Romanov Vlad comments

Results 19 comments of


                                            Romanov Vlad

[SYCL] Add invalid kernel handling

BTW. It would be nice to have a test.

[SYCL][OpenCL] Host accessors wait for barriers when not needed

>> Apparently, the constructor of the host accessor calls clEnqueueMapBuffer with the barrier in its event_wait_list. Could you please tell why you think so? Looking at logs with `SYCL_PI_TRACE` env...

[SYCL] Program reports it does not contain the kernel requested 0 (CL_SUCESS) when using functor on GPU

While the `sycl::program` class is going away, I believe the same problem exists with `sycl::kernel_bundle` as well. @gmlueck Could you please clarify if anything is going to be changed in...

[SYCL] Add some trivial util functions for half type.

Hi, @jinge90 Since the functions are in public namespace could you please clarify if we have an extension document covering them? If not, shouldn't we add/update existing one? Could you...

[SYCL] Add some trivial util functions for half type.

> Hi, @romanovvlad As we discussed, we can proceed with this PR and I will submit a new PR to update the doc, is it OK? > > @intel/llvm-reviewers-runtime Could...

[SYCL] Balance device reference counter

Related github issue https://github.com/intel/llvm/issues/6034

[SYCL] Balance device reference counter

@smaslov-intel Could you please review?

[SYCL] Explicit fill broken when using nested template types

> > > > Yes, the restrictions on the type of `T` in the `handler::fill()` function seem weird to me. Currently, we allow only scalar and vector types, but I...

[SYCL] Explicit fill broken when using nested template types

The removal of types restriction for `handler::fill` is being done here: https://github.com/intel/llvm/pull/6714

[SYCL] Explicit fill broken when using nested template types

> @romanovvlad, should we link this PR to close the issue when it's merged? Not sure, the original issue is that the integration header is incorrectly generated for the case...