[appveyor] use flang from conda-forge

Break out of potentially infinite rescaling loop after 1000 iterations

Inf values in the input vector will survive rescaling, causing an infinite loop. The value of 1000 is arbitrarily chosen as a large but finite value with the intention to never interfere with regular calculations.

Try to handle shmget or shmat failing

also replaces one verbatim sched_yield with the YIELDING macro for consistency as suggested in #1351

This flang will be updated in the future. We leave cmake because it's
not yet released with fortran support

Output an error message when shmat() fails

Observed in #1351 with SELinux as the likely culprit. Without the message, the user saw a segfault with no apparent reason

WIP: Support for Flang on Windows

[appveyor] fixes

MSVC or clang-cl.

(Trivial) optimized dsdot implementation for HASWELL

ARM64: Convert all labels to local labels

While debugging/profiling applications using perf or other tools, the
kernels appear scattered in the profile reports. This is because the labels
within the kernels are not local and each label is shown as a separate
function.

To avoid this, all the labels within the kernels are changed to local
labels.

Fix 32bit HASWELL builds

Make ReLAPACK available in cmake builds

Update README.md to include POWER8

Add POWER 8 to the list of additional architectures.

* 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

2D thread distribution for multi-threaded GEMMs

Fix out-of-bounds memory accesses exposed by xccblat3 testcase

Save and restore VSX registers

...and does not appear to be needed even when it stays within the bounds of the array

CMake improvements

Variable thread count for multi-threaded GEMMs

Allows maximum use of available cores if one of M and N is small and the other is large.

Add trivial smoketest for xpotrf

m, n, and k can be set to arbitrary constants. A and B matrices can be transposed independently.