This is the mail archive of the
gcc-help@gcc.gnu.org
mailing list for the GCC project.
Interprocedural optimizations
- From: charfi asma <charfiasma at yahoo dot fr>
- To: gcc-help at gcc dot gnu dot org
- Date: Tue, 13 Apr 2010 09:11:45 +0000 (GMT)
- Subject: Interprocedural optimizations
Hello,
I have red 4 articles from GCC summit about Interprocedural optimizations, and I do not understand some details:
1. Are IPA performed in High gimple (before SSA convertion) or In SSA form just like other intraprocedural optimization?
If they are performed before SSA did this implementation has an impact on the memory consumption, since SSA is less compact than High Gimple.
2. In -function-at-a-time, each function is compiled immediately after parsing its body. however, in unit-at-a-time, all functions are parsed first then analyzed and compiled.
unit-at-time is a new compilation scheme that was added to perfrom IPA, cause it is impossible to implement IPA in function-at-time. why we can not do it ?
What does parsing refer to ? if parsing means converting from source code to AST (Generic) and compiling means passing the function to the backend (RTL), we can then wait until all functions are parsed and then we perfrom IPA after the parsing. did this sentence: "each function is compiled immediately after parsing its body" means that each function is passed to backend just after parsing and we will not save its GENERIC and Gimple presentation. that's why we can not perfrom IPA that needs all function presentation in GIMPLE??
I seems to be very confused :(
thank you very much for your help
Asma