- The implementation in mempool graphs requires refcounting VA object.
That requires release() to update the map only on the actual destruction.
- Add GPU event tracking for paging operation. Otherwise, runtime
may not always flush IB.
Change-Id: Idf99ffb894321a38e04b490116a7ca435635918d
This reverts commit e83c7a0510195fda6640c22d7ad98f056cf95dcd.
Reason for revert: it is a test issue to consider uuid a string it is array of values
Change-Id: Ifb2ae05a99cdaa80e689b83693fc553dd974a249
The wavelimiter init was unintentionally added in one change.
It wasn't supposed to perform any logic, since LC doesn't support it and
the number of waves can be overwritten only with an environment variable.
Change-Id: I447bd1ad685800f874b6a6fb7409dc67e43640ff
Heap initialization used device queue, but it shoudl be used for
cooperative launches only. Heap initialization must use the same queue
as the current dispatch.
Change-Id: I856621bf82bbdeb1c2d0fbc4970e90d09af805cb
Use static VM ID if available for capture.
Add a delay in RGP trace finish to help RGP server to complete
file operations.
Change-Id: Id0dab2fe4398ac2df79f25608fdd06b1f84e7fe0
Windows kills threads on exit without any notification. However,
runtime can still destroy VirtualGPU object from the host thread with
HostQueue destruction.
This change also forces RGP trace transfer on the last capture without
any delays.
Change-Id: I768e87e99e1d23a021e63c12f36e450817743759
OCL runtime uses WGP mode and total CU count reported in WGP.
Realtime values are still in CUs. That can mislead in the test results.
Report realtime in WGP values and convert to CUs for KMD.
Change-Id: I90b82615640734dd655be2b613ccac3cb8483239
Adobe AE allocates/destroys system memory per each frame
in some effects, which is a very expensive operation.
Disable pinning and force sysmem allocation in VidMM.
That allows to trigger resource cache in runtime and
avoid paging operations, caused by pinning.
The actual benchmark doesn't have significant improvement
from the change, but GPU view traces are much cleaner.
Change-Id: I92cdd87d37c64611788f05250f71d6fe489b4d7e
Pinned copy can cause big performance drops, because slow pinning under Windows.
Use up to 128MB for staging transfers. Change staging buffer size to 4MB.
Linux path should still have the old defaults.
Change-Id: I954edceb3ec89e8e670be116aa2d0a9564c8b11c
Sync between compute and SDMA engines can be very expensive under Windows.
Use CP DMA for tiny transfers (< 1KiB) to avoid syncs and improve performance.
Change-Id: I9db39a2199f7b9e337ed08fd36d9cbc150502f1f
HIP can't rely on the resource tracking, used in OCL and requires different explicit sync.
Make sure ROCCLR syncs compute only when SDMA is used and vise versa.
The new logic will allow to enable CPDMA without unnecessary waits.
Change-Id: Ib9d1788cfd5afa5ea2fec4c96a37d8b9c4d0059d
Jaydeep Patel
Jatin Chaudhary
victzhan
German
pghafari
Konstantin Zhuravlyov
Shadi Dashmiz
Xie.AlexBin
Ajay
Saleel Kudchadker
Todd tiantuo Li
Maneesh Gupta
Anusha GodavarthySurya
Jiabao Xie
jatang
Ioannis Assiouras
Tao Sang
Julia Jiang
kjayapra-amd