[llvm] fe0dee4 - [AArch64] Add tests for extending mul. NFC
David Green via llvm-commits
llvm-commits at lists.llvm.org
Mon Jun 10 03:43:24 PDT 2024
Author: David Green
Date: 2024-06-10T11:43:21+01:00
New Revision: fe0dee4df05e2e0497385e2de73a510b9156e1f4
URL: https://github.com/llvm/llvm-project/commit/fe0dee4df05e2e0497385e2de73a510b9156e1f4
DIFF: https://github.com/llvm/llvm-project/commit/fe0dee4df05e2e0497385e2de73a510b9156e1f4.diff
LOG: [AArch64] Add tests for extending mul. NFC
Added:
llvm/test/CodeGen/AArch64/neon-extmul.ll
Modified:
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/AArch64/neon-extmul.ll b/llvm/test/CodeGen/AArch64/neon-extmul.ll
new file mode 100644
index 0000000000000..502673fa76694
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/neon-extmul.ll
@@ -0,0 +1,477 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple aarch64 -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD
+; RUN: llc < %s -mtriple aarch64 -o - -global-isel | FileCheck %s --check-prefixes=CHECK,CHECK-GI
+
+define <8 x i32> @extmuls_v8i8_i32(<8 x i8> %s0, <8 x i8> %s1) {
+; CHECK-SD-LABEL: extmuls_v8i8_i32:
+; CHECK-SD: // %bb.0: // %entry
+; CHECK-SD-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-SD-NEXT: sshll v2.8h, v1.8b, #0
+; CHECK-SD-NEXT: smull2 v1.4s, v0.8h, v2.8h
+; CHECK-SD-NEXT: smull v0.4s, v0.4h, v2.4h
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: extmuls_v8i8_i32:
+; CHECK-GI: // %bb.0: // %entry
+; CHECK-GI-NEXT: sshll v2.8h, v0.8b, #0
+; CHECK-GI-NEXT: sshll v1.8h, v1.8b, #0
+; CHECK-GI-NEXT: smull v0.4s, v2.4h, v1.4h
+; CHECK-GI-NEXT: smull2 v1.4s, v2.8h, v1.8h
+; CHECK-GI-NEXT: ret
+entry:
+ %s0s = sext <8 x i8> %s0 to <8 x i32>
+ %s1s = sext <8 x i8> %s1 to <8 x i32>
+ %m = mul <8 x i32> %s0s, %s1s
+ ret <8 x i32> %m
+}
+
+define <8 x i32> @extmulu_v8i8_i32(<8 x i8> %s0, <8 x i8> %s1) {
+; CHECK-SD-LABEL: extmulu_v8i8_i32:
+; CHECK-SD: // %bb.0: // %entry
+; CHECK-SD-NEXT: ushll v0.8h, v0.8b, #0
+; CHECK-SD-NEXT: ushll v2.8h, v1.8b, #0
+; CHECK-SD-NEXT: umull2 v1.4s, v0.8h, v2.8h
+; CHECK-SD-NEXT: umull v0.4s, v0.4h, v2.4h
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: extmulu_v8i8_i32:
+; CHECK-GI: // %bb.0: // %entry
+; CHECK-GI-NEXT: ushll v2.8h, v0.8b, #0
+; CHECK-GI-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-GI-NEXT: umull v0.4s, v2.4h, v1.4h
+; CHECK-GI-NEXT: umull2 v1.4s, v2.8h, v1.8h
+; CHECK-GI-NEXT: ret
+entry:
+ %s0s = zext <8 x i8> %s0 to <8 x i32>
+ %s1s = zext <8 x i8> %s1 to <8 x i32>
+ %m = mul <8 x i32> %s0s, %s1s
+ ret <8 x i32> %m
+}
+
+define <8 x i32> @extmulsu_v8i8_i32(<8 x i8> %s0, <8 x i8> %s1) {
+; CHECK-SD-LABEL: extmulsu_v8i8_i32:
+; CHECK-SD: // %bb.0: // %entry
+; CHECK-SD-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-SD-NEXT: ushll v2.8h, v1.8b, #0
+; CHECK-SD-NEXT: smull2 v1.4s, v0.8h, v2.8h
+; CHECK-SD-NEXT: smull v0.4s, v0.4h, v2.4h
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: extmulsu_v8i8_i32:
+; CHECK-GI: // %bb.0: // %entry
+; CHECK-GI-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-GI-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-GI-NEXT: sshll v2.4s, v0.4h, #0
+; CHECK-GI-NEXT: sshll2 v3.4s, v0.8h, #0
+; CHECK-GI-NEXT: ushll v0.4s, v1.4h, #0
+; CHECK-GI-NEXT: ushll2 v1.4s, v1.8h, #0
+; CHECK-GI-NEXT: mul v0.4s, v2.4s, v0.4s
+; CHECK-GI-NEXT: mul v1.4s, v3.4s, v1.4s
+; CHECK-GI-NEXT: ret
+entry:
+ %s0s = sext <8 x i8> %s0 to <8 x i32>
+ %s1s = zext <8 x i8> %s1 to <8 x i32>
+ %m = mul <8 x i32> %s0s, %s1s
+ ret <8 x i32> %m
+}
+
+define <8 x i32> @extmuladds_v8i8_i32(<8 x i8> %s0, <8 x i8> %s1, <8 x i32> %b) {
+; CHECK-SD-LABEL: extmuladds_v8i8_i32:
+; CHECK-SD: // %bb.0: // %entry
+; CHECK-SD-NEXT: sshll v1.8h, v1.8b, #0
+; CHECK-SD-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-SD-NEXT: smlal2 v3.4s, v0.8h, v1.8h
+; CHECK-SD-NEXT: smlal v2.4s, v0.4h, v1.4h
+; CHECK-SD-NEXT: mov v0.16b, v2.16b
+; CHECK-SD-NEXT: mov v1.16b, v3.16b
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: extmuladds_v8i8_i32:
+; CHECK-GI: // %bb.0: // %entry
+; CHECK-GI-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-GI-NEXT: sshll v1.8h, v1.8b, #0
+; CHECK-GI-NEXT: smlal v2.4s, v0.4h, v1.4h
+; CHECK-GI-NEXT: smlal2 v3.4s, v0.8h, v1.8h
+; CHECK-GI-NEXT: mov v0.16b, v2.16b
+; CHECK-GI-NEXT: mov v1.16b, v3.16b
+; CHECK-GI-NEXT: ret
+entry:
+ %s0s = sext <8 x i8> %s0 to <8 x i32>
+ %s1s = sext <8 x i8> %s1 to <8 x i32>
+ %m = mul <8 x i32> %s0s, %s1s
+ %a = add <8 x i32> %m, %b
+ ret <8 x i32> %a
+}
+
+define <8 x i32> @extmuladdu_v8i8_i32(<8 x i8> %s0, <8 x i8> %s1, <8 x i32> %b) {
+; CHECK-SD-LABEL: extmuladdu_v8i8_i32:
+; CHECK-SD: // %bb.0: // %entry
+; CHECK-SD-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-SD-NEXT: ushll v0.8h, v0.8b, #0
+; CHECK-SD-NEXT: umlal2 v3.4s, v0.8h, v1.8h
+; CHECK-SD-NEXT: umlal v2.4s, v0.4h, v1.4h
+; CHECK-SD-NEXT: mov v0.16b, v2.16b
+; CHECK-SD-NEXT: mov v1.16b, v3.16b
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: extmuladdu_v8i8_i32:
+; CHECK-GI: // %bb.0: // %entry
+; CHECK-GI-NEXT: ushll v0.8h, v0.8b, #0
+; CHECK-GI-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-GI-NEXT: umlal v2.4s, v0.4h, v1.4h
+; CHECK-GI-NEXT: umlal2 v3.4s, v0.8h, v1.8h
+; CHECK-GI-NEXT: mov v0.16b, v2.16b
+; CHECK-GI-NEXT: mov v1.16b, v3.16b
+; CHECK-GI-NEXT: ret
+entry:
+ %s0s = zext <8 x i8> %s0 to <8 x i32>
+ %s1s = zext <8 x i8> %s1 to <8 x i32>
+ %m = mul <8 x i32> %s0s, %s1s
+ %a = add <8 x i32> %m, %b
+ ret <8 x i32> %a
+}
+
+define <8 x i32> @extmuladdsu_v8i8_i32(<8 x i8> %s0, <8 x i8> %s1, <8 x i32> %b) {
+; CHECK-SD-LABEL: extmuladdsu_v8i8_i32:
+; CHECK-SD: // %bb.0: // %entry
+; CHECK-SD-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-SD-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-SD-NEXT: smlal2 v3.4s, v0.8h, v1.8h
+; CHECK-SD-NEXT: smlal v2.4s, v0.4h, v1.4h
+; CHECK-SD-NEXT: mov v0.16b, v2.16b
+; CHECK-SD-NEXT: mov v1.16b, v3.16b
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: extmuladdsu_v8i8_i32:
+; CHECK-GI: // %bb.0: // %entry
+; CHECK-GI-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-GI-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-GI-NEXT: sshll v4.4s, v0.4h, #0
+; CHECK-GI-NEXT: sshll2 v0.4s, v0.8h, #0
+; CHECK-GI-NEXT: ushll v5.4s, v1.4h, #0
+; CHECK-GI-NEXT: ushll2 v1.4s, v1.8h, #0
+; CHECK-GI-NEXT: mla v2.4s, v4.4s, v5.4s
+; CHECK-GI-NEXT: mla v3.4s, v0.4s, v1.4s
+; CHECK-GI-NEXT: mov v0.16b, v2.16b
+; CHECK-GI-NEXT: mov v1.16b, v3.16b
+; CHECK-GI-NEXT: ret
+entry:
+ %s0s = sext <8 x i8> %s0 to <8 x i32>
+ %s1s = zext <8 x i8> %s1 to <8 x i32>
+ %m = mul <8 x i32> %s0s, %s1s
+ %a = add <8 x i32> %m, %b
+ ret <8 x i32> %a
+}
+
+
+
+define <8 x i64> @extmuls_v8i8_i64(<8 x i8> %s0, <8 x i8> %s1) {
+; CHECK-SD-LABEL: extmuls_v8i8_i64:
+; CHECK-SD: // %bb.0: // %entry
+; CHECK-SD-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-SD-NEXT: sshll v1.8h, v1.8b, #0
+; CHECK-SD-NEXT: sshll v2.4s, v0.4h, #0
+; CHECK-SD-NEXT: sshll v4.4s, v1.4h, #0
+; CHECK-SD-NEXT: sshll2 v5.4s, v0.8h, #0
+; CHECK-SD-NEXT: sshll2 v6.4s, v1.8h, #0
+; CHECK-SD-NEXT: smull v0.2d, v2.2s, v4.2s
+; CHECK-SD-NEXT: smull2 v1.2d, v2.4s, v4.4s
+; CHECK-SD-NEXT: smull2 v3.2d, v5.4s, v6.4s
+; CHECK-SD-NEXT: smull v2.2d, v5.2s, v6.2s
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: extmuls_v8i8_i64:
+; CHECK-GI: // %bb.0: // %entry
+; CHECK-GI-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-GI-NEXT: sshll v1.8h, v1.8b, #0
+; CHECK-GI-NEXT: sshll v2.4s, v0.4h, #0
+; CHECK-GI-NEXT: sshll v3.4s, v1.4h, #0
+; CHECK-GI-NEXT: sshll2 v4.4s, v0.8h, #0
+; CHECK-GI-NEXT: sshll2 v5.4s, v1.8h, #0
+; CHECK-GI-NEXT: smull v0.2d, v2.2s, v3.2s
+; CHECK-GI-NEXT: smull2 v1.2d, v2.4s, v3.4s
+; CHECK-GI-NEXT: smull v2.2d, v4.2s, v5.2s
+; CHECK-GI-NEXT: smull2 v3.2d, v4.4s, v5.4s
+; CHECK-GI-NEXT: ret
+entry:
+ %s0s = sext <8 x i8> %s0 to <8 x i64>
+ %s1s = sext <8 x i8> %s1 to <8 x i64>
+ %m = mul <8 x i64> %s0s, %s1s
+ ret <8 x i64> %m
+}
+
+define <8 x i64> @extmulu_v8i8_i64(<8 x i8> %s0, <8 x i8> %s1) {
+; CHECK-SD-LABEL: extmulu_v8i8_i64:
+; CHECK-SD: // %bb.0: // %entry
+; CHECK-SD-NEXT: ushll v0.8h, v0.8b, #0
+; CHECK-SD-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-SD-NEXT: ushll v2.4s, v0.4h, #0
+; CHECK-SD-NEXT: ushll v4.4s, v1.4h, #0
+; CHECK-SD-NEXT: ushll2 v5.4s, v0.8h, #0
+; CHECK-SD-NEXT: ushll2 v6.4s, v1.8h, #0
+; CHECK-SD-NEXT: umull v0.2d, v2.2s, v4.2s
+; CHECK-SD-NEXT: umull2 v1.2d, v2.4s, v4.4s
+; CHECK-SD-NEXT: umull2 v3.2d, v5.4s, v6.4s
+; CHECK-SD-NEXT: umull v2.2d, v5.2s, v6.2s
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: extmulu_v8i8_i64:
+; CHECK-GI: // %bb.0: // %entry
+; CHECK-GI-NEXT: ushll v0.8h, v0.8b, #0
+; CHECK-GI-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-GI-NEXT: ushll v2.4s, v0.4h, #0
+; CHECK-GI-NEXT: ushll v3.4s, v1.4h, #0
+; CHECK-GI-NEXT: ushll2 v4.4s, v0.8h, #0
+; CHECK-GI-NEXT: ushll2 v5.4s, v1.8h, #0
+; CHECK-GI-NEXT: umull v0.2d, v2.2s, v3.2s
+; CHECK-GI-NEXT: umull2 v1.2d, v2.4s, v3.4s
+; CHECK-GI-NEXT: umull v2.2d, v4.2s, v5.2s
+; CHECK-GI-NEXT: umull2 v3.2d, v4.4s, v5.4s
+; CHECK-GI-NEXT: ret
+entry:
+ %s0s = zext <8 x i8> %s0 to <8 x i64>
+ %s1s = zext <8 x i8> %s1 to <8 x i64>
+ %m = mul <8 x i64> %s0s, %s1s
+ ret <8 x i64> %m
+}
+
+define <8 x i64> @extaddsu_v8i8_i64(<8 x i8> %s0, <8 x i8> %s1) {
+; CHECK-SD-LABEL: extaddsu_v8i8_i64:
+; CHECK-SD: // %bb.0: // %entry
+; CHECK-SD-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-SD-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-SD-NEXT: sshll v2.4s, v0.4h, #0
+; CHECK-SD-NEXT: ushll v4.4s, v1.4h, #0
+; CHECK-SD-NEXT: sshll2 v5.4s, v0.8h, #0
+; CHECK-SD-NEXT: ushll2 v6.4s, v1.8h, #0
+; CHECK-SD-NEXT: smull v0.2d, v2.2s, v4.2s
+; CHECK-SD-NEXT: smull2 v1.2d, v2.4s, v4.4s
+; CHECK-SD-NEXT: smull2 v3.2d, v5.4s, v6.4s
+; CHECK-SD-NEXT: smull v2.2d, v5.2s, v6.2s
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: extaddsu_v8i8_i64:
+; CHECK-GI: // %bb.0: // %entry
+; CHECK-GI-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-GI-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-GI-NEXT: sshll v2.4s, v0.4h, #0
+; CHECK-GI-NEXT: ushll v3.4s, v1.4h, #0
+; CHECK-GI-NEXT: sshll2 v0.4s, v0.8h, #0
+; CHECK-GI-NEXT: ushll2 v1.4s, v1.8h, #0
+; CHECK-GI-NEXT: sshll v4.2d, v2.2s, #0
+; CHECK-GI-NEXT: ushll v5.2d, v3.2s, #0
+; CHECK-GI-NEXT: sshll2 v2.2d, v2.4s, #0
+; CHECK-GI-NEXT: ushll2 v3.2d, v3.4s, #0
+; CHECK-GI-NEXT: sshll v6.2d, v0.2s, #0
+; CHECK-GI-NEXT: ushll v7.2d, v1.2s, #0
+; CHECK-GI-NEXT: sshll2 v0.2d, v0.4s, #0
+; CHECK-GI-NEXT: ushll2 v1.2d, v1.4s, #0
+; CHECK-GI-NEXT: fmov x8, d4
+; CHECK-GI-NEXT: fmov x9, d5
+; CHECK-GI-NEXT: mov x12, v4.d[1]
+; CHECK-GI-NEXT: fmov x10, d3
+; CHECK-GI-NEXT: fmov x11, d7
+; CHECK-GI-NEXT: mov x13, v5.d[1]
+; CHECK-GI-NEXT: fmov x14, d1
+; CHECK-GI-NEXT: mov x15, v2.d[1]
+; CHECK-GI-NEXT: mov x16, v3.d[1]
+; CHECK-GI-NEXT: mul x8, x8, x9
+; CHECK-GI-NEXT: fmov x9, d2
+; CHECK-GI-NEXT: mov x17, v7.d[1]
+; CHECK-GI-NEXT: mov x18, v1.d[1]
+; CHECK-GI-NEXT: mul x12, x12, x13
+; CHECK-GI-NEXT: mov x13, v0.d[1]
+; CHECK-GI-NEXT: mul x9, x9, x10
+; CHECK-GI-NEXT: fmov x10, d6
+; CHECK-GI-NEXT: mul x15, x15, x16
+; CHECK-GI-NEXT: mul x10, x10, x11
+; CHECK-GI-NEXT: fmov x11, d0
+; CHECK-GI-NEXT: fmov d0, x8
+; CHECK-GI-NEXT: fmov d1, x9
+; CHECK-GI-NEXT: mul x13, x13, x18
+; CHECK-GI-NEXT: mov v0.d[1], x12
+; CHECK-GI-NEXT: mul x11, x11, x14
+; CHECK-GI-NEXT: mov x14, v6.d[1]
+; CHECK-GI-NEXT: mov v1.d[1], x15
+; CHECK-GI-NEXT: fmov d2, x10
+; CHECK-GI-NEXT: mul x14, x14, x17
+; CHECK-GI-NEXT: fmov d3, x11
+; CHECK-GI-NEXT: mov v3.d[1], x13
+; CHECK-GI-NEXT: mov v2.d[1], x14
+; CHECK-GI-NEXT: ret
+entry:
+ %s0s = sext <8 x i8> %s0 to <8 x i64>
+ %s1s = zext <8 x i8> %s1 to <8 x i64>
+ %m = mul <8 x i64> %s0s, %s1s
+ ret <8 x i64> %m
+}
+
+define <8 x i64> @extmuladds_v8i8_i64(<8 x i8> %s0, <8 x i8> %s1, <8 x i64> %b) {
+; CHECK-SD-LABEL: extmuladds_v8i8_i64:
+; CHECK-SD: // %bb.0: // %entry
+; CHECK-SD-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-SD-NEXT: sshll v1.8h, v1.8b, #0
+; CHECK-SD-NEXT: sshll v6.4s, v0.4h, #0
+; CHECK-SD-NEXT: sshll v7.4s, v1.4h, #0
+; CHECK-SD-NEXT: sshll2 v0.4s, v0.8h, #0
+; CHECK-SD-NEXT: sshll2 v1.4s, v1.8h, #0
+; CHECK-SD-NEXT: smlal v2.2d, v6.2s, v7.2s
+; CHECK-SD-NEXT: smlal2 v3.2d, v6.4s, v7.4s
+; CHECK-SD-NEXT: smlal2 v5.2d, v0.4s, v1.4s
+; CHECK-SD-NEXT: smlal v4.2d, v0.2s, v1.2s
+; CHECK-SD-NEXT: mov v0.16b, v2.16b
+; CHECK-SD-NEXT: mov v1.16b, v3.16b
+; CHECK-SD-NEXT: mov v2.16b, v4.16b
+; CHECK-SD-NEXT: mov v3.16b, v5.16b
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: extmuladds_v8i8_i64:
+; CHECK-GI: // %bb.0: // %entry
+; CHECK-GI-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-GI-NEXT: sshll v1.8h, v1.8b, #0
+; CHECK-GI-NEXT: sshll v6.4s, v0.4h, #0
+; CHECK-GI-NEXT: sshll v7.4s, v1.4h, #0
+; CHECK-GI-NEXT: sshll2 v0.4s, v0.8h, #0
+; CHECK-GI-NEXT: sshll2 v1.4s, v1.8h, #0
+; CHECK-GI-NEXT: smlal v2.2d, v6.2s, v7.2s
+; CHECK-GI-NEXT: smlal2 v3.2d, v6.4s, v7.4s
+; CHECK-GI-NEXT: smlal v4.2d, v0.2s, v1.2s
+; CHECK-GI-NEXT: smlal2 v5.2d, v0.4s, v1.4s
+; CHECK-GI-NEXT: mov v0.16b, v2.16b
+; CHECK-GI-NEXT: mov v1.16b, v3.16b
+; CHECK-GI-NEXT: mov v2.16b, v4.16b
+; CHECK-GI-NEXT: mov v3.16b, v5.16b
+; CHECK-GI-NEXT: ret
+entry:
+ %s0s = sext <8 x i8> %s0 to <8 x i64>
+ %s1s = sext <8 x i8> %s1 to <8 x i64>
+ %m = mul <8 x i64> %s0s, %s1s
+ %a = add <8 x i64> %m, %b
+ ret <8 x i64> %a
+}
+
+define <8 x i64> @extmuladdu_v8i8_i64(<8 x i8> %s0, <8 x i8> %s1, <8 x i64> %b) {
+; CHECK-SD-LABEL: extmuladdu_v8i8_i64:
+; CHECK-SD: // %bb.0: // %entry
+; CHECK-SD-NEXT: ushll v0.8h, v0.8b, #0
+; CHECK-SD-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-SD-NEXT: ushll v6.4s, v0.4h, #0
+; CHECK-SD-NEXT: ushll v7.4s, v1.4h, #0
+; CHECK-SD-NEXT: ushll2 v0.4s, v0.8h, #0
+; CHECK-SD-NEXT: ushll2 v1.4s, v1.8h, #0
+; CHECK-SD-NEXT: umlal v2.2d, v6.2s, v7.2s
+; CHECK-SD-NEXT: umlal2 v3.2d, v6.4s, v7.4s
+; CHECK-SD-NEXT: umlal2 v5.2d, v0.4s, v1.4s
+; CHECK-SD-NEXT: umlal v4.2d, v0.2s, v1.2s
+; CHECK-SD-NEXT: mov v0.16b, v2.16b
+; CHECK-SD-NEXT: mov v1.16b, v3.16b
+; CHECK-SD-NEXT: mov v2.16b, v4.16b
+; CHECK-SD-NEXT: mov v3.16b, v5.16b
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: extmuladdu_v8i8_i64:
+; CHECK-GI: // %bb.0: // %entry
+; CHECK-GI-NEXT: ushll v0.8h, v0.8b, #0
+; CHECK-GI-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-GI-NEXT: ushll v6.4s, v0.4h, #0
+; CHECK-GI-NEXT: ushll v7.4s, v1.4h, #0
+; CHECK-GI-NEXT: ushll2 v0.4s, v0.8h, #0
+; CHECK-GI-NEXT: ushll2 v1.4s, v1.8h, #0
+; CHECK-GI-NEXT: umlal v2.2d, v6.2s, v7.2s
+; CHECK-GI-NEXT: umlal2 v3.2d, v6.4s, v7.4s
+; CHECK-GI-NEXT: umlal v4.2d, v0.2s, v1.2s
+; CHECK-GI-NEXT: umlal2 v5.2d, v0.4s, v1.4s
+; CHECK-GI-NEXT: mov v0.16b, v2.16b
+; CHECK-GI-NEXT: mov v1.16b, v3.16b
+; CHECK-GI-NEXT: mov v2.16b, v4.16b
+; CHECK-GI-NEXT: mov v3.16b, v5.16b
+; CHECK-GI-NEXT: ret
+entry:
+ %s0s = zext <8 x i8> %s0 to <8 x i64>
+ %s1s = zext <8 x i8> %s1 to <8 x i64>
+ %m = mul <8 x i64> %s0s, %s1s
+ %a = add <8 x i64> %m, %b
+ ret <8 x i64> %a
+}
+
+define <8 x i64> @extmuladdsu_v8i8_i64(<8 x i8> %s0, <8 x i8> %s1, <8 x i64> %b) {
+; CHECK-SD-LABEL: extmuladdsu_v8i8_i64:
+; CHECK-SD: // %bb.0: // %entry
+; CHECK-SD-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-SD-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-SD-NEXT: sshll v6.4s, v0.4h, #0
+; CHECK-SD-NEXT: ushll v7.4s, v1.4h, #0
+; CHECK-SD-NEXT: sshll2 v0.4s, v0.8h, #0
+; CHECK-SD-NEXT: ushll2 v1.4s, v1.8h, #0
+; CHECK-SD-NEXT: smlal v2.2d, v6.2s, v7.2s
+; CHECK-SD-NEXT: smlal2 v3.2d, v6.4s, v7.4s
+; CHECK-SD-NEXT: smlal2 v5.2d, v0.4s, v1.4s
+; CHECK-SD-NEXT: smlal v4.2d, v0.2s, v1.2s
+; CHECK-SD-NEXT: mov v0.16b, v2.16b
+; CHECK-SD-NEXT: mov v1.16b, v3.16b
+; CHECK-SD-NEXT: mov v2.16b, v4.16b
+; CHECK-SD-NEXT: mov v3.16b, v5.16b
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: extmuladdsu_v8i8_i64:
+; CHECK-GI: // %bb.0: // %entry
+; CHECK-GI-NEXT: sshll v0.8h, v0.8b, #0
+; CHECK-GI-NEXT: ushll v1.8h, v1.8b, #0
+; CHECK-GI-NEXT: sshll v6.4s, v0.4h, #0
+; CHECK-GI-NEXT: ushll v7.4s, v1.4h, #0
+; CHECK-GI-NEXT: sshll2 v0.4s, v0.8h, #0
+; CHECK-GI-NEXT: ushll2 v1.4s, v1.8h, #0
+; CHECK-GI-NEXT: sshll v16.2d, v6.2s, #0
+; CHECK-GI-NEXT: ushll v17.2d, v7.2s, #0
+; CHECK-GI-NEXT: sshll2 v6.2d, v6.4s, #0
+; CHECK-GI-NEXT: ushll2 v7.2d, v7.4s, #0
+; CHECK-GI-NEXT: sshll v18.2d, v0.2s, #0
+; CHECK-GI-NEXT: ushll v19.2d, v1.2s, #0
+; CHECK-GI-NEXT: sshll2 v0.2d, v0.4s, #0
+; CHECK-GI-NEXT: ushll2 v1.2d, v1.4s, #0
+; CHECK-GI-NEXT: fmov x8, d16
+; CHECK-GI-NEXT: fmov x9, d17
+; CHECK-GI-NEXT: mov x12, v16.d[1]
+; CHECK-GI-NEXT: fmov x10, d7
+; CHECK-GI-NEXT: fmov x11, d19
+; CHECK-GI-NEXT: mov x13, v17.d[1]
+; CHECK-GI-NEXT: fmov x14, d1
+; CHECK-GI-NEXT: mov x15, v6.d[1]
+; CHECK-GI-NEXT: mov x16, v7.d[1]
+; CHECK-GI-NEXT: mul x8, x8, x9
+; CHECK-GI-NEXT: fmov x9, d6
+; CHECK-GI-NEXT: mov x17, v19.d[1]
+; CHECK-GI-NEXT: mov x18, v1.d[1]
+; CHECK-GI-NEXT: mul x12, x12, x13
+; CHECK-GI-NEXT: mov x13, v0.d[1]
+; CHECK-GI-NEXT: mul x9, x9, x10
+; CHECK-GI-NEXT: fmov x10, d18
+; CHECK-GI-NEXT: mul x15, x15, x16
+; CHECK-GI-NEXT: mul x10, x10, x11
+; CHECK-GI-NEXT: fmov x11, d0
+; CHECK-GI-NEXT: fmov d0, x8
+; CHECK-GI-NEXT: fmov d1, x9
+; CHECK-GI-NEXT: mul x13, x13, x18
+; CHECK-GI-NEXT: mov v0.d[1], x12
+; CHECK-GI-NEXT: mul x11, x11, x14
+; CHECK-GI-NEXT: mov x14, v18.d[1]
+; CHECK-GI-NEXT: mov v1.d[1], x15
+; CHECK-GI-NEXT: fmov d6, x10
+; CHECK-GI-NEXT: add v0.2d, v0.2d, v2.2d
+; CHECK-GI-NEXT: mul x14, x14, x17
+; CHECK-GI-NEXT: add v1.2d, v1.2d, v3.2d
+; CHECK-GI-NEXT: fmov d7, x11
+; CHECK-GI-NEXT: mov v7.d[1], x13
+; CHECK-GI-NEXT: mov v6.d[1], x14
+; CHECK-GI-NEXT: add v3.2d, v7.2d, v5.2d
+; CHECK-GI-NEXT: add v2.2d, v6.2d, v4.2d
+; CHECK-GI-NEXT: ret
+entry:
+ %s0s = sext <8 x i8> %s0 to <8 x i64>
+ %s1s = zext <8 x i8> %s1 to <8 x i64>
+ %m = mul <8 x i64> %s0s, %s1s
+ %a = add <8 x i64> %m, %b
+ ret <8 x i64> %a
+}
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; CHECK: {{.*}}
More information about the llvm-commits
mailing list