This is unlikely to happen in practice, but if it does, the rogue file would get included instead of the dynamically generated version for each target_core, leading to very confusing errors like "invalid operands (undefined UND and ABS sections)" in compilation of the assembly kernels as macros like PREFETCH would remain undefined
1. Added bfloat16 based dot as new API: shdot
2. Implemented generic kernel and cooperlake-specific (AVX512-BF16) kernel for shdot
3. Added 4 conversion APIs for bfloat16 data type <=> single/double: shstobf16 shdtobf16 sbf16tos dbf16tod
shstobf16 -- convert single float array to bfloat16 array
shdtobf16 -- convert double float array to bfloat16 array
sbf16tos -- convert bfloat16 array to single float array
dbf16tod -- convert bfloat16 array to double float array
4. Implemented generic kernels for all 4 conversion APIs, and cooperlake-specific kernel for shstobf16 and shdtobf16
5. Update level1 thread facilitate functions and macros to support multi-threading for these new APIs
6. Fix Cooperlake platform detection/specify issue when under dynamic-arch building
7. Change the typedef of bfloat16 from unsigned short to more strict uint16_t
Signed-off-by: Chen, Guobing <guobing.chen@intel.com>
Enable new build target platform -- COOPERLAKE. This target platform
supports all the SKYLAKEX supported ISAs + avx512bf16. So all the
SKYLAKEX specific kernels/drivers and related code are now extended
to be also active on COOPERLAKE. Besides, new BF16 related kernels
are active under this target.
* make building the bfloat16 BLAS functions conditional on BUILD_HALF
* pass the BUILD_HALF option to gensymbol
* Pass BUILD_HALF as a compiler define for dynamic_arch builds
This patch adds support for bfloat16 data type matrix multiplication kernel.
For architectures that don't support bfloat16, it is defined as unsigned short
(2 bytes). Default unroll sizes can be changed as per architecture as done for
SGEMM and for now 8 and 4 are used for M and N. Size of ncopy/tcopy can be
changed as per architecture requirement and for now, size 2 is used.
Added shgemm in kernel/power/KERNEL.POWER9 and tested in powerpc64le and
powerpc64. For reference, added a small test compare_sgemm_shgemm.c to compare
sgemm and shgemm output.
This patch does not cover OpenBLAS test, benchmark and lapack tests for shgemm.
Complex type implementation can be discussed and added once this is approved.
The need to call it .tmp existed only when it was generated by a tmpfile call, and the "-x c" option to tell the compiler it is actually a C source is not universally supported (this broke the test with clang-cl at least)
* Update make.inc entries for LAPACK 3.9.0
Reference-LAPACK PR 347 changed some variable names and relative paths
* Update LAPACK to 3.9.0
* Add new functions from LAPACK 3.9.0
* Add new functions from LAPACK 3.9.0
* Restore LOADER command
as it makes it easier to specify pthread as needed
* Restore LOADER
* Restore EIG/LIN prefixes in cmdbase
* add binary path to lapack_testing.py call
* Restore OpenMP version check
* Restore OpenMP version check
* Restore fix for out-of-bounds array accesses
from #2096
* Update DYNAMIC_ARCH list of ARM64 targets for gmake
* Update arm64 cpu list for runtime detection
* Update DYNAMIC_ARCH list of ARM64 targets for cmake and add POWERPC targets
* Fix compiler identification and option setting
* Handle BINARY=32 option on X86_64
* Add xGEMM3M unroll parameters for crossbuild-target CORE2
* Replace bogus mingw64/32bit CI job with actual 32bit build
mingw64 is not multilib-capable, so using an x86_64-mingw with BINARY=32 in the CI was not going to work anyway (but build passed while BINARY=32 was ignored).
1. Supply appropriate values for C/Z GEMM unroll when cross-compiling for CORE2 or ARMV7
2. Add the required xLOCAL_BUFFER_SIZE parameters for cross-compiling CORE2
3. Add -DFORCE_<target> option to getarch when building with -DTARGET=target
for #2245
* clean up a case where CMake endif()
contained the conditional used in the
if(), which is no longer needed /
discouraged since our minimum required
CMake version supports the modern syntax
from #2143, -march=native precludes use of more specific options like -march=skylake-avx512 in individual kernels, and defeats the purpose of dynamic arch anyway.
21eda8b5 introduced a check in getarch.c to test if the compiler is capable of
AVX512. This check currently fails, since the used __AVX2__ macro is only
defined if getarch itself was compiled with AVX2/AVX512 support. Make sure this
is the case by building getarch with -march=native on x86_64. It is only
supposed to run on the build host anyway.
Fixed core naming for DYNAMIC_ARCH. Corrected GEMM_DEFAULT entries and added SYMV_P. Replaced outdated VULCAN define for ThunderX2T99 with ARMV8 to get basic definitions back. For issue #1908
This patch adds the basic infrastructure for adding the SkylakeX (Intel Skylake server)
target. The SkylakeX target will use the AVX512 (AVX512VL level) instruction set,
which brings 2 basic things:
1) 512 bit wide SIMD (2x width of AVX2)
2) 32 SIMD registers (2x the number on AVX2)
This initial patch only contains a trivial transofrmation of the Haswell SGEMM kernel
to AVX512VL; more will follow later but this patch aims to get the infrastructure
in place for this "later".
Full performance tuning has not been done yet; with more registers and wider SIMD
it's in theory possible to retune the kernels but even without that there's an
interesting enough performance increase (30-40% range) with just this change.
Without specifying thread count, NUM_THREADS would not be defined and CMake would fail.
This is because core count cannot be determined when cross-compiling.
* prebuild.cmake: Put quotes around path names that may contain whitespace
(Copied from alexkaratakis' PR #1295)
* kernel/CMakeLists.txt: Fix common_lapack header inclusion and DYNAMIC_ARCH generation of ?neg_tcopy and ?laswp_ncopy files
* lapack/CMakeLists.txt: Use correct template for ?laswp_(plus,minus) functions
Martin Kroeker
Qiyu8
Chen, Guobing
Ashwin Sekhar T K
Rajalakshmi Srinivasaraghavan
Rajalakshmi Srinivasaraghavan
مهدي شينون (Mehdi Chinoune)
Ali Saidi
k.dunikowski
Tyler Reddy
Michael Lass
luz.paz
Sacha
Andrew
Yuri
Arjan van de Ven
Zhiyong Dang
xoviat
Isuru Fernando
martin
Ian Henriksen