Add tests for cl_ext_immutable_memory_objects #2286
Conversation
|
Looks like this is another one where we'll need header changes before we can merge. Would it be possible to get a draft headers PR up with these changes so this won't become a bottleneck once the tests have been reviewed? |
There was a problem hiding this comment.
I don't have a device to test these changes, and I can't even build without the header changes, but the test changes look like a good start.
The command buffer negative testing looks pretty comprehensive. Are there plans to have negative tests for command-queue fills and copies also?
Are there any plans to extend positive testing to confirm that we can read from immutable buffers and images?
|
Thanks for the review. There is a plan to add more tests to cover command queue's fills and copies and reading from buffers and images. |
|
Here's a PR adding definitions to the headers: KhronosGroup/OpenCL-Headers#274 |
MichaelRizkalla-arm
force-pushed
the
cl_ext_immutable_memory_objects_tests_patch_1
branch
from
e0cb8e5 to
41652c3
Compare
|
Please re-add "focused review" once the merge conflicts are resolved. |
|
@MichaelRizkalla-arm Could you resolve the merge conflicts please? |
This change does the following: 1. Create `negative_enqueue_map_image` with tests for mapping images created with `CL_MEM_IMMUTABLE_EXT` 2. Update `enqueue_map_*` with more variants of `cl_mem_flags` Signed-off-by: Michael Rizkalla <[email protected]>
This change updates `cl_khr_command_buffer` test binary to include behaviour with `CL_MEM_IMMUTABLE_EXT` memory flag. Added the following tests: 1. negative_copy_to_immutable_buffer 2. negative_copy_image_to_immutable_buffer 3. negative_copy_to_immutable_buffer_rect 4. negative_copy_to_immutable_image 5. negative_copy_buffer_to_immutable_image 6. negative_fill_immutable_image 7. negative_fill_immutable_buffer Signed-off-by: Michael Rizkalla <[email protected]>
This change adds `negative_set_immutable_memory_to_writeable_kernel_arg` to verify setting memory arguments marked with `CL_MEM_IMMUTABLE_EXT`. Signed-off-by: Michael Rizkalla <[email protected]>
MichaelRizkalla-arm
force-pushed
the
cl_ext_immutable_memory_objects_tests_patch_1
branch
from
41652c3 to
b13aca0
Compare
|
I have updated the PR to resolve the merge conflicts. |
|
Merging as discussed in the June 17th memory subgroup. |
MichaelRizkalla-arm
deleted the
cl_ext_immutable_memory_objects_tests_patch_1
branch
This change provides partial test coverage for KhronosGroup/OpenCL-Docs#1280
Adding CTS tests for:
The bulk of the tests is to make sure that the CL driver does not allow writing to a memory object that is created with
CL_MEM_IMMUTABLE_EXTflag when used with the above APIs.