[vmvx] Relax requirement on fptosi lowering #11553
Merged
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
hanhanW
previously approved these changes
Dec 14, 2022
| auto resultType = getTypeConverter()->convertType(dstType); | ||
| // This uses the resultType rather than dstType as any truncation | ||
| // required will be handled via interpretation by consumer. | ||
| if (resultType.isSignlessInteger(32) || resultType.isSignedInteger(32)) { | ||
| rewriter.replaceOpWithNewOp<IREE::VM::CastF32SI32Op>( | ||
| srcOp, resultType, adaptor.getOperands()[0]); |
There was a problem hiding this comment.
since you're touching the code around this area, could you also update adaptor.getOperands()[0] to adaptor.getIn()?
There was a problem hiding this comment.
I was referring l.935, but it's still good that you fix the other part..
hanhanW
reviewed
Dec 14, 2022
| auto resultType = getTypeConverter()->convertType(dstType); | ||
| // This uses the resultType rather than dstType as any truncation | ||
| // required will be handled via interpretation by consumer. | ||
| if (resultType.isSignlessInteger(32) || resultType.isSignedInteger(32)) { | ||
| rewriter.replaceOpWithNewOp<IREE::VM::CastF32SI32Op>( | ||
| srcOp, resultType, adaptor.getOperands()[0]); |
There was a problem hiding this comment.
I was referring l.935, but it's still good that you fix the other part..
Comment on lines
+80
to
82
| // CHECK: vm.cast.f32.si32 %[[ARG0]] : f32 -> i32 | ||
| %1 = arith.fptosi %arg0 : f32 to i8 | ||
| return %1 : i8 |
There was a problem hiding this comment.
Are there other IRs that truncate i32 to i8? If so, should we check it as well?
There was a problem hiding this comment.
We checked the integer ones yesterday (8a3d270). At least for those individually the result wrt llvm and vmvx was the same.
The consumption of si of small width is already handled successfully by consumers of this op. This was not true for the ui version - although I suspect consumer op there, but keeping it more restricted until checked.
This was referenced Dec 30, 2022
pzread
pushed a commit
to pzread/iree
that referenced
this pull request
Jan 4, 2023
This reverts commit cee9b3c.
pzread
pushed a commit
to pzread/iree
that referenced
this pull request
Jan 11, 2023
This reverts commit cee9b3c.
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.
The consumption of si of small width is already handled successfully by
consumers of this op (at least in limited trial). This was not true for the ui version ---although I suspect consumer op there, but keeping it more restricted until
checked.