jeltz/heptagon
1
0
Fork 0
Code Issues Pull requests Releases Activity
1006 commits 2 branches 0 tags 6.9 MiB
Commit graph

347 commits

Author SHA1 Message Date
Gwenal Delaval
2be4e287b9 Take into account the -nosink option in Sigali 
Corrected bug: do not suppress the last state
variable when -nosink is on.
 2013-01-28 15:54:27 +01:00
Gwenaël Delaval
5a9b27d6a9 Correct handling of local assume/enforce of contracts 2012-12-11 08:20:04 +01:00
Gwenal Delaval
ce2e2bdcd0 Added -nbvars option for performance evaluation 2012-11-26 10:05:20 +01:00
Gwenaël Delaval
ef536f412d Added option -nosink (CS optimisation) 
The -nosink option suppress the sink state of sigali equations.

This optimizes the controller synthesis, but work only
when the synthesis objective instantaneoulsy depends only
on the current state (and not on current inputs).
 2012-11-23 12:23:17 +01:00
Gwenaël Delaval
5bdd891105 Handling of controllables in tomato 
Added controllable variables as "inputs" in the
tomato pass.
 2012-11-22 18:52:21 +01:00
Gwenaël Delaval
e42ff23a4b Corrected abstraction for z3z target 2012-11-22 18:37:32 +01:00
Gwenal Delaval
74a760ee0a Handling of controllables in Normalize_mem 2012-11-17 23:01:49 +01:00
Léonard Gérard
b0e2e41299 Schedule: do only count optimized variables in costs. 2012-10-17 12:02:00 +02:00
Cédric Pasteur
36bfa81b17 Fix for memalloc 
Take the simplified versions of types (i.e. 
with constants instantiated) to check the 
equality of types
 2012-09-14 16:08:26 +02:00
Gwenal Delaval
9334b73ef1 Naive abstraction for sigali 2012-08-08 18:16:33 +02:00
Gwenal Delaval
41fccc66fb Bugs corrections 
- callgraph: add idents used for instantiated nodes
- cgen : added Idents.enter_node
- cmain : removed error when simulated node does not exist (existence
of simulated node was tested for every program, comprising loaded ones)
 2012-08-01 17:08:58 +02:00
Gwenaël Delaval
72b1bd8de3 Typo in error message 2012-07-26 01:33:32 +02:00
Gwenaël Delaval
90dda27a3a Bug correction in Schedule_interf 
Bug due to the fact that a variable can be "defined" and "read"
(in scheduling sense) by the same equation, without being a memory:
e.g., a clock defined as the result of a node application, together
with another result on this same clock.

Bug correction: basically removed the "assert false" on killed_vars,
decr_uses; do not count as "use" the self reads.
 2012-07-26 01:29:22 +02:00
Gwenaël Delaval
e103d60c26 Removed eprintf in schedule_interf 
eprintf -> Interference.print_debug
 2012-07-14 14:31:46 +02:00
Gwenaël Delaval
ef00823cf7 Added Marc as co-author 2012-06-29 01:43:15 +02:00
Gwenaël Delaval
c080ad6cf3 Controller call only when controllables 
Avoid built of dummy empty controllers
 2012-06-29 01:41:13 +02:00
Gwenaël Delaval
58086190eb Headers and license file for GPL 
Headers for every source file (excluding examples), mentioning
authors, copyright and license (GPL)

COPYING file with GPLv3 content.
 2012-06-27 18:14:29 +02:00
Cédric Pasteur
5e1dad630b Force fby to be scheduled at the end 2012-06-20 17:09:30 +02:00
Cédric Pasteur
7a10ba028a Fixed disjoint clock computation 2012-06-20 17:09:17 +02:00
Cédric Pasteur
ffeb81f529 Use idents instead of ivars in scheduling 2012-06-20 16:33:06 +02:00
Cédric Pasteur
8815a2cd03 Better handling of clocks in memalloc 
- We can do a better allocation if we take into
account 'when' in extvalues 
(test/good/memalloc_clocks.ept shows the
improvement)
- Fixed a bug with memalloc on records: if we 
translate:
o = { a with .f = u }
to
o = a; o.f = u
then we cannot share u and o.f
 2012-06-20 09:17:13 +02:00
Cédric Pasteur
ed2c08315b Should be affinity, not copy 2012-06-19 09:33:27 +02:00
Gwenal Delaval
cf22ba3989 Optional block in contracts 
Optional let...tel block in contracts
Sink state in sigali
 2012-06-07 17:48:31 +02:00
Gwenal Delaval
2bd31db883 Causality and scheduling with contracts 
Correction of the causality analysis and scheduling (with interference)
to take contracts into account.
 2012-06-07 15:27:07 +02:00
Gwenal Delaval
2956002f85 Correction and simplification of the sigali pass 
Added a "Contracts" pass, after inlining, taking care of the
contracts of the nodes called in the body of a node. This pass
"inlines" the code and assume/guarantee parts of these subcontracts.

The "Sigali" pass both generates the sigali ("z3z") code and add the call to
the controller (which is a node generated further by the sigali tool).
Therefore this pass has been included into the mls compiler, and removed
from the targets (a "z3z" dummy target has been kept for backward compatibility
reasons).
 2012-06-06 15:59:08 +02:00
Gwenal Delaval
bb0bc8bfe5 Added local assume/guarantee 
Added local assume/guarantee in contracts.

No syntax associated to these local asume/guarantee: internal use only.
 2012-05-29 14:14:46 +02:00
Adrien Guatto
8153bc4eb5 Fixed Tomato: did not reconstruct internal clocks of extvalues 2012-03-30 14:47:47 +02:00
Léonard Gérard
61e14546df Interf scheduling tries harder with arrays 2012-03-19 17:45:06 +01:00
Adrien Guatto
44d3a639e5 Internal error for unknown clocks. 2012-03-02 17:12:30 +01:00
Adrien Guatto
699b3c68e9 Tomato: bug fix, node inputs were not properly considered different. 2012-03-01 14:46:33 +01:00
Léonard Gérard
8a78bc7d7d Add [external] in the signatures. fix callgraph acordingly. 2012-02-21 16:07:29 +01:00
Adrien Guatto
5ac3a7f028 Redisable interference of scalars during scheduling 2012-02-10 11:33:09 +01:00
Cédric Pasteur
6e2e2a9f47 Fixed bug in interference computation 
To know the variables read by an equation, we should
only look at the clock of variables. Otherwise, 
there could be a problem for node calls that 
define new clocks.
 2012-02-09 11:48:36 +01:00
Cédric Pasteur
7866321089 Fixed bug in interference 2012-02-09 11:06:28 +01:00
Adrien Guatto
3aeb499cc2 Re-enable interference for enums 2012-02-08 11:13:02 +01:00
Cédric Pasteur
f66c9045df Special case for merge in the scheduling 
For merge equations, the generated code will be 
on a different rate than the activation clock of
the equation.
 2012-02-01 11:19:25 +01:00
Adrien Guatto
1cb4b1154b Memory allocation: do not share enums (and thus clocks). 2012-01-30 17:48:47 +01:00
Adrien Guatto
946a1f8228 Inline extvalues: do not inline array literals. 2012-01-30 16:37:42 +01:00
Adrien Guatto
84ca123361 Extvalue inlining: rogue debug message. 2012-01-26 13:23:01 +01:00
Adrien Guatto
964f6ca605 Extvalue inlining: fix point computation 2012-01-26 13:21:17 +01:00
Adrien Guatto
53de6cd915 Bug fix in extvalue inlining 2012-01-25 16:11:22 +01:00
Adrien Guatto
bca8664d2f Tomato: do not forget eq_base_ck when reconstructing 2012-01-25 13:19:09 +01:00
Cédric Pasteur
2f993a602c Fixed base clock in code generation 
Put the base clock inside the equation where it 
belongs.
 2012-01-25 09:34:58 +01:00
Cédric Pasteur
3f80f844c7 Schedule with clocks first 
First optimize clocks and then look at life ranges
 2012-01-24 15:33:54 +01:00
Cédric Pasteur
06e997e0c8 Fix for -mall 
Do not share memories and outputs if they are not
arrays, as it might make the register allocation
by the C compiler less efficient.
 2012-01-24 10:29:05 +01:00
Cédric Pasteur
b1bd6dbd57 Compare clock repr 
Not sure this is necessary but it doesn't hurt to 
check twice.
 2012-01-23 16:03:01 +01:00
Cédric Pasteur
2b59ec754a Fix for memories with no uses 
Add a fake use to make sure they interfere with
other memories and outputs
 2012-01-23 16:02:34 +01:00
Adrien Guatto
cad8a0149f Option to perform type inference on all types 2012-01-23 13:36:24 +01:00
Léonard Gérard
c75236b688 miscs 2011-12-12 12:06:46 +01:00
Léonard Gérard
da3147151d Better check signature error message 2011-12-12 11:30:18 +01:00