JIT: Canonicalize '(A & pow2) == pow2' -> '(A & pow2) != 0' by BoyBaykiller · Pull Request #128533 · dotnet/runtime

BoyBaykiller · 2026-05-24T03:31:11Z

Basically so we always have 0 at the right-side:

'(A & pow2) == pow2' -> '(A & pow2) != 0'
'(A & pow2) != pow2' -> '(A & pow2) == 0'

Tthis currently helps optimizeBools, but I am also adding this with the future in mind.
Here is one example:

staticboolExample(intfoo){return((foo&4)==4)||((foo&8)==8);}

;; ------ BASEG_M000_IG02: ;; offset=0x0000testcl,4je SHORT G_M000_IG05G_M000_IG03: ;; offset=0x0005moveax,1G_M000_IG04: ;; offset=0x000AretG_M000_IG05: ;; offset=0x000Btestcl,8 setne almovzxrax,alG_M000_IG06: ;; offset=0x0014ret;; ------ DIFFG_M33938_IG02: ;; offset=0x0000moveax,ecxandeax,4andecx,8oreax,ecx setne almovzxrax,al

Note: With #126852 this would then get transformed into just test 12.

* '(A & pow2) != pow2' -> '(A & pow2) == 0'

dotnet-policy-service · 2026-05-24T03:32:18Z

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

MichalPetryka · 2026-05-24T22:43:16Z

@MihuBot -nuget

…it-test-to-zero-on-right

Copilot

Pull request overview

Canonicalizes (A & pow2) == pow2 to (A & pow2) != 0 (and the inverse for !=) during morph relational-comparison optimization. This puts the comparison into a uniform form with 0 on the right, which immediately helps optimizeBools and sets the stage for further follow-on optimizations (e.g., the distributive transform in #126852).

Changes:

Add a canonicalization in fgOptimizeRelationalComparison that, when matching (A & pow2) ==/!= pow2, reverses the relop and bashes the RHS constant to zero, with VN updated accordingly.

…it-test-to-zero-on-right

BoyBaykiller · 2026-06-02T19:10:06Z

@EgorBo I fixed the merge conflicts.

…TreeDone

BoyBaykiller · 2026-06-04T15:00:06Z

Are we good to merge :) ? I'd like to continue follow up work @EgorBo

EgorBo · 2026-06-04T16:07:57Z

/ba-g known CSharpMissingShebangInFileBasedProgram failure

BoyBaykiller · 2026-05-24T03:31:11Z

Basically so we always have 0 at the right-side:

'(A & pow2) == pow2' -> '(A & pow2) != 0'
'(A & pow2) != pow2' -> '(A & pow2) == 0'

Tthis currently helps optimizeBools, but I am also adding this with the future in mind.
Here is one example:

staticboolExample(intfoo){return((foo&4)==4)||((foo&8)==8);}

;; ------ BASEG_M000_IG02: ;; offset=0x0000testcl,4je SHORT G_M000_IG05G_M000_IG03: ;; offset=0x0005moveax,1G_M000_IG04: ;; offset=0x000AretG_M000_IG05: ;; offset=0x000Btestcl,8 setne almovzxrax,alG_M000_IG06: ;; offset=0x0014ret;; ------ DIFFG_M33938_IG02: ;; offset=0x0000moveax,ecxandeax,4andecx,8oreax,ecx setne almovzxrax,al

Note: With #126852 this would then get transformed into just test 12.

* '(A & pow2) != pow2' -> '(A & pow2) == 0'

dotnet-policy-service · 2026-05-24T03:32:18Z

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

MichalPetryka · 2026-05-24T22:43:16Z

@MihuBot -nuget

…it-test-to-zero-on-right

Copilot

Pull request overview

Canonicalizes (A & pow2) == pow2 to (A & pow2) != 0 (and the inverse for !=) during morph relational-comparison optimization. This puts the comparison into a uniform form with 0 on the right, which immediately helps optimizeBools and sets the stage for further follow-on optimizations (e.g., the distributive transform in #126852).

Changes:

Add a canonicalization in fgOptimizeRelationalComparison that, when matching (A & pow2) ==/!= pow2, reverses the relop and bashes the RHS constant to zero, with VN updated accordingly.

…it-test-to-zero-on-right

BoyBaykiller · 2026-06-02T19:10:06Z

@EgorBo I fixed the merge conflicts.

…TreeDone

BoyBaykiller · 2026-06-04T15:00:06Z

Are we good to merge :) ? I'd like to continue follow up work @EgorBo

EgorBo · 2026-06-04T16:07:57Z

/ba-g known CSharpMissingShebangInFileBasedProgram failure

BoyBaykiller · 2026-05-24T03:31:11Z

Basically so we always have 0 at the right-side:

'(A & pow2) == pow2' -> '(A & pow2) != 0'
'(A & pow2) != pow2' -> '(A & pow2) == 0'

Tthis currently helps optimizeBools, but I am also adding this with the future in mind.
Here is one example:

staticboolExample(intfoo){return((foo&4)==4)||((foo&8)==8);}

;; ------ BASEG_M000_IG02: ;; offset=0x0000testcl,4je SHORT G_M000_IG05G_M000_IG03: ;; offset=0x0005moveax,1G_M000_IG04: ;; offset=0x000AretG_M000_IG05: ;; offset=0x000Btestcl,8 setne almovzxrax,alG_M000_IG06: ;; offset=0x0014ret;; ------ DIFFG_M33938_IG02: ;; offset=0x0000moveax,ecxandeax,4andecx,8oreax,ecx setne almovzxrax,al

Note: With #126852 this would then get transformed into just test 12.

* '(A & pow2) != pow2' -> '(A & pow2) == 0'

dotnet-policy-service · 2026-05-24T03:32:18Z

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

MichalPetryka · 2026-05-24T22:43:16Z

@MihuBot -nuget

…it-test-to-zero-on-right

Copilot

Pull request overview

Canonicalizes (A & pow2) == pow2 to (A & pow2) != 0 (and the inverse for !=) during morph relational-comparison optimization. This puts the comparison into a uniform form with 0 on the right, which immediately helps optimizeBools and sets the stage for further follow-on optimizations (e.g., the distributive transform in #126852).

Changes:

Add a canonicalization in fgOptimizeRelationalComparison that, when matching (A & pow2) ==/!= pow2, reverses the relop and bashes the RHS constant to zero, with VN updated accordingly.

…it-test-to-zero-on-right

BoyBaykiller · 2026-06-02T19:10:06Z

@EgorBo I fixed the merge conflicts.

…TreeDone

BoyBaykiller · 2026-06-04T15:00:06Z

Are we good to merge :) ? I'd like to continue follow up work @EgorBo

EgorBo · 2026-06-04T16:07:57Z

/ba-g known CSharpMissingShebangInFileBasedProgram failure

BoyBaykiller · 2026-05-24T03:31:11Z

Basically so we always have 0 at the right-side:

'(A & pow2) == pow2' -> '(A & pow2) != 0'
'(A & pow2) != pow2' -> '(A & pow2) == 0'

Tthis currently helps optimizeBools, but I am also adding this with the future in mind.
Here is one example:

staticboolExample(intfoo){return((foo&4)==4)||((foo&8)==8);}

;; ------ BASEG_M000_IG02: ;; offset=0x0000testcl,4je SHORT G_M000_IG05G_M000_IG03: ;; offset=0x0005moveax,1G_M000_IG04: ;; offset=0x000AretG_M000_IG05: ;; offset=0x000Btestcl,8 setne almovzxrax,alG_M000_IG06: ;; offset=0x0014ret;; ------ DIFFG_M33938_IG02: ;; offset=0x0000moveax,ecxandeax,4andecx,8oreax,ecx setne almovzxrax,al

Note: With #126852 this would then get transformed into just test 12.

* '(A & pow2) != pow2' -> '(A & pow2) == 0'

dotnet-policy-service · 2026-05-24T03:32:18Z

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

MichalPetryka · 2026-05-24T22:43:16Z

@MihuBot -nuget

…it-test-to-zero-on-right

Copilot

Pull request overview

Canonicalizes (A & pow2) == pow2 to (A & pow2) != 0 (and the inverse for !=) during morph relational-comparison optimization. This puts the comparison into a uniform form with 0 on the right, which immediately helps optimizeBools and sets the stage for further follow-on optimizations (e.g., the distributive transform in #126852).

Changes:

Add a canonicalization in fgOptimizeRelationalComparison that, when matching (A & pow2) ==/!= pow2, reverses the relop and bashes the RHS constant to zero, with VN updated accordingly.

…it-test-to-zero-on-right

BoyBaykiller · 2026-06-02T19:10:06Z

@EgorBo I fixed the merge conflicts.

…TreeDone

BoyBaykiller · 2026-06-04T15:00:06Z

Are we good to merge :) ? I'd like to continue follow up work @EgorBo

EgorBo · 2026-06-04T16:07:57Z

/ba-g known CSharpMissingShebangInFileBasedProgram failure

BoyBaykiller · 2026-05-24T03:31:11Z

Basically so we always have 0 at the right-side:

'(A & pow2) == pow2' -> '(A & pow2) != 0'
'(A & pow2) != pow2' -> '(A & pow2) == 0'

Tthis currently helps optimizeBools, but I am also adding this with the future in mind.
Here is one example:

staticboolExample(intfoo){return((foo&4)==4)||((foo&8)==8);}

;; ------ BASEG_M000_IG02: ;; offset=0x0000testcl,4je SHORT G_M000_IG05G_M000_IG03: ;; offset=0x0005moveax,1G_M000_IG04: ;; offset=0x000AretG_M000_IG05: ;; offset=0x000Btestcl,8 setne almovzxrax,alG_M000_IG06: ;; offset=0x0014ret;; ------ DIFFG_M33938_IG02: ;; offset=0x0000moveax,ecxandeax,4andecx,8oreax,ecx setne almovzxrax,al

Note: With #126852 this would then get transformed into just test 12.

* '(A & pow2) != pow2' -> '(A & pow2) == 0'

dotnet-policy-service · 2026-05-24T03:32:18Z

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

MichalPetryka · 2026-05-24T22:43:16Z

@MihuBot -nuget

…it-test-to-zero-on-right

Copilot

Pull request overview

Canonicalizes (A & pow2) == pow2 to (A & pow2) != 0 (and the inverse for !=) during morph relational-comparison optimization. This puts the comparison into a uniform form with 0 on the right, which immediately helps optimizeBools and sets the stage for further follow-on optimizations (e.g., the distributive transform in #126852).

Changes:

Add a canonicalization in fgOptimizeRelationalComparison that, when matching (A & pow2) ==/!= pow2, reverses the relop and bashes the RHS constant to zero, with VN updated accordingly.

…it-test-to-zero-on-right

BoyBaykiller · 2026-06-02T19:10:06Z

@EgorBo I fixed the merge conflicts.

…TreeDone

BoyBaykiller · 2026-06-04T15:00:06Z

Are we good to merge :) ? I'd like to continue follow up work @EgorBo

EgorBo · 2026-06-04T16:07:57Z

/ba-g known CSharpMissingShebangInFileBasedProgram failure

BoyBaykiller · 2026-05-24T03:31:11Z

Basically so we always have 0 at the right-side:

'(A & pow2) == pow2' -> '(A & pow2) != 0'
'(A & pow2) != pow2' -> '(A & pow2) == 0'

Tthis currently helps optimizeBools, but I am also adding this with the future in mind.
Here is one example:

staticboolExample(intfoo){return((foo&4)==4)||((foo&8)==8);}

;; ------ BASEG_M000_IG02: ;; offset=0x0000testcl,4je SHORT G_M000_IG05G_M000_IG03: ;; offset=0x0005moveax,1G_M000_IG04: ;; offset=0x000AretG_M000_IG05: ;; offset=0x000Btestcl,8 setne almovzxrax,alG_M000_IG06: ;; offset=0x0014ret;; ------ DIFFG_M33938_IG02: ;; offset=0x0000moveax,ecxandeax,4andecx,8oreax,ecx setne almovzxrax,al

Note: With #126852 this would then get transformed into just test 12.

* '(A & pow2) != pow2' -> '(A & pow2) == 0'

dotnet-policy-service · 2026-05-24T03:32:18Z

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

MichalPetryka · 2026-05-24T22:43:16Z

@MihuBot -nuget

…it-test-to-zero-on-right

Copilot

Pull request overview

Canonicalizes (A & pow2) == pow2 to (A & pow2) != 0 (and the inverse for !=) during morph relational-comparison optimization. This puts the comparison into a uniform form with 0 on the right, which immediately helps optimizeBools and sets the stage for further follow-on optimizations (e.g., the distributive transform in #126852).

Changes:

Add a canonicalization in fgOptimizeRelationalComparison that, when matching (A & pow2) ==/!= pow2, reverses the relop and bashes the RHS constant to zero, with VN updated accordingly.

…it-test-to-zero-on-right

BoyBaykiller · 2026-06-02T19:10:06Z

@EgorBo I fixed the merge conflicts.

…TreeDone

BoyBaykiller · 2026-06-04T15:00:06Z

Are we good to merge :) ? I'd like to continue follow up work @EgorBo

EgorBo · 2026-06-04T16:07:57Z

/ba-g known CSharpMissingShebangInFileBasedProgram failure

BoyBaykiller · 2026-05-24T03:31:11Z

Basically so we always have 0 at the right-side:

'(A & pow2) == pow2' -> '(A & pow2) != 0'
'(A & pow2) != pow2' -> '(A & pow2) == 0'

Tthis currently helps optimizeBools, but I am also adding this with the future in mind.
Here is one example:

staticboolExample(intfoo){return((foo&4)==4)||((foo&8)==8);}

;; ------ BASEG_M000_IG02: ;; offset=0x0000testcl,4je SHORT G_M000_IG05G_M000_IG03: ;; offset=0x0005moveax,1G_M000_IG04: ;; offset=0x000AretG_M000_IG05: ;; offset=0x000Btestcl,8 setne almovzxrax,alG_M000_IG06: ;; offset=0x0014ret;; ------ DIFFG_M33938_IG02: ;; offset=0x0000moveax,ecxandeax,4andecx,8oreax,ecx setne almovzxrax,al

Note: With #126852 this would then get transformed into just test 12.

* '(A & pow2) != pow2' -> '(A & pow2) == 0'

dotnet-policy-service · 2026-05-24T03:32:18Z

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

MichalPetryka · 2026-05-24T22:43:16Z

@MihuBot -nuget

…it-test-to-zero-on-right

Copilot

Pull request overview

Canonicalizes (A & pow2) == pow2 to (A & pow2) != 0 (and the inverse for !=) during morph relational-comparison optimization. This puts the comparison into a uniform form with 0 on the right, which immediately helps optimizeBools and sets the stage for further follow-on optimizations (e.g., the distributive transform in #126852).

Changes:

Add a canonicalization in fgOptimizeRelationalComparison that, when matching (A & pow2) ==/!= pow2, reverses the relop and bashes the RHS constant to zero, with VN updated accordingly.

…it-test-to-zero-on-right

BoyBaykiller · 2026-06-02T19:10:06Z

@EgorBo I fixed the merge conflicts.

…TreeDone

BoyBaykiller · 2026-06-04T15:00:06Z

Are we good to merge :) ? I'd like to continue follow up work @EgorBo

EgorBo · 2026-06-04T16:07:57Z

/ba-g known CSharpMissingShebangInFileBasedProgram failure

BoyBaykiller · 2026-05-24T03:31:11Z

Basically so we always have 0 at the right-side:

'(A & pow2) == pow2' -> '(A & pow2) != 0'
'(A & pow2) != pow2' -> '(A & pow2) == 0'

Tthis currently helps optimizeBools, but I am also adding this with the future in mind.
Here is one example:

staticboolExample(intfoo){return((foo&4)==4)||((foo&8)==8);}

;; ------ BASEG_M000_IG02: ;; offset=0x0000testcl,4je SHORT G_M000_IG05G_M000_IG03: ;; offset=0x0005moveax,1G_M000_IG04: ;; offset=0x000AretG_M000_IG05: ;; offset=0x000Btestcl,8 setne almovzxrax,alG_M000_IG06: ;; offset=0x0014ret;; ------ DIFFG_M33938_IG02: ;; offset=0x0000moveax,ecxandeax,4andecx,8oreax,ecx setne almovzxrax,al

Note: With #126852 this would then get transformed into just test 12.

* '(A & pow2) != pow2' -> '(A & pow2) == 0'

dotnet-policy-service · 2026-05-24T03:32:18Z

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

MichalPetryka · 2026-05-24T22:43:16Z

@MihuBot -nuget

…it-test-to-zero-on-right

Copilot

Pull request overview

Canonicalizes (A & pow2) == pow2 to (A & pow2) != 0 (and the inverse for !=) during morph relational-comparison optimization. This puts the comparison into a uniform form with 0 on the right, which immediately helps optimizeBools and sets the stage for further follow-on optimizations (e.g., the distributive transform in #126852).

Changes:

Add a canonicalization in fgOptimizeRelationalComparison that, when matching (A & pow2) ==/!= pow2, reverses the relop and bashes the RHS constant to zero, with VN updated accordingly.

…it-test-to-zero-on-right

BoyBaykiller · 2026-06-02T19:10:06Z

@EgorBo I fixed the merge conflicts.

…TreeDone

BoyBaykiller · 2026-06-04T15:00:06Z

Are we good to merge :) ? I'd like to continue follow up work @EgorBo

EgorBo · 2026-06-04T16:07:57Z

/ba-g known CSharpMissingShebangInFileBasedProgram failure

* '(A & pow2) == pow2' -> '(A & pow2) != 0'
9e47a3d
* '(A & pow2) != pow2' -> '(A & pow2) == 0'

github-actionsBot added the area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI label May 24, 2026

dotnet-policy-serviceBot added the community-contribution Indicates that the PR has been added by a community member label May 24, 2026

BoyBaykiller mentioned this pull request May 24, 2026
JIT: Refactor morph cmp optimizations #127548
Closed

BoyBaykiller marked this pull request as draft May 24, 2026 05:38

BoyBaykiller marked this pull request as ready for review May 24, 2026 05:44

build-analysisBot mentioned this pull request May 24, 2026
The Operation will be canceled. The next steps may not contain expected logs. dotnet/dnceng#3008
Open
3 tasks

MichalPetryka reviewed May 24, 2026
View reviewed changes

Comment threadsrc/coreclr/jit/morph.cpp Outdated

MihuBot mentioned this pull request May 24, 2026
[JitDiff X64] [BoyBaykiller] JIT: Canonicalize '(A & pow2) == pow2' -> '(A & ... MihuBot/runtime-utils#1934
Open

BoyBaykiller added 2 commits May 25, 2026 01:02

Merge branch 'main' ofhttps://github.com/dotnet/runtimeinto canon-b…
cffc026
…it-test-to-zero-on-right

* feedback
bc71b6f

EgorBo requested a review from Copilot June 1, 2026 23:39

Copilotstarted reviewing on behalf of EgorBo June 1, 2026 23:40 View session

CopilotAI reviewed Jun 1, 2026
View reviewed changes

Merge branch 'main' ofhttps://github.com/dotnet/runtimeinto canon-b…
5ed8739
…it-test-to-zero-on-right

EgorBo reviewed Jun 2, 2026
View reviewed changes

Comment threadsrc/coreclr/jit/morph.cpp Outdated

BoyBaykiller added 3 commits June 2, 2026 22:21

* feedback
71ad99d

* add missing fgMorphTreeDone for gtNewIconNodeWithVN
6b7f243

* switch to gtNewZeroConNode + fgUpdateConstTreeValueNumber + fgMorph…
70b794b
…TreeDone

BoyBaykiller closed this Jun 3, 2026

BoyBaykiller reopened this Jun 3, 2026

BoyBaykiller requested a review from EgorBo June 3, 2026 23:47

EgorBo approved these changes Jun 4, 2026
View reviewed changes

EgorBo reviewed Jun 4, 2026
View reviewed changes

Comment threadsrc/coreclr/jit/morph.cpp

EgorBo enabled auto-merge (squash) June 4, 2026 16:08

EgorBo merged commit ebb13c5 into dotnet:mainJun 4, 2026
148 of 151 checks passed

dotnet-maestroBot mentioned this pull request Jun 5, 2026
[main] Source code updates from dotnet/runtime dotnet/dotnet#7066
Merged

dotnet-milestone-botBot added this to the 11.0-preview6 milestone Jun 5, 2026

github-actionsBot locked and limited conversation to collaborators Jul 6, 2026