guidance for GSoC 2016 under GCC
Hello GCC developers, I would like to work on one of the following idea in GSoC 2016 for GCC. Function Reordering (Improvement) with LTO Inter-procedural value range propagation pass Implement tree level section anchors to improve code generation at ARM/PPC. I have done some reading for first and second topic. I would like your guidance. For first topic I have read Martin's master thesis and as far as I understand currently he has implemented function reordering with PGO support but this project would be using LTO support. Am I thinking it right ? For second project I have read the IPCP.c file in gcc source code which implements Inter-procedural constant propagation with call graphs and jump functions. According to Chapter 11, page 664 of Optimizing Compilers for Modern Architectures: A Dependence-based Approach book range propagation pass can be designed by extending IPCP. Here extensions to IPCP would be deciding ranges of variable from for loops, const assignment or if/else statement and modifying jump functions so that ranges can be calculated base on operations. Also we may use data structure for range as used in tree-vrp.c of gcc. For third project I have not started studying about it. Please suggest some readings. Apart from this I have learned how to write simple passes and plugins for gcc and its related data structures ( learned from Diego Novillo's slide ). I have also written some simple optimization passes with LLVM libs. Please provide more information or experimental patches to study. Sincerely, Vivek Pandya P.S : Actually I tried to contact Mr. Jan Hubicka as mentioned on idea page but it seems that he is not reachable on his mail address j...@suse.cz that is why I have mail to gcc dev list.
Re: ipa vrp implementation in gcc
> On Mon, Jan 11, 2016 at 4:07 PM, Richard Biener > wrote: >> >> On Mon, Jan 11, 2016 at 1:38 AM, Kugan >> wrote: >> > Hi All, >> > >> > I am looking at implementing a ipa vrp pass. Jan Hubicka also talks >> > about this in 2013 GNU Cauldron as one of the optimization he would like >> > to see in gcc. So my question is, is any one implementing it. If not we >> > would like to do that. >> > > > Hello I am Vivek Pandya, I am actually working on a GSoC 2016 proposal for > his work and it is very similar to extending ipa-cp pass. I am also in touch > with Jan Hubicka. These comments will certainly help me but if is urgent for > any one you can begin work on this. Jan has shown interest to mentor me for > this project but any help from community is always appreciated. > >> >> > I also looked at the ipa-cp implementation to see how this can be done. >> > Going by this, one of the way to implement this is (skipping all the >> > details): >> > >> > - Have an early tree-vrp so that we can have value ranges for parameters >> > at call sites. > > Actually a tree-vrp pass already exists. But as Jan has suggested me that > ipa-vrp implementation should not be too much costly. So I am also thinking > to include this work in my proposal and also using the analysis to improve > LTO heuristics as the project duration will be around 2.5 months. >> >> >> I'd rather use the IPA analysis phase for this and use a VRP algorithm >> that doesn't require ASSERT_EXPR insertion. >> >> > - Create jump functions that captures the value ranges of call sites >> > propagate the value ranges. In 2013 talk, Jan Hubicka talks about >> > >> > - Modifying ipa-prop.[h|c] to handles this but wouldn't it be easier to >> > have its own and much more simpler implementation ? >> >> No idea. >> >> > - Once we have the value ranges for parameter/return values, we could >> > rely on tree-vrp to use this and do the optimizations >> >> Yep. IPA transform phase should annotate parameter default defs with >> computed ranges. >> >> > Does this make any sense? Any thoughts/suggestions to work on this is >> > highly appreciated. >> >> IPA alignment propagation should already be somewhat similar as in doing >> an intersection step during propagation. >> >> Richard. >> >> > Thanks, >> > Kugan > > Your comments certainly helps me to develop my proposal. Please let me know > any updated to avoid the confusion and duplication of work. > Sincerely, > Vivek
Source Code for Profile Guided Code Positioning
Hello GCC Developers, Are 'Profile Guided Code Positioning' algorithms mentioned in http://dl.acm.org/citation.cfm?id=93550 this paper ( Pettis and Hanse ) implemented in gcc ? If yes kindly help me with code file location in gcc source tree. Sincerely, Vivek Pandya
Re: Source Code for Profile Guided Code Positioning
Thanks Yury for https://gcc.gnu.org/ml/gcc-patches/2011-09/msg01440.html this link. It implements procedure reordering as linker plugin. I have some questions : 1 ) Can you point me to some documentation for "how to write plugin for linkers " I am I have not seen doc for structs with 'ld_' prefix (i.e defined in plugin-api.h ) 2 ) There is one more algorithm for Basic Block ordering with execution frequency count in PH paper . Is there any implementation available for it ? Sincerely, Vivek
Re: ipa vrp implementation in gcc
Vivek Pandya On Mon, Jan 18, 2016 at 4:16 AM, Kugan wrote: > > > > Hello I am Vivek Pandya, I am actually working on a GSoC 2016 proposal > > for his work and it is very similar to extending ipa-cp pass. I am also > > in touch with Jan Hubicka. > > Hi Vivek, > > Glad to know that you are planning to work on this. Could you please put > you plan in an accessible place (or post it here) so that we know what > you plans are. That way we can work on what you are not working. And > also possible contribute to your plan in other ways (like testing and > reviewing). > Hello Kugan, Actually my work will include extending the ipa-cp pass to propagate range information and then integrating this information to improve LTO optimizations (at-least one). But as mentioned by Jan Hubicka the real problem is not to extend ipa-cp pass but tree-vrp it self a big task and scheduling it at early stage will cost a performance lose. So actually I was looking at some alternatives to Patterson's approach and particularly I found this non iterative method: https://www.cs.berkeley.edu/~daw/papers/range-tacas04.pdf which has already implemented in LLVM .http://homepages.dcc.ufmg.br/~fernando/publications/papers/SBLP2011_douglas.pdf So my plan for this is first implementing above mentioned approach till 23 May , 2016 (My college project ) and then use this local pass for Value range analysis and then in my GSoC 2016 project I will use this pass for ipa-vrp pass and improving other ipa optimizations to use this information. Though in particular I have yet not figured implementation details. Currently I am learning about gcc IRs. If you have any further idea ( specially about constraints based method ) please let me know and help building my implementation approach. Sincerely, Vivek > > Thanks, > Kugan
Re: ipa vrp implementation in gcc
Vivek Pandya On Mon, Jan 18, 2016 at 11:35 AM, vivek pandya wrote: > Vivek Pandya > > > On Mon, Jan 18, 2016 at 4:16 AM, Kugan > wrote: >> >> >> > Hello I am Vivek Pandya, I am actually working on a GSoC 2016 proposal >> > for his work and it is very similar to extending ipa-cp pass. I am also >> > in touch with Jan Hubicka. >> >> Hi Vivek, >> >> Glad to know that you are planning to work on this. Could you please put >> you plan in an accessible place (or post it here) so that we know what >> you plans are. That way we can work on what you are not working. And >> also possible contribute to your plan in other ways (like testing and >> reviewing). >> > Hello Kugan, > > Actually my work will include extending the ipa-cp pass to propagate > range information and then integrating this information to improve LTO > optimizations (at-least one). But as mentioned by Jan Hubicka the real > problem is not to extend ipa-cp pass but tree-vrp it self a big task > and scheduling it at early stage will cost a performance lose. > So actually I was looking at some alternatives to Patterson's approach > and particularly I found this non iterative method: > https://www.cs.berkeley.edu/~daw/papers/range-tacas04.pdf which has > already implemented in LLVM > .http://homepages.dcc.ufmg.br/~fernando/publications/papers/SBLP2011_douglas.pdf Also please some one suggest me wether this non iterative method will be good to have as alternative VRP or not ? i.e will it serve our purpose of light weight VRP to be used at earlier stage ?? > So my plan for this is first implementing above mentioned approach > till 23 May , 2016 (My college project ) and then use this local pass > for Value range analysis and then in my GSoC 2016 project I will use > this pass for ipa-vrp pass and improving other ipa optimizations to > use this information. > Though in particular I have yet not figured implementation details. > Currently I am learning about gcc IRs. > If you have any further idea ( specially about constraints based > method ) please let me know and help building my implementation > approach. > > Sincerely, > Vivek >> >> Thanks, >> Kugan
