add a note. Our 64-bit shifts are ~30% slower than gcc's

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@30457 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/README.txt b/lib/Target/X86/README.txt
index 62c8ab1dcc0a65743f6e243e05bb676177794765..dd02948f8701c0c66bd04a3836d4ba73d7ce51f9 100644 (file)
--- a/lib/Target/X86/README.txt
+++ b/lib/Target/X86/README.txt
@@ -59,7 +59,8 @@ One better solution for 1LL << x is:
 
 But that requires good 8-bit subreg support.
 
-
+64-bit shifts (in general) expand to really bad code.  Instead of using
+cmovs, we should expand to a conditional branch like GCC produces.
 
 //===---------------------------------------------------------------------===//