summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorDhruv Chawla <dhruvc@nvidia.com>2024-03-13 13:14:40 +0530
committerDhruv Chawla <dhruvc@nvidia.com>2024-03-13 13:14:40 +0530
commit19e564be216064b829dfc90ad5362ed518212f2c (patch)
treed968c1f37f1065a3f270ee1a6c1a84c2af09ff6c
parent0d98582c8b86644e77f8ddd68fc251e41127b7f4 (diff)
Created using spr 1.3.5 [skip ci]
-rw-r--r--llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp20
-rw-r--r--llvm/test/CodeGen/AArch64/fcmp.ll81
-rw-r--r--llvm/test/CodeGen/AArch64/sext.ll14
-rw-r--r--llvm/test/CodeGen/AArch64/vecreduce-add.ll809
-rw-r--r--llvm/test/CodeGen/AArch64/zext.ll18
5 files changed, 779 insertions, 163 deletions
diff --git a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index bd3ff7265d51..a480c2909077 100644
--- a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -5496,6 +5496,26 @@ LegalizerHelper::moreElementsVector(MachineInstr &MI, unsigned TypeIdx,
return Legalized;
}
+ case TargetOpcode::G_SEXT:
+ case TargetOpcode::G_ZEXT:
+ case TargetOpcode::G_ANYEXT: {
+ LLT SrcTy = MRI.getType(MI.getOperand(1).getReg());
+ LLT DstTy = MRI.getType(MI.getOperand(0).getReg());
+ if (TypeIdx == 0) {
+ DstTy = MoreTy;
+ SrcTy = MoreTy.changeElementType(SrcTy.getElementType());
+ } else if (TypeIdx == 1) {
+ SrcTy = MoreTy;
+ DstTy = MoreTy.changeElementType(DstTy.getElementType());
+ }
+
+ Observer.changingInstr(MI);
+ moreElementsVectorSrc(MI, SrcTy, 1);
+ moreElementsVectorDst(MI, DstTy, 0);
+ Observer.changedInstr(MI);
+ return Legalized;
+ }
+
default:
return UnableToLegalize;
}
diff --git a/llvm/test/CodeGen/AArch64/fcmp.ll b/llvm/test/CodeGen/AArch64/fcmp.ll
index 2d0b5574cdd7..9916aeeab1ca 100644
--- a/llvm/test/CodeGen/AArch64/fcmp.ll
+++ b/llvm/test/CodeGen/AArch64/fcmp.ll
@@ -1108,61 +1108,54 @@ define <7 x i32> @v7f16_i32(<7 x half> %a, <7 x half> %b, <7 x i32> %d, <7 x i32
;
; CHECK-GI-FP16-LABEL: v7f16_i32:
; CHECK-GI-FP16: // %bb.0: // %entry
-; CHECK-GI-FP16-NEXT: fcmgt v1.8h, v1.8h, v0.8h
-; CHECK-GI-FP16-NEXT: mov w12, #31 // =0x1f
-; CHECK-GI-FP16-NEXT: ldr s4, [sp]
-; CHECK-GI-FP16-NEXT: fmov s2, w12
+; CHECK-GI-FP16-NEXT: fcmgt v0.8h, v1.8h, v0.8h
+; CHECK-GI-FP16-NEXT: mov w10, #31 // =0x1f
+; CHECK-GI-FP16-NEXT: ldr s3, [sp]
+; CHECK-GI-FP16-NEXT: fmov s1, w10
; CHECK-GI-FP16-NEXT: fmov s6, w0
-; CHECK-GI-FP16-NEXT: ldr s5, [sp, #8]
+; CHECK-GI-FP16-NEXT: ldr s4, [sp, #8]
; CHECK-GI-FP16-NEXT: ldr s7, [sp, #24]
; CHECK-GI-FP16-NEXT: ldr s16, [sp, #32]
-; CHECK-GI-FP16-NEXT: umov w9, v1.h[4]
-; CHECK-GI-FP16-NEXT: umov w8, v1.h[0]
-; CHECK-GI-FP16-NEXT: umov w11, v1.h[5]
-; CHECK-GI-FP16-NEXT: umov w10, v1.h[1]
-; CHECK-GI-FP16-NEXT: mov v2.s[1], w12
-; CHECK-GI-FP16-NEXT: umov w13, v1.h[2]
+; CHECK-GI-FP16-NEXT: umov w8, v0.h[4]
+; CHECK-GI-FP16-NEXT: umov w9, v0.h[5]
+; CHECK-GI-FP16-NEXT: mov v1.s[1], w10
; CHECK-GI-FP16-NEXT: mov v6.s[1], w1
; CHECK-GI-FP16-NEXT: mov v7.s[1], v16.s[0]
; CHECK-GI-FP16-NEXT: ldr s16, [sp, #40]
-; CHECK-GI-FP16-NEXT: fmov s3, w9
-; CHECK-GI-FP16-NEXT: fmov s0, w8
-; CHECK-GI-FP16-NEXT: umov w8, v1.h[6]
-; CHECK-GI-FP16-NEXT: mov v2.s[2], w12
-; CHECK-GI-FP16-NEXT: umov w9, v1.h[3]
+; CHECK-GI-FP16-NEXT: fmov s2, w8
+; CHECK-GI-FP16-NEXT: umov w8, v0.h[6]
+; CHECK-GI-FP16-NEXT: mov v1.s[2], w10
+; CHECK-GI-FP16-NEXT: ushll v0.4s, v0.4h, #0
; CHECK-GI-FP16-NEXT: mov v6.s[2], w2
; CHECK-GI-FP16-NEXT: mov v7.s[2], v16.s[0]
-; CHECK-GI-FP16-NEXT: mov v3.s[1], w11
-; CHECK-GI-FP16-NEXT: mov v0.s[1], w10
-; CHECK-GI-FP16-NEXT: mov w10, #-1 // =0xffffffff
-; CHECK-GI-FP16-NEXT: fmov s1, w10
-; CHECK-GI-FP16-NEXT: neg v17.4s, v2.4s
+; CHECK-GI-FP16-NEXT: mov v2.s[1], w9
+; CHECK-GI-FP16-NEXT: mov w9, #-1 // =0xffffffff
+; CHECK-GI-FP16-NEXT: fmov s5, w9
+; CHECK-GI-FP16-NEXT: neg v17.4s, v1.4s
+; CHECK-GI-FP16-NEXT: shl v0.4s, v0.4s, #31
; CHECK-GI-FP16-NEXT: mov v6.s[3], w3
+; CHECK-GI-FP16-NEXT: mov v2.s[2], w8
+; CHECK-GI-FP16-NEXT: fmov w8, s3
+; CHECK-GI-FP16-NEXT: fmov s3, w7
+; CHECK-GI-FP16-NEXT: mov v5.s[1], w9
+; CHECK-GI-FP16-NEXT: sshr v0.4s, v0.4s, #31
+; CHECK-GI-FP16-NEXT: mov v3.s[1], w8
+; CHECK-GI-FP16-NEXT: fmov w8, s4
+; CHECK-GI-FP16-NEXT: ldr s4, [sp, #16]
+; CHECK-GI-FP16-NEXT: ushl v1.4s, v2.4s, v1.4s
+; CHECK-GI-FP16-NEXT: fmov s2, w4
+; CHECK-GI-FP16-NEXT: mov v5.s[2], w9
+; CHECK-GI-FP16-NEXT: mov v2.s[1], w5
; CHECK-GI-FP16-NEXT: mov v3.s[2], w8
+; CHECK-GI-FP16-NEXT: sshl v1.4s, v1.4s, v17.4s
; CHECK-GI-FP16-NEXT: fmov w8, s4
-; CHECK-GI-FP16-NEXT: fmov s4, w7
-; CHECK-GI-FP16-NEXT: mov v0.s[2], w13
-; CHECK-GI-FP16-NEXT: mov v1.s[1], w10
-; CHECK-GI-FP16-NEXT: mov v4.s[1], w8
-; CHECK-GI-FP16-NEXT: fmov w8, s5
-; CHECK-GI-FP16-NEXT: ldr s5, [sp, #16]
-; CHECK-GI-FP16-NEXT: ushl v2.4s, v3.4s, v2.4s
-; CHECK-GI-FP16-NEXT: fmov s3, w4
-; CHECK-GI-FP16-NEXT: mov v0.s[3], w9
-; CHECK-GI-FP16-NEXT: mov v1.s[2], w10
-; CHECK-GI-FP16-NEXT: mov v3.s[1], w5
-; CHECK-GI-FP16-NEXT: mov v4.s[2], w8
-; CHECK-GI-FP16-NEXT: sshl v2.4s, v2.4s, v17.4s
-; CHECK-GI-FP16-NEXT: fmov w8, s5
-; CHECK-GI-FP16-NEXT: shl v0.4s, v0.4s, #31
-; CHECK-GI-FP16-NEXT: eor v1.16b, v2.16b, v1.16b
-; CHECK-GI-FP16-NEXT: mov v3.s[2], w6
-; CHECK-GI-FP16-NEXT: mov v4.s[3], w8
-; CHECK-GI-FP16-NEXT: sshr v0.4s, v0.4s, #31
-; CHECK-GI-FP16-NEXT: and v1.16b, v7.16b, v1.16b
-; CHECK-GI-FP16-NEXT: and v2.16b, v3.16b, v2.16b
-; CHECK-GI-FP16-NEXT: bsl v0.16b, v6.16b, v4.16b
-; CHECK-GI-FP16-NEXT: orr v1.16b, v2.16b, v1.16b
+; CHECK-GI-FP16-NEXT: eor v4.16b, v1.16b, v5.16b
+; CHECK-GI-FP16-NEXT: mov v2.s[2], w6
+; CHECK-GI-FP16-NEXT: mov v3.s[3], w8
+; CHECK-GI-FP16-NEXT: and v1.16b, v2.16b, v1.16b
+; CHECK-GI-FP16-NEXT: and v2.16b, v7.16b, v4.16b
+; CHECK-GI-FP16-NEXT: bsl v0.16b, v6.16b, v3.16b
+; CHECK-GI-FP16-NEXT: orr v1.16b, v1.16b, v2.16b
; CHECK-GI-FP16-NEXT: mov s2, v0.s[1]
; CHECK-GI-FP16-NEXT: mov s3, v0.s[2]
; CHECK-GI-FP16-NEXT: mov s4, v0.s[3]
diff --git a/llvm/test/CodeGen/AArch64/sext.ll b/llvm/test/CodeGen/AArch64/sext.ll
index 61f04fbf0484..3e0d5dd87509 100644
--- a/llvm/test/CodeGen/AArch64/sext.ll
+++ b/llvm/test/CodeGen/AArch64/sext.ll
@@ -280,13 +280,12 @@ define <3 x i64> @sext_v3i8_v3i64(<3 x i8> %a) {
;
; CHECK-GI-LABEL: sext_v3i8_v3i64:
; CHECK-GI: // %bb.0: // %entry
-; CHECK-GI-NEXT: // kill: def $w0 killed $w0 def $x0
-; CHECK-GI-NEXT: fmov d0, x0
-; CHECK-GI-NEXT: // kill: def $w1 killed $w1 def $x1
+; CHECK-GI-NEXT: fmov s0, w0
; CHECK-GI-NEXT: // kill: def $w2 killed $w2 def $x2
; CHECK-GI-NEXT: sxtb x8, w2
; CHECK-GI-NEXT: fmov d2, x8
-; CHECK-GI-NEXT: mov v0.d[1], x1
+; CHECK-GI-NEXT: mov v0.s[1], w1
+; CHECK-GI-NEXT: ushll v0.2d, v0.2s, #0
; CHECK-GI-NEXT: shl v0.2d, v0.2d, #56
; CHECK-GI-NEXT: sshr v0.2d, v0.2d, #56
; CHECK-GI-NEXT: mov d1, v0.d[1]
@@ -444,13 +443,12 @@ define <3 x i64> @sext_v3i10_v3i64(<3 x i10> %a) {
;
; CHECK-GI-LABEL: sext_v3i10_v3i64:
; CHECK-GI: // %bb.0: // %entry
-; CHECK-GI-NEXT: // kill: def $w0 killed $w0 def $x0
-; CHECK-GI-NEXT: fmov d0, x0
-; CHECK-GI-NEXT: // kill: def $w1 killed $w1 def $x1
+; CHECK-GI-NEXT: fmov s0, w0
; CHECK-GI-NEXT: // kill: def $w2 killed $w2 def $x2
; CHECK-GI-NEXT: sbfx x8, x2, #0, #10
; CHECK-GI-NEXT: fmov d2, x8
-; CHECK-GI-NEXT: mov v0.d[1], x1
+; CHECK-GI-NEXT: mov v0.s[1], w1
+; CHECK-GI-NEXT: ushll v0.2d, v0.2s, #0
; CHECK-GI-NEXT: shl v0.2d, v0.2d, #54
; CHECK-GI-NEXT: sshr v0.2d, v0.2d, #54
; CHECK-GI-NEXT: mov d1, v0.d[1]
diff --git a/llvm/test/CodeGen/AArch64/vecreduce-add.ll b/llvm/test/CodeGen/AArch64/vecreduce-add.ll
index 66b49466cc73..94ab173e9183 100644
--- a/llvm/test/CodeGen/AArch64/vecreduce-add.ll
+++ b/llvm/test/CodeGen/AArch64/vecreduce-add.ll
@@ -4,11 +4,6 @@
; RUN: llc -mtriple=aarch64-none-linux-gnu -global-isel -global-isel-abort=2 %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI,CHECK-GI-BASE
; RUN: llc -mtriple=aarch64-none-linux-gnu -global-isel -global-isel-abort=2 %s -o - -mattr=+dotprod 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI,CHECK-GI-DOT
-; CHECK-GI-BASE: warning: Instruction selection used fallback path for test_udot_v24i8
-; CHECK-GI-BASE-NEXT: warning: Instruction selection used fallback path for test_udot_v48i8
-; CHECK-GI-BASE-NEXT: warning: Instruction selection used fallback path for test_sdot_v24i8
-; CHECK-GI-BASE-NEXT: warning: Instruction selection used fallback path for test_sdot_v48i8
-
define i32 @addv_v2i32(<2 x i32> %a) {
; CHECK-LABEL: addv_v2i32:
; CHECK: // %bb.0: // %entry
@@ -2068,25 +2063,125 @@ define i32 @test_udot_v24i8(ptr %p1, ptr %p2) {
;
; CHECK-GI-BASE-LABEL: test_udot_v24i8:
; CHECK-GI-BASE: // %bb.0: // %entry
-; CHECK-GI-BASE-NEXT: ldr q0, [x0]
-; CHECK-GI-BASE-NEXT: ldr q1, [x1]
-; CHECK-GI-BASE-NEXT: ldr d4, [x0, #16]
-; CHECK-GI-BASE-NEXT: ldr d5, [x1, #16]
-; CHECK-GI-BASE-NEXT: ushll v2.8h, v0.8b, #0
-; CHECK-GI-BASE-NEXT: ushll v3.8h, v1.8b, #0
-; CHECK-GI-BASE-NEXT: ushll2 v0.8h, v0.16b, #0
-; CHECK-GI-BASE-NEXT: ushll2 v1.8h, v1.16b, #0
-; CHECK-GI-BASE-NEXT: umull v6.4s, v3.4h, v2.4h
-; CHECK-GI-BASE-NEXT: umull2 v2.4s, v3.8h, v2.8h
-; CHECK-GI-BASE-NEXT: ushll v3.8h, v4.8b, #0
-; CHECK-GI-BASE-NEXT: ushll v4.8h, v5.8b, #0
-; CHECK-GI-BASE-NEXT: umlal2 v2.4s, v4.8h, v3.8h
-; CHECK-GI-BASE-NEXT: umlal v6.4s, v4.4h, v3.4h
-; CHECK-GI-BASE-NEXT: umlal2 v2.4s, v1.8h, v0.8h
-; CHECK-GI-BASE-NEXT: umlal v6.4s, v1.4h, v0.4h
-; CHECK-GI-BASE-NEXT: add v0.4s, v6.4s, v2.4s
+; CHECK-GI-BASE-NEXT: ldr b0, [x0]
+; CHECK-GI-BASE-NEXT: ldr b19, [x0, #1]
+; CHECK-GI-BASE-NEXT: ldr b1, [x0, #8]
+; CHECK-GI-BASE-NEXT: ldr b18, [x0, #9]
+; CHECK-GI-BASE-NEXT: ldr b2, [x0, #16]
+; CHECK-GI-BASE-NEXT: ldr b16, [x0, #17]
+; CHECK-GI-BASE-NEXT: mov v0.b[1], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b3, [x1]
+; CHECK-GI-BASE-NEXT: ldr b17, [x1, #1]
+; CHECK-GI-BASE-NEXT: mov v1.b[1], v18.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[1], v16.b[0]
+; CHECK-GI-BASE-NEXT: ldr b4, [x1, #8]
+; CHECK-GI-BASE-NEXT: ldr b7, [x1, #9]
+; CHECK-GI-BASE-NEXT: mov v3.b[1], v17.b[0]
+; CHECK-GI-BASE-NEXT: ldr b5, [x1, #16]
+; CHECK-GI-BASE-NEXT: ldr b6, [x1, #17]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #2]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #18]
+; CHECK-GI-BASE-NEXT: mov v4.b[1], v7.b[0]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #10]
+; CHECK-GI-BASE-NEXT: ldr b18, [x1, #10]
+; CHECK-GI-BASE-NEXT: mov v5.b[1], v6.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[2], v17.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #2]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #18]
+; CHECK-GI-BASE-NEXT: mov v1.b[2], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[2], v16.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[2], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #3]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #11]
+; CHECK-GI-BASE-NEXT: mov v4.b[2], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #3]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #19]
+; CHECK-GI-BASE-NEXT: mov v5.b[2], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x1, #11]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #19]
+; CHECK-GI-BASE-NEXT: mov v1.b[3], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[3], v16.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[3], v17.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[3], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #4]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #12]
+; CHECK-GI-BASE-NEXT: mov v4.b[3], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #4]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #20]
+; CHECK-GI-BASE-NEXT: mov v5.b[3], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x1, #12]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #20]
+; CHECK-GI-BASE-NEXT: mov v1.b[4], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[4], v16.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[4], v17.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[4], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #5]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #13]
+; CHECK-GI-BASE-NEXT: mov v4.b[4], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #5]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #21]
+; CHECK-GI-BASE-NEXT: mov v5.b[4], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x1, #13]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #21]
+; CHECK-GI-BASE-NEXT: mov v1.b[5], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[5], v16.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[5], v17.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[5], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #6]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #14]
+; CHECK-GI-BASE-NEXT: mov v4.b[5], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #6]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #22]
+; CHECK-GI-BASE-NEXT: mov v5.b[5], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x1, #14]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #22]
+; CHECK-GI-BASE-NEXT: mov v1.b[6], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[6], v16.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[6], v17.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[6], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #7]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #15]
+; CHECK-GI-BASE-NEXT: mov v4.b[6], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #7]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #23]
+; CHECK-GI-BASE-NEXT: mov v5.b[6], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x1, #15]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #23]
+; CHECK-GI-BASE-NEXT: mov v1.b[7], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[7], v16.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[7], v17.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[7], v6.b[0]
+; CHECK-GI-BASE-NEXT: mov v4.b[7], v18.b[0]
+; CHECK-GI-BASE-NEXT: mov v5.b[7], v19.b[0]
+; CHECK-GI-BASE-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-GI-BASE-NEXT: ushll v3.8h, v3.8b, #0
+; CHECK-GI-BASE-NEXT: ushll v2.8h, v2.8b, #0
+; CHECK-GI-BASE-NEXT: ushll v0.8h, v0.8b, #0
+; CHECK-GI-BASE-NEXT: ushll v4.8h, v4.8b, #0
+; CHECK-GI-BASE-NEXT: ushll v5.8h, v5.8b, #0
+; CHECK-GI-BASE-NEXT: umull v6.4s, v3.4h, v0.4h
+; CHECK-GI-BASE-NEXT: umull2 v0.4s, v3.8h, v0.8h
+; CHECK-GI-BASE-NEXT: umull v3.4s, v4.4h, v1.4h
+; CHECK-GI-BASE-NEXT: umull2 v1.4s, v4.8h, v1.8h
+; CHECK-GI-BASE-NEXT: umull v4.4s, v5.4h, v2.4h
+; CHECK-GI-BASE-NEXT: umull2 v2.4s, v5.8h, v2.8h
+; CHECK-GI-BASE-NEXT: addv s5, v6.4s
; CHECK-GI-BASE-NEXT: addv s0, v0.4s
-; CHECK-GI-BASE-NEXT: fmov w0, s0
+; CHECK-GI-BASE-NEXT: addv s1, v1.4s
+; CHECK-GI-BASE-NEXT: addv s3, v3.4s
+; CHECK-GI-BASE-NEXT: addv s4, v4.4s
+; CHECK-GI-BASE-NEXT: addv s2, v2.4s
+; CHECK-GI-BASE-NEXT: fmov w8, s5
+; CHECK-GI-BASE-NEXT: fmov w9, s0
+; CHECK-GI-BASE-NEXT: fmov w11, s1
+; CHECK-GI-BASE-NEXT: fmov w10, s3
+; CHECK-GI-BASE-NEXT: fmov w12, s4
+; CHECK-GI-BASE-NEXT: fmov w13, s2
+; CHECK-GI-BASE-NEXT: add w8, w8, w9
+; CHECK-GI-BASE-NEXT: add w8, w8, w10
+; CHECK-GI-BASE-NEXT: add w9, w11, w12
+; CHECK-GI-BASE-NEXT: add w9, w9, w13
+; CHECK-GI-BASE-NEXT: add w0, w8, w9
; CHECK-GI-BASE-NEXT: ret
;
; CHECK-GI-DOT-LABEL: test_udot_v24i8:
@@ -2257,39 +2352,245 @@ define i32 @test_udot_v48i8(ptr %p1, ptr %p2) {
;
; CHECK-GI-BASE-LABEL: test_udot_v48i8:
; CHECK-GI-BASE: // %bb.0: // %entry
-; CHECK-GI-BASE-NEXT: ldp q0, q4, [x1]
-; CHECK-GI-BASE-NEXT: ldr q2, [x0, #32]
-; CHECK-GI-BASE-NEXT: ldp q1, q3, [x0]
-; CHECK-GI-BASE-NEXT: ldr q7, [x1, #32]
-; CHECK-GI-BASE-NEXT: ushll2 v16.8h, v2.16b, #0
-; CHECK-GI-BASE-NEXT: ushll2 v6.8h, v0.16b, #0
+; CHECK-GI-BASE-NEXT: ldr b0, [x0]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #1]
+; CHECK-GI-BASE-NEXT: ldr b1, [x0, #8]
+; CHECK-GI-BASE-NEXT: ldr b3, [x0, #9]
+; CHECK-GI-BASE-NEXT: ldr b2, [x0, #16]
+; CHECK-GI-BASE-NEXT: ldr b4, [x0, #17]
+; CHECK-GI-BASE-NEXT: mov v0.b[1], v7.b[0]
+; CHECK-GI-BASE-NEXT: ldr b5, [x0, #2]
+; CHECK-GI-BASE-NEXT: ldr b19, [x0, #10]
+; CHECK-GI-BASE-NEXT: mov v1.b[1], v3.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[1], v4.b[0]
+; CHECK-GI-BASE-NEXT: ldr b4, [x0, #32]
+; CHECK-GI-BASE-NEXT: ldr b16, [x0, #33]
+; CHECK-GI-BASE-NEXT: ldr b20, [x0, #18]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #3]
+; CHECK-GI-BASE-NEXT: ldr b3, [x0, #24]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #25]
+; CHECK-GI-BASE-NEXT: ldr b24, [x0, #34]
+; CHECK-GI-BASE-NEXT: mov v0.b[2], v5.b[0]
+; CHECK-GI-BASE-NEXT: mov v4.b[1], v16.b[0]
+; CHECK-GI-BASE-NEXT: ldr b5, [x0, #40]
+; CHECK-GI-BASE-NEXT: mov v1.b[2], v19.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[2], v20.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[1], v7.b[0]
+; CHECK-GI-BASE-NEXT: ldr b19, [x0, #41]
+; CHECK-GI-BASE-NEXT: ldr b20, [x0, #11]
+; CHECK-GI-BASE-NEXT: ldr b18, [x0, #4]
+; CHECK-GI-BASE-NEXT: ldr b25, [x0, #42]
+; CHECK-GI-BASE-NEXT: ldr b21, [x0, #12]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #5]
+; CHECK-GI-BASE-NEXT: mov v5.b[1], v19.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[3], v6.b[0]
+; CHECK-GI-BASE-NEXT: mov v4.b[2], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #19]
+; CHECK-GI-BASE-NEXT: mov v1.b[3], v20.b[0]
+; CHECK-GI-BASE-NEXT: ldr b20, [x0, #26]
+; CHECK-GI-BASE-NEXT: ldr b19, [x0, #13]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #6]
+; CHECK-GI-BASE-NEXT: ldr b22, [x0, #14]
+; CHECK-GI-BASE-NEXT: mov v2.b[3], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #20]
+; CHECK-GI-BASE-NEXT: mov v3.b[2], v20.b[0]
+; CHECK-GI-BASE-NEXT: mov v5.b[2], v25.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[4], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x0, #35]
+; CHECK-GI-BASE-NEXT: mov v1.b[4], v21.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x0, #27]
+; CHECK-GI-BASE-NEXT: ldr b20, [x0, #21]
+; CHECK-GI-BASE-NEXT: mov v4.b[3], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x0, #44]
+; CHECK-GI-BASE-NEXT: ldr b24, [x0, #22]
+; CHECK-GI-BASE-NEXT: mov v2.b[4], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #43]
+; CHECK-GI-BASE-NEXT: mov v3.b[3], v25.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[5], v7.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x0, #28]
+; CHECK-GI-BASE-NEXT: ldr b16, [x0, #7]
+; CHECK-GI-BASE-NEXT: mov v5.b[3], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #36]
+; CHECK-GI-BASE-NEXT: mov v1.b[5], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b19, [x0, #37]
+; CHECK-GI-BASE-NEXT: ldr b23, [x0, #15]
+; CHECK-GI-BASE-NEXT: ldr b26, [x0, #29]
+; CHECK-GI-BASE-NEXT: mov v4.b[4], v6.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[5], v20.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[4], v25.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[6], v17.b[0]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #45]
+; CHECK-GI-BASE-NEXT: ldr b20, [x0, #38]
+; CHECK-GI-BASE-NEXT: mov v5.b[4], v18.b[0]
+; CHECK-GI-BASE-NEXT: mov v1.b[6], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x1]
+; CHECK-GI-BASE-NEXT: ldr b21, [x0, #23]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #17]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #2]
+; CHECK-GI-BASE-NEXT: mov v4.b[5], v19.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[6], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #1]
+; CHECK-GI-BASE-NEXT: mov v3.b[5], v26.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[7], v16.b[0]
+; CHECK-GI-BASE-NEXT: ldr b26, [x1, #9]
+; CHECK-GI-BASE-NEXT: mov v5.b[5], v17.b[0]
+; CHECK-GI-BASE-NEXT: ldr b17, [x1, #8]
+; CHECK-GI-BASE-NEXT: mov v1.b[7], v23.b[0]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #16]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #24]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #25]
+; CHECK-GI-BASE-NEXT: mov v4.b[6], v20.b[0]
+; CHECK-GI-BASE-NEXT: ldr b20, [x1, #32]
+; CHECK-GI-BASE-NEXT: mov v6.b[1], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #33]
+; CHECK-GI-BASE-NEXT: mov v2.b[7], v21.b[0]
+; CHECK-GI-BASE-NEXT: mov v17.b[1], v26.b[0]
+; CHECK-GI-BASE-NEXT: ldr b21, [x1, #40]
+; CHECK-GI-BASE-NEXT: mov v16.b[1], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #41]
+; CHECK-GI-BASE-NEXT: mov v19.b[1], v23.b[0]
+; CHECK-GI-BASE-NEXT: mov v20.b[1], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #10]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #18]
+; CHECK-GI-BASE-NEXT: mov v21.b[1], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #26]
+; CHECK-GI-BASE-NEXT: mov v6.b[2], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #34]
+; CHECK-GI-BASE-NEXT: mov v17.b[2], v23.b[0]
+; CHECK-GI-BASE-NEXT: mov v16.b[2], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #42]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #3]
+; CHECK-GI-BASE-NEXT: mov v19.b[2], v25.b[0]
+; CHECK-GI-BASE-NEXT: mov v20.b[2], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #11]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #19]
+; CHECK-GI-BASE-NEXT: mov v21.b[2], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #27]
+; CHECK-GI-BASE-NEXT: mov v6.b[3], v23.b[0]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #35]
+; CHECK-GI-BASE-NEXT: mov v17.b[3], v25.b[0]
+; CHECK-GI-BASE-NEXT: mov v16.b[3], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #43]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #4]
+; CHECK-GI-BASE-NEXT: mov v19.b[3], v24.b[0]
+; CHECK-GI-BASE-NEXT: mov v20.b[3], v23.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #12]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #20]
+; CHECK-GI-BASE-NEXT: mov v21.b[3], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #28]
+; CHECK-GI-BASE-NEXT: mov v6.b[4], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #36]
+; CHECK-GI-BASE-NEXT: mov v17.b[4], v24.b[0]
+; CHECK-GI-BASE-NEXT: mov v16.b[4], v23.b[0]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #44]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #5]
+; CHECK-GI-BASE-NEXT: mov v19.b[4], v22.b[0]
+; CHECK-GI-BASE-NEXT: mov v20.b[4], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #13]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #21]
+; CHECK-GI-BASE-NEXT: mov v21.b[4], v23.b[0]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #29]
+; CHECK-GI-BASE-NEXT: mov v6.b[5], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #37]
+; CHECK-GI-BASE-NEXT: ldr b27, [x0, #30]
+; CHECK-GI-BASE-NEXT: mov v17.b[5], v22.b[0]
+; CHECK-GI-BASE-NEXT: mov v16.b[5], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #45]
+; CHECK-GI-BASE-NEXT: mov v19.b[5], v23.b[0]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #6]
+; CHECK-GI-BASE-NEXT: mov v20.b[5], v24.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[6], v27.b[0]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #14]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #22]
+; CHECK-GI-BASE-NEXT: mov v21.b[5], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #30]
+; CHECK-GI-BASE-NEXT: mov v6.b[6], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #38]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #31]
+; CHECK-GI-BASE-NEXT: mov v17.b[6], v23.b[0]
+; CHECK-GI-BASE-NEXT: mov v16.b[6], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #46]
+; CHECK-GI-BASE-NEXT: mov v19.b[6], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #7]
+; CHECK-GI-BASE-NEXT: mov v20.b[6], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b28, [x0, #46]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #15]
+; CHECK-GI-BASE-NEXT: mov v21.b[6], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #23]
+; CHECK-GI-BASE-NEXT: mov v3.b[7], v7.b[0]
+; CHECK-GI-BASE-NEXT: ldr b7, [x1, #31]
+; CHECK-GI-BASE-NEXT: ldr b18, [x0, #39]
+; CHECK-GI-BASE-NEXT: mov v6.b[7], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #39]
+; CHECK-GI-BASE-NEXT: mov v5.b[6], v28.b[0]
+; CHECK-GI-BASE-NEXT: mov v17.b[7], v22.b[0]
+; CHECK-GI-BASE-NEXT: mov v16.b[7], v24.b[0]
+; CHECK-GI-BASE-NEXT: mov v19.b[7], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v4.b[7], v18.b[0]
+; CHECK-GI-BASE-NEXT: mov v20.b[7], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b23, [x0, #47]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #47]
; CHECK-GI-BASE-NEXT: ushll v0.8h, v0.8b, #0
-; CHECK-GI-BASE-NEXT: ushll2 v17.8h, v7.16b, #0
-; CHECK-GI-BASE-NEXT: ushll2 v5.8h, v1.16b, #0
; CHECK-GI-BASE-NEXT: ushll v1.8h, v1.8b, #0
-; CHECK-GI-BASE-NEXT: umull2 v18.4s, v6.8h, v5.8h
-; CHECK-GI-BASE-NEXT: umull v19.4s, v0.4h, v1.4h
-; CHECK-GI-BASE-NEXT: umull v5.4s, v6.4h, v5.4h
-; CHECK-GI-BASE-NEXT: umull2 v0.4s, v0.8h, v1.8h
-; CHECK-GI-BASE-NEXT: ushll v1.8h, v2.8b, #0
-; CHECK-GI-BASE-NEXT: ushll v2.8h, v7.8b, #0
-; CHECK-GI-BASE-NEXT: ushll2 v6.8h, v3.16b, #0
-; CHECK-GI-BASE-NEXT: ushll2 v7.8h, v4.16b, #0
-; CHECK-GI-BASE-NEXT: umlal2 v18.4s, v17.8h, v16.8h
-; CHECK-GI-BASE-NEXT: umlal v5.4s, v17.4h, v16.4h
-; CHECK-GI-BASE-NEXT: umlal v19.4s, v2.4h, v1.4h
-; CHECK-GI-BASE-NEXT: umlal2 v0.4s, v2.8h, v1.8h
-; CHECK-GI-BASE-NEXT: ushll v1.8h, v3.8b, #0
-; CHECK-GI-BASE-NEXT: ushll v2.8h, v4.8b, #0
-; CHECK-GI-BASE-NEXT: umlal2 v18.4s, v7.8h, v6.8h
-; CHECK-GI-BASE-NEXT: umlal v5.4s, v7.4h, v6.4h
-; CHECK-GI-BASE-NEXT: umlal v19.4s, v2.4h, v1.4h
-; CHECK-GI-BASE-NEXT: umlal2 v0.4s, v2.8h, v1.8h
-; CHECK-GI-BASE-NEXT: add v1.4s, v19.4s, v5.4s
-; CHECK-GI-BASE-NEXT: add v0.4s, v0.4s, v18.4s
-; CHECK-GI-BASE-NEXT: add v0.4s, v1.4s, v0.4s
+; CHECK-GI-BASE-NEXT: mov v5.b[7], v23.b[0]
+; CHECK-GI-BASE-NEXT: ushll v2.8h, v2.8b, #0
+; CHECK-GI-BASE-NEXT: ushll v3.8h, v3.8b, #0
+; CHECK-GI-BASE-NEXT: mov v21.b[7], v22.b[0]
+; CHECK-GI-BASE-NEXT: ushll v6.8h, v6.8b, #0
+; CHECK-GI-BASE-NEXT: ushll v7.8h, v17.8b, #0
+; CHECK-GI-BASE-NEXT: ushll v16.8h, v16.8b, #0
+; CHECK-GI-BASE-NEXT: ushll v17.8h, v19.8b, #0
+; CHECK-GI-BASE-NEXT: ushll v4.8h, v4.8b, #0
+; CHECK-GI-BASE-NEXT: ushll v18.8h, v20.8b, #0
+; CHECK-GI-BASE-NEXT: umull v20.4s, v6.4h, v0.4h
+; CHECK-GI-BASE-NEXT: umull2 v0.4s, v6.8h, v0.8h
+; CHECK-GI-BASE-NEXT: umull v6.4s, v7.4h, v1.4h
+; CHECK-GI-BASE-NEXT: umull2 v1.4s, v7.8h, v1.8h
+; CHECK-GI-BASE-NEXT: umull v7.4s, v16.4h, v2.4h
+; CHECK-GI-BASE-NEXT: umull2 v2.4s, v16.8h, v2.8h
+; CHECK-GI-BASE-NEXT: umull v16.4s, v17.4h, v3.4h
+; CHECK-GI-BASE-NEXT: umull2 v3.4s, v17.8h, v3.8h
+; CHECK-GI-BASE-NEXT: ushll v5.8h, v5.8b, #0
+; CHECK-GI-BASE-NEXT: ushll v19.8h, v21.8b, #0
+; CHECK-GI-BASE-NEXT: umull v17.4s, v18.4h, v4.4h
+; CHECK-GI-BASE-NEXT: umull2 v4.4s, v18.8h, v4.8h
+; CHECK-GI-BASE-NEXT: addv s20, v20.4s
; CHECK-GI-BASE-NEXT: addv s0, v0.4s
-; CHECK-GI-BASE-NEXT: fmov w0, s0
+; CHECK-GI-BASE-NEXT: addv s6, v6.4s
+; CHECK-GI-BASE-NEXT: addv s1, v1.4s
+; CHECK-GI-BASE-NEXT: addv s7, v7.4s
+; CHECK-GI-BASE-NEXT: addv s2, v2.4s
+; CHECK-GI-BASE-NEXT: umull v18.4s, v19.4h, v5.4h
+; CHECK-GI-BASE-NEXT: umull2 v5.4s, v19.8h, v5.8h
+; CHECK-GI-BASE-NEXT: addv s16, v16.4s
+; CHECK-GI-BASE-NEXT: addv s3, v3.4s
+; CHECK-GI-BASE-NEXT: addv s17, v17.4s
+; CHECK-GI-BASE-NEXT: addv s4, v4.4s
+; CHECK-GI-BASE-NEXT: fmov w8, s20
+; CHECK-GI-BASE-NEXT: fmov w9, s0
+; CHECK-GI-BASE-NEXT: fmov w10, s6
+; CHECK-GI-BASE-NEXT: fmov w11, s1
+; CHECK-GI-BASE-NEXT: fmov w12, s7
+; CHECK-GI-BASE-NEXT: fmov w13, s2
+; CHECK-GI-BASE-NEXT: addv s18, v18.4s
+; CHECK-GI-BASE-NEXT: fmov w14, s16
+; CHECK-GI-BASE-NEXT: addv s0, v5.4s
+; CHECK-GI-BASE-NEXT: fmov w15, s3
+; CHECK-GI-BASE-NEXT: fmov w16, s17
+; CHECK-GI-BASE-NEXT: fmov w17, s4
+; CHECK-GI-BASE-NEXT: add w8, w8, w9
+; CHECK-GI-BASE-NEXT: add w9, w11, w12
+; CHECK-GI-BASE-NEXT: add w8, w8, w10
+; CHECK-GI-BASE-NEXT: add w9, w9, w13
+; CHECK-GI-BASE-NEXT: add w11, w14, w15
+; CHECK-GI-BASE-NEXT: fmov w18, s18
+; CHECK-GI-BASE-NEXT: add w8, w8, w9
+; CHECK-GI-BASE-NEXT: add w10, w11, w16
+; CHECK-GI-BASE-NEXT: fmov w11, s0
+; CHECK-GI-BASE-NEXT: add w9, w10, w17
+; CHECK-GI-BASE-NEXT: add w8, w8, w9
+; CHECK-GI-BASE-NEXT: add w9, w18, w11
+; CHECK-GI-BASE-NEXT: add w0, w8, w9
; CHECK-GI-BASE-NEXT: ret
;
; CHECK-GI-DOT-LABEL: test_udot_v48i8:
@@ -2646,25 +2947,125 @@ define i32 @test_sdot_v24i8(ptr %p1, ptr %p2) {
;
; CHECK-GI-BASE-LABEL: test_sdot_v24i8:
; CHECK-GI-BASE: // %bb.0: // %entry
-; CHECK-GI-BASE-NEXT: ldr q0, [x0]
-; CHECK-GI-BASE-NEXT: ldr q1, [x1]
-; CHECK-GI-BASE-NEXT: ldr d4, [x0, #16]
-; CHECK-GI-BASE-NEXT: ldr d5, [x1, #16]
-; CHECK-GI-BASE-NEXT: sshll v2.8h, v0.8b, #0
-; CHECK-GI-BASE-NEXT: sshll v3.8h, v1.8b, #0
-; CHECK-GI-BASE-NEXT: sshll2 v0.8h, v0.16b, #0
-; CHECK-GI-BASE-NEXT: sshll2 v1.8h, v1.16b, #0
-; CHECK-GI-BASE-NEXT: smull v6.4s, v3.4h, v2.4h
-; CHECK-GI-BASE-NEXT: smull2 v2.4s, v3.8h, v2.8h
-; CHECK-GI-BASE-NEXT: sshll v3.8h, v4.8b, #0
-; CHECK-GI-BASE-NEXT: sshll v4.8h, v5.8b, #0
-; CHECK-GI-BASE-NEXT: smlal2 v2.4s, v4.8h, v3.8h
-; CHECK-GI-BASE-NEXT: smlal v6.4s, v4.4h, v3.4h
-; CHECK-GI-BASE-NEXT: smlal2 v2.4s, v1.8h, v0.8h
-; CHECK-GI-BASE-NEXT: smlal v6.4s, v1.4h, v0.4h
-; CHECK-GI-BASE-NEXT: add v0.4s, v6.4s, v2.4s
+; CHECK-GI-BASE-NEXT: ldr b0, [x0]
+; CHECK-GI-BASE-NEXT: ldr b19, [x0, #1]
+; CHECK-GI-BASE-NEXT: ldr b1, [x0, #8]
+; CHECK-GI-BASE-NEXT: ldr b18, [x0, #9]
+; CHECK-GI-BASE-NEXT: ldr b2, [x0, #16]
+; CHECK-GI-BASE-NEXT: ldr b16, [x0, #17]
+; CHECK-GI-BASE-NEXT: mov v0.b[1], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b3, [x1]
+; CHECK-GI-BASE-NEXT: ldr b17, [x1, #1]
+; CHECK-GI-BASE-NEXT: mov v1.b[1], v18.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[1], v16.b[0]
+; CHECK-GI-BASE-NEXT: ldr b4, [x1, #8]
+; CHECK-GI-BASE-NEXT: ldr b7, [x1, #9]
+; CHECK-GI-BASE-NEXT: mov v3.b[1], v17.b[0]
+; CHECK-GI-BASE-NEXT: ldr b5, [x1, #16]
+; CHECK-GI-BASE-NEXT: ldr b6, [x1, #17]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #2]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #18]
+; CHECK-GI-BASE-NEXT: mov v4.b[1], v7.b[0]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #10]
+; CHECK-GI-BASE-NEXT: ldr b18, [x1, #10]
+; CHECK-GI-BASE-NEXT: mov v5.b[1], v6.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[2], v17.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #2]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #18]
+; CHECK-GI-BASE-NEXT: mov v1.b[2], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[2], v16.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[2], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #3]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #11]
+; CHECK-GI-BASE-NEXT: mov v4.b[2], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #3]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #19]
+; CHECK-GI-BASE-NEXT: mov v5.b[2], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x1, #11]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #19]
+; CHECK-GI-BASE-NEXT: mov v1.b[3], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[3], v16.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[3], v17.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[3], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #4]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #12]
+; CHECK-GI-BASE-NEXT: mov v4.b[3], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #4]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #20]
+; CHECK-GI-BASE-NEXT: mov v5.b[3], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x1, #12]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #20]
+; CHECK-GI-BASE-NEXT: mov v1.b[4], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[4], v16.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[4], v17.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[4], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #5]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #13]
+; CHECK-GI-BASE-NEXT: mov v4.b[4], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #5]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #21]
+; CHECK-GI-BASE-NEXT: mov v5.b[4], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x1, #13]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #21]
+; CHECK-GI-BASE-NEXT: mov v1.b[5], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[5], v16.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[5], v17.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[5], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #6]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #14]
+; CHECK-GI-BASE-NEXT: mov v4.b[5], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #6]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #22]
+; CHECK-GI-BASE-NEXT: mov v5.b[5], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x1, #14]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #22]
+; CHECK-GI-BASE-NEXT: mov v1.b[6], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[6], v16.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[6], v17.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[6], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #7]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #15]
+; CHECK-GI-BASE-NEXT: mov v4.b[6], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #7]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #23]
+; CHECK-GI-BASE-NEXT: mov v5.b[6], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x1, #15]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #23]
+; CHECK-GI-BASE-NEXT: mov v1.b[7], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[7], v16.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[7], v17.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[7], v6.b[0]
+; CHECK-GI-BASE-NEXT: mov v4.b[7], v18.b[0]
+; CHECK-GI-BASE-NEXT: mov v5.b[7], v19.b[0]
+; CHECK-GI-BASE-NEXT: sshll v1.8h, v1.8b, #0
+; CHECK-GI-BASE-NEXT: sshll v3.8h, v3.8b, #0
+; CHECK-GI-BASE-NEXT: sshll v2.8h, v2.8b, #0
+; CHECK-GI-BASE-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-GI-BASE-NEXT: sshll v4.8h, v4.8b, #0
+; CHECK-GI-BASE-NEXT: sshll v5.8h, v5.8b, #0
+; CHECK-GI-BASE-NEXT: smull v6.4s, v3.4h, v0.4h
+; CHECK-GI-BASE-NEXT: smull2 v0.4s, v3.8h, v0.8h
+; CHECK-GI-BASE-NEXT: smull v3.4s, v4.4h, v1.4h
+; CHECK-GI-BASE-NEXT: smull2 v1.4s, v4.8h, v1.8h
+; CHECK-GI-BASE-NEXT: smull v4.4s, v5.4h, v2.4h
+; CHECK-GI-BASE-NEXT: smull2 v2.4s, v5.8h, v2.8h
+; CHECK-GI-BASE-NEXT: addv s5, v6.4s
; CHECK-GI-BASE-NEXT: addv s0, v0.4s
-; CHECK-GI-BASE-NEXT: fmov w0, s0
+; CHECK-GI-BASE-NEXT: addv s1, v1.4s
+; CHECK-GI-BASE-NEXT: addv s3, v3.4s
+; CHECK-GI-BASE-NEXT: addv s4, v4.4s
+; CHECK-GI-BASE-NEXT: addv s2, v2.4s
+; CHECK-GI-BASE-NEXT: fmov w8, s5
+; CHECK-GI-BASE-NEXT: fmov w9, s0
+; CHECK-GI-BASE-NEXT: fmov w11, s1
+; CHECK-GI-BASE-NEXT: fmov w10, s3
+; CHECK-GI-BASE-NEXT: fmov w12, s4
+; CHECK-GI-BASE-NEXT: fmov w13, s2
+; CHECK-GI-BASE-NEXT: add w8, w8, w9
+; CHECK-GI-BASE-NEXT: add w8, w8, w10
+; CHECK-GI-BASE-NEXT: add w9, w11, w12
+; CHECK-GI-BASE-NEXT: add w9, w9, w13
+; CHECK-GI-BASE-NEXT: add w0, w8, w9
; CHECK-GI-BASE-NEXT: ret
;
; CHECK-GI-DOT-LABEL: test_sdot_v24i8:
@@ -2835,39 +3236,245 @@ define i32 @test_sdot_v48i8(ptr %p1, ptr %p2) {
;
; CHECK-GI-BASE-LABEL: test_sdot_v48i8:
; CHECK-GI-BASE: // %bb.0: // %entry
-; CHECK-GI-BASE-NEXT: ldp q0, q4, [x1]
-; CHECK-GI-BASE-NEXT: ldr q2, [x0, #32]
-; CHECK-GI-BASE-NEXT: ldp q1, q3, [x0]
-; CHECK-GI-BASE-NEXT: ldr q7, [x1, #32]
-; CHECK-GI-BASE-NEXT: sshll2 v16.8h, v2.16b, #0
-; CHECK-GI-BASE-NEXT: sshll2 v6.8h, v0.16b, #0
+; CHECK-GI-BASE-NEXT: ldr b0, [x0]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #1]
+; CHECK-GI-BASE-NEXT: ldr b1, [x0, #8]
+; CHECK-GI-BASE-NEXT: ldr b3, [x0, #9]
+; CHECK-GI-BASE-NEXT: ldr b2, [x0, #16]
+; CHECK-GI-BASE-NEXT: ldr b4, [x0, #17]
+; CHECK-GI-BASE-NEXT: mov v0.b[1], v7.b[0]
+; CHECK-GI-BASE-NEXT: ldr b5, [x0, #2]
+; CHECK-GI-BASE-NEXT: ldr b19, [x0, #10]
+; CHECK-GI-BASE-NEXT: mov v1.b[1], v3.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[1], v4.b[0]
+; CHECK-GI-BASE-NEXT: ldr b4, [x0, #32]
+; CHECK-GI-BASE-NEXT: ldr b16, [x0, #33]
+; CHECK-GI-BASE-NEXT: ldr b20, [x0, #18]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #3]
+; CHECK-GI-BASE-NEXT: ldr b3, [x0, #24]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #25]
+; CHECK-GI-BASE-NEXT: ldr b24, [x0, #34]
+; CHECK-GI-BASE-NEXT: mov v0.b[2], v5.b[0]
+; CHECK-GI-BASE-NEXT: mov v4.b[1], v16.b[0]
+; CHECK-GI-BASE-NEXT: ldr b5, [x0, #40]
+; CHECK-GI-BASE-NEXT: mov v1.b[2], v19.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[2], v20.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[1], v7.b[0]
+; CHECK-GI-BASE-NEXT: ldr b19, [x0, #41]
+; CHECK-GI-BASE-NEXT: ldr b20, [x0, #11]
+; CHECK-GI-BASE-NEXT: ldr b18, [x0, #4]
+; CHECK-GI-BASE-NEXT: ldr b25, [x0, #42]
+; CHECK-GI-BASE-NEXT: ldr b21, [x0, #12]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #5]
+; CHECK-GI-BASE-NEXT: mov v5.b[1], v19.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[3], v6.b[0]
+; CHECK-GI-BASE-NEXT: mov v4.b[2], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #19]
+; CHECK-GI-BASE-NEXT: mov v1.b[3], v20.b[0]
+; CHECK-GI-BASE-NEXT: ldr b20, [x0, #26]
+; CHECK-GI-BASE-NEXT: ldr b19, [x0, #13]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #6]
+; CHECK-GI-BASE-NEXT: ldr b22, [x0, #14]
+; CHECK-GI-BASE-NEXT: mov v2.b[3], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #20]
+; CHECK-GI-BASE-NEXT: mov v3.b[2], v20.b[0]
+; CHECK-GI-BASE-NEXT: mov v5.b[2], v25.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[4], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x0, #35]
+; CHECK-GI-BASE-NEXT: mov v1.b[4], v21.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x0, #27]
+; CHECK-GI-BASE-NEXT: ldr b20, [x0, #21]
+; CHECK-GI-BASE-NEXT: mov v4.b[3], v18.b[0]
+; CHECK-GI-BASE-NEXT: ldr b18, [x0, #44]
+; CHECK-GI-BASE-NEXT: ldr b24, [x0, #22]
+; CHECK-GI-BASE-NEXT: mov v2.b[4], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #43]
+; CHECK-GI-BASE-NEXT: mov v3.b[3], v25.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[5], v7.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x0, #28]
+; CHECK-GI-BASE-NEXT: ldr b16, [x0, #7]
+; CHECK-GI-BASE-NEXT: mov v5.b[3], v6.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x0, #36]
+; CHECK-GI-BASE-NEXT: mov v1.b[5], v19.b[0]
+; CHECK-GI-BASE-NEXT: ldr b19, [x0, #37]
+; CHECK-GI-BASE-NEXT: ldr b23, [x0, #15]
+; CHECK-GI-BASE-NEXT: ldr b26, [x0, #29]
+; CHECK-GI-BASE-NEXT: mov v4.b[4], v6.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[5], v20.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[4], v25.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[6], v17.b[0]
+; CHECK-GI-BASE-NEXT: ldr b17, [x0, #45]
+; CHECK-GI-BASE-NEXT: ldr b20, [x0, #38]
+; CHECK-GI-BASE-NEXT: mov v5.b[4], v18.b[0]
+; CHECK-GI-BASE-NEXT: mov v1.b[6], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b6, [x1]
+; CHECK-GI-BASE-NEXT: ldr b21, [x0, #23]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #17]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #2]
+; CHECK-GI-BASE-NEXT: mov v4.b[5], v19.b[0]
+; CHECK-GI-BASE-NEXT: mov v2.b[6], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #1]
+; CHECK-GI-BASE-NEXT: mov v3.b[5], v26.b[0]
+; CHECK-GI-BASE-NEXT: mov v0.b[7], v16.b[0]
+; CHECK-GI-BASE-NEXT: ldr b26, [x1, #9]
+; CHECK-GI-BASE-NEXT: mov v5.b[5], v17.b[0]
+; CHECK-GI-BASE-NEXT: ldr b17, [x1, #8]
+; CHECK-GI-BASE-NEXT: mov v1.b[7], v23.b[0]
+; CHECK-GI-BASE-NEXT: ldr b16, [x1, #16]
+; CHECK-GI-BASE-NEXT: ldr b19, [x1, #24]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #25]
+; CHECK-GI-BASE-NEXT: mov v4.b[6], v20.b[0]
+; CHECK-GI-BASE-NEXT: ldr b20, [x1, #32]
+; CHECK-GI-BASE-NEXT: mov v6.b[1], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #33]
+; CHECK-GI-BASE-NEXT: mov v2.b[7], v21.b[0]
+; CHECK-GI-BASE-NEXT: mov v17.b[1], v26.b[0]
+; CHECK-GI-BASE-NEXT: ldr b21, [x1, #40]
+; CHECK-GI-BASE-NEXT: mov v16.b[1], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #41]
+; CHECK-GI-BASE-NEXT: mov v19.b[1], v23.b[0]
+; CHECK-GI-BASE-NEXT: mov v20.b[1], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #10]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #18]
+; CHECK-GI-BASE-NEXT: mov v21.b[1], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #26]
+; CHECK-GI-BASE-NEXT: mov v6.b[2], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #34]
+; CHECK-GI-BASE-NEXT: mov v17.b[2], v23.b[0]
+; CHECK-GI-BASE-NEXT: mov v16.b[2], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #42]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #3]
+; CHECK-GI-BASE-NEXT: mov v19.b[2], v25.b[0]
+; CHECK-GI-BASE-NEXT: mov v20.b[2], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #11]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #19]
+; CHECK-GI-BASE-NEXT: mov v21.b[2], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #27]
+; CHECK-GI-BASE-NEXT: mov v6.b[3], v23.b[0]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #35]
+; CHECK-GI-BASE-NEXT: mov v17.b[3], v25.b[0]
+; CHECK-GI-BASE-NEXT: mov v16.b[3], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #43]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #4]
+; CHECK-GI-BASE-NEXT: mov v19.b[3], v24.b[0]
+; CHECK-GI-BASE-NEXT: mov v20.b[3], v23.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #12]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #20]
+; CHECK-GI-BASE-NEXT: mov v21.b[3], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #28]
+; CHECK-GI-BASE-NEXT: mov v6.b[4], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #36]
+; CHECK-GI-BASE-NEXT: mov v17.b[4], v24.b[0]
+; CHECK-GI-BASE-NEXT: mov v16.b[4], v23.b[0]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #44]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #5]
+; CHECK-GI-BASE-NEXT: mov v19.b[4], v22.b[0]
+; CHECK-GI-BASE-NEXT: mov v20.b[4], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #13]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #21]
+; CHECK-GI-BASE-NEXT: mov v21.b[4], v23.b[0]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #29]
+; CHECK-GI-BASE-NEXT: mov v6.b[5], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #37]
+; CHECK-GI-BASE-NEXT: ldr b27, [x0, #30]
+; CHECK-GI-BASE-NEXT: mov v17.b[5], v22.b[0]
+; CHECK-GI-BASE-NEXT: mov v16.b[5], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #45]
+; CHECK-GI-BASE-NEXT: mov v19.b[5], v23.b[0]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #6]
+; CHECK-GI-BASE-NEXT: mov v20.b[5], v24.b[0]
+; CHECK-GI-BASE-NEXT: mov v3.b[6], v27.b[0]
+; CHECK-GI-BASE-NEXT: ldr b23, [x1, #14]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #22]
+; CHECK-GI-BASE-NEXT: mov v21.b[5], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #30]
+; CHECK-GI-BASE-NEXT: mov v6.b[6], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #38]
+; CHECK-GI-BASE-NEXT: ldr b7, [x0, #31]
+; CHECK-GI-BASE-NEXT: mov v17.b[6], v23.b[0]
+; CHECK-GI-BASE-NEXT: mov v16.b[6], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #46]
+; CHECK-GI-BASE-NEXT: mov v19.b[6], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #7]
+; CHECK-GI-BASE-NEXT: mov v20.b[6], v22.b[0]
+; CHECK-GI-BASE-NEXT: ldr b28, [x0, #46]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #15]
+; CHECK-GI-BASE-NEXT: mov v21.b[6], v24.b[0]
+; CHECK-GI-BASE-NEXT: ldr b24, [x1, #23]
+; CHECK-GI-BASE-NEXT: mov v3.b[7], v7.b[0]
+; CHECK-GI-BASE-NEXT: ldr b7, [x1, #31]
+; CHECK-GI-BASE-NEXT: ldr b18, [x0, #39]
+; CHECK-GI-BASE-NEXT: mov v6.b[7], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b25, [x1, #39]
+; CHECK-GI-BASE-NEXT: mov v5.b[6], v28.b[0]
+; CHECK-GI-BASE-NEXT: mov v17.b[7], v22.b[0]
+; CHECK-GI-BASE-NEXT: mov v16.b[7], v24.b[0]
+; CHECK-GI-BASE-NEXT: mov v19.b[7], v7.b[0]
+; CHECK-GI-BASE-NEXT: mov v4.b[7], v18.b[0]
+; CHECK-GI-BASE-NEXT: mov v20.b[7], v25.b[0]
+; CHECK-GI-BASE-NEXT: ldr b23, [x0, #47]
+; CHECK-GI-BASE-NEXT: ldr b22, [x1, #47]
; CHECK-GI-BASE-NEXT: sshll v0.8h, v0.8b, #0
-; CHECK-GI-BASE-NEXT: sshll2 v17.8h, v7.16b, #0
-; CHECK-GI-BASE-NEXT: sshll2 v5.8h, v1.16b, #0
; CHECK-GI-BASE-NEXT: sshll v1.8h, v1.8b, #0
-; CHECK-GI-BASE-NEXT: smull2 v18.4s, v6.8h, v5.8h
-; CHECK-GI-BASE-NEXT: smull v19.4s, v0.4h, v1.4h
-; CHECK-GI-BASE-NEXT: smull v5.4s, v6.4h, v5.4h
-; CHECK-GI-BASE-NEXT: smull2 v0.4s, v0.8h, v1.8h
-; CHECK-GI-BASE-NEXT: sshll v1.8h, v2.8b, #0
-; CHECK-GI-BASE-NEXT: sshll v2.8h, v7.8b, #0
-; CHECK-GI-BASE-NEXT: sshll2 v6.8h, v3.16b, #0
-; CHECK-GI-BASE-NEXT: sshll2 v7.8h, v4.16b, #0
-; CHECK-GI-BASE-NEXT: smlal2 v18.4s, v17.8h, v16.8h
-; CHECK-GI-BASE-NEXT: smlal v5.4s, v17.4h, v16.4h
-; CHECK-GI-BASE-NEXT: smlal v19.4s, v2.4h, v1.4h
-; CHECK-GI-BASE-NEXT: smlal2 v0.4s, v2.8h, v1.8h
-; CHECK-GI-BASE-NEXT: sshll v1.8h, v3.8b, #0
-; CHECK-GI-BASE-NEXT: sshll v2.8h, v4.8b, #0
-; CHECK-GI-BASE-NEXT: smlal2 v18.4s, v7.8h, v6.8h
-; CHECK-GI-BASE-NEXT: smlal v5.4s, v7.4h, v6.4h
-; CHECK-GI-BASE-NEXT: smlal v19.4s, v2.4h, v1.4h
-; CHECK-GI-BASE-NEXT: smlal2 v0.4s, v2.8h, v1.8h
-; CHECK-GI-BASE-NEXT: add v1.4s, v19.4s, v5.4s
-; CHECK-GI-BASE-NEXT: add v0.4s, v0.4s, v18.4s
-; CHECK-GI-BASE-NEXT: add v0.4s, v1.4s, v0.4s
+; CHECK-GI-BASE-NEXT: mov v5.b[7], v23.b[0]
+; CHECK-GI-BASE-NEXT: sshll v2.8h, v2.8b, #0
+; CHECK-GI-BASE-NEXT: sshll v3.8h, v3.8b, #0
+; CHECK-GI-BASE-NEXT: mov v21.b[7], v22.b[0]
+; CHECK-GI-BASE-NEXT: sshll v6.8h, v6.8b, #0
+; CHECK-GI-BASE-NEXT: sshll v7.8h, v17.8b, #0
+; CHECK-GI-BASE-NEXT: sshll v16.8h, v16.8b, #0
+; CHECK-GI-BASE-NEXT: sshll v17.8h, v19.8b, #0
+; CHECK-GI-BASE-NEXT: sshll v4.8h, v4.8b, #0
+; CHECK-GI-BASE-NEXT: sshll v18.8h, v20.8b, #0
+; CHECK-GI-BASE-NEXT: smull v20.4s, v6.4h, v0.4h
+; CHECK-GI-BASE-NEXT: smull2 v0.4s, v6.8h, v0.8h
+; CHECK-GI-BASE-NEXT: smull v6.4s, v7.4h, v1.4h
+; CHECK-GI-BASE-NEXT: smull2 v1.4s, v7.8h, v1.8h
+; CHECK-GI-BASE-NEXT: smull v7.4s, v16.4h, v2.4h
+; CHECK-GI-BASE-NEXT: smull2 v2.4s, v16.8h, v2.8h
+; CHECK-GI-BASE-NEXT: smull v16.4s, v17.4h, v3.4h
+; CHECK-GI-BASE-NEXT: smull2 v3.4s, v17.8h, v3.8h
+; CHECK-GI-BASE-NEXT: sshll v5.8h, v5.8b, #0
+; CHECK-GI-BASE-NEXT: sshll v19.8h, v21.8b, #0
+; CHECK-GI-BASE-NEXT: smull v17.4s, v18.4h, v4.4h
+; CHECK-GI-BASE-NEXT: smull2 v4.4s, v18.8h, v4.8h
+; CHECK-GI-BASE-NEXT: addv s20, v20.4s
; CHECK-GI-BASE-NEXT: addv s0, v0.4s
-; CHECK-GI-BASE-NEXT: fmov w0, s0
+; CHECK-GI-BASE-NEXT: addv s6, v6.4s
+; CHECK-GI-BASE-NEXT: addv s1, v1.4s
+; CHECK-GI-BASE-NEXT: addv s7, v7.4s
+; CHECK-GI-BASE-NEXT: addv s2, v2.4s
+; CHECK-GI-BASE-NEXT: smull v18.4s, v19.4h, v5.4h
+; CHECK-GI-BASE-NEXT: smull2 v5.4s, v19.8h, v5.8h
+; CHECK-GI-BASE-NEXT: addv s16, v16.4s
+; CHECK-GI-BASE-NEXT: addv s3, v3.4s
+; CHECK-GI-BASE-NEXT: addv s17, v17.4s
+; CHECK-GI-BASE-NEXT: addv s4, v4.4s
+; CHECK-GI-BASE-NEXT: fmov w8, s20
+; CHECK-GI-BASE-NEXT: fmov w9, s0
+; CHECK-GI-BASE-NEXT: fmov w10, s6
+; CHECK-GI-BASE-NEXT: fmov w11, s1
+; CHECK-GI-BASE-NEXT: fmov w12, s7
+; CHECK-GI-BASE-NEXT: fmov w13, s2
+; CHECK-GI-BASE-NEXT: addv s18, v18.4s
+; CHECK-GI-BASE-NEXT: fmov w14, s16
+; CHECK-GI-BASE-NEXT: addv s0, v5.4s
+; CHECK-GI-BASE-NEXT: fmov w15, s3
+; CHECK-GI-BASE-NEXT: fmov w16, s17
+; CHECK-GI-BASE-NEXT: fmov w17, s4
+; CHECK-GI-BASE-NEXT: add w8, w8, w9
+; CHECK-GI-BASE-NEXT: add w9, w11, w12
+; CHECK-GI-BASE-NEXT: add w8, w8, w10
+; CHECK-GI-BASE-NEXT: add w9, w9, w13
+; CHECK-GI-BASE-NEXT: add w11, w14, w15
+; CHECK-GI-BASE-NEXT: fmov w18, s18
+; CHECK-GI-BASE-NEXT: add w8, w8, w9
+; CHECK-GI-BASE-NEXT: add w10, w11, w16
+; CHECK-GI-BASE-NEXT: fmov w11, s0
+; CHECK-GI-BASE-NEXT: add w9, w10, w17
+; CHECK-GI-BASE-NEXT: add w8, w8, w9
+; CHECK-GI-BASE-NEXT: add w9, w18, w11
+; CHECK-GI-BASE-NEXT: add w0, w8, w9
; CHECK-GI-BASE-NEXT: ret
;
; CHECK-GI-DOT-LABEL: test_sdot_v48i8:
diff --git a/llvm/test/CodeGen/AArch64/zext.ll b/llvm/test/CodeGen/AArch64/zext.ll
index 54b29be2132c..716d2398996b 100644
--- a/llvm/test/CodeGen/AArch64/zext.ll
+++ b/llvm/test/CodeGen/AArch64/zext.ll
@@ -305,15 +305,14 @@ define <3 x i64> @zext_v3i8_v3i64(<3 x i8> %a) {
;
; CHECK-GI-LABEL: zext_v3i8_v3i64:
; CHECK-GI: // %bb.0: // %entry
-; CHECK-GI-NEXT: // kill: def $w0 killed $w0 def $x0
-; CHECK-GI-NEXT: fmov d1, x0
-; CHECK-GI-NEXT: // kill: def $w1 killed $w1 def $x1
-; CHECK-GI-NEXT: movi v0.2d, #0x000000000000ff
+; CHECK-GI-NEXT: fmov s0, w0
+; CHECK-GI-NEXT: movi v1.2d, #0x000000000000ff
; CHECK-GI-NEXT: // kill: def $w2 killed $w2 def $x2
; CHECK-GI-NEXT: and x8, x2, #0xff
; CHECK-GI-NEXT: fmov d2, x8
-; CHECK-GI-NEXT: mov v1.d[1], x1
-; CHECK-GI-NEXT: and v0.16b, v1.16b, v0.16b
+; CHECK-GI-NEXT: mov v0.s[1], w1
+; CHECK-GI-NEXT: ushll v0.2d, v0.2s, #0
+; CHECK-GI-NEXT: and v0.16b, v0.16b, v1.16b
; CHECK-GI-NEXT: mov d1, v0.d[1]
; CHECK-GI-NEXT: // kill: def $d0 killed $d0 killed $q0
; CHECK-GI-NEXT: ret
@@ -470,15 +469,14 @@ define <3 x i64> @zext_v3i10_v3i64(<3 x i10> %a) {
;
; CHECK-GI-LABEL: zext_v3i10_v3i64:
; CHECK-GI: // %bb.0: // %entry
-; CHECK-GI-NEXT: // kill: def $w0 killed $w0 def $x0
-; CHECK-GI-NEXT: fmov d0, x0
-; CHECK-GI-NEXT: // kill: def $w1 killed $w1 def $x1
+; CHECK-GI-NEXT: fmov s0, w0
; CHECK-GI-NEXT: adrp x8, .LCPI27_0
; CHECK-GI-NEXT: // kill: def $w2 killed $w2 def $x2
; CHECK-GI-NEXT: ldr q1, [x8, :lo12:.LCPI27_0]
; CHECK-GI-NEXT: and x8, x2, #0x3ff
; CHECK-GI-NEXT: fmov d2, x8
-; CHECK-GI-NEXT: mov v0.d[1], x1
+; CHECK-GI-NEXT: mov v0.s[1], w1
+; CHECK-GI-NEXT: ushll v0.2d, v0.2s, #0
; CHECK-GI-NEXT: and v0.16b, v0.16b, v1.16b
; CHECK-GI-NEXT: mov d1, v0.d[1]
; CHECK-GI-NEXT: // kill: def $d0 killed $d0 killed $q0