[GHA] Uplift Linux GPU RT version to 23.09.25812.14

[bb-sycl]

Scheduled drivers uplift

[cperkinsintel]

On L0 and OpenCL GPU we have a test unexpectedly passing. That's fine.

On both L0 and OpenCL GPU, the Printf/char.cpp test is failing. Does anyone know anything about that?

And the L0 queue priority/profling tests are failing. Is this known? Or need to be addressed?

********************
Failed Tests (3):
  SYCL :: Plugin/level_zero_queue_priority.cpp
  SYCL :: Plugin/level_zero_queue_profiling.cpp
  SYCL :: Printf/char.cpp

********************
Unexpectedly Passed Tests (1):
  SYCL :: ESIMD/thread_id_test.cpp

[smaslov-intel]

And the L0 queue priority/profling tests are failing. Is this known? Or need to be addressed?

I am trying to address them in #9166

Check file: /__w/llvm/llvm/llvm/sycl/test-e2e/Plugin/level_zero_queue_priority.cpp
Input was:
<<<<<<
1: ZE ---> zeInit(0)
check:20'0 X~~~~~~~~~~~~~~~~~ error: no match found
2: zeInit: Level Zero initialization failure

[bader]

And the L0 queue priority/profling tests are failing. Is this known? Or need to be addressed?

I am trying to address them in #9166

Check file: /__w/llvm/llvm/llvm/sycl/test-e2e/Plugin/level_zero_queue_priority.cpp
Input was:
<<<<<<
1: ZE ---> zeInit(0)
check:20'0 X~~~~~~~~~~~~~~~~~ error: no match found
2: zeInit: Level Zero initialization failure

@smaslov-intel, do you have any updates on this? I see that #9166 is merged.

[smaslov-intel]

@bader : yes, it is merged and I expected the following GPU driver update to be successful. however I see this in a yesterday's attempt: https://github.com/intel/llvm/actions/runs/4987433999/jobs/8929192727

Any idea?

[bader]

@bader : yes, it is merged and I expected the following GPU driver update to be successful. however I see this in a yesterday's attempt: https://github.com/intel/llvm/actions/runs/4987433999/jobs/8929192727

Any idea?

The job fails to push a branch with the driver version update change because the branch already exists. Please, merge the fix to the https://github.com/intel/llvm/tree/ci/update_gpu_driver-linux-23.09.25812.14 branch.

[bader]

---

Failed Tests (1):
SYCL :: Printf/char.cpp

---

Unexpectedly Passed Tests (1):
SYCL :: ESIMD/thread_id_test.cpp

[smaslov-intel]

Failed Tests (1):
SYCL :: Printf/char.cpp

I can reproduce fail locally it is a segfault in L0 RT. Following up with @jandres742

Unexpectedly Passed Tests (1):
SYCL :: ESIMD/thread_id_test.cpp

@v-klochkov : can you see if this is a reasonable new pass?

[v-klochkov]

Unexpectedly Passed Tests (1):
SYCL :: ESIMD/thread_id_test.cpp

@v-klochkov : can you see if this is a reasonable new pass

Yes, we were waiting for GPU driver update to enable that test.

[smaslov-intel]

Unexpectedly Passed Tests (1):
SYCL :: ESIMD/thread_id_test.cpp

@v-klochkov : can you see if this is a reasonable new pass

Yes, we were waiting for GPU driver update to enable that test.

Great, thanks for confirming.

[bader]

Failed Tests (1):
SYCL :: Printf/char.cpp

I can reproduce fail locally it is a segfault in L0 RT. Following up with @jandres742

@smaslov-intel, do you have any updates on this issue? In case it might be helpful: this issue appeared with 22.49.* driver (see #8156 (comment)).

[jandres742]

Failed Tests (1):
SYCL :: Printf/char.cpp

I can reproduce fail locally it is a segfault in L0 RT. Following up with @jandres742

@smaslov-intel, do you have any updates on this issue? In case it might be helpful: this issue appeared with 22.49.* driver (see #8156 (comment)).

this has been fixed in IGC. We need to wait for fix to be promoted to compute-runtime GPU repo to use it.

[bader]

Failed Tests (1):
SYCL :: Printf/char.cpp

I can reproduce fail locally it is a segfault in L0 RT. Following up with @jandres742

@smaslov-intel, do you have any updates on this issue? In case it might be helpful: this issue appeared with 22.49.* driver (see #8156 (comment)).

this has been fixed in IGC. We need to wait for fix to be promoted to compute-runtime GPU repo to use it.

I'd like to update the driver ASAP as old version gating some important changes in DPC++ compiler like switching to opaque pointers. The risk to update to 23.09.* version is that we disable Printf/char.cpp test and we might miss a regression while waiting for the fixed new version. I talked to @smaslov-intel and we agreed to take that risk.
@jandres742, does it sound okay to you?

[jandres742]

Failed Tests (1):
SYCL :: Printf/char.cpp

I can reproduce fail locally it is a segfault in L0 RT. Following up with @jandres742

@smaslov-intel, do you have any updates on this issue? In case it might be helpful: this issue appeared with 22.49.* driver (see #8156 (comment)).

this has been fixed in IGC. We need to wait for fix to be promoted to compute-runtime GPU repo to use it.

I'd like to update the driver ASAP as old version gating some important changes in DPC++ compiler like switching to opaque pointers. The risk to update to 23.09.* version is that we disable Printf/char.cpp test and we might miss a regression while waiting for the fixed new version. I talked to @smaslov-intel and we agreed to take that risk. @jandres742, does it sound okay to you?

thanks @bader. +1 on my side.

[bader]

@intel/llvm-reviewers-runtime, @KseniyaTikhomirova, ping.

[smaslov-intel]

Failed Tests (1):
SYCL :: Printf/char.cpp

I can reproduce fail locally it is a segfault in L0 RT. Following up with @jandres742

@smaslov-intel, do you have any updates on this issue? In case it might be helpful: this issue appeared with 22.49.* driver (see #8156 (comment)).

this has been fixed in IGC. We need to wait for fix to be promoted to compute-runtime GPU repo to use it.

I'd like to update the driver ASAP as old version gating some important changes in DPC++ compiler like switching to opaque pointers. The risk to update to 23.09.* version is that we disable Printf/char.cpp test and we might miss a regression while waiting for the fixed new version. I talked to @smaslov-intel and we agreed to take that risk. @jandres742, does it sound okay to you?

thanks @bader. +1 on my side.

@bader : did you create an issue to re-enable the test when the fixed driver is available?