Jinyang He

ggml: optimize some vec dot functions for LoongArch ASX

2

* ggml : optimize ggml_vec_dot_iq4_xs_q8_K for LoongArch ASX * ggml : optimize mul_sum_i8_pairs_float for LoongArch ASX * ggml : optimize ggml_vec_dot_q2_K_q8_K for LoongArch ASX * ggml : optimize ggml_vec_dot_q5_K_q8_K for...

Prepare for LoongArch SX and ASX

1

1, Add an option named XNNPACK_ENABLE_CPUINFO so that avoid compiling cpuinfo for LoongArch (and Hexagon). 2, Add "&& XNN_ARCH_RISCV" in transpose-config for RISCV. 3, Others are prepared for LoongArch SX...

Add LoongArch LASX support

1

### PR devices LoongArch ### PR types New features ### PR changes API ### Description Add LoongArch LASX support Try `./lite/tools/build_linux.sh --arch=loongarch --with_extra=on` to build on LoongArch. Co-authored-by: @junchao-loongson Cc:...

The implementation is referred to x86_avx2 and neon. The loongarch simd intrinsics can be found at [1]. [1] https://jia.je/unofficial-loongarch-intrinsics-guide