Adding front-end support to several intrinsics (bit scanning, conversion and state reading intrinsics)

[opaparo]

Summary

Adding LLVM front-end support to two intrinsics dealing with bit scan: _bit_scan_forward and _bit_scan_reverse.
Their functionality is as described in Intel intrinsics guide:
https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_bit_scan_forward&expand=371,370
https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_bit_scan_reverse&expand=371,370

Furthermore, adding clang front-end support to these conversion intrinsics: _mm256_cvtsd_f64, _mm256_cvtsi256_si32 and _mm256_cvtss_f32.

Finally, adding tests to all of the above, as well as to the state reading intrinsics _rdpmc and _rdtsc.
Their functionality is also specified in the Intel intrinsics guide.

[m_zuckerman]

test/CodeGen/avx-builtins.c

192

Add more one line

test/CodeGen/bitscan-builtins.c

18

Add more one line

test/CodeGen/rd-builtins.c

19

Add more one line

[m_zuckerman]

include/clang/Basic/BuiltinsX86.def

39

convention: add _ia32

40

convention: add _ia32

[m_zuckerman]

Add Dependencies in file http://reviews.llvm.org/D19915

[m_zuckerman]

include/clang/Basic/BuiltinsX86.def

39

Why use BUILTIN and not target_builtin?

[AsafBadouh]

lib/Headers/immintrin.h

112

is it only for X86_64?
I think it isn't.

test/CodeGen/bitscan-builtins.c

1

you don't need all the target features

test/CodeGen/rd-builtins.c

1

I think that the target features are incorrect.

[AsafBadouh]

LGTM

[Eugene.Zelenko]

Committed in r271387.

include/clang/Basic/BuiltinsX86.def

 // can use it?
 BUILTIN(__builtin_cpu_supports, "bcC*", "nc")
 
 // Win64-compatible va_list functions
 BUILTIN(__builtin_ms_va_start, "vc*&.", "nt")
 BUILTIN(__builtin_ms_va_end, "vc*&", "n")
 BUILTIN(__builtin_ms_va_copy, "vc*&c*&", "n")
 
+// Bit scan
+BUILTIN(__builtin_bit_scan_forward, "ii", "")
+BUILTIN(__builtin_bit_scan_reverse, "ii", "")
+
 // Undefined Values
 //
 TARGET_BUILTIN(__builtin_ia32_undef128, "V2d", "nc", "")
 TARGET_BUILTIN(__builtin_ia32_undef256, "V4d", "nc", "")
 TARGET_BUILTIN(__builtin_ia32_undef512, "V8d", "nc", "")
 
 // FLAGS
 //

lib/Headers/avxintrin.h

 }
 
 static __inline __m256i __DEFAULT_FN_ATTRS
 _mm256_cvttps_epi32(__m256 __a)
 {
   return (__m256i)__builtin_ia32_cvttps2dq256((__v8sf) __a);
 }
 
+static __inline double __DEFAULT_FN_ATTRS
+_mm256_cvtsd_f64(__m256d __a)
+{
+  return __a[0];
+}
+
+static __inline int __DEFAULT_FN_ATTRS
+_mm256_cvtsi256_si32(__m256i __a)
+{
+  __v8si __b = (__v8si)__a;
+  return __b[0];
+}
+
+static __inline float __DEFAULT_FN_ATTRS
+_mm256_cvtss_f32(__m256 __a)
+{
+  return __a[0];
+}
+
 /* Vector replicate */
 static __inline __m256 __DEFAULT_FN_ATTRS
 _mm256_movehdup_ps(__m256 __a)
 {
   return __builtin_shufflevector(__a, __a, 1, 1, 3, 3, 5, 5, 7, 7);
 }
 
 static __inline __m256 __DEFAULT_FN_ATTRS

lib/Headers/ia32intrin.h

 /* __rdtscp */
 static __inline__ unsigned long long __attribute__((__always_inline__, __nodebug__))
 __rdtscp(unsigned int *__A) {
   return __builtin_ia32_rdtscp(__A);
 }
 
 #define _rdtsc() __rdtsc()
 
+#define _rdpmc(A) __rdpmc(A)
+
 #endif /* __IA32INTRIN_H */

lib/Headers/immintrin.h

 #ifdef __x86_64__
 static __inline__ int __attribute__((__always_inline__, __nodebug__, __target__("rdrnd")))
 _rdrand64_step(unsigned long long *__p)
 {
   return __builtin_ia32_rdrand64_step(__p);
 }
 #endif
 
 #ifdef __x86_64__
 static __inline__ unsigned int __attribute__((__always_inline__, __nodebug__, __target__("fsgsbase")))
 _readfsbase_u32(void)
 {
   return __builtin_ia32_rdfsbase32();
 }
 
 static __inline__ unsigned long long __attribute__((__always_inline__, __nodebug__, __target__("fsgsbase")))
 _readfsbase_u64(void)
_writegsbase_u32(unsigned int __V)
   return __builtin_ia32_wrgsbase32(__V);
 }
 
 static __inline__ void __attribute__((__always_inline__, __nodebug__, __target__("fsgsbase")))
 _writegsbase_u64(unsigned long long __V)
 {
   return __builtin_ia32_wrgsbase64(__V);
 }
+
+/* __bit_scan_forward */
+static __inline__ int __attribute__((__always_inline__, __nodebug__))
+_bit_scan_forward(int __A) {
+  return __builtin_bit_scan_forward(__A);
+}
+
+/* __bit_scan_reverse */
+static __inline__ int __attribute__((__always_inline__, __nodebug__))
+_bit_scan_reverse(int __A) {
+  return __builtin_bit_scan_reverse(__A);
+}
+
 #endif
 
 #include <rtmintrin.h>
 
 #include <xtestintrin.h>
 
 #include <shaintrin.h>
 

test/CodeGen/avx-builtins.c

__m256d test_mm256_undefined_pd() {
   return _mm256_undefined_pd();
 }
 
 __m256i test_mm256_undefined_si256() {
   // CHECK-LABEL: @test_mm256_undefined_si256
   // CHECK: ret <4 x i64> undef
   return _mm256_undefined_si256();
 }
+
+double test_mm256_cvtsd_f64(__m256d __a)
+{
+  // CHECK-LABEL: @test_mm256_cvtsd_f64
+  // CHECK: extractelement <4 x double> %{{.*}}, i32 0
+  return _mm256_cvtsd_f64(__a);
+}
+
+int test_mm256_cvtsi256_si32(__m256i __a)
+{
+  // CHECK-LABEL: @test_mm256_cvtsi256_si32
+  // CHECK: extractelement <8 x i32> %{{.*}}, i32 0
+  return _mm256_cvtsi256_si32(__a);
+}
+
+float test_mm256_cvtss_f32(__m256 __a)
+{
+  // CHECK-LABEL: @test_mm256_cvtss_f32
+  // CHECK: extractelement <8 x float> %{{.*}}, i32 0
+  return _mm256_cvtss_f32(__a);
+}
 No newline at end of file

test/CodeGen/bitscan-builtins.c

+// RUN: %clang_cc1 -triple x86_64-unknown-unknown -target-feature +rdrnd -target-feature +rdseed -emit-llvm -o - %s | FileCheck %s
+
+// Don't include mm_malloc.h, it's system specific.
+#define __MM_MALLOC_H
+#include <immintrin.h>
+
+int test_bit_scan_forward(int a) {
+  return _bit_scan_forward(a);
+// CHECK: @test_bit_scan_forward
+// CHECK: call i32 @llvm.x86.bit.scan.forward
+}
+
+int test_bit_scan_reverse(int a) {
+  return _bit_scan_reverse(a);
+// CHECK: @test_bit_scan_reverse
+// CHECK: call i32 @llvm.x86.bit.scan.reverse
+}
 No newline at end of file

test/CodeGen/rd-builtins.c

+// RUN: %clang_cc1 -triple x86_64-unknown-unknown -target-feature +rdrnd -target-feature +rdseed -emit-llvm -o - %s | FileCheck %s
+
+// Don't include mm_malloc.h, it's system specific.
+#define __MM_MALLOC_H
+
+#include <x86intrin.h>
+
+unsigned long long test_rdpmc(int a) {
+  return _rdpmc(a);
+// CHECK: @test_rdpmc
+// CHECK: call i64 @llvm.x86.rdpmc
+}
+
+int test_rdtsc() {
+  return _rdtsc();
+// CHECK: @test_rdtsc
+// CHECK: call i64 @llvm.x86.rdtsc
+}
 No newline at end of file