-
Notifications
You must be signed in to change notification settings - Fork 124
Batch adapter changes: spec motivated refactors #1212
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Batch adapter changes: spec motivated refactors #1212
Conversation
Codecov ReportAttention:
❗ Your organization needs to install the Codecov GitHub app to enable full functionality. Additional details and impacted files@@ Coverage Diff @@
## main #1212 +/- ##
==========================================
- Coverage 15.44% 15.38% -0.07%
==========================================
Files 239 240 +1
Lines 33970 34135 +165
Branches 3758 3781 +23
==========================================
+ Hits 5246 5250 +4
- Misses 28673 28834 +161
Partials 51 51 ☔ View full report in Codecov by Sentry. |
5f073f3 to
e66d5bc
Compare
100798b to
16e8095
Compare
a7c04c0 to
43313b0
Compare
7e5fd51 to
69ddf01
Compare
9ead86f to
8089d80
Compare
8089d80 to
be51455
Compare
|
the e2e tests aren't going to work here since this requires a change on the LLVM side to build |
5749dea to
15b8ce2
Compare
15b8ce2 to
719e62d
Compare
Also removes these checks from the cuda and hip adapters since they should be caught by the validation layer now. The new validation rules bring image_desc_t in line with the CL struct it was derived from. We already had CTS coverage for them.
This brings it in line with OpenCL, and with SYCL's expectations. Addresses oneapi-src#1038
This gates support for the EnqueueDeviceGlobalVariableRead and Write operations. The CTS tests for these now check for support before running, and all adapters report their support correctly.
The native CPU adapter has been added to CI since some of these PRs were reviewed, so inevitably we need to tack some match file fixes on here post scriptum.
* Return UNSUPPORTED_ENUMERATION in device info query when FP16
extension isn't supported
* Add a mapping from CL_INVALID_ARG_SIZE to
UR_RESULT_ERROR_INVALID_KERNEL_ARGUMENT_SIZE
* Remove a bunch of {{OPT}} entries from the enqueue match file, these
should all be passing since bounds checking was added to the
validation layer
* Write in a special case for PROGRAM_INFO_BINARIES test - this query
works pretty differently to the others
This lines up better with the original PI error code and what it was meant to convey. Also clarify the spec around return codes for urProgramGetFunctionPointer and update the relevant CTS test.
719e62d to
2f121cd
Compare
Pulls in changes from oneapi-src/unified-runtime#1212 --------- Co-authored-by: Kenneth Benzie (Benie) <[email protected]>
This is a batch PR containing a number of spec-motivated refactors/small changes. Changes include:
llvm testing intel/llvm#12681