Arm Backend: Improve VGF runtime and update MLSDK setup #13155
Conversation
* Adds portability extensions to instance_info * Explicitly sets poolSizes in descriptor pool * Update Arm MLSDK manifest url to default to newly publish repo during setup.sh * Add new exports to setup_path.sh for mlsdk components On behalf of Ryan O'Shea Signed-off-by: Ryan O'Shea <[email protected]> Change-Id: I3b69108c99b7a5144898a564d519d72b8ca7f702
YufengShi-dudu
requested review from
AdrianLundell,
Sebastian-Larsson,
robell,
tom-arm and
zingo
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/13155
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ❌ 2 New Failures, 2 Unrelated FailuresAs of commit c1f05f3 with merge base 414fc32 ( NEW FAILURES - The following jobs have failed:
BROKEN TRUNK - The following jobs failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
YufengShi-dudu
added
partner: arm
meta-cla
bot
added
the
CLA Signed
robell
left a comment
robell
left a comment
There was a problem hiding this comment.
Looks good - was internally reviewed.
| enable_vgf_lib=0 # vgf reader - runtime backend dependency | ||
| enable_emulation_layer=0 # Vulkan layer driver - emulates Vulkan ML extensions | ||
| mlsdk_manifest_url="" | ||
| mlsdk_manifest_url="https://github.com/arm/ai-ml-sdk-manifest.git" |
There was a problem hiding this comment.
From the manifest, most of them are on github.com which is awesome but I do see one on mlplatforms.org (the tosa one). FYI in the past we have seen a lot of reliability issues on that domain, keep an eye out of that if we are enabling this on the CI with high freq like pull.yaml for example.
| VkInstanceCreateFlags instance_flags = 0; | ||
|
|
||
| #ifdef __APPLE__ | ||
| instance_flags |= VK_INSTANCE_CREATE_ENUMERATE_PORTABILITY_BIT_KHR; |
There was a problem hiding this comment.
Curious what will this do on Apple platforms?
There was a problem hiding this comment.
it's ensuring we have a bit better vulkan functional compatibility in situations where we deploy on moltenvk or other layer implementations. See https://www.lunarg.com/wp-content/uploads/2022/04/Portability-Enumeration-Extension-APR2022.pdf for the rough overview.
|
Fails are unrelated |
* Adds portability extensions to instance_info * Explicitly sets poolSizes in descriptor pool * Update Arm MLSDK manifest url to default to newly publish repo during setup.sh * Add new exports to setup_path.sh for mlsdk components Signed-off-by: Ryan O'Shea <[email protected]> Co-authored-by: Ryan O'Shea <[email protected]>
6 participants
On behalf of Ryan O'Shea
Change-Id: I3b69108c99b7a5144898a564d519d72b8ca7f702
cc @digantdesai @freddan80 @per @zingo @oscarandersson8218