[rocm-libraries] ROCm/rocm-libraries#4406 (commit 61f9f90)

[CK] CK Tile grouped convolution direct load

## Motivation

CK Tile grouped convolution forward direct load support.

## Technical Details

Basic pipeline for direct load and new instances for forward for v1 and
v4 pipelines.

## Test Plan

test_grouped_convnd_fwd_tile

## Test Result

CI pending

## Submission Checklist

- [x] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
AICK-130
This commit is contained in:
Bartłomiej Kocot
2026-02-09 21:09:42 +00:00
committed by assistant-librarian[bot]
parent 0cafa68b6f
commit 27e0a34e0f
29 changed files with 739 additions and 56 deletions

View File

@@ -234,4 +234,30 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 64, 64, Filter1x1Stride
DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 16, 128, 64, Filter1x1Stride1Pad0, 16, 16, 1, 4, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2>
DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 128, 64, Filter1x1Stride1Pad0, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2>
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<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, 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<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<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, 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<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<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>

View File

@@ -225,4 +225,30 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 64, 64, Filter1x1Stride
DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 16, 128, 64, Filter1x1Stride1Pad0, 16, 16, 1, 4, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2>
DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 128, 64, Filter1x1Stride1Pad0, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2>
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<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, 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<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<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, 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<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<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>

View File

@@ -234,4 +234,30 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 64, 64, Filter1x1Stride
DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 16, 128, 64, Filter1x1Stride1Pad0, 16, 16, 1, 4, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2>
DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 128, 64, Filter1x1Stride1Pad0, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2>
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<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, 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<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<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, 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<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<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>

View File

@@ -225,4 +225,30 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 64, 64, Filter1x1Stride
DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 16, 128, 64, Filter1x1Stride1Pad0, 16, 16, 1, 4, 8, 8, 4, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2>
DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 128, 64, Filter1x1Stride1Pad0, 32, 32, 1, 2, 8, 8, 8, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v2>
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<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, 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<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<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, 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<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<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>

View File

@@ -38,4 +38,12 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 256, 32, 64, Default, 32, 3
DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 16, 64, Filter1x1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v1>
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<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<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<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>

View File

@@ -38,4 +38,12 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 256, 32, 64, Default, 32, 3
DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 16, 64, Filter1x1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v1>
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<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<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<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>

View File

@@ -38,4 +38,12 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 256, 32, 64, Default, 32, 3
DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 16, 64, Filter1x1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v1>
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<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<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<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>

View File

@@ -38,4 +38,12 @@ DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<256, 256, 32, 64, Default, 32, 3
DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle_V3<128, 32, 16, 64, Filter1x1Pad0, 16, 16, 1, 1, 8, 8, 2, 1, 1, BlkGemmPipelineScheduler: Interwave, BlkGemmPipelineVersion: v1>
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<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<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<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>

View File

@@ -60,7 +60,6 @@ class ConvInstanceTemplateParams:
def get_block_gemm_desc(self):
double_smem_buffer = "true" if self.double_smem_buffer else "false"
pipeline_version = self.pipeline_version[-1:]
scheduler = (
"INTRAWAVE" if self.scheduler.find("Intrawave") != -1 else "INTERWAVE"
)
@@ -69,7 +68,7 @@ class ConvInstanceTemplateParams:
.warp_tile = {{.m = {self.warp_tile[0]}, .n = {self.warp_tile[1]}, .k = {self.warp_tile[2]}}},
.double_smem_buffer = {double_smem_buffer},
.num_wave_groups = {self.num_wave_groups},
.pipeline_version = ckb::PipelineVersion::V{pipeline_version},
.pipeline_version = ckb::PipelineVersion::{self.pipeline_version},
.scheduler = ckb::PipelineScheduler::{scheduler}}}"""
def get_block_transfer(self):
@@ -180,6 +179,16 @@ def parse_fwd_instances(instances, problem_name):
pipeline_version = (
"v1" if instance.find("BlkGemmPipelineVersion") == -1 else args[15]
)
# Replace pipeline if Direct Load
if instance.find("DirectLoad") != -1:
if instance.find("BlkGemmPipelineVersion: v1") != -1:
pipeline_version = "ASYNC_V1"
elif instance.find("BlkGemmPipelineVersion: v4") != -1:
pipeline_version = "ASYNC_V4"
else:
raise RuntimeError("not supported pipeline for direct load")
else:
pipeline_version = f"""V{pipeline_version[-1:]}"""
m_warp = int(m_per_block / (m_per_xdl * m_xdl_per_wave))
n_warp = int(n_per_block / (n_per_xdl * n_xdl_per_wave))