Array based clUpdateMutableCommandsKHR changes#1984
Array based clUpdateMutableCommandsKHR changes#1984bashbaug merged 2 commits intoKhronosGroup:mainfrom
Conversation
Implementation change to prototype specification changes proposed in KhronosGroup/OpenCL-Docs#1041 Uses OpenCL changes from * KhronosGroup/OpenCL-Docs#1045 * KhronosGroup/OpenCL-Headers#245 * KhronosGroup/OpenCL-CTS#1984
Implementation change to prototype specification changes proposed in KhronosGroup/OpenCL-Docs#1041 Uses OpenCL changes from * KhronosGroup/OpenCL-Docs#1045 * KhronosGroup/OpenCL-Headers#245 * KhronosGroup/OpenCL-CTS#1984
Implementation change to prototype specification changes proposed in KhronosGroup/OpenCL-Docs#1041 Uses OpenCL changes from * KhronosGroup/OpenCL-Docs#1045 * KhronosGroup/OpenCL-Headers#245 * KhronosGroup/OpenCL-CTS#1984
Implementation change to prototype specification changes proposed in KhronosGroup/OpenCL-Docs#1041 Uses OpenCL changes from * KhronosGroup/OpenCL-Docs#1045 * KhronosGroup/OpenCL-Headers#245 * KhronosGroup/OpenCL-CTS#1984
bashbaug
left a comment
There was a problem hiding this comment.
I don't have an easy way to test these changes (unless I pull and build the latest OCK) but they all look reasonable.
I think it's a good proof-point that the versioned extension headers work, but for the CTS tests I think we've generally been OK tracking the latest headers. Since we don't regularly test (or even build) with older headers I think the older code would eventually break anyhow.
I think it would be a good idea to skip these tests for older versions of the extension though, otherwise they'll likely crash unpredictably if the newer functions run on an older implementation. Could this be added?
CTS change to reflect proposed change from KhronosGroup/OpenCL-Docs#1041 Taking advantage of the extension version macros to avoid breaking existing implementations immediately. This reflects PRs: * KhronosGroup/OpenCL-Docs#1045 * KhronosGroup/OpenCL-Headers#245
c34b6ce to
ec8193c
Compare
I've removed the #ifdefs so this PR will require the matching headers to work. Also added in a CL_DEVICE_EXTENSIONS_WITH_VERSION runtime check and skipped testing when a device doesn't support at least 0.9.2 of mutable-dispatch. Although devices which don't support that query will still run into issues |
|
Merging as discussed in the September 3rd teleconference + email. |
fixes a few minor issues introduced by #1984. edit: Tested with the updated command buffer emulation layer on a device that supports out-of-order queues and SVM and all of the `test_cl_khr_mutable_dispatch` tests pass.
CTS changes to reflect the spec changes merged in KhronosGroup/OpenCL-Docs#1045 and requires header updates from KhronosGroup/OpenCL-Headers#245
Tested using OCK implementation from uxlfoundation/oneapi-construction-kit#501