Re: [GRAPHITE, PATCH] Loop unroll and jam optimization

2014-11-13 Thread Mircea Namolaru
...@kotnet.org To: Mircea Namolaru mircea.namol...@inria.fr Cc: gcc-patches@gcc.gnu.org, Tobias Grosser tob...@grosser.es, Richard Biener richard.guent...@gmail.com, Albert Cohen albert.co...@inria.fr Sent: Wednesday, November 12, 2014 12:24:40 AM Subject: Re: [GRAPHITE, PATCH] Loop unroll

Re: [GRAPHITE, PATCH] Loop unroll and jam optimization

2014-11-12 Thread Mircea Namolaru
At this point pbb-transformed has two basic maps, one is the mapping for unroll and jam, and one for the full tile for the striped dimension. Introduce a check that differentiate between them as the image of one maps should be included in the other. I didn't do a full review (and I

Re: [GRAPHITE, PATCH] Loop unroll and jam optimization

2014-11-11 Thread Mircea Namolaru
Many thanks. Here is the new patch that fixes the main problem of the previous one (i.e separation of the loop after unroll and jam) as well as the problems raised by you (see comments below). Now the code with the separation class option looks: ISL AST generated by ISL: { for (int c0 = 0;

Re: [GRAPHITE, PATCH] Loop unroll and jam optimization

2014-11-11 Thread Mircea Namolaru
Changed the option to -floop-unroll-and jam as you suggested. The patch takes advantage of the new isl based code generator introduced recently in GCC (in fact of the possible options for building the AST). The code generated for this optimization in the case of non-constant loop

Re: [GRAPHITE, PATCH] Loop unroll and jam optimization

2014-11-11 Thread Sven Verdoolaege
On Tue, Nov 11, 2014 at 04:05:57PM +0100, Mircea Namolaru wrote: I'm not sure if Tobi or Albert have told you, but the separation_class option is going to be phased out since its design is fundamentally flawed. If you can't wait until isl-0.15, then I guess you have no choice but to use

Re: [GRAPHITE, PATCH] Loop unroll and jam optimization

2014-11-08 Thread Sven Verdoolaege
On Sat, Nov 08, 2014 at 12:32:05AM +0100, Mircea Namolaru wrote: Hello, This is the patch for unroll and jam optimizations. It is based on the code written by Tobias from graphite-optimize-isl.c (the code was unreachable till now) extended and enabled it via a new option

Re: [GRAPHITE, PATCH] Loop unroll and jam optimization

2014-11-08 Thread Richard Biener
On Sat, Nov 8, 2014 at 12:32 AM, Mircea Namolaru mircea.namol...@inria.fr wrote: Hello, This is the patch for unroll and jam optimizations. It is based on the code written by Tobias from graphite-optimize-isl.c (the code was unreachable till now) extended and enabled it via a new option

[GRAPHITE, PATCH] Loop unroll and jam optimization

2014-11-07 Thread Mircea Namolaru
Hello, This is the patch for unroll and jam optimizations. It is based on the code written by Tobias from graphite-optimize-isl.c (the code was unreachable till now) extended and enabled it via a new option -floop-unroll-jam. The patch takes advantage of the new isl based code generator