hipBLAS
hipBLAS copied to clipboard
ROCm
ROCm BLAS marshalling library
resolves:[ lwpmlse-644](https://ontrack-internal.amd.com/browse/LWPMLSE-644)
Summary of proposed changes: An upstream llvm change enables ```LLVM_ENABLE_PER_TARGET_RUNTIME_DIR``` by default for the openmp build. This installs the openmp libraries into ```/opt/rocm-ver/llvm/lib/x86_64-unknown-linux-gnu``` instead of ```/opt/rocm-ver/llvm/lib```. Currenty, hipBLAS only looks...
An upstream llvm change enables LLVM_ENABLE_PER_TARGET_RUNTIME_DIR by default for the openmp build. This installs the openmp libraries into `/opt/rocm-ver/llvm/lib/x86_64-unknown-linux-gnu` instead of `/opt/rocm-ver/llvm/lib`. Currenty, hipBLAS only uses /lib. Since hipBLAS uses...
Would the mixed precision "dot" be added to the library ? For mixed precision version (inputs are float while result is double), dot product is computed with double precision. Reference...
Running a CUDA program shows that cublasGemmEx supports compute type CUBLAS_COMPUTE_32F_FAST_TF32 and CUBLAS_GEMM_DEFAULT_TENSOR_OP. The type is not available in hipBLAS. Thank you for your discussion. status = cublasGemmEx(handle, CUBLAS_OP_N, CUBLAS_OP_N,...
Would "axpby" (https://oneapi-src.github.io/oneMKL/domains/blas/axpby.html) be added to hip/roc-BLAS ?
Does the BLAS (e.g. hipblasIdamax) function support 64-bit interface ? Thanks. https://www.intel.com/content/www/us/en/docs/onemkl/developer-reference-dpcpp/2023-0/iamax.html
### Suggestion Description There is a `BUILD_WITH_SOLVER` CMake option, but it would be nice if the rocsolver library could be loaded at run-time with dlopen when `BUILD_WITH_SOLVER` is `OFF`. This...
resolves: hipBLASLt build dependency on hipBLAS by utilizing the hipBLAS-common package. Summary of proposed changes: - Removes datatypes that exist within hipBLAS-common package from hipBLAS - Adds hipblas-common as a...
