For instance:
type metres = int
type metres = MyLib.longueur
type matrice = metres^10^100
Code generation in C (with typedef) included. The
code uses the aliases for traceability.
Same as fold but the iterated function takes an
integer as the last but one argument.
This is not something that is necessarily
important to have in trunk but the patch is so
small that it doesn't hurt.
- Most transformation passes expect that list of
equations are always nested inside a block.
This fixes a problem with a duplication of
equations in Automata_mapfold.
Use a block instead of variable list + equation list
for contract and node.
The new program transformations based on the
mapfold iterator are now enabled by default.
- Many changes to make Hept2mls, mls2obc, etc
compile with the api changes
- Added Callgraph_mapfold: starting from a main
program, generates the list of instances of each
node necessary and creates them.
- Mls2seq deals with giving to the code generators
the correct source (mls or obc, wit or without
static parameters)
It is now possible to use parametrized nodes that
are defined in other files. For that to work, the
first file has to be compiled to an object file:
heptc -c mylib.ept
which creates a mylib.epo file. Compiling the main
file will then generate all the instances of
parametrized nodes from the lib (only the called
nodes will be compiled, but all the nodes in the
main file are compiled).
- Static are used for consts in Heptagon and Minils.
For now, node static parameters remain int only
(ie they are type parameters). Do we need more ?
- Also updated Parsetree AST to the recent changes
in API
There are now 3 kinds of nodes: Eprim : primitive
nodes, Enode: nodes with memory, Efun: nodes
without memory.
Typing now sets correct eq_statefull and b_statefull
for equations and blocks. Resets are only added
when a node with memory is called.
The calling convention is the generated code needs
to be modified so that we can remove the context
for nodes without memory.
