--- /dev/null
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=aarch64-none-eabi -mattr=+sve2 | FileCheck %s
+
+define i64 @smlsl_i64(i64 %a, i32 %b, i32 %c, i32 %d, i32 %e) {
+; CHECK-LABEL: smlsl_i64:
+; CHECK: // %bb.0:
+; CHECK-NEXT: smull x8, w4, w3
+; CHECK-NEXT: smaddl x8, w2, w1, x8
+; CHECK-NEXT: sub x0, x0, x8
+; CHECK-NEXT: ret
+ %be = sext i32 %b to i64
+ %ce = sext i32 %c to i64
+ %de = sext i32 %d to i64
+ %ee = sext i32 %e to i64
+ %m1.neg = mul nsw i64 %ce, %be
+ %m2.neg = mul nsw i64 %ee, %de
+ %reass.add = add i64 %m2.neg, %m1.neg
+ %s2 = sub i64 %a, %reass.add
+ ret i64 %s2
+}
+
+define i64 @umlsl_i64(i64 %a, i32 %b, i32 %c, i32 %d, i32 %e) {
+; CHECK-LABEL: umlsl_i64:
+; CHECK: // %bb.0:
+; CHECK-NEXT: umull x8, w4, w3
+; CHECK-NEXT: umaddl x8, w2, w1, x8
+; CHECK-NEXT: sub x0, x0, x8
+; CHECK-NEXT: ret
+ %be = zext i32 %b to i64
+ %ce = zext i32 %c to i64
+ %de = zext i32 %d to i64
+ %ee = zext i32 %e to i64
+ %m1.neg = mul nuw i64 %ce, %be
+ %m2.neg = mul nuw i64 %ee, %de
+ %reass.add = add i64 %m2.neg, %m1.neg
+ %s2 = sub i64 %a, %reass.add
+ ret i64 %s2
+}
+
+define i64 @mls_i64(i64 %a, i64 %b, i64 %c, i64 %d, i64 %e) {
+; CHECK-LABEL: mls_i64:
+; CHECK: // %bb.0:
+; CHECK-NEXT: mul x8, x2, x1
+; CHECK-NEXT: madd x8, x4, x3, x8
+; CHECK-NEXT: sub x0, x0, x8
+; CHECK-NEXT: ret
+ %m1.neg = mul i64 %c, %b
+ %m2.neg = mul i64 %e, %d
+ %reass.add = add i64 %m2.neg, %m1.neg
+ %s2 = sub i64 %a, %reass.add
+ ret i64 %s2
+}
+
+define i16 @mls_i16(i16 %a, i16 %b, i16 %c, i16 %d, i16 %e) {
+; CHECK-LABEL: mls_i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: mul w8, w2, w1
+; CHECK-NEXT: madd w8, w4, w3, w8
+; CHECK-NEXT: sub w0, w0, w8
+; CHECK-NEXT: ret
+ %m1.neg = mul i16 %c, %b
+ %m2.neg = mul i16 %e, %d
+ %reass.add = add i16 %m2.neg, %m1.neg
+ %s2 = sub i16 %a, %reass.add
+ ret i16 %s2
+}
+
+define i64 @mla_i64(i64 %a, i64 %b, i64 %c, i64 %d, i64 %e) {
+; CHECK-LABEL: mla_i64:
+; CHECK: // %bb.0:
+; CHECK-NEXT: madd x8, x2, x1, x0
+; CHECK-NEXT: madd x0, x4, x3, x8
+; CHECK-NEXT: ret
+ %m1 = mul i64 %c, %b
+ %m2 = mul i64 %e, %d
+ %s1 = add i64 %m1, %a
+ %s2 = add i64 %s1, %m2
+ ret i64 %s2
+}
+
+define i64 @mls_i64_C(i64 %a, i64 %b, i64 %c, i64 %d, i64 %e) {
+; CHECK-LABEL: mls_i64_C:
+; CHECK: // %bb.0:
+; CHECK-NEXT: mul x8, x2, x1
+; CHECK-NEXT: mov w9, #10
+; CHECK-NEXT: madd x8, x4, x3, x8
+; CHECK-NEXT: sub x0, x9, x8
+; CHECK-NEXT: ret
+ %m1.neg = mul i64 %c, %b
+ %m2.neg = mul i64 %e, %d
+ %reass.add = add i64 %m2.neg, %m1.neg
+ %s2 = sub i64 10, %reass.add
+ ret i64 %s2
+}
+
+
+define <8 x i16> @smlsl_v8i16(<8 x i16> %a, <8 x i8> %b, <8 x i8> %c, <8 x i8> %d, <8 x i8> %e) {
+; CHECK-LABEL: smlsl_v8i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: smull v3.8h, v4.8b, v3.8b
+; CHECK-NEXT: smlal v3.8h, v2.8b, v1.8b
+; CHECK-NEXT: sub v0.8h, v0.8h, v3.8h
+; CHECK-NEXT: ret
+ %be = sext <8 x i8> %b to <8 x i16>
+ %ce = sext <8 x i8> %c to <8 x i16>
+ %de = sext <8 x i8> %d to <8 x i16>
+ %ee = sext <8 x i8> %e to <8 x i16>
+ %m1.neg = mul nsw <8 x i16> %ce, %be
+ %m2.neg = mul nsw <8 x i16> %ee, %de
+ %reass.add = add <8 x i16> %m2.neg, %m1.neg
+ %s2 = sub <8 x i16> %a, %reass.add
+ ret <8 x i16> %s2
+}
+
+define <8 x i16> @umlsl_v8i16(<8 x i16> %a, <8 x i8> %b, <8 x i8> %c, <8 x i8> %d, <8 x i8> %e) {
+; CHECK-LABEL: umlsl_v8i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: umull v3.8h, v4.8b, v3.8b
+; CHECK-NEXT: umlal v3.8h, v2.8b, v1.8b
+; CHECK-NEXT: sub v0.8h, v0.8h, v3.8h
+; CHECK-NEXT: ret
+ %be = zext <8 x i8> %b to <8 x i16>
+ %ce = zext <8 x i8> %c to <8 x i16>
+ %de = zext <8 x i8> %d to <8 x i16>
+ %ee = zext <8 x i8> %e to <8 x i16>
+ %m1.neg = mul nuw <8 x i16> %ce, %be
+ %m2.neg = mul nuw <8 x i16> %ee, %de
+ %reass.add = add <8 x i16> %m2.neg, %m1.neg
+ %s2 = sub <8 x i16> %a, %reass.add
+ ret <8 x i16> %s2
+}
+
+define <8 x i16> @mls_v8i16(<8 x i16> %a, <8 x i16> %b, <8 x i16> %c, <8 x i16> %d, <8 x i16> %e) {
+; CHECK-LABEL: mls_v8i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: mul v1.8h, v2.8h, v1.8h
+; CHECK-NEXT: mla v1.8h, v4.8h, v3.8h
+; CHECK-NEXT: sub v0.8h, v0.8h, v1.8h
+; CHECK-NEXT: ret
+ %m1.neg = mul <8 x i16> %c, %b
+ %m2.neg = mul <8 x i16> %e, %d
+ %reass.add = add <8 x i16> %m2.neg, %m1.neg
+ %s2 = sub <8 x i16> %a, %reass.add
+ ret <8 x i16> %s2
+}
+
+define <8 x i16> @mla_v8i16(<8 x i16> %a, <8 x i16> %b, <8 x i16> %c, <8 x i16> %d, <8 x i16> %e) {
+; CHECK-LABEL: mla_v8i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: mla v0.8h, v2.8h, v1.8h
+; CHECK-NEXT: mla v0.8h, v4.8h, v3.8h
+; CHECK-NEXT: ret
+ %m1 = mul <8 x i16> %c, %b
+ %m2 = mul <8 x i16> %e, %d
+ %s1 = add <8 x i16> %m1, %a
+ %s2 = add <8 x i16> %s1, %m2
+ ret <8 x i16> %s2
+}
+
+
+define <vscale x 8 x i16> @smlsl_nxv8i16(<vscale x 8 x i16> %a, <vscale x 8 x i8> %b, <vscale x 8 x i8> %c, <vscale x 8 x i8> %d, <vscale x 8 x i8> %e) {
+; CHECK-LABEL: smlsl_nxv8i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: ptrue p0.h
+; CHECK-NEXT: sxtb z3.h, p0/m, z3.h
+; CHECK-NEXT: sxtb z4.h, p0/m, z4.h
+; CHECK-NEXT: sxtb z1.h, p0/m, z1.h
+; CHECK-NEXT: sxtb z2.h, p0/m, z2.h
+; CHECK-NEXT: mul z3.h, z4.h, z3.h
+; CHECK-NEXT: mla z3.h, p0/m, z2.h, z1.h
+; CHECK-NEXT: sub z0.h, z0.h, z3.h
+; CHECK-NEXT: ret
+ %be = sext <vscale x 8 x i8> %b to <vscale x 8 x i16>
+ %ce = sext <vscale x 8 x i8> %c to <vscale x 8 x i16>
+ %de = sext <vscale x 8 x i8> %d to <vscale x 8 x i16>
+ %ee = sext <vscale x 8 x i8> %e to <vscale x 8 x i16>
+ %m1.neg = mul nsw <vscale x 8 x i16> %ce, %be
+ %m2.neg = mul nsw <vscale x 8 x i16> %ee, %de
+ %reass.add = add <vscale x 8 x i16> %m2.neg, %m1.neg
+ %s2 = sub <vscale x 8 x i16> %a, %reass.add
+ ret <vscale x 8 x i16> %s2
+}
+
+define <vscale x 8 x i16> @umlsl_nxv8i16(<vscale x 8 x i16> %a, <vscale x 8 x i8> %b, <vscale x 8 x i8> %c, <vscale x 8 x i8> %d, <vscale x 8 x i8> %e) {
+; CHECK-LABEL: umlsl_nxv8i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: and z3.h, z3.h, #0xff
+; CHECK-NEXT: and z4.h, z4.h, #0xff
+; CHECK-NEXT: ptrue p0.h
+; CHECK-NEXT: and z1.h, z1.h, #0xff
+; CHECK-NEXT: and z2.h, z2.h, #0xff
+; CHECK-NEXT: mul z3.h, z4.h, z3.h
+; CHECK-NEXT: mla z3.h, p0/m, z2.h, z1.h
+; CHECK-NEXT: sub z0.h, z0.h, z3.h
+; CHECK-NEXT: ret
+ %be = zext <vscale x 8 x i8> %b to <vscale x 8 x i16>
+ %ce = zext <vscale x 8 x i8> %c to <vscale x 8 x i16>
+ %de = zext <vscale x 8 x i8> %d to <vscale x 8 x i16>
+ %ee = zext <vscale x 8 x i8> %e to <vscale x 8 x i16>
+ %m1.neg = mul nuw <vscale x 8 x i16> %ce, %be
+ %m2.neg = mul nuw <vscale x 8 x i16> %ee, %de
+ %reass.add = add <vscale x 8 x i16> %m2.neg, %m1.neg
+ %s2 = sub <vscale x 8 x i16> %a, %reass.add
+ ret <vscale x 8 x i16> %s2
+}
+
+define <vscale x 8 x i16> @mls_nxv8i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c, <vscale x 8 x i16> %d, <vscale x 8 x i16> %e) {
+; CHECK-LABEL: mls_nxv8i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: ptrue p0.h
+; CHECK-NEXT: mul z3.h, z4.h, z3.h
+; CHECK-NEXT: mla z3.h, p0/m, z2.h, z1.h
+; CHECK-NEXT: sub z0.h, z0.h, z3.h
+; CHECK-NEXT: ret
+ %m1.neg = mul <vscale x 8 x i16> %c, %b
+ %m2.neg = mul <vscale x 8 x i16> %e, %d
+ %reass.add = add <vscale x 8 x i16> %m2.neg, %m1.neg
+ %s2 = sub <vscale x 8 x i16> %a, %reass.add
+ ret <vscale x 8 x i16> %s2
+}
+
+define <vscale x 8 x i16> @mla_nxv8i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c, <vscale x 8 x i16> %d, <vscale x 8 x i16> %e) {
+; CHECK-LABEL: mla_nxv8i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: ptrue p0.h
+; CHECK-NEXT: mla z0.h, p0/m, z2.h, z1.h
+; CHECK-NEXT: mla z0.h, p0/m, z4.h, z3.h
+; CHECK-NEXT: ret
+ %m1 = mul <vscale x 8 x i16> %c, %b
+ %m2 = mul <vscale x 8 x i16> %e, %d
+ %s1 = add <vscale x 8 x i16> %m1, %a
+ %s2 = add <vscale x 8 x i16> %s1, %m2
+ ret <vscale x 8 x i16> %s2
+}