<html><head><meta http-equiv="Content-Type" content="text/html charset=windows-1252"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;"><font size="5">Hi Rafael</font><div><font size="5"><br></font></div><div><font size="5">it seems that one of your commits modifies stack alignment (and consequently instruction selection) of not just APCS and results in a performance regression on sphereflake. Could you please take a look? </font></div><div><font size="5"><br></font></div><div><font size="5">Thanks</font></div><div><font size="5">Gerolf</font></div><div><font size="5"><br></font></div><div><font size="5">— svn rev 197526 </font> </div><div><font size="5">Author: Rafael Espindola <<a href="mailto:rafael.espindola@gmail.com">rafael.espindola@gmail.com</a>></font></div><div><font size="5">Date: Tue Dec 17 21:36:54 2013 +0000</font></div><div><font size="5"><br></font></div><div><font size="5"> On APCS, only try to align aggregates to 32 bits instead of 64.</font></div><div><font size="5"> </font></div><div><font size="5"> This matches clang's behavior and since it is only a preference, it is not</font></div><div><font size="5"> an ABI issue.</font></div><div><font size="5">—</font></div><div><font size="5"><br></font></div><div><font size="5">Sphereflake sample snippet (main()) with 32b (after) and 64b (before your change) stack alignment:</font></div><div><br></div><div><img apple-inline="yes" id="418D29AB-89CA-412D-BF1B-647EDCBAC911" height="342" width="1392" apple-width="yes" apple-height="yes" src="cid:08CFA4A1-AD4B-48D8-8BBD-8991155F1A7D@apple.com"></div></body></html>