d54bca3067
Get basic superclass method calls working.
2013-12-13 08:37:49 -08:00
4957728893
Move OP_CLOSURE and OP_LIST down.
...
Since they're less frequently used, pushing them down speeds
things up a little bit.
2013-12-13 06:47:34 -08:00
5aaaa33552
Revise benchmarks:
...
- Ditch JS since it's in a different league.
- Make binary_trees and fib run faster.
- Compare using best time instead of mean.
2013-12-12 16:59:57 -08:00
8e71660ce6
Reorder the method types.
...
Seems to slightly improve perf.
2013-12-12 07:39:27 -08:00
4ba5c38bbf
Generate default constructors at compile time.
...
Also re-ordered a few of the cases in the interpret loop.
Removing a bit of code from CODE_CLASS and CODE_CALL triggered
some branch alignment issues which had a negative perf impact.
The new ordering seems to cancel that out.
2013-12-12 07:34:16 -08:00
e191424a70
Make "super" a reserved word.
2013-12-11 08:02:17 -08:00
6e16df08f0
Add note about default constructor.
2013-12-11 08:00:25 -08:00
71c26c6d89
Add support for just control which language benchmarks are run.
2013-12-10 23:02:24 -08:00
510e3fc383
Add test for default constructor.
2013-12-10 21:58:02 -08:00
03c7b01b0f
Start writing performance doc.
2013-12-10 21:51:58 -08:00
c1cfe1b9f6
Handle inherited fields correctly.
2013-12-10 16:13:25 -08:00
1e2449893e
Add baseline comparison to benchmark script.
...
This way, I can track performance improvements and regressions.
2013-12-07 22:15:25 -08:00
42b04c6c90
Fix bug in loops.
...
It was still popping the stack for the body even though loop
bodies are statements now.
2013-12-07 18:47:40 -08:00
eb1e5b48d6
Fix up benchmarks and add method call one.
2013-12-07 18:43:39 -08:00
3f24515496
Don't allow method names to overflow.
2013-12-06 20:09:43 -08:00
6a0d550562
Bitwise ~ operator.
2013-12-04 22:09:31 -08:00
7355dd3dc7
Fill in some more basic syntax docs.
2013-12-04 21:51:23 -08:00
c91fcd12cc
Treat ";" like newlines.
2013-12-04 19:12:21 -08:00
8521d5ad79
Bunch more work on docs.
2013-12-04 07:46:41 -08:00
047cfacede
Include "include" directory in source code metrics.
2013-12-04 07:46:23 -08:00
157944aa27
Get closures working!
...
In the process, I had to change the grammar. There is now a strong
separation between statements and expressions. The code was just wrong
before when it popped locals at the end of a block scope because there
could be temporaries on the stack if the block was in expression
position. This fixes that.
Still need to implement closing over `this`.
2013-12-04 07:43:50 -08:00
c14b115c02
Use char* instead of offsets for tokens.
...
Simplifies a bunch of code that works with token strings.
2013-12-01 15:22:24 -08:00
e4043d69b4
Unify code for ending compiler.
...
There's now a single code path for the end of a chunk of bytecode, so
we can eventually put code there for capturing closures.
2013-12-01 14:59:56 -08:00
19811143a0
Always use provided allocator for allocation.
2013-12-01 10:22:32 -08:00
9188c00bc0
Start adding navigation to the docs.
2013-12-01 09:54:51 -08:00
d2e82d8967
Clean up parsing a bit.
2013-12-01 00:04:46 -08:00
60c97b5c2f
Make sure we don't go over the maximum number of locals.
2013-11-30 18:51:27 -08:00
88852960ca
Handle local variable shadowing.
2013-11-30 18:28:01 -08:00
74c414937c
Add JS version of binary tree benchmark.
2013-11-29 20:25:00 -08:00
f6c9848706
Ignore scratch script.
2013-11-29 16:20:44 -08:00
edb9032052
Get binary_trees benchmark working.
...
Wren is actually doing well in it:
wren mean: 1.9441 median: 1.9428 std_dev: 0.0260
lua mean: 3.5992 median: 3.6033 std_dev: 0.0156
python mean: 3.6667 median: 3.7097 std_dev: 0.1340
ruby mean: 1.3941 median: 1.3914 std_dev: 0.0091
2013-11-29 16:19:13 -08:00
ce89dd9c2f
Fix compiling constructors.
2013-11-29 16:17:43 -08:00
8c9ce024ee
Move debug code into separate module.
2013-11-29 16:17:33 -08:00
b390bdf604
Clean up constructor tests.
2013-11-29 15:21:36 -08:00
325bb281a5
Allow 16 arguments, and handle going over that correctly.
2013-11-29 15:08:27 -08:00
5f1d5cfa85
Do not allow fields outside of classes.
2013-11-29 10:53:56 -08:00
1b353e9380
Make pinned list a linked list instead of hardcoded array.
2013-11-29 09:18:13 -08:00
715fecdc7c
Rename files and add some docs.
2013-11-28 08:11:50 -08:00
9d9f9c11ca
Clean up "primitive" code.
2013-11-28 08:00:55 -08:00
1fc9fe131d
Clean up value code.
2013-11-27 20:00:03 -08:00
c2f4345af4
Remove stale TODO.
2013-11-27 11:34:35 -08:00
1bfa252a70
List.removeAt().
2013-11-27 10:20:32 -08:00
f8ded27376
List.clear and List.insert().
2013-11-26 23:11:11 -08:00
56449cdbef
Reorganize tests.
2013-11-26 22:52:00 -08:00
897a396599
List.add() method.
2013-11-26 18:02:10 -08:00
eaeec01dc6
Get inheritance and core classes more correctly wired up.
2013-11-26 07:52:37 -08:00
af694053a5
Add JavaScript to benchmark runner.
2013-11-26 07:50:12 -08:00
af6b73208c
Start defining actual API for embedding Wren.
2013-11-25 07:47:02 -08:00
f9b2aed4dd
Unify to a single reallocation function.
2013-11-25 07:14:51 -08:00
ece3f3b83b
Add optional support for computed gotos.
...
In my benchmark, this gives like a 5% speed boost. But I'm not sure
it's taking full advantage of it. The XCode project is using LLVM,
which has support for it, but may not optimize it as much as GCC.
There are also some GCC flags I may need to set (-fno-gcse?) to get the
best perf on GCC. Needs more investigation.
2013-11-24 21:48:47 -08:00
