This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH 1/3][rs6000] x86-compat vector intrinsics fixes for BE, 32bit
- From: Segher Boessenkool <segher at kernel dot crashing dot org>
- To: Paul Clarke <pc at us dot ibm dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Tue, 4 Dec 2018 14:16:04 -0600
- Subject: Re: [PATCH 1/3][rs6000] x86-compat vector intrinsics fixes for BE, 32bit
- References: <fd7571d1-1ce6-be2c-b474-116485a9674c@us.ibm.com>
Hi!
On Tue, Dec 04, 2018 at 08:59:03AM -0600, Paul Clarke wrote:
> Fix general endian and 32-bit mode issues found in the
> compatibility implementations of the x86 vector intrinsics when running the
> associated test suite tests. (The tests had been inadvertently made to PASS
> without actually running the test code. A later patch fixes this issue.)
>
> In a few cases, the opportunity was taken to update the vector API used in
> the implementations to the preferred functions from the
> OpenPOWER 64-Bit ELF V2 ABI Specification.
>
> [gcc]
>
> 2018-12-03 Paul A. Clarke <pc@us.ibm.com>
>
> PR target/88316
> * config/rs6000/mmintrin.h (_mm_unpackhi_pi8): Fix for big-endian.
> (_mm_unpacklo_pi8): Likewise.
> (_mm_mulhi_pi16): Likewise.
> (_mm_packs_pi16): Fix for big-endian. Use preferred API.
> (_mm_packs_pi32): Likewise.
> (_mm_packs_pu16): Likewise.
> * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix for big-endian.
> (_mm_cvtss_si64): Likewise.
> (_mm_cvtpi32x2_ps): Likewise.
> (_mm_shuffle_ps): Likewise.
> (_mm_movemask_pi8): Likewise.
> (_mm_mulhi_pu16): Likewise.
> (_mm_sad_pu8): Likewise.
> (_mm_sad_pu8): Likewise.
> (_mm_cvtpu16_ps): Fix for big-endian. Use preferred API.
> (_mm_cvtpu8_ps): Likewise.
> * config/rs6000/emmintrin.h (_mm_movemask_pd): Fix for big-endian.
> (_mm_mul_epu32): Likewise.
> (_mm_bsrli_si128): Likewise.
> (_mm_movemask_epi8): Likewise.
> (_mm_shufflehi_epi16): Likewise.
> (_mm_shufflelo_epi16): Likewise.
> (_mm_shuffle_epi32): Likewise.
> * config/rs6000/pmmintrin.h (_mm_hadd_ps): Fix for big-endian.
> (_mm_sub_ps): Likewise.
> * config/rs6000/mmintrin.h (_mm_cmpeq_pi8): Fix for 32-bit mode.
> @@ -1612,7 +1608,8 @@ _mm_bsrli_si128 (__m128i __A, const int __N)
> const __v16qu zeros = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
>
> if (__N < 16)
> - if (__builtin_constant_p(__N))
> + if (__builtin_constant_p(__N) &&
> + __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
Please just use __LITTLE_ENDIAN__, as the rest of these files already does.
(More times in this patch; also BIG).
Okay for trunk with that fixed. Thanks!
Do you have new testcases, too? Or is all this caught by existing
testcases?
Segher