zjing14

@iq136boy Could you specify which DeviceOp exactly?

Synced with @bghimireamd, CK already has CUBLASLT_EPILOGUE_GELU_AUX, CUBLASLT_EPILOGUE_BIAS, CUBLASLT_EPILOGUE_GELU_AUX_BIAS. We can quickly add CUBLASLT_EPILOGUE_DGELU. For CUBLASLT_EPILOGUE_BGRADB, we need double-check.

@illsilin Could you take care of it?

In #30, we had to increase the threshold of cache search time, since we added many entries into cachetxt. Otherwise, the MIOpenGEMM failed in Jenkins.

@atamazov Checked with ROCm 4.3. Confirmed that the issue has been resolved.