Qualcomm AI Engine Direct - support SW6100 #16003
Conversation
Summary - Support SOC SW6100 Test plan - Use unit test with -m SW6100
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/16003
Note: Links to docs will display an error until the docs builds have been completed. ❗ 2 Active SEVsThere are 2 currently active SEVs. If your PR is affected, please view them below: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
|
@pytorchbot label "release notes: qualcomm" |
pytorch-bot
bot
added
the
release notes: qualcomm
|
@cccclai @haowhsu-quic |
| QcomChipset.SXR2330P: SocInfo(QcomChipset.SXR2330P, HtpInfo(HtpArch.V79, 8)), | ||
| QcomChipset.QCS9100: SocInfo(QcomChipset.QCS9100, HtpInfo(HtpArch.V73, 8)), | ||
| QcomChipset.SAR2230P: SocInfo(QcomChipset.SAR2230P, HtpInfo(HtpArch.V81, 4)), | ||
| QcomChipset.SW6100: SocInfo(QcomChipset.SW6100, HtpInfo(HtpArch.V81, 4)), |
There was a problem hiding this comment.
@billmguo mentioned in the email saying it should be v79 HTP + 2MB vtcm, can we double check?
There was a problem hiding this comment.
The data was collected in internal QNN codebase. Since we don't have the corresponding device, could you try this configuration on your side?
There was a problem hiding this comment.
The data was collected in internal QNN codebase. Since we don't have the corresponding device, could you try this configuration on your side?
Referring to QC doc 80-74841-45 Rev. AA Qualcomm® Hexagon Tensor Processor Overview for SW6100, it states v79 HTP, but doesn't mention VTCM size. Is the doc inaccurate or outdated?
4 participants
Summary
Test plan