[X86] Add gcc rotate intrinsics to ia32intrin.h

[craig.topper]

Summary

This is another attempt at what Erich Keane tried to do in r355322.a

This adds rolb, rolw, rold, rolq and their ror equivalent as always_inline wrappers around __builtin_rotate* which will lower to funnel shift intrinsics in IR.

Additionally, when _MSC_VER is not defined we will define _rotl, _lrotl, _rotr, _lrotr as macros to one of the always_inline intrinsics mentioned above. Making sure that _lrotl/_lrotr use either 32 or 64 bit based on the size of long. These need to be macros because we have builtins with the same name for MS compatibility, but _MSC_VER isn't always defined when those builtins are enabled.

We also define _rotwl and _rotwr as macros aliasing to rolw/rorw just like gcc to complete the set. These don't need to be gated with _MSC_VER because these aren't MS builtins.

I've added tests both for non-MS and -ms-extensions with and without _MSC_VER being defined.

[craig.topper]

Add the test file

[craig.topper]

Fix bad comment copy/paste

[rnk]

Looks good to me. Let's wait for @jyknight though.

lib/Headers/ia32intrin.h

120

Can you add comments here indicating that these are already provided as builtins for MSVC, so these macros are unnecessary?

133

And, add a comment here indicating that these are not builtins, so they are provided in all modes.

[craig.topper]

Add comments based on rnk's review

[jyknight]

Looks good.

lib/Headers/ia32intrin.h

 
 #define _rdpmc(A) __rdpmc(A)
 
 static __inline__ void __attribute__((__always_inline__, __nodebug__))
 _wbinvd(void) {
   __builtin_ia32_wbinvd();
 }
 
+static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
+__rolb(unsigned char __X, int __C) {
+  return __builtin_rotateleft8(__X, __C);
+}
+
+static __inline__ unsigned char __attribute__((__always_inline__, __nodebug__))
+__rorb(unsigned char __X, int __C) {
+  return __builtin_rotateright8(__X, __C);
+}
+
+static __inline__ unsigned short __attribute__((__always_inline__, __nodebug__))
+__rolw(unsigned short __X, int __C) {
+  return __builtin_rotateleft16(__X, __C);
+}
+
+static __inline__ unsigned short __attribute__((__always_inline__, __nodebug__))
+__rorw(unsigned short __X, int __C) {
+  return __builtin_rotateright16(__X, __C);
+}
+
+static __inline__ unsigned int __attribute__((__always_inline__, __nodebug__))
+__rold(unsigned int __X, int __C) {
+  return __builtin_rotateleft32(__X, __C);
+}
+
+static __inline__ unsigned int __attribute__((__always_inline__, __nodebug__))
+__rord(unsigned int __X, int __C) {
+  return __builtin_rotateright32(__X, __C);
+}
+
+#ifdef __x86_64__
+static __inline__ unsigned long long __attribute__((__always_inline__, __nodebug__))
+__rolq(unsigned long long __X, int __C) {
+  return __builtin_rotateleft64(__X, __C);
+}
+
+static __inline__ unsigned long long __attribute__((__always_inline__, __nodebug__))
+__rorq(unsigned long long __X, int __C) {
+  return __builtin_rotateright64(__X, __C);
+}
+#endif /* __x86_64__ */
+
+#ifndef _MSC_VER
+/* These are already provided as builtins for MSVC. */
+/* Select the correct function based on the size of long. */
+#ifdef __LP64__
+#define _lrotl(a,b) __rolq((a), (b))
+#define _lrotr(a,b) __rorq((a), (b))
+#else
+#define _lrotl(a,b) __rold((a), (b))
+#define _lrotr(a,b) __rord((a), (b))
+#endif
+#define _rotl(a,b) __rold((a), (b))
+#define _rotr(a,b) __rord((a), (b))
+#endif // _MSC_VER
+
+/* These are not builtins so need to be provided in all modes. */
+#define _rotwl(a,b) __rolw((a), (b))
+#define _rotwr(a,b) __rorw((a), (b))
+
 #endif /* __IA32INTRIN_H */

test/CodeGen/rot-intrinsics.c

+// RUN: %clang_cc1 -ffreestanding -triple i686--linux -emit-llvm %s -o - | FileCheck %s --check-prefixes CHECK,CHECK-32BIT-LONG
+// RUN: %clang_cc1 -ffreestanding -triple x86_64--linux -emit-llvm %s -o - | FileCheck %s --check-prefixes CHECK,CHECK-64BIT-LONG
+// RUN: %clang_cc1 -fms-extensions -fms-compatibility -ffreestanding %s -triple=i686-windows-msvc -target-feature +sse2 -emit-llvm -o - -Wall -Werror | FileCheck %s --check-prefixes CHECK,CHECK-32BIT-LONG
+// RUN: %clang_cc1 -fms-extensions -fms-compatibility -ffreestanding %s -triple=x86_64-windows-msvc -target-feature +sse2 -emit-llvm -o - -Wall -Werror | FileCheck %s --check-prefixes CHECK,CHECK-32BIT-LONG
+// RUN: %clang_cc1 -fms-extensions -fms-compatibility -fms-compatibility-version=17.00 -ffreestanding %s -triple=i686-windows-msvc -target-feature +sse2 -emit-llvm -o - -Wall -Werror | FileCheck %s --check-prefixes CHECK,CHECK-32BIT-LONG
+// RUN: %clang_cc1 -fms-extensions -fms-compatibility -fms-compatibility-version=17.00 -ffreestanding %s -triple=x86_64-windows-msvc -target-feature +sse2 -emit-llvm -o - -Wall -Werror | FileCheck %s --check-prefixes CHECK,CHECK-32BIT-LONG
+
+#include <x86intrin.h>
+
+unsigned char test__rolb(unsigned char value, int shift) {
+// CHECK-LABEL: i8 @test__rolb
+// CHECK:   [[R:%.*]] = call i8 @llvm.fshl.i8(i8 [[X:%.*]], i8 [[X]], i8 [[Y:%.*]])
+// CHECK:   ret i8 [[R]]
+  return __rolb(value, shift);
+}
+
+unsigned short test__rolw(unsigned short value, int shift) {
+// CHECK-LABEL: i16 @test__rolw
+// CHECK:   [[R:%.*]] = call i16 @llvm.fshl.i16(i16 [[X:%.*]], i16 [[X]], i16 [[Y:%.*]])
+// CHECK:   ret i16 [[R]]
+  return __rolw(value, shift);
+}
+
+unsigned int test__rold(unsigned int value, int shift) {
+// CHECK-LABEL: i32 @test__rold
+// CHECK:   [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[X]], i32 [[Y:%.*]])
+// CHECK:   ret i32 [[R]]
+  return __rold(value, shift);
+}
+
+#if defined(__x86_64__)
+unsigned long test__rolq(unsigned long value, int shift) {
+// CHECK-LONG-LABEL: i64 @test__rolq
+// CHECK-LONG:   [[R:%.*]] = call i64 @llvm.fshl.i64(i64 [[X:%.*]], i64 [[X]], i64 [[Y:%.*]])
+// CHECK-LONG:   ret i64 [[R]]
+  return __rolq(value, shift);
+}
+#endif
+
+unsigned char test__rorb(unsigned char value, int shift) {
+// CHECK-LABEL: i8 @test__rorb
+// CHECK:   [[R:%.*]] = call i8 @llvm.fshr.i8(i8 [[X:%.*]], i8 [[X]], i8 [[Y:%.*]])
+// CHECK:   ret i8 [[R]]
+  return __rorb(value, shift);
+}
+
+unsigned short test__rorw(unsigned short value, int shift) {
+// CHECK-LABEL: i16 @test__rorw
+// CHECK:   [[R:%.*]] = call i16 @llvm.fshr.i16(i16 [[X:%.*]], i16 [[X]], i16 [[Y:%.*]])
+// CHECK:   ret i16 [[R]]
+  return __rorw(value, shift);
+}
+
+unsigned int test__rord(unsigned int value, int shift) {
+// CHECK-LABEL: i32 @test__rord
+// CHECK:   [[R:%.*]] = call i32 @llvm.fshr.i32(i32 [[X:%.*]], i32 [[X]], i32 [[Y:%.*]])
+// CHECK:   ret i32 [[R]]
+  return __rord(value, shift);
+}
+
+#if defined(__x86_64__)
+unsigned long test__rorq(unsigned long value, int shift) {
+// CHECK-LONG-LABEL: i64 @test__rorq
+// CHECK-LONG:   [[R:%.*]] = call i64 @llvm.fshr.i64(i64 [[X:%.*]], i64 [[X]], i64 [[Y:%.*]])
+// CHECK-LONG:   ret i64 [[R]]
+  return __rorq(value, shift);
+}
+#endif
+
+unsigned short test_rotwl(unsigned short value, int shift) {
+// CHECK-LABEL: i16 @test_rotwl
+// CHECK:   [[R:%.*]] = call i16 @llvm.fshl.i16(i16 [[X:%.*]], i16 [[X]], i16 [[Y:%.*]])
+// CHECK:   ret i16 [[R]]
+  return _rotwl(value, shift);
+}
+
+unsigned int test_rotl(unsigned int value, int shift) {
+// CHECK-LABEL: i32 @test_rotl
+// CHECK:   [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[X]], i32 [[Y:%.*]])
+// CHECK:   ret i32 [[R]]
+  return _rotl(value, shift);
+}
+
+unsigned long test_lrotl(unsigned long value, int shift) {
+// CHECK-32BIT-LONG-LABEL: i32 @test_lrotl
+// CHECK-32BIT-LONG:   [[R:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[X]], i32 [[Y:%.*]])
+// CHECK-32BIT-LONG:   ret i32 [[R]]
+//
+// CHECK-64BIT-LONG-LABEL: i64 @test_lrotl
+// CHECK-64BIT-LONG:   [[R:%.*]] = call i64 @llvm.fshl.i64(i64 [[X:%.*]], i64 [[X]], i64 [[Y:%.*]])
+// CHECK-64BIT-LONG:   ret i64 [[R]]
+  return _lrotl(value, shift);
+}
+
+
+unsigned short test_rotwr(unsigned short value, int shift) {
+// CHECK-LABEL: i16 @test_rotwr
+// CHECK:   [[R:%.*]] = call i16 @llvm.fshr.i16(i16 [[X:%.*]], i16 [[X]], i16 [[Y:%.*]])
+// CHECK:   ret i16 [[R]]
+  return _rotwr(value, shift);
+}
+
+unsigned int test_rotr(unsigned int value, int shift) {
+// CHECK-LABEL: i32 @test_rotr
+// CHECK:   [[R:%.*]] = call i32 @llvm.fshr.i32(i32 [[X:%.*]], i32 [[X]], i32 [[Y:%.*]])
+// CHECK:   ret i32 [[R]]
+  return _rotr(value, shift);
+}
+
+unsigned long test_lrotr(unsigned long value, int shift) {
+// CHECK-32BIT-LONG-LABEL: i32 @test_lrotr
+// CHECK-32BIT-LONG:   [[R:%.*]] = call i32 @llvm.fshr.i32(i32 [[X:%.*]], i32 [[X]], i32 [[Y:%.*]])
+// CHECK-32BIT-LONG:   ret i32 [[R]]
+//
+// CHECK-64BIT-LONG-LABEL: i64 @test_lrotr
+// CHECK-64BIT-LONG:   [[R:%.*]] = call i64 @llvm.fshr.i64(i64 [[X:%.*]], i64 [[X]], i64 [[Y:%.*]])
+// CHECK-64BIT-LONG:   ret i64 [[R]]
+  return _lrotr(value, shift);
+}
+