Ye Luo

Results 358 comments of Ye Luo

@cgmb This remains super confusing what is going on with 5.2 I noticed ``` /opt/rocm-5.2.0/include/hipblas.h ``` gcc prints ": warning : This file is deprecated. Use the header file from...

@saadrahim besides library issue, I still struggle with the HIP as a CMake language https://github.com/ROCm-Developer-Tools/HIP/pull/2776

I no more have issues with recent ROCm releases.

I will close this ticket when the first bits of progress shows up in the release.

My first ride is pretty rough. Here is the list of issues to track https://github.com/ROCm-Developer-Tools/HIP/issues/2433 ~~https://github.com/ROCm-Developer-Tools/HIP/pull/2434~~ superseded by https://github.com/ROCm-Developer-Tools/HIP/pull/2776 ~~https://github.com/ROCm-Developer-Tools/hipamd/pull/7~~ https://github.com/RadeonOpenCompute/ROCm/issues/1636 https://gitlab.kitware.com/cmake/cmake/-/issues/23006 My goal is to get the following working...

I just noticed another trouble https://github.com/RadeonOpenCompute/ROCm/issues/1717

Here is my understanding. ROCm is a software suite with compilers, runtime libraries, accelerated numerical libraries, AI related libraries and more. "Support" simply means given hardware are validated at AMD...

@nartmada this was the source of the pain described https://github.com/ROCm/ROCm/issues/2084

ROCm 5.7.0, the situation seems improved. Using the default SVM implementation, the timing improved. ``` yeluo@epyc-server:~/temp/many_transfers$ OMP_NUM_THREADS=8 ./many_transfer.omp.x Function many_transfer.omp HostRegistering takes 1.46004e+07 us Function many_transfer.omp thread 0 takes 40738.9...