[DAGCombiner] insert_vector_elt: Avoid building a vector twice.

This patch prevents the following combine when the input vector is used more
than once.
insert_vector_elt (build_vector elt0, ..., eltN), NewEltIdx, idx
=>
build_vector elt0, ..., NewEltIdx, ..., eltN 

The reasons are:
- Building a vector may be expensive, so try to reuse the existing part of a
  vector instead of creating a new one (think big vectors).
- elt0 to eltN now have two users instead of one. This may prevent some other
  optimizations.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187396 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 6 insertions, 6 deletions

diff --git a/test/CodeGen/X86/vshift-3.ll b/test/CodeGen/X86/vshift-3.ll
index 9b484a71d1..0bdb32fcb8 100644
--- a/test/CodeGen/X86/vshift-3.ll
+++ b/test/CodeGen/X86/vshift-3.ll
@@ -55,12 +55,12 @@ entry:
 ; CHECK: psraw
   %0 = insertelement <8 x i16> undef, i16 %amt, i32 0
   %1 = insertelement <8 x i16> %0, i16 %amt, i32 1
-  %2 = insertelement <8 x i16> %0, i16 %amt, i32 2
-  %3 = insertelement <8 x i16> %0, i16 %amt, i32 3
-  %4 = insertelement <8 x i16> %0, i16 %amt, i32 4
-  %5 = insertelement <8 x i16> %0, i16 %amt, i32 5
-  %6 = insertelement <8 x i16> %0, i16 %amt, i32 6
-  %7 = insertelement <8 x i16> %0, i16 %amt, i32 7
+  %2 = insertelement <8 x i16> %1, i16 %amt, i32 2
+  %3 = insertelement <8 x i16> %2, i16 %amt, i32 3
+  %4 = insertelement <8 x i16> %3, i16 %amt, i32 4
+  %5 = insertelement <8 x i16> %4, i16 %amt, i32 5
+  %6 = insertelement <8 x i16> %5, i16 %amt, i32 6
+  %7 = insertelement <8 x i16> %6, i16 %amt, i32 7
   %ashr = ashr <8 x i16> %val, %7
   store <8 x i16> %ashr, <8 x i16>* %dst
   ret void