From 54c7432e2259a35cd3656fc732c4ff187d0dcf6f Mon Sep 17 00:00:00 2001
From: Jim Grosbach <grosbach@apple.com>
Date: Mon, 15 Oct 2012 21:23:40 +0000
Subject: [PATCH] ARM: v1i64 and v2i64 VBSL intrinsic support.

rdar://12502028

llvm-svn: 165981
---
 llvm/lib/Target/ARM/ARMInstrNEON.td | 17 +++++++++++++++++
 llvm/test/CodeGen/ARM/vbsl.ll       | 30 ++++++++++++++++++++++++++++++
 2 files changed, 47 insertions(+)

diff --git a/llvm/lib/Target/ARM/ARMInstrNEON.td b/llvm/lib/Target/ARM/ARMInstrNEON.td
index 837c37d..ede4def 100644
--- a/llvm/lib/Target/ARM/ARMInstrNEON.td
+++ b/llvm/lib/Target/ARM/ARMInstrNEON.td
@@ -4504,12 +4504,21 @@ def : Pat<(v2f32 (int_arm_neon_vbsl (v2f32 DPR:$src1),
                                     (v2f32 DPR:$Vn), (v2f32 DPR:$Vm))),
           (VBSLd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
         Requires<[HasNEON]>;
+def : Pat<(v1i64 (int_arm_neon_vbsl (v1i64 DPR:$src1),
+                                    (v1i64 DPR:$Vn), (v1i64 DPR:$Vm))),
+          (VBSLd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
+        Requires<[HasNEON]>;
 
 def : Pat<(v2i32 (or (and DPR:$Vn, DPR:$Vd),
                      (and DPR:$Vm, (vnotd DPR:$Vd)))),
           (VBSLd DPR:$Vd, DPR:$Vn, DPR:$Vm)>,
         Requires<[HasNEON]>;
 
+def : Pat<(v1i64 (or (and DPR:$Vn, DPR:$Vd),
+                     (and DPR:$Vm, (vnotd DPR:$Vd)))),
+          (VBSLd DPR:$Vd, DPR:$Vn, DPR:$Vm)>,
+        Requires<[HasNEON]>;
+
 def  VBSLq    : N3VX<1, 0, 0b01, 0b0001, 1, 1, (outs QPR:$Vd),
                      (ins QPR:$src1, QPR:$Vn, QPR:$Vm),
                      N3RegFrm, IIC_VCNTiQ,
@@ -4533,11 +4542,19 @@ def : Pat<(v4f32 (int_arm_neon_vbsl (v4f32 QPR:$src1),
                                     (v4f32 QPR:$Vn), (v4f32 QPR:$Vm))),
           (VBSLq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
         Requires<[HasNEON]>;
+def : Pat<(v2i64 (int_arm_neon_vbsl (v2i64 QPR:$src1),
+                                    (v2i64 QPR:$Vn), (v2i64 QPR:$Vm))),
+          (VBSLq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
+        Requires<[HasNEON]>;
 
 def : Pat<(v4i32 (or (and QPR:$Vn, QPR:$Vd),
                      (and QPR:$Vm, (vnotq QPR:$Vd)))),
           (VBSLq QPR:$Vd, QPR:$Vn, QPR:$Vm)>,
         Requires<[HasNEON]>;
+def : Pat<(v2i64 (or (and QPR:$Vn, QPR:$Vd),
+                     (and QPR:$Vm, (vnotq QPR:$Vd)))),
+          (VBSLq QPR:$Vd, QPR:$Vn, QPR:$Vm)>,
+        Requires<[HasNEON]>;
 
 //   VBIF     : Vector Bitwise Insert if False
 //              like VBSL but with: "vbif $dst, $src3, $src1", "$src2 = $dst",
diff --git a/llvm/test/CodeGen/ARM/vbsl.ll b/llvm/test/CodeGen/ARM/vbsl.ll
index dca30eb..750fb0d 100644
--- a/llvm/test/CodeGen/ARM/vbsl.ll
+++ b/llvm/test/CodeGen/ARM/vbsl.ll
@@ -162,6 +162,34 @@ define <4 x float> @g4(<4 x float> %a, <4 x float> %b, <4 x float> %c) nounwind
   ret <4 x float> %vbsl4.i
 }
 
+define <1 x i64> @test_vbsl_s64(<1 x i64> %a, <1 x i64> %b, <1 x i64> %c) nounwind readnone optsize ssp {
+; CHECK: test_vbsl_s64:
+; CHECK: vbsl d
+  %vbsl3.i = tail call <1 x i64> @llvm.arm.neon.vbsl.v1i64(<1 x i64> %a, <1 x i64> %b, <1 x i64> %c) nounwind
+  ret <1 x i64> %vbsl3.i
+}
+
+define <1 x i64> @test_vbsl_u64(<1 x i64> %a, <1 x i64> %b, <1 x i64> %c) nounwind readnone optsize ssp {
+; CHECK: test_vbsl_u64:
+; CHECK: vbsl d
+  %vbsl3.i = tail call <1 x i64> @llvm.arm.neon.vbsl.v1i64(<1 x i64> %a, <1 x i64> %b, <1 x i64> %c) nounwind
+  ret <1 x i64> %vbsl3.i
+}
+
+define <2 x i64> @test_vbslq_s64(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) nounwind readnone optsize ssp {
+; CHECK: test_vbslq_s64:
+; CHECK: vbsl q
+  %vbsl3.i = tail call <2 x i64> @llvm.arm.neon.vbsl.v2i64(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) nounwind
+  ret <2 x i64> %vbsl3.i
+}
+
+define <2 x i64> @test_vbslq_u64(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) nounwind readnone optsize ssp {
+; CHECK: test_vbslq_u64:
+; CHECK: vbsl q
+  %vbsl3.i = tail call <2 x i64> @llvm.arm.neon.vbsl.v2i64(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) nounwind
+  ret <2 x i64> %vbsl3.i
+}
+
 declare <4 x i32> @llvm.arm.neon.vbsl.v4i32(<4 x i32>, <4 x i32>, <4 x i32>) nounwind readnone
 declare <8 x i16> @llvm.arm.neon.vbsl.v8i16(<8 x i16>, <8 x i16>, <8 x i16>) nounwind readnone
 declare <16 x i8> @llvm.arm.neon.vbsl.v16i8(<16 x i8>, <16 x i8>, <16 x i8>) nounwind readnone
@@ -170,3 +198,5 @@ declare <4 x i16> @llvm.arm.neon.vbsl.v4i16(<4 x i16>, <4 x i16>, <4 x i16>) nou
 declare <8 x i8> @llvm.arm.neon.vbsl.v8i8(<8 x i8>, <8 x i8>, <8 x i8>) nounwind readnone
 declare <2 x float> @llvm.arm.neon.vbsl.v2f32(<2 x float>, <2 x float>, <2 x float>) nounwind readnone
 declare <4 x float> @llvm.arm.neon.vbsl.v4f32(<4 x float>, <4 x float>, <4 x float>) nounwind readnone
+declare <2 x i64> @llvm.arm.neon.vbsl.v2i64(<2 x i64>, <2 x i64>, <2 x i64>) nounwind readnone
+declare <1 x i64> @llvm.arm.neon.vbsl.v1i64(<1 x i64>, <1 x i64>, <1 x i64>) nounwind readnone
-- 
2.7.4