _mm512_maskz_div_ps(__mmask16 __U, __m512 __A, __m512 __B) {
                                           (__mmask8)(A), \
                                           _MM_FROUND_CUR_DIRECTION); })
 
+#define _mm512_mask_roundscale_round_pd( __A, __B, __C, __imm ,__R) __extension__ ({ \
+  (__m512d)__builtin_ia32_rndscalepd_mask ((__v8df)( __C),(int)__imm,\
+                                           (__v8df)( __A),\
+                                           (__mmask8)( __B),(int)__R);\
+})
+
+#define _mm512_maskz_roundscale_round_pd( __A, __B, __imm, __R) __extension__ ({ \
+  (__m512d)__builtin_ia32_rndscalepd_mask ((__v8df)( __B),(int)__imm,\
+                                            (__v8df)_mm512_setzero_pd (),\
+                                            (__mmask8)( __A),(int) __R);\
+})
+
+#define _mm512_roundscale_round_pd( __A, __imm , __R) __extension__ ({ \
+  (__m512d)__builtin_ia32_rndscalepd_mask ((__v8df)( __A),(int) __imm,\
+                                            (__v8df)_mm512_undefined_pd (),\
+                                            (__mmask8) -1,(int) __R);\
+})
+
 #define _mm512_fmadd_round_pd(A, B, C, R) __extension__ ({ \
   (__m512d)__builtin_ia32_vfmaddpd512_mask((__v8df)(__m512d)(A), \
                                            (__v8df)(__m512d)(B), \

diff --git a/test/CodeGen/avx512f-builtins.c b/test/CodeGen/avx512f-builtins.c
index 05e7aeaabec2547671ce194f3bb987b782b998df..50734c08b809db477c7a888db06c52f8492c46cb 100644 (file)
--- a/test/CodeGen/avx512f-builtins.c
+++ b/test/CodeGen/avx512f-builtins.c
@@ -6653,6 +6653,27 @@ __m512 test_mm512_maskz_roundscale_ps(__mmask16 __U, __m512 __A)
   return _mm512_maskz_roundscale_ps(__U,__A, 1);
 }
 
+__m512 test_mm512_mask_roundscale_round_ps(__m512 __A,__mmask16 __U,__m512 __C)
+{
+  // CHECK-LABEL: @test_mm512_mask_roundscale_round_ps
+  // CHECK: @llvm.x86.avx512.mask.rndscale.ps.512
+  return _mm512_mask_roundscale_round_ps(__A,__U,__C,3,_MM_FROUND_CUR_DIRECTION);
+}
+
+__m512 test_mm512_maskz_roundscale_round_ps(__m512 __A,__mmask16 __U) 
+{
+  // CHECK-LABEL: @test_mm512_maskz_roundscale_round_ps
+  // CHECK: @llvm.x86.avx512.mask.rndscale.ps.512
+  return _mm512_maskz_roundscale_round_ps(__U,__A,3,_MM_FROUND_CUR_DIRECTION);
+}
+
+__m512 test_mm512_roundscale_round_ps(__m512 __A)
+{
+  // CHECK-LABEL: @test_mm512_roundscale_round_ps
+  // CHECK: @llvm.x86.avx512.mask.rndscale.ps.512
+  return _mm512_roundscale_round_ps(__A,3,_MM_FROUND_CUR_DIRECTION);
+}
+
 __m512d test_mm512_mask_roundscale_pd(__m512d __W, __mmask8 __U, __m512d __A) 
 {
   // CHECK-LABEL: @test_mm512_mask_roundscale_pd
@@ -6667,6 +6688,27 @@ __m512d test_mm512_maskz_roundscale_pd(__mmask8 __U, __m512d __A)
   return _mm512_maskz_roundscale_pd(__U,__A, 1);
 }
 
+__m512d test_mm512_mask_roundscale_round_pd(__m512d __A,__mmask8 __U,__m512d __C)
+{
+  // CHECK-LABEL: @test_mm512_mask_roundscale_round_pd
+  // CHECK: @llvm.x86.avx512.mask.rndscale.pd.512
+  return _mm512_mask_roundscale_round_pd(__A,__U,__C,3,_MM_FROUND_CUR_DIRECTION);
+}
+
+__m512d test_mm512_maskz_roundscale_round_pd(__m512d __A,__mmask8 __U)
+{
+  // CHECK-LABEL: @test_mm512_maskz_roundscale_round_pd
+  // CHECK: @llvm.x86.avx512.mask.rndscale.pd.512
+  return _mm512_maskz_roundscale_round_pd(__U,__A,3,_MM_FROUND_CUR_DIRECTION);
+}
+
+__m512d test_mm512_roundscale_round_pd(__m512d __A)
+{
+  // CHECK-LABEL: @test_mm512_roundscale_round_pd
+  // CHECK: @llvm.x86.avx512.mask.rndscale.pd.512
+  return _mm512_roundscale_round_pd(__A,3,_MM_FROUND_CUR_DIRECTION);
+}
+
 __m512i test_mm512_mask_max_epi32 (__m512i __W, __mmask16 __M, __m512i __A, __m512i __B)
 {
   // CHECK-LABEL: @test_mm512_mask_max_epi32