From 3b9f777f4769611043e5c48ea190711fc89cc59b Mon Sep 17 00:00:00 2001
From: Evan Cheng <evan.cheng@apple.com>
Date: Mon, 10 Sep 2007 22:11:18 +0000
Subject: Observation of rematerialization.

llvm-svn: 41809
---
 llvm/lib/CodeGen/README.txt | 5 +++++
 1 file changed, 5 insertions(+)

(limited to 'llvm/lib/CodeGen')

diff --git a/llvm/lib/CodeGen/README.txt b/llvm/lib/CodeGen/README.txt
index 4fff15c1e3f..87671913209 100644
--- a/llvm/lib/CodeGen/README.txt
+++ b/llvm/lib/CodeGen/README.txt
@@ -143,3 +143,8 @@ load [T + 7]
 ...
 load [T + 15]
 //===---------------------------------------------------------------------===//
+
+It's not always a good idea to choose rematerialization over spilling. If all
+the load / store instructions would be folded then spilling is cheaper because
+it won't require new live intervals / registers. See 2003-05-31-LongShifts for
+an example.
-- 
cgit v1.2.3