cd28decba9
Make sse2 functions const
2025-12-02 03:56:18 +05:30
7d4bcdd3fd
Make sse functions const
2025-12-02 03:56:18 +05:30
233574ace3
Make abm, bmi1, bmi2, bswap and tbm functions const
2025-12-02 03:56:18 +05:30
742e70fe8f
Modify #[simd_test] to enable const-testing
2025-12-02 03:56:17 +05:30
453377b0ff
Make some helpers const
2025-12-02 03:56:17 +05:30
b963d29852
Use ordered reduction intrinsics for integer reductions
...
only ordered intrinsics have implementation in rustc-const-eval
2025-12-02 03:56:15 +05:30
eb49e65e2a
Merge pull request #1952 from madhav-madhusoodanan/intrinsic-test-updates
...
`intrinsic-test`: Coalescing all intrinsic tests into a single run of the test-file binaries
2025-12-01 21:58:40 +00:00
d5c2590972
chore: logging to trace the time taken by binaries to complete a run
2025-12-02 02:22:26 +05:30
060b4d347c
feat: reduce the intrinsics coverage for x86_64 intrinsics
2025-12-02 02:22:26 +05:30
7215e078f5
feat: extract the delimiter to a const value for uniformity
2025-12-02 02:22:26 +05:30
998aff3dc5
feat: add equality checks to reduce computation spent on equal outputs
2025-12-02 02:22:26 +05:30
99fbd39a7d
fix: update the comparison algorithm for cleaner comparisons
2025-12-02 02:22:21 +05:30
09a67dfab0
feat: increase TEST_SAMPLE_INTRINSICS_PERCENTAGE of x86_64 to 30%
2025-12-02 02:20:28 +05:30
a472b5dcd9
chore: formatting checks
2025-12-02 02:20:28 +05:30
344acad879
feat: modified the compare function to check one huge output stream
...
instead of len(intrinsic) output streams
2025-12-02 02:20:28 +05:30
f0f0b8b222
feat: configure test-files to run all intrinsics in the same process
2025-12-02 02:20:28 +05:30
e282dc1fb0
Merge pull request #1969 from sayantn/fix-kadd
...
use `wrapping_add` for `_kadd` intrinsics
2025-12-01 19:21:50 +00:00
9f1724d3ac
use wrapping_add for _kadd intrinsics
2025-12-02 00:22:05 +05:30
21f714bf38
Merge pull request #1965 from folkertdev/s390x-shuffle
...
s390x: use `simd_shuffle!` macro
2025-11-30 10:46:01 +00:00
46efc1b356
Merge pull request #1935 from sayantn/safe-cpuid
...
Make `cpuid` safe and update docs
2025-11-30 02:39:24 +00:00
ac05d0d945
Merge pull request #1968 from sayantn/intrinsic-test
...
Format `f16` values correctly in intrinsic-test
2025-11-29 20:59:18 +00:00
3e8185f155
Make cpuid safe and update docs
2025-11-30 02:23:29 +05:30
ae78a301e0
fix intrinsic-test to use core_arch::arch rather than std::arch
2025-11-30 02:16:43 +05:30
8b35ef74a5
correct behavior of some intrinsics
2025-11-30 02:16:42 +05:30
0c0de0742a
Format f16 values normally, rather than use the bit patterns
2025-11-30 02:16:42 +05:30
00ee70d216
systematically exclude svml, mmx and mpx rather than having an ad-hoc list
2025-11-30 02:16:39 +05:30
f7808acb3e
Merge pull request #1963 from valadaptive/relaxed-minmax-docs
...
Correct the docs for WASM relaxed_simd's `relaxed_min`/`relaxed_max` operations
2025-11-28 15:13:33 +00:00
989aff20e0
Merge pull request #1966 from sayantn/saturation
...
Use generic SIMD intrinsics for cvtepi intrinsics
2025-11-28 10:27:36 +00:00
98a95454b0
Use generic SIMD intrinsics for AVX512BW cvtepi
2025-11-28 04:52:50 +05:30
4c49a7a22f
Refactor max/min intrinsics using a helper function
2025-11-28 03:36:44 +05:30
daa2cd340a
s390x: use simd_shuffle! macro
2025-11-27 22:36:54 +01:00
3bbbddf4b9
Merge pull request #1964 from CrooseGit/dev/reucru01/arm-intrinsics-debug
...
Fixes issue with arm intrinsics not compiling in dev profile
2025-11-27 17:23:42 +00:00
73ad94ab86
Provides work-around for vreinterpret inline fail
2025-11-27 14:06:27 +00:00
1c2a0aaa31
Fixes differences found between clang & rust
...
The intrinsics test was flagging differences in aarch64_be between rust in debug and clang in O2.
It was found that rust was correct in debug, but incorrect in release,
and in both cases were being compared against clang in O2 which was also
incorrect.
The vdot intrinsics were fixed and are now correct in rust for both
release and debug. However the vcmla ones could not be as the issue lies
with LLVM. Both the vdot and vcmla intrinsics were added to the skiplist
as clang is still incorrect for both.
LLVM issue: https://github.com/llvm/llvm-project/issues/166190
2025-11-27 14:06:27 +00:00
b3f67c73d0
Correct the docs for WASM relaxed_simd's min/max
2025-11-24 14:19:12 -05:00
d7860a5107
Fixes generator, neon intrinics now build in debug
...
This fixes build issues associated with failing LLVM const param assertions
2025-11-24 15:12:44 +00:00
2922cbdd47
Merge pull request #1961 from folkertdev/pmadd-correct-signedness
...
correct signedness of pmadd arguments
2025-11-17 15:52:10 +00:00
2788686995
Merge pull request #1957 from marcoieni/rename-default-branch-to-main
...
rename default branch to main
2025-11-17 15:02:03 +00:00
ac2d97254e
correct signedness of pmadd arguments
2025-11-17 16:01:02 +01:00
57436fe950
Merge pull request #1960 from maurer/remove-tme
...
aarch64: Remove withdrawn FEAT_TME
2025-11-17 14:29:49 +00:00
7f3fcb347b
Merge pull request #1956 from sayantn/fix-tests
...
correct some `#[simd_test]` attributes
2025-11-16 23:00:42 +00:00
8fe87e9623
correct some #[simd_test] attributes
2025-11-17 01:24:03 +05:30
0882a6e02a
aarch64: Remove withdrawn FEAT_TME
...
ARM has withdrawn FEAT_TME
https://developer.arm.com/documentation/102105/lb-05/
LLVM has also dropped support for enabling the feature.
2025-11-14 19:25:34 +00:00
d84c69548d
Merge pull request #1955 from sayantn/vector-shifts
...
Use SIMD intrinsics for vector shifts
2025-11-13 11:10:25 +00:00
84c44c449d
Merge pull request #1959 from folkertdev/ternary-logic-tests-2
...
improve ternary logic tests
2025-11-13 09:41:34 +00:00
0ab7c9e3da
Use SIMD intrinsics for vector shifts
2025-11-13 10:49:31 +05:30
849ace0685
Merge pull request #1953 from sayantn/masked-load-store
...
Use generic SIMD masked load/stores for avx512 masked load/stores
2025-11-12 12:54:10 +00:00
e94ac6b638
improve ternary logic tests
2025-11-11 17:04:11 +01:00
b85138793e
Merge pull request #1958 from folkertdev/ternary-logic-tests
...
add logic tests for ternarylogic
2025-11-11 11:21:57 +00:00
148a7509a1
add logic tests for ternarylogic
...
previously the output would just always be all zeroes
2025-11-11 11:30:14 +01:00
sayantn