This is an archive of the discontinued LLVM Phabricator instance.

Differential D155422

[MemCpyOpt] precommit test for D155406 (NFC)
ClosedPublic

Authored by khei4 on Jul 16 2023, 11:21 PM.

Details

Reviewers
nikic
pcwalton
Commits
rGe0911b98d19e: [MemCpyOpt] precommit test for D155406 (NFC)
Summary

pre-commit test for https://reviews.llvm.org/D155406

Diff Detail

Event Timeline

khei4 created this revision.Jul 16 2023, 11:21 PM
Herald added a project: Restricted Project. · View Herald TranscriptJul 16 2023, 11:21 PM
khei4 requested review of this revision.Jul 16 2023, 11:21 PM
Herald added a project: Restricted Project. · View Herald TranscriptJul 16 2023, 11:21 PM
Herald added a subscriber: llvm-commits. · View Herald Transcript
khei4 mentioned this in D155406: [MemCpyOpt] implement multi BB stack-move optimization.Jul 16 2023, 11:21 PM
khei4 added reviewers: nikic, pcwalton.
Herald added a subscriber: StephenFan. · View Herald TranscriptJul 16 2023, 11:22 PM
khei4 updated this revision to Diff 540872.Jul 16 2023, 11:31 PM

fix wrong diffs for TODO comments

Harbormaster completed remote builds in B245736: Diff 540872.Jul 17 2023, 12:09 AM
khei4 retitled this revision from [MemCpyOpt] precommit test for D155406 (NFC) to (WIP)[MemCpyOpt] precommit test for D155406 (NFC).Jul 17 2023, 1:54 AM
khei4 updated this revision to Diff 542301.Jul 19 2023, 9:40 PM

rebase

Harbormaster completed remote builds in B246760: Diff 542301.Jul 19 2023, 9:50 PM
khei4 updated this revision to Diff 542791.Jul 21 2023, 12:57 AM
khei4 retitled this revision from (WIP)[MemCpyOpt] precommit test for D155406 (NFC) to [MemCpyOpt] precommit test for D155406 (NFC).

add

  • loop positive/negative cases
  • pcwalton's test cases for multi-bb https://reviews.llvm.org/D140089
  • pcwalton's test cases which require Def like concepts to reduce
khei4 edited the summary of this revision. (Show Details)Jul 21 2023, 12:58 AM
Harbormaster completed remote builds in B247126: Diff 542791.Jul 21 2023, 3:04 AM
khei4 updated this revision to Diff 551405.Aug 18 2023, 12:19 AM

add

  • a test for BB, Instruction mixed post-dominator finding
Harbormaster completed remote builds in B253414: Diff 551405.Aug 18 2023, 1:39 AM
khei4 updated this revision to Diff 551903.Aug 20 2023, 10:42 PM

add phi node test

Harbormaster completed remote builds in B253765: Diff 551903.Aug 20 2023, 11:35 PM
nikic added a child revision: D155406: [MemCpyOpt] implement multi BB stack-move optimization.Aug 21 2023, 7:20 AM
nikic added a comment.Aug 21 2023, 7:24 AM

Is there a negative test where %dest is written before the memcpy (with some control flow)? Maybe I missed it, but it seems like most of the tests here work on %src.

khei4 added a comment.Aug 21 2023, 3:01 PM

Is there a negative test where %dest is written before the memcpy (with some control flow)? Maybe I missed it, but it seems like most of the tests here work on %src.

Thank you for the good catch! I should have added the dest mod before copying for the branch case!

khei4 updated this revision to Diff 552149.Aug 21 2023, 3:01 PM

add a case for branched dest mod before copy

Harbormaster completed remote builds in B253936: Diff 552149.Aug 21 2023, 4:31 PM
nikic accepted this revision.Aug 22 2023, 3:50 AM

LGTM

This revision is now accepted and ready to land.Aug 22 2023, 3:50 AM
Closed by commit rGe0911b98d19e: [MemCpyOpt] precommit test for D155406 (NFC) (authored by khei4). · Explain WhyAug 24 2023, 6:19 AM
This revision was automatically updated to reflect the committed changes.
khei4 added a commit: rGe0911b98d19e: [MemCpyOpt] precommit test for D155406 (NFC).

Revision Contents

PathSize
llvm/
test/
Transforms/
MemCpyOpt/
86 lines

Diff 542301

llvm/test/Transforms/MemCpyOpt/stack-move.ll

Show First 20 Lines • Show All 482 Lines • ▼ Show 20 Linesbb1:
br label %bb2 br label %bb2
bb2: bb2:
call void @llvm.lifetime.end.p0(i64 12, ptr nocapture %src) call void @llvm.lifetime.end.p0(i64 12, ptr nocapture %src)
call void @llvm.lifetime.end.p0(i64 12, ptr nocapture %dest) call void @llvm.lifetime.end.p0(i64 12, ptr nocapture %dest)
ret void ret void
} }
; TODO: merge allocas for multi basicblocks, s.t. all copy-dominated
define ptr @multi_bb_loop(i32 %n) { define ptr @multi_bb_loop(i32 %n) {
; CHECK-LABEL: define ptr @multi_bb_loop ; CHECK-LABEL: define ptr @multi_bb_loop
; CHECK-SAME: (i32 [[N:%.*]]) { ; CHECK-SAME: (i32 [[N:%.*]]) {
; CHECK-NEXT: entry: ; CHECK-NEXT: entry:
; CHECK-NEXT: [[NLT1:%.*]] = icmp slt i32 [[N]], 1 ; CHECK-NEXT: [[NLT1:%.*]] = icmp slt i32 [[N]], 1
; CHECK-NEXT: [[SRC:%.*]] = alloca [[STRUCT_FOO:%.*]], align 8 ; CHECK-NEXT: [[SRC:%.*]] = alloca [[STRUCT_FOO:%.*]], align 8
; CHECK-NEXT: [[DEST:%.*]] = alloca [[STRUCT_FOO]], align 8 ; CHECK-NEXT: [[DEST:%.*]] = alloca [[STRUCT_FOO]], align 8
; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 12, ptr nocapture [[SRC]]) ; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 12, ptr nocapture [[SRC]])
Show All 26 Linesloop_body:
store i32 %new_i, ptr %dest store i32 %new_i, ptr %dest
%igtn = icmp sgt i32 %new_i, %n %igtn = icmp sgt i32 %new_i, %n
br i1 %igtn, label %loop_exit, label %loop_body br i1 %igtn, label %loop_exit, label %loop_body
loop_exit: loop_exit:
ret ptr %dest ret ptr %dest
} }
; TODO: merge allocas for multi basicblocks, s.t. some modref which is unreachable from copy exists.
define void @multi_bb_unreachable_modref(i1 %b0) {
; CHECK-LABEL: define void @multi_bb_unreachable_modref
; CHECK-SAME: (i1 [[B0:%.*]]) {
; CHECK-NEXT: [[SRC:%.*]] = alloca [[STRUCT_FOO:%.*]], align 4
; CHECK-NEXT: [[DEST:%.*]] = alloca [[STRUCT_FOO]], align 4
; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 12, ptr nocapture [[SRC]])
; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 12, ptr nocapture [[DEST]])
; CHECK-NEXT: store [[STRUCT_FOO]] { i32 10, i32 20, i32 30 }, ptr [[SRC]], align 4
; CHECK-NEXT: [[TMP1:%.*]] = call i32 @use_nocapture(ptr nocapture noundef [[SRC]])
; CHECK-NEXT: br i1 [[B0]], label [[BB0:%.*]], label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: [[TMP2:%.*]] = call i32 @use_nocapture(ptr nocapture noundef [[SRC]])
; CHECK-NEXT: ret void
; CHECK: bb0:
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[DEST]], ptr align 4 [[SRC]], i64 12, i1 false)
; CHECK-NEXT: call void @llvm.lifetime.end.p0(i64 12, ptr nocapture [[SRC]])
; CHECK-NEXT: call void @llvm.lifetime.end.p0(i64 12, ptr nocapture [[DEST]])
; CHECK-NEXT: ret void
;
%src = alloca %struct.Foo, align 4
%dest = alloca %struct.Foo, align 4
call void @llvm.lifetime.start.p0(i64 12, ptr nocapture %src)
call void @llvm.lifetime.start.p0(i64 12, ptr nocapture %dest)
store %struct.Foo { i32 10, i32 20, i32 30 }, ptr %src
%1 = call i32 @use_nocapture(ptr noundef nocapture %src)
br i1 %b0, label %bb0, label %exit
exit:
%2 = call i32 @use_nocapture(ptr noundef nocapture %src)
ret void
bb0:
call void @llvm.memcpy.p0.p0.i64(ptr align 4 %dest, ptr align 4 %src, i64 12, i1 false)
call void @llvm.lifetime.end.p0(i64 12, ptr nocapture %src)
call void @llvm.lifetime.end.p0(i64 12, ptr nocapture %dest)
ret void
}
; TODO: merge allocas for multi basicblocks, s.t. memcpy doesn't dominate the uses.
define void @multi_bb_non_dominated(i1 %b0, i1 %b1) {
; CHECK-LABEL: define void @multi_bb_non_dominated
; CHECK-SAME: (i1 [[B0:%.*]], i1 [[B1:%.*]]) {
; CHECK-NEXT: [[SRC:%.*]] = alloca [[STRUCT_FOO:%.*]], align 4
; CHECK-NEXT: [[DEST:%.*]] = alloca [[STRUCT_FOO]], align 4
; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 12, ptr nocapture [[SRC]])
; CHECK-NEXT: call void @llvm.lifetime.start.p0(i64 12, ptr nocapture [[DEST]])
; CHECK-NEXT: store [[STRUCT_FOO]] { i32 10, i32 20, i32 30 }, ptr [[SRC]], align 4
; CHECK-NEXT: [[TMP1:%.*]] = call i32 @use_nocapture(ptr nocapture noundef [[SRC]])
; CHECK-NEXT: br i1 [[B0]], label [[BB0:%.*]], label [[BB1:%.*]]
; CHECK: bb0:
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[DEST]], ptr align 4 [[SRC]], i64 12, i1 false)
; CHECK-NEXT: br label [[BB2:%.*]]
; CHECK: bb1:
; CHECK-NEXT: [[TMP2:%.*]] = call i32 @use_nocapture(ptr nocapture noundef [[SRC]])
; CHECK-NEXT: br label [[BB2]]
; CHECK: bb2:
; CHECK-NEXT: call void @llvm.lifetime.end.p0(i64 12, ptr nocapture [[SRC]])
; CHECK-NEXT: call void @llvm.lifetime.end.p0(i64 12, ptr nocapture [[DEST]])
; CHECK-NEXT: ret void
;
%src = alloca %struct.Foo, align 4
%dest = alloca %struct.Foo, align 4
call void @llvm.lifetime.start.p0(i64 12, ptr nocapture %src)
call void @llvm.lifetime.start.p0(i64 12, ptr nocapture %dest)
store %struct.Foo { i32 10, i32 20, i32 30 }, ptr %src
%1 = call i32 @use_nocapture(ptr noundef nocapture %src)
br i1 %b0, label %bb0, label %bb1
bb0:
call void @llvm.memcpy.p0.p0.i64(ptr align 4 %dest, ptr align 4 %src, i64 12, i1 false)
br label %bb2
bb1:
%2 = call i32 @use_nocapture(ptr noundef nocapture %src)
; %3 = call i32 @use_writeonly(ptr noundef nocapture %dest)
br label %bb2
bb2:
; %4 = call i32 @use_nocapture(ptr noundef nocapture %src)
call void @llvm.lifetime.end.p0(i64 12, ptr nocapture %src)
call void @llvm.lifetime.end.p0(i64 12, ptr nocapture %dest)
ret void
}
; Optimization failures follow: ; Optimization failures follow:
; Tests that dynamically-sized allocas are never merged. ; Tests that dynamically-sized allocas are never merged.
define void @dynamically_sized_alloca(i64 %i) { define void @dynamically_sized_alloca(i64 %i) {
; CHECK-LABEL: define void @dynamically_sized_alloca ; CHECK-LABEL: define void @dynamically_sized_alloca
; CHECK-SAME: (i64 [[I:%.*]]) { ; CHECK-SAME: (i64 [[I:%.*]]) {
; CHECK-NEXT: [[SRC:%.*]] = alloca i8, i64 [[I]], align 4 ; CHECK-NEXT: [[SRC:%.*]] = alloca i8, i64 [[I]], align 4
; CHECK-NEXT: [[DEST:%.*]] = alloca i8, i64 [[I]], align 4 ; CHECK-NEXT: [[DEST:%.*]] = alloca i8, i64 [[I]], align 4
▲ Show 20 Lines • Show All 532 Lines • Show Last 20 Lines