Li, Jiang e4d61d0d22 [CPU] Add AMX-only high-performance MLA backend for DeepSeek V2/V3/R1 (#52616) ...

Signed-off-by: jiang1.li <jiang1.li@intel.com>

2026-08-19 10:28:52 +08:00

..

micro_gemm

[Hardware][Power] Unqualized MoE Backend for Power (VSX) (#51624)

2026-08-13 10:56:09 +08:00

sgl-kernels

[CPU] Add AMX-only high-performance MLA backend for DeepSeek V2/V3/R1 (#52616)

2026-08-19 10:28:52 +08:00

activation_lut_bf16.cpp

…

activation.cpp

…

cpu_arch_macros.h

[CPU][s390x] Optimize inference perf and add oneDNN INT8 GEMM for s390x (#50219)

2026-07-31 17:49:31 +08:00

cpu_attn_amx.hpp

…

cpu_attn_fp8.hpp

[Refactor] Remove kernel dead code (#51051)

2026-08-07 01:20:19 -07:00

cpu_attn_impl.hpp

…

cpu_attn_neon_bfmmla.hpp

…

cpu_attn_neon.hpp

…

cpu_attn_rvv.hpp

…

cpu_attn_vec16.hpp

…

cpu_attn_vec.hpp

[Bugfix][CPU] Fix Clang OpenMP build on macOS (#49021)

2026-07-21 09:58:52 +00:00

cpu_attn_vsx.hpp

[HARDWARE][POWER] optimize math functions of VSX power (#47321)

2026-07-07 09:35:47 +00:00

cpu_attn_vxe.hpp

…

cpu_attn.cpp

[CPU] Fix FP8 attention scratchpad sizing (#50194)

2026-07-29 15:34:37 +08:00

cpu_fused_moe_activations.hpp

[CPU] Migrate unquantized MoE to the modular-kernel experts structure (#50133)

2026-08-03 14:55:04 +08:00

cpu_fused_moe_int8.cpp

[CPU][Perf] INT8 Fused MoE Kernel for Arm CPUs (#48637)

2026-07-27 05:53:09 +00:00

cpu_fused_moe.cpp

[Hardware][Power] Unqualized MoE Backend for Power (VSX) (#51624)

2026-08-13 10:56:09 +08:00

cpu_tanhf_neon.hpp

…

cpu_types_arm.hpp

[CPU] Add MLA backend so DeepSeek-V2/V3 can run on CPU (#49453)

2026-08-06 16:51:41 +08:00

cpu_types_riscv_defs.hpp

…

cpu_types_riscv_impl.hpp

[Bugfix][CPU][RISC-V] Fix build: make FP32Vec copy constructors non-explicit (#51099)

2026-08-14 04:59:01 +00:00

cpu_types_riscv.hpp

…

cpu_types_scalar.hpp

[CPU] Migrate unquantized MoE to the modular-kernel experts structure (#50133)

2026-08-03 14:55:04 +08:00

cpu_types_vsx.hpp

[CPU] Migrate unquantized MoE to the modular-kernel experts structure (#50133)

2026-08-03 14:55:04 +08:00

cpu_types_vxe.hpp

[CPU] Enable GPTQ and AWQ quantization for s390x (#51148)

2026-08-10 11:41:43 +08:00

cpu_types_x86.hpp

[CPU] Migrate unquantized MoE to the modular-kernel experts structure (#50133)

2026-08-03 14:55:04 +08:00

cpu_types.hpp

[HARDWARE][POWER] optimize math functions of VSX power (#47321)

2026-07-07 09:35:47 +00:00

cpu_wna16.cpp

…

dnnl_helper.cpp

…

dnnl_helper.h

…

dnnl_kernels.cpp

…

float_convert.hpp

…

generate_cpu_attn_dispatch.py

…

layernorm.cpp

…

mamba_cpu.cpp

[Hardware][CPU] Enable granite-4 model on cpu (#47641)

2026-07-20 06:15:16 +00:00

mamba_kernels.hpp

[Hardware][CPU] Enable granite-4 model on cpu (#47641)

2026-07-20 06:15:16 +00:00

mla_decode.cpp

[CPU] Add MLA backend so DeepSeek-V2/V3 can run on CPU (#49453)

2026-08-06 16:51:41 +08:00

pos_encoding.cpp

…

shm.cpp

…

spec_decode_utils.cpp

…

torch_bindings.cpp

[CPU] Add AMX-only high-performance MLA backend for DeepSeek V2/V3/R1 (#52616)

2026-08-19 10:28:52 +08:00

utils.cpp

…

utils.hpp

[Hardware][Power] Unqualized MoE Backend for Power (VSX) (#51624)

2026-08-13 10:56:09 +08:00