rebase

Optimize the performance of dot by using universal intrinsics in X86/ARM

Default BLAS3_MEM_ALLOC_THRESHOLD on all platforms to 32

rebase

Avoid out of bounds access on invalid memory free

Lazyly reinit threads after a fork in OMP mode

fix clang std=c18 compilation on aarch64

This initializes the per-thread memory buffers which get
cleared/released on a fork via pthread_at_fork. Not doing so leads to
each thread calling blas_memory_alloc on almost every execution which
slows down the code significantly as the threads race for the memory
allocation using locks to serialize that.

Check for __linux rather than linux in cpuid code and benchmarks

Consolidate usage of backticks for build options

and add a fallback for unsupported operating systems in detect()

rebase

Optimize dcopy/zcopy for POWER10

This patch makes use of new POWER10 vector pair instructions for
loads and stores. Tested in simulator and no new failures.

There were some build options in the README that were not
highlighted. Now all are highlighted.

Optimize the performance of daxpy by using universal intrinsics

FIx underflow/rounding errors in LAPACK (S,D)LANV2

Reference-LAPACK PR 445, fixing their issue 263

Readmefixes

rebase

fix a bug of trmm

Add an AWS-Graviton2 build to Travis CI

rebase

Add an OSX build with xcode12

s390x: fix cscal and zscal implementations

The implementation of complex scalar * vector multiplication for Z14
makes some LAPACK tests fail because the numerical differences to the
reference implementation exceed the threshold (as can be seen by running
make lapack-test and replacing kernel/zarch/cscal.c with a generic
implementation for comparison).

The complex multiplication uses terms of the form a * b + c * d for both
real and imaginary parts. The assembly code (and compiler-emitted code
as well) uses fused multiply add operations for the second product and
sum. The results can be "surprising", for example when both terms in the
imaginary part nearly cancel each other out. In that case, the second
product contributes more digits to the sum than the first product that
has been rounded before.

One option is to use separate multiplications (which then round the same
way) and a distinct add. Change the code to pursue that path, by (1)
requesting the compiler not to contract the operations into FMAs and (2)
replacing the assembly kernel with corresponding vectorized C code
(where change 1 also applies).

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

Make clang slightly more cautious when contracting floating-point
operations (e.g., when applying fused multiply add) by setting
-ffp-contract=on (instead of fast).

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

... to facilitate reuse beyond gemm_vec.c and avoid code duplication.

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

Fix workspace query in LAPACK xGELQ (Reference-LAPACK 443)