forked from qt/qtbase
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
CMake: update the x86 intrinsic checks
Merge all the existing checks into a single one, which is a simple pass or fail, since all our supported compilers support all the intrinsics up to Cannon Lake. The two I've recently added (AVX512VBMI2 and VAES) aren't yet supported everywhere, so they stay. For some reason, all intrinsics seem to be disabled on Android. It looks like some support was missing during the CMake port and this was never again looked at. I'm leaving it be. As for WASM, discussion with maintainers is that the WASM emulation of x86 intrinsics is too hit-and-miss. No one is testing the performance, particularly the person writing such code (me). They also have some non-obvious selection of what is supported natively and what is merely emulated. Using the actual WASM intrinsics is preferred, but someone else's job. Change-Id: Ib42b3adc93bf4d43bd55fffd16c10d66208e8384 Reviewed-by: Tor Arne Vestbø <[email protected]> Reviewed-by: Morten Johan Sørvig <[email protected]> Reviewed-by: Lorn Potter <[email protected]> Reviewed-by: Kai Koehne <[email protected]>
- Loading branch information
1 parent
500c116
commit db342f4
Showing
8 changed files
with
158 additions
and
381 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,9 @@ | ||
| cmake_minimum_required(VERSION 3.16) | ||
| project(x86intrin LANGUAGES CXX) | ||
| add_executable(x86intrin main.cpp) | ||
| if(CMAKE_CXX_COMPILER_ID MATCHES "Clang|GNU|IntelLLVM|QCC") | ||
| target_compile_options(x86intrin PUBLIC | ||
| "-march=cannonlake" "-mrdrnd" "-mrdseed" "-maes" "-msha" "-w") | ||
| elseif(CMAKE_CXX_COMPILER_ID STREQUAL "MSVC") | ||
| target_compile_options(x86intrin PUBLIC "-arch:AVX512" "-W0") | ||
| endif() |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,46 @@ | ||
| // Copyright (C) 2022 Intel Corporation. | ||
| // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only | ||
|
|
||
| #include <immintrin.h> | ||
|
|
||
| // Skylake AVX512 was added to GCC 4.9, Clang 3.7, and MSVC 2015. | ||
| // Cannon Lake was added to GCC 5, Clang 3.8, and MSVC 2017 15.7, | ||
| // so that's our minimum. | ||
| // Ice Lake was completed with GCC 8, Clang 6, and MSVC 2017 15.8. | ||
|
|
||
| int test(int argc, char **argv) | ||
| { | ||
| unsigned randomvalue; | ||
| _rdrand32_step(&randomvalue); // RDRND (IVB) | ||
| _rdseed32_step(&randomvalue); // RDSEED (BDW) | ||
| unsigned mask = _blsmsk_u32(argc); // BMI (HSW) | ||
| int clz = _lzcnt_u32(mask); // LZCNT (HSW) | ||
| int ctz = _tzcnt_u32(mask); // BMI (HSW) | ||
| mask = _bzhi_u32(-1, argc); // BMI2 (HSW) | ||
|
|
||
| __m128d d = _mm_setzero_pd(); // SSE2 | ||
| d = _mm_cvtsi32_sd(d, argc); // SSE2 | ||
| __m256d d2 = _mm256_broadcastsd_pd(d); // AVX (SNB) | ||
| d2 = _mm256_fmadd_pd(d2, d2, d2); // FMA (HSW) | ||
|
|
||
| __m128 f = _mm256_cvtpd_ps(d2); // AVX (SNB) | ||
| __m128i a = _mm_cvtps_ph(f, 0); // F16C (IVB) | ||
| __m128i b = _mm_aesenc_si128(a, a); // AESNI (WSM) | ||
| __m128i c = _mm_sha1rnds4_epu32(a, a, 0); // SHA (CNL) | ||
| __m128i e = _mm_sha1msg1_epu32(a, b); // SHA (CNL) | ||
| __m128i g = _mm_sha256msg2_epu32(b, c); // SHA (CNL) | ||
|
|
||
| __m512i zero = _mm512_setzero_si512(); // AVX512F (SKX) | ||
| __m512i data = _mm512_maskz_loadu_epi8(mask, argv[0]); // AVX512BW (SKX) | ||
| __m256i ptrs = _mm256_maskz_loadu_epi64(mask, argv); // AVX512VL (SKX) | ||
| __m512i data2 = _mm512_broadcast_i64x4(ptrs); // AVX512DQ (SKX) | ||
| __m256i data3 = _mm256_madd52lo_epu64(ptrs, ptrs, ptrs);// AVX512IFMA (CNL) | ||
| data2 = _mm512_multishift_epi64_epi8(data, data2); // AVX512VBMI (CNL) | ||
|
|
||
| return _mm256_extract_epi32(data3, 0); // AVX2 (HSW) | ||
| } | ||
|
|
||
| int main(int argc, char **argv) | ||
| { | ||
| return test(argc, argv); | ||
| } |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.