[InstCombine] Fold (ct{t,l}z Pow2) -> Log2(Pow2)
#122620
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
|
@llvm/pr-subscribers-llvm-transforms Author: None (goldsteinn) Changes
Full diff: https://github.com/llvm/llvm-project/pull/122620.diff 2 Files Affected:
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index 7454382412369f..94240773f46a80 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -588,6 +588,22 @@ static Instruction *foldCttzCtlz(IntrinsicInst &II, InstCombinerImpl &IC) {
}
}
+ // cttz(Pow2) -> Log2(Pow2)
+ // ctlz(Pow2) -> BitWidth - 1 - Log2(Pow2)
+ if (IsTZ || II.hasOneUse()) {
+ if (auto *R = IC.tryGetLog2(Op0, match(Op1, m_One()))) {
+ if (IsTZ)
+ return IC.replaceInstUsesWith(II, R);
+ BinaryOperator *BO = BinaryOperator::CreateSub(
+ ConstantInt::get(R->getType(),
+ R->getType()->getScalarSizeInBits() - 1),
+ R);
+ BO->setHasNoSignedWrap();
+ BO->setHasNoUnsignedWrap();
+ return BO;
+ }
+ }
+
KnownBits Known = IC.computeKnownBits(Op0, 0, &II);
// Create a mask for bits above (ctlz) or below (cttz) the first known one.
diff --git a/llvm/test/Transforms/InstCombine/cttz.ll b/llvm/test/Transforms/InstCombine/cttz.ll
index cb0bc59ae79958..5717e352c81e13 100644
--- a/llvm/test/Transforms/InstCombine/cttz.ll
+++ b/llvm/test/Transforms/InstCombine/cttz.ll
@@ -297,3 +297,72 @@ define i16 @cttz_assume(i16 %x) {
%cttz = call i16 @llvm.cttz.i16(i16 %x, i1 false)
ret i16 %cttz
}
+
+
+declare void @use.i8(i8)
+define i8 @fold_ctz_log2(i8 %x) {
+; CHECK-LABEL: @fold_ctz_log2(
+; CHECK-NEXT: [[R:%.*]] = call i8 @llvm.umin.i8(i8 [[X:%.*]], i8 5)
+; CHECK-NEXT: ret i8 [[R]]
+;
+ %p2 = shl i8 1, %x
+ %v = call i8 @llvm.umin(i8 %p2, i8 32)
+ %r = call i8 @llvm.cttz(i8 %v, i1 false)
+ ret i8 %r
+}
+
+define i8 @fold_ctz_log2_maybe_z(i8 %x, i8 %y, i1 %c) {
+; CHECK-LABEL: @fold_ctz_log2_maybe_z(
+; CHECK-NEXT: [[V:%.*]] = shl i8 2, [[V_V:%.*]]
+; CHECK-NEXT: [[P2_2:%.*]] = shl i8 4, [[Y:%.*]]
+; CHECK-NEXT: [[V1:%.*]] = select i1 [[C:%.*]], i8 [[V]], i8 [[P2_2]]
+; CHECK-NEXT: [[R:%.*]] = call range(i8 1, 9) i8 @llvm.cttz.i8(i8 [[V1]], i1 false)
+; CHECK-NEXT: ret i8 [[R]]
+;
+ %p2 = shl i8 2, %x
+ %p2_2 = shl i8 4, %y
+ %v = select i1 %c, i8 %p2, i8 %p2_2
+ %r = call i8 @llvm.cttz(i8 %v, i1 false)
+ ret i8 %r
+}
+
+define i8 @fold_ctz_log2_maybe_z_okay(i8 %x, i8 %y, i1 %c) {
+; CHECK-LABEL: @fold_ctz_log2_maybe_z_okay(
+; CHECK-NEXT: [[X:%.*]] = add i8 [[X1:%.*]], 1
+; CHECK-NEXT: [[Y:%.*]] = add i8 [[Y1:%.*]], 2
+; CHECK-NEXT: [[V_V:%.*]] = select i1 [[C:%.*]], i8 [[X]], i8 [[Y]]
+; CHECK-NEXT: ret i8 [[V_V]]
+;
+ %p2 = shl i8 2, %x
+ %p2_2 = shl i8 4, %y
+ %v = select i1 %c, i8 %p2, i8 %p2_2
+ %r = call i8 @llvm.cttz(i8 %v, i1 true)
+ ret i8 %r
+}
+
+define i8 @fold_clz_log2(i8 %x) {
+; CHECK-LABEL: @fold_clz_log2(
+; CHECK-NEXT: [[TMP1:%.*]] = call i8 @llvm.umin.i8(i8 [[X:%.*]], i8 5)
+; CHECK-NEXT: [[R:%.*]] = xor i8 [[TMP1]], 7
+; CHECK-NEXT: ret i8 [[R]]
+;
+ %p2 = shl i8 1, %x
+ %v = call i8 @llvm.umin(i8 %p2, i8 32)
+ %r = call i8 @llvm.ctlz(i8 %v, i1 false)
+ ret i8 %r
+}
+
+define i8 @fold_clz_log2_fail_multi_use(i8 %x) {
+; CHECK-LABEL: @fold_clz_log2_fail_multi_use(
+; CHECK-NEXT: [[P2:%.*]] = shl nuw i8 1, [[X:%.*]]
+; CHECK-NEXT: [[V:%.*]] = call i8 @llvm.umin.i8(i8 [[P2]], i8 32)
+; CHECK-NEXT: [[R:%.*]] = call range(i8 2, 9) i8 @llvm.ctlz.i8(i8 [[V]], i1 true)
+; CHECK-NEXT: call void @use.i8(i8 [[R]])
+; CHECK-NEXT: ret i8 [[R]]
+;
+ %p2 = shl i8 1, %x
+ %v = call i8 @llvm.umin(i8 %p2, i8 32)
+ %r = call i8 @llvm.ctlz(i8 %v, i1 false)
+ call void @use.i8(i8 %r)
+ ret i8 %r
+}
|
goldsteinn
changed the title
(ct{t,l}z Pow2) -> Log2(Pow2)
This was referenced Jan 12, 2025
nikic
reviewed
Jan 12, 2025
| @@ -588,6 +588,22 @@ static Instruction *foldCttzCtlz(IntrinsicInst &II, InstCombinerImpl &IC) { | |||
| } | |||
| } | |||
|
|
|||
| // cttz(Pow2) -> Log2(Pow2) | |||
| // ctlz(Pow2) -> BitWidth - 1 - Log2(Pow2) | |||
| if (IsTZ || II.hasOneUse()) { | |||
There was a problem hiding this comment.
Why does the number of uses of the intrinsic matter?
There was a problem hiding this comment.
Oops. I guess @goldsteinn means Op0->hasOneUse().
There was a problem hiding this comment.
should probably be no multi-use check. We are replacing ctlz with sub which is just worth it.
There was a problem hiding this comment.
Your new condition effectively drops ctlz entirely :)
goldsteinn
force-pushed
the
goldsteinn/cttz-ctlz-of-p2
branch
from
aa09334 to
3599487
Compare
goldsteinn
force-pushed
the
goldsteinn/cttz-ctlz-of-p2
branch
from
3599487 to
72ef809
Compare
nikic
approved these changes
Jan 12, 2025
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/59/builds/10999 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/190/builds/12725 Here is the relevant piece of the build log for the reference |
kazutakahirata
pushed a commit
to kazutakahirata/llvm-project
that referenced
this pull request
Jan 13, 2025
- **[InstCombine] Add tests for folding `(ct{t,l}z Pow2)`; NFC**
- **[InstCombine] Fold `(ct{t,l}z Pow2)` -> `Log2(Pow2)`**
Do so we can find `Log2(Pow2)` for "free" with `takeLog2`
https://alive2.llvm.org/ce/z/CL77fo
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.
(ct{t,l}z Pow2); NFC(ct{t,l}z Pow2)->Log2(Pow2)Do so we can find
Log2(Pow2)for "free" withtakeLog2https://alive2.llvm.org/ce/z/CL77fo