Skip to content
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

[LoongArch] Pre-commit tests for absolute difference #132898

Open
wants to merge 6 commits into
base: main
Choose a base branch
from
Open

[LoongArch] Pre-commit tests for absolute difference #132898

wants to merge 6 commits into from
+1,114 −0

Conversation

tangaac
Copy link
Contributor

@tangaac tangaac commented Mar 25, 2025

No description provided.

@tangaac tangaac requested review from heiher and SixWeining March 25, 2025 08:57
@llvmbot
Copy link
Member

llvmbot commented Mar 25, 2025

@llvm/pr-subscribers-backend-loongarch

Author: None (tangaac)

Changes

Patch is 23.56 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/132898.diff

2 Files Affected:

  • (added) llvm/test/CodeGen/LoongArch/lasx/ir-instruction/absd.ll (+326)
  • (added) llvm/test/CodeGen/LoongArch/lsx/ir-instruction/absd.ll (+326)
diff --git a/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/absd.ll b/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/absd.ll
new file mode 100644
index 0000000000000..42152bc7f5fcf
--- /dev/null
+++ b/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/absd.ll
@@ -0,0 +1,326 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=loongarch64 -mattr=+lasx < %s | FileCheck %s
+
+;; Mostly copy from AArch64/neon-abd.ll
+
+;
+; VABDS_[B/H/W/D]
+;
+define <32 x i8> @xvabsd_b(<32 x i8> %a, <32 x i8> %b) #0 {
+; CHECK-LABEL: xvabsd_b:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.b $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.b $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.b $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a.sext = sext <32 x i8> %a to <32 x i16>
+  %b.sext = sext <32 x i8> %b to <32 x i16>
+  %sub = sub <32 x i16> %a.sext, %b.sext
+  %abs = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %sub, i1 true)
+  %trunc = trunc <32 x i16> %abs to <32 x i8>
+  ret <32 x i8> %trunc
+}
+
+define <16 x i16> @xvabsd_h(<16 x i16> %a, <16 x i16> %b) #0 {
+; CHECK-LABEL: xvabsd_h:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.h $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.h $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.h $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a.sext = sext <16 x i16> %a to <16 x i32>
+  %b.sext = sext <16 x i16> %b to <16 x i32>
+  %sub = sub <16 x i32> %a.sext, %b.sext
+  %abs = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %sub, i1 true)
+  %trunc = trunc <16 x i32> %abs to <16 x i16>
+  ret <16 x i16> %trunc
+}
+
+define <8 x i32> @xvabsd_w(<8 x i32> %a, <8 x i32> %b) #0 {
+; CHECK-LABEL: xvabsd_w:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.w $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.w $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.w $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a.sext = sext <8 x i32> %a to <8 x i64>
+  %b.sext = sext <8 x i32> %b to <8 x i64>
+  %sub = sub <8 x i64> %a.sext, %b.sext
+  %abs = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %sub, i1 true)
+  %trunc = trunc <8 x i64> %abs to <8 x i32>
+  ret <8 x i32> %trunc
+}
+
+define <4 x i64> @xvabsd_d(<4 x i64> %a, <4 x i64> %b) #0 {
+; CHECK-LABEL: xvabsd_d:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.d $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.d $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.d $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a.sext = sext <4 x i64> %a to <4 x i128>
+  %b.sext = sext <4 x i64> %b to <4 x i128>
+  %sub = sub <4 x i128> %a.sext, %b.sext
+  %abs = call <4 x i128> @llvm.abs.v4i128(<4 x i128> %sub, i1 true)
+  %trunc = trunc <4 x i128> %abs to <4 x i64>
+  ret <4 x i64> %trunc
+}
+
+;
+; xvabsd_[B/H/W/D]U
+;
+
+define <32 x i8> @xvabsd_bu(<32 x i8> %a, <32 x i8> %b) #0 {
+; CHECK-LABEL: xvabsd_bu:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.bu $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.bu $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.b $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a.zext = zext <32 x i8> %a to <32 x i16>
+  %b.zext = zext <32 x i8> %b to <32 x i16>
+  %sub = sub <32 x i16> %a.zext, %b.zext
+  %abs = call <32 x i16> @llvm.abs.v32i16(<32 x i16> %sub, i1 true)
+  %trunc = trunc <32 x i16> %abs to <32 x i8>
+  ret <32 x i8> %trunc
+}
+
+define <16 x i16> @xvabsd_hu(<16 x i16> %a, <16 x i16> %b) #0 {
+; CHECK-LABEL: xvabsd_hu:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.hu $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.hu $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.h $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a.zext = zext <16 x i16> %a to <16 x i32>
+  %b.zext = zext <16 x i16> %b to <16 x i32>
+  %sub = sub <16 x i32> %a.zext, %b.zext
+  %abs = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %sub, i1 true)
+  %trunc = trunc <16 x i32> %abs to <16 x i16>
+  ret <16 x i16> %trunc
+}
+
+define <8 x i32> @xvabsd_wu(<8 x i32> %a, <8 x i32> %b) #0 {
+; CHECK-LABEL: xvabsd_wu:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.wu $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.wu $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.w $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a.zext = zext <8 x i32> %a to <8 x i64>
+  %b.zext = zext <8 x i32> %b to <8 x i64>
+  %sub = sub <8 x i64> %a.zext, %b.zext
+  %abs = call <8 x i64> @llvm.abs.v8i64(<8 x i64> %sub, i1 true)
+  %trunc = trunc <8 x i64> %abs to <8 x i32>
+  ret <8 x i32> %trunc
+}
+
+define <4 x i64> @xvabsd_du(<4 x i64> %a, <4 x i64> %b) #0 {
+; CHECK-LABEL: xvabsd_du:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.du $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.du $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.d $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a.zext = zext <4 x i64> %a to <4 x i128>
+  %b.zext = zext <4 x i64> %b to <4 x i128>
+  %sub = sub <4 x i128> %a.zext, %b.zext
+  %abs = call <4 x i128> @llvm.abs.v4i128(<4 x i128> %sub, i1 true)
+  %trunc = trunc <4 x i128> %abs to <4 x i64>
+  ret <4 x i64> %trunc
+}
+
+define <32 x i8> @xvabsd_v32i8_nsw(<32 x i8> %a, <32 x i8> %b) #0 {
+; CHECK-LABEL: xvabsd_v32i8_nsw:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvsub.b $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvneg.b $xr1, $xr0
+; CHECK-NEXT:    xvmax.b $xr0, $xr0, $xr1
+; CHECK-NEXT:    ret
+  %sub = sub nsw <32 x i8> %a, %b
+  %abs = call <32 x i8> @llvm.abs.v32i8(<32 x i8> %sub, i1 true)
+  ret <32 x i8> %abs
+}
+
+define <16 x i16> @xvabsd_v16i16_nsw(<16 x i16> %a, <16 x i16> %b) #0 {
+; CHECK-LABEL: xvabsd_v16i16_nsw:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvsub.h $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvneg.h $xr1, $xr0
+; CHECK-NEXT:    xvmax.h $xr0, $xr0, $xr1
+; CHECK-NEXT:    ret
+  %sub = sub nsw <16 x i16> %a, %b
+  %abs = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %sub, i1 true)
+  ret <16 x i16> %abs
+}
+
+define <8 x i32> @xvabsd_v8i32_nsw(<8 x i32> %a, <8 x i32> %b) #0 {
+; CHECK-LABEL: xvabsd_v8i32_nsw:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvsub.w $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvneg.w $xr1, $xr0
+; CHECK-NEXT:    xvmax.w $xr0, $xr0, $xr1
+; CHECK-NEXT:    ret
+  %sub = sub nsw <8 x i32> %a, %b
+  %abs = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %sub, i1 true)
+  ret <8 x i32> %abs
+}
+
+define <4 x i64> @xvabsd_v4i64_nsw(<4 x i64> %a, <4 x i64> %b) #0 {
+; CHECK-LABEL: xvabsd_v4i64_nsw:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvsub.d $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvneg.d $xr1, $xr0
+; CHECK-NEXT:    xvmax.d $xr0, $xr0, $xr1
+; CHECK-NEXT:    ret
+  %sub = sub nsw <4 x i64> %a, %b
+  %abs = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %sub, i1 true)
+  ret <4 x i64> %abs
+}
+
+define <32 x i8> @smaxmin_v32i8(<32 x i8> %0, <32 x i8> %1) {
+; CHECK-LABEL: smaxmin_v32i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.b $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.b $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.b $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a = tail call <32 x i8> @llvm.smax.v32i8(<32 x i8> %0, <32 x i8> %1)
+  %b = tail call <32 x i8> @llvm.smin.v32i8(<32 x i8> %0, <32 x i8> %1)
+  %sub = sub <32 x i8> %a, %b
+  ret <32 x i8> %sub
+}
+
+define <16 x i16> @smaxmin_v16i16(<16 x i16> %0, <16 x i16> %1) {
+; CHECK-LABEL: smaxmin_v16i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.h $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.h $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.h $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a = tail call <16 x i16> @llvm.smax.v16i16(<16 x i16> %0, <16 x i16> %1)
+  %b = tail call <16 x i16> @llvm.smin.v16i16(<16 x i16> %0, <16 x i16> %1)
+  %sub = sub <16 x i16> %a, %b
+  ret <16 x i16> %sub
+}
+
+define <8 x i32> @smaxmin_v8i32(<8 x i32> %0, <8 x i32> %1) {
+; CHECK-LABEL: smaxmin_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.w $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.w $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.w $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a = tail call <8 x i32> @llvm.smax.v8i32(<8 x i32> %0, <8 x i32> %1)
+  %b = tail call <8 x i32> @llvm.smin.v8i32(<8 x i32> %0, <8 x i32> %1)
+  %sub = sub <8 x i32> %a, %b
+  ret <8 x i32> %sub
+}
+
+define <4 x i64> @smaxmin_v4i64(<4 x i64> %0, <4 x i64> %1) {
+; CHECK-LABEL: smaxmin_v4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.d $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.d $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.d $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a = tail call <4 x i64> @llvm.smax.v4i64(<4 x i64> %0, <4 x i64> %1)
+  %b = tail call <4 x i64> @llvm.smin.v4i64(<4 x i64> %0, <4 x i64> %1)
+  %sub = sub <4 x i64> %a, %b
+  ret <4 x i64> %sub
+}
+
+define <32 x i8> @umaxmin_v32i8(<32 x i8> %0, <32 x i8> %1) {
+; CHECK-LABEL: umaxmin_v32i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.bu $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.bu $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.b $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a = tail call <32 x i8> @llvm.umax.v32i8(<32 x i8> %0, <32 x i8> %1)
+  %b = tail call <32 x i8> @llvm.umin.v32i8(<32 x i8> %0, <32 x i8> %1)
+  %sub = sub <32 x i8> %a, %b
+  ret <32 x i8> %sub
+}
+
+define <16 x i16> @umaxmin_v16i16(<16 x i16> %0, <16 x i16> %1) {
+; CHECK-LABEL: umaxmin_v16i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.hu $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.hu $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.h $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a = tail call <16 x i16> @llvm.umax.v16i16(<16 x i16> %0, <16 x i16> %1)
+  %b = tail call <16 x i16> @llvm.umin.v16i16(<16 x i16> %0, <16 x i16> %1)
+  %sub = sub <16 x i16> %a, %b
+  ret <16 x i16> %sub
+}
+
+define <8 x i32> @umaxmin_v8i32(<8 x i32> %0, <8 x i32> %1) {
+; CHECK-LABEL: umaxmin_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.wu $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.wu $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.w $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a = tail call <8 x i32> @llvm.umax.v8i32(<8 x i32> %0, <8 x i32> %1)
+  %b = tail call <8 x i32> @llvm.umin.v8i32(<8 x i32> %0, <8 x i32> %1)
+  %sub = sub <8 x i32> %a, %b
+  ret <8 x i32> %sub
+}
+
+define <4 x i64> @umaxmin_v4i64(<4 x i64> %0, <4 x i64> %1) {
+; CHECK-LABEL: umaxmin_v4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.du $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.du $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.d $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a = tail call <4 x i64> @llvm.umax.v4i64(<4 x i64> %0, <4 x i64> %1)
+  %b = tail call <4 x i64> @llvm.umin.v4i64(<4 x i64> %0, <4 x i64> %1)
+  %sub = sub <4 x i64> %a, %b
+  ret <4 x i64> %sub
+}
+
+define <32 x i8> @umaxmin_v32i8_com1(<32 x i8> %0, <32 x i8> %1) {
+; CHECK-LABEL: umaxmin_v32i8_com1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmin.bu $xr2, $xr0, $xr1
+; CHECK-NEXT:    xvmax.bu $xr0, $xr0, $xr1
+; CHECK-NEXT:    xvsub.b $xr0, $xr0, $xr2
+; CHECK-NEXT:    ret
+  %a = tail call <32 x i8> @llvm.umax.v32i8(<32 x i8> %0, <32 x i8> %1)
+  %b = tail call <32 x i8> @llvm.umin.v32i8(<32 x i8> %1, <32 x i8> %0)
+  %sub = sub <32 x i8> %a, %b
+  ret <32 x i8> %sub
+}
+
+declare <32 x i8> @llvm.abs.v32i8(<32 x i8>, i1)
+
+declare <16 x i16> @llvm.abs.v16i16(<16 x i16>, i1)
+declare <32 x i16> @llvm.abs.v32i16(<32 x i16>, i1)
+
+declare <8 x i32> @llvm.abs.v8i32(<8 x i32>, i1)
+declare <16 x i32> @llvm.abs.v16i32(<16 x i32>, i1)
+
+declare <4 x i64> @llvm.abs.v4i64(<4 x i64>, i1)
+declare <8 x i64> @llvm.abs.v8i64(<8 x i64>, i1)
+
+declare <4 x i128> @llvm.abs.v4i128(<4 x i128>, i1)
+
+declare <32 x i8> @llvm.smax.v32i8(<32 x i8>, <32 x i8>)
+declare <16 x i16> @llvm.smax.v16i16(<16 x i16>, <16 x i16>)
+declare <8 x i32> @llvm.smax.v8i32(<8 x i32>, <8 x i32>)
+declare <4 x i64> @llvm.smax.v4i64(<4 x i64>, <4 x i64>)
+declare <32 x i8> @llvm.smin.v32i8(<32 x i8>, <32 x i8>)
+declare <16 x i16> @llvm.smin.v16i16(<16 x i16>, <16 x i16>)
+declare <8 x i32> @llvm.smin.v8i32(<8 x i32>, <8 x i32>)
+declare <4 x i64> @llvm.smin.v4i64(<4 x i64>, <4 x i64>)
+declare <32 x i8> @llvm.umax.v32i8(<32 x i8>, <32 x i8>)
+declare <16 x i16> @llvm.umax.v16i16(<16 x i16>, <16 x i16>)
+declare <8 x i32> @llvm.umax.v8i32(<8 x i32>, <8 x i32>)
+declare <4 x i64> @llvm.umax.v4i64(<4 x i64>, <4 x i64>)
+declare <32 x i8> @llvm.umin.v32i8(<32 x i8>, <32 x i8>)
+declare <16 x i16> @llvm.umin.v16i16(<16 x i16>, <16 x i16>)
+declare <8 x i32> @llvm.umin.v8i32(<8 x i32>, <8 x i32>)
+declare <4 x i64> @llvm.umin.v4i64(<4 x i64>, <4 x i64>)
diff --git a/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/absd.ll b/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/absd.ll
new file mode 100644
index 0000000000000..3ecfaed8e1505
--- /dev/null
+++ b/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/absd.ll
@@ -0,0 +1,326 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=loongarch64 -mattr=+lsx < %s | FileCheck %s
+
+;; Mostly copy from AArch64/neon-abd.ll
+
+;
+; VABDS_[B/H/W/D]
+;
+define <16 x i8> @vabsd_b(<16 x i8> %a, <16 x i8> %b) #0 {
+; CHECK-LABEL: vabsd_b:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmin.b $vr2, $vr0, $vr1
+; CHECK-NEXT:    vmax.b $vr0, $vr0, $vr1
+; CHECK-NEXT:    vsub.b $vr0, $vr0, $vr2
+; CHECK-NEXT:    ret
+  %a.sext = sext <16 x i8> %a to <16 x i16>
+  %b.sext = sext <16 x i8> %b to <16 x i16>
+  %sub = sub <16 x i16> %a.sext, %b.sext
+  %abs = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %sub, i1 true)
+  %trunc = trunc <16 x i16> %abs to <16 x i8>
+  ret <16 x i8> %trunc
+}
+
+define <8 x i16> @vabsd_h(<8 x i16> %a, <8 x i16> %b) #0 {
+; CHECK-LABEL: vabsd_h:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmin.h $vr2, $vr0, $vr1
+; CHECK-NEXT:    vmax.h $vr0, $vr0, $vr1
+; CHECK-NEXT:    vsub.h $vr0, $vr0, $vr2
+; CHECK-NEXT:    ret
+  %a.sext = sext <8 x i16> %a to <8 x i32>
+  %b.sext = sext <8 x i16> %b to <8 x i32>
+  %sub = sub <8 x i32> %a.sext, %b.sext
+  %abs = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %sub, i1 true)
+  %trunc = trunc <8 x i32> %abs to <8 x i16>
+  ret <8 x i16> %trunc
+}
+
+define <4 x i32> @vabsd_w(<4 x i32> %a, <4 x i32> %b) #0 {
+; CHECK-LABEL: vabsd_w:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmin.w $vr2, $vr0, $vr1
+; CHECK-NEXT:    vmax.w $vr0, $vr0, $vr1
+; CHECK-NEXT:    vsub.w $vr0, $vr0, $vr2
+; CHECK-NEXT:    ret
+  %a.sext = sext <4 x i32> %a to <4 x i64>
+  %b.sext = sext <4 x i32> %b to <4 x i64>
+  %sub = sub <4 x i64> %a.sext, %b.sext
+  %abs = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %sub, i1 true)
+  %trunc = trunc <4 x i64> %abs to <4 x i32>
+  ret <4 x i32> %trunc
+}
+
+define <2 x i64> @vabsd_d(<2 x i64> %a, <2 x i64> %b) #0 {
+; CHECK-LABEL: vabsd_d:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmin.d $vr2, $vr0, $vr1
+; CHECK-NEXT:    vmax.d $vr0, $vr0, $vr1
+; CHECK-NEXT:    vsub.d $vr0, $vr0, $vr2
+; CHECK-NEXT:    ret
+  %a.sext = sext <2 x i64> %a to <2 x i128>
+  %b.sext = sext <2 x i64> %b to <2 x i128>
+  %sub = sub <2 x i128> %a.sext, %b.sext
+  %abs = call <2 x i128> @llvm.abs.v2i128(<2 x i128> %sub, i1 true)
+  %trunc = trunc <2 x i128> %abs to <2 x i64>
+  ret <2 x i64> %trunc
+}
+
+;
+; VABSD_[B/H/W/D]U
+;
+
+define <16 x i8> @vabsd_bu(<16 x i8> %a, <16 x i8> %b) #0 {
+; CHECK-LABEL: vabsd_bu:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmin.bu $vr2, $vr0, $vr1
+; CHECK-NEXT:    vmax.bu $vr0, $vr0, $vr1
+; CHECK-NEXT:    vsub.b $vr0, $vr0, $vr2
+; CHECK-NEXT:    ret
+  %a.zext = zext <16 x i8> %a to <16 x i16>
+  %b.zext = zext <16 x i8> %b to <16 x i16>
+  %sub = sub <16 x i16> %a.zext, %b.zext
+  %abs = call <16 x i16> @llvm.abs.v16i16(<16 x i16> %sub, i1 true)
+  %trunc = trunc <16 x i16> %abs to <16 x i8>
+  ret <16 x i8> %trunc
+}
+
+define <8 x i16> @vabsd_hu(<8 x i16> %a, <8 x i16> %b) #0 {
+; CHECK-LABEL: vabsd_hu:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmin.hu $vr2, $vr0, $vr1
+; CHECK-NEXT:    vmax.hu $vr0, $vr0, $vr1
+; CHECK-NEXT:    vsub.h $vr0, $vr0, $vr2
+; CHECK-NEXT:    ret
+  %a.zext = zext <8 x i16> %a to <8 x i32>
+  %b.zext = zext <8 x i16> %b to <8 x i32>
+  %sub = sub <8 x i32> %a.zext, %b.zext
+  %abs = call <8 x i32> @llvm.abs.v8i32(<8 x i32> %sub, i1 true)
+  %trunc = trunc <8 x i32> %abs to <8 x i16>
+  ret <8 x i16> %trunc
+}
+
+define <4 x i32> @vabsd_wu(<4 x i32> %a, <4 x i32> %b) #0 {
+; CHECK-LABEL: vabsd_wu:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmin.wu $vr2, $vr0, $vr1
+; CHECK-NEXT:    vmax.wu $vr0, $vr0, $vr1
+; CHECK-NEXT:    vsub.w $vr0, $vr0, $vr2
+; CHECK-NEXT:    ret
+  %a.zext = zext <4 x i32> %a to <4 x i64>
+  %b.zext = zext <4 x i32> %b to <4 x i64>
+  %sub = sub <4 x i64> %a.zext, %b.zext
+  %abs = call <4 x i64> @llvm.abs.v4i64(<4 x i64> %sub, i1 true)
+  %trunc = trunc <4 x i64> %abs to <4 x i32>
+  ret <4 x i32> %trunc
+}
+
+define <2 x i64> @vabsd_du(<2 x i64> %a, <2 x i64> %b) #0 {
+; CHECK-LABEL: vabsd_du:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmin.du $vr2, $vr0, $vr1
+; CHECK-NEXT:    vmax.du $vr0, $vr0, $vr1
+; CHECK-NEXT:    vsub.d $vr0, $vr0, $vr2
+; CHECK-NEXT:    ret
+  %a.zext = zext <2 x i64> %a to <2 x i128>
+  %b.zext = zext <2 x i64> %b to <2 x i128>
+  %sub = sub <2 x i128> %a.zext, %b.zext
+  %abs = call <2 x i128> @llvm.abs.v2i128(<2 x i128> %sub, i1 true)
+  %trunc = trunc <2 x i128> %abs to <2 x i64>
+  ret <2 x i64> %trunc
+}
+
+define <16 x i8> @vabsd_v16i8_nsw(<16 x i8> %a, <16 x i8> %b) #0 {
+; CHECK-LABEL: vabsd_v16i8_nsw:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsub.b $vr0, $vr0, $vr1
+; CHECK-NEXT:    vneg.b $vr1, $vr0
+; CHECK-NEXT:    vmax.b $vr0, $vr0, $vr1
+; CHECK-NEXT:    ret
+  %sub = sub nsw <16 x i8> %a, %b
+  %abs = call <16 x i8> @llvm.abs.v16i8(<16 x i8> %sub, i1 true)
+  ret <16 x i8> %abs
+}
+
+define <8 x i16> @vabsd_v8i16_nsw(<8 x i16> %a, <8 x i16> %b) #0 {
+; CHECK-LABEL: vabsd_v8i16_nsw:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsub.h $vr0, $vr0, $vr1
+; CHECK-NEXT:    vneg.h $vr1, $vr0
+; CHECK-NEXT:    vmax.h $vr0, $vr0, $vr1
+; CHECK-NEXT:    ret
+  %sub = sub nsw <8 x i16> %a, %b
+  %abs = call <8 x i16> @llvm.abs.v8i16(<8 x i16> %sub, i1 true)
+  ret <8 x i16> %abs
+}
+
+define <4 x i32> @vabsd_v4i32_nsw(<4 x i32> %a, <4 x i32> %b) #0 {
+; CHECK-LABEL: vabsd_v4i32_nsw:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsub.w $vr0, $vr0, $vr1
+; CHECK-NEXT:    vneg.w $vr1, $vr0
+; CHECK-NEXT:    vmax.w $vr0, $vr0, $vr1
+; CHECK-NEXT:    ret
+  %sub = sub nsw <4 x i32> %a, %b
+  %abs = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %sub, i1 true)
+  ret <4 x i32> %abs
+}
+
+define <2 x i64> @vabsd_v2i64_nsw(<2 x i64> %a, <2 x i64> %b) #0 {
+; CHECK-LABEL: vabsd_v2i64_nsw:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsub.d $vr0, $vr0, $vr1
+; CHECK-NEXT:    vneg.d $vr1, $vr0
+; CHECK-NEXT:    vmax.d $vr0, $vr0, $vr1
+; CHECK-NEXT:    ret
+  %sub = sub nsw <2 x i64> %a, %b
+  %abs = call <2 x i64> @llvm.abs.v2i64(<2 x i64> %sub, i1 true)
+  ret <2 x i64> %abs
+}
+
+define <16 x i8> @smaxmin_v16i8(<16 x i8> %0, <16 x i8> %1) {
+; CHECK-LABEL: smaxmin_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmin.b $vr2, $vr0, $vr1
+; CHECK-NEXT:    vmax.b $vr0, $vr0, $vr1
+; CHECK-NEXT:    vsub.b $vr0, $vr0, $vr2
+; CHECK-NEXT:    ret
+  %a = tail call <16 x i8> @llvm.smax.v16i8(<16 x i8> %0, <16 x i8> %1)
+  %b = tail call <16 x i8> @llvm.smin.v16i8(<16 x i8> %0, <16 x i8> %1)
+  %sub = sub <16 x i8> %a, %b
+  ret <16 x i8> %sub
+}
+
+define <8 x i16> @smaxmin_v8i16(<8 x i16> %0, <8 x i16> %1) {
+; CHECK-LABEL: smaxmin_v8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmin.h $vr2, $vr0, $vr1
+; CHECK-NEXT:    vmax.h $vr0, $vr0, $vr1
+; CHECK-NEXT:    vsub.h $vr0, $vr0, $vr2
+; CHECK-NEXT:    ret
+  %a = tail call <8 x i16> @llvm.smax.v8i16(<8 x i16> %0, <8 x i16> %1)
+  %b = tail call <8 x i16> @llvm.smin.v8i16(<8 x i16> %0, <8 x i16> %1)
+  %sub = sub <8 x i16> %a, %b
+  ret <8 x i16> %sub
+}
+
+define <4 x i32> @smaxmin_v4i32(<4 x i32> %0, <4 x i32> %1) {
+; CHECK-LABEL: smaxmin_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmin.w $vr2, $vr0, $vr1
+; CHECK-NEXT:    vmax.w $vr0, $vr0, $vr1
+; CHECK-NEXT:    vsub.w $vr0, $vr0, $vr2
+; CHECK-NEXT:    ret
+  %a = tail call <4 x i32> @llvm.smax.v4i32(<4 x i32> %0, <4 x i32> %1)
+  %b = tail call <4 x i32> @llvm.smin.v4i32(<4 x i32> %0, <4 x i32> %1)
+  %sub = sub <4 x i32> %a, %b
+  ret <4 x i32> %sub
+}
+
+define <2 x i64> @smaxmin_v2i64(<2 x i64> %0, <2 x i64> %1) {
+; CHECK-LABEL: smaxmin_v2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmin.d $vr2, $vr0, $vr1
+; CHECK-NEXT:    vmax.d $vr0, $vr0, $vr1
+; CHECK-NEXT:    vsub.d $vr0, $vr0, $vr2
+; CHECK-NEXT:    ret
+  %a = tail call <2 x i64> @llvm.smax.v2i64(<2 x i64> %0, <2 x i64> %1)
+  %b = tail call <2 x i64> @llvm.smin.v2i64(<2 x i64> %0, <2 x i64> %1)
+  %sub = sub <2 x i64> %a, %b
+  ret <2 x i64> %sub
+}
+
+define <16 x i8> @umaxmin_v16i8(<16 x i8> %0, <16 x i8> %1) {
+; CHECK-LAB...
[truncated]

SixWeining
SixWeining reviewed Mar 26, 2025
View reviewed changes
llvm/test/CodeGen/LoongArch/lasx/ir-instruction/absd.ll
;; 1. trunc(abs(sub(sext(a),sext(b)))) -> abds(a,b) or abdu(a,b)
;; 2. abs(sub_nsw(x, y)) -> abds(a,b)
;; 3. sub(smax(a,b),smin(a,b)) -> abds(a,b) or abdu(a,b)
;; 4. select(icmp(a,b),sub(a,b),sub(b,a)) -> abds(a,b) or abdu(a,b)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
;; 4. select(icmp(a,b),sub(a,b),sub(b,a)) -> abds(a,b) or abdu(a,b)
;; 4. select(icmp(a,b, slt|ult),sub(a,b),sub(b,a)) -> abds(a,b) or abdu(a,b)

SixWeining
SixWeining reviewed Mar 26, 2025
View reviewed changes
llvm/test/CodeGen/LoongArch/lasx/ir-instruction/absd.ll
;; 2. abs(sub_nsw(x, y)) -> abds(a,b)
;; 3. sub(smax(a,b),smin(a,b)) -> abds(a,b) or abdu(a,b)
;; 4. select(icmp(a,b),sub(a,b),sub(b,a)) -> abds(a,b) or abdu(a,b)
;; 5. sub(select(icmp(a,b),a,b),select(icmp(a,b),b,a)) -> abds(a,b) or abdu(a,b)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ditto

llvm/test/CodeGen/LoongArch/lasx/ir-instruction/absd.ll
ret <32 x i8> %sub
}

;; select(icmp(a,b),sub(a,b),sub(b,a)) -> abds(a,b)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
;; select(icmp(a,b),sub(a,b),sub(b,a)) -> abds(a,b)
;; select(icmp(a,b,slt),sub(a,b),sub(b,a)) -> abds(a,b)

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

@SixWeining SixWeining

@heiher heiher

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
@tangaac @llvmbot @SixWeining