consider having a main node like this one:
node main<<i : int; b : bool>> () returns...
Then if to call the generated program, you'll need ot give an int and a bool.
Furthermore, if you give an additional arg, it'll be used as the wanted
iteration number.
Conflicts:
compiler/obc/java/java_main.ml
( = ) in pervasives is a stub, it will be typed in a polymorphic way.
This is necessary to have a simple way to transform exp into a static_exp
even when there is the = operator.
I introduced a notion of extended values in Obc expressions,
replacing the Epattern constructor. Patterns may now only
occur at their rightful place, on the left of an assignment.
This change allows to index global constant arrays.
Ewhen is now the only case of possible recursion for minils exps.
This add was motivated by equations like :
(y,z) = f(x) when c
This equation to be correctly normalized in minils before needed :
y',z' = f(x)
y = y' when c
z = z' when c
But this new variables where needless since the final translation of when c
is the identity.
Added graphical simulator hepts
Added option -hepts to the compiler: this option formats the output of the simulation loop
to fit with hepts (currently working only on C target code). Typically, suppresses all
"decoration" output, like input or output names, and prints output on single lines.
Currently not implemented in hepts: array types.
Usage: to simulate the node Modulename.f:
heptc -target c -s f -hepts modulename.ept
gcc modulename_c/*.c -o ./f_sim
hepts -mod Modulename -node f -exec ./f_sim
The order switch then reset was wrong, since some reset reset slower inner blocks and equations, to have reset correct after switch it would have been necessary to sample the reset condition correctly (use r when c) using the level_ck... anyway the order seems now irrelevant considering code size.
