Skip to content

Commit

Permalink
[AArch64][GlobalISel] Select SHL(ZEXT, DUP imm) into {U/S}HLL imm
Browse files Browse the repository at this point in the history
  • Loading branch information
chuongg3 committed Jun 26, 2024
1 parent f29cb3b commit c0a6902
Show file tree
Hide file tree
Showing 2 changed files with 50 additions and 72 deletions.
14 changes: 14 additions & 0 deletions llvm/lib/Target/AArch64/AArch64InstrInfo.td
Original file line number Diff line number Diff line change
Expand Up @@ -7879,6 +7879,20 @@ def : Pat<(v4i32 (concat_vectors (v2i32 V64:$Rd),
(SHRNv4i32_shift (INSERT_SUBREG (IMPLICIT_DEF), V64:$Rd, dsub),
V128:$Rn, vecshiftR32Narrow:$imm)>;

def : Pat<(shl (v8i16 (zext (v8i8 V64:$Rm))), (v8i16 (AArch64dup (i32 imm:$size)))),
(USHLLv8i8_shift V64:$Rm, (i32 imm:$size))>;
def : Pat<(shl (v4i32 (zext (v4i16 V64:$Rm))), (v4i32 (AArch64dup (i32 imm:$size)))),
(USHLLv4i16_shift V64:$Rm, (i32 imm:$size))>;
def : Pat<(shl (v2i64 (zext (v2i32 V64:$Rm))), (v2i64 (AArch64dup (i64 imm:$size)))),
(USHLLv2i32_shift V64:$Rm, (trunc_imm imm:$size))>;

def : Pat<(shl (v8i16 (sext (v8i8 V64:$Rm))), (v8i16 (AArch64dup (i32 imm:$size)))),
(SSHLLv8i8_shift V64:$Rm, (i32 imm:$size))>;
def : Pat<(shl (v4i32 (sext (v4i16 V64:$Rm))), (v4i32 (AArch64dup (i32 imm:$size)))),
(SSHLLv4i16_shift V64:$Rm, (i32 imm:$size))>;
def : Pat<(shl (v2i64 (sext (v2i32 V64:$Rm))), (v2i64 (AArch64dup (i64 imm:$size)))),
(SSHLLv2i32_shift V64:$Rm, (trunc_imm imm:$size))>;

// Vector sign and zero extensions are implemented with SSHLL and USSHLL.
// Anyexts are implemented as zexts.
def : Pat<(v8i16 (sext (v8i8 V64:$Rn))), (SSHLLv8i8_shift V64:$Rn, (i32 0))>;
Expand Down
108 changes: 36 additions & 72 deletions llvm/test/CodeGen/AArch64/neon-shift-left-long.ll
Original file line number Diff line number Diff line change
Expand Up @@ -3,96 +3,60 @@
; RUN: llc < %s -verify-machineinstrs -mtriple=aarch64-none-linux-gnu -mattr=+neon -global-isel | FileCheck %s --check-prefixes=CHECK,CHECK-GI

define <8 x i16> @test_sshll_v8i8(<8 x i8> %a) {
; CHECK-SD-LABEL: test_sshll_v8i8:
; CHECK-SD: // %bb.0:
; CHECK-SD-NEXT: sshll v0.8h, v0.8b, #3
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: test_sshll_v8i8:
; CHECK-GI: // %bb.0:
; CHECK-GI-NEXT: sshll v0.8h, v0.8b, #0
; CHECK-GI-NEXT: shl v0.8h, v0.8h, #3
; CHECK-GI-NEXT: ret
; CHECK-LABEL: test_sshll_v8i8:
; CHECK: // %bb.0:
; CHECK-NEXT: sshll v0.8h, v0.8b, #3
; CHECK-NEXT: ret
%1 = sext <8 x i8> %a to <8 x i16>
%tmp = shl <8 x i16> %1, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
ret <8 x i16> %tmp
}

define <4 x i32> @test_sshll_v4i16(<4 x i16> %a) {
; CHECK-SD-LABEL: test_sshll_v4i16:
; CHECK-SD: // %bb.0:
; CHECK-SD-NEXT: sshll v0.4s, v0.4h, #9
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: test_sshll_v4i16:
; CHECK-GI: // %bb.0:
; CHECK-GI-NEXT: sshll v0.4s, v0.4h, #0
; CHECK-GI-NEXT: shl v0.4s, v0.4s, #9
; CHECK-GI-NEXT: ret
; CHECK-LABEL: test_sshll_v4i16:
; CHECK: // %bb.0:
; CHECK-NEXT: sshll v0.4s, v0.4h, #9
; CHECK-NEXT: ret
%1 = sext <4 x i16> %a to <4 x i32>
%tmp = shl <4 x i32> %1, <i32 9, i32 9, i32 9, i32 9>
ret <4 x i32> %tmp
}

define <2 x i64> @test_sshll_v2i32(<2 x i32> %a) {
; CHECK-SD-LABEL: test_sshll_v2i32:
; CHECK-SD: // %bb.0:
; CHECK-SD-NEXT: sshll v0.2d, v0.2s, #19
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: test_sshll_v2i32:
; CHECK-GI: // %bb.0:
; CHECK-GI-NEXT: sshll v0.2d, v0.2s, #0
; CHECK-GI-NEXT: shl v0.2d, v0.2d, #19
; CHECK-GI-NEXT: ret
; CHECK-LABEL: test_sshll_v2i32:
; CHECK: // %bb.0:
; CHECK-NEXT: sshll v0.2d, v0.2s, #19
; CHECK-NEXT: ret
%1 = sext <2 x i32> %a to <2 x i64>
%tmp = shl <2 x i64> %1, <i64 19, i64 19>
ret <2 x i64> %tmp
}

define <8 x i16> @test_ushll_v8i8(<8 x i8> %a) {
; CHECK-SD-LABEL: test_ushll_v8i8:
; CHECK-SD: // %bb.0:
; CHECK-SD-NEXT: ushll v0.8h, v0.8b, #3
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: test_ushll_v8i8:
; CHECK-GI: // %bb.0:
; CHECK-GI-NEXT: ushll v0.8h, v0.8b, #0
; CHECK-GI-NEXT: shl v0.8h, v0.8h, #3
; CHECK-GI-NEXT: ret
; CHECK-LABEL: test_ushll_v8i8:
; CHECK: // %bb.0:
; CHECK-NEXT: ushll v0.8h, v0.8b, #3
; CHECK-NEXT: ret
%1 = zext <8 x i8> %a to <8 x i16>
%tmp = shl <8 x i16> %1, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
ret <8 x i16> %tmp
}

define <4 x i32> @test_ushll_v4i16(<4 x i16> %a) {
; CHECK-SD-LABEL: test_ushll_v4i16:
; CHECK-SD: // %bb.0:
; CHECK-SD-NEXT: ushll v0.4s, v0.4h, #9
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: test_ushll_v4i16:
; CHECK-GI: // %bb.0:
; CHECK-GI-NEXT: ushll v0.4s, v0.4h, #0
; CHECK-GI-NEXT: shl v0.4s, v0.4s, #9
; CHECK-GI-NEXT: ret
; CHECK-LABEL: test_ushll_v4i16:
; CHECK: // %bb.0:
; CHECK-NEXT: ushll v0.4s, v0.4h, #9
; CHECK-NEXT: ret
%1 = zext <4 x i16> %a to <4 x i32>
%tmp = shl <4 x i32> %1, <i32 9, i32 9, i32 9, i32 9>
ret <4 x i32> %tmp
}

define <2 x i64> @test_ushll_v2i32(<2 x i32> %a) {
; CHECK-SD-LABEL: test_ushll_v2i32:
; CHECK-SD: // %bb.0:
; CHECK-SD-NEXT: ushll v0.2d, v0.2s, #19
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: test_ushll_v2i32:
; CHECK-GI: // %bb.0:
; CHECK-GI-NEXT: ushll v0.2d, v0.2s, #0
; CHECK-GI-NEXT: shl v0.2d, v0.2d, #19
; CHECK-GI-NEXT: ret
; CHECK-LABEL: test_ushll_v2i32:
; CHECK: // %bb.0:
; CHECK-NEXT: ushll v0.2d, v0.2s, #19
; CHECK-NEXT: ret
%1 = zext <2 x i32> %a to <2 x i64>
%tmp = shl <2 x i64> %1, <i64 19, i64 19>
ret <2 x i64> %tmp
Expand All @@ -106,8 +70,8 @@ define <8 x i16> @test_sshll2_v16i8(<16 x i8> %a) {
;
; CHECK-GI-LABEL: test_sshll2_v16i8:
; CHECK-GI: // %bb.0:
; CHECK-GI-NEXT: sshll2 v0.8h, v0.16b, #0
; CHECK-GI-NEXT: shl v0.8h, v0.8h, #3
; CHECK-GI-NEXT: mov d0, v0.d[1]
; CHECK-GI-NEXT: sshll v0.8h, v0.8b, #3
; CHECK-GI-NEXT: ret
%1 = shufflevector <16 x i8> %a, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
%2 = sext <8 x i8> %1 to <8 x i16>
Expand All @@ -123,8 +87,8 @@ define <4 x i32> @test_sshll2_v8i16(<8 x i16> %a) {
;
; CHECK-GI-LABEL: test_sshll2_v8i16:
; CHECK-GI: // %bb.0:
; CHECK-GI-NEXT: sshll2 v0.4s, v0.8h, #0
; CHECK-GI-NEXT: shl v0.4s, v0.4s, #9
; CHECK-GI-NEXT: mov d0, v0.d[1]
; CHECK-GI-NEXT: sshll v0.4s, v0.4h, #9
; CHECK-GI-NEXT: ret
%1 = shufflevector <8 x i16> %a, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
%2 = sext <4 x i16> %1 to <4 x i32>
Expand All @@ -140,8 +104,8 @@ define <2 x i64> @test_sshll2_v4i32(<4 x i32> %a) {
;
; CHECK-GI-LABEL: test_sshll2_v4i32:
; CHECK-GI: // %bb.0:
; CHECK-GI-NEXT: sshll2 v0.2d, v0.4s, #0
; CHECK-GI-NEXT: shl v0.2d, v0.2d, #19
; CHECK-GI-NEXT: mov d0, v0.d[1]
; CHECK-GI-NEXT: sshll v0.2d, v0.2s, #19
; CHECK-GI-NEXT: ret
%1 = shufflevector <4 x i32> %a, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
%2 = sext <2 x i32> %1 to <2 x i64>
Expand All @@ -157,8 +121,8 @@ define <8 x i16> @test_ushll2_v16i8(<16 x i8> %a) {
;
; CHECK-GI-LABEL: test_ushll2_v16i8:
; CHECK-GI: // %bb.0:
; CHECK-GI-NEXT: ushll2 v0.8h, v0.16b, #0
; CHECK-GI-NEXT: shl v0.8h, v0.8h, #3
; CHECK-GI-NEXT: mov d0, v0.d[1]
; CHECK-GI-NEXT: ushll v0.8h, v0.8b, #3
; CHECK-GI-NEXT: ret
%1 = shufflevector <16 x i8> %a, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
%2 = zext <8 x i8> %1 to <8 x i16>
Expand All @@ -174,8 +138,8 @@ define <4 x i32> @test_ushll2_v8i16(<8 x i16> %a) {
;
; CHECK-GI-LABEL: test_ushll2_v8i16:
; CHECK-GI: // %bb.0:
; CHECK-GI-NEXT: ushll2 v0.4s, v0.8h, #0
; CHECK-GI-NEXT: shl v0.4s, v0.4s, #9
; CHECK-GI-NEXT: mov d0, v0.d[1]
; CHECK-GI-NEXT: ushll v0.4s, v0.4h, #9
; CHECK-GI-NEXT: ret
%1 = shufflevector <8 x i16> %a, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
%2 = zext <4 x i16> %1 to <4 x i32>
Expand All @@ -191,8 +155,8 @@ define <2 x i64> @test_ushll2_v4i32(<4 x i32> %a) {
;
; CHECK-GI-LABEL: test_ushll2_v4i32:
; CHECK-GI: // %bb.0:
; CHECK-GI-NEXT: ushll2 v0.2d, v0.4s, #0
; CHECK-GI-NEXT: shl v0.2d, v0.2d, #19
; CHECK-GI-NEXT: mov d0, v0.d[1]
; CHECK-GI-NEXT: ushll v0.2d, v0.2s, #19
; CHECK-GI-NEXT: ret
%1 = shufflevector <4 x i32> %a, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
%2 = zext <2 x i32> %1 to <2 x i64>
Expand Down

0 comments on commit c0a6902

Please sign in to comment.