Add support for ARM's specialized vector-compare-against-zero instructions.

llvm-svn: 118453

4 files changed, 66 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/ARM/vceq.ll b/llvm/test/CodeGen/ARM/vceq.ll
index e4787518e73..051c349a06a 100644
--- a/llvm/test/CodeGen/ARM/vceq.ll
+++ b/llvm/test/CodeGen/ARM/vceq.ll
@@ -79,3 +79,14 @@ define <4 x i32> @vceqQf32(<4 x float>* %A, <4 x float>* %B) nounwind {
         %tmp4 = sext <4 x i1> %tmp3 to <4 x i32>
 	ret <4 x i32> %tmp4
 }
+
+define <8 x i8> @vceqi8Z(<8 x i8>* %A) nounwind {
+;CHECK: vceqi8Z:
+;CHECK-NOT: vmov
+;CHECK-NOT: vmvn
+;CHECK: vceq.i8
+	%tmp1 = load <8 x i8>* %A
+	%tmp3 = icmp eq <8 x i8> %tmp1, <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
+        %tmp4 = sext <8 x i1> %tmp3 to <8 x i8>
+	ret <8 x i8> %tmp4
+}
diff --git a/llvm/test/CodeGen/ARM/vcge.ll b/llvm/test/CodeGen/ARM/vcge.ll
index 2c161113c11..f190931f1bf 100644
--- a/llvm/test/CodeGen/ARM/vcge.ll
+++ b/llvm/test/CodeGen/ARM/vcge.ll
@@ -160,3 +160,25 @@ define <4 x i32> @vacgeQf32(<4 x float>* %A, <4 x float>* %B) nounwind {
 
 declare <2 x i32> @llvm.arm.neon.vacged(<2 x float>, <2 x float>) nounwind readnone
 declare <4 x i32> @llvm.arm.neon.vacgeq(<4 x float>, <4 x float>) nounwind readnone
+
+define <8 x i8> @vcgei8Z(<8 x i8>* %A) nounwind {
+;CHECK: vcgei8Z:
+;CHECK-NOT: vmov
+;CHECK-NOT: vmvn
+;CHECK: vcge.s8
+	%tmp1 = load <8 x i8>* %A
+	%tmp3 = icmp sge <8 x i8> %tmp1, <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
+        %tmp4 = sext <8 x i1> %tmp3 to <8 x i8>
+	ret <8 x i8> %tmp4
+}
+
+define <8 x i8> @vclei8Z(<8 x i8>* %A) nounwind {
+;CHECK: vclei8Z:
+;CHECK-NOT: vmov
+;CHECK-NOT: vmvn
+;CHECK: vcle.s8
+	%tmp1 = load <8 x i8>* %A
+	%tmp3 = icmp sle <8 x i8> %tmp1, <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
+        %tmp4 = sext <8 x i1> %tmp3 to <8 x i8>
+	ret <8 x i8> %tmp4
+}
diff --git a/llvm/test/CodeGen/ARM/vcgt.ll b/llvm/test/CodeGen/ARM/vcgt.ll
index e3318cd2069..7663da3c612 100644
--- a/llvm/test/CodeGen/ARM/vcgt.ll
+++ b/llvm/test/CodeGen/ARM/vcgt.ll
@@ -173,3 +173,25 @@ define <4 x i32> @vcgt_zext(<4 x float>* %A, <4 x float>* %B) nounwind {
 
 declare <2 x i32> @llvm.arm.neon.vacgtd(<2 x float>, <2 x float>) nounwind readnone
 declare <4 x i32> @llvm.arm.neon.vacgtq(<4 x float>, <4 x float>) nounwind readnone
+
+define <8 x i8> @vcgti8Z(<8 x i8>* %A) nounwind {
+;CHECK: vcgti8Z:
+;CHECK-NOT: vmov
+;CHECK-NOT: vmvn
+;CHECK: vcgt.s8
+	%tmp1 = load <8 x i8>* %A
+	%tmp3 = icmp sgt <8 x i8> %tmp1, <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
+        %tmp4 = sext <8 x i1> %tmp3 to <8 x i8>
+	ret <8 x i8> %tmp4
+}
+
+define <8 x i8> @vclti8Z(<8 x i8>* %A) nounwind {
+;CHECK: vclti8Z:
+;CHECK-NOT: vmov
+;CHECK-NOT: vmvn
+;CHECK: vclt.s8
+	%tmp1 = load <8 x i8>* %A
+	%tmp3 = icmp slt <8 x i8> %tmp1, <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
+        %tmp4 = sext <8 x i1> %tmp3 to <8 x i8>
+	ret <8 x i8> %tmp4
+}
diff --git a/llvm/test/MC/ARM/neon-cmp-encoding.s b/llvm/test/MC/ARM/neon-cmp-encoding.s
index 27ebc8a9363..da59ee7209c 100644
--- a/llvm/test/MC/ARM/neon-cmp-encoding.s
+++ b/llvm/test/MC/ARM/neon-cmp-encoding.s
@@ -102,3 +102,14 @@
 	vtst.16	q8, q8, q9
 @ CHECK: vtst.32	q8, q8, q9              @ encoding: [0xf2,0x08,0x60,0xf2]
 	vtst.32	q8, q8, q9
+
+@ CHECK: vceq.i8	d16, d16, #0            @ encoding: [0x20,0x01,0xf1,0xf3]
+  vceq.i8	d16, d16, #0
+@ CHECK: vcge.s8	d16, d16, #0            @ encoding: [0xa0,0x00,0xf1,0xf3]
+  vcge.s8	d16, d16, #0
+@ CHECK: vcle.s8	d16, d16, #0            @ encoding: [0xa0,0x01,0xf1,0xf3]
+  vcle.s8	d16, d16, #0
+@ CHECK: vcgt.s8	d16, d16, #0            @ encoding: [0x20,0x00,0xf1,0xf3]
+  vcgt.s8	d16, d16, #0
+@ CHECK: vclt.s8	d16, d16, #0            @ encoding: [0x20,0x02,0xf1,0xf3]
+  vclt.s8	d16, d16, #0