[PATCH] [AArch64 NEON] Add patterns for concat_vector on v2i32.

[Kevin Qin]

Hi,

Please review.

http://llvm-reviews.chandlerc.com/D2612

Files:
  lib/Target/AArch64/AArch64InstrNEON.td
  test/CodeGen/AArch64/neon-copy.ll

Index: lib/Target/AArch64/AArch64InstrNEON.td
===================================================================
--- lib/Target/AArch64/AArch64InstrNEON.td
+++ lib/Target/AArch64/AArch64InstrNEON.td
@@ -7070,10 +7070,18 @@
 defm : Concat_Vector_Pattern<v4f32, v2f32>;
 defm : Concat_Vector_Pattern<v2f64, v1f64>;
 
+def : Pat<(v2i32 (concat_vectors (v1i32 FPR32:$Rn), undef)),
+          (v2i32 (SUBREG_TO_REG(i64 0), $Rn, sub_32))>;
+def : Pat<(v2i32 (concat_vectors (v1i32 FPR32:$Rn), (v1i32 FPR32:$Rm))),
+          (EXTRACT_SUBREG 
+            (v4i32 (INSELs
+              (v4i32 (SUBREG_TO_REG (i64 0), FPR32:$Rn, sub_32)),
+              (v4i32 (SUBREG_TO_REG (i64 0), FPR32:$Rm, sub_32)),
+              (i64 1),
+              (i64 0))),
+            sub_64)>;
 def : Pat<(v2i32 (concat_vectors (v1i32 FPR32:$Rn), (v1i32 FPR32:$Rn))),
           (DUPELT2s (v4i32 (SUBREG_TO_REG(i64 0), $Rn, sub_32)), 0)>;
-def : Pat<(v2i32 (concat_vectors undef, (v1i32 FPR32:$Rn))),
-          (DUPELT2s (v4i32 (SUBREG_TO_REG(i64 0), $Rn, sub_32)), 0)>;
 
 //patterns for EXTRACT_SUBVECTOR
 def : Pat<(v8i8 (extract_subvector (v16i8 VPR128:$Rn), (i64 0))),
Index: test/CodeGen/AArch64/neon-copy.ll
===================================================================
--- test/CodeGen/AArch64/neon-copy.ll
+++ test/CodeGen/AArch64/neon-copy.ll
@@ -948,25 +948,6 @@
   ret <2 x i32> %vecinit1.i
 }
 
-define <2 x i32> @test_concat_undef_v1i32(<1 x i32> %a) {
-; CHECK-LABEL: test_concat_undef_v1i32:
-; CHECK: dup v{{[0-9]+}}.2s, v{{[0-9]+}}.s[0]
-entry:
-  %0 = extractelement <1 x i32> %a, i32 0
-  %vecinit1.i = insertelement <2 x i32> undef, i32 %0, i32 1
-  ret <2 x i32> %vecinit1.i
-}
-
-define <2 x i32> @test_concat_v1i32_v1i32(<1 x i32> %a) {
-; CHECK-LABEL: test_concat_v1i32_v1i32:
-; CHECK: dup v{{[0-9]+}}.2s, v{{[0-9]+}}.s[0]
-entry:
-  %0 = extractelement <1 x i32> %a, i32 0
-  %vecinit.i = insertelement <2 x i32> undef, i32 %0, i32 0
-  %vecinit1.i = insertelement <2 x i32> %vecinit.i, i32 %0, i32 1
-  ret <2 x i32> %vecinit1.i
-}
-
 
 define <2 x float> @test_scalar_to_vector_f32_to_v2f32(<2 x float> %a) {
 ; CHECK-LABEL: test_scalar_to_vector_f32_to_v2f32:
@@ -1002,6 +983,52 @@
   ret <16 x i8> %vecinit30
 }
 
+define <2 x i32> @test_concat_undef_v1i32(<1 x i32> %a) {
+; CHECK-LABEL: test_concat_undef_v1i32:
+; CHECK: ins v{{[0-9]+}}.s[1], v{{[0-9]+}}.s[0]
+entry:
+  %0 = extractelement <1 x i32> %a, i32 0
+  %vecinit1.i = insertelement <2 x i32> undef, i32 %0, i32 1
+  ret <2 x i32> %vecinit1.i
+}
+
+declare <1 x i32> @llvm.arm.neon.vqabs.v1i32(<1 x i32>) #4
+
+define <2 x i32> @test_concat_v1i32_undef(<1 x i32> %a) {
+; CHECK-LABEL: test_concat_v1i32_undef:
+; CHECK: sqabs s{{[0-9]+}}, s{{[0-9]+}}
+; CHECK-NEXT: ret
+entry:
+  %b = tail call <1 x i32> @llvm.arm.neon.vqabs.v1i32(<1 x i32> %a)
+  %0 = extractelement <1 x i32> %b, i32 0
+  %vecinit.i432 = insertelement <2 x i32> undef, i32 %0, i32 0
+  ret <2 x i32> %vecinit.i432
+}
+
+define <2 x i32> @test_concat_same_v1i32_v1i32(<1 x i32> %a) {
+; CHECK-LABEL: test_concat_same_v1i32_v1i32:
+; CHECK: dup v{{[0-9]+}}.2s, v{{[0-9]+}}.s[0]
+entry:
+  %0 = extractelement <1 x i32> %a, i32 0
+  %vecinit.i = insertelement <2 x i32> undef, i32 %0, i32 0
+  %vecinit1.i = insertelement <2 x i32> %vecinit.i, i32 %0, i32 1
+  ret <2 x i32> %vecinit1.i
+}
+
+define <2 x i32> @test_concat_diff_v1i32_v1i32(<1 x i32> %a, <1 x i32> %b) {
+; CHECK-LABEL: test_concat_diff_v1i32_v1i32:
+; CHECK: sqabs s{{[0-9]+}}, s{{[0-9]+}}
+; CHECK-NEXT: sqabs s{{[0-9]+}}, s{{[0-9]+}}
+; CHECK-NEXT: ins v0.s[1], v1.s[0]
+entry:
+  %c = tail call <1 x i32> @llvm.arm.neon.vqabs.v1i32(<1 x i32> %a)
+  %d = extractelement <1 x i32> %c, i32 0
+  %e = tail call <1 x i32> @llvm.arm.neon.vqabs.v1i32(<1 x i32> %b)
+  %f = extractelement <1 x i32> %e, i32 0
+  %h = shufflevector <1 x i32> %c, <1 x i32> %e, <2 x i32> <i32 0, i32 1>
+  ret <2 x i32> %h
+}
+
 define <16 x i8> @test_concat_v16i8_v8i8_v16i8(<8 x i8> %x, <16 x i8> %y) #0 {
 ; CHECK-LABEL: test_concat_v16i8_v8i8_v16i8:
 ; CHECK: ins {{v[0-9]+}}.d[1], {{v[0-9]+}}.d[0]
@@ -1262,4 +1289,5 @@
   %vecext1 = extractelement <1 x i64> %y, i32 0
   %vecinit2 = insertelement <2 x i64> %vecinit, i64 %vecext1, i32 1
   ret <2 x i64> %vecinit2
-}
\ No newline at end of file
+}
+
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D2612.1.patch
Type: text/x-patch
Size: 4295 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140124/8021e8e2/attachment.bin>