[LoongArch][NFC] Pre-commit tests for LSX/LASX bswap codegen

[xen0n]

No description provided.

[llvmbot]

@llvm/pr-subscribers-backend-loongarch

Author: WÁNG Xuěruì (xen0n)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/114170.diff

2 Files Affected:

• (added) llvm/test/CodeGen/LoongArch/lasx/bswap.ll (+86)
• (added) llvm/test/CodeGen/LoongArch/lsx/bswap.ll (+86)

diff --git a/llvm/test/CodeGen/LoongArch/lasx/bswap.ll b/llvm/test/CodeGen/LoongArch/lasx/bswap.ll
new file mode 100644
index 00000000000000..4f6d49c7a79db5
--- /dev/null
+++ b/llvm/test/CodeGen/LoongArch/lasx/bswap.ll
@@ -0,0 +1,86 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc --mtriple=loongarch64 --mattr=+lasx < %s | FileCheck %s
+
+define void @bswap_v16i16(ptr %src, ptr %dst) nounwind {
+; CHECK-LABEL: bswap_v16i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a0, 0
+; CHECK-NEXT:    xvsrli.h $xr1, $xr0, 8
+; CHECK-NEXT:    xvslli.h $xr0, $xr0, 8
+; CHECK-NEXT:    xvor.v $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvst $xr0, $a1, 0
+; CHECK-NEXT:    ret
+  %v = load <16 x i16>, ptr %src
+  %res = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> %v)
+  store <16 x i16> %res, ptr %dst
+  ret void
+}
+
+define void @bswap_v8i32(ptr %src, ptr %dst) nounwind {
+; CHECK-LABEL: bswap_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a0, 0
+; CHECK-NEXT:    lu12i.w $a0, 15
+; CHECK-NEXT:    ori $a0, $a0, 3840
+; CHECK-NEXT:    xvreplgr2vr.w $xr1, $a0
+; CHECK-NEXT:    xvsrli.w $xr2, $xr0, 8
+; CHECK-NEXT:    xvand.v $xr2, $xr2, $xr1
+; CHECK-NEXT:    xvsrli.w $xr3, $xr0, 24
+; CHECK-NEXT:    xvor.v $xr2, $xr2, $xr3
+; CHECK-NEXT:    xvand.v $xr1, $xr0, $xr1
+; CHECK-NEXT:    xvslli.w $xr1, $xr1, 8
+; CHECK-NEXT:    xvslli.w $xr0, $xr0, 24
+; CHECK-NEXT:    xvor.v $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvor.v $xr0, $xr0, $xr2
+; CHECK-NEXT:    xvst $xr0, $a1, 0
+; CHECK-NEXT:    ret
+  %v = load <8 x i32>, ptr %src
+  %res = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> %v)
+  store <8 x i32> %res, ptr %dst
+  ret void
+}
+
+define void @bswap_v4i64(ptr %src, ptr %dst) nounwind {
+; CHECK-LABEL: bswap_v4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvld $xr0, $a0, 0
+; CHECK-NEXT:    lu12i.w $a0, 4080
+; CHECK-NEXT:    xvreplgr2vr.d $xr1, $a0
+; CHECK-NEXT:    xvsrli.d $xr2, $xr0, 24
+; CHECK-NEXT:    xvand.v $xr2, $xr2, $xr1
+; CHECK-NEXT:    lu12i.w $a0, -4096
+; CHECK-NEXT:    lu32i.d $a0, 0
+; CHECK-NEXT:    xvreplgr2vr.d $xr3, $a0
+; CHECK-NEXT:    xvsrli.d $xr4, $xr0, 8
+; CHECK-NEXT:    xvand.v $xr4, $xr4, $xr3
+; CHECK-NEXT:    xvor.v $xr2, $xr4, $xr2
+; CHECK-NEXT:    lu12i.w $a0, 15
+; CHECK-NEXT:    ori $a0, $a0, 3840
+; CHECK-NEXT:    xvreplgr2vr.d $xr4, $a0
+; CHECK-NEXT:    xvsrli.d $xr5, $xr0, 40
+; CHECK-NEXT:    xvand.v $xr5, $xr5, $xr4
+; CHECK-NEXT:    xvsrli.d $xr6, $xr0, 56
+; CHECK-NEXT:    xvor.v $xr5, $xr5, $xr6
+; CHECK-NEXT:    xvor.v $xr2, $xr2, $xr5
+; CHECK-NEXT:    xvand.v $xr1, $xr0, $xr1
+; CHECK-NEXT:    xvslli.d $xr1, $xr1, 24
+; CHECK-NEXT:    xvand.v $xr3, $xr0, $xr3
+; CHECK-NEXT:    xvslli.d $xr3, $xr3, 8
+; CHECK-NEXT:    xvor.v $xr1, $xr1, $xr3
+; CHECK-NEXT:    xvand.v $xr3, $xr0, $xr4
+; CHECK-NEXT:    xvslli.d $xr3, $xr3, 40
+; CHECK-NEXT:    xvslli.d $xr0, $xr0, 56
+; CHECK-NEXT:    xvor.v $xr0, $xr0, $xr3
+; CHECK-NEXT:    xvor.v $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvor.v $xr0, $xr0, $xr2
+; CHECK-NEXT:    xvst $xr0, $a1, 0
+; CHECK-NEXT:    ret
+  %v = load <4 x i64>, ptr %src
+  %res = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> %v)
+  store <4 x i64> %res, ptr %dst
+  ret void
+}
+
+declare <16 x i16> @llvm.bswap.v16i16(<16 x i16>)
+declare <8 x i32> @llvm.bswap.v8i32(<8 x i32>)
+declare <4 x i64> @llvm.bswap.v4i64(<4 x i64>)
diff --git a/llvm/test/CodeGen/LoongArch/lsx/bswap.ll b/llvm/test/CodeGen/LoongArch/lsx/bswap.ll
new file mode 100644
index 00000000000000..ce7af9d33f1501
--- /dev/null
+++ b/llvm/test/CodeGen/LoongArch/lsx/bswap.ll
@@ -0,0 +1,86 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s
+
+define void @bswap_v8i16(ptr %src, ptr %dst) nounwind {
+; CHECK-LABEL: bswap_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vld $vr0, $a0, 0
+; CHECK-NEXT:    vsrli.h $vr1, $vr0, 8
+; CHECK-NEXT:    vslli.h $vr0, $vr0, 8
+; CHECK-NEXT:    vor.v $vr0, $vr0, $vr1
+; CHECK-NEXT:    vst $vr0, $a1, 0
+; CHECK-NEXT:    ret
+  %v = load <8 x i16>, ptr %src
+  %res = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> %v)
+  store <8 x i16> %res, ptr %dst
+  ret void
+}
+
+define void @bswap_v4i32(ptr %src, ptr %dst) nounwind {
+; CHECK-LABEL: bswap_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vld $vr0, $a0, 0
+; CHECK-NEXT:    lu12i.w $a0, 15
+; CHECK-NEXT:    ori $a0, $a0, 3840
+; CHECK-NEXT:    vreplgr2vr.w $vr1, $a0
+; CHECK-NEXT:    vsrli.w $vr2, $vr0, 8
+; CHECK-NEXT:    vand.v $vr2, $vr2, $vr1
+; CHECK-NEXT:    vsrli.w $vr3, $vr0, 24
+; CHECK-NEXT:    vor.v $vr2, $vr2, $vr3
+; CHECK-NEXT:    vand.v $vr1, $vr0, $vr1
+; CHECK-NEXT:    vslli.w $vr1, $vr1, 8
+; CHECK-NEXT:    vslli.w $vr0, $vr0, 24
+; CHECK-NEXT:    vor.v $vr0, $vr0, $vr1
+; CHECK-NEXT:    vor.v $vr0, $vr0, $vr2
+; CHECK-NEXT:    vst $vr0, $a1, 0
+; CHECK-NEXT:    ret
+  %v = load <4 x i32>, ptr %src
+  %res = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %v)
+  store <4 x i32> %res, ptr %dst
+  ret void
+}
+
+define void @bswap_v2i64(ptr %src, ptr %dst) nounwind {
+; CHECK-LABEL: bswap_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vld $vr0, $a0, 0
+; CHECK-NEXT:    lu12i.w $a0, 4080
+; CHECK-NEXT:    vreplgr2vr.d $vr1, $a0
+; CHECK-NEXT:    vsrli.d $vr2, $vr0, 24
+; CHECK-NEXT:    vand.v $vr2, $vr2, $vr1
+; CHECK-NEXT:    lu12i.w $a0, -4096
+; CHECK-NEXT:    lu32i.d $a0, 0
+; CHECK-NEXT:    vreplgr2vr.d $vr3, $a0
+; CHECK-NEXT:    vsrli.d $vr4, $vr0, 8
+; CHECK-NEXT:    vand.v $vr4, $vr4, $vr3
+; CHECK-NEXT:    vor.v $vr2, $vr4, $vr2
+; CHECK-NEXT:    lu12i.w $a0, 15
+; CHECK-NEXT:    ori $a0, $a0, 3840
+; CHECK-NEXT:    vreplgr2vr.d $vr4, $a0
+; CHECK-NEXT:    vsrli.d $vr5, $vr0, 40
+; CHECK-NEXT:    vand.v $vr5, $vr5, $vr4
+; CHECK-NEXT:    vsrli.d $vr6, $vr0, 56
+; CHECK-NEXT:    vor.v $vr5, $vr5, $vr6
+; CHECK-NEXT:    vor.v $vr2, $vr2, $vr5
+; CHECK-NEXT:    vand.v $vr1, $vr0, $vr1
+; CHECK-NEXT:    vslli.d $vr1, $vr1, 24
+; CHECK-NEXT:    vand.v $vr3, $vr0, $vr3
+; CHECK-NEXT:    vslli.d $vr3, $vr3, 8
+; CHECK-NEXT:    vor.v $vr1, $vr1, $vr3
+; CHECK-NEXT:    vand.v $vr3, $vr0, $vr4
+; CHECK-NEXT:    vslli.d $vr3, $vr3, 40
+; CHECK-NEXT:    vslli.d $vr0, $vr0, 56
+; CHECK-NEXT:    vor.v $vr0, $vr0, $vr3
+; CHECK-NEXT:    vor.v $vr0, $vr0, $vr1
+; CHECK-NEXT:    vor.v $vr0, $vr0, $vr2
+; CHECK-NEXT:    vst $vr0, $a1, 0
+; CHECK-NEXT:    ret
+  %v = load <2 x i64>, ptr %src
+  %res = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %v)
+  store <2 x i64> %res, ptr %dst
+  ret void
+}
+
+declare <8 x i16> @llvm.bswap.v8i16(<8 x i16>)
+declare <4 x i32> @llvm.bswap.v4i32(<4 x i32>)
+declare <2 x i64> @llvm.bswap.v2i64(<2 x i64>)

[heiher]

LGTM. Thanks.