r307524 - X86 Intrinsics: _bit_scan_forward should not be under #ifdef __RDRND__

2017-07-10 Thread Zvi Rackover via cfe-commits 
Author: zvi
Date: Mon Jul 10 00:13:56 2017
New Revision: 307524

URL: http://llvm.org/viewvc/llvm-project?rev=307524=rev
Log:
X86 Intrinsics: _bit_scan_forward should not be under #ifdef __RDRND__

Summary:
The _bit_scan_forward and _bit_scan_reverse intrinsics were accidentally
masked under the preprocessor checks that prune intrinsics definitions for the
benefit of faster compile-time on Windows. This patch moves the
definitons out of that region.

Fixes pr33722

Reviewers: craig.topper, aaboud, thakis

Reviewed By: craig.topper

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D35184

Modified:
cfe/trunk/lib/Headers/immintrin.h
cfe/trunk/test/CodeGen/bitscan-builtins.c

Modified: cfe/trunk/lib/Headers/immintrin.h
URL: 
http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Headers/immintrin.h?rev=307524=307523=307524=diff
==
--- cfe/trunk/lib/Headers/immintrin.h (original)
+++ cfe/trunk/lib/Headers/immintrin.h Mon Jul 10 00:13:56 2017
@@ -212,6 +212,15 @@ _rdrand32_step(unsigned int *__p)
   return __builtin_ia32_rdrand32_step(__p);
 }
 
+#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
+#endif /* __RDRND__ */
+
 /* __bit_scan_forward */
 static __inline__ int __attribute__((__always_inline__, __nodebug__))
 _bit_scan_forward(int __A) {
@@ -224,15 +233,6 @@ _bit_scan_reverse(int __A) {
   return 31 - __builtin_clz(__A);
 }
 
-#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
-#endif /* __RDRND__ */
-
 #if !defined(_MSC_VER) || __has_feature(modules) || defined(__FSGSBASE__)
 #ifdef __x86_64__
 static __inline__ unsigned int __attribute__((__always_inline__, __nodebug__, 
__target__("fsgsbase")))

Modified: cfe/trunk/test/CodeGen/bitscan-builtins.c
URL: 
http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGen/bitscan-builtins.c?rev=307524=307523=307524=diff
==
--- cfe/trunk/test/CodeGen/bitscan-builtins.c (original)
+++ cfe/trunk/test/CodeGen/bitscan-builtins.c Mon Jul 10 00:13:56 2017
@@ -1,5 +1,8 @@
 // RUN: %clang_cc1 -ffreestanding -triple x86_64-unknown-unknown -emit-llvm -o 
- %s | FileCheck %s
 
+// PR33722
+// RUN: %clang_cc1 -ffreestanding -triple x86_64-unknown-unknown -D_MSC_VER 
-emit-llvm -o - %s | FileCheck %s
+
 #include 
 
 int test_bit_scan_forward(int a) {


___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D26306: [AVX-512] Make VBMI instruction set enabling imply that the BWI instruction set is also enabled.

2016-11-05 Thread Zvi Rackover via cfe-commits 
zvi added inline comments.



Comment at: lib/Basic/Targets.cpp:3353
   setSSELevel(Features, AVX512F, Enabled);
+// Enable BWI instruction is VBMI is being enabled.
+if (Name == "avx512vbmi" && Enabled)

is -> if



Comment at: test/Preprocessor/x86_target_features.c:212
 
+// RUN: %clang -target i386-unknown-unknown -march=atom -mavx512vbmi 
-mno-avx512vbmi -x c -E -dM -o - %s | FileCheck -match-full-lines 
--check-prefix=AVX512VBMINOAVX512BW %s
+

Test is turning on and then off the same feature. Did you mean -mavx512bw 
-mno-avx512vbmi ?

If the latter, I would prefer the compiler shout at me for providing 
contradicting flags than let the last flag win. Consider enforcing this in 
handleTargetFeatures().




https://reviews.llvm.org/D26306



___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D26306: [AVX-512] Make VBMI instruction set enabling imply that the BWI instruction set is also enabled.

2016-11-05 Thread Zvi Rackover via cfe-commits 
zvi accepted this revision.
zvi added a comment.
This revision is now accepted and ready to land.

LGTM


https://reviews.llvm.org/D26306



___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

r273095 - [X86] _MM_ALIGN16 attribute support for non-windows targets

2016-06-18 Thread Zvi Rackover via cfe-commits 
Author: zvi
Date: Sat Jun 18 15:01:07 2016
New Revision: 273095

URL: http://llvm.org/viewvc/llvm-project?rev=273095=rev
Log:
[X86]  _MM_ALIGN16 attribute support for non-windows targets

Summary:
This patch adds support for the _MM_ALIGN16 attribute on non-windows targets. 
This aligns Clang with ICC which supports the attribute on all targets.

Fixes PR28056

Reviewers: aaboud, echristo, cfe-commits, mkuper

Subscribers: zvi, mehdi_amini

Projects: #clang-c

Differential Revision: http://reviews.llvm.org/D21173

Modified:
cfe/trunk/lib/Headers/xmmintrin.h
cfe/trunk/test/Headers/xmmintrin.c

Modified: cfe/trunk/lib/Headers/xmmintrin.h
URL: 
http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Headers/xmmintrin.h?rev=273095=273094=273095=diff
==
--- cfe/trunk/lib/Headers/xmmintrin.h (original)
+++ cfe/trunk/lib/Headers/xmmintrin.h Sat Jun 18 15:01:07 2016
@@ -2821,9 +2821,7 @@ _mm_movemask_ps(__m128 __a)
 }
 
 
-#ifdef _MSC_VER
-#define _MM_ALIGN16 __declspec(align(16))
-#endif
+#define _MM_ALIGN16 __attribute__((aligned(16)))
 
 #define _MM_SHUFFLE(z, y, x, w) (((z) << 6) | ((y) << 4) | ((x) << 2) | (w))
 

Modified: cfe/trunk/test/Headers/xmmintrin.c
URL: 
http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Headers/xmmintrin.c?rev=273095=273094=273095=diff
==
--- cfe/trunk/test/Headers/xmmintrin.c (original)
+++ cfe/trunk/test/Headers/xmmintrin.c Sat Jun 18 15:01:07 2016
@@ -7,6 +7,9 @@
 // REQUIRES: x86-registered-target
 #include 
 
+// CHECK: @c = common global i8 0, align 16
+_MM_ALIGN16 char c;
+
 // Make sure the last step of _mm_cvtps_pi16 converts <4 x i32> to <4 x i16> by
 // checking that clang emits PACKSSDW instead of PACKSSWB.
 


___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Re: [PATCH] D21173: [X86] _MM_ALIGN16 attribute support for non-windows targets

2016-06-09 Thread Zvi Rackover via cfe-commits 
zvi updated this revision to Diff 60217.
zvi added a comment.

Following David Majnemer's suggestion on cfe-commits: I would just use the 
__attribute__ spelling, no need for two definitions


Repository:
  rL LLVM

http://reviews.llvm.org/D21173

Files:
  lib/Headers/xmmintrin.h
  test/Headers/xmmintrin.c

Index: test/Headers/xmmintrin.c
===
--- test/Headers/xmmintrin.c
+++ test/Headers/xmmintrin.c
@@ -7,6 +7,9 @@
 // REQUIRES: x86-registered-target
 #include 
 
+// CHECK: @c = common global i8 0, align 16
+_MM_ALIGN16 char c;
+
 // Make sure the last step of _mm_cvtps_pi16 converts <4 x i32> to <4 x i16> by
 // checking that clang emits PACKSSDW instead of PACKSSWB.
 
Index: lib/Headers/xmmintrin.h
===
--- lib/Headers/xmmintrin.h
+++ lib/Headers/xmmintrin.h
@@ -2821,9 +2821,7 @@
 }
 
 
-#ifdef _MSC_VER
-#define _MM_ALIGN16 __declspec(align(16))
-#endif
+#define _MM_ALIGN16 __attribute__((aligned(16)))
 
 #define _MM_SHUFFLE(z, y, x, w) (((z) << 6) | ((y) << 4) | ((x) << 2) | (w))
 


Index: test/Headers/xmmintrin.c
===
--- test/Headers/xmmintrin.c
+++ test/Headers/xmmintrin.c
@@ -7,6 +7,9 @@
 // REQUIRES: x86-registered-target
 #include 
 
+// CHECK: @c = common global i8 0, align 16
+_MM_ALIGN16 char c;
+
 // Make sure the last step of _mm_cvtps_pi16 converts <4 x i32> to <4 x i16> by
 // checking that clang emits PACKSSDW instead of PACKSSWB.
 
Index: lib/Headers/xmmintrin.h
===
--- lib/Headers/xmmintrin.h
+++ lib/Headers/xmmintrin.h
@@ -2821,9 +2821,7 @@
 }
 
 
-#ifdef _MSC_VER
-#define _MM_ALIGN16 __declspec(align(16))
-#endif
+#define _MM_ALIGN16 __attribute__((aligned(16)))
 
 #define _MM_SHUFFLE(z, y, x, w) (((z) << 6) | ((y) << 4) | ((x) << 2) | (w))
 
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D21173: [X86] _MM_ALIGN16 attribute support for non-windows targets

2016-06-09 Thread Zvi Rackover via cfe-commits 
zvi created this revision.
zvi added reviewers: aaboud, mkuper, echristo, cfe-commits.
zvi set the repository for this revision to rL LLVM.
zvi added a project: clang-c.
Herald added a subscriber: mehdi_amini.

This patch adds support for the _MM_ALIGN16 attribute on non-windows targets. 
This aligns Clang with ICC which supports the attribute on all targets.

Fixes PR28056

Repository:
  rL LLVM

http://reviews.llvm.org/D21173

Files:
  lib/Headers/xmmintrin.h
  test/Headers/xmmintrin.c

Index: test/Headers/xmmintrin.c
===
--- test/Headers/xmmintrin.c
+++ test/Headers/xmmintrin.c
@@ -7,6 +7,9 @@
 // REQUIRES: x86-registered-target
 #include 
 
+// CHECK: @c = common global i8 0, align 16
+_MM_ALIGN16 char c;
+
 // Make sure the last step of _mm_cvtps_pi16 converts <4 x i32> to <4 x i16> by
 // checking that clang emits PACKSSDW instead of PACKSSWB.
 
Index: lib/Headers/xmmintrin.h
===
--- lib/Headers/xmmintrin.h
+++ lib/Headers/xmmintrin.h
@@ -2823,6 +2823,8 @@
 
 #ifdef _MSC_VER
 #define _MM_ALIGN16 __declspec(align(16))
+#else
+#define _MM_ALIGN16 __attribute__((aligned(16)))
 #endif
 
 #define _MM_SHUFFLE(z, y, x, w) (((z) << 6) | ((y) << 4) | ((x) << 2) | (w))


Index: test/Headers/xmmintrin.c
===
--- test/Headers/xmmintrin.c
+++ test/Headers/xmmintrin.c
@@ -7,6 +7,9 @@
 // REQUIRES: x86-registered-target
 #include 
 
+// CHECK: @c = common global i8 0, align 16
+_MM_ALIGN16 char c;
+
 // Make sure the last step of _mm_cvtps_pi16 converts <4 x i32> to <4 x i16> by
 // checking that clang emits PACKSSDW instead of PACKSSWB.
 
Index: lib/Headers/xmmintrin.h
===
--- lib/Headers/xmmintrin.h
+++ lib/Headers/xmmintrin.h
@@ -2823,6 +2823,8 @@
 
 #ifdef _MSC_VER
 #define _MM_ALIGN16 __declspec(align(16))
+#else
+#define _MM_ALIGN16 __attribute__((aligned(16)))
 #endif
 
 #define _MM_SHUFFLE(z, y, x, w) (((z) << 6) | ((y) << 4) | ((x) << 2) | (w))
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits