Add support for ARM's Advanced SIMD (NEON) instruction set.

This is still a work in progress but most of the NEON instruction set
is supported.

llvm-svn: 73919

1 files changed, 141 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/ARM/vqsub.ll b/llvm/test/CodeGen/ARM/vqsub.ll
new file mode 100644
index 00000000000..07052f78dbe
--- /dev/null
+++ b/llvm/test/CodeGen/ARM/vqsub.ll
@@ -0,0 +1,141 @@
+; RUN: llvm-as < %s | llc -march=arm -mattr=+neon > %t
+; RUN: grep {vqsub\\.s8} %t | count 2
+; RUN: grep {vqsub\\.s16} %t | count 2
+; RUN: grep {vqsub\\.s32} %t | count 2
+; RUN: grep {vqsub\\.s64} %t | count 2
+; RUN: grep {vqsub\\.u8} %t | count 2
+; RUN: grep {vqsub\\.u16} %t | count 2
+; RUN: grep {vqsub\\.u32} %t | count 2
+; RUN: grep {vqsub\\.u64} %t | count 2
+
+define <8 x i8> @vqsubs8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
+	%tmp1 = load <8 x i8>* %A
+	%tmp2 = load <8 x i8>* %B
+	%tmp3 = call <8 x i8> @llvm.arm.neon.vqsubs.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
+	ret <8 x i8> %tmp3
+}
+
+define <4 x i16> @vqsubs16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
+	%tmp1 = load <4 x i16>* %A
+	%tmp2 = load <4 x i16>* %B
+	%tmp3 = call <4 x i16> @llvm.arm.neon.vqsubs.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
+	ret <4 x i16> %tmp3
+}
+
+define <2 x i32> @vqsubs32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
+	%tmp1 = load <2 x i32>* %A
+	%tmp2 = load <2 x i32>* %B
+	%tmp3 = call <2 x i32> @llvm.arm.neon.vqsubs.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
+	ret <2 x i32> %tmp3
+}
+
+define <1 x i64> @vqsubs64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
+	%tmp1 = load <1 x i64>* %A
+	%tmp2 = load <1 x i64>* %B
+	%tmp3 = call <1 x i64> @llvm.arm.neon.vqsubs.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2)
+	ret <1 x i64> %tmp3
+}
+
+define <8 x i8> @vqsubu8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
+	%tmp1 = load <8 x i8>* %A
+	%tmp2 = load <8 x i8>* %B
+	%tmp3 = call <8 x i8> @llvm.arm.neon.vqsubu.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
+	ret <8 x i8> %tmp3
+}
+
+define <4 x i16> @vqsubu16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
+	%tmp1 = load <4 x i16>* %A
+	%tmp2 = load <4 x i16>* %B
+	%tmp3 = call <4 x i16> @llvm.arm.neon.vqsubu.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
+	ret <4 x i16> %tmp3
+}
+
+define <2 x i32> @vqsubu32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
+	%tmp1 = load <2 x i32>* %A
+	%tmp2 = load <2 x i32>* %B
+	%tmp3 = call <2 x i32> @llvm.arm.neon.vqsubu.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
+	ret <2 x i32> %tmp3
+}
+
+define <1 x i64> @vqsubu64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
+	%tmp1 = load <1 x i64>* %A
+	%tmp2 = load <1 x i64>* %B
+	%tmp3 = call <1 x i64> @llvm.arm.neon.vqsubu.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2)
+	ret <1 x i64> %tmp3
+}
+
+define <16 x i8> @vqsubQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
+	%tmp1 = load <16 x i8>* %A
+	%tmp2 = load <16 x i8>* %B
+	%tmp3 = call <16 x i8> @llvm.arm.neon.vqsubs.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2)
+	ret <16 x i8> %tmp3
+}
+
+define <8 x i16> @vqsubQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
+	%tmp1 = load <8 x i16>* %A
+	%tmp2 = load <8 x i16>* %B
+	%tmp3 = call <8 x i16> @llvm.arm.neon.vqsubs.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2)
+	ret <8 x i16> %tmp3
+}
+
+define <4 x i32> @vqsubQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
+	%tmp1 = load <4 x i32>* %A
+	%tmp2 = load <4 x i32>* %B
+	%tmp3 = call <4 x i32> @llvm.arm.neon.vqsubs.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2)
+	ret <4 x i32> %tmp3
+}
+
+define <2 x i64> @vqsubQs64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
+	%tmp1 = load <2 x i64>* %A
+	%tmp2 = load <2 x i64>* %B
+	%tmp3 = call <2 x i64> @llvm.arm.neon.vqsubs.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2)
+	ret <2 x i64> %tmp3
+}
+
+define <16 x i8> @vqsubQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
+	%tmp1 = load <16 x i8>* %A
+	%tmp2 = load <16 x i8>* %B
+	%tmp3 = call <16 x i8> @llvm.arm.neon.vqsubu.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2)
+	ret <16 x i8> %tmp3
+}
+
+define <8 x i16> @vqsubQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
+	%tmp1 = load <8 x i16>* %A
+	%tmp2 = load <8 x i16>* %B
+	%tmp3 = call <8 x i16> @llvm.arm.neon.vqsubu.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2)
+	ret <8 x i16> %tmp3
+}
+
+define <4 x i32> @vqsubQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
+	%tmp1 = load <4 x i32>* %A
+	%tmp2 = load <4 x i32>* %B
+	%tmp3 = call <4 x i32> @llvm.arm.neon.vqsubu.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2)
+	ret <4 x i32> %tmp3
+}
+
+define <2 x i64> @vqsubQu64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
+	%tmp1 = load <2 x i64>* %A
+	%tmp2 = load <2 x i64>* %B
+	%tmp3 = call <2 x i64> @llvm.arm.neon.vqsubu.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2)
+	ret <2 x i64> %tmp3
+}
+
+declare <8 x i8>  @llvm.arm.neon.vqsubs.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
+declare <4 x i16> @llvm.arm.neon.vqsubs.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
+declare <2 x i32> @llvm.arm.neon.vqsubs.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
+declare <1 x i64> @llvm.arm.neon.vqsubs.v1i64(<1 x i64>, <1 x i64>) nounwind readnone
+
+declare <8 x i8>  @llvm.arm.neon.vqsubu.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
+declare <4 x i16> @llvm.arm.neon.vqsubu.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
+declare <2 x i32> @llvm.arm.neon.vqsubu.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
+declare <1 x i64> @llvm.arm.neon.vqsubu.v1i64(<1 x i64>, <1 x i64>) nounwind readnone
+
+declare <16 x i8> @llvm.arm.neon.vqsubs.v16i8(<16 x i8>, <16 x i8>) nounwind readnone
+declare <8 x i16> @llvm.arm.neon.vqsubs.v8i16(<8 x i16>, <8 x i16>) nounwind readnone
+declare <4 x i32> @llvm.arm.neon.vqsubs.v4i32(<4 x i32>, <4 x i32>) nounwind readnone
+declare <2 x i64> @llvm.arm.neon.vqsubs.v2i64(<2 x i64>, <2 x i64>) nounwind readnone
+
+declare <16 x i8> @llvm.arm.neon.vqsubu.v16i8(<16 x i8>, <16 x i8>) nounwind readnone
+declare <8 x i16> @llvm.arm.neon.vqsubu.v8i16(<8 x i16>, <8 x i16>) nounwind readnone
+declare <4 x i32> @llvm.arm.neon.vqsubu.v4i32(<4 x i32>, <4 x i32>) nounwind readnone
+declare <2 x i64> @llvm.arm.neon.vqsubu.v2i64(<2 x i64>, <2 x i64>) nounwind readnone