Current RVV x280 target depends on vlen=512-bits for Level 3 operations.
Commit adds generic target that supports vlen=128-bits.
New target uses the same scalable kernels as x280 for Level 1&2 operations, and autogenerated kernels for Level 3 operations.
Functional correctness of Level 3 operations tested on vlen=128-bits using QEMU v8.1.1 for ctests and BLAS-Tester.
POWER10 optimizations are disabled when using default AIX assembler.
As we have fixed many issues recently, enabling optimization path
for default assembler.
This patch removes the prefetches from cgemm/zgemm which improves the performance similar to sgemm/dgemm did in #3868, this means I'm happy to enable this on any applicable cores.
I also replicated the unrolling the copies from sgemm and dgemm.
In order to enable support for future cores which have similar tunings
(in this case I'm doing this for the Arm(R) Neoverse(TM) V2 core), this generically detects SVE support and enables it. This should better manage the size and complexity of dynamic dispatch rather than just copy pasting the same parameters.
To make `ARMV8SVE` more representive of the common 128-bit SVE case,
I've split it and similar parameters from A64FX which has the wider
512-bit SVE.
This re-spins #3869 with some additional copy unrolling which helps maintain SYRK performance.
After #3868, the SVE kernels represent a pretty good boost.
This re-uses ARMV8SVE as a base and I'm going to incrementally move everything to use ARMV8SVE in additional patches (as well as fix up anything that's not already in ARMV8SVE).
This seems like a good balance of values for reasonably sized matrices. With `SWITCH_RATIO=16` the DGEMM scales better to bigger sizes but the better solution would be some kind of
thread throttling so I've gone with `SWITCH_RATIO=8`.
Previously dynamic builds were either using the default SWITCH_RATIO
or one from the higher level architecture; this patch ensures the
dynamic builds can use this parameter as well.
* fix multiple numerical stability and corner case issues
* add a script to generate arbitrary gemm kernel shapes
* add a generic zvl256b target to demonstrate large gemm kernel unrolls
