Skip to content

[VectorCombine] foldShuffleOfShuffles - fold shuffle(shuffle(x,y),poison) length changing masks #144690

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Jun 22, 2025
Merged

[VectorCombine] foldShuffleOfShuffles - fold shuffle(shuffle(x,y),poison) length changing masks #144690

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
ed15ee0
[VectorCombine] foldShuffleOfShuffles - fold shuffle(shuffle(x,y),poi…
RKSimon Jun 18, 2025
5eb843f
Merge branch 'main' into vectorcombine-permute-shuffle
RKSimon Jun 19, 2025
f0317f1
Merge branch 'main' into vectorcombine-permute-shuffle
RKSimon Jun 21, 2025
fb89ac2
Drop superfluous Match0 check
RKSimon Jun 21, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 13 additions & 2 deletions llvm/lib/Transforms/Vectorize/VectorCombine.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2282,6 +2282,17 @@ bool VectorCombine::foldShuffleOfShuffles(Instruction &I) {
if (!Match0 && !Match1)
return false;

// If the outer shuffle is a permute, then create a fake inner all-poison
// shuffle. This is easier than accounting for length-changing shuffles below.
SmallVector<int, 16> PoisonMask1;
if (!Match1 && isa<PoisonValue>(OuterV1)) {
X1 = X0;
Y1 = Y0;
PoisonMask1.append(InnerMask0.size(), PoisonMaskElem);
InnerMask1 = PoisonMask1;
Match1 = true; // fake match
}

X0 = Match0 ? X0 : OuterV0;
Y0 = Match0 ? Y0 : OuterV0;
X1 = Match1 ? X1 : OuterV1;
Expand Down Expand Up @@ -2356,11 +2367,11 @@ bool VectorCombine::foldShuffleOfShuffles(Instruction &I) {
// Try to merge the shuffles if the new shuffle is not costly.
InstructionCost InnerCost0 = 0;
if (Match0)
InnerCost0 = TTI.getInstructionCost(cast<Instruction>(OuterV0), CostKind);
InnerCost0 = TTI.getInstructionCost(cast<User>(OuterV0), CostKind);

InstructionCost InnerCost1 = 0;
if (Match1)
InnerCost1 = TTI.getInstructionCost(cast<Instruction>(OuterV1), CostKind);
InnerCost1 = TTI.getInstructionCost(cast<User>(OuterV1), CostKind);

InstructionCost OuterCost = TTI.getInstructionCost(&I, CostKind);

Expand Down
6 changes: 2 additions & 4 deletions llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -262,10 +262,8 @@ define <8 x half> @splatandidentity(<8 x half> %a, <8 x half> %b) {

define <8 x half> @splattwice(<8 x half> %a, <8 x half> %b) {
; CHECK-LABEL: @splattwice(
; CHECK-NEXT: [[AS:%.*]] = shufflevector <8 x half> [[A:%.*]], <8 x half> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[BS:%.*]] = shufflevector <8 x half> [[B:%.*]], <8 x half> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x half> [[AS]], <4 x half> poison, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0>
; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <4 x half> [[BS]], <4 x half> poison, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0>
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <8 x half> [[A:%.*]], <8 x half> poison, <8 x i32> zeroinitializer
; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <8 x half> [[B:%.*]], <8 x half> poison, <8 x i32> zeroinitializer
; CHECK-NEXT: [[R:%.*]] = fadd <8 x half> [[TMP1]], [[TMP2]]
; CHECK-NEXT: ret <8 x half> [[R]]
;
Expand Down
36 changes: 16 additions & 20 deletions llvm/test/Transforms/VectorCombine/X86/extract-insert-poison.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,8 +32,7 @@ define <4 x double> @src_ins2_v4f64_ext0_v2f64(<4 x double> %a, <2 x double> %b)
; SSE-NEXT: ret <4 x double> [[INS]]
;
; AVX-LABEL: @src_ins2_v4f64_ext0_v2f64(
; AVX-NEXT: [[TMP1:%.*]] = shufflevector <2 x double> [[B:%.*]], <2 x double> poison, <4 x i32> <i32 0, i32 poison, i32 poison, i32 poison>
; AVX-NEXT: [[INS:%.*]] = shufflevector <4 x double> [[TMP1]], <4 x double> poison, <4 x i32> <i32 poison, i32 poison, i32 0, i32 poison>
; AVX-NEXT: [[INS:%.*]] = shufflevector <2 x double> [[B:%.*]], <2 x double> poison, <4 x i32> <i32 poison, i32 poison, i32 0, i32 poison>
; AVX-NEXT: ret <4 x double> [[INS]]
;
%ext = extractelement <2 x double> %b, i32 0
Expand All @@ -48,8 +47,7 @@ define <4 x double> @src_ins3_v4f64_ext0_v2f64(<4 x double> %a, <2 x double> %b)
; SSE-NEXT: ret <4 x double> [[INS]]
;
; AVX-LABEL: @src_ins3_v4f64_ext0_v2f64(
; AVX-NEXT: [[TMP1:%.*]] = shufflevector <2 x double> [[B:%.*]], <2 x double> poison, <4 x i32> <i32 0, i32 poison, i32 poison, i32 poison>
; AVX-NEXT: [[INS:%.*]] = shufflevector <4 x double> [[TMP1]], <4 x double> poison, <4 x i32> <i32 poison, i32 poison, i32 poison, i32 0>
; AVX-NEXT: [[INS:%.*]] = shufflevector <2 x double> [[B:%.*]], <2 x double> poison, <4 x i32> <i32 poison, i32 poison, i32 poison, i32 0>
; AVX-NEXT: ret <4 x double> [[INS]]
;
%ext = extractelement <2 x double> %b, i32 0
Expand Down Expand Up @@ -86,8 +84,7 @@ define <4 x double> @src_ins2_v4f64_ext1_v2f64(<4 x double> %a, <2 x double> %b)
; SSE-NEXT: ret <4 x double> [[INS]]
;
; AVX-LABEL: @src_ins2_v4f64_ext1_v2f64(
; AVX-NEXT: [[TMP1:%.*]] = shufflevector <2 x double> [[B:%.*]], <2 x double> poison, <4 x i32> <i32 poison, i32 1, i32 poison, i32 poison>
; AVX-NEXT: [[INS:%.*]] = shufflevector <4 x double> [[TMP1]], <4 x double> poison, <4 x i32> <i32 poison, i32 poison, i32 1, i32 poison>
; AVX-NEXT: [[INS:%.*]] = shufflevector <2 x double> [[B:%.*]], <2 x double> poison, <4 x i32> <i32 poison, i32 poison, i32 1, i32 poison>
; AVX-NEXT: ret <4 x double> [[INS]]
;
%ext = extractelement <2 x double> %b, i32 1
Expand All @@ -96,10 +93,14 @@ define <4 x double> @src_ins2_v4f64_ext1_v2f64(<4 x double> %a, <2 x double> %b)
}

define <4 x double> @src_ins3_v4f64_ext1_v2f64(<4 x double> %a, <2 x double> %b) #0 {
; CHECK-LABEL: @src_ins3_v4f64_ext1_v2f64(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <2 x double> [[B:%.*]], <2 x double> poison, <4 x i32> <i32 poison, i32 1, i32 poison, i32 poison>
; CHECK-NEXT: [[INS:%.*]] = shufflevector <4 x double> [[TMP1]], <4 x double> poison, <4 x i32> <i32 poison, i32 poison, i32 poison, i32 1>
; CHECK-NEXT: ret <4 x double> [[INS]]
; SSE-LABEL: @src_ins3_v4f64_ext1_v2f64(
; SSE-NEXT: [[TMP1:%.*]] = shufflevector <2 x double> [[B:%.*]], <2 x double> poison, <4 x i32> <i32 poison, i32 1, i32 poison, i32 poison>
; SSE-NEXT: [[INS:%.*]] = shufflevector <4 x double> [[TMP1]], <4 x double> poison, <4 x i32> <i32 poison, i32 poison, i32 poison, i32 1>
; SSE-NEXT: ret <4 x double> [[INS]]
;
; AVX-LABEL: @src_ins3_v4f64_ext1_v2f64(
; AVX-NEXT: [[INS:%.*]] = shufflevector <2 x double> [[B:%.*]], <2 x double> poison, <4 x i32> <i32 poison, i32 poison, i32 poison, i32 1>
; AVX-NEXT: ret <4 x double> [[INS]]
;
%ext = extractelement <2 x double> %b, i32 1
%ins = insertelement <4 x double> poison, double %ext, i32 3
Expand All @@ -119,8 +120,7 @@ define <2 x double> @src_ins0_v2f64_ext0_v4f64(<2 x double> %a, <4 x double> %b)

define <2 x double> @src_ins0_v2f64_ext1_v4f64(<2 x double> %a, <4 x double> %b) {
; CHECK-LABEL: @src_ins0_v2f64_ext1_v4f64(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x double> [[B:%.*]], <4 x double> poison, <2 x i32> <i32 poison, i32 1>
; CHECK-NEXT: [[INS:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> poison, <2 x i32> <i32 1, i32 poison>
; CHECK-NEXT: [[INS:%.*]] = shufflevector <4 x double> [[B:%.*]], <4 x double> poison, <2 x i32> <i32 1, i32 poison>
; CHECK-NEXT: ret <2 x double> [[INS]]
;
%ext = extractelement <4 x double> %b, i32 1
Expand Down Expand Up @@ -152,8 +152,7 @@ define <2 x double> @src_ins0_v2f64_ext3_v4f64(<2 x double> %a, <4 x double> %b)

define <2 x double> @src_ins1_v2f64_ext0_v4f64(<2 x double> %a, <4 x double> %b) {
; CHECK-LABEL: @src_ins1_v2f64_ext0_v4f64(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x double> [[B:%.*]], <4 x double> poison, <2 x i32> <i32 0, i32 poison>
; CHECK-NEXT: [[INS:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> poison, <2 x i32> <i32 poison, i32 0>
; CHECK-NEXT: [[INS:%.*]] = shufflevector <4 x double> [[B:%.*]], <4 x double> poison, <2 x i32> <i32 poison, i32 0>
; CHECK-NEXT: ret <2 x double> [[INS]]
;
%ext = extractelement <4 x double> %b, i32 0
Expand All @@ -164,8 +163,7 @@ define <2 x double> @src_ins1_v2f64_ext0_v4f64(<2 x double> %a, <4 x double> %b)
define <2 x double> @src_ins1_v2f64_ext1_v4f64(<2 x double> %a, <4 x double> %b) {
; CHECK-LABEL: @src_ins1_v2f64_ext1_v4f64(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x double> [[B:%.*]], <4 x double> poison, <2 x i32> <i32 poison, i32 1>
; CHECK-NEXT: [[INS:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> poison, <2 x i32> <i32 poison, i32 1>
; CHECK-NEXT: ret <2 x double> [[INS]]
; CHECK-NEXT: ret <2 x double> [[TMP1]]
;
%ext = extractelement <4 x double> %b, i32 1
%ins = insertelement <2 x double> poison, double %ext, i32 1
Expand All @@ -174,8 +172,7 @@ define <2 x double> @src_ins1_v2f64_ext1_v4f64(<2 x double> %a, <4 x double> %b)

define <2 x double> @src_ins1_v2f64_ext2_v4f64(<2 x double> %a, <4 x double> %b) {
; SSE-LABEL: @src_ins1_v2f64_ext2_v4f64(
; SSE-NEXT: [[TMP1:%.*]] = shufflevector <4 x double> [[B:%.*]], <4 x double> poison, <2 x i32> <i32 2, i32 poison>
; SSE-NEXT: [[INS:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> poison, <2 x i32> <i32 poison, i32 0>
; SSE-NEXT: [[INS:%.*]] = shufflevector <4 x double> [[B:%.*]], <4 x double> poison, <2 x i32> <i32 poison, i32 2>
; SSE-NEXT: ret <2 x double> [[INS]]
;
; AVX-LABEL: @src_ins1_v2f64_ext2_v4f64(
Expand All @@ -190,8 +187,7 @@ define <2 x double> @src_ins1_v2f64_ext2_v4f64(<2 x double> %a, <4 x double> %b)

define <2 x double> @src_ins1_v2f64_ext3_v4f64(<2 x double> %a, <4 x double> %b) {
; CHECK-LABEL: @src_ins1_v2f64_ext3_v4f64(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x double> [[B:%.*]], <4 x double> poison, <2 x i32> <i32 3, i32 poison>
; CHECK-NEXT: [[INS:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> poison, <2 x i32> <i32 poison, i32 0>
; CHECK-NEXT: [[INS:%.*]] = shufflevector <4 x double> [[B:%.*]], <4 x double> poison, <2 x i32> <i32 poison, i32 3>
; CHECK-NEXT: ret <2 x double> [[INS]]
;
%ext = extractelement <4 x double> %b, i32 3
Expand Down
3 changes: 1 addition & 2 deletions llvm/test/Transforms/VectorCombine/X86/load-inseltpoison.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -578,8 +578,7 @@ define <8 x i32> @load_v1i32_extract_insert_v8i32_extra_use(ptr align 16 derefer
; CHECK-NEXT: [[L:%.*]] = load <1 x i32>, ptr [[P:%.*]], align 4
; CHECK-NEXT: store <1 x i32> [[L]], ptr [[STORE_PTR:%.*]], align 4
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <1 x i32> [[L]], <1 x i32> poison, <8 x i32> <i32 0, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; CHECK-NEXT: [[R:%.*]] = shufflevector <8 x i32> [[TMP1]], <8 x i32> poison, <8 x i32> <i32 0, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; CHECK-NEXT: ret <8 x i32> [[R]]
; CHECK-NEXT: ret <8 x i32> [[TMP1]]
;
%l = load <1 x i32>, ptr %p, align 4
store <1 x i32> %l, ptr %store_ptr
Expand Down
Loading