Skip to content

Fix Driver Ext Tracing Legacy driver compatability - #507

Merged
rwmcguir merged 1 commit into
oneapi-src:masterfrom
nrspruit:fix_ext_tracing_compat
Aug 26, 2026
Merged

Fix Driver Ext Tracing Legacy driver compatability#507
rwmcguir merged 1 commit into
oneapi-src:masterfrom
nrspruit:fix_ext_tracing_compat

Conversation

@nrspruit

Copy link
Copy Markdown
Contributor

No description provided.

@nrspruit
nrspruit requested a review from rwmcguir August 26, 2026 18:28
@nrspruit
nrspruit force-pushed the fix_ext_tracing_compat branch from e10a48c to 64f8fb0 Compare August 26, 2026 18:44
Signed-off-by: Neil R. Spruit <neil.r.spruit@intel.com>
@nrspruit
nrspruit force-pushed the fix_ext_tracing_compat branch from 64f8fb0 to e19c6de Compare August 26, 2026 18:51
@rwmcguir

Copy link
Copy Markdown
Contributor

Two small things.

pfnGet is called twice, but the first call's count is only used for the 0 == driverCount check — which the second call already reports. One call would do.

Bigger question: the loader keys extension registration and fan-out per handle (ExtensionFunctionKey{hDriver, functionName} in tracing_imp.h), but this gates per library using one cached handle. That's fine if zelDriverEnableTracing is defined as library-global — all handles in a driver share a code base, so one handle can speak for all of them. Could the header say that explicitly? Otherwise a driver that keeps the flag per handle silently traces only the first one.

@rwmcguir
rwmcguir merged commit 6142c13 into oneapi-src:master Aug 26, 2026
53 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants