IntrinsicGemm

xDNN v1.5.6

• fix issue of BF16 pack issue when transpose=true

xDNN v1.5.5

• bf16fp8bf16 gemv

xDNN v1.5.4

• Optimize BF16 pack performance
• Add bf16bf16bf16 gemm

xDNN v1.5.3

• Add beta parameter to AMX gemm

xDNN v1.5.2

• Add AMX_FP16 support for small gemm.
• Built with GCC 13.2.1

xDNN v1.5.1

• Add xdnn_hgemm_f32f16f32_packb_block.

xDNN v1.5.0

• Add hgemm w/ fp32 bias.

xDNN v1.4.6

• Add alpha and beta param to small_sgemm_f32f16bf16.

xDNN v1.4.5

• Add post op gelu activation.

xDNN v1.4.4

• Fix AMX illegal instruction issue.

xDNN v1.4.3

• Add small_sgemm_f32bf16bf16.
• Add small_sgemm_f32f16bf16.

xDNN v1.4.2

• Support amx_gemm_bf16bf16bf16 kernel w/ any shapes.

xDNN v1.4.1

• Add sgemm_bf16bf16f32 and sgemm_f32bf16bf16 kernels.
• Add softmax kernels.

xDNN v1.4.0

• Add hgemm_f32u4f32 kernels.
• Add sgemm_f32nf4f32 kernels.

xDNN v1.3.1

• Fix sgemm_f32u4f32 kernels parallel bug.

xDNN v1.3.0

• Add sgemm_f32u4f32 kernels

xDNN v1.2.1

• Add xdnn_small_amx_sgemm_bf16bf16bf16_packb implemention with transposed weight.

xDNN v1.2

• Add xdnn_small_amx_sgemm_bf16bf16bf16_packb implemention.
• Add xdnn_small_amx_sgemm_bf16bf16bf16_compute implemention.

xDNN v1.1

• Add bgemm_f32bf16f32_packb weight format BA16a64b2a.
• Add intrinsic extension api.

xDNN v1.0

• Add sgemm kernels
• Add sgemm_f32f16f32 kernels
• Add sgemm_f32i8f32 kernels
• Add hgemm_f32f16f32 kernels
• Add hgemm_f16f16f32 kernels
• Add hgemm_f32i8f32 kernels
• Add bgemm_f32bf16f32 kernels