This change makes two big adjustments.
authorChris Lattner <sabre@nondot.org>
Sat, 13 Mar 2004 23:15:45 +0000 (23:15 +0000)
committerChris Lattner <sabre@nondot.org>
Sat, 13 Mar 2004 23:15:45 +0000 (23:15 +0000)
commit619d3544b19dfdc0006bb0f036253d76488fc212
tree828eaf8ec4737c97e623243293239bb7a6c3f625
parent786c5646e9cd15f18a6a7938673f74b02a05adb8
This change makes two big adjustments.
 * Be a lot more accurate about what the effects will be when inlining a call
   to a function when an argument is an alloca.
 * Dramatically reduce the penalty for inlining a call in a large function.
   This heuristic made it almost impossible to inline a function into a large
   function, no matter how small the callee is.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@12363 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Transforms/IPO/InlineSimple.cpp