gwenn
fcf106e685
ssse3 ( #224 )
...
* ssse3: _mm_abs_pi8 failing
Intrinsic has incorrect return type!
<8 x i8> (<8 x i8>)* @llvm.x86.ssse3.pabs.b
* Introduce a x86_mmx type
And make it compatible with i8x8 and u8x8.
Alex suggested to change the i8x8 declaration as:
```
struct i8x8(i64);
```
But I don't see how to make it compatible with the
existing code/macros.
* ssse3: _mm_abs_pi16, _mm_abs_pi32, _mm_shuffle_pi8
* ssse3: _mm_abs_pi16, _mm_abs_pi32, _mm_shuffle_pi8 tests
* Replace x86_mmx by __m64
* ssse3: _mm_sign_pi8, _mm_sign_pi16, _mm_sign_pi32
* ssse3: _mm_mulhrs_pi16
* ssse3: _mm_maddubs_pi16
* ssse3: _mm_hsub_pi16, _mm_hsub_pi32, _mm_hsubs_pi16
* ssse3: _mm_hadd_pi16, _mm_hadd_pi32, _mm_hadds_pi16
* Move some ssse3 intrinsics from i586 to i686
2017-12-03 11:53:36 -06:00
gnzlbg
6461312210
[ci] test i686-apple-darwin ( #221 )
...
* [ci] test i686-apple-darwin
* fix overflow on i686-apple-darwin
2017-11-28 17:09:38 -07:00
gnzlbg
8a92a566c9
[sse] _mm_stream_{ps,pi} ( #219 )
2017-11-28 07:48:26 -08:00
gnzlbg
288a30a93e
add mmx module, mmx run-time detection, intrinsics ( #220 )
...
* [sse] _mm_cvtps_pi32, _mm_cvt_ps2pi
* [mmx] run-time detection support
* [x86] add mmx module
* [x86] make __m64 public
* [sse] add _mm_cvtps_pi{8,16}, _mm_cvttps_pi32, _mm_cvtt_ps2pi
* move new intrinsics from i586 to i686 module
* mmx requires i686
2017-11-28 07:45:41 -08:00
gnzlbg
ef847ac83b
[sse] add _mm_{min, max}_{pi16, pu8} ( #218 )
...
* [sse] add _mm_{min, max}_{pi16, pu8}
* format docs
2017-11-27 14:54:28 -08:00
gnzlbg
b8a4b397ad
update docs ( #217 )
...
* update docs
* cargo clean deletes previous docs
* remove stdsimd from coresimd examples
* use stdsimd instead of coresimd in core docs
* add stdsimd as a dev-dependency of coresimd
2017-11-27 10:47:23 -08:00
Tony Sifkarovski
40a0b1cc92
[avx2] add shuffle, insert/extract i128, permute* ( #210 )
...
* [x86][avx2] add _mm256_shuffle{hi,lo}_epi16
* [x86][avx2] add _mm256_{insert,extract}i128_si256
* [x86][avx2] add remaining permute intrinsics
2017-11-26 17:40:26 +01:00
gnzlbg
426621f021
Add FXSAVE/FXRSTOR, update Intel SDE, fix xsave tests ( #205 )
...
* [x86] add run-time detection for fxsr
* [x86] add i386 fxsr intrinsics: FXSAVE,FXRSTOR
* [x86_64] add x86_64 fxsr intrinsics: FXSAVE64/FXRSTOR64
* [x86-runtime]: document xsave detection further
* [x86] disable xsaves and xsaves64 tests
2017-11-22 15:25:15 +01:00
gnzlbg
20529701d8
Fix clippy and rust-fmt.
2017-11-22 13:42:58 +01:00
Alex Crichton
922345c005
Use workspaces and fix tests
...
* Enable a Cargo workspace for the repo
* Disable tests for proc-macro crates
* Move back to mounting source directory read-only
* Refactor test invocation to only test one crate with `--all`
2017-11-22 13:42:58 +01:00
gnzlbg
86fa377cea
Only coresimd depends on stdsimd-test.
2017-11-22 13:42:58 +01:00
gnzlbg
cb9888f802
[ci] flag the documentation build bot
2017-11-22 13:42:58 +01:00
gnzlbg
b940d3311a
fix doc script
2017-11-22 13:42:58 +01:00
gnzlbg
6a0a55f01a
c_void -> *mut u8
2017-11-22 13:42:58 +01:00
gnzlbg
14d0903309
refactor no_std components into the coresimd crate
2017-11-22 13:42:58 +01:00
Adam Niederer
dc9f076480
Add AVX2 gathers ( #202 )
...
* Add _mm_[mask_]gatheri32_epi32
* Add _mm[256][_mask]_i32gather_{epi64, pd}
* Add _mm[256][_mask]_gather_ps
* Add _mm[256][_mask]_i64gather_{epi32, epi64, ps, pd}
2017-11-22 09:26:12 +01:00
gnzlbg
2faf11ab44
[readme] point always to latests docs ( #206 )
2017-11-21 15:05:46 -06:00
gnzlbg
0129d3be76
[nvptx] enable nvptx only when all other targets are disabled ( #208 )
...
Closes #207 .
2017-11-21 15:05:05 -06:00
Alex Crichton
8356754fe7
Fix hygiene in various macros ( #204 )
2017-11-21 12:54:06 -06:00
gnzlbg
bd629147a1
Upgrade to cupid 0.0.5 and cleanup duplicated code in x86 run-time ( #203 )
...
* [ci] upgrade to cupid 0.0.5
* [runtime x86] cleanup duplicated code
2017-11-21 08:46:36 -06:00
Jonathan Goodman
f236ef8f6b
Fix comments for _mm_cvtepu8_epi{32, 64} ( #200 )
2017-11-20 16:55:48 -06:00
Alex Crichton
738312d17c
Unconditionally flag as #![no_std] ( #196 )
...
This is more idiomatic for no-std-compatible crates where imports are
unconditionally rewritten to `core` and then only when necessary `std` is pulled
in explicitly.
2017-11-19 19:53:02 +01:00
gnzlbg
0d11a78a0e
refactor the x86 module ( #195 )
...
* refactor the x86 module
* document the i686 check
* document strict and intel_sde feature
* document nvptx module
2017-11-19 19:51:53 +01:00
gnzlbg
ff1b88d721
[clippy] fix missing doc on pub item
2017-11-17 17:41:23 +01:00
gnzlbg
ceef91aaba
[arm] runtime-detection support
2017-11-17 17:41:23 +01:00
gnzlbg
fe7da57403
[ci] add intel_sde feature
2017-11-17 17:41:23 +01:00
gnzlbg
9e7242ecad
[cpuid] Improve docs, implement __get_cpuid_max
...
Closes #174 .
2017-11-17 17:41:23 +01:00
gnzlbg
00cf3c05eb
[x86] cleanup run-time; add SSE4a, AVX-512, and xsave
2017-11-17 17:41:23 +01:00
gnzlbg
9ac630245d
[x86] implement cpuid intrinsics
2017-11-17 17:41:23 +01:00
gnzlbg
2fc4c25972
[x86] implement __read/write eflags
2017-11-17 17:41:23 +01:00
gnzlbg
05dd98c643
[x86] implement xsave intrinsics
2017-11-17 17:41:23 +01:00
gnzlbg
7c593d9857
[stdsimd-test] testing conditional on more than one feature
2017-11-17 17:41:23 +01:00
gnzlbg
2136214934
add support for no_std
2017-11-17 16:52:05 +01:00
gnzlbg
fda2ead377
add nvptx architecture
2017-11-17 16:52:05 +01:00
André Oliveira
33e26c0b4a
Formatting
2017-11-17 16:42:35 +01:00
André Oliveira
093029a6c3
Make test intrinsics use __m128i
2017-11-17 16:42:35 +01:00
André Oliveira
9d8c2639c1
Change _mm_mpsadbw_epu8 to work with unsigned integers
2017-11-17 16:42:35 +01:00
André Oliveira
7b0e7c6f52
Add _mm_mpsadbw_epu8
2017-11-17 16:42:35 +01:00
André Oliveira
b85d4f799c
Add _mm_minpos_epu16
2017-11-17 16:42:35 +01:00
André Oliveira
b67e9dfe5d
Add _mm_test_all_zeros, _mm_test_all_ones and _mm_test_mix_ones_zeros
2017-11-17 16:42:35 +01:00
André Oliveira
93c76381b7
Add documentation for testz, testc and testnzc
2017-11-17 16:42:35 +01:00
André Oliveira
4ce80f138b
Add _mm_testz_si128, _mm_testc_si128 and _mm_testnzc_si128
...
This should work for any 128 bit sized vector, but it only accepts i64x2 for now
2017-11-17 16:42:35 +01:00
André Oliveira
38f6087b9a
Add _mm_mul_epi32 and _mm_mullo_epi32
2017-11-17 16:42:35 +01:00
André Oliveira
613cacb317
Add remaining _mm_cvtep* intrinsics
2017-11-17 16:42:35 +01:00
André Oliveira
ac11d6941d
Add _mm_cvtepu8_epi{16, 32, 64}
2017-11-17 16:42:35 +01:00
André Oliveira
48027e994b
Add _mm_cvtepi32_epi64 and fix typo
2017-11-17 16:42:35 +01:00
Tony Sifkarovski
60c2608cce
[avx2] add _mm_256_cvtepu{8,16,32}_epi{16,32,64} ( #192 )
2017-11-17 09:22:18 +01:00
crypto-universe
1842e36d00
[x86][sse4.1] Add phminposuw & pmul* instructions
...
pmulld is implemented via multiplication.
2017-11-16 07:12:14 -05:00
gnzlbg
955fd849ff
implement missing std::ops
2017-11-13 06:42:49 -05:00
gnzlbg
6ed424a848
syn API breaking change ( #189 )
2017-11-11 23:35:00 +01:00
