Skip to content

[LoongArch] Add patterns for [X]VSTELM_W to adapt to LA32 #165224

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

Jump to bottom
Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[LoongArch] Add patterns for [X]VSTELM_W to adapt to LA32 #165224

wants to merge 1 commit into from

Conversation

@zhaoqi5
Copy link
Contributor

@zhaoqi5 zhaoqi5 commented Oct 27, 2025

No description provided.

@llvmbot
Copy link
Member

llvmbot commented Oct 27, 2025

@llvm/pr-subscribers-backend-loongarch

Author: ZhaoQi (zhaoqi5)

Changes

Patch is 38.06 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/165224.diff

21 Files Affected:

  • (modified) llvm/lib/Target/LoongArch/LoongArchLASXInstrInfo.td (+4-1)
  • (modified) llvm/lib/Target/LoongArch/LoongArchLSXInstrInfo.td (+4)
  • (modified) llvm/test/CodeGen/LoongArch/lasx/ir-instruction/extractelement.ll (+20-40)
  • (modified) llvm/test/CodeGen/LoongArch/lasx/vec-reduce-add.ll (+1-2)
  • (modified) llvm/test/CodeGen/LoongArch/lasx/vec-reduce-and.ll (+11-24)
  • (modified) llvm/test/CodeGen/LoongArch/lasx/vec-reduce-or.ll (+11-24)
  • (modified) llvm/test/CodeGen/LoongArch/lasx/vec-reduce-smax.ll (+11-24)
  • (modified) llvm/test/CodeGen/LoongArch/lasx/vec-reduce-smin.ll (+11-24)
  • (modified) llvm/test/CodeGen/LoongArch/lasx/vec-reduce-umax.ll (+11-24)
  • (modified) llvm/test/CodeGen/LoongArch/lasx/vec-reduce-umin.ll (+11-24)
  • (modified) llvm/test/CodeGen/LoongArch/lasx/vec-reduce-xor.ll (+11-24)
  • (modified) llvm/test/CodeGen/LoongArch/lsx/ir-instruction/extractelement.ll (+9-18)
  • (modified) llvm/test/CodeGen/LoongArch/lsx/vec-reduce-add.ll (+2-4)
  • (modified) llvm/test/CodeGen/LoongArch/lsx/vec-reduce-and.ll (+10-22)
  • (modified) llvm/test/CodeGen/LoongArch/lsx/vec-reduce-or.ll (+10-22)
  • (modified) llvm/test/CodeGen/LoongArch/lsx/vec-reduce-smax.ll (+10-22)
  • (modified) llvm/test/CodeGen/LoongArch/lsx/vec-reduce-smin.ll (+10-22)
  • (modified) llvm/test/CodeGen/LoongArch/lsx/vec-reduce-umax.ll (+10-22)
  • (modified) llvm/test/CodeGen/LoongArch/lsx/vec-reduce-umin.ll (+10-22)
  • (modified) llvm/test/CodeGen/LoongArch/lsx/vec-reduce-xor.ll (+10-22)
  • (modified) llvm/test/CodeGen/LoongArch/lsx/vec-trunc.ll (+10-20)
diff --git a/llvm/lib/Target/LoongArch/LoongArchLASXInstrInfo.td b/llvm/lib/Target/LoongArch/LoongArchLASXInstrInfo.td
index 613dea6093f5f..6a3717bd6dda7 100644
--- a/llvm/lib/Target/LoongArch/LoongArchLASXInstrInfo.td
+++ b/llvm/lib/Target/LoongArch/LoongArchLASXInstrInfo.td
@@ -1920,10 +1920,13 @@ foreach vt = [v32i8, v16i16, v8i32, v4i64, v8f32, v4f64] in
   def : Pat<(vt (loongarch_xvreplve0q LASX256:$xj)),
             (XVREPLVE0_Q LASX256:$xj)>;
 
-// VSTELM
+// XVSTELM
 defm : VstelmPat<truncstorei8, v32i8, XVSTELM_B, simm8, uimm5>;
 defm : VstelmPat<truncstorei16, v16i16, XVSTELM_H, simm8_lsl1, uimm4>;
 defm : VstelmPat<truncstorei32, v8i32, XVSTELM_W, simm8_lsl2, uimm3>;
+let Predicates = [IsLA32] in {
+  defm : VstelmPat<store, v8i32, XVSTELM_W, simm8_lsl2, uimm3>;
+} // Predicates = [IsLA32]
 defm : VstelmPat<store, v4i64, XVSTELM_D, simm8_lsl3, uimm2>;
 defm : VstelmPat<store, v8f32, XVSTELM_W, simm8_lsl2, uimm3, f32>;
 defm : VstelmPat<store, v4f64, XVSTELM_D, simm8_lsl3, uimm2, f64>;
diff --git a/llvm/lib/Target/LoongArch/LoongArchLSXInstrInfo.td b/llvm/lib/Target/LoongArch/LoongArchLSXInstrInfo.td
index 4619c6bd248a6..1d60cb9360289 100644
--- a/llvm/lib/Target/LoongArch/LoongArchLSXInstrInfo.td
+++ b/llvm/lib/Target/LoongArch/LoongArchLSXInstrInfo.td
@@ -2061,9 +2061,13 @@ def : Pat<(lsxsplatf32 FPR32:$fj),
 def : Pat<(lsxsplatf64 FPR64:$fj),
           (VREPLVEI_D (SUBREG_TO_REG (i64 0), FPR64:$fj, sub_64), 0)>;
 
+// VSTELM
 defm : VstelmPat<truncstorei8, v16i8, VSTELM_B, simm8, uimm4>;
 defm : VstelmPat<truncstorei16, v8i16, VSTELM_H, simm8_lsl1, uimm3>;
 defm : VstelmPat<truncstorei32, v4i32, VSTELM_W, simm8_lsl2, uimm2>;
+let Predicates = [IsLA32] in {
+  defm : VstelmPat<store, v4i32, VSTELM_W, simm8_lsl2, uimm2>;
+} // Predicates = [IsLA32]
 defm : VstelmPat<store, v2i64, VSTELM_D, simm8_lsl3, uimm1>;
 defm : VstelmPat<store, v4f32, VSTELM_W, simm8_lsl2, uimm2, f32>;
 defm : VstelmPat<store, v2f64, VSTELM_D, simm8_lsl3, uimm1, f64>;
diff --git a/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/extractelement.ll b/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/extractelement.ll
index 60b51755681a4..d9d1f19726e62 100644
--- a/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/extractelement.ll
+++ b/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/extractelement.ll
@@ -27,18 +27,11 @@ define void @extract_16xi16(ptr %src, ptr %dst) nounwind {
 }
 
 define void @extract_8xi32(ptr %src, ptr %dst) nounwind {
-; LA32-LABEL: extract_8xi32:
-; LA32:       # %bb.0:
-; LA32-NEXT:    xvld $xr0, $a0, 0
-; LA32-NEXT:    xvpickve2gr.w $a0, $xr0, 1
-; LA32-NEXT:    st.w $a0, $a1, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: extract_8xi32:
-; LA64:       # %bb.0:
-; LA64-NEXT:    xvld $xr0, $a0, 0
-; LA64-NEXT:    xvstelm.w $xr0, $a1, 0, 1
-; LA64-NEXT:    ret
+; CHECK-LABEL: extract_8xi32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a0, 0
+; CHECK-NEXT:    xvstelm.w $xr0, $a1, 0, 1
+; CHECK-NEXT:    ret
   %v = load volatile <8 x i32>, ptr %src
   %e = extractelement <8 x i32> %v, i32 1
   store i32 %e, ptr %dst
@@ -49,10 +42,8 @@ define void @extract_4xi64(ptr %src, ptr %dst) nounwind {
 ; LA32-LABEL: extract_4xi64:
 ; LA32:       # %bb.0:
 ; LA32-NEXT:    xvld $xr0, $a0, 0
-; LA32-NEXT:    xvpickve2gr.w $a0, $xr0, 2
-; LA32-NEXT:    xvpickve2gr.w $a2, $xr0, 3
-; LA32-NEXT:    st.w $a2, $a1, 4
-; LA32-NEXT:    st.w $a0, $a1, 0
+; LA32-NEXT:    xvstelm.w $xr0, $a1, 4, 3
+; LA32-NEXT:    xvstelm.w $xr0, $a1, 0, 2
 ; LA32-NEXT:    ret
 ;
 ; LA64-LABEL: extract_4xi64:
@@ -139,22 +130,13 @@ define void @extract_16xi16_idx(ptr %src, ptr %dst, i32 %idx) nounwind {
 }
 
 define void @extract_8xi32_idx(ptr %src, ptr %dst, i32 %idx) nounwind {
-; LA32-LABEL: extract_8xi32_idx:
-; LA32:       # %bb.0:
-; LA32-NEXT:    xvld $xr0, $a0, 0
-; LA32-NEXT:    xvreplgr2vr.w $xr1, $a2
-; LA32-NEXT:    xvperm.w $xr0, $xr0, $xr1
-; LA32-NEXT:    xvpickve2gr.w $a0, $xr0, 0
-; LA32-NEXT:    st.w $a0, $a1, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: extract_8xi32_idx:
-; LA64:       # %bb.0:
-; LA64-NEXT:    xvld $xr0, $a0, 0
-; LA64-NEXT:    xvreplgr2vr.w $xr1, $a2
-; LA64-NEXT:    xvperm.w $xr0, $xr0, $xr1
-; LA64-NEXT:    xvstelm.w $xr0, $a1, 0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: extract_8xi32_idx:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a0, 0
+; CHECK-NEXT:    xvreplgr2vr.w $xr1, $a2
+; CHECK-NEXT:    xvperm.w $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvstelm.w $xr0, $a1, 0, 0
+; CHECK-NEXT:    ret
   %v = load volatile <8 x i32>, ptr %src
   %e = extractelement <8 x i32> %v, i32 %idx
   store i32 %e, ptr %dst
@@ -169,12 +151,10 @@ define void @extract_4xi64_idx(ptr %src, ptr %dst, i32 %idx) nounwind {
 ; LA32-NEXT:    addi.w $a2, $a0, 1
 ; LA32-NEXT:    xvreplgr2vr.w $xr1, $a2
 ; LA32-NEXT:    xvperm.w $xr1, $xr0, $xr1
-; LA32-NEXT:    xvpickve2gr.w $a2, $xr1, 0
-; LA32-NEXT:    xvreplgr2vr.w $xr1, $a0
-; LA32-NEXT:    xvperm.w $xr0, $xr0, $xr1
-; LA32-NEXT:    xvpickve2gr.w $a0, $xr0, 0
-; LA32-NEXT:    st.w $a0, $a1, 0
-; LA32-NEXT:    st.w $a2, $a1, 4
+; LA32-NEXT:    xvreplgr2vr.w $xr2, $a0
+; LA32-NEXT:    xvperm.w $xr0, $xr0, $xr2
+; LA32-NEXT:    xvstelm.w $xr0, $a1, 0, 0
+; LA32-NEXT:    xvstelm.w $xr1, $a1, 4, 0
 ; LA32-NEXT:    ret
 ;
 ; LA64-LABEL: extract_4xi64_idx:
@@ -233,8 +213,8 @@ define void @eliminate_frame_index(<8 x i32> %a) nounwind {
 ; LA32-LABEL: eliminate_frame_index:
 ; LA32:       # %bb.0:
 ; LA32-NEXT:    addi.w $sp, $sp, -1040
-; LA32-NEXT:    xvpickve2gr.w $a0, $xr0, 1
-; LA32-NEXT:    st.w $a0, $sp, 524
+; LA32-NEXT:    addi.w $a0, $sp, 524
+; LA32-NEXT:    xvstelm.w $xr0, $a0, 0, 1
 ; LA32-NEXT:    addi.w $sp, $sp, 1040
 ; LA32-NEXT:    ret
 ;
diff --git a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-add.ll b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-add.ll
index 3e815a174d232..a6d1b23e083e9 100644
--- a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-add.ll
+++ b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-add.ll
@@ -72,8 +72,7 @@ define void @vec_reduce_add_v8i32(ptr %src, ptr %dst) nounwind {
 ; LA32-NEXT:    xvhaddw.q.d $xr0, $xr0, $xr0
 ; LA32-NEXT:    xvpermi.d $xr1, $xr0, 2
 ; LA32-NEXT:    xvadd.d $xr0, $xr1, $xr0
-; LA32-NEXT:    xvpickve2gr.w $a0, $xr0, 0
-; LA32-NEXT:    st.w $a0, $a1, 0
+; LA32-NEXT:    xvstelm.w $xr0, $a1, 0, 0
 ; LA32-NEXT:    ret
 ;
 ; LA64-LABEL: vec_reduce_add_v8i32:
diff --git a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-and.ll b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-and.ll
index 23cc230f04503..614dcf4b500ff 100644
--- a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-and.ll
+++ b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-and.ll
@@ -45,30 +45,17 @@ define void @vec_reduce_and_v16i16(ptr %src, ptr %dst) nounwind {
 }
 
 define void @vec_reduce_and_v8i32(ptr %src, ptr %dst) nounwind {
-; LA32-LABEL: vec_reduce_and_v8i32:
-; LA32:       # %bb.0:
-; LA32-NEXT:    xvld $xr0, $a0, 0
-; LA32-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA32-NEXT:    vand.v $vr0, $vr0, $vr1
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA32-NEXT:    vand.v $vr0, $vr1, $vr0
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA32-NEXT:    vand.v $vr0, $vr1, $vr0
-; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
-; LA32-NEXT:    st.w $a0, $a1, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: vec_reduce_and_v8i32:
-; LA64:       # %bb.0:
-; LA64-NEXT:    xvld $xr0, $a0, 0
-; LA64-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA64-NEXT:    vand.v $vr0, $vr0, $vr1
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA64-NEXT:    vand.v $vr0, $vr1, $vr0
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA64-NEXT:    vand.v $vr0, $vr1, $vr0
-; LA64-NEXT:    vstelm.w $vr0, $a1, 0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: vec_reduce_and_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a0, 0
+; CHECK-NEXT:    xvpermi.q $xr1, $xr0, 1
+; CHECK-NEXT:    vand.v $vr0, $vr0, $vr1
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 8
+; CHECK-NEXT:    vand.v $vr0, $vr1, $vr0
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 4
+; CHECK-NEXT:    vand.v $vr0, $vr1, $vr0
+; CHECK-NEXT:    vstelm.w $vr0, $a1, 0, 0
+; CHECK-NEXT:    ret
   %v = load <8 x i32>, ptr %src
   %res = call i32 @llvm.vector.reduce.and.v8i32(<8 x i32> %v)
   store i32 %res, ptr %dst
diff --git a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-or.ll b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-or.ll
index d7d3afc6dd1da..e3e02b5ba8b7d 100644
--- a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-or.ll
+++ b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-or.ll
@@ -45,30 +45,17 @@ define void @vec_reduce_or_v16i16(ptr %src, ptr %dst) nounwind {
 }
 
 define void @vec_reduce_or_v8i32(ptr %src, ptr %dst) nounwind {
-; LA32-LABEL: vec_reduce_or_v8i32:
-; LA32:       # %bb.0:
-; LA32-NEXT:    xvld $xr0, $a0, 0
-; LA32-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA32-NEXT:    vor.v $vr0, $vr0, $vr1
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA32-NEXT:    vor.v $vr0, $vr1, $vr0
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA32-NEXT:    vor.v $vr0, $vr1, $vr0
-; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
-; LA32-NEXT:    st.w $a0, $a1, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: vec_reduce_or_v8i32:
-; LA64:       # %bb.0:
-; LA64-NEXT:    xvld $xr0, $a0, 0
-; LA64-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA64-NEXT:    vor.v $vr0, $vr0, $vr1
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA64-NEXT:    vor.v $vr0, $vr1, $vr0
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA64-NEXT:    vor.v $vr0, $vr1, $vr0
-; LA64-NEXT:    vstelm.w $vr0, $a1, 0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: vec_reduce_or_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a0, 0
+; CHECK-NEXT:    xvpermi.q $xr1, $xr0, 1
+; CHECK-NEXT:    vor.v $vr0, $vr0, $vr1
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 8
+; CHECK-NEXT:    vor.v $vr0, $vr1, $vr0
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 4
+; CHECK-NEXT:    vor.v $vr0, $vr1, $vr0
+; CHECK-NEXT:    vstelm.w $vr0, $a1, 0, 0
+; CHECK-NEXT:    ret
   %v = load <8 x i32>, ptr %src
   %res = call i32 @llvm.vector.reduce.or.v8i32(<8 x i32> %v)
   store i32 %res, ptr %dst
diff --git a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-smax.ll b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-smax.ll
index 8cbbb52884865..8024b5906e98b 100644
--- a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-smax.ll
+++ b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-smax.ll
@@ -45,30 +45,17 @@ define void @vec_reduce_smax_v16i16(ptr %src, ptr %dst) nounwind {
 }
 
 define void @vec_reduce_smax_v8i32(ptr %src, ptr %dst) nounwind {
-; LA32-LABEL: vec_reduce_smax_v8i32:
-; LA32:       # %bb.0:
-; LA32-NEXT:    xvld $xr0, $a0, 0
-; LA32-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA32-NEXT:    vmax.w $vr0, $vr0, $vr1
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA32-NEXT:    vmax.w $vr0, $vr1, $vr0
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA32-NEXT:    vmax.w $vr0, $vr1, $vr0
-; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
-; LA32-NEXT:    st.w $a0, $a1, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: vec_reduce_smax_v8i32:
-; LA64:       # %bb.0:
-; LA64-NEXT:    xvld $xr0, $a0, 0
-; LA64-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA64-NEXT:    vmax.w $vr0, $vr0, $vr1
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA64-NEXT:    vmax.w $vr0, $vr1, $vr0
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA64-NEXT:    vmax.w $vr0, $vr1, $vr0
-; LA64-NEXT:    vstelm.w $vr0, $a1, 0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: vec_reduce_smax_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a0, 0
+; CHECK-NEXT:    xvpermi.q $xr1, $xr0, 1
+; CHECK-NEXT:    vmax.w $vr0, $vr0, $vr1
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 8
+; CHECK-NEXT:    vmax.w $vr0, $vr1, $vr0
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 4
+; CHECK-NEXT:    vmax.w $vr0, $vr1, $vr0
+; CHECK-NEXT:    vstelm.w $vr0, $a1, 0, 0
+; CHECK-NEXT:    ret
   %v = load <8 x i32>, ptr %src
   %res = call i32 @llvm.vector.reduce.smax.v8i32(<8 x i32> %v)
   store i32 %res, ptr %dst
diff --git a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-smin.ll b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-smin.ll
index c34852aa8a28f..f1ce9b20adcc2 100644
--- a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-smin.ll
+++ b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-smin.ll
@@ -45,30 +45,17 @@ define void @vec_reduce_smin_v16i16(ptr %src, ptr %dst) nounwind {
 }
 
 define void @vec_reduce_smin_v8i32(ptr %src, ptr %dst) nounwind {
-; LA32-LABEL: vec_reduce_smin_v8i32:
-; LA32:       # %bb.0:
-; LA32-NEXT:    xvld $xr0, $a0, 0
-; LA32-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA32-NEXT:    vmin.w $vr0, $vr0, $vr1
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA32-NEXT:    vmin.w $vr0, $vr1, $vr0
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA32-NEXT:    vmin.w $vr0, $vr1, $vr0
-; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
-; LA32-NEXT:    st.w $a0, $a1, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: vec_reduce_smin_v8i32:
-; LA64:       # %bb.0:
-; LA64-NEXT:    xvld $xr0, $a0, 0
-; LA64-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA64-NEXT:    vmin.w $vr0, $vr0, $vr1
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA64-NEXT:    vmin.w $vr0, $vr1, $vr0
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA64-NEXT:    vmin.w $vr0, $vr1, $vr0
-; LA64-NEXT:    vstelm.w $vr0, $a1, 0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: vec_reduce_smin_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a0, 0
+; CHECK-NEXT:    xvpermi.q $xr1, $xr0, 1
+; CHECK-NEXT:    vmin.w $vr0, $vr0, $vr1
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 8
+; CHECK-NEXT:    vmin.w $vr0, $vr1, $vr0
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 4
+; CHECK-NEXT:    vmin.w $vr0, $vr1, $vr0
+; CHECK-NEXT:    vstelm.w $vr0, $a1, 0, 0
+; CHECK-NEXT:    ret
   %v = load <8 x i32>, ptr %src
   %res = call i32 @llvm.vector.reduce.smin.v8i32(<8 x i32> %v)
   store i32 %res, ptr %dst
diff --git a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-umax.ll b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-umax.ll
index c44f83a909a68..040e25a99cd5c 100644
--- a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-umax.ll
+++ b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-umax.ll
@@ -45,30 +45,17 @@ define void @vec_reduce_umax_v16i16(ptr %src, ptr %dst) nounwind {
 }
 
 define void @vec_reduce_umax_v8i32(ptr %src, ptr %dst) nounwind {
-; LA32-LABEL: vec_reduce_umax_v8i32:
-; LA32:       # %bb.0:
-; LA32-NEXT:    xvld $xr0, $a0, 0
-; LA32-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA32-NEXT:    vmax.wu $vr0, $vr0, $vr1
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA32-NEXT:    vmax.wu $vr0, $vr1, $vr0
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA32-NEXT:    vmax.wu $vr0, $vr1, $vr0
-; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
-; LA32-NEXT:    st.w $a0, $a1, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: vec_reduce_umax_v8i32:
-; LA64:       # %bb.0:
-; LA64-NEXT:    xvld $xr0, $a0, 0
-; LA64-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA64-NEXT:    vmax.wu $vr0, $vr0, $vr1
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA64-NEXT:    vmax.wu $vr0, $vr1, $vr0
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA64-NEXT:    vmax.wu $vr0, $vr1, $vr0
-; LA64-NEXT:    vstelm.w $vr0, $a1, 0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: vec_reduce_umax_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a0, 0
+; CHECK-NEXT:    xvpermi.q $xr1, $xr0, 1
+; CHECK-NEXT:    vmax.wu $vr0, $vr0, $vr1
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 8
+; CHECK-NEXT:    vmax.wu $vr0, $vr1, $vr0
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 4
+; CHECK-NEXT:    vmax.wu $vr0, $vr1, $vr0
+; CHECK-NEXT:    vstelm.w $vr0, $a1, 0, 0
+; CHECK-NEXT:    ret
   %v = load <8 x i32>, ptr %src
   %res = call i32 @llvm.vector.reduce.umax.v8i32(<8 x i32> %v)
   store i32 %res, ptr %dst
diff --git a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-umin.ll b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-umin.ll
index f91a1b34dffe9..a20b903eb5e69 100644
--- a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-umin.ll
+++ b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-umin.ll
@@ -45,30 +45,17 @@ define void @vec_reduce_umin_v16i16(ptr %src, ptr %dst) nounwind {
 }
 
 define void @vec_reduce_umin_v8i32(ptr %src, ptr %dst) nounwind {
-; LA32-LABEL: vec_reduce_umin_v8i32:
-; LA32:       # %bb.0:
-; LA32-NEXT:    xvld $xr0, $a0, 0
-; LA32-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA32-NEXT:    vmin.wu $vr0, $vr0, $vr1
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA32-NEXT:    vmin.wu $vr0, $vr1, $vr0
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA32-NEXT:    vmin.wu $vr0, $vr1, $vr0
-; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
-; LA32-NEXT:    st.w $a0, $a1, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: vec_reduce_umin_v8i32:
-; LA64:       # %bb.0:
-; LA64-NEXT:    xvld $xr0, $a0, 0
-; LA64-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA64-NEXT:    vmin.wu $vr0, $vr0, $vr1
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA64-NEXT:    vmin.wu $vr0, $vr1, $vr0
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA64-NEXT:    vmin.wu $vr0, $vr1, $vr0
-; LA64-NEXT:    vstelm.w $vr0, $a1, 0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: vec_reduce_umin_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a0, 0
+; CHECK-NEXT:    xvpermi.q $xr1, $xr0, 1
+; CHECK-NEXT:    vmin.wu $vr0, $vr0, $vr1
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 8
+; CHECK-NEXT:    vmin.wu $vr0, $vr1, $vr0
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 4
+; CHECK-NEXT:    vmin.wu $vr0, $vr1, $vr0
+; CHECK-NEXT:    vstelm.w $vr0, $a1, 0, 0
+; CHECK-NEXT:    ret
   %v = load <8 x i32>, ptr %src
   %res = call i32 @llvm.vector.reduce.umin.v8i32(<8 x i32> %v)
   store i32 %res, ptr %dst
diff --git a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-xor.ll b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-xor.ll
index af1a66b574c03..2145947523fa9 100644
--- a/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-xor.ll
+++ b/llvm/test/CodeGen/LoongArch/lasx/vec-reduce-xor.ll
@@ -45,30 +45,17 @@ define void @vec_reduce_xor_v16i16(ptr %src, ptr %dst) nounwind {
 }
 
 define void @vec_reduce_xor_v8i32(ptr %src, ptr %dst) nounwind {
-; LA32-LABEL: vec_reduce_xor_v8i32:
-; LA32:       # %bb.0:
-; LA32-NEXT:    xvld $xr0, $a0, 0
-; LA32-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA32-NEXT:    vxor.v $vr0, $vr0, $vr1
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA32-NEXT:    vxor.v $vr0, $vr1, $vr0
-; LA32-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA32-NEXT:    vxor.v $vr0, $vr1, $vr0
-; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 0
-; LA32-NEXT:    st.w $a0, $a1, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: vec_reduce_xor_v8i32:
-; LA64:       # %bb.0:
-; LA64-NEXT:    xvld $xr0, $a0, 0
-; LA64-NEXT:    xvpermi.q $xr1, $xr0, 1
-; LA64-NEXT:    vxor.v $vr0, $vr0, $vr1
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 8
-; LA64-NEXT:    vxor.v $vr0, $vr1, $vr0
-; LA64-NEXT:    vbsrl.v $vr1, $vr0, 4
-; LA64-NEXT:    vxor.v $vr0, $vr1, $vr0
-; LA64-NEXT:    vstelm.w $vr0, $a1, 0, 0
-; LA64-NEXT:    ret
+; CHECK-LABEL: vec_reduce_xor_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a0, 0
+; CHECK-NEXT:    xvpermi.q $xr1, $xr0, 1
+; CHECK-NEXT:    vxor.v $vr0, $vr0, $vr1
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 8
+; CHECK-NEXT:    vxor.v $vr0, $vr1, $vr0
+; CHECK-NEXT:    vbsrl.v $vr1, $vr0, 4
+; CHECK-NEXT:    vxor.v $vr0, $vr1, $vr0
+; CHECK-NEXT:    vstelm.w $vr0, $a1, 0, 0
+; CHECK-NEXT:    ret
   %v = load <8 x i32>, ptr %src
   %res = call i32 @llvm.vector.reduce.xor.v8i32(<8 x i32> %v)
   store i32 %res, ptr %dst
diff --git a/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/extractelement.ll b/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/extractelement.ll
index b17a90e71e85a..144835d92d0fc 100644
--- a/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/extractelement.ll
+++ b/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/extractelement.ll
@@ -27,18 +27,11 @@ define void @extract_8xi16(ptr %src, ptr %dst) nounwind {
 }
 
 define void @extract_4xi32(ptr %src, ptr %dst) nounwind {
-; LA32-LABEL: extract_4xi32:
-; LA32:       # %bb.0:
-; LA32-NEXT:    vld $vr0, $a0, 0
-; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 1
-; LA32-NEXT:    st.w $a0, $a1, 0
-; LA32-NEXT:    ret
-;
-; LA64-LABEL: extract_4xi32:
-; LA64:       # %bb.0:
-; LA64-NEXT:    vld $vr0, $a0, 0
-; LA64-NEXT:    vstelm.w $vr0, $a1, 0, 1
-; LA64-NEXT:    ret
+; CHECK-LABEL: extract_4xi32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vld $vr0, $a0, 0
+; CHECK-NEXT:    vstelm.w $vr0, $a1, 0, 1
+; CHECK-NEXT:    ret
   %v = load volatile <4 x i32>, ptr %src
   %e = extractelement <4 x i32> %v, i32 1
   store i32 %e, ptr %dst
@@ -49,10 +42,8 @@ define void @extract_2xi64(ptr %src, ptr %dst) nounwind {
 ; LA32-LABEL: extract_2xi64:
 ; LA32:       # %bb.0:
 ; LA32-NEXT:    vld $vr0, $a0, 0
-; LA32-NEXT:    vpickve2gr.w $a0, $vr0, 2
-; LA32-NEXT:    vpickve2gr.w $a2, $vr0, 3
-; LA32-NEXT:    st.w $a2, $a1, 4
-; LA32-NEXT:    st.w $a0, $a1, 0
+; LA32-NEXT:    vstelm.w $vr0, $a1, 4, 3
+; LA32-NE...
[truncated]

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

backend:loongarch

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@zhaoqi5 @llvmbot