The ARM disassembler did not handle the alignment correctly for VLD*DUP* instructions

[oota-llvm.git] / test / CodeGen / ARM / vcvt.ll

diff --git a/test/CodeGen/ARM/vcvt.ll b/test/CodeGen/ARM/vcvt.ll
index 795caa418d5c29f692b76f08c5d44d30aa789c17..c078f493094bea3bbb08e804b0fd2f2b32f9932e 100644 (file)
--- a/test/CodeGen/ARM/vcvt.ll
+++ b/test/CodeGen/ARM/vcvt.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s
+; RUN: llc < %s -march=arm -mattr=+neon,+fp16 | FileCheck %s
 
 define <2 x i32> @vcvt_f32tos32(<2 x float>* %A) nounwind {
 ;CHECK: vcvt_f32tos32:
@@ -63,3 +63,96 @@ define <4 x float> @vcvtQ_u32tof32(<4 x i32>* %A) nounwind {
        %tmp2 = uitofp <4 x i32> %tmp1 to <4 x float>
        ret <4 x float> %tmp2
 }
+
+define <2 x i32> @vcvt_n_f32tos32(<2 x float>* %A) nounwind {
+;CHECK: vcvt_n_f32tos32:
+;CHECK: vcvt.s32.f32
+       %tmp1 = load <2 x float>* %A
+       %tmp2 = call <2 x i32> @llvm.arm.neon.vcvtfp2fxs.v2i32.v2f32(<2 x float> %tmp1, i32 1)
+       ret <2 x i32> %tmp2
+}
+
+define <2 x i32> @vcvt_n_f32tou32(<2 x float>* %A) nounwind {
+;CHECK: vcvt_n_f32tou32:
+;CHECK: vcvt.u32.f32
+       %tmp1 = load <2 x float>* %A
+       %tmp2 = call <2 x i32> @llvm.arm.neon.vcvtfp2fxu.v2i32.v2f32(<2 x float> %tmp1, i32 1)
+       ret <2 x i32> %tmp2
+}
+
+define <2 x float> @vcvt_n_s32tof32(<2 x i32>* %A) nounwind {
+;CHECK: vcvt_n_s32tof32:
+;CHECK: vcvt.f32.s32
+       %tmp1 = load <2 x i32>* %A
+       %tmp2 = call <2 x float> @llvm.arm.neon.vcvtfxs2fp.v2f32.v2i32(<2 x i32> %tmp1, i32 1)
+       ret <2 x float> %tmp2
+}
+
+define <2 x float> @vcvt_n_u32tof32(<2 x i32>* %A) nounwind {
+;CHECK: vcvt_n_u32tof32:
+;CHECK: vcvt.f32.u32
+       %tmp1 = load <2 x i32>* %A
+       %tmp2 = call <2 x float> @llvm.arm.neon.vcvtfxu2fp.v2f32.v2i32(<2 x i32> %tmp1, i32 1)
+       ret <2 x float> %tmp2
+}
+
+declare <2 x i32> @llvm.arm.neon.vcvtfp2fxs.v2i32.v2f32(<2 x float>, i32) nounwind readnone
+declare <2 x i32> @llvm.arm.neon.vcvtfp2fxu.v2i32.v2f32(<2 x float>, i32) nounwind readnone
+declare <2 x float> @llvm.arm.neon.vcvtfxs2fp.v2f32.v2i32(<2 x i32>, i32) nounwind readnone
+declare <2 x float> @llvm.arm.neon.vcvtfxu2fp.v2f32.v2i32(<2 x i32>, i32) nounwind readnone
+
+define <4 x i32> @vcvtQ_n_f32tos32(<4 x float>* %A) nounwind {
+;CHECK: vcvtQ_n_f32tos32:
+;CHECK: vcvt.s32.f32
+       %tmp1 = load <4 x float>* %A
+       %tmp2 = call <4 x i32> @llvm.arm.neon.vcvtfp2fxs.v4i32.v4f32(<4 x float> %tmp1, i32 1)
+       ret <4 x i32> %tmp2
+}
+
+define <4 x i32> @vcvtQ_n_f32tou32(<4 x float>* %A) nounwind {
+;CHECK: vcvtQ_n_f32tou32:
+;CHECK: vcvt.u32.f32
+       %tmp1 = load <4 x float>* %A
+       %tmp2 = call <4 x i32> @llvm.arm.neon.vcvtfp2fxu.v4i32.v4f32(<4 x float> %tmp1, i32 1)
+       ret <4 x i32> %tmp2
+}
+
+define <4 x float> @vcvtQ_n_s32tof32(<4 x i32>* %A) nounwind {
+;CHECK: vcvtQ_n_s32tof32:
+;CHECK: vcvt.f32.s32
+       %tmp1 = load <4 x i32>* %A
+       %tmp2 = call <4 x float> @llvm.arm.neon.vcvtfxs2fp.v4f32.v4i32(<4 x i32> %tmp1, i32 1)
+       ret <4 x float> %tmp2
+}
+
+define <4 x float> @vcvtQ_n_u32tof32(<4 x i32>* %A) nounwind {
+;CHECK: vcvtQ_n_u32tof32:
+;CHECK: vcvt.f32.u32
+       %tmp1 = load <4 x i32>* %A
+       %tmp2 = call <4 x float> @llvm.arm.neon.vcvtfxu2fp.v4f32.v4i32(<4 x i32> %tmp1, i32 1)
+       ret <4 x float> %tmp2
+}
+
+declare <4 x i32> @llvm.arm.neon.vcvtfp2fxs.v4i32.v4f32(<4 x float>, i32) nounwind readnone
+declare <4 x i32> @llvm.arm.neon.vcvtfp2fxu.v4i32.v4f32(<4 x float>, i32) nounwind readnone
+declare <4 x float> @llvm.arm.neon.vcvtfxs2fp.v4f32.v4i32(<4 x i32>, i32) nounwind readnone
+declare <4 x float> @llvm.arm.neon.vcvtfxu2fp.v4f32.v4i32(<4 x i32>, i32) nounwind readnone
+
+define <4 x float> @vcvt_f16tof32(<4 x i16>* %A) nounwind {
+;CHECK: vcvt_f16tof32:
+;CHECK: vcvt.f32.f16
+       %tmp1 = load <4 x i16>* %A
+       %tmp2 = call <4 x float> @llvm.arm.neon.vcvthf2fp(<4 x i16> %tmp1)
+       ret <4 x float> %tmp2
+}
+
+define <4 x i16> @vcvt_f32tof16(<4 x float>* %A) nounwind {
+;CHECK: vcvt_f32tof16:
+;CHECK: vcvt.f16.f32
+       %tmp1 = load <4 x float>* %A
+       %tmp2 = call <4 x i16> @llvm.arm.neon.vcvtfp2hf(<4 x float> %tmp1)
+       ret <4 x i16> %tmp2
+}
+
+declare <4 x float> @llvm.arm.neon.vcvthf2fp(<4 x i16>) nounwind readnone
+declare <4 x i16> @llvm.arm.neon.vcvtfp2hf(<4 x float>) nounwind readnone