I have mixed feelings - it's great to see they're trying to improve the performance, but sad to see the attempts appear to bring down performance in other areas.
Does anyone know what XE6's backend currently is? I am still hoping for them to bring the LLVM backend (and thus the LLVM optimiser) to Windows.
I have mixed feelings - it's great to see they're trying to improve the performance, but sad to see the attempts appear to bring down performance in other areas.
ReplyDeleteDoes anyone know what XE6's backend currently is? I am still hoping for them to bring the LLVM backend (and thus the LLVM optimiser) to Windows.
Delphi compiler does a lot of stack juggling as is, doesn't need to do more...
ReplyDeletedang that comparison of FPC to XE:
ReplyDeletehttp://goo.gl/4shkCG
Reported as QC124652 "x32 compiler regression for floating point expressions" http://qc.embarcadero.com/wc/qcmain.aspx?d=124652
ReplyDeleteSo this is supposedly fixed now in XE6 update 1. Has anyone tested and checked how/if the code generation improved?
ReplyDeleteVille Krumlinde Yes tested and fixed. Performance on pair with and slightly better than XE5.
ReplyDelete