o
    it                     @   sD   d dl mZ d dlmZmZmZ d dlmZ d
de	defddZ
d	S )   )env)JitSpecgen_jit_specsm90a_nvcc_flags)is_cuda_version_at_leastFuse_fast_buildreturnc                 C   s   t dddgtdrdnd }tdtjd tjd	 tjd
 tjd tjd tjd tjd tjd g|| r8dgng ddgtjd tjd d tjd d d d tjd d d d d tjd d d d gdS )z<Generate JIT spec for FP8 block scale GEMM on SM90 (Hopper).z-DCOMPILE_HOPPER_TMA_GEMMSz-DENABLE_BF16z-DENABLE_FP8z12.8)z-DENABLE_FP8_BLOCK_SCALE fp8_blockscale_gemm_90z[nv_internal/tensorrt_llm/kernels/cutlass_kernels/fp8_blockscale_gemm/fp8_blockscale_gemm.cuz#fp8_blockscale_gemm_sm90_binding.cuz/fused_moe/cutlass_backend/deepgemm_jit_setup.cuz#nv_internal/cpp/common/envUtils.cppz!nv_internal/cpp/common/logger.cppz&nv_internal/cpp/common/stringUtils.cppz(nv_internal/cpp/common/tllmException.cppz%nv_internal/cpp/common/memoryUtils.cuz-DFAST_BUILDz-lnvrtcz-lcudanv_internalincludetensorrt_llmcutlass_extensionskernelscutlass_kernels)extra_cuda_cflagsextra_cflagsextra_ldflagsextra_include_paths)r   r   r   jit_envFLASHINFER_CSRC_DIR)r   
nvcc_flagsr	   r	   X/home/ubuntu/vllm_env/lib/python3.10/site-packages/flashinfer/jit/gemm/fp8_blockscale.py#gen_fp8_blockscale_gemm_sm90_module
   st   r   N)F) r   r   corer   r   r   cpp_extr   boolr   r	   r	   r	   r   <module>   s    