This adds all the relevant bits and pieces to add a `shgemv` path as
well as a future `hgemm`/`hgemv` path in a similar model to `sb` and `b`
interfaces.
I've also fixed a few bits and pieces around `shgemm` which didn't build
in a few situations.
There were a couple of issues with the detection code used to check
for RVV 1.0 on kernels that do not support hwprobe.
1. The vtype clobber was missing
2. The wrong form of vsetvli was being used. The vsetvli x0, x0 form
is inappropriate for this use case as it can only be safely used
in code where the value of vtype is known. The use of vsetvli
x0, x0 here can lead to a failure to detect RVV 1.0, if,
for example, the vill bit happens to be set before
detect_riscv64_rvv100 is called.
We fix both issues by adding the missing clobber and replacing the
first parameter to vsetvli with t0 (which we add to our clobbers).
The compiler options that enable 16 bit floating point instructions
should not be enabled by default when building the RISCV64_ZVL128B
and RISCV64_ZVL256B targets. The zfh and zvfh extensions are not part
of the 'V' extension and are not required by any of the RVA profiles.
There's no guarantee that kernels built with zfh and zvfh will work
correctly on fully compliant RVA23U64 devices.
To fix the issue we only build the RISCV64_ZVL128B and RISCV64_ZVL256B
kernels with the half float flags if BUILD_HFLOAT16=1. We also update
the RISC-V dynamic detection code to disable the RISCV64_ZVL128B and
RISCV64_ZVL256B kernels at runtime if we've built with DYNAMIC_ARCH=1
and BUILD_HFLOAT16=1 and are running on a device that does not support
both Zfh and Zvfh.
Fixes: https://github.com/OpenMathLib/OpenBLAS/issues/5428
- Sets up all the various entrypoints for `bgemv`
- Adds `bscal` for use in the `bgemv` interface
- Adds test cases for comparing `sgemv` and `bgemv`
- Adds generic kernels for `bgemv_n` and `bgemv_t` which are accurate
enough to pass above tests
When there are multiple NUMA nodes and hyper-threading causes adjacent logical cores to share a physical core (e.g., common -> avail[i] = 0x5555555555555555UL), the numa_mapping function should not use a bitmask for filtering, as this would lead to redundant masking with the subsequent local_cpu_map function.
Added HFLOAT16 support for RISCV64
Added shgemm_kernel_8x8 for RISCV64_ZVL128B and shgemm_kernel_16x8 for RISCV64_ZVL256B based on HFLOAT16
The instruction sets used are ZVFH and ZFH, which need to be supported by RVV1.0
Related to issue #5279
Co-authored-by Linjin Li <linjin_li@163.com>
* Fix ARMV9SME target and add support_sme1 code for MacOS
* make sgemm_direct unconditionally available on all arm64
* build a (dummy) sgemm_direct kernel on all arm64
* Update dynamic_arm64.c
Using `OpenMP::OpenMP_LANG` targets for CMake is less error-prone than
passing the compiler and linker flags manually. Furthermore, it allows
the user to customize those flags by setting `OpenMP_LANG_FLAGS`,
`OpenMP_LANG_LIB_NAMES`, and `OpenMP_omp_LIBRARY`.
