[llvm] r180604 - ARM/NEON: Pattern match vector integer abs to vabs.

[Benjamin Kramer]

Author: d0k
Date: Fri Apr 26 10:00:57 2013
New Revision: 180604

URL: http://llvm.org/viewvc/llvm-project?rev=180604&view=rev
Log:
ARM/NEON: Pattern match vector integer abs to vabs.

Added:
    llvm/trunk/test/CodeGen/ARM/neon_vabs.ll
Modified:
    llvm/trunk/lib/Target/ARM/ARMInstrNEON.td

Modified: llvm/trunk/lib/Target/ARM/ARMInstrNEON.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrNEON.td?rev=180604&r1=180603&r2=180604&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrNEON.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrNEON.td Fri Apr 26 10:00:57 2013
@@ -4907,6 +4907,29 @@ def  VABSfq   : N2VQ<0b11, 0b11, 0b10, 0
                      "vabs", "f32",
                       v4f32, v4f32, fabs>;
 
+def : Pat<(xor (v2i32 (bitconvert (v8i8 (NEONvshrs DPR:$src, (i32 7))))),
+               (v2i32 (bitconvert (v8i8 (add DPR:$src,
+                                             (NEONvshrs DPR:$src, (i32 7))))))),
+          (VABSv8i8 DPR:$src)>;
+def : Pat<(xor (v2i32 (bitconvert (v4i16 (NEONvshrs DPR:$src, (i32 15))))),
+               (v2i32 (bitconvert (v4i16 (add DPR:$src,
+                                            (NEONvshrs DPR:$src, (i32 15))))))),
+          (VABSv4i16 DPR:$src)>;
+def : Pat<(xor (v2i32 (NEONvshrs DPR:$src, (i32 31))),
+               (v2i32 (add DPR:$src, (NEONvshrs DPR:$src, (i32 31))))),
+          (VABSv2i32 DPR:$src)>;
+def : Pat<(xor (v4i32 (bitconvert (v16i8 (NEONvshrs QPR:$src, (i32 7))))),
+               (v4i32 (bitconvert (v16i8 (add QPR:$src,
+                                             (NEONvshrs QPR:$src, (i32 7))))))),
+          (VABSv16i8 QPR:$src)>;
+def : Pat<(xor (v4i32 (bitconvert (v8i16 (NEONvshrs QPR:$src, (i32 15))))),
+               (v4i32 (bitconvert (v8i16 (add QPR:$src,
+                                            (NEONvshrs QPR:$src, (i32 15))))))),
+          (VABSv8i16 QPR:$src)>;
+def : Pat<(xor (v4i32 (NEONvshrs QPR:$src, (i32 31))),
+               (v4i32 (add QPR:$src, (NEONvshrs QPR:$src, (i32 31))))),
+          (VABSv4i32 QPR:$src)>;
+
 def : Pat<(v2f32 (int_arm_neon_vabs (v2f32 DPR:$src))), (VABSfd DPR:$src)>;
 def : Pat<(v4f32 (int_arm_neon_vabs (v4f32 QPR:$src))), (VABSfq QPR:$src)>;
 

Added: llvm/trunk/test/CodeGen/ARM/neon_vabs.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/neon_vabs.ll?rev=180604&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/ARM/neon_vabs.ll (added)
+++ llvm/trunk/test/CodeGen/ARM/neon_vabs.ll Fri Apr 26 10:00:57 2013
@@ -0,0 +1,91 @@
+; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s
+
+define <4 x i32> @test1(<4 x i32> %a) nounwind {
+; CHECK: test1:
+; CHECK: vabs.s32 q
+        %tmp1neg = sub <4 x i32> zeroinitializer, %a
+        %b = icmp sgt <4 x i32> %a, <i32 -1, i32 -1, i32 -1, i32 -1>
+        %abs = select <4 x i1> %b, <4 x i32> %a, <4 x i32> %tmp1neg
+        ret <4 x i32> %abs
+}
+
+define <4 x i32> @test2(<4 x i32> %a) nounwind {
+; CHECK: test2:
+; CHECK: vabs.s32 q
+        %tmp1neg = sub <4 x i32> zeroinitializer, %a
+        %b = icmp sge <4 x i32> %a, zeroinitializer
+        %abs = select <4 x i1> %b, <4 x i32> %a, <4 x i32> %tmp1neg
+        ret <4 x i32> %abs
+}
+
+define <8 x i16> @test3(<8 x i16> %a) nounwind {
+; CHECK: test3:
+; CHECK: vabs.s16 q
+        %tmp1neg = sub <8 x i16> zeroinitializer, %a
+        %b = icmp sgt <8 x i16> %a, zeroinitializer
+        %abs = select <8 x i1> %b, <8 x i16> %a, <8 x i16> %tmp1neg
+        ret <8 x i16> %abs
+}
+
+define <16 x i8> @test4(<16 x i8> %a) nounwind {
+; CHECK: test4:
+; CHECK: vabs.s8 q
+        %tmp1neg = sub <16 x i8> zeroinitializer, %a
+        %b = icmp slt <16 x i8> %a, zeroinitializer
+        %abs = select <16 x i1> %b, <16 x i8> %tmp1neg, <16 x i8> %a
+        ret <16 x i8> %abs
+}
+
+define <4 x i32> @test5(<4 x i32> %a) nounwind {
+; CHECK: test5:
+; CHECK: vabs.s32 q
+        %tmp1neg = sub <4 x i32> zeroinitializer, %a
+        %b = icmp sle <4 x i32> %a, zeroinitializer
+        %abs = select <4 x i1> %b, <4 x i32> %tmp1neg, <4 x i32> %a
+        ret <4 x i32> %abs
+}
+
+define <2 x i32> @test6(<2 x i32> %a) nounwind {
+; CHECK: test6:
+; CHECK: vabs.s32 d
+        %tmp1neg = sub <2 x i32> zeroinitializer, %a
+        %b = icmp sgt <2 x i32> %a, <i32 -1, i32 -1>
+        %abs = select <2 x i1> %b, <2 x i32> %a, <2 x i32> %tmp1neg
+        ret <2 x i32> %abs
+}
+
+define <2 x i32> @test7(<2 x i32> %a) nounwind {
+; CHECK: test7:
+; CHECK: vabs.s32 d
+        %tmp1neg = sub <2 x i32> zeroinitializer, %a
+        %b = icmp sge <2 x i32> %a, zeroinitializer
+        %abs = select <2 x i1> %b, <2 x i32> %a, <2 x i32> %tmp1neg
+        ret <2 x i32> %abs
+}
+
+define <4 x i16> @test8(<4 x i16> %a) nounwind {
+; CHECK: test8:
+; CHECK: vabs.s16 d
+        %tmp1neg = sub <4 x i16> zeroinitializer, %a
+        %b = icmp sgt <4 x i16> %a, zeroinitializer
+        %abs = select <4 x i1> %b, <4 x i16> %a, <4 x i16> %tmp1neg
+        ret <4 x i16> %abs
+}
+
+define <8 x i8> @test9(<8 x i8> %a) nounwind {
+; CHECK: test9:
+; CHECK: vabs.s8 d
+        %tmp1neg = sub <8 x i8> zeroinitializer, %a
+        %b = icmp slt <8 x i8> %a, zeroinitializer
+        %abs = select <8 x i1> %b, <8 x i8> %tmp1neg, <8 x i8> %a
+        ret <8 x i8> %abs
+}
+
+define <2 x i32> @test10(<2 x i32> %a) nounwind {
+; CHECK: test10:
+; CHECK: vabs.s32 d
+        %tmp1neg = sub <2 x i32> zeroinitializer, %a
+        %b = icmp sle <2 x i32> %a, zeroinitializer
+        %abs = select <2 x i1> %b, <2 x i32> %tmp1neg, <2 x i32> %a
+        ret <2 x i32> %abs
+}