The separately compiled occam PROCs now use #PRAGMA OCCAMEXTERNAL, which also discards the "= number" thing at the end. These PROCs then need to be processed differently when adding on the sizes (C externals have one size per dimension, occam externals have the normal array of sizes).
We also now record which processes were originally at the top-level, and keep their original names (i.e. minus the _u43 suffixes) plus an "occam_" prefix to avoid collisions.
This shows that my current scheme is a bit of a hack; if an occam programmer starts a normal PROC with B or C, it will be treated funny during separate compilation. In future I probably need some new form of #PRAGMA to support separate compilation, different from that used to interface with C.
I changed a little bit of the code, but mainly the tests. Several of the remaining failures are actually real failures, so I need to dig through the rest carefully. A lot are failing because the C++ backend is broken.
All the changes in this patch should be undone as soon as possible (building the list of workspaces, and how we sort out the enrollment counts -- that should be done by pulling up PAR replicator counts outside the whole PAR), but for now, it works.
This is along the lines of the Tock ticket. cgtest90 (recursive FUNCTIONs) now passes, but cgtest91 (recursive PROCs) doesn't, because it uses barriers.
Also known as communication semantics, I think. The pass adds an extra channel parameter per mobile (perhaps in future this could be a single extra channel?) that is used to send back the mobile value, and hacked the backend so that the communications to receive these mobiles are done in the right place (after the processes have been run, but before waiting on the barrier for them to complete).
cgtest83 now compiles, runs and passes without a segfault.
Surprisingly, the generated code seems to be compiled by GCC without warnings. Array subscripts are currently unimplemented (a star, i.e. equivalent of subscript 0) is used wherever there should be a subscript.
The new mechanism is based on working out the C type of the original variable/abbrev-mode, the C type of the desired variable/abbrev-mode, and works out how many *s or &s to insert to coerce it to the right type.
Previously, such a function was an IntrinsicFunctionCall inside one expression of an ExpressionList, which the type-checker rejected. I've had to add a new constructor to ExpressionList, and I've quickly hacked together the line in the C backend to make it work -- but it does seem to work.
Fixes#37. The change itself was very easy, but the main problem was to then go and change everything in GenerateCTest that had been written using A.Int as a test type.
This is mostly straightforward: modify the parser to allow direction
decorators in the right places, and extend the type checker to match.
There's some slight awkwardness in that some of the Types functions
have to perform the same checks as the type checker (e.g. directing a
non-channel), so I've tidied up their error messages a bit.
At the backend, I've just added a little pass to strip out all the
DirectedVariables, since the other backend passes don't handle them
gracefully. From the occam/C point of view this is fine, but I'm not
sure if it's going to cause problems for C++.
I'm pretty sure this test used to fail and got fixed, but I think the
test was fixed to match the incorrect code rather than the other way
around. What it's testing is the abbreviation of a channel end as an
identically-typed channel end, which (since they're both pointers)
should look something like:
Channel *c;
Channel *d = c;
Instead it was generating "d = &c", which broke cgtest87.
The change made to GenerateC to fix this feels a bit awkward to me --
it's very simple, but it's yet another special case in genVariableAM.
All of the genVariable code could really use reworking (again),
perhaps to use some internal representation of the C expression...
All the passes now have their information (name, pre-requisites and post- properties) stored at the point where the pass is declared, which means the pass lists are just a simple list of pass functions.
The main consequence of this change was that the tests had to be changed. Now, instead of taking a "pass applied to data" item (type: PassM b), they take both the pass (type: Pass) and source data (type: b), and apply them later. This was the decision that involved the simplest changes to the existing tests (simply unbracketing the application of the pass to the source). I also had to include a few old-style versions though (testPass', testPassShouldFail') for where the functions were being used to test things that weren't actually passes (mainly StructureOccam).
Fixes#48
NameType is only really needed in the parser, so this takes it out of
NameDef, meaning that later passes defining names no longer need to
set an arbitrary NameType for them. The parser gets slightly more
complicated (because some productions now have to return a SpecType
and a NameType too), but lots of other code gets simpler.
The code that removed free names was the only thing outside the parser
using NameType, and it now makes a more sensible decision based on the
SpecType. Since unscoped names previously didn't have a SpecType at
all, I've added an Unscoped constructor to it and arranged matters
such that unscoped names now get a proper entry in csNames.
Fixes#61.
