- remove some obsolete stdout prints
- first version since a while that actually does some useful native calculation. It still has java threads and the pthreads stuff is yet untested.
- the optimization of variables with dirty-flag was invalid because it is not safe to assume that getValue() always occurs later in the phase.
- prepare a lot of stuff for the upcoming almost completely native, multithreaded AdvancerRegistry
- prepare creating JNI threads from native part
- even more profiling preparation
- add more profiling info functions
- add more profiling info functions
- AdvancerRegistry (native) now uses the new List features. Next step, eliminate java part of AdvReg (java ProcessElement implementation classes called via reverse JNI then); after that, output gathered profiling info after finishing processing, then split native AdvancerRegistry in threads
- both list types failed to update the element count in the case that space had to be allocated first
- for the planned transition to pure native and multithreaded AdvancerRegistry incl. profiling, we'll need a thread safe implementation of a List as well as one that orders elements by a given priority.
- add complexity member to ScBase to allow ordering advanceState calls in AdvancerRegistry (more complex cases first is advantageous for multithread situations). Note: the type specific implementations are all missing.
- enter a type info method to ScBase descendants (will be used for resource consumption analysis)
-
- there is no real difference. It's probably not MingW, but eclipse 4.8 code checker that complains, and I think it is the complex code structure that is confusing it. The code should always have been ok in this respect. MingW compiles it fine.
- fix an issue MingW 8.2 complains about a pure virtual function not being implemented in derived class (not sure why, it worked before and I don't really see a significant difference between original and seemingly fixed version)
- a few minor changes and fixes
Allow Lists to create a read only copy for permitting modification in parallel to read loops
- minor optimization: replace 2 divisions by one multiplication
- a lot of optimizations, tests run about 30% faster now. Not yet completely checked for semantic correctness
- minor changes
- minor fixes
- again fix some build problems in exotic c++ setups
- fix 6581 native adapter
- fix a resource conflict bug leading to random crashes
- fixed a 32/64 portability issue
- fixed : not every c++ compiler understood some constructs in monkey's audio code. Now it should work with at least C99/C++98
- fixed the preprocessor tests
- ParserTest now uses the "testfiles/parsertest" test file
- removed redundant stuff
- comment
- some work on the parser
- some additions to the language description
- initial version of language description document
- fix: a lot of grammar symbols didn't get source location
- added a missing initialization
- TDHS works again, the first time with native SVF
- prepare getting TDHS to run again after SVF native transition: inner classes now also need native counterpart. Not yet working
- fix the 880HzSyncSaw test case
- InputFile now works without memory access problem in the "cached" variant. The uncached variant is still broken.
- fixed a parameter mistake done during native transformation of InputFile
- make InputFile run again - but there still is a memory handling bug that keeps us from freeing all used memory completely.
- Panning element now functional
- add a test case for the panner. The level panning works, panning delay is still broken
- attempt to improve panning (not yet tested): in addition to level panning, also implement a delay of the remote side. Needs a test case.
- remove 2 files that were not meant to be versioned
I hope this helps... at least I haven't seen a newer one for german. I also hope that the texts are correct. Some translations would be context dependent and some english words (I took english as a reference) can have multiple meanings. I didn't have a chance to really play it through to see all context.
SoundComp (https://sourceforge.net/p/soundcomp) will be a compiler tool that translates a textual description of a piece of sound into respective sound files. There was a longer almost inactive phase but we are getting active again. SoundComp mainly consists of 3 parts: - 1. a scanner+parser that creates a tree structure of sound related objects from the textual description - 2. a collection of signal processing objects with a uniform interface that can be combined to produce almost any kind of sound...
- PinkFilter now native
- added Panning element. Not yet complete; it would still need to implement a slight (few microsec) delay on the opposite side, proportional to the displacement from center. Presently it only does level panning.
- InputFile now native; fixed a Mixer bug leading to squared output
- Variable now native
- Third now native
- wrong return signature
- make new functions atanh, acosh, asinh available to the parser
- WhiteNoise now native
- propagate samplerate to native without reflection
- a bit more work on math functions
- rename a lot of potentially conflicting class names commit 3
- rename a lot of potentially conflicting class names
- forgot to set PWM as native preferred
-
- some more native variable initialization
- continued linux native name unclashing
- remove some "unused warning" warnings
- svf now works in native
- svf shited to native, compiles and runs but doesn't yet produce correct output
- eliminate some compiler warnings
- missing "return"
- some fixes for "logaritm functions now native"
- logaritm functions now native
- basic math operators +-*/ are now native
- some generator code is now native. Still some problems with some parts, some do already work.
- unregistering must not include reverse JNI reflection code
- further steps to native signal processing
- no more crash also under linux, the cpp list allocations now appear to be memory safe; BUT still no correct output from the native mixers
- native Mixer and AdvancerRegistry no longer crash but do not give output either now.
- fixed runtime link error, but still crashes
- fixing the last cpp build issue, build is again ok, now comes debugging the native mixer, still has memory corruption
- fixing the last cpp build issue, build is again ok, now comes debugging the native mixer
- next step towards a native Mixer element
- next step towards a native Mixer element
- next step towards a native Mixer element
- fixing a lot of c++ issues
- prepare making Mixer native
- some native util classes
- some native util classes
- some improvements for procss references in the parser
- prepare multithreaded calculation, enabled by setting thread counts in the advancer registry before adding the first element. Note that for simple settings the cost of parallelization is much higher than the benefit; it will probably make sense only in extremely complex cases.As an example, Test TS2 becomes about 11 times slower by enabling a second core, instead of the expected speedup. The problem appears to be the internal cost of synchronized/wait/notify, which has to be called about tw...
- some parser problems fixed and some of the errors in the example text corrected
- start building a small but real compiler test case
- note events need not only gate and frequency, but also an intensity (velocity, volume?)