[AArch64] Handle any extend whilst lowering addw/addl/subw/subl

This adds an extra tablegen PatFrag, zanyext, which matches either any
extend or zext and uses that in the aarch64 backend to handle any
extends in addw/addl/subw/subl patterns.

Differential Revision: https://reviews.llvm.org/D93833

diff --git a/llvm/include/llvm/Target/TargetSelectionDAG.td b/llvm/include/llvm/Target/TargetSelectionDAG.td
index 0c6eef9..a1e961a 100644 (file)
--- a/llvm/include/llvm/Target/TargetSelectionDAG.td
+++ b/llvm/include/llvm/Target/TargetSelectionDAG.td
@@ -920,6 +920,10 @@ def not  : PatFrag<(ops node:$in), (xor node:$in, -1)>;
 def vnot : PatFrag<(ops node:$in), (xor node:$in, immAllOnesV)>;
 def ineg : PatFrag<(ops node:$in), (sub 0, node:$in)>;
 
+def zanyext : PatFrags<(ops node:$op),
+                       [(zext node:$op),
+                        (anyext node:$op)]>;
+
 // null_frag - The null pattern operator is used in multiclass instantiations
 // which accept an SDPatternOperator for use in matching patterns for internal
 // definitions. When expanding a pattern, if the null fragment is referenced

diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.td b/llvm/lib/Target/AArch64/AArch64InstrInfo.td
index 7e9f2fb..6209f51 100644 (file)
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.td
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.td
@@ -4765,18 +4765,18 @@ defm SSUBW   : SIMDWideThreeVectorBHS<0, 0b0011, "ssubw",
 defm UABAL   : SIMDLongThreeVectorTiedBHSabal<1, 0b0101, "uabal",
                                               AArch64uabd>;
 defm UADDL   : SIMDLongThreeVectorBHS<1, 0b0000, "uaddl",
-                 BinOpFrag<(add (zext node:$LHS), (zext node:$RHS))>>;
+                 BinOpFrag<(add (zanyext node:$LHS), (zanyext node:$RHS))>>;
 defm UADDW   : SIMDWideThreeVectorBHS<1, 0b0001, "uaddw",
-                 BinOpFrag<(add node:$LHS, (zext node:$RHS))>>;
+                 BinOpFrag<(add node:$LHS, (zanyext node:$RHS))>>;
 defm UMLAL   : SIMDLongThreeVectorTiedBHS<1, 0b1000, "umlal",
     TriOpFrag<(add node:$LHS, (int_aarch64_neon_umull node:$MHS, node:$RHS))>>;
 defm UMLSL   : SIMDLongThreeVectorTiedBHS<1, 0b1010, "umlsl",
     TriOpFrag<(sub node:$LHS, (int_aarch64_neon_umull node:$MHS, node:$RHS))>>;
 defm UMULL   : SIMDLongThreeVectorBHS<1, 0b1100, "umull", int_aarch64_neon_umull>;
 defm USUBL   : SIMDLongThreeVectorBHS<1, 0b0010, "usubl",
-                 BinOpFrag<(sub (zext node:$LHS), (zext node:$RHS))>>;
+                 BinOpFrag<(sub (zanyext node:$LHS), (zanyext node:$RHS))>>;
 defm USUBW   : SIMDWideThreeVectorBHS<   1, 0b0011, "usubw",
-                 BinOpFrag<(sub node:$LHS, (zext node:$RHS))>>;
+                 BinOpFrag<(sub node:$LHS, (zanyext node:$RHS))>>;
 
 // Additional patterns for SMULL and UMULL
 multiclass Neon_mul_widen_patterns<SDPatternOperator opnode,

diff --git a/llvm/test/CodeGen/AArch64/arm64-neon-3vdiff.ll b/llvm/test/CodeGen/AArch64/arm64-neon-3vdiff.ll
index 344b047..17c70ac 100644 (file)
--- a/llvm/test/CodeGen/AArch64/arm64-neon-3vdiff.ll
+++ b/llvm/test/CodeGen/AArch64/arm64-neon-3vdiff.ll
@@ -103,9 +103,7 @@ entry:
 define <8 x i16> @test_vaddl_a8(<8 x i8> %a, <8 x i8> %b) {
 ; CHECK-LABEL: test_vaddl_a8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v0.8h, v0.8b, #0
-; CHECK-NEXT:    ushll v1.8h, v1.8b, #0
-; CHECK-NEXT:    add v0.8h, v0.8h, v1.8h
+; CHECK-NEXT:    uaddl v0.8h, v0.8b, v1.8b
 ; CHECK-NEXT:    bic v0.8h, #255, lsl #8
 ; CHECK-NEXT:    ret
 entry:
@@ -119,9 +117,7 @@ entry:
 define <4 x i32> @test_vaddl_a16(<4 x i16> %a, <4 x i16> %b) {
 ; CHECK-LABEL: test_vaddl_a16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v0.4s, v0.4h, #0
-; CHECK-NEXT:    ushll v1.4s, v1.4h, #0
-; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    uaddl v0.4s, v0.4h, v1.4h
 ; CHECK-NEXT:    movi v1.2d, #0x00ffff0000ffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -136,9 +132,7 @@ entry:
 define <2 x i64> @test_vaddl_a32(<2 x i32> %a, <2 x i32> %b) {
 ; CHECK-LABEL: test_vaddl_a32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v0.2d, v0.2s, #0
-; CHECK-NEXT:    ushll v1.2d, v1.2s, #0
-; CHECK-NEXT:    add v0.2d, v0.2d, v1.2d
+; CHECK-NEXT:    uaddl v0.2d, v0.2s, v1.2s
 ; CHECK-NEXT:    movi v1.2d, #0x000000ffffffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -237,9 +231,7 @@ entry:
 define <8 x i16> @test_vaddl_high_a8(<16 x i8> %a, <16 x i8> %b) {
 ; CHECK-LABEL: test_vaddl_high_a8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll2 v0.8h, v0.16b, #0
-; CHECK-NEXT:    ushll2 v1.8h, v1.16b, #0
-; CHECK-NEXT:    add v0.8h, v0.8h, v1.8h
+; CHECK-NEXT:    uaddl2 v0.8h, v0.16b, v1.16b
 ; CHECK-NEXT:    bic v0.8h, #255, lsl #8
 ; CHECK-NEXT:    ret
 entry:
@@ -255,9 +247,7 @@ entry:
 define <4 x i32> @test_vaddl_high_a16(<8 x i16> %a, <8 x i16> %b) {
 ; CHECK-LABEL: test_vaddl_high_a16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll2 v0.4s, v0.8h, #0
-; CHECK-NEXT:    ushll2 v1.4s, v1.8h, #0
-; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    uaddl2 v0.4s, v0.8h, v1.8h
 ; CHECK-NEXT:    movi v1.2d, #0x00ffff0000ffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -274,9 +264,7 @@ entry:
 define <2 x i64> @test_vaddl_high_a32(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: test_vaddl_high_a32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll2 v0.2d, v0.4s, #0
-; CHECK-NEXT:    ushll2 v1.2d, v1.4s, #0
-; CHECK-NEXT:    add v0.2d, v0.2d, v1.2d
+; CHECK-NEXT:    uaddl2 v0.2d, v0.4s, v1.4s
 ; CHECK-NEXT:    movi v1.2d, #0x000000ffffffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -359,8 +347,7 @@ entry:
 define <8 x i16> @test_vaddw_a8(<8 x i16> %a, <8 x i8> %b) {
 ; CHECK-LABEL: test_vaddw_a8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v1.8h, v1.8b, #0
-; CHECK-NEXT:    add v0.8h, v1.8h, v0.8h
+; CHECK-NEXT:    uaddw v0.8h, v0.8h, v1.8b
 ; CHECK-NEXT:    bic v0.8h, #255, lsl #8
 ; CHECK-NEXT:    ret
 entry:
@@ -373,8 +360,7 @@ entry:
 define <4 x i32> @test_vaddw_a16(<4 x i32> %a, <4 x i16> %b) {
 ; CHECK-LABEL: test_vaddw_a16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v1.4s, v1.4h, #0
-; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
+; CHECK-NEXT:    uaddw v0.4s, v0.4s, v1.4h
 ; CHECK-NEXT:    movi v1.2d, #0x00ffff0000ffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -388,8 +374,7 @@ entry:
 define <2 x i64> @test_vaddw_a32(<2 x i64> %a, <2 x i32> %b) {
 ; CHECK-LABEL: test_vaddw_a32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v1.2d, v1.2s, #0
-; CHECK-NEXT:    add v0.2d, v1.2d, v0.2d
+; CHECK-NEXT:    uaddw v0.2d, v0.2d, v1.2s
 ; CHECK-NEXT:    movi v1.2d, #0x000000ffffffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -475,8 +460,7 @@ entry:
 define <8 x i16> @test_vaddw_high_a8(<8 x i16> %a, <16 x i8> %b) {
 ; CHECK-LABEL: test_vaddw_high_a8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll2 v1.8h, v1.16b, #0
-; CHECK-NEXT:    add v0.8h, v1.8h, v0.8h
+; CHECK-NEXT:    uaddw2 v0.8h, v0.8h, v1.16b
 ; CHECK-NEXT:    bic v0.8h, #255, lsl #8
 ; CHECK-NEXT:    ret
 entry:
@@ -490,8 +474,7 @@ entry:
 define <4 x i32> @test_vaddw_high_a16(<4 x i32> %a, <8 x i16> %b) {
 ; CHECK-LABEL: test_vaddw_high_a16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll2 v1.4s, v1.8h, #0
-; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
+; CHECK-NEXT:    uaddw2 v0.4s, v0.4s, v1.8h
 ; CHECK-NEXT:    movi v1.2d, #0x00ffff0000ffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -506,8 +489,7 @@ entry:
 define <2 x i64> @test_vaddw_high_a32(<2 x i64> %a, <4 x i32> %b) {
 ; CHECK-LABEL: test_vaddw_high_a32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll2 v1.2d, v1.4s, #0
-; CHECK-NEXT:    add v0.2d, v1.2d, v0.2d
+; CHECK-NEXT:    uaddw2 v0.2d, v0.2d, v1.4s
 ; CHECK-NEXT:    movi v1.2d, #0x000000ffffffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -594,9 +576,7 @@ entry:
 define <8 x i16> @test_vsubl_a8(<8 x i8> %a, <8 x i8> %b) {
 ; CHECK-LABEL: test_vsubl_a8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v0.8h, v0.8b, #0
-; CHECK-NEXT:    ushll v1.8h, v1.8b, #0
-; CHECK-NEXT:    sub v0.8h, v0.8h, v1.8h
+; CHECK-NEXT:    usubl v0.8h, v0.8b, v1.8b
 ; CHECK-NEXT:    bic v0.8h, #255, lsl #8
 ; CHECK-NEXT:    ret
 entry:
@@ -610,9 +590,7 @@ entry:
 define <4 x i32> @test_vsubl_a16(<4 x i16> %a, <4 x i16> %b) {
 ; CHECK-LABEL: test_vsubl_a16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v0.4s, v0.4h, #0
-; CHECK-NEXT:    ushll v1.4s, v1.4h, #0
-; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    usubl v0.4s, v0.4h, v1.4h
 ; CHECK-NEXT:    movi v1.2d, #0x00ffff0000ffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -627,9 +605,7 @@ entry:
 define <2 x i64> @test_vsubl_a32(<2 x i32> %a, <2 x i32> %b) {
 ; CHECK-LABEL: test_vsubl_a32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v0.2d, v0.2s, #0
-; CHECK-NEXT:    ushll v1.2d, v1.2s, #0
-; CHECK-NEXT:    sub v0.2d, v0.2d, v1.2d
+; CHECK-NEXT:    usubl v0.2d, v0.2s, v1.2s
 ; CHECK-NEXT:    movi v1.2d, #0x000000ffffffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -728,9 +704,7 @@ entry:
 define <8 x i16> @test_vsubl_high_a8(<16 x i8> %a, <16 x i8> %b) {
 ; CHECK-LABEL: test_vsubl_high_a8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll2 v0.8h, v0.16b, #0
-; CHECK-NEXT:    ushll2 v1.8h, v1.16b, #0
-; CHECK-NEXT:    sub v0.8h, v0.8h, v1.8h
+; CHECK-NEXT:    usubl2 v0.8h, v0.16b, v1.16b
 ; CHECK-NEXT:    bic v0.8h, #255, lsl #8
 ; CHECK-NEXT:    ret
 entry:
@@ -746,9 +720,7 @@ entry:
 define <4 x i32> @test_vsubl_high_a16(<8 x i16> %a, <8 x i16> %b) {
 ; CHECK-LABEL: test_vsubl_high_a16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll2 v0.4s, v0.8h, #0
-; CHECK-NEXT:    ushll2 v1.4s, v1.8h, #0
-; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    usubl2 v0.4s, v0.8h, v1.8h
 ; CHECK-NEXT:    movi v1.2d, #0x00ffff0000ffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -765,9 +737,7 @@ entry:
 define <2 x i64> @test_vsubl_high_a32(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: test_vsubl_high_a32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll2 v0.2d, v0.4s, #0
-; CHECK-NEXT:    ushll2 v1.2d, v1.4s, #0
-; CHECK-NEXT:    sub v0.2d, v0.2d, v1.2d
+; CHECK-NEXT:    usubl2 v0.2d, v0.4s, v1.4s
 ; CHECK-NEXT:    movi v1.2d, #0x000000ffffffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -850,8 +820,7 @@ entry:
 define <8 x i16> @test_vsubw_a8(<8 x i16> %a, <8 x i8> %b) {
 ; CHECK-LABEL: test_vsubw_a8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v1.8h, v1.8b, #0
-; CHECK-NEXT:    sub v0.8h, v0.8h, v1.8h
+; CHECK-NEXT:    usubw v0.8h, v0.8h, v1.8b
 ; CHECK-NEXT:    bic v0.8h, #255, lsl #8
 ; CHECK-NEXT:    ret
 entry:
@@ -864,8 +833,7 @@ entry:
 define <4 x i32> @test_vsubw_a16(<4 x i32> %a, <4 x i16> %b) {
 ; CHECK-LABEL: test_vsubw_a16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v1.4s, v1.4h, #0
-; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    usubw v0.4s, v0.4s, v1.4h
 ; CHECK-NEXT:    movi v1.2d, #0x00ffff0000ffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -879,8 +847,7 @@ entry:
 define <2 x i64> @test_vsubw_a32(<2 x i64> %a, <2 x i32> %b) {
 ; CHECK-LABEL: test_vsubw_a32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v1.2d, v1.2s, #0
-; CHECK-NEXT:    sub v0.2d, v0.2d, v1.2d
+; CHECK-NEXT:    usubw v0.2d, v0.2d, v1.2s
 ; CHECK-NEXT:    movi v1.2d, #0x000000ffffffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -966,8 +933,7 @@ entry:
 define <8 x i16> @test_vsubw_high_a8(<8 x i16> %a, <16 x i8> %b) {
 ; CHECK-LABEL: test_vsubw_high_a8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll2 v1.8h, v1.16b, #0
-; CHECK-NEXT:    sub v0.8h, v0.8h, v1.8h
+; CHECK-NEXT:    usubw2 v0.8h, v0.8h, v1.16b
 ; CHECK-NEXT:    bic v0.8h, #255, lsl #8
 ; CHECK-NEXT:    ret
 entry:
@@ -981,8 +947,7 @@ entry:
 define <4 x i32> @test_vsubw_high_a16(<4 x i32> %a, <8 x i16> %b) {
 ; CHECK-LABEL: test_vsubw_high_a16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll2 v1.4s, v1.8h, #0
-; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    usubw2 v0.4s, v0.4s, v1.8h
 ; CHECK-NEXT:    movi v1.2d, #0x00ffff0000ffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -997,8 +962,7 @@ entry:
 define <2 x i64> @test_vsubw_high_a32(<2 x i64> %a, <4 x i32> %b) {
 ; CHECK-LABEL: test_vsubw_high_a32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll2 v1.2d, v1.4s, #0
-; CHECK-NEXT:    sub v0.2d, v0.2d, v1.2d
+; CHECK-NEXT:    usubw2 v0.2d, v0.2d, v1.4s
 ; CHECK-NEXT:    movi v1.2d, #0x000000ffffffff
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret

diff --git a/llvm/test/CodeGen/AArch64/lowerMUL-newload.ll b/llvm/test/CodeGen/AArch64/lowerMUL-newload.ll
index b8d0afc..116db96 100644 (file)
--- a/llvm/test/CodeGen/AArch64/lowerMUL-newload.ll
+++ b/llvm/test/CodeGen/AArch64/lowerMUL-newload.ll
@@ -21,10 +21,10 @@ entry:
 define <4 x i32> @mlai16_and(<4 x i16> %vec0, <4 x i16> %vec1, <4 x i16> %vec2) {
 ; CHECK-LABEL: mlai16_and:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v2.4s, v2.4h, #0
-; CHECK-NEXT:    umlal v2.4s, v1.4h, v0.4h
-; CHECK-NEXT:    movi v0.2d, #0x00ffff0000ffff
-; CHECK-NEXT:    and v0.16b, v2.16b, v0.16b
+; CHECK-NEXT:    umull v0.4s, v1.4h, v0.4h
+; CHECK-NEXT:    uaddw v0.4s, v0.4s, v2.4h
+; CHECK-NEXT:    movi v1.2d, #0x00ffff0000ffff
+; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
 entry:
   %v0 = sext <4 x i16> %vec0 to <4 x i32>
@@ -157,10 +157,10 @@ entry:
 define <2 x i64> @mlai32_and(<2 x i32> %vec0, <2 x i32> %vec1, <2 x i32> %vec2) {
 ; CHECK-LABEL: mlai32_and:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v2.2d, v2.2s, #0
-; CHECK-NEXT:    umlal v2.2d, v1.2s, v0.2s
-; CHECK-NEXT:    movi v0.2d, #0x000000ffffffff
-; CHECK-NEXT:    and v0.16b, v2.16b, v0.16b
+; CHECK-NEXT:    umull v0.2d, v1.2s, v0.2s
+; CHECK-NEXT:    uaddw v0.2d, v0.2d, v2.2s
+; CHECK-NEXT:    movi v1.2d, #0x000000ffffffff
+; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
 entry:
   %v0 = sext <2 x i32> %vec0 to <2 x i64>