[LLVM] Create TBAA information based on the unrelying buffer type #6046
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
Currently, the TBAA information is based on the access type, i.e. the data type from the load or store instruction. When the same memory area is accessed with different types, the corresponding load/store instruction may end up not being aliased to each other. This could lead to incorrect code being generated. An example of when such a situation can occur is when two different buffer_decl's are created for the same buffer: ba = buffer_decl(... dtype = 'int16' ...) bb = buffer_decl(data = ba.data, dtype = 'int32x32' ...) Then instructions ba[x] = 0 ... = bb[x] may be reordered in the final code due to the alias info indicating that they are not aliased.
tqchen
approved these changes
Jul 13, 2020
|
Thanks @kparzysz-quic ! |
trevor-m
pushed a commit
to trevor-m/tvm
that referenced
this pull request
Jul 14, 2020
…ache#6046) Currently, the TBAA information is based on the access type, i.e. the data type from the load or store instruction. When the same memory area is accessed with different types, the corresponding load/store instruction may end up not being aliased to each other. This could lead to incorrect code being generated. An example of when such a situation can occur is when two different buffer_decl's are created for the same buffer: ba = buffer_decl(... dtype = 'int16' ...) bb = buffer_decl(data = ba.data, dtype = 'int32x32' ...) Then instructions ba[x] = 0 ... = bb[x] may be reordered in the final code due to the alias info indicating that they are not aliased.
trevor-m
pushed a commit
to neo-ai/tvm
that referenced
this pull request
Jul 14, 2020
…ache#6046) Currently, the TBAA information is based on the access type, i.e. the data type from the load or store instruction. When the same memory area is accessed with different types, the corresponding load/store instruction may end up not being aliased to each other. This could lead to incorrect code being generated. An example of when such a situation can occur is when two different buffer_decl's are created for the same buffer: ba = buffer_decl(... dtype = 'int16' ...) bb = buffer_decl(data = ba.data, dtype = 'int32x32' ...) Then instructions ba[x] = 0 ... = bb[x] may be reordered in the final code due to the alias info indicating that they are not aliased.
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.
Currently, the TBAA information is based on the access type, i.e. the data type from the load or store instruction. When the same
memory area is accessed with different types, the corresponding load/store instruction may end up not being aliased to each other. This could lead to incorrect code being generated.
An example of when such a situation can occur is when two different
buffer_decls are created for the same buffer:Then instructions
may be reordered in the final code due to the alias info indicating that they are not aliased.