[llvm] 4ca9a58 - [LoongArch][NFC] Add horizontal widening add/sub tests (#201487)
via llvm-commits
llvm-commits at lists.llvm.org
Tue Jun 16 18:15:26 PDT 2026
Author: hev
Date: 2026-06-17T09:15:21+08:00
New Revision: 4ca9a58546c6a1c070f1fc9079da46b340047632
URL: https://github.com/llvm/llvm-project/commit/4ca9a58546c6a1c070f1fc9079da46b340047632
DIFF: https://github.com/llvm/llvm-project/commit/4ca9a58546c6a1c070f1fc9079da46b340047632.diff
LOG: [LoongArch][NFC] Add horizontal widening add/sub tests (#201487)
Added:
llvm/test/CodeGen/LoongArch/lasx/vhaddw.ll
llvm/test/CodeGen/LoongArch/lasx/vhsubw.ll
llvm/test/CodeGen/LoongArch/lsx/vhaddw.ll
llvm/test/CodeGen/LoongArch/lsx/vhsubw.ll
Modified:
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/LoongArch/lasx/vhaddw.ll b/llvm/test/CodeGen/LoongArch/lasx/vhaddw.ll
new file mode 100644
index 0000000000000..1f405d640b133
--- /dev/null
+++ b/llvm/test/CodeGen/LoongArch/lasx/vhaddw.ll
@@ -0,0 +1,547 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lasx < %s | FileCheck %s
+; RUN: llc --mtriple=loongarch64 --mattr=+lasx < %s | FileCheck %s
+
+define void @xvhaddw_h_b_illegal(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhaddw_h_b_illegal:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr4, $a0, 0
+; CHECK-NEXT: xvld $xr3, $a0, 32
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: xvld $xr0, $a1, 32
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 1
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 3
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 5
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 7
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 9
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 11
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 13
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 15
+; CHECK-NEXT: xvpermi.d $xr4, $xr4, 14
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 7
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 1
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 3
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 5
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 7
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 9
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 11
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 13
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 15
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 15
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 1
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 3
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 5
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 7
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 9
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 11
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 13
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 15
+; CHECK-NEXT: xvpermi.d $xr3, $xr3, 14
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 7
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 1
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 3
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 5
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 7
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 9
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 11
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 13
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 15
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 15
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 8
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 10
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 12
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 14
+; CHECK-NEXT: xvpermi.d $xr1, $xr1, 14
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 7
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 8
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 10
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 12
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 14
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 15
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 0
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 2
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 4
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 6
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 8
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 10
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 12
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 14
+; CHECK-NEXT: xvpermi.d $xr0, $xr0, 14
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 7
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 0
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 2
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 4
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 6
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 8
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 10
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 12
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 14
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 15
+; CHECK-NEXT: vext2xv.h.b $xr0, $xr2
+; CHECK-NEXT: vext2xv.h.b $xr2, $xr4
+; CHECK-NEXT: vext2xv.h.b $xr3, $xr3
+; CHECK-NEXT: vext2xv.h.b $xr1, $xr1
+; CHECK-NEXT: xvadd.h $xr1, $xr2, $xr1
+; CHECK-NEXT: xvadd.h $xr0, $xr0, $xr3
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: xvst $xr1, $a2, 32
+; CHECK-NEXT: ret
+entry:
+ %0 = load <64 x i8>, ptr %a
+ %1 = load <64 x i8>, ptr %b
+ %2 = shufflevector <64 x i8> %0, <64 x i8> poison,
+ <32 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15,
+ i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31,
+ i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47,
+ i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63>
+ %3 = shufflevector <64 x i8> %1, <64 x i8> poison,
+ <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14,
+ i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30,
+ i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46,
+ i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62>
+ %4 = sext <32 x i8> %2 to <32 x i16>
+ %5 = sext <32 x i8> %3 to <32 x i16>
+ %6 = add <32 x i16> %4, %5
+ store <32 x i16> %6, ptr %r
+ ret void
+}
+
+define void @xvhaddw_h_b(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhaddw_h_b:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr2, $a0, 0
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 1
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 3
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 5
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 7
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 9
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 11
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 13
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 15
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 7
+; CHECK-NEXT: xvpermi.d $xr2, $xr2, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 1
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 3
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 5
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 7
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 9
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 11
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 13
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 15
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 15
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 8
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 10
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 12
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 14
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 7
+; CHECK-NEXT: xvpermi.d $xr1, $xr1, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 8
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 10
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 12
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 14
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 15
+; CHECK-NEXT: vext2xv.h.b $xr0, $xr0
+; CHECK-NEXT: vext2xv.h.b $xr1, $xr2
+; CHECK-NEXT: xvadd.h $xr0, $xr0, $xr1
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <32 x i8>, ptr %a
+ %1 = load <32 x i8>, ptr %b
+ %2 = shufflevector <32 x i8> %0, <32 x i8> poison,
+ <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15,
+ i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+ %3 = shufflevector <32 x i8> %1, <32 x i8> poison,
+ <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14,
+ i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+ %4 = sext <16 x i8> %2 to <16 x i16>
+ %5 = sext <16 x i8> %3 to <16 x i16>
+ %6 = add <16 x i16> %4, %5
+ store <16 x i16> %6, ptr %r
+ ret void
+}
+
+define void @xvhaddw_w_h(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhaddw_w_h:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr0, $a0, 0
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 1
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 0
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 3
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 1
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 5
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 2
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 7
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 3
+; CHECK-NEXT: xvpermi.d $xr0, $xr0, 14
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 1
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 4
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 3
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 5
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 5
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 6
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 7
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 7
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 0
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 1
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 2
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 3
+; CHECK-NEXT: xvpermi.d $xr1, $xr1, 14
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 4
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 5
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 6
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 7
+; CHECK-NEXT: vext2xv.w.h $xr1, $xr2
+; CHECK-NEXT: vext2xv.w.h $xr0, $xr0
+; CHECK-NEXT: xvadd.w $xr0, $xr1, $xr0
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <16 x i16>, ptr %a
+ %1 = load <16 x i16>, ptr %b
+ %2 = shufflevector <16 x i16> %0, <16 x i16> poison,
+ <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+ %3 = shufflevector <16 x i16> %1, <16 x i16> poison,
+ <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+ %4 = sext <8 x i16> %2 to <8 x i32>
+ %5 = sext <8 x i16> %3 to <8 x i32>
+ %6 = add <8 x i32> %4, %5
+ store <8 x i32> %6, ptr %r
+ ret void
+}
+
+define void @xvhaddw_d_w(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhaddw_d_w:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr0, $a0, 0
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 1
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 0
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 3
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 1
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 5
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 2
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 7
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 3
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 0
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 0
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 2
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 1
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 4
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 2
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 6
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 3
+; CHECK-NEXT: vext2xv.d.w $xr1, $xr2
+; CHECK-NEXT: vext2xv.d.w $xr0, $xr0
+; CHECK-NEXT: xvadd.d $xr0, $xr1, $xr0
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <8 x i32>, ptr %a
+ %1 = load <8 x i32>, ptr %b
+ %2 = shufflevector <8 x i32> %0, <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+ %3 = shufflevector <8 x i32> %1, <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+ %4 = sext <4 x i32> %2 to <4 x i64>
+ %5 = sext <4 x i32> %3 to <4 x i64>
+ %6 = add <4 x i64> %4, %5
+ store <4 x i64> %6, ptr %r
+ ret void
+}
+
+define void @xvhaddw_hu_bu(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhaddw_hu_bu:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr2, $a0, 0
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 1
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 3
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 5
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 7
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 9
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 11
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 13
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 15
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 7
+; CHECK-NEXT: xvpermi.d $xr2, $xr2, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 1
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 3
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 5
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 7
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 9
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 11
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 13
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 15
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 15
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 8
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 10
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 12
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 14
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 7
+; CHECK-NEXT: xvpermi.d $xr1, $xr1, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 8
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 10
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 12
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 14
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 15
+; CHECK-NEXT: vext2xv.hu.bu $xr0, $xr0
+; CHECK-NEXT: vext2xv.hu.bu $xr1, $xr2
+; CHECK-NEXT: xvadd.h $xr0, $xr0, $xr1
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <32 x i8>, ptr %a
+ %1 = load <32 x i8>, ptr %b
+ %2 = shufflevector <32 x i8> %0, <32 x i8> poison,
+ <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15,
+ i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+ %3 = shufflevector <32 x i8> %1, <32 x i8> poison,
+ <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14,
+ i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+ %4 = zext <16 x i8> %2 to <16 x i16>
+ %5 = zext <16 x i8> %3 to <16 x i16>
+ %6 = add <16 x i16> %4, %5
+ store <16 x i16> %6, ptr %r
+ ret void
+}
+
+define void @xvhaddw_wu_hu(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhaddw_wu_hu:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr0, $a0, 0
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 1
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 0
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 3
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 1
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 5
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 2
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 7
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 3
+; CHECK-NEXT: xvpermi.d $xr0, $xr0, 14
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 1
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 4
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 3
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 5
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 5
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 6
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 7
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 7
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 0
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 1
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 2
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 3
+; CHECK-NEXT: xvpermi.d $xr1, $xr1, 14
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 4
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 5
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 6
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 7
+; CHECK-NEXT: vext2xv.wu.hu $xr1, $xr2
+; CHECK-NEXT: vext2xv.wu.hu $xr0, $xr0
+; CHECK-NEXT: xvadd.w $xr0, $xr1, $xr0
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <16 x i16>, ptr %a
+ %1 = load <16 x i16>, ptr %b
+ %2 = shufflevector <16 x i16> %0, <16 x i16> poison,
+ <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+ %3 = shufflevector <16 x i16> %1, <16 x i16> poison,
+ <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+ %4 = zext <8 x i16> %2 to <8 x i32>
+ %5 = zext <8 x i16> %3 to <8 x i32>
+ %6 = add <8 x i32> %4, %5
+ store <8 x i32> %6, ptr %r
+ ret void
+}
+
+define void @xvhaddw_du_wu(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhaddw_du_wu:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr0, $a0, 0
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 1
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 0
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 3
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 1
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 5
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 2
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 7
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 3
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 0
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 0
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 2
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 1
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 4
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 2
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 6
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 3
+; CHECK-NEXT: vext2xv.du.wu $xr1, $xr2
+; CHECK-NEXT: vext2xv.du.wu $xr0, $xr0
+; CHECK-NEXT: xvadd.d $xr0, $xr1, $xr0
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <8 x i32>, ptr %a
+ %1 = load <8 x i32>, ptr %b
+ %2 = shufflevector <8 x i32> %0, <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+ %3 = shufflevector <8 x i32> %1, <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+ %4 = zext <4 x i32> %2 to <4 x i64>
+ %5 = zext <4 x i32> %3 to <4 x i64>
+ %6 = add <4 x i64> %4, %5
+ store <4 x i64> %6, ptr %r
+ ret void
+}
diff --git a/llvm/test/CodeGen/LoongArch/lasx/vhsubw.ll b/llvm/test/CodeGen/LoongArch/lasx/vhsubw.ll
new file mode 100644
index 0000000000000..856a7bf52e159
--- /dev/null
+++ b/llvm/test/CodeGen/LoongArch/lasx/vhsubw.ll
@@ -0,0 +1,547 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lasx < %s | FileCheck %s
+; RUN: llc --mtriple=loongarch64 --mattr=+lasx < %s | FileCheck %s
+
+define void @xvhsubw_h_b_illegal(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhsubw_h_b_illegal:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr4, $a0, 0
+; CHECK-NEXT: xvld $xr3, $a0, 32
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: xvld $xr0, $a1, 32
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 1
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 3
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 5
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 7
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 9
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 11
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 13
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 15
+; CHECK-NEXT: xvpermi.d $xr4, $xr4, 14
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 7
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 1
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 3
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 5
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 7
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 9
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 11
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 13
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr4, 15
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 15
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 1
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 3
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 5
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 7
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 9
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 11
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 13
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 15
+; CHECK-NEXT: xvpermi.d $xr3, $xr3, 14
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 7
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 1
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 3
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 5
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 7
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 9
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 11
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 13
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr3, 15
+; CHECK-NEXT: vinsgr2vr.b $vr4, $a0, 15
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 8
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 10
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 12
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 14
+; CHECK-NEXT: xvpermi.d $xr1, $xr1, 14
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 7
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 8
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 10
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 12
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 14
+; CHECK-NEXT: vinsgr2vr.b $vr3, $a0, 15
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 0
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 2
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 4
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 6
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 8
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 10
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 12
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 14
+; CHECK-NEXT: xvpermi.d $xr0, $xr0, 14
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 7
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 0
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 2
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 4
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 6
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 8
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 10
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 12
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr0, 14
+; CHECK-NEXT: vinsgr2vr.b $vr1, $a0, 15
+; CHECK-NEXT: vext2xv.h.b $xr0, $xr2
+; CHECK-NEXT: vext2xv.h.b $xr2, $xr4
+; CHECK-NEXT: vext2xv.h.b $xr3, $xr3
+; CHECK-NEXT: vext2xv.h.b $xr1, $xr1
+; CHECK-NEXT: xvsub.h $xr1, $xr2, $xr1
+; CHECK-NEXT: xvsub.h $xr0, $xr0, $xr3
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: xvst $xr1, $a2, 32
+; CHECK-NEXT: ret
+entry:
+ %0 = load <64 x i8>, ptr %a
+ %1 = load <64 x i8>, ptr %b
+ %2 = shufflevector <64 x i8> %0, <64 x i8> poison,
+ <32 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15,
+ i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31,
+ i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47,
+ i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63>
+ %3 = shufflevector <64 x i8> %1, <64 x i8> poison,
+ <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14,
+ i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30,
+ i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46,
+ i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62>
+ %4 = sext <32 x i8> %2 to <32 x i16>
+ %5 = sext <32 x i8> %3 to <32 x i16>
+ %6 = sub <32 x i16> %4, %5
+ store <32 x i16> %6, ptr %r
+ ret void
+}
+
+define void @xvhsubw_h_b(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhsubw_h_b:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr2, $a0, 0
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 1
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 3
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 5
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 7
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 9
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 11
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 13
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 15
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 7
+; CHECK-NEXT: xvpermi.d $xr2, $xr2, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 1
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 3
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 5
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 7
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 9
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 11
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 13
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 15
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 15
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 8
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 10
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 12
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 14
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 7
+; CHECK-NEXT: xvpermi.d $xr1, $xr1, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 8
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 10
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 12
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 14
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 15
+; CHECK-NEXT: vext2xv.h.b $xr0, $xr0
+; CHECK-NEXT: vext2xv.h.b $xr1, $xr2
+; CHECK-NEXT: xvsub.h $xr0, $xr0, $xr1
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <32 x i8>, ptr %a
+ %1 = load <32 x i8>, ptr %b
+ %2 = shufflevector <32 x i8> %0, <32 x i8> poison,
+ <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15,
+ i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+ %3 = shufflevector <32 x i8> %1, <32 x i8> poison,
+ <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14,
+ i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+ %4 = sext <16 x i8> %2 to <16 x i16>
+ %5 = sext <16 x i8> %3 to <16 x i16>
+ %6 = sub <16 x i16> %4, %5
+ store <16 x i16> %6, ptr %r
+ ret void
+}
+
+define void @xvhsubw_w_h(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhsubw_w_h:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr0, $a0, 0
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 1
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 0
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 3
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 1
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 5
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 2
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 7
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 3
+; CHECK-NEXT: xvpermi.d $xr0, $xr0, 14
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 1
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 4
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 3
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 5
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 5
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 6
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 7
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 7
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 0
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 1
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 2
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 3
+; CHECK-NEXT: xvpermi.d $xr1, $xr1, 14
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 4
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 5
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 6
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 7
+; CHECK-NEXT: vext2xv.w.h $xr1, $xr2
+; CHECK-NEXT: vext2xv.w.h $xr0, $xr0
+; CHECK-NEXT: xvsub.w $xr0, $xr1, $xr0
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <16 x i16>, ptr %a
+ %1 = load <16 x i16>, ptr %b
+ %2 = shufflevector <16 x i16> %0, <16 x i16> poison,
+ <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+ %3 = shufflevector <16 x i16> %1, <16 x i16> poison,
+ <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+ %4 = sext <8 x i16> %2 to <8 x i32>
+ %5 = sext <8 x i16> %3 to <8 x i32>
+ %6 = sub <8 x i32> %4, %5
+ store <8 x i32> %6, ptr %r
+ ret void
+}
+
+define void @xvhsubw_d_w(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhsubw_d_w:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr0, $a0, 0
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 1
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 0
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 3
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 1
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 5
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 2
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 7
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 3
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 0
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 0
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 2
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 1
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 4
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 2
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 6
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 3
+; CHECK-NEXT: vext2xv.d.w $xr1, $xr2
+; CHECK-NEXT: vext2xv.d.w $xr0, $xr0
+; CHECK-NEXT: xvsub.d $xr0, $xr1, $xr0
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <8 x i32>, ptr %a
+ %1 = load <8 x i32>, ptr %b
+ %2 = shufflevector <8 x i32> %0, <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+ %3 = shufflevector <8 x i32> %1, <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+ %4 = sext <4 x i32> %2 to <4 x i64>
+ %5 = sext <4 x i32> %3 to <4 x i64>
+ %6 = sub <4 x i64> %4, %5
+ store <4 x i64> %6, ptr %r
+ ret void
+}
+
+define void @xvhsubw_hu_bu(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhsubw_hu_bu:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr2, $a0, 0
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 1
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 3
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 5
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 7
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 9
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 11
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 13
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 15
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 7
+; CHECK-NEXT: xvpermi.d $xr2, $xr2, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 1
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 3
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 5
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 7
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 9
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 11
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 13
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr2, 15
+; CHECK-NEXT: vinsgr2vr.b $vr0, $a0, 15
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 0
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 1
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 2
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 3
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 8
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 4
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 10
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 5
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 12
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 6
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 14
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 7
+; CHECK-NEXT: xvpermi.d $xr1, $xr1, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 8
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 9
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 10
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 11
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 8
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 12
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 10
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 13
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 12
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 14
+; CHECK-NEXT: vpickve2gr.b $a0, $vr1, 14
+; CHECK-NEXT: vinsgr2vr.b $vr2, $a0, 15
+; CHECK-NEXT: vext2xv.hu.bu $xr0, $xr0
+; CHECK-NEXT: vext2xv.hu.bu $xr1, $xr2
+; CHECK-NEXT: xvsub.h $xr0, $xr0, $xr1
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <32 x i8>, ptr %a
+ %1 = load <32 x i8>, ptr %b
+ %2 = shufflevector <32 x i8> %0, <32 x i8> poison,
+ <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15,
+ i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+ %3 = shufflevector <32 x i8> %1, <32 x i8> poison,
+ <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14,
+ i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+ %4 = zext <16 x i8> %2 to <16 x i16>
+ %5 = zext <16 x i8> %3 to <16 x i16>
+ %6 = sub <16 x i16> %4, %5
+ store <16 x i16> %6, ptr %r
+ ret void
+}
+
+define void @xvhsubw_wu_hu(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhsubw_wu_hu:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr0, $a0, 0
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 1
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 0
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 3
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 1
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 5
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 2
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 7
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 3
+; CHECK-NEXT: xvpermi.d $xr0, $xr0, 14
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 1
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 4
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 3
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 5
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 5
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 6
+; CHECK-NEXT: vpickve2gr.h $a0, $vr0, 7
+; CHECK-NEXT: vinsgr2vr.h $vr2, $a0, 7
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 0
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 1
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 2
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 3
+; CHECK-NEXT: xvpermi.d $xr1, $xr1, 14
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 0
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 4
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 2
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 5
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 4
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 6
+; CHECK-NEXT: vpickve2gr.h $a0, $vr1, 6
+; CHECK-NEXT: vinsgr2vr.h $vr0, $a0, 7
+; CHECK-NEXT: vext2xv.wu.hu $xr1, $xr2
+; CHECK-NEXT: vext2xv.wu.hu $xr0, $xr0
+; CHECK-NEXT: xvsub.w $xr0, $xr1, $xr0
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <16 x i16>, ptr %a
+ %1 = load <16 x i16>, ptr %b
+ %2 = shufflevector <16 x i16> %0, <16 x i16> poison,
+ <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+ %3 = shufflevector <16 x i16> %1, <16 x i16> poison,
+ <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+ %4 = zext <8 x i16> %2 to <8 x i32>
+ %5 = zext <8 x i16> %3 to <8 x i32>
+ %6 = sub <8 x i32> %4, %5
+ store <8 x i32> %6, ptr %r
+ ret void
+}
+
+define void @xvhsubw_du_wu(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: xvhsubw_du_wu:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xvld $xr0, $a0, 0
+; CHECK-NEXT: xvld $xr1, $a1, 0
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 1
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 0
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 3
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 1
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 5
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 2
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr0, 7
+; CHECK-NEXT: vinsgr2vr.w $vr2, $a0, 3
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 0
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 0
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 2
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 1
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 4
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 2
+; CHECK-NEXT: xvpickve2gr.w $a0, $xr1, 6
+; CHECK-NEXT: vinsgr2vr.w $vr0, $a0, 3
+; CHECK-NEXT: vext2xv.du.wu $xr1, $xr2
+; CHECK-NEXT: vext2xv.du.wu $xr0, $xr0
+; CHECK-NEXT: xvsub.d $xr0, $xr1, $xr0
+; CHECK-NEXT: xvst $xr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <8 x i32>, ptr %a
+ %1 = load <8 x i32>, ptr %b
+ %2 = shufflevector <8 x i32> %0, <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+ %3 = shufflevector <8 x i32> %1, <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+ %4 = zext <4 x i32> %2 to <4 x i64>
+ %5 = zext <4 x i32> %3 to <4 x i64>
+ %6 = sub <4 x i64> %4, %5
+ store <4 x i64> %6, ptr %r
+ ret void
+}
diff --git a/llvm/test/CodeGen/LoongArch/lsx/vhaddw.ll b/llvm/test/CodeGen/LoongArch/lsx/vhaddw.ll
new file mode 100644
index 0000000000000..5a89e26de741f
--- /dev/null
+++ b/llvm/test/CodeGen/LoongArch/lsx/vhaddw.ll
@@ -0,0 +1,217 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lsx < %s | FileCheck %s --check-prefixes=CHECK,LA32
+; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s --check-prefixes=CHECK,LA64
+
+define void @vhaddw_h_b_illegal(ptr %a, ptr %b, ptr %r) nounwind {
+; LA32-LABEL: vhaddw_h_b_illegal:
+; LA32: # %bb.0: # %entry
+; LA32-NEXT: ld.w $a3, $a0, 0
+; LA32-NEXT: ld.w $a0, $a0, 4
+; LA32-NEXT: ld.w $a4, $a1, 0
+; LA32-NEXT: ld.w $a1, $a1, 4
+; LA32-NEXT: vinsgr2vr.w $vr0, $a3, 0
+; LA32-NEXT: vinsgr2vr.w $vr0, $a0, 1
+; LA32-NEXT: vinsgr2vr.w $vr1, $a4, 0
+; LA32-NEXT: vinsgr2vr.w $vr1, $a1, 1
+; LA32-NEXT: vpickod.b $vr0, $vr0, $vr0
+; LA32-NEXT: vpickev.b $vr1, $vr1, $vr1
+; LA32-NEXT: vslti.b $vr2, $vr0, 0
+; LA32-NEXT: vilvl.b $vr0, $vr2, $vr0
+; LA32-NEXT: vslti.b $vr2, $vr1, 0
+; LA32-NEXT: vilvl.b $vr1, $vr2, $vr1
+; LA32-NEXT: vadd.h $vr0, $vr0, $vr1
+; LA32-NEXT: vpickve2gr.w $a0, $vr0, 0
+; LA32-NEXT: st.w $a0, $a2, 0
+; LA32-NEXT: vpickve2gr.w $a0, $vr0, 1
+; LA32-NEXT: st.w $a0, $a2, 4
+; LA32-NEXT: ret
+;
+; LA64-LABEL: vhaddw_h_b_illegal:
+; LA64: # %bb.0: # %entry
+; LA64-NEXT: ld.d $a0, $a0, 0
+; LA64-NEXT: ld.d $a1, $a1, 0
+; LA64-NEXT: vinsgr2vr.d $vr0, $a0, 0
+; LA64-NEXT: vinsgr2vr.d $vr1, $a1, 0
+; LA64-NEXT: vpickod.b $vr0, $vr0, $vr0
+; LA64-NEXT: vpickev.b $vr1, $vr1, $vr1
+; LA64-NEXT: vslti.b $vr2, $vr0, 0
+; LA64-NEXT: vilvl.b $vr0, $vr2, $vr0
+; LA64-NEXT: vslti.b $vr2, $vr1, 0
+; LA64-NEXT: vilvl.b $vr1, $vr2, $vr1
+; LA64-NEXT: vadd.h $vr0, $vr0, $vr1
+; LA64-NEXT: vstelm.d $vr0, $a2, 0, 0
+; LA64-NEXT: ret
+entry:
+ %0 = load <8 x i8>, ptr %a
+ %1 = load <8 x i8>, ptr %b
+ %2 = shufflevector <8 x i8> %0, <8 x i8> poison,
+ <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+ %3 = shufflevector <8 x i8> %1, <8 x i8> poison,
+ <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+ %4 = sext <4 x i8> %2 to <4 x i16>
+ %5 = sext <4 x i8> %3 to <4 x i16>
+ %6 = add <4 x i16> %4, %5
+ store <4 x i16> %6, ptr %r
+ ret void
+}
+
+define void @vhaddw_h_b(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: vhaddw_h_b:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vld $vr0, $a0, 0
+; CHECK-NEXT: vld $vr1, $a1, 0
+; CHECK-NEXT: vpickod.b $vr0, $vr0, $vr0
+; CHECK-NEXT: vpickev.b $vr1, $vr1, $vr1
+; CHECK-NEXT: vslti.b $vr2, $vr0, 0
+; CHECK-NEXT: vilvl.b $vr0, $vr2, $vr0
+; CHECK-NEXT: vslti.b $vr2, $vr1, 0
+; CHECK-NEXT: vilvl.b $vr1, $vr2, $vr1
+; CHECK-NEXT: vadd.h $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <16 x i8>, ptr %a
+ %1 = load <16 x i8>, ptr %b
+ %2 = shufflevector <16 x i8> %0, <16 x i8> poison,
+ <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+ %3 = shufflevector <16 x i8> %1, <16 x i8> poison,
+ <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+ %4 = sext <8 x i8> %2 to <8 x i16>
+ %5 = sext <8 x i8> %3 to <8 x i16>
+ %6 = add <8 x i16> %4, %5
+ store <8 x i16> %6, ptr %r
+ ret void
+}
+
+define void @vhaddw_w_h(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: vhaddw_w_h:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vld $vr0, $a0, 0
+; CHECK-NEXT: vld $vr1, $a1, 0
+; CHECK-NEXT: vpickod.h $vr0, $vr0, $vr0
+; CHECK-NEXT: vpickev.h $vr1, $vr1, $vr1
+; CHECK-NEXT: vslti.h $vr2, $vr0, 0
+; CHECK-NEXT: vilvl.h $vr0, $vr2, $vr0
+; CHECK-NEXT: vslti.h $vr2, $vr1, 0
+; CHECK-NEXT: vilvl.h $vr1, $vr2, $vr1
+; CHECK-NEXT: vadd.w $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <8 x i16>, ptr %a
+ %1 = load <8 x i16>, ptr %b
+ %2 = shufflevector <8 x i16> %0, <8 x i16> poison,
+ <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+ %3 = shufflevector <8 x i16> %1, <8 x i16> poison,
+ <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+ %4 = sext <4 x i16> %2 to <4 x i32>
+ %5 = sext <4 x i16> %3 to <4 x i32>
+ %6 = add <4 x i32> %4, %5
+ store <4 x i32> %6, ptr %r
+ ret void
+}
+
+define void @vhaddw_d_w(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: vhaddw_d_w:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vld $vr0, $a0, 0
+; CHECK-NEXT: vld $vr1, $a1, 0
+; CHECK-NEXT: vshuf4i.w $vr0, $vr0, 13
+; CHECK-NEXT: vshuf4i.w $vr1, $vr1, 8
+; CHECK-NEXT: vslti.w $vr2, $vr0, 0
+; CHECK-NEXT: vilvl.w $vr0, $vr2, $vr0
+; CHECK-NEXT: vslti.w $vr2, $vr1, 0
+; CHECK-NEXT: vilvl.w $vr1, $vr2, $vr1
+; CHECK-NEXT: vadd.d $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <4 x i32>, ptr %a
+ %1 = load <4 x i32>, ptr %b
+ %2 = shufflevector <4 x i32> %0, <4 x i32> poison, <2 x i32> <i32 1, i32 3>
+ %3 = shufflevector <4 x i32> %1, <4 x i32> poison, <2 x i32> <i32 0, i32 2>
+ %4 = sext <2 x i32> %2 to <2 x i64>
+ %5 = sext <2 x i32> %3 to <2 x i64>
+ %6 = add <2 x i64> %4, %5
+ store <2 x i64> %6, ptr %r
+ ret void
+}
+
+define void @vhaddw_hu_bu(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: vhaddw_hu_bu:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vld $vr0, $a0, 0
+; CHECK-NEXT: vld $vr1, $a1, 0
+; CHECK-NEXT: pcalau12i $a0, %pc_hi20(.LCPI4_0)
+; CHECK-NEXT: vld $vr2, $a0, %pc_lo12(.LCPI4_0)
+; CHECK-NEXT: vrepli.b $vr3, 0
+; CHECK-NEXT: vpackod.b $vr0, $vr3, $vr0
+; CHECK-NEXT: vshuf.b $vr1, $vr3, $vr1, $vr2
+; CHECK-NEXT: vadd.h $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <16 x i8>, ptr %a
+ %1 = load <16 x i8>, ptr %b
+ %2 = shufflevector <16 x i8> %0, <16 x i8> poison,
+ <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+ %3 = shufflevector <16 x i8> %1, <16 x i8> poison,
+ <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+ %4 = zext <8 x i8> %2 to <8 x i16>
+ %5 = zext <8 x i8> %3 to <8 x i16>
+ %6 = add <8 x i16> %4, %5
+ store <8 x i16> %6, ptr %r
+ ret void
+}
+
+define void @vhaddw_wu_hu(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: vhaddw_wu_hu:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vld $vr0, $a0, 0
+; CHECK-NEXT: vld $vr1, $a1, 0
+; CHECK-NEXT: pcalau12i $a0, %pc_hi20(.LCPI5_0)
+; CHECK-NEXT: vld $vr2, $a0, %pc_lo12(.LCPI5_0)
+; CHECK-NEXT: vrepli.b $vr3, 0
+; CHECK-NEXT: vpackod.h $vr0, $vr3, $vr0
+; CHECK-NEXT: vshuf.h $vr2, $vr3, $vr1
+; CHECK-NEXT: vadd.w $vr0, $vr0, $vr2
+; CHECK-NEXT: vst $vr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <8 x i16>, ptr %a
+ %1 = load <8 x i16>, ptr %b
+ %2 = shufflevector <8 x i16> %0, <8 x i16> poison,
+ <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+ %3 = shufflevector <8 x i16> %1, <8 x i16> poison,
+ <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+ %4 = zext <4 x i16> %2 to <4 x i32>
+ %5 = zext <4 x i16> %3 to <4 x i32>
+ %6 = add <4 x i32> %4, %5
+ store <4 x i32> %6, ptr %r
+ ret void
+}
+
+define void @vhaddw_du_wu(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: vhaddw_du_wu:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vld $vr0, $a0, 0
+; CHECK-NEXT: vld $vr1, $a1, 0
+; CHECK-NEXT: pcalau12i $a0, %pc_hi20(.LCPI6_0)
+; CHECK-NEXT: vld $vr2, $a0, %pc_lo12(.LCPI6_0)
+; CHECK-NEXT: vrepli.b $vr3, 0
+; CHECK-NEXT: vpackod.w $vr0, $vr3, $vr0
+; CHECK-NEXT: vshuf.w $vr2, $vr3, $vr1
+; CHECK-NEXT: vadd.d $vr0, $vr0, $vr2
+; CHECK-NEXT: vst $vr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <4 x i32>, ptr %a
+ %1 = load <4 x i32>, ptr %b
+ %2 = shufflevector <4 x i32> %0, <4 x i32> poison, <2 x i32> <i32 1, i32 3>
+ %3 = shufflevector <4 x i32> %1, <4 x i32> poison, <2 x i32> <i32 0, i32 2>
+ %4 = zext <2 x i32> %2 to <2 x i64>
+ %5 = zext <2 x i32> %3 to <2 x i64>
+ %6 = add <2 x i64> %4, %5
+ store <2 x i64> %6, ptr %r
+ ret void
+}
diff --git a/llvm/test/CodeGen/LoongArch/lsx/vhsubw.ll b/llvm/test/CodeGen/LoongArch/lsx/vhsubw.ll
new file mode 100644
index 0000000000000..79d02daa0648f
--- /dev/null
+++ b/llvm/test/CodeGen/LoongArch/lsx/vhsubw.ll
@@ -0,0 +1,217 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lsx < %s | FileCheck %s --check-prefixes=CHECK,LA32
+; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s --check-prefixes=CHECK,LA64
+
+define void @vhsubw_h_b_illegal(ptr %a, ptr %b, ptr %r) nounwind {
+; LA32-LABEL: vhsubw_h_b_illegal:
+; LA32: # %bb.0: # %entry
+; LA32-NEXT: ld.w $a3, $a0, 0
+; LA32-NEXT: ld.w $a0, $a0, 4
+; LA32-NEXT: ld.w $a4, $a1, 0
+; LA32-NEXT: ld.w $a1, $a1, 4
+; LA32-NEXT: vinsgr2vr.w $vr0, $a3, 0
+; LA32-NEXT: vinsgr2vr.w $vr0, $a0, 1
+; LA32-NEXT: vinsgr2vr.w $vr1, $a4, 0
+; LA32-NEXT: vinsgr2vr.w $vr1, $a1, 1
+; LA32-NEXT: vpickod.b $vr0, $vr0, $vr0
+; LA32-NEXT: vpickev.b $vr1, $vr1, $vr1
+; LA32-NEXT: vslti.b $vr2, $vr0, 0
+; LA32-NEXT: vilvl.b $vr0, $vr2, $vr0
+; LA32-NEXT: vslti.b $vr2, $vr1, 0
+; LA32-NEXT: vilvl.b $vr1, $vr2, $vr1
+; LA32-NEXT: vsub.h $vr0, $vr0, $vr1
+; LA32-NEXT: vpickve2gr.w $a0, $vr0, 0
+; LA32-NEXT: st.w $a0, $a2, 0
+; LA32-NEXT: vpickve2gr.w $a0, $vr0, 1
+; LA32-NEXT: st.w $a0, $a2, 4
+; LA32-NEXT: ret
+;
+; LA64-LABEL: vhsubw_h_b_illegal:
+; LA64: # %bb.0: # %entry
+; LA64-NEXT: ld.d $a0, $a0, 0
+; LA64-NEXT: ld.d $a1, $a1, 0
+; LA64-NEXT: vinsgr2vr.d $vr0, $a0, 0
+; LA64-NEXT: vinsgr2vr.d $vr1, $a1, 0
+; LA64-NEXT: vpickod.b $vr0, $vr0, $vr0
+; LA64-NEXT: vpickev.b $vr1, $vr1, $vr1
+; LA64-NEXT: vslti.b $vr2, $vr0, 0
+; LA64-NEXT: vilvl.b $vr0, $vr2, $vr0
+; LA64-NEXT: vslti.b $vr2, $vr1, 0
+; LA64-NEXT: vilvl.b $vr1, $vr2, $vr1
+; LA64-NEXT: vsub.h $vr0, $vr0, $vr1
+; LA64-NEXT: vstelm.d $vr0, $a2, 0, 0
+; LA64-NEXT: ret
+entry:
+ %0 = load <8 x i8>, ptr %a
+ %1 = load <8 x i8>, ptr %b
+ %2 = shufflevector <8 x i8> %0, <8 x i8> poison,
+ <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+ %3 = shufflevector <8 x i8> %1, <8 x i8> poison,
+ <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+ %4 = sext <4 x i8> %2 to <4 x i16>
+ %5 = sext <4 x i8> %3 to <4 x i16>
+ %6 = sub <4 x i16> %4, %5
+ store <4 x i16> %6, ptr %r
+ ret void
+}
+
+define void @vhsubw_h_b(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: vhsubw_h_b:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vld $vr0, $a0, 0
+; CHECK-NEXT: vld $vr1, $a1, 0
+; CHECK-NEXT: vpickod.b $vr0, $vr0, $vr0
+; CHECK-NEXT: vpickev.b $vr1, $vr1, $vr1
+; CHECK-NEXT: vslti.b $vr2, $vr0, 0
+; CHECK-NEXT: vilvl.b $vr0, $vr2, $vr0
+; CHECK-NEXT: vslti.b $vr2, $vr1, 0
+; CHECK-NEXT: vilvl.b $vr1, $vr2, $vr1
+; CHECK-NEXT: vsub.h $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <16 x i8>, ptr %a
+ %1 = load <16 x i8>, ptr %b
+ %2 = shufflevector <16 x i8> %0, <16 x i8> poison,
+ <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+ %3 = shufflevector <16 x i8> %1, <16 x i8> poison,
+ <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+ %4 = sext <8 x i8> %2 to <8 x i16>
+ %5 = sext <8 x i8> %3 to <8 x i16>
+ %6 = sub <8 x i16> %4, %5
+ store <8 x i16> %6, ptr %r
+ ret void
+}
+
+define void @vhsubw_w_h(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: vhsubw_w_h:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vld $vr0, $a0, 0
+; CHECK-NEXT: vld $vr1, $a1, 0
+; CHECK-NEXT: vpickod.h $vr0, $vr0, $vr0
+; CHECK-NEXT: vpickev.h $vr1, $vr1, $vr1
+; CHECK-NEXT: vslti.h $vr2, $vr0, 0
+; CHECK-NEXT: vilvl.h $vr0, $vr2, $vr0
+; CHECK-NEXT: vslti.h $vr2, $vr1, 0
+; CHECK-NEXT: vilvl.h $vr1, $vr2, $vr1
+; CHECK-NEXT: vsub.w $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <8 x i16>, ptr %a
+ %1 = load <8 x i16>, ptr %b
+ %2 = shufflevector <8 x i16> %0, <8 x i16> poison,
+ <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+ %3 = shufflevector <8 x i16> %1, <8 x i16> poison,
+ <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+ %4 = sext <4 x i16> %2 to <4 x i32>
+ %5 = sext <4 x i16> %3 to <4 x i32>
+ %6 = sub <4 x i32> %4, %5
+ store <4 x i32> %6, ptr %r
+ ret void
+}
+
+define void @vhsubw_d_w(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: vhsubw_d_w:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vld $vr0, $a0, 0
+; CHECK-NEXT: vld $vr1, $a1, 0
+; CHECK-NEXT: vshuf4i.w $vr0, $vr0, 13
+; CHECK-NEXT: vshuf4i.w $vr1, $vr1, 8
+; CHECK-NEXT: vslti.w $vr2, $vr0, 0
+; CHECK-NEXT: vilvl.w $vr0, $vr2, $vr0
+; CHECK-NEXT: vslti.w $vr2, $vr1, 0
+; CHECK-NEXT: vilvl.w $vr1, $vr2, $vr1
+; CHECK-NEXT: vsub.d $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <4 x i32>, ptr %a
+ %1 = load <4 x i32>, ptr %b
+ %2 = shufflevector <4 x i32> %0, <4 x i32> poison, <2 x i32> <i32 1, i32 3>
+ %3 = shufflevector <4 x i32> %1, <4 x i32> poison, <2 x i32> <i32 0, i32 2>
+ %4 = sext <2 x i32> %2 to <2 x i64>
+ %5 = sext <2 x i32> %3 to <2 x i64>
+ %6 = sub <2 x i64> %4, %5
+ store <2 x i64> %6, ptr %r
+ ret void
+}
+
+define void @vhsubw_hu_bu(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: vhsubw_hu_bu:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vld $vr0, $a0, 0
+; CHECK-NEXT: vld $vr1, $a1, 0
+; CHECK-NEXT: pcalau12i $a0, %pc_hi20(.LCPI4_0)
+; CHECK-NEXT: vld $vr2, $a0, %pc_lo12(.LCPI4_0)
+; CHECK-NEXT: vrepli.b $vr3, 0
+; CHECK-NEXT: vpackod.b $vr0, $vr3, $vr0
+; CHECK-NEXT: vshuf.b $vr1, $vr3, $vr1, $vr2
+; CHECK-NEXT: vsub.h $vr0, $vr0, $vr1
+; CHECK-NEXT: vst $vr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <16 x i8>, ptr %a
+ %1 = load <16 x i8>, ptr %b
+ %2 = shufflevector <16 x i8> %0, <16 x i8> poison,
+ <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+ %3 = shufflevector <16 x i8> %1, <16 x i8> poison,
+ <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+ %4 = zext <8 x i8> %2 to <8 x i16>
+ %5 = zext <8 x i8> %3 to <8 x i16>
+ %6 = sub <8 x i16> %4, %5
+ store <8 x i16> %6, ptr %r
+ ret void
+}
+
+define void @vhsubw_wu_hu(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: vhsubw_wu_hu:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vld $vr0, $a0, 0
+; CHECK-NEXT: vld $vr1, $a1, 0
+; CHECK-NEXT: pcalau12i $a0, %pc_hi20(.LCPI5_0)
+; CHECK-NEXT: vld $vr2, $a0, %pc_lo12(.LCPI5_0)
+; CHECK-NEXT: vrepli.b $vr3, 0
+; CHECK-NEXT: vpackod.h $vr0, $vr3, $vr0
+; CHECK-NEXT: vshuf.h $vr2, $vr3, $vr1
+; CHECK-NEXT: vsub.w $vr0, $vr0, $vr2
+; CHECK-NEXT: vst $vr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <8 x i16>, ptr %a
+ %1 = load <8 x i16>, ptr %b
+ %2 = shufflevector <8 x i16> %0, <8 x i16> poison,
+ <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+ %3 = shufflevector <8 x i16> %1, <8 x i16> poison,
+ <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+ %4 = zext <4 x i16> %2 to <4 x i32>
+ %5 = zext <4 x i16> %3 to <4 x i32>
+ %6 = sub <4 x i32> %4, %5
+ store <4 x i32> %6, ptr %r
+ ret void
+}
+
+define void @vhsubw_du_wu(ptr %a, ptr %b, ptr %r) nounwind {
+; CHECK-LABEL: vhsubw_du_wu:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vld $vr0, $a0, 0
+; CHECK-NEXT: vld $vr1, $a1, 0
+; CHECK-NEXT: pcalau12i $a0, %pc_hi20(.LCPI6_0)
+; CHECK-NEXT: vld $vr2, $a0, %pc_lo12(.LCPI6_0)
+; CHECK-NEXT: vrepli.b $vr3, 0
+; CHECK-NEXT: vpackod.w $vr0, $vr3, $vr0
+; CHECK-NEXT: vshuf.w $vr2, $vr3, $vr1
+; CHECK-NEXT: vsub.d $vr0, $vr0, $vr2
+; CHECK-NEXT: vst $vr0, $a2, 0
+; CHECK-NEXT: ret
+entry:
+ %0 = load <4 x i32>, ptr %a
+ %1 = load <4 x i32>, ptr %b
+ %2 = shufflevector <4 x i32> %0, <4 x i32> poison, <2 x i32> <i32 1, i32 3>
+ %3 = shufflevector <4 x i32> %1, <4 x i32> poison, <2 x i32> <i32 0, i32 2>
+ %4 = zext <2 x i32> %2 to <2 x i64>
+ %5 = zext <2 x i32> %3 to <2 x i64>
+ %6 = sub <2 x i64> %4, %5
+ store <2 x i64> %6, ptr %r
+ ret void
+}
More information about the llvm-commits
mailing list