Progress on the C backend #671
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
and it makes things much more complicated, so let's do without for now.
Slowly getting there
* make Eq an overloaded operator * no effort to support compound types has been made * z3 backend impl not done
Including mallocs, this is much cleaner in the resulting code.
Note that this may affect the logs
It's more tidy, and was required anyway for e.g. division operators
This makes the C runtime interface for date calculations work properly (although we have placeholders instead of `dates_calc` at the moment) Includes a few changes: - use exprs instead of naked_exprs in statements (it's just easier to manipulate: passing a naked_expr when an expr is expected is annoying, while the opposite is trivial) - add position to the `Add_dat_dur` operator, which can fail if no specific rounding mode is set (in the OCaml and C backends) - inline closure calls a bit more in `closure_conversion`, for readability
This adds an optional pass that recursively expands equality tests across structures and enums, on lcalc. NOTE: this is a temporary solution. - all tests are completely inlined, which may be a bit bloated - due to the lack of primitives (and expressive pattern-matching), checking equality on enums generates a 2-level pattern matching, quadratic in the number of constructors - this is completely separate from the monomorphisation pass, which morally should take care of generating this code (as specific functions rather than inlined code) So, while this should work as a place-holder for now, it actually seems more reasonable mid-term (before we do it through monomorphisation) to do this translation at the backend level, i.e. when generating the C code, when we have full access to the representation of enums.
These should be moved to the lcalc->scalc translation
and remove special handling from the backends (in particular, lifting in the C backend)
Note: we don't really detect the changes, maybe a hash could help; but at least this avoids an ugly `Index out of bounds` in the case where the changes caused a line to be shorter than expected.
closure conversion was disabled, but hoisting remained, which was causing trouble.
This concerns operators like Map, etc. This solution helps with testing, but may be temporary if we choose to expand closure application in operator arguments as a part of closure conversion ; it may be better to let the backends relying on it to handle the closures directly though.
it now follows a pattern that was generalised
(of course, their output values aren't tested yet for all modules that don't have asserts; but the testing framework is the next step)
we want our pointers to "const" (in the C sense, i.e. read-only) values in
general; but structures need to be initialised field by field. We handle the
absence of `const` in this case (the pointer target will be made `const` upon
function return).
When we compile with `-O`, though, intermediate variables are removed, and we
have branchings with one side getting an already allocated (`const`) structure,
while the other one will need to init through allocation of individual fields.
In the current state of this patch, we define the structure as non-const and
malloc it when this happens, like in the basic definition case. It works, but in
the branch of the already allocated structure, the malloc is discarded (wasted),
and there is a legitimate warning because we strip the `const` flag from the
source pointer.
I see only two solutions:
- remove all consts (beh, they are very good for understanding what's going on,
and I'm sure that without them we'll fall into more traps later)
- in the branch(es) needing malloc, allocate and initialise to a temporary
non-const structure, then affect it to the return pointer ("sealing" it as
const after initialisation). This basically what happens without `-O` because
we have more intermediate variables... But we want to target that specifically
at cases that need malloc.
A variant of the 2nd option could be to define initialiser functions for all
structs, but we loose in readability since the arguments would be unlabelled.
The cleaner way to do it would probably be at the scalc level, depending on the
`--no-struct-literals` flag.
This happens when there are two branches that are bound to an existing struct, and to a literal one that needs a malloc+non-const; the fix refines the handling of `no_struct_literals` in scalc, to ensure that `SLocalInit` is used on a local temporary variable for the constructs needing malloc.
Also includes some fixes to the propagation of visibility info of types.
now all artefacts related to the module Foo are capitalised, like the module name.
This was referenced Sep 24, 2024
|
Don't forget to check out the changes in related PRs CatalaLang/catala-examples#14 and CatalaLang/french-law#3 — and merge in sync |
denismerigoux
approved these changes
Sep 24, 2024
| @@ -15,10 +15,17 @@ | |||
| (python/** with_prefix runtime_python))) | |||
| (section lib)) | |||
|
|
|||
| ; Rescript runtime | |||
| ;; ; Rescript runtime -- disabled | |||
There was a problem hiding this comment.
Why disabled ? Should we reenable it?
There was a problem hiding this comment.
I don't think so, we could remove the whole directory; the only difference is on the visibility of the thing I guess.
In any case, probably better left for a specific PR.
|
Fixes #204 |
Closed
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Along with many small amelioration / fixes, mainly in the scalc pass and ident renaming processing.