Commit graph

33 commits

Author SHA1 Message Date
Leonardo de Moura
dcd15f3424 refactor(runtime): C backend 2019-08-24 07:40:38 -07:00
Leonardo de Moura
30a6a2ade8 feat(library/init/data, runtime): remove parray support from runtime, and implement them in Lean using Scala/Clojure Radix trees
The Scala/Clojure approach for persistent arrays works great with our
`reset/reuse`. We seem to be much more efficient than their
implementations because of `reset/reuse`. The new approach also seems
better than the old one implemented in the runtime, and has a few
advantages:
1- The reroot procedure used in the old approach required
synchronization for multi-threaded code, or we would need to perform
deep copies when sending `parray` objects between threads.
2- We don't need any runtime extension for the new approach.
3- The old approach used "trail lists" for undoing array updates.
This works well for bactracking search use cases, but it is bad
in use cases where we are simultaneously updating the persistent
arrays that have shared nodes.
2019-06-02 09:18:19 -07:00
Leonardo de Moura
a32fcf33c7 feat(util/array_ref): simple wrapper for creating Lean array objects in C++ 2019-04-29 17:10:35 -07:00
Leonardo de Moura
0d5ac5288a feat(runtime): increase small nat size
In 64-bit machines, the max small nat value should now be (2^63 - 1), and on 32-bit
machines (2^32 - 1).

The main motivation for this modification are the array indexing
operations. With the new representation, if a Nat index is not small,
then it must not be a valid index. This was not true in 64-bit
machines. Example: an array of size 2^33 would fit in memory, and but
an index `i` > 2^32 - 1 would not be a small nat value.
2019-03-26 14:21:03 -07:00
Leonardo de Moura
efa703d2b5 feat(runtime): implement string.iterator primitives in the new runtime
Some of the primitives do not have optimal implementation.

@Kha Could you please check if everything we use in the parser has a
reasonable implementation?
2018-11-15 10:42:23 -08:00
Leonardo de Moura
23202bada1 chore(runtime/object): allow shared objects at string_append and string_push 2018-11-14 16:30:23 -08:00
Leonardo de Moura
a368ed2185 test(tests/util/object): add map test 2018-10-27 20:50:09 -07:00
Leonardo de Moura
46d6f7bfb5 chore(runtime/object): store function pointer as void * inside closure 2018-09-11 14:27:45 -07:00
Leonardo de Moura
7c7eccf6ad test(tests/util/object): add perf test 2018-09-11 08:27:25 -07:00
Leonardo de Moura
0573d7e1d5 fix(runtime/object): parray RC bugs 2018-09-09 12:04:27 -07:00
Leonardo de Moura
36423e4389 fix(runtime): parray memory leaks 2018-09-09 10:33:15 -07:00
Leonardo de Moura
2a24f4a09b test(tests/util/object): add parray test 2018-09-09 10:21:55 -07:00
Leonardo de Moura
4863ca071a chore(runtime): make sure we use the same naming convention for getters and setters 2018-09-09 10:07:00 -07:00
Leonardo de Moura
8f195515a6 feat(runtime): add persistent arrays to runtime 2018-09-09 09:44:38 -07:00
Leonardo de Moura
3ab1ebcb3f feat(init/core): add task 2018-08-21 16:10:07 -07:00
Leonardo de Moura
0b349f1abf chore(*): fix style 2018-08-21 09:32:01 -07:00
Leonardo de Moura
dc1f5c0aa6 feat(runtime/object): task API functions can take thunks as arguments 2018-08-20 09:13:35 -07:00
Leonardo de Moura
684085d93f refactor(runtime/object): delete data needed to execute task after it finishes 2018-08-18 14:33:27 -07:00
Leonardo de Moura
a0b5502821 fix(runtime/object): memory leak and simplify task_object
We remove per task condition_variable and use m_task_finished_cv.
The same condition_variable used to implement `wait_any`.
2018-08-18 10:29:12 -07:00
Leonardo de Moura
4bc8414d2b feat(runtime/object): use "weak pointers" in the task manager, and interrupt tasks at GC time 2018-08-17 15:35:00 -07:00
Leonardo de Moura
24444d65c4 refactor(runtime/object): do not use Lean runtime lists to implement the reverse dependency list in task objects 2018-08-17 14:42:43 -07:00
Leonardo de Moura
1d5411f455 feat(runtime/object): add support for io.wait_any 2018-08-17 13:04:06 -07:00
Leonardo de Moura
5f78087b08 feat(runtime/object): add support for io.has_finished 2018-08-17 12:36:48 -07:00
Leonardo de Moura
ae9eac6781 feat(runtime/object): simplify and more tests 2018-08-17 09:41:22 -07:00
Leonardo de Moura
c863e86429 feat(runtime/object): primitives for interrupting threads 2018-08-17 09:25:40 -07:00
Leonardo de Moura
7a7d443ad5 test(tests/util/object): improve tests 2018-08-17 09:04:08 -07:00
Leonardo de Moura
510a5ffeaa chore(tests/util/object): adjust test 2018-08-17 08:48:11 -07:00
Leonardo de Moura
018542e2e1 fix(runtime/object): task bugs 2018-08-16 21:28:58 -07:00
Leonardo de Moura
c322ef0cab fix(tests/util/object): use new calling convention for apply 2018-08-15 21:17:50 -07:00
Leonardo de Moura
f9bad161c9 chore(tests/util/object): fix "unused" variables warnings in release mode 2018-08-14 15:19:07 -07:00
Leonardo de Moura
bcb37ef862 feat(runtime): thunk serialization 2018-08-14 15:15:12 -07:00
Leonardo de Moura
f036a7ad16 chore(runtime/object): mk_thunk(c) should not modify c's RC
This is useful when we are generating a function that does not use the
"borrow semantics" for an argument `c` which is used in `mk_thunk(c)`.
2018-08-10 18:11:29 -07:00
Leonardo de Moura
ea8e1075d2 feat(runtime/thunk): add runtime support for thunks
We did not use constructor objects for implementing thunks because we
wanted to use `atomic<object *>` to implement the cached result.
2018-08-10 18:11:29 -07:00