[X86] Add __popcntd and __popcntq to ia32intrin.h to match gcc and icc. Remove popcnt feature flag from _popcnt32/_popcnt64 and move to ia32intrin.h to match gcc

[craig.topper]

Summary

gcc and icc both implement popcntd and popcntq which we did not. gcc doesn't seem to require a feature flag for the _popcnt32/_popcnt64 spelling and will use a libcall if its not supported.

[RKSimon]

IIRC we don't use libcalls for popcnt - we just expand

clang/lib/Headers/ia32intrin.h

35 ↗

(On Diff #191411)

Change this as we don't require POPCNT any more?

[craig.topper]

Update doxygen comments

[craig.topper]

In D59567#1436035, @RKSimon wrote:

IIRC we don't use libcalls for popcnt - we just expand

Yeah we expand. gcc uses libcall. I think icc also expands.

[RKSimon]

LGTM with some minor typos

lib/Headers/ia32intrin.h

36

calculate

57

calculate

lib/Headers/ia32intrin.h

 
 #ifndef __X86INTRIN_H
 #error "Never use <ia32intrin.h> directly; include <x86intrin.h> instead."
 #endif
 
 #ifndef __IA32INTRIN_H
 #define __IA32INTRIN_H
 
+/** Counts the number of bits in the source operand having a value of 1.
+ *
+ *  \headerfile <x86intrin.h>
+ *
+ *  This intrinsic corresponds to the <c> POPCNT </c> instruction or a
+ *  a sequence of arithmetic and logic ops to calculate it.
+ *
+ *  \param __A
+ *     An unsigned 32-bit integer operand.
+ *  \returns A 32-bit integer containing the number of bits with value 1 in the
+ *     source operand.
+ */
+static __inline__ int __attribute__((__always_inline__, __nodebug__))
+__popcntd(unsigned int __A)
+{
+  return __builtin_popcount(__A);
+}
+
+#define _popcnt32(A) __popcntd((A))
+
+#ifdef __x86_64__
+/** Counts the number of bits in the source operand having a value of 1.
+ *
+ *  \headerfile <x86intrin.h>
+ *
+ *  This intrinsic corresponds to the <c> POPCNT </c> instruction or a
+ *  a sequence of arithmetic and logic ops to calculate it.
+ *
+ *  \param __A
+ *     An unsigned 64-bit integer operand.
+ *  \returns A 64-bit integer containing the number of bits with value 1 in the
+ *     source operand.
+ */
+static __inline__ long long __attribute__((__always_inline__, __nodebug__))
+__popcntq(unsigned long long __A)
+{
+  return __builtin_popcountll(__A);
+}
+
+#define _popcnt64(A) __popcntq((A))
+#endif /* __x86_64__ */
+
 #ifdef __x86_64__
 static __inline__ unsigned long long __attribute__((__always_inline__, __nodebug__))
 __readeflags(void)
 {
   return __builtin_ia32_readeflags_u64();
 }
 
 static __inline__ void __attribute__((__always_inline__, __nodebug__))

lib/Headers/popcntintrin.h

 /// \returns A 32-bit integer containing the number of bits with value 1 in the
 ///    source operand.
 static __inline__ int __DEFAULT_FN_ATTRS
 _mm_popcnt_u32(unsigned int __A)
 {
   return __builtin_popcount(__A);
 }
 
-/// Counts the number of bits in the source operand having a value of 1.
-///
-/// \headerfile <x86intrin.h>
-///
-/// This intrinsic corresponds to the <c> POPCNT </c> instruction.
-///
-/// \param __A
-///    A signed 32-bit integer operand.
-/// \returns A 32-bit integer containing the number of bits with value 1 in the
-///    source operand.
-static __inline__ int __DEFAULT_FN_ATTRS
-_popcnt32(int __A)
-{
-  return __builtin_popcount(__A);
-}
-
 #ifdef __x86_64__
 /// Counts the number of bits in the source operand having a value of 1.
 ///
 /// \headerfile <x86intrin.h>
 ///
 /// This intrinsic corresponds to the <c> POPCNT </c> instruction.
 ///
 /// \param __A
 ///    An unsigned 64-bit integer operand.
 /// \returns A 64-bit integer containing the number of bits with value 1 in the
 ///    source operand.
 static __inline__ long long __DEFAULT_FN_ATTRS
 _mm_popcnt_u64(unsigned long long __A)
 {
   return __builtin_popcountll(__A);
 }
-
-/// Counts the number of bits in the source operand having a value of 1.
-///
-/// \headerfile <x86intrin.h>
-///
-/// This intrinsic corresponds to the <c> POPCNT </c> instruction.
-///
-/// \param __A
-///    A signed 64-bit integer operand.
-/// \returns A 64-bit integer containing the number of bits with value 1 in the
-///    source operand.
-static __inline__ long long __DEFAULT_FN_ATTRS
-_popcnt64(long long __A)
-{
-  return __builtin_popcountll(__A);
-}
 #endif /* __x86_64__ */
 
 #undef __DEFAULT_FN_ATTRS
 
 #endif /* __POPCNTINTRIN_H */

test/CodeGen/popcnt-builtins.c

-// RUN: %clang_cc1 -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +popcnt -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +popcnt -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK,CHECK-POPCNT
+// RUN: %clang_cc1 -ffreestanding %s -triple=x86_64-apple-darwin -emit-llvm -o - | FileCheck %s
 
 
-#include <immintrin.h>
+#include <x86intrin.h>
 
-unsigned int test_mm_popcnt_u32(unsigned int __X) {
-  //CHECK: call i32 @llvm.ctpop.i32
+#ifdef __POPCNT__
+int test_mm_popcnt_u32(unsigned int __X) {
+  //CHECK-POPCNT: call i32 @llvm.ctpop.i32
   return _mm_popcnt_u32(__X);
 }
+#endif
 
-unsigned int test_popcnt_32(int __X) {
+int test_popcnt32(unsigned int __X) {
   //CHECK: call i32 @llvm.ctpop.i32
   return _popcnt32(__X);
 }
 
-unsigned long long test_mm_popcnt_u64(unsigned long long __X) {
-  //CHECK: call i64 @llvm.ctpop.i64
+int test__popcntd(unsigned int __X) {
+  //CHECK: call i32 @llvm.ctpop.i32
+  return __popcntd(__X);
+}
+
+#ifdef __POPCNT__
+long long test_mm_popcnt_u64(unsigned long long __X) {
+  //CHECK-POPCNT: call i64 @llvm.ctpop.i64
   return _mm_popcnt_u64(__X);
 }
+#endif
 
-unsigned long long test_popcnt_64(long long __X) {
+long long test_popcnt64(unsigned long long __X) {
   //CHECK: call i64 @llvm.ctpop.i64
   return _popcnt64(__X);
 }
+
+long long test__popcntq(unsigned long long __X) {
+  //CHECK: call i64 @llvm.ctpop.i64
+  return __popcntq(__X);
+}