[Hexagon] Treat floats as float32 when passing args to offloaded kernels #9010
Conversation
This file contains 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
`TVMArg` can hold a floating point value, but it's stored as `double`. In Hexagon ABI doubles are passed in a register pair, but if the offloaded function was using floats (i.e. float32), it will expect values being passed in single registers. Since floats are much more common on Hexagon, assume all scalar floating point values are floats. This is only an issue with offloading, and can be treated as a limitation (we do something analogous for integers already).
kparzysz-quic
requested review from
areusch,
junrushao,
kazum,
liangfu,
masahi,
tmoreau89,
tqchen,
vinx13 and
ZihengJiang
as code owners
|
Thank you @kparzysz-quic - the PR has been merged |
AndrewZhaoLuo
added a commit
to AndrewZhaoLuo/tvm
that referenced
this pull request
Sep 16, 2021
* main: (102 commits) Implementation of relay_to_tir target hook (apache#8423) [Onnx] Fix NLL Loss tests (apache#8971) [Bugfix] Fix other div zero errors also in rewrite_simplify (apache#8983) [ONNX] enable the onnx tests after PR apache#8274 merged (apache#9019) [Hexagon] Disable `thread_local` on Hexagon (apache#9025) [Hexagon] Allow undefined symbols in libtvm_runtime.so on Hexagon (apache#9024) [Onnx] Add momentum (apache#9000) fix (apache#9021) [Community] @AndrewZhaoLuo -> Reviewer (apache#9020) [Hexagon] Implement model launcher (apache#8986) [Relay][Pass] Add ExtractOperators pass (apache#8996) [BYOC][TensorRT] Add TensorRT own int8 calibration support to TensorRT BYOC integration (apache#8808) [ONNX] Add Einsum converter (apache#8985) Add standalone_crt/ to be part of the wheel package, when available. (apache#9005) [Relay] Remove memory planing from LowerTEPass (apache#8974) [Hexagon] Treat floats as float32 when passing args to offloaded kernels (apache#9010) [Runtime] Pipeline Executor Initial patch. (apache#8702) [Hexagon] `llvm-options` attribute is an array of strings (apache#9011) disable cuda int8 schedule for non-cuda gpu target (apache#9014) [Torch] Add an option to make imported models compatible with the Relay text parser (apache#9015) ...
|
sorry i missed this @tmoreau89 @kparzysz-quic . i think this makes sense for now. it seems like in the post-fused-IRModule world, we could do some analysis of the calling/called function to eliminate this limitation. what do you think? also, seems like we should potentially document the TVM port in a README/docs.tvm.ai somewhere. what are your thoughts about that? |
ylc
pushed a commit
to ylc/tvm
that referenced
this pull request
Sep 29, 2021
…els (apache#9010) `TVMArg` can hold a floating point value, but it's stored as `double`. In Hexagon ABI doubles are passed in a register pair, but if the offloaded function was using floats (i.e. float32), it will expect values being passed in single registers. Since floats are much more common on Hexagon, assume all scalar floating point values are floats. This is only an issue with offloading, and can be treated as a limitation (we do something analogous for integers already).
ylc
pushed a commit
to ylc/tvm
that referenced
this pull request
Jan 13, 2022
…els (apache#9010) `TVMArg` can hold a floating point value, but it's stored as `double`. In Hexagon ABI doubles are passed in a register pair, but if the offloaded function was using floats (i.e. float32), it will expect values being passed in single registers. Since floats are much more common on Hexagon, assume all scalar floating point values are floats. This is only an issue with offloading, and can be treated as a limitation (we do something analogous for integers already).
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
TVMArgcan hold a floating point value, but it's stored asdouble. In Hexagon ABI doubles are passed in a register pair, but if the offloaded function was using floats (i.e. float32), it will expect values being passed in single registers. Since floats are much more common on Hexagon, assume all scalar floating point values are floats. This is only an issue with offloading, and can be treated as a limitation (we do something analogous for integers already).