[Clang][Intrinsics][avx512] Continue Adding round cvt to clang

[AsafBadouh]

minor fixes, LGTM

lib/Headers/avx512fintrin.h

3658 ↗

(On Diff #59048)

please add (__mmask16)

3669 ↗

(On Diff #59048)

please add casting to return type (__m256i)

same for the rest

test/CodeGen/avx512f-builtins.c

3121 ↗

(On Diff #59048)

remove empty line

3129 ↗

(On Diff #59048)

remove empty line

cfe/trunk/lib/Headers/avx512fintrin.h

 #define _mm512_cmp_pd_mask(A, B, P) \
   _mm512_cmp_round_pd_mask((A), (B), (P), _MM_FROUND_CUR_DIRECTION)
 
 #define _mm512_mask_cmp_pd_mask(U, A, B, P) \
   _mm512_mask_cmp_round_pd_mask((U), (A), (B), (P), _MM_FROUND_CUR_DIRECTION)
 
 /* Conversion */
 
+#define _mm512_cvtt_roundps_epu32( __A, __R) __extension__ ({ \
+__builtin_ia32_cvttps2udq512_mask ((__v16sf)( __A),\
+                 (__v16si)\
+                 _mm512_undefined_epi32 (),\
+                 (__mmask16) -1,( __R));\
+})
+
+#define _mm512_mask_cvtt_roundps_epu32( __W, __U, __A, __R) __extension__ ({ \
+__builtin_ia32_cvttps2udq512_mask ((__v16sf)( __A),\
+                 (__v16si)( __W),\
+                 (__mmask16)( __U),( __R));\
+})
+
+#define _mm512_maskz_cvtt_roundps_epu32( __U, __A, __R) __extension__ ({ \
+__builtin_ia32_cvttps2udq512_mask ((__v16sf)( __A),\
+                 (__v16si)\
+                 _mm512_setzero_si512 (),\
+                 (__mmask16)( __U),( __R));\
+})
+
+
 static __inline __m512i __DEFAULT_FN_ATTRS
 _mm512_cvttps_epu32(__m512 __A)
 {
   return (__m512i) __builtin_ia32_cvttps2udq512_mask ((__v16sf) __A,
                   (__v16si)
                   _mm512_setzero_si512 (),
                   (__mmask16) -1,
                   _MM_FROUND_CUR_DIRECTION);

 _mm512_maskz_cvtpd_ps (__mmask8 __U, __m512d __A)
 {
   return (__m256) __builtin_ia32_cvtpd2ps512_mask ((__v8df) __A,
                 (__v8sf) _mm256_setzero_ps (),
                 (__mmask8) __U,
                 _MM_FROUND_CUR_DIRECTION);
 }
 
+#define _mm512_cvt_roundps_ph( __A, __I) __extension__ ({ \
+  (__m256i)__builtin_ia32_vcvtps2ph512_mask ((__v16sf)( __A),\
+                                             (__I),\
+                                             (__v16hi)_mm256_undefined_si256 (),\
+                                             (__mmask16) -1);\
+})
+
+#define _mm512_mask_cvt_roundps_ph( __U, __W, __A, __I) __extension__ ({ \
+  (__m256i)__builtin_ia32_vcvtps2ph512_mask ((__v16sf)( __A),\
+                                             (__I),\
+                                             (__v16hi)( __U),\
+                                             (__mmask16)( __W));\
+})
+
+#define _mm512_maskz_cvt_roundps_ph( __W, __A, __I) __extension__ ({ \
+  (__m256i)__builtin_ia32_vcvtps2ph512_mask ((__v16sf)( __A),\
+                                             (__I),\
+                                             (__v16hi)\
+                                             _mm256_setzero_si256 (),\
+                                             (__mmask16)( __W));\
+})
+
 #define _mm512_cvtps_ph(A, I) __extension__ ({ \
   (__m256i)__builtin_ia32_vcvtps2ph512_mask((__v16sf)(__m512)(A), (int)(I), \
                                             (__v16hi)_mm256_setzero_si256(), \
                                             (__mmask16)-1); })
 
 #define _mm512_mask_cvtps_ph(U, W, A, I) __extension__ ({ \
   (__m256i)__builtin_ia32_vcvtps2ph512_mask((__v16sf)(__m512)(A), (int)(I), \
                                             (__v16hi)(__m256i)(U), \
                                             (__mmask16)(W)); })
 
 #define _mm512_maskz_cvtps_ph(W, A, I) __extension__ ({\
   (__m256i)__builtin_ia32_vcvtps2ph512_mask((__v16sf)(__m512)(A), (int)(I), \
                                             (__v16hi)_mm256_setzero_si256(), \
                                             (__mmask16)(W)); })
-                   
+#define _mm512_cvt_roundph_ps( __A, __R) __extension__ ({ \
+__builtin_ia32_vcvtph2ps512_mask ((__v16hi)( __A),\
+               (__v16sf)\
+               _mm512_undefined_ps (),\
+               (__mmask16) -1,( __R));\
+})
+
+#define _mm512_mask_cvt_roundph_ps( __W, __U, __A, __R) __extension__ ({ \
+__builtin_ia32_vcvtph2ps512_mask ((__v16hi)( __A),\
+               (__v16sf)( __W),\
+               (__mmask16)( __U),( __R));\
+})
+
+#define _mm512_maskz_cvt_roundph_ps( __U, __A, __R) __extension__ ({ \
+__builtin_ia32_vcvtph2ps512_mask ((__v16hi)( __A),\
+               (__v16sf)\
+               _mm512_setzero_ps (),\
+               (__mmask16)( __U),( __R));\
+})
+                     
+                     
 static  __inline __m512 __DEFAULT_FN_ATTRS
 _mm512_cvtph_ps(__m256i __A)
 {
   return (__m512) __builtin_ia32_vcvtph2ps512_mask ((__v16hi) __A,
                 (__v16sf)
                 _mm512_setzero_ps (),
                 (__mmask16) -1,
                 _MM_FROUND_CUR_DIRECTION);

cfe/trunk/test/CodeGen/avx512f-builtins.c

 __m512 test_mm512_maskz_unpacklo_ps(__mmask16 __U, __m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_maskz_unpacklo_ps
   // CHECK: @llvm.x86.avx512.mask.unpckl.ps.512
   return _mm512_maskz_unpacklo_ps(__U, __A, __B); 
 }
 int test_mm_cvt_roundsd_si32(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvt_roundsd_si32
   // CHECK: @llvm.x86.avx512.vcvtsd2si32
-  return _mm_cvt_roundsd_si32(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvt_roundsd_si32(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 int test_mm_cvt_roundsd_i32(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvt_roundsd_i32
   // CHECK: @llvm.x86.avx512.vcvtsd2si32
-  return _mm_cvt_roundsd_i32(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvt_roundsd_i32(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 unsigned test_mm_cvt_roundsd_u32(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvt_roundsd_u32
   // CHECK: @llvm.x86.avx512.vcvtsd2usi32
-  return _mm_cvt_roundsd_u32(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvt_roundsd_u32(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 unsigned test_mm_cvtsd_u32(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvtsd_u32
   // CHECK: @llvm.x86.avx512.vcvtsd2usi32
   return _mm_cvtsd_u32(__A); 
 }
 
 unsigned long long test_mm_cvt_roundsd_u64(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvt_roundsd_u64
   // CHECK: @llvm.x86.avx512.vcvtsd2usi64
-  return _mm_cvt_roundsd_u64(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvt_roundsd_u64(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 unsigned long long test_mm_cvtsd_u64(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvtsd_u64
   // CHECK: @llvm.x86.avx512.vcvtsd2usi64
   return _mm_cvtsd_u64(__A); 
 }
 
 int test_mm_cvt_roundss_si32(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvt_roundss_si32
   // CHECK: @llvm.x86.avx512.vcvtss2si32
-  return _mm_cvt_roundss_si32(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvt_roundss_si32(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 int test_mm_cvt_roundss_i32(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvt_roundss_i32
   // CHECK: @llvm.x86.avx512.vcvtss2si32
-  return _mm_cvt_roundss_i32(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvt_roundss_i32(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 int test_mm_cvt_roundss_si64(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvt_roundss_si64
   // CHECK: @llvm.x86.avx512.vcvtss2si64
-  return _mm_cvt_roundss_si64(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvt_roundss_si64(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 long long test_mm_cvt_roundss_i64(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvt_roundss_i64
   // CHECK: @llvm.x86.avx512.vcvtss2si64
-  return _mm_cvt_roundss_i64(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvt_roundss_i64(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 unsigned test_mm_cvt_roundss_u32(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvt_roundss_u32
   // CHECK: @llvm.x86.avx512.vcvtss2usi32
-  return _mm_cvt_roundss_u32(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvt_roundss_u32(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 unsigned test_mm_cvtss_u32(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvtss_u32
   // CHECK: @llvm.x86.avx512.vcvtss2usi32
   return _mm_cvtss_u32(__A); 
 }
 
 unsigned long long test_mm_cvt_roundss_u64(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvt_roundss_u64
   // CHECK: @llvm.x86.avx512.vcvtss2usi64
-  return _mm_cvt_roundss_u64(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvt_roundss_u64(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 unsigned long long test_mm_cvtss_u64(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvtss_u64
   // CHECK: @llvm.x86.avx512.vcvtss2usi64
   return _mm_cvtss_u64(__A); 
 }
 
 int test_mm_cvtt_roundsd_i32(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvtt_roundsd_i32
   // CHECK: @llvm.x86.avx512.cvttsd2si
-  return _mm_cvtt_roundsd_i32(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvtt_roundsd_i32(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 int test_mm_cvtt_roundsd_si32(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvtt_roundsd_si32
   // CHECK: @llvm.x86.avx512.cvttsd2si
-  return _mm_cvtt_roundsd_si32(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvtt_roundsd_si32(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 int test_mm_cvttsd_i32(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvttsd_i32
   // CHECK: @llvm.x86.avx512.cvttsd2si
   return _mm_cvttsd_i32(__A); 
 }
 
 unsigned long long test_mm_cvtt_roundsd_si64(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvtt_roundsd_si64
   // CHECK: @llvm.x86.avx512.cvttsd2si64
-  return _mm_cvtt_roundsd_si64(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvtt_roundsd_si64(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 long long test_mm_cvtt_roundsd_i64(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvtt_roundsd_i64
   // CHECK: @llvm.x86.avx512.cvttsd2si64
-  return _mm_cvtt_roundsd_i64(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvtt_roundsd_i64(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 long long test_mm_cvttsd_i64(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvttsd_i64
   // CHECK: @llvm.x86.avx512.cvttsd2si64
   return _mm_cvttsd_i64(__A); 
 }
 
 unsigned test_mm_cvtt_roundsd_u32(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvtt_roundsd_u32
   // CHECK: @llvm.x86.avx512.cvttsd2usi
-  return _mm_cvtt_roundsd_u32(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvtt_roundsd_u32(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 unsigned test_mm_cvttsd_u32(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvttsd_u32
   // CHECK: @llvm.x86.avx512.cvttsd2usi
   return _mm_cvttsd_u32(__A); 
 }
 
 unsigned long long test_mm_cvtt_roundsd_u64(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvtt_roundsd_u64
   // CHECK: @llvm.x86.avx512.cvttsd2usi64
-  return _mm_cvtt_roundsd_u64(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvtt_roundsd_u64(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 unsigned long long test_mm_cvttsd_u64(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvttsd_u64
   // CHECK: @llvm.x86.avx512.cvttsd2usi64
   return _mm_cvttsd_u64(__A); 
 }
 
 int test_mm_cvtt_roundss_i32(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvtt_roundss_i32
   // CHECK: @llvm.x86.avx512.cvttss2si
-  return _mm_cvtt_roundss_i32(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvtt_roundss_i32(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 int test_mm_cvtt_roundss_si32(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvtt_roundss_si32
   // CHECK: @llvm.x86.avx512.cvttss2si
-  return _mm_cvtt_roundss_si32(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvtt_roundss_si32(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 int test_mm_cvttss_i32(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvttss_i32
   // CHECK: @llvm.x86.avx512.cvttss2si
   return _mm_cvttss_i32(__A); 
 }
 
 float test_mm_cvtt_roundss_i64(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvtt_roundss_i64
   // CHECK: @llvm.x86.avx512.cvttss2si64
-  return _mm_cvtt_roundss_i64(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvtt_roundss_i64(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 long long test_mm_cvtt_roundss_si64(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvtt_roundss_si64
   // CHECK: @llvm.x86.avx512.cvttss2si64
-  return _mm_cvtt_roundss_si64(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvtt_roundss_si64(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 long long test_mm_cvttss_i64(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvttss_i64
   // CHECK: @llvm.x86.avx512.cvttss2si64
   return _mm_cvttss_i64(__A); 
 }
 
 unsigned test_mm_cvtt_roundss_u32(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvtt_roundss_u32
   // CHECK: @llvm.x86.avx512.cvttss2usi
-  return _mm_cvtt_roundss_u32(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvtt_roundss_u32(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 unsigned test_mm_cvttss_u32(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvttss_u32
   // CHECK: @llvm.x86.avx512.cvttss2usi
   return _mm_cvttss_u32(__A); 
 }
 
 unsigned long long test_mm_cvtt_roundss_u64(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvtt_roundss_u64
   // CHECK: @llvm.x86.avx512.cvttss2usi64
-  return _mm_cvtt_roundss_u64(__A, _MM_FROUND_CUR_DIRECTION); 
+  return _mm_cvtt_roundss_u64(__A, _MM_FROUND_CUR_DIRECTION);
 }
 
 unsigned long long test_mm_cvttss_u64(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvttss_u64
   // CHECK: @llvm.x86.avx512.cvttss2usi64
   return _mm_cvttss_u64(__A); 
 }
 
+__m512i test_mm512_cvtt_roundps_epu32(__m512 __A) 
+{
+    // CHECK-LABEL: @test_mm512_cvtt_roundps_epu32
+    // CHECK: @llvm.x86.avx512.mask.cvttps2udq.512
+    return _mm512_cvtt_roundps_epu32(__A, _MM_FROUND_CUR_DIRECTION);
+}
+
+__m512i test_mm512_mask_cvtt_roundps_epu32(__m512i __W, __mmask16 __U, __m512 __A)
+{
+    // CHECK-LABEL: @test_mm512_mask_cvtt_roundps_epu32
+    // CHECK: @llvm.x86.avx512.mask.cvttps2udq.512
+    return _mm512_mask_cvtt_roundps_epu32(__W, __U, __A, _MM_FROUND_CUR_DIRECTION);
+}
+
+__m512i test_mm512_maskz_cvtt_roundps_epu32( __mmask16 __U, __m512 __A)
+{
+    // CHECK-LABEL: @test_mm512_maskz_cvtt_roundps_epu32
+    // CHECK: @llvm.x86.avx512.mask.cvttps2udq.512
+
+    return _mm512_maskz_cvtt_roundps_epu32(__U, __A, _MM_FROUND_CUR_DIRECTION);
+}
+
+__m256i test_mm512_cvt_roundps_ph(__m512  __A)
+{
+    // CHECK-LABEL: @test_mm512_cvt_roundps_ph
+    // CHECK: @llvm.x86.avx512.mask.vcvtps2ph.512
+    return _mm512_cvt_roundps_ph(__A, _MM_FROUND_CUR_DIRECTION);
+}
+
+__m256i test_mm512_mask_cvt_roundps_ph(__m256i __W , __mmask16 __U, __m512  __A)
+{
+    // CHECK-LABEL: @test_mm512_mask_cvt_roundps_ph
+    // CHECK: @llvm.x86.avx512.mask.vcvtps2ph.512
+    return _mm512_mask_cvt_roundps_ph(__W, __U, __A, _MM_FROUND_CUR_DIRECTION);
+}
+
+__m256i test_mm512_maskz_cvt_roundps_ph(__mmask16 __U, __m512  __A)
+{
+    // CHECK-LABEL: @test_mm512_maskz_cvt_roundps_ph
+    // CHECK: @llvm.x86.avx512.mask.vcvtps2ph.512
+    return _mm512_maskz_cvt_roundps_ph(__U, __A, _MM_FROUND_CUR_DIRECTION);
+}
+
+__m512 test_mm512_cvt_roundph_ps(__m256i __A)
+{
+    // CHECK-LABEL: @test_mm512_cvt_roundph_ps
+    // CHECK: @llvm.x86.avx512.mask.vcvtph2ps.512
+    return _mm512_cvt_roundph_ps(__A, _MM_FROUND_CUR_DIRECTION);
+}
+
+__m512 test_mm512_mask_cvt_roundph_ps(__m512 __W, __mmask16 __U, __m256i __A)
+{
+    // CHECK-LABEL: @test_mm512_mask_cvt_roundph_ps
+    // CHECK: @llvm.x86.avx512.mask.vcvtph2ps.512
+    return _mm512_mask_cvt_roundph_ps(__W, __U, __A, _MM_FROUND_CUR_DIRECTION);
+}
+
+__m512 test_mm512_maskz_cvt_roundph_ps(__mmask16 __U, __m256i __A)
+{
+    // CHECK-LABEL: @test_mm512_maskz_cvt_roundph_ps
+    // CHECK: @llvm.x86.avx512.mask.vcvtph2ps.512
+    return _mm512_maskz_cvt_roundph_ps(__U, __A, _MM_FROUND_CUR_DIRECTION);
+}
+
 __m512 test_mm512_mask_cvt_roundepi32_ps(__m512 __W, __mmask16 __U, __m512i __A)
 {
   // CHECK-LABEL: @test_mm512_mask_cvt_roundepi32_ps
   // CHECK: @llvm.x86.avx512.mask.cvtdq2ps.512
   return _mm512_mask_cvt_roundepi32_ps(__W,__U,__A,4);
 }
 
 __m512 test_mm512_maskz_cvt_roundepi32_ps(__mmask16 __U, __m512i __A)