[InstCombine] Add CTLZ -> CTTZ simplification #164733

MihailMihov · 2025-10-22T23:55:54Z

This PR adds the simplification ctlz(~x & (x - 1)) -> bitwidth - cttz(x, false) (Alive2).

For the implementation I initially referenced D76568, which added ctpop(~x & (x - 1)) -> cttz(x, false) and ctpop(x | -x) -> bitwidth - cttz(x, false). I also saw that later fdcb271 made both of the previous only apply when the and/or is one use and later a1fe6be made that only apply to the case that emits more than one instruction. From this and reading the contributor guide I concluded that the current simplification should also check for one use. While writing the tests however I noticed that checking one use on Op0 (%and) only ignored the case where %and is used again, but I don't see why %not and %dec shouldn't be checked too, so I've checked for the one use of those too. Is this incorrect?

For the tests I couldn't find a good place to add them, so I made a new llvm/test/Transforms/InstCombine/ctlz-cttz.ll, but if they have a good place already I'll move them there. Also, while attempting to follow the contributor guide, I added a few different tests, but I'm not sure if they're too little/too much and I don't know if the sub in the simplificaton requires any flag tests?

issue #164436

github-actions · 2025-10-22T23:56:20Z

Thank you for submitting a Pull Request (PR) to the LLVM Project!

This PR will be automatically labeled and the relevant teams will be notified.

If you wish to, you can add reviewers by using the "Reviewers" section on this page.

If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using @ followed by their GitHub username.

If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers.

If you have further questions, they may be answered by the LLVM GitHub User Guide.

You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums.

llvmbot · 2025-10-22T23:57:03Z

@llvm/pr-subscribers-llvm-transforms

Author: Mihail Mihov (MihailMihov)

Changes

This PR adds the simplification ctlz(~x & (x - 1)) -> bitwidth - cttz(x, false) (Alive2).

For the implementation I initially referenced D76568, which added ctpop(~x & (x - 1)) -> cttz(x, false) and ctpop(x | -x) -> bitwidth - cttz(x, false). I also saw that later fdcb271 made both of the previous only apply when the and/or is one use and later a1fe6be made that only apply to the case that emits more than one instruction. From this and reading the contributor guide I concluded that the current simplification should also check for one use. While writing the tests however I noticed that checking one use on Op0 (%and) only ignored the case where %and is used again, but I don't see why %not and %dec shouldn't be checked too, so I've checked for the one use of those too. Is this incorrect?

For the tests I couldn't find a good place to add them, so I made a new llvm/test/Transforms/InstCombine/ctlz-cttz.ll, but if they have a good place already I'll move them there. Also, while attempting to follow the contributor guide, I added a few different tests, but I'm not sure if they're too little/too much and I don't know if the sub in the simplificaton requires any flag tests?

issue #164436

Full diff: https://github.com/llvm/llvm-project/pull/164733.diff

2 Files Affected:

(modified) llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp (+12)
(added) llvm/test/Transforms/InstCombine/ctlz-cttz.ll (+101)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index e1e24a99d0474..2855263ab09c5 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -585,6 +585,18 @@ static Instruction *foldCttzCtlz(IntrinsicInst &II, InstCombinerImpl &IC) {
           IC.Builder.CreateBinaryIntrinsic(Intrinsic::ctlz, C, Op1);
       return BinaryOperator::CreateSub(ConstCtlz, X);
     }
+
+    // ctlz(~x & (x - 1)) -> bitwidth - cttz(x, false)
+    if (Op0->hasOneUse() &&
+        match(Op0, m_c_And(m_OneUse(m_Not(m_Value(X))),
+                           m_OneUse(m_Add(m_Deferred(X), m_AllOnes()))))) {
+      Type *Ty = II.getType();
+      unsigned BitWidth = Ty->getScalarSizeInBits();
+      auto *Cttz = IC.Builder.CreateIntrinsic(Intrinsic::cttz, Ty,
+                                              {X, IC.Builder.getFalse()});
+      auto *Bw = ConstantInt::get(Ty, APInt(BitWidth, BitWidth));
+      return IC.replaceInstUsesWith(II, IC.Builder.CreateSub(Bw, Cttz));
+    }
   }
 
   // cttz(Pow2) -> Log2(Pow2)
diff --git a/llvm/test/Transforms/InstCombine/ctlz-cttz.ll b/llvm/test/Transforms/InstCombine/ctlz-cttz.ll
new file mode 100644
index 0000000000000..2a97c1efd9f16
--- /dev/null
+++ b/llvm/test/Transforms/InstCombine/ctlz-cttz.ll
@@ -0,0 +1,101 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
+; RUN: opt < %s -S -passes=instcombine | FileCheck %s
+
+; ctpop(~i & (i - 1)) -> bitwidth - cttz(i, false)
+define i8 @ctlz_to_sub_bw_cttz(i8 %a0) {
+; CHECK-LABEL: define i8 @ctlz_to_sub_bw_cttz(
+; CHECK-SAME: i8 [[A0:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = call range(i8 0, 9) i8 @llvm.cttz.i8(i8 [[A0]], i1 false)
+; CHECK-NEXT:    [[CLZ:%.*]] = sub nuw nsw i8 8, [[TMP1]]
+; CHECK-NEXT:    ret i8 [[CLZ]]
+;
+  %dec = add i8 %a0, -1
+  %not = xor i8 %a0, -1
+  %and = and i8 %dec, %not
+  %clz = tail call i8 @llvm.ctlz.i8(i8 %and, i1 false)
+  ret i8 %clz
+}
+
+declare void @use(i8)
+
+define i8 @ctlz_to_sub_bw_cttz_multi_use_dec(i8 %a0) {
+; CHECK-LABEL: define i8 @ctlz_to_sub_bw_cttz_multi_use_dec(
+; CHECK-SAME: i8 [[A0:%.*]]) {
+; CHECK-NEXT:    [[DEC:%.*]] = add i8 [[A0]], -1
+; CHECK-NEXT:    call void @use(i8 [[DEC]])
+; CHECK-NEXT:    [[NOT:%.*]] = xor i8 [[A0]], -1
+; CHECK-NEXT:    [[AND:%.*]] = and i8 [[DEC]], [[NOT]]
+; CHECK-NEXT:    [[CLZ:%.*]] = tail call range(i8 0, 9) i8 @llvm.ctlz.i8(i8 [[AND]], i1 false)
+; CHECK-NEXT:    ret i8 [[CLZ]]
+;
+  %dec = add i8 %a0, -1
+  call void @use(i8 %dec)
+  %not = xor i8 %a0, -1
+  %and = and i8 %dec, %not
+  %clz = tail call i8 @llvm.ctlz.i8(i8 %and, i1 false)
+  ret i8 %clz
+}
+
+define i8 @ctlz_to_sub_bw_cttz_multi_use_not(i8 %a0) {
+; CHECK-LABEL: define i8 @ctlz_to_sub_bw_cttz_multi_use_not(
+; CHECK-SAME: i8 [[A0:%.*]]) {
+; CHECK-NEXT:    [[DEC:%.*]] = add i8 [[A0]], -1
+; CHECK-NEXT:    [[NOT:%.*]] = xor i8 [[A0]], -1
+; CHECK-NEXT:    call void @use(i8 [[NOT]])
+; CHECK-NEXT:    [[AND:%.*]] = and i8 [[DEC]], [[NOT]]
+; CHECK-NEXT:    [[CLZ:%.*]] = tail call range(i8 0, 9) i8 @llvm.ctlz.i8(i8 [[AND]], i1 false)
+; CHECK-NEXT:    ret i8 [[CLZ]]
+;
+  %dec = add i8 %a0, -1
+  %not = xor i8 %a0, -1
+  call void @use(i8 %not)
+  %and = and i8 %dec, %not
+  %clz = tail call i8 @llvm.ctlz.i8(i8 %and, i1 false)
+  ret i8 %clz
+}
+
+define i8 @ctlz_to_sub_bw_cttz_multi_use_and(i8 %a0) {
+; CHECK-LABEL: define i8 @ctlz_to_sub_bw_cttz_multi_use_and(
+; CHECK-SAME: i8 [[A0:%.*]]) {
+; CHECK-NEXT:    [[DEC:%.*]] = add i8 [[A0]], -1
+; CHECK-NEXT:    [[NOT:%.*]] = xor i8 [[A0]], -1
+; CHECK-NEXT:    [[AND:%.*]] = and i8 [[DEC]], [[NOT]]
+; CHECK-NEXT:    call void @use(i8 [[AND]])
+; CHECK-NEXT:    [[CLZ:%.*]] = tail call range(i8 0, 9) i8 @llvm.ctlz.i8(i8 [[AND]], i1 false)
+; CHECK-NEXT:    ret i8 [[CLZ]]
+;
+  %dec = add i8 %a0, -1
+  %not = xor i8 %a0, -1
+  %and = and i8 %dec, %not
+  call void @use(i8 %and)
+  %clz = tail call i8 @llvm.ctlz.i8(i8 %and, i1 false)
+  ret i8 %clz
+}
+
+define i8 @ctlz_to_sub_bw_cttz_commute_and(i8 %a0) {
+; CHECK-LABEL: define i8 @ctlz_to_sub_bw_cttz_commute_and(
+; CHECK-SAME: i8 [[A0:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = call range(i8 0, 9) i8 @llvm.cttz.i8(i8 [[A0]], i1 false)
+; CHECK-NEXT:    [[CLZ:%.*]] = sub nuw nsw i8 8, [[TMP1]]
+; CHECK-NEXT:    ret i8 [[CLZ]]
+;
+  %dec = add i8 %a0, -1
+  %not = xor i8 %a0, -1
+  %and = and i8 %not, %dec
+  %clz = tail call i8 @llvm.ctlz.i8(i8 %and, i1 false)
+  ret i8 %clz
+}
+
+define <2 x i8> @ctlz_to_sub_bw_cttz_vec_splat(<2 x i8> %a0) {
+; CHECK-LABEL: define <2 x i8> @ctlz_to_sub_bw_cttz_vec_splat(
+; CHECK-SAME: <2 x i8> [[A0:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = call range(i8 0, 9) <2 x i8> @llvm.cttz.v2i8(<2 x i8> [[A0]], i1 false)
+; CHECK-NEXT:    [[CLZ:%.*]] = sub nuw nsw <2 x i8> splat (i8 8), [[TMP1]]
+; CHECK-NEXT:    ret <2 x i8> [[CLZ]]
+;
+  %dec = add <2 x i8> %a0, <i8 -1, i8 -1>
+  %not = xor <2 x i8> %a0, <i8 -1, i8 -1>
+  %and = and <2 x i8> %dec, %not
+  %clz = tail call <2 x i8>@llvm.ctlz.v2i8(<2 x i8> %and, i1 false)
+  ret <2 x i8> %clz
+}

nikic · 2025-10-23T08:34:37Z

@zyw-bot mfuzz

nikic

This looks good to me.

llvm/test/Transforms/InstCombine/ctlz-cttz.ll

nikic

LGTM

MihailMihov · 2025-10-23T20:24:49Z

@nikic thanks for the review. I don't have commit access, so if I understood things correctly, I should ask you to squash and merge this on my behalf.

dtcxzyw

LGTM. Thanks.

github-actions · 2025-10-24T16:40:34Z

@MihailMihov Congratulations on having your first Pull Request (PR) merged into the LLVM Project!

Your changes will be combined with recent changes from other authors, then tested by our build bots. If there is a problem with a build, you may receive a report in an email or a comment on this PR.

Please check whether problems have been caused by your change specifically, as the builds can include changes from many authors. It is not uncommon for your change to be included in a build that fails due to someone else's changes, or infrastructure issues.

How to do this, and the rest of the post-merge process, is covered in detail here.

If your change does cause a problem, it may be reverted, or you can revert it yourself. This is a normal part of LLVM development. You can fix your changes and open a new PR to merge them again.

If you don't get any reports, no action is required from you. Your changes are working as expected, well done!

This PR adds the simplification `ctlz(~x & (x - 1)) -> bitwidth - cttz(x, false)` ([Alive2](https://alive2.llvm.org/ce/z/vVDRCu)). Closes issue llvm#164436

MihailMihov requested a review from nikic as a code owner October 22, 2025 23:55

llvmbot added llvm:instcombine Covers the InstCombine, InstSimplify and AggressiveInstCombine passes llvm:transforms labels Oct 22, 2025

nikic mentioned this pull request Oct 23, 2025

Task submission dtcxzyw/llvm-opt-benchmark#1312

Open

This was referenced Oct 23, 2025

pre-commit: PR164733 dtcxzyw/llvm-opt-benchmark#2970

Closed

Fuzz PR164733 dtcxzyw/llvm-mutation-based-fuzz-service#114

Closed

nikic reviewed Oct 23, 2025

View reviewed changes

llvm/test/Transforms/InstCombine/ctlz-cttz.ll Show resolved Hide resolved

llvm/test/Transforms/InstCombine/ctlz-cttz.ll Show resolved Hide resolved

llvm/test/Transforms/InstCombine/ctlz-cttz.ll Show resolved Hide resolved

MihailMihov added 2 commits October 23, 2025 16:47

[InstCombine] Add failing test for CTLZ -> CTTZ simplification

f422303

[InstCombine] Add CTLZ -> CTTZ simplification

171ee2b

MihailMihov force-pushed the instcombine-ctlz-cttz-simplification branch from 2d8559d to 171ee2b Compare October 23, 2025 15:11

MihailMihov requested a review from nikic October 23, 2025 15:18

nikic approved these changes Oct 23, 2025

View reviewed changes

Merge branch 'main' into instcombine-ctlz-cttz-simplification

fee6a0e

dtcxzyw linked an issue Oct 24, 2025 that may be closed by this pull request

[InstCombine] Failure to recognise CTTZ pattern using CTLZ #164436

Closed

dtcxzyw approved these changes Oct 24, 2025

View reviewed changes

dtcxzyw merged commit 6034ab3 into llvm:main Oct 24, 2025
10 checks passed

MihailMihov deleted the instcombine-ctlz-cttz-simplification branch October 24, 2025 17:03

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[InstCombine] Add CTLZ -> CTTZ simplification #164733

[InstCombine] Add CTLZ -> CTTZ simplification #164733

Uh oh!

MihailMihov commented Oct 22, 2025

Uh oh!

github-actions bot commented Oct 22, 2025

Uh oh!

llvmbot commented Oct 22, 2025

Uh oh!

nikic commented Oct 23, 2025

Uh oh!

nikic left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

nikic left a comment

Uh oh!

MihailMihov commented Oct 23, 2025

Uh oh!

dtcxzyw left a comment

Uh oh!

Uh oh!

github-actions bot commented Oct 24, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

[InstCombine] Add CTLZ -> CTTZ simplification #164733

[InstCombine] Add CTLZ -> CTTZ simplification #164733

Uh oh!

Conversation

MihailMihov commented Oct 22, 2025

Uh oh!

github-actions bot commented Oct 22, 2025

Uh oh!

llvmbot commented Oct 22, 2025

Uh oh!

nikic commented Oct 23, 2025

Uh oh!

nikic left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

nikic left a comment

Choose a reason for hiding this comment

Uh oh!

MihailMihov commented Oct 23, 2025

Uh oh!

dtcxzyw left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

github-actions bot commented Oct 24, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants