questions about it. I'll send them in a new message.
–
Cheers, Roman Gareev
tically all the code of a CLAST generation
through GLooG and started consideration of translate_clast. However, I
don't understand the idea behind if_region creation. Do you know
something about it?
--
Cheers, Roman Gareev
owing link http://romangareev.blogspot.com
--
Cheers, Roman Gareev
then bound_two? (We also consider two cases
in “type_for_interval”)
10. Why do we assume that new bounds are min(bound_one, bound_two) and
min(b1, b2) in case of clast_red_min?
--
Cheers, Roman Gareev
7;ve registered as
RomanGareev)
--
Cheers, Roman Gareev
Thank you for your answers! They saved me the time of code consideration.
--
Cheers, Roman Gareev
e. I've asked about this issue.
--
Cheers, Roman Gareev
Thank you!
--
Cheers, Roman Gareev
ion in GCC.
I've set up a second code generation in parallel that generates ISL
AST and can be enabled by a command line flag. Could you please advise
me how to verify the results of this generation?
Below is the code of this generation.
--
Cheers, Roman Gareev
code
Description: Binary data
Hi Tobias,
I tried to incorporate all your comments in the following patch. It
also contains traversing of ISL AST and its dump to a file. You can
find out more about this at the following link
http://romangareev.blogspot.ru/2014/05/gsoc-report-i.html
--
Cheers, Roman Gareev
patch
Description
ink http://romangareev.blogspot.com/2014/05/gsoc-report-ii.html
--
Cheers, Roman Gareev
patch
Description: Binary data
change log to this message.
--
Cheers, Roman Gareev
patch
Description: Binary data
ChangeLog_entry
Description: Binary data
to the community.
--
Cheers, Roman Gareev
ChangeLog_entry1
Description: Binary data
ChangeLog_entry2
Description: Binary data
patch1
Description: Binary data
patch2
Description: Binary data
n
? for \\(int c3 = 0; c3 < n; c3 \\+= 1\\)\n
?S_5\\(c1, c3\\);" 1 "graphite"}
? } */
Is it possible? Where can I find the description of regular
expressions used in scan-tree-dump-times? I would be very grateful for
your comments.
--
Cheers, Roman Gareev
I used trunk and compiled these patches only with isl 0.12 and ClooG
0.18.1. Which versions of these libraries are need to be checked for
compatibility?
--
Cheers, Roman Gareev
ast_op_or_else, isl_ast_op_call,
isl_ast_op_member, isl_ast_op_access, isl_ast_op_pdiv_q,
isl_ast_op_pdiv_r, isl_ast_op_div, isl_ast_op_fdiv_q.
The first draft of generation of GCC expression trees from isl ast
expressions can be found below:
--
Cheers, Roman Gareev
pat
here exists a unit-test infrastructure in
> gcc.
Yes, I've started to working on simple DejaGnu test cases for these
expressions, which will possibly use the previous ones.
--
Cheers, Roman Gareev
Dear gcc contributors,
could you please answer a few questions about unit tests? Is it
possible to use them in gcc? Or maybe there is some analogue? I would
be very grateful for your comments.
--
Cheers, Roman Gareev
the semantics of the GCC and isl expressions.
--
Cheers, Roman Gareev
his generation (It still uses isl_int for
generation of isl_expr_int, because the error related to isl/val_gmp.h
still arises. I've tried to use isl 0.12.2 and 0.13, but gotten the
same error).
--
Cheers, Roman Gareev
Index: gcc/graphite-isl-ast-to-gi
Thank you for the answer!
--
Cheers, Roman Gareev
.
--
Cheers, Roman Gareev
or this. As was mentioned above, I
haven't found another way for generation of loops.
> Why are the previous two functions necessary?
Yes, they are unimportant. I've replaced them with
add_parameters_to_ivs_params, which add tree representations and name
ssert (isl_ast_expr_get_type (name_expr) == isl_ast_expr_id);
isl_id *name_id = isl_ast_expr_get_id (name_expr);
poly_bb_p pbb = (poly_bb_p) isl_id_get_user (name_id);
(I'm getting this information, while we're handling isl_ast_node_user)
--
Cheers, Roman Gareev
No, I don't (at least for now).
--
Cheers, Roman Gareev
th vec. Is there a way to use it with
vec?
I would be very grateful for your comments.
--
Cheers, Roman Gareev.
}
Do you know anything about this issue? Should we use some options to
avoid illegal transformations (I haven't fount them)? Is it possible
to manually update SSA form to use the last versions of variables (I
haven't found such possibilities)?
P.S. There are other examples of wrong generati
0] -> [0, i0, 0] : exists (e0 = [(-1 + n.0)/4294967296]: i0 >= 0
and 4294967296e0 <= -1 + n.0 and 4294967296e0 >= -4294967296 + n.0 and
4294967296e0 <= -1 + n.0 - i0 and i0 <= 2147483646) }
--
Cheers, Roman Gareev.
It seems that the problem is solved now. Thank you! I've sent
corresponding patches to gcc-patches.
--
Cheers, Roman Gareev.
> Yes, you need to use .create() to initialize vec instances.
Thank you for the answer!
--
Cheers, Roman Gareev.
unity. If I'm not mistaken “2 * (
level + 1) “ is also not suitable for all cases.
--
Cheers, Roman Gareev.
"
--
Cheers, Roman Gareev.
r comments.
--
Cheers, Roman Gareev.
utation of types (in case of
completion of its integration into ISL). 2) Profile the generator. 3)
Make execution-time and compile-time performance comparisons between
CLooG and ISL code generation in Graphite. 4) Use the full/partial
tile separation for the tiles generated by the isl scheduler.
--
Cheers, Roman Gareev.
ce, but can not do
> the full move myself)
I could find time for this. What do you mean by ‘switch’? If I’m not
mistaken, ISL generator is already used by default. Should we remove
support of CLooG generator and all files related to it?
--
Cheers, Roman Gareev.
es
related to CLooG code generator, the second one removes support of
CLooG library from configure and make files. What do you think about
this?
--
Cheers, Roman Gareev.
me how to fix this? If I’m not mistaken,
r216735 is the only commit related to graphite-optimize-isl.c which
has been made since my latest patch.
--
Cheers, Roman Gareev.
I would be very grateful for your comments, feedback and ideas about
its improvement.
-
Roman Gareev
garding the implementation I have a question - why a new AST-like
> representation is needed ?
> It is not possible to generate the code directly from the ISL AST (with
> possible addition of
> new attributes/transformations) ?
>
> Regards, Mircea
>
> - Original
presentation in the improved
version of the proposal.
--
Cheers, Roman Gareev
40 matches
Mail list logo
