Allow native vector intrinsics with fixed component types #7597
Conversation
This file contains hidden or 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
Several intrinsics that support native vectors return the same element type regardless of overload. For native vectors, these need to return a vector of size matching the overload with the fixed element type. This provides the source generation to allow it and the changes to the related lowering operations to generate the correct DXIL. This includes some queries that return boolean information about the vector components in a boolean vector output and also the unary bit operations which operate on integers and return integers containing information about the component bits. Fixes microsoft#7342
pow2clk
force-pushed
the
nvec-shape-intrinsics
branch
from
5843483 to
99e1364
Compare
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
added 2 commits
June 30, 2025 13:21
Enables native vectors for isspecial operations. Left out of previous commit. Test verifies current behavior, but will need to change when microsoft#7588 is closed.
llvm-beanz
reviewed
Jul 7, 2025
| // CHECK: call <13 x i32> @dx.op.unaryBits.v13i64(i32 32, <13 x i64> [[lvec2]]) ; FirstbitLo(value) | ||
| uRes += firstbitlow(lVec2); | ||
|
|
||
| // CHECK: call <13 x i32> @dx.op.unaryBits.v13i32(i32 33, <13 x i32> [[uvec1]]) ; FirstbitHi(value) |
There was a problem hiding this comment.
This isn't testing the full expansion.
tex3d
reviewed
Jul 7, 2025
| @@ -1406,6 +1406,7 @@ def populate_dxil_operations(self): | |||
| # $o in a parameter type means the overload type | |||
| # $r in a parameter type means the resource type | |||
| # $cb in a parameter type means cbuffer legacy load return type | |||
| # $otpl{component} in a return type means the overload template shape with the specified component type | |||
There was a problem hiding this comment.
I have a hard time reading overloads with this name. What about just $o_{component} for overload, but with fixed component type? I think it would be easier to read.
If necessary, this pattern could be extended to the $x0 and $x1 pattern, with $x0_{component}.
tex3d
approved these changes
Jul 7, 2025
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.
Several intrinsics that support native vectors return the same element type regardless of overload. For native vectors, these need to return a vector of size matching the overload with the fixed element type. This provides the source generation to allow it and the changes to the related lowering operations to generate the correct DXIL.
This includes some queries that return boolean information about the vector components in a boolean vector output and also the unary bit operations which operate on integers and return integers containing information about the component bits.
Incidentally moves some intrinsics that have long since not been scalarized out of the scalarized test. They were updated to expect the new output, but were in the wrong location and not tested as thoroughly as their native vector counterparts.
Fixes #7342