| Commit message (Collapse) | Author | Files | Lines |
|---|
|
MSVC is not fully C11-compliant, even when compiling with /std:c11.
Some changes were needed to make the codebase compatible. Notably, the
notation a[static N] and a[n] for function parameters of array type is
not supported, so that had to be hidden behind a macro. Atomic types
are also an experimental feature, apparently, but at least they work
with the correct compiler flag.
One thing that MSVC does well, however, is warning on integer conversions
on /W4 level. I am not sure if Clang and GCC have something similar,
so I took this chance to fix some of these.
|
|
|
|
- Use multithreading (works with a sufficiently recent version of
the Microsoft developer tools / C SDK).
- Detect CPU architecture and use AVX2 or NEON when appropriate.
- Automatically detect python installation path.
|
|
With this commit we re-structure how the node expansion in the
H48 solution search works to allow prefetching of pruning values,
showing performance improvements in the range of 30-45% on x86,
depending on table size and solution length.
A small bug fix related to appending solutions is included in this commit.
|
|
Add a trivial check to H48 solver. The benchmarks show, surprisingly,
a more noticeable improvement for larger solvers (h10, h11). I was expecting
it to be more noticeable for smaller solvers.
It is likely that this change is irrelevant to performance and all changes
in the benchmark results are just random chance.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
for H48.
This commit fixes two bugs:
- A bug that caused duplicates solutions for symmetric scrambles.
- An overflow in the maxsols parameter for the H48 solver, which
caused it to find much fewer solutions than existed.
Moreover, the H48 solvers has been improved by reducing by symmetry not
only from the starting position, but also up to the first 4 moves.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
I found out that this gives undefined behavior when then size is 0.
Better not to have it at all, it is confusing for other developers anyway.
|
|
This is a small amendment to c6a77f30f64be73a5e55e06336975f2ecfbb2324,
which changed the way we log solutions while the h48 solver is running.
With the method recently introduced, the main thread checks for solutions
to log every 0.5 seconds, resulting in a possible slowdown of at most
0.5s per solve. The solutions are also logged when all worker threads are
completed. With this new method, when the poll_status() callback function
is NULL, which likely means nissy is not run interactively, we rely
only on the final log on completion of the worker threads. This means
less frequent logging, but at no performance cost.
|
|
|
|
|
|
Before this committ, the solver (via the generic solution-appender
routines in src/solve/solutions.h) and the H48 data generator did some
logging in the worker threads, without using any locks. This was not nice,
but in practice it did not cause any problem, because the log messages
were rare.
However, this turned out to be a problem when building to WASM, because
web workers do not have access to the main JS memory, and therefore
they cannot call functions from the main JS. This includes not only the
callback functions for logging, but also those for polling the status
of the solver (run / pause / stop).
This commit fixes this at the cost or being somewhat inelegant: the
solutions are not logged as they are found, but only every 500ms.
|
|
Relevant changes include:
- Changed the signature of nissy_checkdata().
- Removed expected_distribution.h from tools; this data is now included
in each solver's src/ code.
- Removed distribution check for cocsep; may add back later.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Before this commit I was inconsistently using one of void *, char *
and uint8_t *.
|
|
|
|
|
|
I wanted to make the "optimal" and "maxsolutions" options mutually
exclusive, but in the end I decided there is value in keeping both
(e.g. for specifying a limit to the number of solutions when asking
for "all" optimal").
Now optimal cannot be negative anymore, for the same reason of maxsolutions.
The interface user (shell, UI) will have to take care of handling this
in a way that makes sense for the user. Usually this means setting
the maximum number of solutions to UINT_MAX (or a similar very high
number) when the user wants "all optimal".
|
|
|
|
- Simplified logger to accept only a string, not a variadic list
of args like printf(). This can still use some improvement,
but now it is easier to use from other languages.
- Fixed some misuses of the logger (wrong types etc)
- Renamed some constants
- Fixed some typos in comments.
|
|
|
|
|
|
With this PR the appendsolution routine is extracted from the h48
solver and the new coordinate solver and made generic. This has
many advantages:
- less repetition (even if the two versions are different enough that
*for now* it was not a big deal)
- smaller h48/solve.h file, which is already a big beast
- easier to test the appendsolution() routine separately
|
|
|
|
|
|
|
|
|
|
|
