This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[PATCH, rs6000] Use unaligned vector types for some pointer casts
- From: Bill Schmidt <wschmidt at linux dot ibm dot com>
- To: GCC Patches <gcc-patches at gcc dot gnu dot org>
- Cc: Segher Boessenkool <segher at kernel dot crashing dot org>
- Date: Fri, 19 Oct 2018 16:27:27 -0500
- Subject: [PATCH, rs6000] Use unaligned vector types for some pointer casts
Hi,
The x86 intrinsic compatibility headers contain a couple of instances of
undefined behavior where a cast to an aligned type is used when that
alignment is not guaranteed by the expression to be cast from. This
patch fixes that problem by replacing the aligned types with unaligned
versions of the same type.
Bootstrapped and tested on powerpc64le-linux-gnu with no regressions.
Is this okay for trunk?
(I also cleaned up a badly formatted comment in the neighborhood.)
Thanks,
Bill
2018-10-19 Bill Schmidt <wschmidt@linux.ibm.com>
Jinsong Ji <jji@us.ibm.com>
* config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
(_mm_store_pd): Use unaligned vector type for pointer cast.
(_mm_maskmoveu_si128): Likewise.
* config/rs6000/xmmintrin.h (__m128_u): New typedef.
(_mm_store_ps): Use unaligned vector type for pointer cast.
Index: gcc/config/rs6000/emmintrin.h
===================================================================
--- gcc/config/rs6000/emmintrin.h (revision 265318)
+++ gcc/config/rs6000/emmintrin.h (working copy)
@@ -85,7 +85,7 @@ typedef double __m128d __attribute__ ((__vector_si
typedef long long __m128i_u __attribute__ ((__vector_size__ (16), __may_alias__, __aligned__ (1)));
typedef double __m128d_u __attribute__ ((__vector_size__ (16), __may_alias__, __aligned__ (1)));
-/* Define two value permute mask */
+/* Define two value permute mask. */
#define _MM_SHUFFLE2(x,y) (((x) << 1) | (y))
/* Create a vector with element 0 as F and the rest zero. */
@@ -201,7 +201,7 @@ _mm_store_pd (double *__P, __m128d __A)
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_storeu_pd (double *__P, __m128d __A)
{
- *(__m128d *)__P = __A;
+ *(__m128d_u *)__P = __A;
}
/* Stores the lower DPFP value. */
@@ -2175,7 +2175,7 @@ _mm_maskmoveu_si128 (__m128i __A, __m128i __B, cha
{
__v2du hibit = { 0x7f7f7f7f7f7f7f7fUL, 0x7f7f7f7f7f7f7f7fUL};
__v16qu mask, tmp;
- __m128i *p = (__m128i*)__C;
+ __m128i_u *p = (__m128i_u*)__C;
tmp = (__v16qu)_mm_loadu_si128(p);
mask = (__v16qu)vec_cmpgt ((__v16qu)__B, (__v16qu)hibit);
Index: gcc/config/rs6000/xmmintrin.h
===================================================================
--- gcc/config/rs6000/xmmintrin.h (revision 265318)
+++ gcc/config/rs6000/xmmintrin.h (working copy)
@@ -85,6 +85,9 @@
vector types, and their scalar components. */
typedef float __m128 __attribute__ ((__vector_size__ (16), __may_alias__));
+/* Unaligned version of the same type. */
+typedef float __m128_u __attribute__ ((__vector_size__ (16), __may_alias__));
+
/* Internal data types for implementing the intrinsics. */
typedef float __v4sf __attribute__ ((__vector_size__ (16)));
@@ -172,7 +175,7 @@ _mm_store_ps (float *__P, __m128 __A)
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_storeu_ps (float *__P, __m128 __A)
{
- *(__m128 *)__P = __A;
+ *(__m128_u *)__P = __A;
}
/* Store four SPFP values in reverse order. The address must be aligned. */