- Refactor blit code and clean ASAN instrumentation
- Use unified function for rocr copy
- Enable shader copy path for unpinned writeBuffer/readBuffer paths
- Set GPU_FORCE_BLIT_COPY_SIZE=16 which means we will use BLIT copy for
pinned copies or unpinned H2D/D2H copies < 16KB
Change-Id: I42045cca79234b340dbf53dafb93044199736ae4
The early return if the thread is not alive causes memory leaks.
Neither doorbell_ or urilocator are released if the thread is not alive.
This change alters the logic so regardless of the thread status the
HostcallListener releases its memory.
Change-Id: Ie912360ec0e2ee257de9937b1a8d7375e6aebd83
hsa_amd_profiling_async_copy_enable is taking 45us for the first call. Disable sdma profiling for enqueuing captured kernel packets and for accumulate command.
Change-Id: I80b51a58c46bccc9c1025e9331515f57c97b5a2a
This reverts commit efce2f77c4.
Reason for revert: Even though this change is valid, this would break backward compatibility.
Change-Id: I9c7cab83198c8d5c8485b11194099162e3e7a874
Currently amd::Monitor can work in FILO mode for the active waits
and cause a delay in wakeup of some threads. That may have a problem
with the current sysmem pool design.
Change-Id: I145081478d1e0b282d8838855c5718f09cf54b69
Runtime may use checkGpuTime() for the wait and not just for the GPU time queries. Hence, the call can't be skipped if profiling isn't enabled.
More changes are required for this optimization.
Change-Id: I79e8918312e755d75f0d26685f2fdc604a8ffb18
- Remove binning logic, although useful it doesnt work in current
scenario as there is no upper limit on the size of allocation. If an
app or framework uses entire VRAM and then creates suballocs, binning
would result in failure.
Change-Id: Icc27c13e433bb4a1f03e82028d8718488b43bfa5
- Default values are being assigned causing occupancy calculation to go
wrong without the right values defined for gfx12 ASICs
- Also added the these values for gfx1105
Change-Id: I611cc3a8ed8c57f2def637310ce1c3a48c16a574
- Device can have multiple isas as per HSA spec
- First isa is most specific one, so this change is sort of a NOP
Change-Id: Ib332af21745f2e6a7c25db8986bf7717501059bc
1. Fix LDSSize type to be uint32_t.
2. Prevent clWaitForEvents running on complete events whose
HostQueue have been destructed.
Change-Id: I829e915f56b37db2ba76bb876c9656166534f154
- Create bins each with its own map and lock. This would help cases
where the hash of a VA is differnet than ther one which falls in
different bin, and there is no lock contention
- Use STL shared mutexes, that way we can unique_lock for map updates
vs simple reads which can use shared_lock
Change-Id: I118818be65c6373700f5e511045babb6a398938a
Add an atomic counter to track the outstanding HSA handlers.
Wait on CPU for the callbacks if the number exceeds the value
in DEBUG_HIP_BLOCK_SYNC env variable.
Change-Id: I95dc8c4bf0258c7e59411b7504220709ed6898c5
Windows alings fields to 8 bytes even with 32bit builds.
Add BUG_CLR_SYSMEM_POOL to cotnrol sysmempool.
Change-Id: I8622aabc9f7391ed7dd8583b252ce9eb41d62293
- Remove the list of all chunks and use embedded chunk
information in each allocation. That simplifies Free() logic,
avoiding expensive loop if for some reason the number of
outstanding allocations significantly grew.
Change-Id: I9ea84d314320ce356ed24dd3180f262e2116c59b
Applications may submit commands withoout waits
for GPU. That causes a growth of SW unreleased commands.
Make sure runtime flushes SW queue, if it grows over some
threshold, controlled by DEBUG_CLR_MAX_BATCH_SIZE.
Change-Id: Ia4d85c24210ef91c394f638ab6b53b14323a0396
- Don't generate callbacks for HIP events
- Don't process profiling info in the callback for HIP events
- Wait for CPU status update of the submitted commands
every 50 calls. That will allow to drain the commands and
destroy HSA signals.
Change-Id: Ib601a350e7e7c2b6c6209a172385389baccf73a9
* In a scenario where kernel is launched with hipExtLaunchKernelGGL and stop event is used, hipGraphInstantiate leaks. Since stop event is used, profiling is enabled and Timestamp (ReferencedCountedObject) is created, but it doesn't get released.
* The idea behind this solution is that profiling should be disabled when command is captured, hence the timestamp should not be created. Because information about capturing isn't available when kernel command is created, packet capturing state is used to determine whether to create a timestamp or not.
Change-Id: Ia23adac4592ded4fb5e236acf99e12e729f63692
Saleel Kudchadker
Jatin Chaudhary
Jimbo Xie
Sourabh Betigeri
Aidan Belton-Schure
taosang2
Anusha GodavarthySurya
Satyanvesh Dittakavi
Jaydeep Patel
Ioannis Assiouras
Maneesh Gupta
German Andryeyev
Rahul Manocha
Konstantin Zhuravlyov
Alex Xie
Julia Jiang
Todd tiantuo Li
kjayapra-amd
Branislav Brzak
Vladana Stojiljkovic