-
**Bug summary**
The current implementation of `joint_exclusive_scan` (at least HIP-like; SSCP does not support them yet) does not seem to support in-place operations, even though the standard requi…
-
When running a program with multiple kernels having arguments mapped on different DDR bank the program fail due to buffer allocated on wrong DDR bank.
**To Reproduce**
Run the `sycl/test/on-device…
-
Hi,
for a SYCL program I'm writing I am requiring some calculations and a FFT to be submitted to the backend.
I have to do this multiple times, so the following code snippet is encapsulated in a…
-
**Bug summary**
Calling `store()` on `sycl::atomic_ref` unexpectedly defaults to `__ATOMIC_SEQ_CST`, causing the following error:
```
fatal error: error in backend: Cannot select: 0x7ee560: ch = …
-
```
sycl::buffer buf_target(target, coeff_count*decomp_mod_count);
sycl::buffer buf_lsp_0(local_small_poly_0.get(), coeff_count);
sycl::buffer buf_lsp_1(local_small_poly_1.get(), coeff_count); …
-
In this issue I want to make the case for an API that allows submitting SYCL kernels directly onto existing CUDA streams (or equivalent native backend queue types). This would allow some applications …
-
In [`671a20e`](https://github.com/windstatus/Control-D-DNS-status-uptime-monitor/commit/671a20e7cf7794fc7c5a7a400f566b486e475a64
), CGH - Proxy (https://controld-network-api.ncprelief.eu.org?iata_code…
-
In cases where a user wants to create an asynchronous function which creates some USM there is currently no good way to release the memory. It would be useful to have a method to enqueue the release o…
-
Hi,
I saw that we have common reducers in hipSYCL/SYCL e.g. [found here](https://github.com/illuhad/hipSYCL/blob/develop/tests/sycl/reduction.cpp#L338)
I need to use the sycl::maximum reducer b…
-
Currently, if we call asynchronous API within a host task the event that is tied to the host task completes when the host task completes, rather than when the operation submitted by the asynchronous A…