[llvm] [LoongArch] Fix broadcast load with extension. (PR #155960)
via llvm-commits
llvm-commits at lists.llvm.org
Thu Aug 28 19:03:56 PDT 2025
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-loongarch
Author: None (tangaac)
<details>
<summary>Changes</summary>
PR #<!-- -->135896 introduces [x]vldrepl instructions without handling extension.
This patch will fix that.
---
Full diff: https://github.com/llvm/llvm-project/pull/155960.diff
3 Files Affected:
- (modified) llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp (+2-1)
- (modified) llvm/test/CodeGen/LoongArch/lasx/broadcast-load.ll (+38-6)
- (modified) llvm/test/CodeGen/LoongArch/lsx/broadcast-load.ll (+39-6)
``````````diff
diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
index ffb6c2980026f..478c335c3f07e 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
@@ -2471,8 +2471,9 @@ static SDValue lowerBUILD_VECTORAsBroadCastLoad(BuildVectorSDNode *BVOp,
if (!IdentitySrc || !BVOp->isOnlyUserOf(IdentitySrc.getNode()))
return SDValue();
- if (IsIdeneity) {
+ if (IsIdeneity && ISD::isNON_EXTLoad(IdentitySrc.getNode())) {
auto *LN = cast<LoadSDNode>(IdentitySrc);
+
SDVTList Tys =
LN->isIndexed()
? DAG.getVTList(VT, LN->getBasePtr().getValueType(), MVT::Other)
diff --git a/llvm/test/CodeGen/LoongArch/lasx/broadcast-load.ll b/llvm/test/CodeGen/LoongArch/lasx/broadcast-load.ll
index 976924bdca686..89592a0886cc1 100644
--- a/llvm/test/CodeGen/LoongArch/lasx/broadcast-load.ll
+++ b/llvm/test/CodeGen/LoongArch/lasx/broadcast-load.ll
@@ -18,6 +18,32 @@ define <4 x i64> @should_not_be_optimized(ptr %ptr, ptr %dst) {
ret <4 x i64> %tmp2
}
+define <16 x i16> @should_not_be_optimized_sext_load(ptr %ptr) {
+; CHECK-LABEL: should_not_be_optimized_sext_load:
+; CHECK: # %bb.0:
+; CHECK-NEXT: ld.b $a0, $a0, 0
+; CHECK-NEXT: xvreplgr2vr.h $xr0, $a0
+; CHECK-NEXT: ret
+ %tmp = load i8, ptr %ptr
+ %tmp1 = sext i8 %tmp to i16
+ %tmp2 = insertelement <16 x i16> zeroinitializer, i16 %tmp1, i32 0
+ %tmp3 = shufflevector <16 x i16> %tmp2, <16 x i16> poison, <16 x i32> zeroinitializer
+ ret <16 x i16> %tmp3
+}
+
+define <16 x i16> @should_not_be_optimized_zext_load(ptr %ptr) {
+; CHECK-LABEL: should_not_be_optimized_zext_load:
+; CHECK: # %bb.0:
+; CHECK-NEXT: ld.bu $a0, $a0, 0
+; CHECK-NEXT: xvreplgr2vr.h $xr0, $a0
+; CHECK-NEXT: ret
+ %tmp = load i8, ptr %ptr
+ %tmp1 = zext i8 %tmp to i16
+ %tmp2 = insertelement <16 x i16> zeroinitializer, i16 %tmp1, i32 0
+ %tmp3 = shufflevector <16 x i16> %tmp2, <16 x i16> poison, <16 x i32> zeroinitializer
+ ret <16 x i16> %tmp3
+}
+
define <4 x i64> @xvldrepl_d_unaligned_offset(ptr %ptr) {
; CHECK-LABEL: xvldrepl_d_unaligned_offset:
; CHECK: # %bb.0:
@@ -34,7 +60,8 @@ define <4 x i64> @xvldrepl_d_unaligned_offset(ptr %ptr) {
define <32 x i8> @xvldrepl_b(ptr %ptr) {
; CHECK-LABEL: xvldrepl_b:
; CHECK: # %bb.0:
-; CHECK-NEXT: xvldrepl.b $xr0, $a0, 0
+; CHECK-NEXT: ld.b $a0, $a0, 0
+; CHECK-NEXT: xvreplgr2vr.b $xr0, $a0
; CHECK-NEXT: ret
%tmp = load i8, ptr %ptr
%tmp1 = insertelement <32 x i8> zeroinitializer, i8 %tmp, i32 0
@@ -45,7 +72,8 @@ define <32 x i8> @xvldrepl_b(ptr %ptr) {
define <32 x i8> @xvldrepl_b_offset(ptr %ptr) {
; CHECK-LABEL: xvldrepl_b_offset:
; CHECK: # %bb.0:
-; CHECK-NEXT: xvldrepl.b $xr0, $a0, 33
+; CHECK-NEXT: ld.b $a0, $a0, 33
+; CHECK-NEXT: xvreplgr2vr.b $xr0, $a0
; CHECK-NEXT: ret
%p = getelementptr i8, ptr %ptr, i64 33
%tmp = load i8, ptr %p
@@ -58,7 +86,8 @@ define <32 x i8> @xvldrepl_b_offset(ptr %ptr) {
define <16 x i16> @xvldrepl_h(ptr %ptr) {
; CHECK-LABEL: xvldrepl_h:
; CHECK: # %bb.0:
-; CHECK-NEXT: xvldrepl.h $xr0, $a0, 0
+; CHECK-NEXT: ld.h $a0, $a0, 0
+; CHECK-NEXT: xvreplgr2vr.h $xr0, $a0
; CHECK-NEXT: ret
%tmp = load i16, ptr %ptr
%tmp1 = insertelement <16 x i16> zeroinitializer, i16 %tmp, i32 0
@@ -69,7 +98,8 @@ define <16 x i16> @xvldrepl_h(ptr %ptr) {
define <16 x i16> @xvldrepl_h_offset(ptr %ptr) {
; CHECK-LABEL: xvldrepl_h_offset:
; CHECK: # %bb.0:
-; CHECK-NEXT: xvldrepl.h $xr0, $a0, 66
+; CHECK-NEXT: ld.h $a0, $a0, 66
+; CHECK-NEXT: xvreplgr2vr.h $xr0, $a0
; CHECK-NEXT: ret
%p = getelementptr i16, ptr %ptr, i64 33
%tmp = load i16, ptr %p
@@ -81,7 +111,8 @@ define <16 x i16> @xvldrepl_h_offset(ptr %ptr) {
define <8 x i32> @xvldrepl_w(ptr %ptr) {
; CHECK-LABEL: xvldrepl_w:
; CHECK: # %bb.0:
-; CHECK-NEXT: xvldrepl.w $xr0, $a0, 0
+; CHECK-NEXT: ld.w $a0, $a0, 0
+; CHECK-NEXT: xvreplgr2vr.w $xr0, $a0
; CHECK-NEXT: ret
%tmp = load i32, ptr %ptr
%tmp1 = insertelement <8 x i32> zeroinitializer, i32 %tmp, i32 0
@@ -92,7 +123,8 @@ define <8 x i32> @xvldrepl_w(ptr %ptr) {
define <8 x i32> @xvldrepl_w_offset(ptr %ptr) {
; CHECK-LABEL: xvldrepl_w_offset:
; CHECK: # %bb.0:
-; CHECK-NEXT: xvldrepl.w $xr0, $a0, 132
+; CHECK-NEXT: ld.w $a0, $a0, 132
+; CHECK-NEXT: xvreplgr2vr.w $xr0, $a0
; CHECK-NEXT: ret
%p = getelementptr i32, ptr %ptr, i64 33
%tmp = load i32, ptr %p
diff --git a/llvm/test/CodeGen/LoongArch/lsx/broadcast-load.ll b/llvm/test/CodeGen/LoongArch/lsx/broadcast-load.ll
index c46747ef30509..a8cddbf9e6400 100644
--- a/llvm/test/CodeGen/LoongArch/lsx/broadcast-load.ll
+++ b/llvm/test/CodeGen/LoongArch/lsx/broadcast-load.ll
@@ -18,6 +18,32 @@ define <2 x i64> @should_not_be_optimized(ptr %ptr, ptr %dst){
ret <2 x i64> %tmp2
}
+define <8 x i16> @should_not_be_optimized_sext_load(ptr %ptr) {
+; CHECK-LABEL: should_not_be_optimized_sext_load:
+; CHECK: # %bb.0:
+; CHECK-NEXT: ld.b $a0, $a0, 0
+; CHECK-NEXT: vreplgr2vr.h $vr0, $a0
+; CHECK-NEXT: ret
+ %tmp = load i8, ptr %ptr
+ %tmp1 = sext i8 %tmp to i16
+ %tmp2 = insertelement <8 x i16> zeroinitializer, i16 %tmp1, i32 0
+ %tmp3 = shufflevector <8 x i16> %tmp2, <8 x i16> poison, <8 x i32> zeroinitializer
+ ret <8 x i16> %tmp3
+}
+
+define <8 x i16> @should_not_be_optimized_zext_load(ptr %ptr) {
+; CHECK-LABEL: should_not_be_optimized_zext_load:
+; CHECK: # %bb.0:
+; CHECK-NEXT: ld.bu $a0, $a0, 0
+; CHECK-NEXT: vreplgr2vr.h $vr0, $a0
+; CHECK-NEXT: ret
+ %tmp = load i8, ptr %ptr
+ %tmp1 = zext i8 %tmp to i16
+ %tmp2 = insertelement <8 x i16> zeroinitializer, i16 %tmp1, i32 0
+ %tmp3 = shufflevector <8 x i16> %tmp2, <8 x i16> poison, <8 x i32> zeroinitializer
+ ret <8 x i16> %tmp3
+}
+
define <2 x i64> @vldrepl_d_unaligned_offset(ptr %ptr) {
; CHECK-LABEL: vldrepl_d_unaligned_offset:
; CHECK: # %bb.0:
@@ -34,7 +60,8 @@ define <2 x i64> @vldrepl_d_unaligned_offset(ptr %ptr) {
define <16 x i8> @vldrepl_b(ptr %ptr) {
; CHECK-LABEL: vldrepl_b:
; CHECK: # %bb.0:
-; CHECK-NEXT: vldrepl.b $vr0, $a0, 0
+; CHECK-NEXT: ld.b $a0, $a0, 0
+; CHECK-NEXT: vreplgr2vr.b $vr0, $a0
; CHECK-NEXT: ret
%tmp = load i8, ptr %ptr
%tmp1 = insertelement <16 x i8> zeroinitializer, i8 %tmp, i32 0
@@ -45,7 +72,8 @@ define <16 x i8> @vldrepl_b(ptr %ptr) {
define <16 x i8> @vldrepl_b_offset(ptr %ptr) {
; CHECK-LABEL: vldrepl_b_offset:
; CHECK: # %bb.0:
-; CHECK-NEXT: vldrepl.b $vr0, $a0, 33
+; CHECK-NEXT: ld.b $a0, $a0, 33
+; CHECK-NEXT: vreplgr2vr.b $vr0, $a0
; CHECK-NEXT: ret
%p = getelementptr i8, ptr %ptr, i64 33
%tmp = load i8, ptr %p
@@ -58,7 +86,8 @@ define <16 x i8> @vldrepl_b_offset(ptr %ptr) {
define <8 x i16> @vldrepl_h(ptr %ptr) {
; CHECK-LABEL: vldrepl_h:
; CHECK: # %bb.0:
-; CHECK-NEXT: vldrepl.h $vr0, $a0, 0
+; CHECK-NEXT: ld.h $a0, $a0, 0
+; CHECK-NEXT: vreplgr2vr.h $vr0, $a0
; CHECK-NEXT: ret
%tmp = load i16, ptr %ptr
%tmp1 = insertelement <8 x i16> zeroinitializer, i16 %tmp, i32 0
@@ -69,7 +98,8 @@ define <8 x i16> @vldrepl_h(ptr %ptr) {
define <8 x i16> @vldrepl_h_offset(ptr %ptr) {
; CHECK-LABEL: vldrepl_h_offset:
; CHECK: # %bb.0:
-; CHECK-NEXT: vldrepl.h $vr0, $a0, 66
+; CHECK-NEXT: ld.h $a0, $a0, 66
+; CHECK-NEXT: vreplgr2vr.h $vr0, $a0
; CHECK-NEXT: ret
%p = getelementptr i16, ptr %ptr, i64 33
%tmp = load i16, ptr %p
@@ -81,7 +111,8 @@ define <8 x i16> @vldrepl_h_offset(ptr %ptr) {
define <4 x i32> @vldrepl_w(ptr %ptr) {
; CHECK-LABEL: vldrepl_w:
; CHECK: # %bb.0:
-; CHECK-NEXT: vldrepl.w $vr0, $a0, 0
+; CHECK-NEXT: ld.w $a0, $a0, 0
+; CHECK-NEXT: vreplgr2vr.w $vr0, $a0
; CHECK-NEXT: ret
%tmp = load i32, ptr %ptr
%tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 0
@@ -92,7 +123,8 @@ define <4 x i32> @vldrepl_w(ptr %ptr) {
define <4 x i32> @vldrepl_w_offset(ptr %ptr) {
; CHECK-LABEL: vldrepl_w_offset:
; CHECK: # %bb.0:
-; CHECK-NEXT: vldrepl.w $vr0, $a0, 132
+; CHECK-NEXT: ld.w $a0, $a0, 132
+; CHECK-NEXT: vreplgr2vr.w $vr0, $a0
; CHECK-NEXT: ret
%p = getelementptr i32, ptr %ptr, i64 33
%tmp = load i32, ptr %p
@@ -169,3 +201,4 @@ define <2 x double> @vldrepl_d_dbl_offset(ptr %ptr) {
%tmp2 = shufflevector <2 x double> %tmp1, <2 x double> poison, <2 x i32> zeroinitializer
ret <2 x double> %tmp2
}
+
``````````
</details>
https://github.com/llvm/llvm-project/pull/155960
More information about the llvm-commits
mailing list