[PATCH] [SelectionDAG] Unary vector constant folding integer legality fixes

[Ahmed Bougacha]

Is there a way to further reduce the testcase, or even rewrite it from scratch?  I shiver whenever I see autogen ;)

The change LGTM, but you might want to wait for other reviews.


REPOSITORY
  rL LLVM

================
Comment at: test/CodeGen/X86/fold-vector-bv-crash.ll:1-2
@@ +1,3 @@
+; RUN: llc < %s -mtriple=i686-unknown -mcpu=corei7
+; RUN: llc < %s -mtriple=x86_64-unknown -mcpu=corei7
+
----------------
cpu -> attr?

================
Comment at: test/CodeGen/X86/fold-vector-trunc-sitofp.ll:1-2
@@ +1,3 @@
+; RUN: llc < %s -mtriple=i686-unknown -mattr=+avx | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx | FileCheck %s
+
----------------
Isn't SSE2 enough?

http://reviews.llvm.org/D9282

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/