[CostModel] Mark ssa_copy as free #75294
Merged
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
llvmbot
added
the
llvm:analysis
|
@llvm/pr-subscribers-llvm-analysis Author: David Green (davemgreen) ChangesThese are intrinsics are only used ephemerally and be should be given a zero cost. Full diff: https://github.com/llvm/llvm-project/pull/75294.diff 2 Files Affected:
diff --git a/llvm/include/llvm/CodeGen/BasicTTIImpl.h b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
index e05ce2890a08c8..33db350a12102a 100644
--- a/llvm/include/llvm/CodeGen/BasicTTIImpl.h
+++ b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
@@ -1510,6 +1510,8 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {
Intrinsic::ID IID = ICA.getID();
if (Function::isTargetIntrinsic(IID))
return TargetTransformInfo::TCC_Basic;
+ if (IID == Intrinsic::ssa_copy)
+ return TargetTransformInfo::TCC_Free;
if (ICA.isTypeBasedOnly())
return getTypeBasedIntrinsicInstrCost(ICA, CostKind);
diff --git a/llvm/test/Analysis/CostModel/ARM/intrinsic-cost-kinds.ll b/llvm/test/Analysis/CostModel/ARM/intrinsic-cost-kinds.ll
index 9a9193b36de029..d58f3348e41fed 100644
--- a/llvm/test/Analysis/CostModel/ARM/intrinsic-cost-kinds.ll
+++ b/llvm/test/Analysis/CostModel/ARM/intrinsic-cost-kinds.ll
@@ -341,27 +341,27 @@ define void @memcpy(ptr %a, ptr %b, i32 %c) {
define void @ssa_copy() {
; CHECK: %{{.*}} = llvm.intr.ssa.copy %{{.*}} : f32
; THRU-LABEL: 'ssa_copy'
-; THRU-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %i = call i32 @llvm.ssa.copy.i32(i32 undef)
-; THRU-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %f = call float @llvm.ssa.copy.f32(float undef)
-; THRU-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %p = call ptr @llvm.ssa.copy.p0(ptr undef)
+; THRU-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %i = call i32 @llvm.ssa.copy.i32(i32 undef)
+; THRU-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %f = call float @llvm.ssa.copy.f32(float undef)
+; THRU-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %p = call ptr @llvm.ssa.copy.p0(ptr undef)
; THRU-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
;
; LATE-LABEL: 'ssa_copy'
-; LATE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %i = call i32 @llvm.ssa.copy.i32(i32 undef)
-; LATE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %f = call float @llvm.ssa.copy.f32(float undef)
-; LATE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %p = call ptr @llvm.ssa.copy.p0(ptr undef)
+; LATE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %i = call i32 @llvm.ssa.copy.i32(i32 undef)
+; LATE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %f = call float @llvm.ssa.copy.f32(float undef)
+; LATE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %p = call ptr @llvm.ssa.copy.p0(ptr undef)
; LATE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
;
; SIZE-LABEL: 'ssa_copy'
-; SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %i = call i32 @llvm.ssa.copy.i32(i32 undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %f = call float @llvm.ssa.copy.f32(float undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %p = call ptr @llvm.ssa.copy.p0(ptr undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %i = call i32 @llvm.ssa.copy.i32(i32 undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %f = call float @llvm.ssa.copy.f32(float undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %p = call ptr @llvm.ssa.copy.p0(ptr undef)
; SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
;
; SIZE_LATE-LABEL: 'ssa_copy'
-; SIZE_LATE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %i = call i32 @llvm.ssa.copy.i32(i32 undef)
-; SIZE_LATE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %f = call float @llvm.ssa.copy.f32(float undef)
-; SIZE_LATE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %p = call ptr @llvm.ssa.copy.p0(ptr undef)
+; SIZE_LATE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %i = call i32 @llvm.ssa.copy.i32(i32 undef)
+; SIZE_LATE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %f = call float @llvm.ssa.copy.f32(float undef)
+; SIZE_LATE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %p = call ptr @llvm.ssa.copy.p0(ptr undef)
; SIZE_LATE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
;
%i = call i32 @llvm.ssa.copy.i32(i32 undef)
|
|
I am curious if you encountered cases where those intrinsics are still around after the middle-end optimization pipeline. I think all users of those intrinsics should clean them up, but there might be some places that got missed. |
|
No the costing is done during a pass that adds and removes them. This could be handled specially in that pass, but it is more general fix to give a zero cost to everywhere. |
nikic
reviewed
Dec 13, 2023
| @@ -1510,6 +1510,8 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> { | |||
| Intrinsic::ID IID = ICA.getID(); | |||
| if (Function::isTargetIntrinsic(IID)) | |||
| return TargetTransformInfo::TCC_Basic; | |||
| if (IID == Intrinsic::ssa_copy) | |||
| return TargetTransformInfo::TCC_Free; | |||
There was a problem hiding this comment.
This should be in TargetTransformInfoImpl.h.
There was a problem hiding this comment.
Ah I see. That is better, thanks.
These are intrinsics are only used ephemerally and be should be given a zero cost.
davemgreen
force-pushed
the
gh-ssacopyfree
branch
from
99621ee to
fad4c27
Compare
hazzlim
added a commit
that referenced
this pull request
Nov 4, 2024
Look through ssa_copy intrinsic calls when computing codesize bonus for a specialization. Also remove redundant logic to skip computing codesize bonus for ssa_copy intrinsics, now these are considered zero-cost by TTI (in PR #75294).
PhilippRados
pushed a commit
to PhilippRados/llvm-project
that referenced
this pull request
Nov 6, 2024
…14247) Look through ssa_copy intrinsic calls when computing codesize bonus for a specialization. Also remove redundant logic to skip computing codesize bonus for ssa_copy intrinsics, now these are considered zero-cost by TTI (in PR llvm#75294).
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.
These are intrinsics are only used ephemerally and be should be given a zero cost.