Commit Graph

2676 Commits

Author SHA1 Message Date
Torsten Ruger
6606593169 add parallel testing possibility
3 time speedup on 8 cores, 2,5 on 4. with

NCPU=8 RUBYOPT=-rminitest/parallel_fork ruby test/test_all.rb
2018-09-07 16:08:53 +03:00
Torsten Ruger
7fc104a019 Use minitest-profile to find long tests
Only found one big fibo
All else due to large amount of objects
(now with factories it jumped from hundreds to thousands)
2018-09-02 13:58:42 +03:00
Torsten Ruger
8a81d41d5e Move booting to RubyXCompiler init
Also pass the source into the compile method.
This way compiler can be reused for subsequent compile.
Does remove some double boots, but no major time save
2018-09-02 13:57:19 +03:00
Torsten Ruger
d73e1526cd Some docs and to_s testing
somewhat code_climate inspired
2018-09-01 15:54:25 +03:00
Torsten Ruger
2bb6ad5f61 remove unused code
Arrays and Hashes have to resolve to object creation calls
Not to constants. In the future off course
2018-09-01 15:14:07 +03:00
Torsten Ruger
f798173a09 change to_risc and builtin code according to last commit
Wherever space was loaded to get to the next_message
we now load the Message factory.
Otherwise much the same, only the attribute is next_object, not next_message
The binary is growing with 1k objects per factory, so i had to fix (hack) arm to handle bigger constants
close #14
2018-09-01 11:28:53 +03:00
Torsten Ruger
d964e9ea9d let spce keep the messages in a factory #14
Like Integers and addresses before, messages are now in a factory
Factories keep allocated (uninitialised) objects, had to make init public to call it
2018-09-01 11:24:16 +03:00
Torsten Ruger
0a390cc5a9 shaves an instruction off resolve_method
by loading nil directly,  not space first
2018-09-01 11:20:59 +03:00
Torsten Ruger
c3b026a180 expand constant load
slightly hacky, but in good tradition
previous implementation only worked until 16 significant bits, which is getting to little
this one just keeps adding more instructions to arrive at the constant by force
There are surely cleverer ways of doing this, ie by using the barrel shifter
A start on #15, admittedly a hack
2018-08-31 23:28:31 +03:00
Torsten Ruger
b2339dc330 fix address nil values 2018-08-30 16:38:00 +03:00
Torsten Ruger
dc12c1d70b add the addresses from labels as constants 2018-08-29 21:06:29 +03:00
Torsten Ruger
ea7f3c9653 remove the last_object from chain
also tried to keep the first around as was done in space
partially to not have to add the consrtants seperately
didn't work, as chain gets broken
also this has to be redone anyway
2018-08-29 21:05:54 +03:00
Torsten Ruger
c983dcf0eb move return address generation to factory
removes the list from space
adds a ReturnAddress factory instead
and uses these throughout
2018-08-29 21:02:49 +03:00
Torsten Ruger
f993ccefe3 litte bit of docs 2018-08-24 18:49:44 +03:00
Torsten Ruger
71ab369c71 use factory to generte intergers in space
start with just integer factory in space
change all the hand-out code
still #14
2018-08-24 18:49:21 +03:00
Torsten Ruger
d396da16e3 start with #14 by implementing factory
page was maybe a too low level name
pages may be the unit of the syscall, but after that objects desolve (maybe later to be added on from different pages)
Factory has the job of handing out a new instance of a type
it keeps a freelist for that and a reserve
2018-08-23 19:55:06 +03:00
Torsten Ruger
9687d6611f avoid adding risc instructions twice
that causes loops in the chain
infinite loops in the code that are hard to debug
closes #11
2018-08-19 17:29:04 +03:00
Torsten Ruger
f85fe8a2cb fix bug in slot_load and definition
move parfait helper for reuse
2018-08-19 15:36:51 +03:00
Torsten Ruger
253d2fead8 change all to_risc functions to add directly
still #11
test not working yet
2018-08-19 13:18:25 +03:00
Torsten Ruger
57dc6c45bb remove the code_builder
this is core of #11
rename compiler_builder to just builder
and change all builder uses to use that
some test change as code is not returned anymore
2018-08-19 13:16:07 +03:00
Torsten Ruger
b294208025 continue with #11
slots are the most tricky, especially testing
2018-08-19 13:06:00 +03:00
Torsten Ruger
047a36178f start with #11
by adding striaght to compiler
2018-08-19 12:56:44 +03:00
Torsten Ruger
5b2c7745fe move the methods test to mains
previous commit made the mains tests more general
this joins methods tests here
so we can run them on arm too
fix #11
2018-08-18 20:06:15 +03:00
Torsten Ruger
80264c5322 mains tests can include any code
not wrapping as main anymore
(must still include mains)
first part of #11
2018-08-18 19:42:14 +03:00
Torsten Ruger
233b83510f fixed some disabled tests 2018-08-17 22:43:17 +03:00
Torsten Ruger
176b12d896 last two to_risc converted to builder
no change of tests, good sign
2018-08-16 20:28:42 +03:00
Torsten Ruger
da9dc30c20 move dynamic_call to builder
no change of test, great
also remove some duplicate code that was hanging around in method_compiler
2018-08-16 10:43:41 +03:00
Torsten Ruger
ce157ffa94 change block_yield to_risc to use builder
only changes the order of two instructions
2018-08-16 08:58:49 +03:00
Torsten Ruger
1dabe0fda1 finish the idea behind #8, conditionally creating variables
by using space? , the ? makes for conditional creation of the variable
This is especially useful for constants (ie space)
2018-08-15 19:59:17 +03:00
Torsten Ruger
01752aab05 remove the last use of an add_xx function, closing #9 2018-08-15 19:37:03 +03:00
Torsten Ruger
43fa7ccbcc redid div10 with builder
div10 is right at the edge of what can be understood
no matter the (assmbler) syntax
2018-08-15 19:30:40 +03:00
Torsten Ruger
252ae6de72 finally get rid of the fixme in div10
create (load/reduce) the int once and transfer.
Save a cruicial 2 instructions

Also expanded the variable name possibilities with _self, __const , _1 and _2
2018-08-15 18:18:21 +03:00
Torsten Ruger
15337e10be move putstring to use builder 2018-08-15 17:52:21 +03:00
Torsten Ruger
e953cc90d5 remove some more add_xx functions
also load labels by dsl now
2018-08-14 20:24:48 +03:00
Torsten Ruger
520dc7b41f remove ad_transfer in save_message
towards removing all add_
change to set_builder and chaining in reg values
2018-08-14 20:08:58 +03:00
Torsten Ruger
37461a1727 add the ! syntax to create variables in builder, fix #8
now a variable has to be created before being used
thus it is save to develop contracts where a certain name
must exist in the scope
Maybe the syntax starts getting a bit weird, but at least the ! is a common symbol in ruby
2018-08-14 19:39:46 +03:00
Torsten Ruger
0bf008b351 bit of documentation 2018-08-14 11:23:19 +03:00
Torsten Ruger
fb54d68020 redo restore after syscall with builder 2018-08-13 18:48:54 +03:00
Torsten Ruger
55bc9c5273 rewrite new_int with builder 2018-08-13 18:02:34 +03:00
Torsten Ruger
86462e238a large test changes due to change in cc
calling convention does affect
less than before, but still a LOT
2018-08-12 15:02:23 +03:00
Torsten Ruger
fee9e261a5 still some names that needed changing 2018-08-12 14:48:20 +03:00
Torsten Ruger
25d7b8bd83 chaning calling convention not to add frame/arg types
these types are only needed to debug and can be gotten from the method (also in the mesage)
just saving the 6 instructions for every call
This was made possible through previous commits on fake_memory access
2018-08-12 14:47:05 +03:00
Torsten Ruger
28b4a2b373 add line numbers to array output 2018-08-12 13:58:58 +03:00
Torsten Ruger
3d113b4d83 fix more of the changed names
add a few self.
2018-08-12 13:10:44 +03:00
Torsten Ruger
66e0d4ae26 cache index resolution
resolve once and reuse
also use non checking version of get/set
as index per definition must be ok
This creates some speedup, but mostly avoids some weird hang in Thread::Queue.pop for many seconds
2018-08-12 13:09:34 +03:00
Torsten Ruger
84de400529 some test change because of renaming
next is a keyword, can't be used as instance writer
fake_memory gets the pobject for debug
2018-08-11 19:17:20 +03:00
Torsten Ruger
b0aefe31fe make all instances attr read writers
unfortunately the writers have to have self.var =
otherwise it is just a local var
Also need to make the type explicit for all
Protocol included memory_length on the class for now
2018-08-11 19:15:34 +03:00
Torsten Ruger
e6df473647 changing the adapter to have fake memory for every object
Before instance variables were used to store the actual data
While this worked it was a double mental loop and (more importantly) did not allow the same memory access
Ie, in the interpreter the types for args had to be set correctly
But they don't need to be until we walk the list, when we can get the types from the method.
In short, now the fake memory may be used as memory, by indexing into it

We now have to use attr reader/writers that map to the fake memory
And while this is possible, it means a lot more self. than i would like
Also the compiler whill have to undo those attr to instance acess at some point
2018-08-11 19:13:09 +03:00
Torsten Ruger
d74e9c2c40 removed some obsolete builder helpers
cleaner code with dsl,
just div10 undone
2018-08-09 21:10:05 +03:00
Torsten Ruger
99a95d40a6 also redid the get/set internal byte with builder
had to pimp register value to use a semblance of the dsl
using <= for bytes and << for words
2018-08-09 20:16:55 +03:00