These IRgen improvements have been done.

Technically we could still do a bit more to avoid deferred generation of statics
which we know are used, but I seriously doubt this is important.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@76844 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/README.txt b/lib/CodeGen/README.txt
index f60cd03ad6490fe476dbe45bcdcef1ba10498daf..e6d61095bf23434e351470ed6b2028d2e551fffa 100644 (file)
--- a/lib/CodeGen/README.txt
+++ b/lib/CodeGen/README.txt
@@ -45,21 +45,3 @@ On 176.gcc:expr.ll, it looks like over 12% of basic blocks are just
 direct branches!
 
 //===---------------------------------------------------------------------===//
-
-There are some more places where we could avoid generating unreachable code. For
-example:
-  void f0(int a) { abort(); if (a) printf("hi"); }
-still generates a call to printf. This doesn't occur much in real
-code, but would still be nice to clean up.
-
-//===---------------------------------------------------------------------===//
-
-Deferred generation of statics incurs some additional
-overhead. Currently it is even possible to construct test cases with
-O(N^2) behavior! For at least simple cases where we can tell a global
-is used, it is probably not worth deferring it. This doesn't solve the
-O(N^2) cases, ,though...
-
-PR3810
-
-//===---------------------------------------------------------------------===//