mmoadeli

Assert in Mem2Reg.cpp due to pointer address space mismatch

1

``` func @_ZN2cl4sycl6detail7declptrINS0_4itemILi2ELb1EEEEEPT_v() -> !llvm.ptr attributes {llvm.linkage = #llvm.linkage} { %false = arith.constant false %true = arith.constant true %0 = memref.alloca() : memref memref.store %true, %0[] : memref %1 =...

Address [test-e2e/Basic/bit_cast/bit_cast.cpp](https://github.com/intel/llvm/blob/sycl/sycl/test-e2e/Basic/bit_cast/bit_cast.cpp) build's missing triple on Windows.

Running the test on Windows failed due to missing support of `ls`. Replacing `ls` with `cat` made the test pass on Windows. Additionally, the patch removes checking the case when...

Modified return types of the following device attributes for CUDA and HIP backends to ensure compliance with [specification](https://oneapi-src.github.io/unified-runtime/core/api.html#ur-device-info-t) - UR_DEVICE_INFO_IMAGE_MAX_ARRAY_SIZE - UR_DEVICE_INFO_MAX_PARAMETER_SIZE - UR_DEVICE_INFO_PROFILING_TIMER_RESOLUTION - UR_DEVICE_INFO_PRINTF_BUFFER_SIZE There were failures on...

[CUDA] Remove the support of concurrent atomic access to host allocated pinned memory.

3

Addresses the removal of `UR_DEVICE_USM_ACCESS_CAPABILITY_FLAG_ATOMIC_CONCURRENT_ACCESS` from the supported features of host memory. This change is prompted by the incompatibility of CUDA with concurrent atomic access between host and device to...

[TEST-E2E][CUDA] Do not allow adding an event as a BlockedUser if it has the current event as a dependent.

1

Adding an event as a BlockedUser if it has the current event as a dependent could potentially result in failure as seen in related issue. Using the patch, there is...

### Describe the bug Below is simplified reproducer of the failure seen in [e2e/InOrderEventsExt/get_last_event.cpp](https://github.com/intel/llvm/blob/sycl/sycl/test-e2e/InOrderEventsExt/get_last_event.cpp) when built/ran on Windows/CUDA(GTX 1650 / sm_75). The failure is seen intermittently and not on all...

`prefetch` on sycl with cuda backend on Windows

1

[Windows doesn't support cudaMemPrefetchAsync()](https://github.com/TimDettmers/bitsandbytes/issues/453) which is used in the call to `prefetch`. Would be a need to investigate alternatives for Windows. https://github.com/intel/llvm/pull/13957 may be used to further information.