This is an archive of the discontinued LLVM Phabricator instance.

Differential D132553

[instcombine] Test for zero initialisation optimisation of integer product
ClosedPublic

Authored by zjaffal on Aug 24 2022, 6:17 AM.

Details

Reviewers
fhahn
Commits
rG23a1c0a7790a: [instcombine] Test for zero initialisation optimisation of integer product
Summary

Following the work on D131672 we do the same optimisations for integer products.
We add tests to check if a loop gets removed if we repeatdly multiply an array elements with an accumulator initalised to zero

Diff Detail

Event Timeline

zjaffal created this revision.Aug 24 2022, 6:17 AM
Herald added a project: Restricted Project. · View Herald TranscriptAug 24 2022, 6:17 AM
zjaffal requested review of this revision.Aug 24 2022, 6:17 AM
Herald added a project: Restricted Project. · View Herald TranscriptAug 24 2022, 6:17 AM
Herald added a subscriber: llvm-commits. · View Herald Transcript
fhahn added a comment.Aug 24 2022, 6:27 AM

Thanks for covering the integer case as well. It might be good to rename llvm/test/Transforms/InstCombine/remove-loop-phi-fastmul.ll to something like llvm/test/Transforms/InstCombine/remove-loop-phi-multiply-by-zero.ll or something, as -fastmul still carries a hint to fast FP math.

Harbormaster completed remote builds in B183091: Diff 455180.Aug 24 2022, 6:57 AM
zjaffal updated this revision to Diff 455210.Aug 24 2022, 7:47 AM

rename file

fhahn accepted this revision.Aug 24 2022, 7:48 AM

LGTM, thanks!

This revision is now accepted and ready to land.Aug 24 2022, 7:48 AM
This revision was landed with ongoing or failed builds.Aug 24 2022, 7:54 AM
Closed by commit rG23a1c0a7790a: [instcombine] Test for zero initialisation optimisation of integer product (authored by zjaffal, committed by fhahn). · Explain Why
This revision was automatically updated to reflect the committed changes.
fhahn added a commit: rG23a1c0a7790a: [instcombine] Test for zero initialisation optimisation of integer product.
Harbormaster completed remote builds in B183107: Diff 455210.Aug 24 2022, 8:49 AM

Revision Contents

PathSize
llvm/
test/
Transforms/
InstCombine/
remove-loop-phi-multiply-by-zero.ll
remove-loop-phi-fastmul.ll
146 lines

Diff 455214

llvm/test/Transforms/InstCombine/remove-loop-phi-fastmul.ll

  • This file was moved to llvm/test/Transforms/InstCombine/remove-loop-phi-multiply-by-zero.ll.

llvm/test/Transforms/InstCombine/remove-loop-phi-multiply-by-zero.ll

  • This file was moved from llvm/test/Transforms/InstCombine/remove-loop-phi-fastmul.ll.
Show First 20 Lines • Show All 267 Lines • ▼ Show 20 Linesfor.body: ; preds = %entry, %for.body
%inc = add i64 %i.02, 1 %inc = add i64 %i.02, 1
%cmp = icmp ult i64 %inc, 1000 %cmp = icmp ult i64 %inc, 1000
br i1 %cmp, label %for.body, label %end br i1 %cmp, label %for.body, label %end
end: ; preds = %for.body end: ; preds = %for.body
%f_prod.0.lcssa = phi double [ %mul, %for.body ] %f_prod.0.lcssa = phi double [ %mul, %for.body ]
ret double %f_prod.0.lcssa ret double %f_prod.0.lcssa
} }
define i32 @test_int_phi_operands(i32* %arr_d) {
; CHECK-LABEL: @test_int_phi_operands(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[I_02:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[F_PROD_01:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[MUL:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[ARR_D:%.*]], i64 [[I_02]]
; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[MUL]] = mul nsw i32 [[F_PROD_01]], [[TMP0]]
; CHECK-NEXT: [[INC]] = add i64 [[I_02]], 1
; CHECK-NEXT: [[CMP:%.*]] = icmp ult i64 [[INC]], 1000
; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[END:%.*]]
; CHECK: end:
; CHECK-NEXT: ret i32 [[MUL]]
;
entry:
br label %for.body
for.body: ; preds = %entry, %for.body
%i.02 = phi i64 [ 0, %entry ], [ %inc, %for.body ]
%f_prod.01 = phi i32 [ 0, %entry ], [ %mul, %for.body ]
%arrayidx = getelementptr inbounds i32, i32* %arr_d, i64 %i.02
%0 = load i32, i32* %arrayidx, align 4
%mul = mul nsw i32 %f_prod.01, %0
%inc = add i64 %i.02, 1
%cmp = icmp ult i64 %inc, 1000
br i1 %cmp, label %for.body, label %end
end: ; preds = %for.body
%f_prod.0.lcssa = phi i32 [ %mul, %for.body ]
ret i32 %f_prod.0.lcssa
}
define i32 @test_int_phi_operands_initalise_to_non_zero(i32* %arr_d) {
; CHECK-LABEL: @test_int_phi_operands_initalise_to_non_zero(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[I_02:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[F_PROD_01:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ [[MUL:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[ARR_D:%.*]], i64 [[I_02]]
; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[MUL]] = mul i32 [[F_PROD_01]], [[TMP0]]
; CHECK-NEXT: [[INC]] = add i64 [[I_02]], 1
; CHECK-NEXT: [[CMP:%.*]] = icmp ult i64 [[INC]], 1000
; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[END:%.*]]
; CHECK: end:
; CHECK-NEXT: ret i32 [[MUL]]
;
entry:
br label %for.body
for.body: ; preds = %entry, %for.body
%i.02 = phi i64 [ 0, %entry ], [ %inc, %for.body ]
%f_prod.01 = phi i32 [ 1, %entry ], [ %mul, %for.body ]
%arrayidx = getelementptr inbounds i32, i32* %arr_d, i64 %i.02
%0 = load i32, i32* %arrayidx, align 4
%mul = mul i32 %f_prod.01, %0
%inc = add i64 %i.02, 1
%cmp = icmp ult i64 %inc, 1000
br i1 %cmp, label %for.body, label %end
end: ; preds = %for.body
%f_prod.0.lcssa = phi i32 [ %mul, %for.body ]
ret i32 %f_prod.0.lcssa
}
define i32 @test_multiple_int_phi_operands(i32* %arr_d, i1 %entry_cond) {
; CHECK-LABEL: @test_multiple_int_phi_operands(
; CHECK-NEXT: entry:
; CHECK-NEXT: br i1 [[ENTRY_COND:%.*]], label [[FOR_BODY:%.*]], label [[ENTRY_2:%.*]]
; CHECK: entry_2:
; CHECK-NEXT: br label [[FOR_BODY]]
; CHECK: for.body:
; CHECK-NEXT: [[I_02:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 0, [[ENTRY_2]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[F_PROD_01:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ 0, [[ENTRY_2]] ], [ [[MUL:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[ARR_D:%.*]], i64 [[I_02]]
; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[MUL]] = mul i32 [[F_PROD_01]], [[TMP0]]
; CHECK-NEXT: [[INC]] = add nuw nsw i64 [[I_02]], 1
; CHECK-NEXT: [[CMP:%.*]] = icmp ult i64 [[I_02]], 999
; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[END:%.*]]
; CHECK: end:
; CHECK-NEXT: ret i32 [[MUL]]
;
entry:
br i1 %entry_cond, label %for.body, label %entry_2
entry_2:
br label %for.body
for.body: ; preds = %entry, %for.body
%i.02 = phi i64 [ 0, %entry ], [0, %entry_2], [ %inc, %for.body ]
%f_prod.01 = phi i32 [ 0, %entry ], [0, %entry_2], [ %mul, %for.body ]
%arrayidx = getelementptr inbounds i32, i32* %arr_d, i64 %i.02
%0 = load i32, i32* %arrayidx, align 4
%mul = mul i32 %f_prod.01, %0
%inc = add i64 %i.02, 1
%cmp = icmp ult i64 %inc, 1000
br i1 %cmp, label %for.body, label %end
end: ; preds = %for.body
%f_prod.0.lcssa = phi i32 [ %mul, %for.body ]
ret i32 %f_prod.0.lcssa
}
define i32 @test_multiple_int_phi_operands_initalise_to_non_zero(i32* %arr_d, i1 %entry_cond) {
; CHECK-LABEL: @test_multiple_int_phi_operands_initalise_to_non_zero(
; CHECK-NEXT: entry:
; CHECK-NEXT: br i1 [[ENTRY_COND:%.*]], label [[FOR_BODY:%.*]], label [[ENTRY_2:%.*]]
; CHECK: entry_2:
; CHECK-NEXT: br label [[FOR_BODY]]
; CHECK: for.body:
; CHECK-NEXT: [[I_02:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 0, [[ENTRY_2]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[F_PROD_01:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ 1, [[ENTRY_2]] ], [ [[MUL:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[ARR_D:%.*]], i64 [[I_02]]
; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[MUL]] = mul i32 [[F_PROD_01]], [[TMP0]]
; CHECK-NEXT: [[INC]] = add nuw nsw i64 [[I_02]], 1
; CHECK-NEXT: [[CMP:%.*]] = icmp ult i64 [[I_02]], 999
; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[END:%.*]]
; CHECK: end:
; CHECK-NEXT: ret i32 [[MUL]]
;
entry:
br i1 %entry_cond, label %for.body, label %entry_2
entry_2:
br label %for.body
for.body: ; preds = %entry, %for.body
%i.02 = phi i64 [ 0, %entry ], [0, %entry_2], [ %inc, %for.body ]
%f_prod.01 = phi i32 [ 0, %entry ], [1, %entry_2], [ %mul, %for.body ]
%arrayidx = getelementptr inbounds i32, i32* %arr_d, i64 %i.02
%0 = load i32, i32* %arrayidx, align 4
%mul = mul i32 %f_prod.01, %0
%inc = add i64 %i.02, 1
%cmp = icmp ult i64 %inc, 1000
br i1 %cmp, label %for.body, label %end
end: ; preds = %for.body
%f_prod.0.lcssa = phi i32 [ %mul, %for.body ]
ret i32 %f_prod.0.lcssa
}