<div dir="ltr"><div>Good point! I removed those calls, and the existing tests for the fold (just above the ones that I added) all still pass. <br></div><div><br></div><div>The existing code was already strange to be calling isKnownNonNegative on a constant.</div><div>I'll let this bake a bit then remove that clause and update the comment.<br></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Tue, Apr 19, 2022 at 3:30 PM Nikita Popov <<a href="mailto:nikita.ppv@gmail.com">nikita.ppv@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Tue, Apr 19, 2022 at 9:21 PM Sanjay Patel via llvm-commits <<a href="mailto:llvm-commits@lists.llvm.org" target="_blank">llvm-commits@lists.llvm.org</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><br>
Author: Sanjay Patel<br>
Date: 2022-04-19T15:21:34-04:00<br>
New Revision: 8a9c70fc01e6c900f060f3c23d96ee444be33a9a<br>
<br>
URL: <a href="https://github.com/llvm/llvm-project/commit/8a9c70fc01e6c900f060f3c23d96ee444be33a9a" rel="noreferrer" target="_blank">https://github.com/llvm/llvm-project/commit/8a9c70fc01e6c900f060f3c23d96ee444be33a9a</a><br>
DIFF: <a href="https://github.com/llvm/llvm-project/commit/8a9c70fc01e6c900f060f3c23d96ee444be33a9a.diff" rel="noreferrer" target="_blank">https://github.com/llvm/llvm-project/commit/8a9c70fc01e6c900f060f3c23d96ee444be33a9a.diff</a><br>
<br>
LOG: [InstCombine] C0 shift (X add nuw C) --> (C0 shift C) shift X<br>
<br>
With 'nuw' we can convert the increment of the shift amount<br>
into a pre-shift (constant fold) of the shifted constant:<br>
<a href="https://alive2.llvm.org/ce/z/FkTyR2" rel="noreferrer" target="_blank">https://alive2.llvm.org/ce/z/FkTyR2</a><br>
<br>
Fixes issue #41976<br>
<br>
Added: <br>
<br>
<br>
Modified: <br>
llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp<br>
llvm/test/Transforms/InstCombine/shift-add.ll<br>
<br>
Removed: <br>
<br>
<br>
<br>
################################################################################<br>
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp<br>
index 26fe8e08d066d..cc4d53d27f031 100644<br>
--- a/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp<br>
+++ b/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp<br>
@@ -399,13 +399,14 @@ Instruction *InstCombinerImpl::commonShiftTransforms(BinaryOperator &I) {<br>
reassociateShiftAmtsOfTwoSameDirectionShifts(&I, SQ)))<br>
return NewShift;<br>
<br>
- // (C1 shift (A add C2)) -> (C1 shift C2) shift A)<br>
- // iff A and C2 are both positive.<br>
+ // C0 shift (A add C) -> (C0 shift C) shift A<br>
+ // iff A and C2 are both positive or the add has 'nuw'.<br>
Value *A;<br>
Constant *C;<br>
if (match(Op0, m_Constant()) && match(Op1, m_Add(m_Value(A), m_Constant(C))))<br>
- if (isKnownNonNegative(A, DL, 0, &AC, &I, &DT) &&<br>
- isKnownNonNegative(C, DL, 0, &AC, &I, &DT))<br>
+ if (cast<BinaryOperator>(Op1)->hasNoUnsignedWrap() ||<br>
+ (isKnownNonNegative(A, DL, 0, &AC, &I, &DT) &&<br>
+ isKnownNonNegative(C, DL, 0, &AC, &I, &DT)))<br>
return BinaryOperator::Create(<br>
I.getOpcode(), Builder.CreateBinOp(I.getOpcode(), Op0, C), A);<br></blockquote><div><br></div><div>Just wondering, are those isKnownNonNegative() actually still necessary? I'd assume that we'd infer "add nuw" from those anyway.<br></div><div><br></div><div>Nikita<br></div><div><br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
<br>
diff --git a/llvm/test/Transforms/InstCombine/shift-add.ll b/llvm/test/Transforms/InstCombine/shift-add.ll<br>
index d9a1c3b1878b3..f75f6a95f8f8a 100644<br>
--- a/llvm/test/Transforms/InstCombine/shift-add.ll<br>
+++ b/llvm/test/Transforms/InstCombine/shift-add.ll<br>
@@ -123,8 +123,7 @@ define <4 x i32> @lshr_C1_add_A_C2_v4i32_splat(i16 %I) {<br>
<br>
define i32 @shl_add_nuw(i32 %x) {<br>
; CHECK-LABEL: @shl_add_nuw(<br>
-; CHECK-NEXT: [[A:%.*]] = add nuw i32 [[X:%.*]], 5<br>
-; CHECK-NEXT: [[R:%.*]] = shl i32 6, [[A]]<br>
+; CHECK-NEXT: [[R:%.*]] = shl i32 192, [[X:%.*]]<br>
; CHECK-NEXT: ret i32 [[R]]<br>
;<br>
%a = add nuw i32 %x, 5<br>
@@ -132,10 +131,11 @@ define i32 @shl_add_nuw(i32 %x) {<br>
ret i32 %r<br>
}<br>
<br>
+; vectors with arbitrary constants work too<br>
+<br>
define <2 x i12> @lshr_add_nuw(<2 x i12> %x) {<br>
; CHECK-LABEL: @lshr_add_nuw(<br>
-; CHECK-NEXT: [[A:%.*]] = add nuw <2 x i12> [[X:%.*]], <i12 5, i12 1><br>
-; CHECK-NEXT: [[R:%.*]] = lshr <2 x i12> <i12 6, i12 42>, [[A]]<br>
+; CHECK-NEXT: [[R:%.*]] = lshr <2 x i12> <i12 0, i12 21>, [[X:%.*]]<br>
; CHECK-NEXT: ret <2 x i12> [[R]]<br>
;<br>
%a = add nuw <2 x i12> %x, <i12 5, i12 1><br>
@@ -143,12 +143,13 @@ define <2 x i12> @lshr_add_nuw(<2 x i12> %x) {<br>
ret <2 x i12> %r<br>
}<br>
<br>
+; extra use is ok and in this case the result can be simplified to a constant<br>
+<br>
define i32 @ashr_add_nuw(i32 %x, i32* %p) {<br>
; CHECK-LABEL: @ashr_add_nuw(<br>
; CHECK-NEXT: [[A:%.*]] = add nuw i32 [[X:%.*]], 5<br>
; CHECK-NEXT: store i32 [[A]], i32* [[P:%.*]], align 4<br>
-; CHECK-NEXT: [[R:%.*]] = ashr i32 -6, [[A]]<br>
-; CHECK-NEXT: ret i32 [[R]]<br>
+; CHECK-NEXT: ret i32 -1<br>
;<br>
%a = add nuw i32 %x, 5<br>
store i32 %a, i32* %p<br>
@@ -156,6 +157,8 @@ define i32 @ashr_add_nuw(i32 %x, i32* %p) {<br>
ret i32 %r<br>
}<br>
<br>
+; negative test - must have 'nuw'<br>
+<br>
define i32 @shl_add_nsw(i32 %x) {<br>
; CHECK-LABEL: @shl_add_nsw(<br>
; CHECK-NEXT: [[A:%.*]] = add nsw i32 [[X:%.*]], 5<br>
<br>
<br>
<br>
_______________________________________________<br>
llvm-commits mailing list<br>
<a href="mailto:llvm-commits@lists.llvm.org" target="_blank">llvm-commits@lists.llvm.org</a><br>
<a href="https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits" rel="noreferrer" target="_blank">https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits</a><br>
</blockquote></div></div>
</blockquote></div>