Merge branch 'openblas-open-910' of git://github.com/damonyu1989/OpenBLAS into damonyu1989-openblas-open-910

Changing the unroll factors for dgemm to 8 shows improved performance with
POWER10 MMA feature.   Also made some minor changes in sgemm for edge cases.

Change-Id: Iae7800a32f5af3903c330882cdf6f292d885f266

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.

We recently changed the register blocking for SGEMM on s390x to 16x4.
However, we did not adjust Q to a multiple of 16 and thus fell back to
the 8x4 kernel at each block's margin, without need. Adjust P and Q to
multiples of 16 to employ the faster 16x4 kernel for complete full-sized
blocks.

Signed-off-by: Marius Hillenbrand <mhillen@linux.ibm.com>

This patch introduces new optimized version of SHGEMM kernel
using power10 Matrix-Multiply Assist (MMA) feature introduced in
POWER ISA v3.1. This patch makes use of new POWER10 compute instructions
for matrix multiplication operation.

Tested on simulator and there are no new test failures.

This is the initial patch to support build infrastructure
for POWER10 architecture.

Change register blocking for SGEMM (and STRMM) on z14 from 8x4 to 16x4
by adjusting SGEMM_DEFAULT_UNROLL_M and choosing the appropriate copy
implementations. Actually make KERNEL.Z14 more flexible, so that the
change in param.h suffices. As a result, performance for SGEMM improves
by around 30% on z15.

On z14, FP SIMD instructions can operate on float-sized scalars in
vector registers, while z13 could do that for double-sized scalars only.
Thus, we can double the amount of elements of C that are held in
registers in an SGEMM kernel.

Signed-off-by: Marius Hillenbrand <mhillen@linux.ibm.com>

fixes lapack-test zger failures seen in #2299 after application of my PR #2551

and allow enforcing TARGET=1004K as well (omission from earlier 1004K merge and later introduction of TARGET check)

As shown in #2538, default buffersizes on some platforms were smaller than required in memory.c
and the requirement could never be fulfilled for a calculated GEMM_R on PPC given the fomula used

Fix avx512 sgemm performance bug when ldc is a multiple of 1024

Adjust SkylakeX GEMM3M parameters, add an AVX512 STRMM kernel and fix performance bugs in AVX2 s/c/z GEMM

The implementation is a hybird of the ARMV8 one with some of the
improved TX2 rountines along with specifying -march=v8.2-a

e.g. make CC=riscv64-unknown-linux-gnu-gcc FC=riscv64-unknown-linux-gnu-gfortran TARGET=RISCV64_GENERIC HOSTCC=gcc

Signed-off-by: Wang,Long <long1.wang@intel.com>

as the respective PPC970 "altivec" kernels give wrong results when compiled for big endian

Optimize SGEMM on SKYLAKEX CPUs

There is currently no simple way to query cache sizes on ARMV8, so this takes the number of cores as a trivial indication if the target is a server-class device with a big cache, or just a single-board toy or smartphone.

power9 cgemm/ctrmm. new sgemm 8x16

FreeBSD needs those macros too.