Add move vector to the integrated ARM assembler.

tests_lit/assembler/arm32/vec-move.ll

Index: tests_lit/assembler/arm32/vec-move.ll
diff --git a/tests_lit/assembler/arm32/vec-move.ll b/tests_lit/assembler/arm32/vec-move.ll
new file mode 100644
index 0000000000000000000000000000000000000000..ca2ebe30203df1dbfbb6c934aa43fdec9f58b4c1
--- /dev/null
+++ b/tests_lit/assembler/arm32/vec-move.ll
@@ -0,0 +1,39 @@
+; Test that we handle a vector move.
+
+; NOTE: We use -O2 to force a vector move for the return value.
+
+; REQUIRES: allow_dump
+
+; Compile using standalone assembler.
+; RUN: %p2i --filetype=asm -i %s --target=arm32 --args -O2 \
+; RUN:   | FileCheck %s --check-prefix=ASM
+
+; Show bytes in assembled standalone code.
+; RUN: %p2i --filetype=asm -i %s --target=arm32 --assemble --disassemble \
+; RUN:   --args -O2 \
+; RUN:   | FileCheck %s --check-prefix=DIS
+
+; Compile using integrated assembler.
+; RUN: %p2i --filetype=iasm -i %s --target=arm32 --args -O2 \
+; RUN:   | FileCheck %s --check-prefix=IASM
+
+; Show bytes in assembled integrated code.
+; RUN: %p2i --filetype=iasm -i %s --target=arm32 --assemble --disassemble \
+; RUN:   --args -O2 \
+; RUN:   | FileCheck %s --check-prefix=DIS
+
+
+define internal <4 x float> @testMoveVector(<4 x i32> %a) {
+; ASM-LABEL: testMoveVector:
+; DIS-LABEL:{{.+}} <testMoveVector>:
+; IASM-LABEL: testMoveVector:
+
+entry:
+  %0 = sitofp <4 x i32> %a to <4 x float>

[Eric Holk, 2016/02/11 19:24:22]

This may not always create a move. Right now this cast works by scalarization
which leads to bad code. We'll hopefully do better in the future, although if we
do I suppose we can fix the test then.

[Karl, 2016/02/11 22:10:28]

Unless you can come up with a better example, I suggest we use this for now. If
some change causes the instruction to disappear, this test will fail, and it can
be fixed then.

[Eric Holk, 2016/02/11 22:51:39]

I agree. This seems like the best course of action.

+  ret <4 x float> %0
+
+; ASM:  vmov.f32        q0, q1
+; DIS:  3c:     eef03a40
+; IASM-NOT: vmov.f32    q0, q1
+
+}