* clr: Use graph segment scheduling to process HIP Graphs
* Add a broader path to use capture packet capture for all topologies
* Refactor code
* Use DEBUG_HIP_GRAPH_SEGMENT_SCHEDULING to toggle new vs classic path,
Enabled by default
* clr: Few fixes and improvements
* clr: Detect complex graphs to take classic path
* Use DEBUG_HIP_GRAPH_SEGMENT_SCHEDULING=2 to force segment scheduling
path
* clr: Fix a cornercase stack corruption
* clr: Track commands of segments instead of snapshots
* clr: Fix Batch dispatch logic
* Track fence_dirty_ flag for command of other streams
* Dependency resolution markers can now accomodate dirty fence on cross
streams
---------
Co-authored-by: Ioannis Assiouras <Ioannis.Assiouras@amd.com>
Co-authored-by: Godavarthy Surya, Anusha <agodavar@amd.com>
- The graph nodes have been updated to capture the device ID from the capture stream or the current device when explicitly added.
- Update the device ID for the memcpy node, ensuring that the device where the memory is allocated is taken into account for H2D and D2H pinned operations.
Co-authored-by: Anusha GodavarthySurya <Anusha.GodavarthySurya@amd.com>
- Added missing validation as graph node should not be created
if parameters are invalid
- Fix conversion of input params to graphNode params
Change-Id: I37ab04942b5fb2eb07386850cb7dbbf26f9ca967
[ROCm/clr commit: db8527f655]
After setting the new params in hipDrvGraphExecMemsetNodeSetParams, we
need to update the AQL packet as well, otherwise during the graph launch
it still dispatches the packet which has the original params and not the
updated one.
Change-Id: Ie49a641ba3f66c8085a29f92d88ac6ea6a1c0534
[ROCm/clr commit: ba2ebb3b99]
* When hipMemset3dAsync is captured, a 3d extent can set be as a parameter (depth > 1). That worked on nvidia, but on amd wrong portion of array was filled because when creating Memset3D command, extent dimensions were used to create pitchedPtr, instead of original array width and height.
* Also, when capturing hipMemset3dAsync, nvidia allows any of the extent dimension to be 0, and in that case, no work should be done.
Change-Id: I46a605bf9ae801cd3348e98d528c21263a8eefce
[ROCm/clr commit: ec60bb1aed]
=> GraphExec instance is destroyed before async launch completes,
destroy after all pending graph launches
=> Remove GraphExec destroy during next sync point(hipStreamSync,
hipDeviceSync etc..)
Change-Id: I4df682aae5787fd6e5240a7be936ce50361345d0
[ROCm/clr commit: f9f995c6d0]
- Added the optimized multi stream path in graph execution. It uses a fixed number of async streams in the execution
- Optimize the launch latency, where commands
creation and execution is done at the same time
- Optimize the scheduling to use less barriers and waiting signals if
the same queue can be detected
- The new path is controlled by DEBUG_HIP_FORCE_GRAPH_QUEUES
environment variable, where 0 will use the original path and any other
value will force the number of asynchronous queues for execution
- DEBUG_HIP_FORCE_ASYNC_QUEUE can force single queue async
execution in graphs(applicable for Navi families only)
Change-Id: I7eb40bc15c45f508d6911868a6f6d4c3598d380e
[ROCm/clr commit: 9db52f9a46]
=> Added support to capture multiple AQL Packets.
=> Added Interface to callback to hip runtime from rocclr to allocate
kernel args from the graph kernel arg pool.
=> Enabled Support to capture memset node.
Change-Id: I7e1c2ba06927459e024653058af142bd82192c43
[ROCm/clr commit: bd3a35bde1]
Godavarthy Surya, Anusha
SaleelK
vstojilj
Jaydeep
systems-assistant[bot]
Danylo Lytovchenko
Dittakavi, Satyanvesh
Brzak, Branislav
Betigeri, Sourabh
Assiouras, Ioannis
Xie, Pengda
Godavarthy Surya, Anusha
Marko Arandjelovic
Vladana Stojiljkovic
Marko Arandjelovic
Rahul Manocha
Jaydeep Patel
kjayapra-amd
German Andryeyev