The generated code for compressed oops shows the klass field being fetched and decompressed a lot, but apparently it's doing that to compare against prior inlining decisions. If it compared narrow oops rather than decompressing them, that would generate better code. Tom can fill in the details. thanks!
- duplicates
-
JDK-6703890 Compressed Oops: add LoadNKlass node to generate narrow oops (32-bits) compare instructions
-
- Resolved
-
- relates to
-
JDK-6459804 Want client (c1) compiler for x86_64 (amd64) for faster start-up
-
- Closed
-