Improving Load Imbalance in Thread-Parallel GEMM

Optimize gemv_n_sve kernel

Fix aarch64 sbgemv_t compilation error for GCC < 13

* Tighten memory orders for C11 atomic operations

Add dispatch of SBGEMVNKERNEL for NEOVERSEN2 and NEOVERSEV2

Redefined threading logic for GESV and GEMV on WoA

Expose the option to build without LAPACKE to ccmake

fix minior issues of redeclaration of float x0,x1 in sbgemv_n_neon.c

add beta and alpha testcase for sbgemv

Add SBGEMVN Kernel for ARM64

Add sbgemv_t_bfdot kernel for ARM64

This improves performance for sbgemv_t by up to 100x on NEOVERSEV1.
The geometric mean speedup is ~61x for M=N=[2,512].

Make the Neoverse GEMM/GEMV throttling code conditional on SMP

dispatch NEOVERSEV2 to NEOVERSEN2 under dynamic setting

Add throttling profile for SGEMM and SGEMV on `NEOVERSEV2`

Fix "dummy2" flag reading in PPC970 S/DSCAL

Fix GEMMT transforming the input array B in some complex cases

Add ARM64 options for NVIDIA HPC

Support for SGEMM_DIRECT Kernel based on SME1

Ensure that gmake builds with flang-new link the flang runtime into the shared library

Fix the CMake build to define USE_TRMM for RISCV64 targets as well

Revert "Fix potential inaccuracy in multithreaded level3 related to SWITCH_RATIO"

CI: remove the express NeoverseN2 target from the Cobalt100 job in the gh workflow

* Added ARMV9SME target
* Added SGEMM_DIRECT kernel based on SME1

Refined GESV Parallelization Logic for Windows on ARM64

Add -O2 to flang flags when building on WoA in Release mode