Skip to content

[LoongArch] Optimize inserting element to high part of 256bits vector #146816

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
zhaoqi5 merged 5 commits into from
Jul 17, 2025
Merged

[LoongArch] Optimize inserting element to high part of 256bits vector #146816

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
fb473a4
[LoongArch] Optimize inserting element to high part of 256bits vector
zhaoqi5 Jul 3, 2025
2f2c103
Merge branch 'main' into users/zhaoqi5/opt-insert-hi128-element
zhaoqi5 Jul 3, 2025
db8f1fc
Merge branch 'main' into users/zhaoqi5/opt-insert-hi128-element
zhaoqi5 Jul 4, 2025
526717c
Merge branch 'main' into users/zhaoqi5/opt-insert-hi128-element
zhaoqi5 Jul 14, 2025
d30224c
Merge branch 'main' into users/zhaoqi5/opt-insert-hi128-element
zhaoqi5 Jul 17, 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
5 changes: 2 additions & 3 deletions llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6000,10 +6000,9 @@ emitPseudoXVINSGR2VR(MachineInstr &MI, MachineBasicBlock *BB,
Register ScratchReg1 = XSrc;
if (Idx >= HalfSize) {
ScratchReg1 = MRI.createVirtualRegister(RC);
BuildMI(*BB, MI, DL, TII->get(LoongArch::XVPERMI_Q), ScratchReg1)
BuildMI(*BB, MI, DL, TII->get(LoongArch::XVPERMI_D), ScratchReg1)
.addReg(XSrc)
.addReg(XSrc)
.addImm(1);
.addImm(14);
}

Register ScratchSubReg1 = MRI.createVirtualRegister(SubRC);
Expand Down
154 changes: 65 additions & 89 deletions llvm/test/CodeGen/LoongArch/lasx/build-vector.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -250,84 +250,68 @@ define void @buildvector_v32i8(ptr %dst, i8 %a0, i8 %a1, i8 %a2, i8 %a3, i8 %a4,
; CHECK-NEXT: vinsgr2vr.b $vr0, $a2, 14
; CHECK-NEXT: ld.b $a1, $sp, 72
; CHECK-NEXT: vinsgr2vr.b $vr0, $a3, 15
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: ld.b $a2, $sp, 80
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 0
; CHECK-NEXT: ld.b $a1, $sp, 80
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 1
; CHECK-NEXT: ld.b $a1, $sp, 88
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 2
; CHECK-NEXT: ld.b $a1, $sp, 96
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a2, 1
; CHECK-NEXT: ld.b $a2, $sp, 96
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 3
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 2
; CHECK-NEXT: ld.b $a1, $sp, 104
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 4
; CHECK-NEXT: ld.b $a1, $sp, 112
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a2, 3
; CHECK-NEXT: ld.b $a2, $sp, 112
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 5
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 4
; CHECK-NEXT: ld.b $a1, $sp, 120
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 6
; CHECK-NEXT: ld.b $a1, $sp, 128
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a2, 5
; CHECK-NEXT: ld.b $a2, $sp, 128
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 7
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 6
; CHECK-NEXT: ld.b $a1, $sp, 136
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 8
; CHECK-NEXT: ld.b $a1, $sp, 144
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a2, 7
; CHECK-NEXT: ld.b $a2, $sp, 144
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 9
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 8
; CHECK-NEXT: ld.b $a1, $sp, 152
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 10
; CHECK-NEXT: ld.b $a1, $sp, 160
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a2, 9
; CHECK-NEXT: ld.b $a2, $sp, 160
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 11
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 10
; CHECK-NEXT: ld.b $a1, $sp, 168
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 12
; CHECK-NEXT: ld.b $a1, $sp, 176
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a2, 11
; CHECK-NEXT: ld.b $a2, $sp, 176
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 13
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 12
; CHECK-NEXT: ld.b $a1, $sp, 184
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a2, 13
; CHECK-NEXT: ld.b $a2, $sp, 192
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 14
; CHECK-NEXT: ld.b $a1, $sp, 192
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.b $vr1, $a1, 15
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a2, 15
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvst $xr0, $a0, 0
; CHECK-NEXT: ret
Expand Down Expand Up @@ -371,54 +355,46 @@ entry:
define void @buildvector_v16i16(ptr %dst, i16 %a0, i16 %a1, i16 %a2, i16 %a3, i16 %a4, i16 %a5, i16 %a6, i16 %a7, i16 %a8, i16 %a9, i16 %a10, i16 %a11, i16 %a12, i16 %a13, i16 %a14, i16 %a15) nounwind {
; CHECK-LABEL: buildvector_v16i16:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: ld.h $t0, $sp, 8
; CHECK-NEXT: ld.h $t1, $sp, 0
; CHECK-NEXT: ld.h $t0, $sp, 64
; CHECK-NEXT: ld.h $t1, $sp, 56
; CHECK-NEXT: ld.h $t2, $sp, 48
; CHECK-NEXT: ld.h $t3, $sp, 40
; CHECK-NEXT: ld.h $t4, $sp, 32
; CHECK-NEXT: ld.h $t5, $sp, 24
; CHECK-NEXT: ld.h $t6, $sp, 16
; CHECK-NEXT: ld.h $t7, $sp, 8
; CHECK-NEXT: ld.h $t8, $sp, 0
; CHECK-NEXT: vinsgr2vr.h $vr0, $a1, 0
; CHECK-NEXT: vinsgr2vr.h $vr0, $a2, 1
; CHECK-NEXT: vinsgr2vr.h $vr0, $a3, 2
; CHECK-NEXT: vinsgr2vr.h $vr0, $a4, 3
; CHECK-NEXT: vinsgr2vr.h $vr0, $a5, 4
; CHECK-NEXT: vinsgr2vr.h $vr0, $a6, 5
; CHECK-NEXT: vinsgr2vr.h $vr0, $a7, 6
; CHECK-NEXT: vinsgr2vr.h $vr0, $t1, 7
; CHECK-NEXT: ld.h $a1, $sp, 16
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.h $vr1, $t0, 0
; CHECK-NEXT: vinsgr2vr.h $vr0, $t8, 7
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.h $vr1, $t7, 0
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: ld.h $a2, $sp, 24
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.h $vr1, $a1, 1
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.h $vr1, $t6, 1
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: ld.h $a1, $sp, 32
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.h $vr1, $a2, 2
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.h $vr1, $t5, 2
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: ld.h $a2, $sp, 40
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.h $vr1, $a1, 3
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.h $vr1, $t4, 3
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: ld.h $a1, $sp, 48
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.h $vr1, $a2, 4
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.h $vr1, $t3, 4
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: ld.h $a2, $sp, 56
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.h $vr1, $a1, 5
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.h $vr1, $t2, 5
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: ld.h $a1, $sp, 64
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.h $vr1, $a2, 6
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.h $vr1, $t1, 6
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: vinsgr2vr.h $vr1, $a1, 7
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.h $vr1, $t0, 7
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvst $xr0, $a0, 0
; CHECK-NEXT: ret
Expand Down
6 changes: 2 additions & 4 deletions llvm/test/CodeGen/LoongArch/lasx/ir-instruction/insertelement.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,8 +18,7 @@ define void @insert_32xi8_upper(ptr %src, ptr %dst, i8 %in) nounwind {
; CHECK-LABEL: insert_32xi8_upper:
; CHECK: # %bb.0:
; CHECK-NEXT: xvld $xr0, $a0, 0
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.b $vr1, $a2, 0
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvst $xr0, $a1, 0
Expand Down Expand Up @@ -47,8 +46,7 @@ define void @insert_16xi16_upper(ptr %src, ptr %dst, i16 %in) nounwind {
; CHECK-LABEL: insert_16xi16_upper:
; CHECK: # %bb.0:
; CHECK-NEXT: xvld $xr0, $a0, 0
; CHECK-NEXT: xvori.b $xr1, $xr0, 0
; CHECK-NEXT: xvpermi.q $xr1, $xr0, 1
; CHECK-NEXT: xvpermi.d $xr1, $xr0, 14
; CHECK-NEXT: vinsgr2vr.h $vr1, $a2, 0
; CHECK-NEXT: xvpermi.q $xr0, $xr1, 2
; CHECK-NEXT: xvst $xr0, $a1, 0
Expand Down
Loading