[OPENCL][TEXTURE] Improved texture memory planning #17571
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Except the relay backend changes everything is reusable for Texture support in Relax. |
srkreddy1238
force-pushed
the
texture-memory-plan
branch
3 times, most recently
from
36e9ea7 to
56d0739
Compare
|
@tvm-bot rerun |
1 similar comment
|
@tvm-bot rerun |
tqchen
requested changes
Jan 19, 2025
include/tvm/runtime/device_api.h
Outdated
| * \param mem_scope The memory scope of allocated tensor. | ||
| * \return The allocated device pointer. | ||
| */ | ||
| virtual void* AllocDataSpaceView(Device dev, void* data, ShapeTuple shape, DLDataType dtype, |
There was a problem hiding this comment.
given this is specific to OpenCL, let us still strive to keep it within opencl allocator interface instead of going through the DeviceAPI
There was a problem hiding this comment.
I see this is the clean way to keep minimal changes in other modules (graph runtime, ndarray, memory manager ..etc).
In Relax also , I am mapping alloc_storage to allocate cl_buffer and alloc_tensor does a view over it by this device API call. (WIP Ref. srkreddy1238@a6376b9#diff-847ee73fb0b77db96cce920da6cbae223f6bdb026ea125514122e96630356c9b)
Later, this also allows easy path for CLML memory management going through TVM memory_manager interface and also features like GMEM (on chip memory of AdrenoGPU) support for TVM ..etc.
Let me know if you have different advice, I can explore the possibilities.
There was a problem hiding this comment.
Would be great to start with thinking along the direction of special allocator https://github.com/apache/tvm/blob/main/include/tvm/runtime/memory/memory_manager.h
My reading is that seems the main issue lies in the need to get Tensor from existing Buffer in an customized fashion, perhaps we can extend Allocator interface to enable such view
There was a problem hiding this comment.
My reading is that seems the main issue lies in the need to get Tensor from existing Buffer in an customized fashion, perhaps we can extend Allocator interface to enable such view
True, where the backing buffer is used as it as or many image views created over based on memory plan.
With view over NDArray or special Allocator we need to reach opencl device api for final view creation which happen by OpenCL call clCreateImage . We create a new cl_mem (image) from an existing cl_mem(buffer) as backing buffer.
Current flow is
storage_pool populated by: Allocator->Empty => NDArray.
Data_entry_ populated by: NDArray => NDArray::CreateView => DeviceAPI::AllocDataSpaceView => NDArray
We can change this to Allocator interface by
Special Allocator (Extended from Allocator with new call for View) registered from OpenCL Device API at Init.
storage_pool by : Allocator->Alloc => StorageObj
data_entry_ : StorageObj => AllocNDArrayWithScope => Allocator::CreateView (access OpenCLWorkSpace and create view) => NDArray
Is my understanding correct here ?
There was a problem hiding this comment.
In case of VM (Relax)
alloc_storage : Allocator->Alloc (always scope "global") => DeviceAPI::AllocDataSpace =>StorageObj
alloc_tensor : StorageObj::AllocNDArrayScoped => DeviceAPI::AllocDataSpaceView => NDArray
Ref. AllocNDArrayScoped with destructor that calls FreeDataSpaceView for clean up.
srkreddy1238@a6376b9#diff-847ee73fb0b77db96cce920da6cbae223f6bdb026ea125514122e96630356c9b
There was a problem hiding this comment.
We might be able to in tihs case have StorageObj to refer back to a raw ptr of allocator, in this case
StorageObj => AllocNDArrayWithScope => allocator_-> CreateView(Storage storage). The main thing is you need an allocator specific dispatch to create such view, while previously we don't have to.
That does mean allocator needs to have a new virtual function(CreateView)
There was a problem hiding this comment.
Up for another review where, we have new allocator for OpenCL and now the allocation goes by the suggested flow
Have added a packed helper function like DeviceAllocator.opencl where the memory manager can query if the the device has specialized Allocator for a given type like Naive or Pooled.
We are now clear from DeviceAPI interface modification. Let me know how does it look now.
srkreddy1238
force-pushed
the
texture-memory-plan
branch
from
753835e to
2f7864f
Compare
srkreddy1238
force-pushed
the
texture-memory-plan
branch
4 times, most recently
from
061ec2f to
4c4a81d
Compare
|
@tvm-bot rerun |
srkreddy1238
force-pushed
the
texture-memory-plan
branch
from
4120af7 to
3571aa5
Compare
|
@tvm-bot rerun |
|
Failed to re-run CI in https://github.com/apache/tvm/actions/runs/13300072214 with response |
|
@tvm-bot rerun |
|
Failed to re-run CI in https://github.com/apache/tvm/actions/runs/13300093925 with response |
Motivated form the fact that textures can be allocated over a clBuffer object and the size of backing clBuffer can be computed based on hardware image pitch alignment. This optimizes the overall memory allocation on device and helps greately the models with large memory requirements. Improvised the graph memory planner to not differentiate buffer and texture storage tokens and reuse them across. The texture pool in OpenCL runtime is rebranded as memory pool that handles allocation for both buffer and image objects. NDArray to DeviceAPI interface is extended with AllocDataSpaceView and FreeDataSpaceView. These new API's acommodates accessing same physical memory as clBuffer / clImage objects.
srkreddy1238
force-pushed
the
texture-memory-plan
branch
from
4c2e3c0 to
ba9b2fd
Compare
|
@tqchen good for review now :) |
tqchen
approved these changes
Feb 14, 2025
|
thanks @srkreddy1238 , merging so we can have changes in before relay phasing out |
ShiboXing
pushed a commit
to ShiboXing/tvm
that referenced
this pull request
Aug 10, 2025
* [OPENCL][TEXTURE] Improved texture memory planning Motivated form the fact that textures can be allocated over a clBuffer object and the size of backing clBuffer can be computed based on hardware image pitch alignment. This optimizes the overall memory allocation on device and helps greately the models with large memory requirements. Improvised the graph memory planner to not differentiate buffer and texture storage tokens and reuse them across. The texture pool in OpenCL runtime is rebranded as memory pool that handles allocation for both buffer and image objects. NDArray to DeviceAPI interface is extended with AllocDataSpaceView and FreeDataSpaceView. These new API's acommodates accessing same physical memory as clBuffer / clImage objects.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Motivated form the fact that textures can be allocated over a clBuffer object and the size of backing clBuffer can be computed based on hardware image pitch alignment.
This optimizes the overall memory allocation on device and helps greately the models with large memory requirements.
Improvised the graph memory planner to not differentiate buffer and texture storage tokens and reuse them across. The texture pool in OpenCL runtime is rebranded as memory pool that handles allocation for both buffer and image objects.
NDArray to DeviceAPI interface is extended with AllocDataSpaceView and FreeDataSpaceView. These new API's acommodates accessing same physical memory as clBuffer / clImage objects.