diff --git a/experimental/grouped_convolution_tile_instances/configs/profiler/ndhwgc_bf16.conf b/experimental/grouped_convolution_tile_instances/configs/profiler/ndhwgc_bf16.conf index db7d362d38..d04973f05d 100644 --- a/experimental/grouped_convolution_tile_instances/configs/profiler/ndhwgc_bf16.conf +++ b/experimental/grouped_convolution_tile_instances/configs/profiler/ndhwgc_bf16.conf @@ -236,28 +236,28 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 128, 64, Filter1x1Strid DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 16, 256, 64, Filter1x1Stride1Pad0, 16, 16, 1, 4, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 32, 256, 64, Filter1x1Stride1Pad0, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 64, 64, Default, 16, 16, 1, 2, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 256, 32, 64, Default, 32, 32, 2, 1, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 128, 64, Default, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 64, 64, Filter1x1Stride1Pad0, 16, 16, 1, 2, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 256, 32, 64, Filter1x1Stride1Pad0, 32, 32, 2, 1, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 128, 64, Filter1x1Stride1Pad0, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> \ No newline at end of file +DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> diff --git a/experimental/grouped_convolution_tile_instances/configs/profiler/ndhwgc_fp16.conf b/experimental/grouped_convolution_tile_instances/configs/profiler/ndhwgc_fp16.conf index dc0908c3d5..f273002e04 100644 --- a/experimental/grouped_convolution_tile_instances/configs/profiler/ndhwgc_fp16.conf +++ b/experimental/grouped_convolution_tile_instances/configs/profiler/ndhwgc_fp16.conf @@ -227,28 +227,28 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 128, 64, Filter1x1Strid DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 16, 256, 64, Filter1x1Stride1Pad0, 16, 16, 1, 4, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 32, 256, 64, Filter1x1Stride1Pad0, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 64, 64, Default, 16, 16, 1, 2, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 256, 32, 64, Default, 32, 32, 2, 1, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 128, 64, Default, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 64, 64, Filter1x1Stride1Pad0, 16, 16, 1, 2, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 256, 32, 64, Filter1x1Stride1Pad0, 32, 32, 2, 1, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 128, 64, Filter1x1Stride1Pad0, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> \ No newline at end of file diff --git a/experimental/grouped_convolution_tile_instances/configs/profiler/nhwgc_bf16.conf b/experimental/grouped_convolution_tile_instances/configs/profiler/nhwgc_bf16.conf index 8cbb0b78df..d5a4a28beb 100644 --- a/experimental/grouped_convolution_tile_instances/configs/profiler/nhwgc_bf16.conf +++ b/experimental/grouped_convolution_tile_instances/configs/profiler/nhwgc_bf16.conf @@ -236,28 +236,28 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 128, 64, Filter1x1Strid DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 16, 256, 64, Filter1x1Stride1Pad0, 16, 16, 1, 4, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 32, 256, 64, Filter1x1Stride1Pad0, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 64, 64, Default, 16, 16, 1, 2, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 256, 32, 64, Default, 32, 32, 2, 1, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 128, 64, Default, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 64, 64, Filter1x1Stride1Pad0, 16, 16, 1, 2, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 256, 32, 64, Filter1x1Stride1Pad0, 32, 32, 2, 1, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 128, 64, Filter1x1Stride1Pad0, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> \ No newline at end of file diff --git a/experimental/grouped_convolution_tile_instances/configs/profiler/nhwgc_fp16.conf b/experimental/grouped_convolution_tile_instances/configs/profiler/nhwgc_fp16.conf index c8bf3698cf..e20fe50689 100644 --- a/experimental/grouped_convolution_tile_instances/configs/profiler/nhwgc_fp16.conf +++ b/experimental/grouped_convolution_tile_instances/configs/profiler/nhwgc_fp16.conf @@ -227,28 +227,28 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 128, 64, Filter1x1Strid DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 16, 256, 64, Filter1x1Stride1Pad0, 16, 16, 1, 4, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 32, 256, 64, Filter1x1Stride1Pad0, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 64, 64, Default, 16, 16, 1, 2, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 256, 32, 64, Default, 32, 32, 2, 1, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 128, 64, Default, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Default, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 64, 64, Filter1x1Stride1Pad0, 16, 16, 1, 2, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 256, 32, 64, Filter1x1Stride1Pad0, 32, 32, 2, 1, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 128, 64, Filter1x1Stride1Pad0, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 16, 32, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<128, 32, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> \ No newline at end of file diff --git a/experimental/grouped_convolution_tile_instances/configs/tests/ndhwgc_bf16.conf b/experimental/grouped_convolution_tile_instances/configs/tests/ndhwgc_bf16.conf index 5b39beabc9..fb44310793 100644 --- a/experimental/grouped_convolution_tile_instances/configs/tests/ndhwgc_bf16.conf +++ b/experimental/grouped_convolution_tile_instances/configs/tests/ndhwgc_bf16.conf @@ -39,11 +39,11 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 16, 64, Filter1x1Pad0, DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 256, 32, 64, Filter1x1Pad0, 32, 32, 2, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 256, 32, 64, Filter1x1Stride1Pad0, 32, 32, 2, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> \ No newline at end of file diff --git a/experimental/grouped_convolution_tile_instances/configs/tests/ndhwgc_fp16.conf b/experimental/grouped_convolution_tile_instances/configs/tests/ndhwgc_fp16.conf index 5b39beabc9..fb44310793 100644 --- a/experimental/grouped_convolution_tile_instances/configs/tests/ndhwgc_fp16.conf +++ b/experimental/grouped_convolution_tile_instances/configs/tests/ndhwgc_fp16.conf @@ -39,11 +39,11 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 16, 64, Filter1x1Pad0, DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 256, 32, 64, Filter1x1Pad0, 32, 32, 2, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 256, 32, 64, Filter1x1Stride1Pad0, 32, 32, 2, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> \ No newline at end of file diff --git a/experimental/grouped_convolution_tile_instances/configs/tests/nhwgc_bf16.conf b/experimental/grouped_convolution_tile_instances/configs/tests/nhwgc_bf16.conf index 5b39beabc9..fb44310793 100644 --- a/experimental/grouped_convolution_tile_instances/configs/tests/nhwgc_bf16.conf +++ b/experimental/grouped_convolution_tile_instances/configs/tests/nhwgc_bf16.conf @@ -39,11 +39,11 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 16, 64, Filter1x1Pad0, DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 256, 32, 64, Filter1x1Pad0, 32, 32, 2, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 256, 32, 64, Filter1x1Stride1Pad0, 32, 32, 2, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Default, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Default, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> -DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> +# DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 128, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v4> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<64, 16, 16, 64, Filter1x1Stride1Pad0, 16, 16, 1, 1, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3_DirectLoad<256, 64, 64, 64, Filter1x1Stride1Pad0, 16, 16, 2, 2, 2, 2, 4, 1, 1, BlkGemmPipelineScheduler: Intrawave, BlkGemmPipelineVersion: v1> \ No newline at end of file