unified-runtime
unified-runtime copied to clipboard
oneapi-src
Building on top of https://github.com/intel/llvm/pull/12604 + https://github.com/oneapi-src/unified-runtime/pull/1318 which adds `handleOutOfResources` to dpcpp and returns `UR_RESULT_ERROR_OUT_OF_RESOURCES`, the local mem size check: https://github.com/oneapi-src/unified-runtime/blob/f086f369cab557bf2a589e22bfc37e18d7de5fa8/source/adapters/cuda/enqueue.cpp#L294-L298 should also return `UR_RESULT_ERROR_OUT_OF_RESOURCES` and have dedicated error handling...
Currently, the error code of `urVirtualMemMap` is invalid argument when the VA is already mapped. But this error code is not clear enough, since returning invalid argument could also be...
There are at least 3 places where RAII wrappers for UR handles have been either implemented or requested recently: * https://github.com/oneapi-src/unified-runtime/blob/main/test/conformance/testing/include/uur/raii.h * https://github.com/oneapi-src/unified-runtime/pull/1286/files#diff-d0ae1ee0edc109c505db620df503c365a6daf95b5f4ef9f12ba06a7d16e0b3d8 * https://github.com/intel/llvm/pull/12589#discussion_r1475997415 Given this is a common...
During discussion of #537 it came up that we don't have a well defined notion of optionality in the UR spec. A specific example of this is that due to...
piProgramBuild receives a list of devices, while urProgramBuild does not. This produces a series of issues when a UR program needs to be created for a specific device. So we...
While testing adapter third bump [PR](https://github.com/oneapi-src/unified-runtime/pull/1072), L0 enqueue tests was failing sometimes with a heap APIs problems like [here](https://github.com/oneapi-src/unified-runtime/actions/runs/6893152000/job/18781929391), it have aborted with `free(): corrupted unsorted chunks`, on a rerun...
We've been investigating changing the PI interface for memory allocations and also to some extent for memory transfers, which in turns also changes some of the meaning of the PI...
To ensure https://github.com/oneapi-src/unified-runtime/pull/1206 doesn't break. This includes: * CUDA * OpenCL * Native CPU (see https://github.com/oneapi-src/unified-runtime/issues/1069 ) HIP is also missing, but is not officially supported on Windows yet.
In https://github.com/oneapi-src/unified-runtime/pull/1139 I proposed testing the availability of the python `mako` module during configuration, but as we use several other python modules such as yaml / sphinx / breathe etc,...
