3 Commits

Author	SHA1	Message	Date
Bartłomiej Kocot	28f2966762	[rocm-libraries] ROCm/rocm-libraries#7734 (commit 03ffb9d) ... [CK] Grouped Convolution Global Load/Store instances ## Motivation Support global load and store in grouped convolutions using instance factory. ## Technical Details - add new instances for each direction - add new tests for large cases ## Test Plan New test for large cases ## Test Result pending ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests. AICK-1255	2026-06-06 22:52:59 +00:00
Johannes Graner	c190d8d61f	[CK tests] Extend conv GPU reference (#3539) ... * test_convnd_fwd * test_convnd_bwd_data * test_conv_bwd_data_scale * test_grouped_convnd_fwd_clamp * test_grouped_convnd_fwd_scale * multiple A/B tensors and D tensor for fwd GPU ref * test_grouped_convnd_fwd_scaleadd_ab * test_grouped_convnd_fwd_bias_clamp * test_grouped_convnd_fwd_bilinear * test_grouped_convnd_fwd_gk_bias_clamp * Extend GPU reference to enable batchnorm epilogue * test_grouped_convnd_fwd{,_gk}_bias_bnorm_clamp * test_grouped_conv_bwd_data_bilinear * test_grouped_convnd_bwd_weight_bilinear * Add missing template instantiation * Perform operations in float in reference * Slightly increase tolerance for batchnorm profiler * Revert "Slightly increase tolerance for batchnorm profiler" This reverts commit a3b2475229. * Revert "test_grouped_convnd_fwd{,_gk}_bias_bnorm_clamp" This reverts commit 6da4576060. * Revert "Extend GPU reference to enable batchnorm epilogue" This reverts commit e2f75fa10e. * Clarify variable names * Refactor elementwise ops into helper functions * Make helpers C++17-compatible	2026-01-27 09:49:42 +01:00
Johannes Graner	bb8445dca8	[CK] Integrate GPU reference into ckProfiler for convolutions (#3379) ... Refactor and integrate CK GPU references into ckProfiler. - All convolution layouts and groupings supported for all three directions - Unit tests verifying GPU and CPU reference is the same - Support added to profiler (do_verification = 2 enables GPU reference) - One profiler-based test per direction changed to GPU reference to demonstrate usag Closes AICK-427	2025-12-18 07:59:45 +01:00