[SystemZ] Add CodeGen support for v4f32

The architecture doesn't really have any native v4f32 operations except
v4f32->v2f64 and v2f64->v4f32 conversions, with only half of the v4f32
elements being used.  Even so, using vector registers for <4 x float>
and scalarising individual operations is much better than generating
completely scalar code, since there's much less register pressure.
It's also more efficient to do v4f32 comparisons by extending to 2
v2f64s, comparing those, then packing the result.

This particularly helps with llvmpipe.

Based on a patch by Richard Sandiford.

llvm-svn: 236523
diff --git a/llvm/test/CodeGen/SystemZ/vec-perm-08.ll b/llvm/test/CodeGen/SystemZ/vec-perm-08.ll
index b5220ab..a18ca7b 100644
--- a/llvm/test/CodeGen/SystemZ/vec-perm-08.ll
+++ b/llvm/test/CodeGen/SystemZ/vec-perm-08.ll
@@ -129,6 +129,26 @@
   ret <2 x i64> %ret
 }
 
+; Test a high1/low2 permute for v4f32.
+define <4 x float> @f12(<4 x float> %val1, <4 x float> %val2) {
+; CHECK-LABEL: f12:
+; CHECK: vpdi %v24, %v24, %v26, 1
+; CHECK: br %r14
+  %ret = shufflevector <4 x float> %val1, <4 x float> %val2,
+                       <4 x i32> <i32 0, i32 1, i32 6, i32 7>
+  ret <4 x float> %ret
+}
+
+; Test a low2/high1 permute for v4f32.
+define <4 x float> @f13(<4 x float> %val1, <4 x float> %val2) {
+; CHECK-LABEL: f13:
+; CHECK: vpdi %v24, %v26, %v24, 4
+; CHECK: br %r14
+  %ret = shufflevector <4 x float> %val1, <4 x float> %val2,
+                       <4 x i32> <i32 6, i32 7, i32 0, i32 1>
+  ret <4 x float> %ret
+}
+
 ; Test a high1/low2 permute for v2f64.
 define <2 x double> @f14(<2 x double> %val1, <2 x double> %val2) {
 ; CHECK-LABEL: f14: