ext-torch/configs/E=8,N=3584,device_name=NVIDIA_H100_80GB_HBM3.json

{
    "1": {
        "BLOCK_SIZE_M": 16,
        "BLOCK_SIZE_N": 64,
        "BLOCK_SIZE_K": 256,
        "GROUP_SIZE_M": 64,
        "num_warps": 4,
        "num_stages": 4
    },
    "2": {
        "BLOCK_SIZE_M": 16,
        "BLOCK_SIZE_N": 256,
        "BLOCK_SIZE_K": 64,
        "GROUP_SIZE_M": 32,
        "num_warps": 4,
        "num_stages": 4
    },
    "4": {
        "BLOCK_SIZE_M": 16,
        "BLOCK_SIZE_N": 64,
        "BLOCK_SIZE_K": 256,
        "GROUP_SIZE_M": 16,
        "num_warps": 4,
        "num_stages": 4
    },
    "8": {
        "BLOCK_SIZE_M": 16,
        "BLOCK_SIZE_N": 64,
        "BLOCK_SIZE_K": 128,
        "GROUP_SIZE_M": 1,
        "num_warps": 4,
        "num_stages": 4
    },
    "16": {
        "BLOCK_SIZE_M": 16,
        "BLOCK_SIZE_N": 64,
        "BLOCK_SIZE_K": 128,
        "GROUP_SIZE_M": 1,
        "num_warps": 4,
        "num_stages": 4
    },
    "24": {
        "BLOCK_SIZE_M": 16,
        "BLOCK_SIZE_N": 64,
        "BLOCK_SIZE_K": 128,
        "GROUP_SIZE_M": 1,
        "num_warps": 4,
        "num_stages": 4
    },
    "32": {
        "BLOCK_SIZE_M": 16,
        "BLOCK_SIZE_N": 64,
        "BLOCK_SIZE_K": 128,
        "GROUP_SIZE_M": 1,
        "num_warps": 4,
        "num_stages": 4
    },
    "48": {
        "BLOCK_SIZE_M": 32,
        "BLOCK_SIZE_N": 128,
        "BLOCK_SIZE_K": 128,
        "GROUP_SIZE_M": 1,
        "num_warps": 4,
        "num_stages": 4
    },
    "64": {
        "BLOCK_SIZE_M": 32,
        "BLOCK_SIZE_N": 128,
        "BLOCK_SIZE_K": 128,
        "GROUP_SIZE_M": 1,
        "num_warps": 4,
        "num_stages": 4
    },
    "96": {
        "BLOCK_SIZE_M": 32,
        "BLOCK_SIZE_N": 128,
        "BLOCK_SIZE_K": 128,
        "GROUP_SIZE_M": 32,
        "num_warps": 4,
        "num_stages": 4
    },
    "128": {
        "BLOCK_SIZE_M": 64,
        "BLOCK_SIZE_N": 128,
        "BLOCK_SIZE_K": 128,
        "GROUP_SIZE_M": 1,
        "num_warps": 4,
        "num_stages": 4
    },
    "256": {
        "BLOCK_SIZE_M": 128,
        "BLOCK_SIZE_N": 256,
        "BLOCK_SIZE_K": 64,
        "GROUP_SIZE_M": 1,
        "num_warps": 8,
        "num_stages": 4
    },
    "512": {
        "BLOCK_SIZE_M": 64,
        "BLOCK_SIZE_N": 128,
        "BLOCK_SIZE_K": 64,
        "GROUP_SIZE_M": 32,
        "num_warps": 4,
        "num_stages": 4
    },
    "1024": {
        "BLOCK_SIZE_M": 64,
        "BLOCK_SIZE_N": 256,
        "BLOCK_SIZE_K": 64,
        "GROUP_SIZE_M": 64,
        "num_warps": 4,
        "num_stages": 4
    },
    "1536": {
        "BLOCK_SIZE_M": 128,
        "BLOCK_SIZE_N": 256,
        "BLOCK_SIZE_K": 64,
        "GROUP_SIZE_M": 64,
        "num_warps": 8,
        "num_stages": 4
    },
    "2048": {
        "BLOCK_SIZE_M": 128,
        "BLOCK_SIZE_N": 256,
        "BLOCK_SIZE_K": 64,
        "GROUP_SIZE_M": 64,
        "num_warps": 8,
        "num_stages": 4
    },
    "3072": {
        "BLOCK_SIZE_M": 128,
        "BLOCK_SIZE_N": 256,
        "BLOCK_SIZE_K": 64,
        "GROUP_SIZE_M": 64,
        "num_warps": 8,
        "num_stages": 4
    },
    "4096": {
        "BLOCK_SIZE_M": 128,
        "BLOCK_SIZE_N": 256,
        "BLOCK_SIZE_K": 64,
        "GROUP_SIZE_M": 1,
        "num_warps": 8,
        "num_stages": 4
    }
}