lcoogan
/
flucoma-sc
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add fdNRTBase: Base class for Non Real-Time SC commands

Browse Source
nix
Owen Green 7 years ago
parent 41ed298279
commit f4a9560e9c
1 changed files with 187 additions and 142 deletions
Show Stats Download Patch File Download Diff File

329
include/fdNRTBase.hpp
Unescape Escape View File

@ -1,168 +1,213 @@
//Can I reallocate buffers on the server? Yes I can.
#pragma once
#include "data/FluidTensor.hpp"
#include "clients/common/FluidParams.hpp"
#include "SC_PlugIn.h"
#include <boost/align/aligned_alloc.hpp>
#include <vector>
#include "data/FluidTensor.hpp"
static InterfaceTable *ft;
//static InterfaceTable *ft;
namespace fluid {
namespace sc{
using ViewType = fluid::FluidTensorView<float, 2>;
/**
A descendent of SndBuf that will populate itself
from the NRT mirror buffers given a world and a bufnum
**/
struct NRTBuf: public SndBuf
{
NRTBuf(SndBuf& b):SndBuf(b){}
NRTBuf(World* world,long bufnum):
NRTBuf(*World_GetNRTBuf(world,bufnum))
{
this->samplerate = world->mFullRate.mSampleRate;
}
};
/**
A combination of SndBuf and parameter::BufferAdaptor (which, in turn, exposes FluidTensorView<float,2>), for simple transfer of data
Given a World* and a buffer number, this will populate its SndBuf stuff
from the NRT mirror buffers, and create a FluidTensorView wrapper of
appropriate dimensions.
The SndBuf can then be 'transferred' back to the RT buffers once we're done with it,
and SC notified of the update. (In the context of SequencedCommands, in which this is meant
to be used, this would happen at Stage3() on the real-time thread)
nSamps = rows
nChans = columns
**/
class SCBufferView: public NRTBuf, public parameter::BufferAdaptor
{
public:
SCBufferView() = delete;
SCBufferView(SCBufferView&) = delete;
SCBufferView operator=(SCBufferView&) = delete;
SCBufferView(long bufnum,World* world):
NRTBuf(world,bufnum),
BufferAdaptor({0,{static_cast<size_t>(frames),static_cast<size_t>(channels)}},NRTBuf::data),
mBufnum(bufnum), mWorld(world)
{}
/**
A descendent of SndBuf that will populate itself
from the NRT mirror buffers given a world and a bufnum
**/
struct NRTBuf: public SndBuf
~SCBufferView() = default;
void assignToRT(World* rtWorld)
{
NRTBuf(SndBuf& b):SndBuf(b){}
NRTBuf(World* world,size_t bufnum):
NRTBuf(*World_GetNRTBuf(world,bufnum))
{}
};
SndBuf* rtBuf = World_GetBuf(rtWorld,mBufnum);
*rtBuf = static_cast<SndBuf>(*this);
rtWorld->mSndBufUpdates[mBufnum].writes++;
}
//No locks in (vanilla) SC, so no-ops for these
void acquire() override {}
void release() override {}
/**
A combination of SndBuf and FluidTensorView, for simple transfer of data
Given a World* and a buffer number, this will populate its SndBuf stuff
from the NRT mirror buffers, and create a FluidTensorView wrapper of
appropriate dimensions.
The SndBuf can then be 'transferred' back to the RT buffers once we're done with it,
and SC notified of the update. (In the context of SequencedCommands, in which this is meant
to be used, this would happen at Stage3() on the real-time thread)
**/
class SCBufferView: public NRTBuf,ViewType
//Validity is based on whether this buffer is within the range the server knows about
bool valid() const override {
return (mBufnum >=0 && mBufnum < mWorld->mNumSndBufs);
}
void resize(size_t frames, size_t channels, size_t rank) override {
SndBuf* thisThing = static_cast<SndBuf*>(this);
float* oldData = thisThing->data;
mWorld->ft->fBufAlloc(this, channels * rank, frames, this->samplerate);
FluidTensorView<float,2> v= FluidTensorView<float,2>(NRTBuf::data,0,static_cast<size_t>(frames),static_cast<size_t>(channels * rank));
static_cast<FluidTensorView<float,2>&>(*this) = std::move(v);
if(oldData)
boost::alignment::aligned_free(oldData);
}
protected:
bool equal(BufferAdaptor* rhs) const override
{
public:
SCBufferView() = delete;
SCBufferView(SCBufferView&) = delete;
SCBufferView operator=(SCBufferView&) = delete;
SCBufferView(size_t bufnum,World* world):
NRTBuf(world,bufnum),
ViewType({0,{static_cast<size_t>(frames),
static_cast<size_t>(channels)}},NRTBuf::data),
mBufnum(bufnum), mWorld(world)
{}
void assignToRT()
{
SndBuf* rtBuf = World_GetBuf(mWorld,mBufnum);
*rtBuf = static_cast<SndBuf>(*this);
mWorld->mSndBufUpdates[mBufnum].writes++;
}
private:
size_t mBufnum;
World * mWorld;
};
SCBufferView* x = dynamic_cast<SCBufferView*>(rhs);
if(x)
{
return mBufnum == x->mBufnum;
}
return false;
}
class NRTCommandBase{
template <typename T>
using AsyncFn = bool (T::*)();
template <typename T>
using AsyncCleanup = void (T::*) ();
template<typename T, AsyncFn<T> F>
static bool call(World*,void* x)
{
return (static_cast<T*>(x)->*F)();
}
template<typename T, AsyncCleanup<T> F>
static void call(World*, void* x)
long mBufnum;
World* mWorld;
};
class NRTCommandBase{
using param_type = fluid::parameter::Instance;
public:
NRTCommandBase() = delete;
NRTCommandBase(NRTCommandBase&) = delete;
NRTCommandBase& operator=(NRTCommandBase&) = delete;
NRTCommandBase(std::vector<param_type> params,
void* inUserData):
// mWorld(inWorld),mReplyAddr(replyAddr), mCompletionMsgData(completionMsgData), mCompletionMsgSize(completionMsgSize),
mParams(params)
{}
~NRTCommandBase() = default;
template <typename T> using AsyncFn = bool (T::*)(World* w);
template <typename T> using AsyncCleanup = void (T::*)();
template <typename T, AsyncFn<T> F>
static bool call(World* w,void* x){return (static_cast<T*>(x)->*F)(w);}
template<typename T>
static void call(World*, void* x){delete static_cast<T*>(x);}
template<typename T, AsyncFn<T> Stage2, AsyncFn<T> Stage3, AsyncFn<T> Stage4>
void cmd(World* world, std::string name, void* replyAddr, char* completionMsgData, size_t completionMsgSize)
{
(world->ft->fDoAsynchronousCommand)(world, replyAddr,name.c_str(),this,
call<T,Stage2>, call<T,Stage3>, call<T,Stage4>,call<T>,
completionMsgSize,completionMsgData);
}
protected:
// World * mWorld;
// InterfaceTable *ft;
long bufNUm;
void* mReplyAddr;
const char* cmdName;
void *cmdData;
char* mCompletionMsgData;
size_t mCompletionMsgSize;
std::vector<param_type> mParams;
};
//This wraps a class instance in a function call to pass to SC
template<typename NRT_Plug>
void command(World *inWorld, void* inUserData, struct sc_msg_iter *args, void *replyAddr)
{
//Iterate over parameter descriptions associated with this client object, fill with data from language side
std::vector<parameter::Instance> params = NRT_Plug::client_type::newParameterSet();
for (auto&& p: params)
{
switch(p.getDescriptor().getType())
{
case parameter::Type::Buffer:
{
(static_cast<T*>(x)->*F)();
long bufNum = static_cast<long>(args->geti());
if(bufNum >= 0){
SCBufferView* buf = new SCBufferView(bufNum,inWorld);
p.setBuffer(buf);
}
break;
}
template<typename T, AsyncFn<T> Stage2, AsyncFn<T> Stage3, AsyncFn<T> Stage4, AsyncCleanup<T> Cleanup>
void cmd(std::string name)
case parameter::Type::Long:
{
(*ft->fDoAsynchronousCommand)( mWorld, mReplyAddr,name.c_str(),this,
call<T,Stage2>, call<T,Stage3>, call<T,Stage4>,call<T,Cleanup>,
mCompletionMsgSize,mCompletionMsgData);
p.setLong(static_cast<long>(args->geti()));
break;
}
public:
NRTCommandBase() = delete;
NRTCommandBase(NRTCommandBase&) = delete;
NRTCommandBase& operator=(NRTCommandBase&) = delete;
NRTCommandBase(World *inWorld, void* inUserData, struct sc_msg_iter *args, void *replyAddr):
mWorld(inWorld),mReplyAddr(replyAddr){}
virtual ~NRTCommandBase() = default;
/**Override these**/
virtual bool process() { return true; } //NRT
virtual bool post_processing() { return true; } //RT
virtual bool post_complete() { return true; } //NRT
void cleanup() {}
/**Probably not this though**/
void runCommand(std::string name)
case parameter::Type::Float:
{
cmd<NRTCommandBase, &NRTCommandBase::process, &NRTCommandBase::post_processing, &NRTCommandBase::post_complete, &NRTCommandBase::cleanup> (name);
p.setFloat(args->getf());
break;
}
private:
protected:
World * mWorld;
void* mReplyAddr;
const char* cmdName;
void *cmdData;
size_t mCompletionMsgSize;
char* mCompletionMsgData;
void handleCompletionMessage(struct sc_msg_iter *args)
default:
{
mCompletionMsgSize = args->getbsize();
mCompletionMsgData = 0;
if(mCompletionMsgSize)
{
//allocate string
mCompletionMsgData = (char*)RTAlloc(mWorld,mCompletionMsgSize);
args->getb(mCompletionMsgData,mCompletionMsgSize);
}
p.setLong(static_cast<long>(args->geti()));
}
};
} //namespace supercollider
}
}
//Deal with the completion message at the end, if any
size_t completionMsgSize = args->getbsize();
char* completionMsgData = 0;
if(completionMsgSize)
{
//allocate string
completionMsgData = (char*)inWorld->ft->fRTAlloc(inWorld,completionMsgSize);
args->getb(completionMsgData,completionMsgSize);
}
//Make a new pointer for our plugin, and set it going
NRT_Plug* cmd = new NRT_Plug(params, inUserData);
cmd->runCommand(inWorld, replyAddr, completionMsgData, completionMsgSize);
}
} //namespace sc
}//namespace fluid
template<typename NRT_Plug>
void command(World *inWorld, void* inUserData, struct sc_msg_iter *args, void *replyAddr)
{
NRT_Plug cmd(inWorld, inUserData, args, replyAddr);
cmd.runCommand("AysncCommand");
}
template <typename NRT_Plug>
template <typename NRT_Plug,typename NRT_Client>
void registerCommand(InterfaceTable* ft, const char* name)
{
//(World *inWorld, void* inUserData, struct sc_msg_iter *args, void *replyAddr);
PlugInCmdFunc cmd = command<NRT_Plug>;
(*ft->fDefinePlugInCmd)(name,cmd,nullptr);
PlugInCmdFunc cmd = fluid::sc::command<NRT_Plug>;
(*ft->fDefinePlugInCmd)(name,cmd,nullptr);
}
//PluginLoad(BufferFunTime) {
//
// using fluid::sc::NRTCommandBase;
//
// registerCommand<NRTCommandBase>(inTable, "ASyncBufMatch");
//
//// ft = inTable;
//// //BufGen version: all in the NRT thread
////// DefineBufGen("BufMatch", BufferMatch);
//// //ASync version: swaps between NRT and RT threads
//// DefinePlugInCmd("AsyncBufMatch", ASyncBufferFun_Main, nullptr);
////
//}

Write Preview
Loading…
Cancel
Save