Introduce PlatformMemoryAllocator #5121
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/5121
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 5e0c38a with merge base 657789e ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D60601742 |
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D60601742 |
manuelcandales
added a commit
to manuelcandales/executorch-1
that referenced
this pull request
Sep 6, 2024
Summary: Pull Request resolved: pytorch#5121 We want to establish a pattern where portable ops fall back to some default allocator if the temp allocator isn't available. This would let users override how the memory is allocated if they care about it, but provide a reasonable default if they don't. We add new PAL hooks for allocation, which default to calling malloc and free. This lets bare metal users inject their own allocator if malloc isn't defined for them. This diff: - Adds et_pal_allocate() and et_pal_free() to platform.h - Provides default implementations of those in posix.cpp and minimal.cpp. Posix uses malloc/free, and minimal returns nullptr when trying to allocate. - Adds a MemoryAllocator subclass that uses the new PAL fallback allocator We allocate a little more than requested and use that memory as a node in a linked list, pushing the allocated buffers onto a list that's iterated and freed when the KernelRuntimeContext is destroyed. The runtime installs the default temp allocator (that uses the PAL hook to allocate) if one isn't provided Differential Revision: D60601742
manuelcandales
force-pushed
the
export-D60601742
branch
from
6c3a18f to
62ca667
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D60601742 |
manuelcandales
added a commit
to manuelcandales/executorch-1
that referenced
this pull request
Sep 6, 2024
Summary: Pull Request resolved: pytorch#5121 We want to establish a pattern where portable ops fall back to some default allocator if the temp allocator isn't available. This would let users override how the memory is allocated if they care about it, but provide a reasonable default if they don't. We add new PAL hooks for allocation, which default to calling malloc and free. This lets bare metal users inject their own allocator if malloc isn't defined for them. This diff: - Adds et_pal_allocate() and et_pal_free() to platform.h - Provides default implementations of those in posix.cpp and minimal.cpp. Posix uses malloc/free, and minimal returns nullptr when trying to allocate. - Adds a MemoryAllocator subclass that uses the new PAL fallback allocator We allocate a little more than requested and use that memory as a node in a linked list, pushing the allocated buffers onto a list that's iterated and freed when the KernelRuntimeContext is destroyed. The runtime installs the default temp allocator (that uses the PAL hook to allocate) if one isn't provided Differential Revision: D60601742
manuelcandales
force-pushed
the
export-D60601742
branch
from
62ca667 to
2c9c6b0
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D60601742 |
manuelcandales
force-pushed
the
export-D60601742
branch
from
2c9c6b0 to
c2f9dfa
Compare
manuelcandales
added a commit
to manuelcandales/executorch-1
that referenced
this pull request
Sep 10, 2024
Summary: Pull Request resolved: pytorch#5121 We want to establish a pattern where portable ops fall back to some default allocator if the temp allocator isn't available. This would let users override how the memory is allocated if they care about it, but provide a reasonable default if they don't. We add new PAL hooks for allocation, which default to calling malloc and free. This lets bare metal users inject their own allocator if malloc isn't defined for them. This diff: - Adds et_pal_allocate() and et_pal_free() to platform.h - Provides default implementations of those in posix.cpp and minimal.cpp. Posix uses malloc/free, and minimal returns nullptr when trying to allocate. - Adds a MemoryAllocator subclass that uses the new PAL fallback allocator We allocate a little more than requested and use that memory as a node in a linked list, pushing the allocated buffers onto a list that's iterated and freed when the KernelRuntimeContext is destroyed. The runtime installs the default temp allocator (that uses the PAL hook to allocate) if one isn't provided Differential Revision: D60601742
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D60601742 |
manuelcandales
added a commit
to manuelcandales/executorch-1
that referenced
this pull request
Sep 10, 2024
Summary: Pull Request resolved: pytorch#5121 We want to establish a pattern where portable ops fall back to some default allocator if the temp allocator isn't available. This would let users override how the memory is allocated if they care about it, but provide a reasonable default if they don't. We add new PAL hooks for allocation, which default to calling malloc and free. This lets bare metal users inject their own allocator if malloc isn't defined for them. This diff: - Adds et_pal_allocate() and et_pal_free() to platform.h - Provides default implementations of those in posix.cpp and minimal.cpp. Posix uses malloc/free, and minimal returns nullptr when trying to allocate. - Adds a MemoryAllocator subclass that uses the new PAL fallback allocator We allocate a little more than requested and use that memory as a node in a linked list, pushing the allocated buffers onto a list that's iterated and freed when the KernelRuntimeContext is destroyed. The runtime installs the default temp allocator (that uses the PAL hook to allocate) if one isn't provided Differential Revision: D60601742
manuelcandales
force-pushed
the
export-D60601742
branch
from
c2f9dfa to
d53fb22
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D60601742 |
manuelcandales
added a commit
to manuelcandales/executorch-1
that referenced
this pull request
Sep 10, 2024
Summary: Pull Request resolved: pytorch#5121 We want to establish a pattern where portable ops fall back to some default allocator if the temp allocator isn't available. This would let users override how the memory is allocated if they care about it, but provide a reasonable default if they don't. We add new PAL hooks for allocation, which default to calling malloc and free. This lets bare metal users inject their own allocator if malloc isn't defined for them. This diff: - Adds et_pal_allocate() and et_pal_free() to platform.h - Provides default implementations of those in posix.cpp and minimal.cpp. Posix uses malloc/free, and minimal returns nullptr when trying to allocate. - Adds a MemoryAllocator subclass that uses the new PAL fallback allocator We allocate a little more than requested and use that memory as a node in a linked list, pushing the allocated buffers onto a list that's iterated and freed when the KernelRuntimeContext is destroyed. The runtime installs the default temp allocator (that uses the PAL hook to allocate) if one isn't provided Differential Revision: D60601742
manuelcandales
force-pushed
the
export-D60601742
branch
from
d53fb22 to
eeafc29
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D60601742 |
manuelcandales
added a commit
to manuelcandales/executorch-1
that referenced
this pull request
Sep 10, 2024
Summary: Pull Request resolved: pytorch#5121 We want to establish a pattern where portable ops fall back to some default allocator if the temp allocator isn't available. This would let users override how the memory is allocated if they care about it, but provide a reasonable default if they don't. We add new PAL hooks for allocation, which default to calling malloc and free. This lets bare metal users inject their own allocator if malloc isn't defined for them. This diff: - Adds et_pal_allocate() and et_pal_free() to platform.h - Provides default implementations of those in posix.cpp and minimal.cpp. Posix uses malloc/free, and minimal returns nullptr when trying to allocate. - Adds a MemoryAllocator subclass that uses the new PAL fallback allocator We allocate a little more than requested and use that memory as a node in a linked list, pushing the allocated buffers onto a list that's iterated and freed when the KernelRuntimeContext is destroyed. The runtime installs the default temp allocator (that uses the PAL hook to allocate) if one isn't provided Differential Revision: D60601742
manuelcandales
force-pushed
the
export-D60601742
branch
from
eeafc29 to
888a55e
Compare
Summary: Pull Request resolved: pytorch#5121 We want to establish a pattern where portable ops fall back to some default allocator if the temp allocator isn't available. This would let users override how the memory is allocated if they care about it, but provide a reasonable default if they don't. We add new PAL hooks for allocation, which default to calling malloc and free. This lets bare metal users inject their own allocator if malloc isn't defined for them. This diff: - Adds et_pal_allocate() and et_pal_free() to platform.h - Provides default implementations of those in posix.cpp and minimal.cpp. Posix uses malloc/free, and minimal returns nullptr when trying to allocate. - Adds a MemoryAllocator subclass that uses the new PAL fallback allocator We allocate a little more than requested and use that memory as a node in a linked list, pushing the allocated buffers onto a list that's iterated and freed when the KernelRuntimeContext is destroyed. The runtime installs the default temp allocator (that uses the PAL hook to allocate) if one isn't provided Differential Revision: D60601742
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D60601742 |
manuelcandales
force-pushed
the
export-D60601742
branch
from
888a55e to
5e0c38a
Compare
dbort
approved these changes
Sep 10, 2024
per
added a commit
to per/executorch
that referenced
this pull request
Sep 12, 2024
The introduction of the PlatformMemoryAllocator in pytorch#5121 introduced a hard fault in arm_executor_runner when not supplying the temp_allocator in the MemoryManager, so add it to get tests running again. Signed-off-by: Per Åstrand <per.astrand@arm.com> Change-Id: I357244a15c1b1fdee567ad828e09e3607e43e2e7
facebook-github-bot
pushed a commit
that referenced
this pull request
Sep 12, 2024
Summary: The introduction of the PlatformMemoryAllocator in #5121 introduced a hard fault in arm_executor_runner when not supplying the temp_allocator in the MemoryManager, so add it to get tests running again. Change-Id: I357244a15c1b1fdee567ad828e09e3607e43e2e7 Pull Request resolved: #5302 Reviewed By: manuelcandales Differential Revision: D62582281 Pulled By: digantdesai fbshipit-source-id: c650771d5ad2eddf80b30fcbafaf6e731ca9eed3
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
Summary:
We want to establish a pattern where portable ops fall back to some default allocator if the temp allocator isn't available. This would let users override how the memory is allocated if they care about it, but provide a reasonable default if they don't.
We add new PAL hooks for allocation, which default to calling malloc and free. This lets bare metal users inject their own allocator if malloc isn't defined for them.
This diff:
We allocate a little more than requested and use that memory as a node in a linked list, pushing the allocated buffers onto a list that's iterated and freed when the KernelRuntimeContext is destroyed.
The runtime installs the default temp allocator (that uses the PAL hook to allocate) if one isn't provided
Differential Revision: D60601742