Fix BMv2 variable-offset slice compilation (#5558) - #5716
Open
anovotarsk wants to merge 4 commits into
Open
Conversation
There was a problem hiding this comment.
Pull request overview
This PR addresses BMv2 compilation failures for variable-offset +: slices (e.g., value[offset+:width]) by adding an opt-in strength-reduction transform that lowers these expressions into a form BMv2 can codegen: (value >> offset)[width-1:0]. The transform is enabled for BMv2 midends and includes targeted unit and regression coverage.
Changes:
- Add a
StrengthReductionPolicyoption to lower variable-offsetIR::PlusSliceintoShr+ constantSlice, while preserving write-context behavior and handling signed sources. - Enable the new transform in BMv2 midend pipelines (simple_switch, psa_switch, pna_nic).
- Add gtest coverage and a new sample/regression program with expected outputs.
Reviewed changes
Copilot reviewed 13 out of 14 changed files in this pull request and generated 3 comments.
Show a summary per file
| File | Description |
|---|---|
| frontends/p4/strengthReduction.h | Extends StrengthReductionPolicy and updates DoStrengthReduction to support write-context awareness. |
| frontends/p4/strengthReduction.cpp | Implements the opt-in lowering of variable-offset PlusSlice into >> plus constant Slice. |
| backends/bmv2/simple_switch/midend.cpp | Enables the new strength-reduction policy for the simple_switch BMv2 midend pipeline. |
| backends/bmv2/psa_switch/midend.cpp | Enables the new strength-reduction policy for the PSA BMv2 midend pipeline. |
| backends/bmv2/pna_nic/midend.cpp | Enables the new strength-reduction policy for the PNA NIC BMv2 midend pipeline. |
| test/gtest/strength_reduction.cpp | Adds unit tests verifying default-off behavior, enabled lowering, write-context preservation, and signed-source handling. |
| testdata/p4_16_samples/issue5558-bmv2.p4 | Adds a regression P4 program reproducing issue #5558 for BMv2. |
| testdata/p4_16_samples_outputs/issue5558-bmv2.p4.p4info.txtpb | Expected P4Info output for the new BMv2 regression sample. |
| testdata/p4_16_samples_outputs/issue5558-bmv2.p4.entries.txtpb | Expected P4Runtime entries output for the new BMv2 regression sample. |
| testdata/p4_16_samples_outputs/issue5558-bmv2.p4-stderr | Expected stderr output for the new BMv2 regression sample. |
| testdata/p4_16_samples_outputs/issue5558-bmv2.p4 | Expected post-parse P4 output for the new regression sample. |
| testdata/p4_16_samples_outputs/issue5558-bmv2-first.p4 | Expected early-frontend dump output for the new regression sample. |
| testdata/p4_16_samples_outputs/issue5558-bmv2-frontend.p4 | Expected end-of-frontend dump output for the new regression sample. |
| testdata/p4_16_samples_outputs/issue5558-bmv2-midend.p4 | Expected end-of-midend dump output for the new regression sample. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
ChrisDodd
reviewed
Jul 28, 2026
| if (sourceType == nullptr) return expr; | ||
| if (sourceType->isSigned) { | ||
| auto *unsignedType = IR::Type_Bits::get(sourceType->size, false); | ||
| source = new IR::Cast(source->srcInfo, unsignedType, source); |
Contributor
There was a problem hiding this comment.
Adding a cast here seems wrong -- if the value is negative and the slice goes off the top, the sign bit should be replicated. That's what happens with a signed shift.
Author
There was a problem hiding this comment.
@ChrisDodd thanks for the review. Removed the unsigned cast to preserve signed right-shift semantics and updated the test
anovotarsk
marked this pull request as draft
July 28, 2026 12:35
anovotarsk
force-pushed
the
anovotarsk/bmv2_variable_slice
branch
3 times, most recently
from
July 28, 2026 13:09
186720f to
850b0a1
Compare
anovotarsk
marked this pull request as ready for review
July 28, 2026 13:13
Signed-off-by: Artur Novotarskyi <artur.novotarskyi@gmail.com>
anovotarsk
force-pushed
the
anovotarsk/bmv2_variable_slice
branch
2 times, most recently
from
July 28, 2026 13:15
850b0a1 to
02af277
Compare
Signed-off-by: Artur Novotarskyi <artur.novotarskyi@gmail.com>
Collaborator
|
The failing container test is fixed in #5715. For P4Testgen add the failures to xfails and file an issue, they might have to be fixed separately. |
fruffy
reviewed
Jul 28, 2026
| @@ -0,0 +1,51 @@ | |||
| #include <core.p4> | |||
Collaborator
There was a problem hiding this comment.
Consider adding an stf test to ensure the transformation works correctly.
fruffy
reviewed
Jul 28, 2026
Signed-off-by: Artur Novotarskyi <artur.novotarskyi@gmail.com>
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
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.
Fixes #5558.
Adds an opt-in strength-reduction transform that lowers variable-offset slices such as value[offset+:width] into (value >> offset)[width-1:0], which BMv2 can generate code for. The transform is enabled for all BMv2 targets and preserves write-context behavior.
Includes unit tests and a BMv2 regression test.