VOID WINAPI UnregisterNotificationWork(PTP_CALLBACK_INSTANCE Instance,
PVOID Context, PTP_WORK Work)
{
HCMNOTIFICATION Handle = static_cast<HCMNOTIFICATION>(Context);
CM_Unregister_Notification(Handle);
CloseThreadpoolWork(Work);
}
NamedPipe::~NamedPipe() {
Close();
base::AutoLock guard(lock_);
if (work_ != nullptr) {
auto local_work = work_;
work_ = nullptr;
base::AutoUnlock unlock(lock_);
WaitForThreadpoolWorkCallbacks(local_work, FALSE);
CloseThreadpoolWork(local_work);
}
}
virtual void schedule(pplx::TaskProc_t proc, void* param)
{
pplx::details::atomic_increment(s_flag);
auto schedulerParam = std::unique_ptr<_Scheduler_Param>(new _Scheduler_Param(proc, param));
auto work = CreateThreadpoolWork(DefaultWorkCallbackTest, schedulerParam.get(), NULL);
if (work == nullptr)
{
throw utility::details::create_system_error(GetLastError());
}
SubmitThreadpoolWork(work);
CloseThreadpoolWork(work);
schedulerParam.release();
}
void
EasyIocp::FreeThreadPool()
{
int i;
for(i = 0; i < worksNum_; ++i)
PostQueuedCompletionStatus(hCompPort_, 0, (ULONG_PTR)0, NULL);
for(i = 0; i < worksNum_; ++i) {
WaitForThreadpoolWorkCallbacks(ptpWorks_[i], FALSE);
CloseThreadpoolWork(ptpWorks_[i]);
}
SysFree(ptpWorks_);
ptpWorks_ = NULL;
print("EasyIocp::FreeThreadPool: free thread pool.");
}
void Server::Destroy()
{
m_ShuttingDown = true;
if (m_AcceptTPWORK != NULL)
{
WaitForThreadpoolWorkCallbacks(m_AcceptTPWORK, true);
CloseThreadpoolWork(m_AcceptTPWORK);
m_AcceptTPWORK = NULL;
}
if (m_listenSocket != INVALID_SOCKET)
{
Network::CloseSocket(m_listenSocket);
CancelIoEx(reinterpret_cast<HANDLE>(m_listenSocket), NULL);
m_listenSocket = INVALID_SOCKET;
}
if (m_pTPIO != NULL)
{
WaitForThreadpoolIoCallbacks(m_pTPIO, true);
CloseThreadpoolIo(m_pTPIO);
m_pTPIO = NULL;
}
if (m_ClientTPCLEAN != NULL)
{
CloseThreadpoolCleanupGroupMembers(m_ClientTPCLEAN, false, NULL);
CloseThreadpoolCleanupGroup(m_ClientTPCLEAN);
DestroyThreadpoolEnvironment(&m_ClientTPENV);
m_ClientTPCLEAN = NULL;
}
EnterCriticalSection(&m_CSForClients);
for (auto client : m_Clients)
{
delete client;
}
m_Clients.clear();
LeaveCriticalSection(&m_CSForClients);
DeleteCriticalSection(&m_CSForClients);
}
__declspec(noinline) bool benchmark_ntp_fs_stat() {
TP_CALLBACK_ENVIRON env;
InitializeThreadpoolEnvironment(&env);
PTP_POOL pool{nullptr};
pool = CreateThreadpool(nullptr);
SetThreadpoolThreadMaximum(pool, 48);
SetThreadpoolThreadMinimum(pool, 12);
PTP_CLEANUP_GROUP group = CreateThreadpoolCleanupGroup();
SetThreadpoolCallbackPool(&env, pool);
SetThreadpoolCallbackCleanupGroup(&env, group, nullptr);
PTP_WORK work_fs_stat = CreateThreadpoolWork(WorkCallback_fs_stat, nullptr, &env);
SubmitThreadpoolWork(work_fs_stat);
WaitForThreadpoolWorkCallbacks(work_fs_stat, false);
CloseThreadpoolWork(work_fs_stat);
CloseThreadpool(pool);
return false;
}
void *LLW_train_thread(void *th_data) {
// Recover data
struct ThreadData *data = (struct ThreadData *)th_data;
const int thread_id = data->thread_id;
const int nprocs = data->nprocs;
struct Model *model = data->model;
struct KernelCache *kernelcache = data->kernelcache;
long chunk_size = data->chunk_size;
const double accuracy = data->accuracy;
double **gradient = data->gradient;
double **H_alpha = data->H_alpha;
double *best_primal_upper_bound = data->best_primal_upper_bound;
int *activeset = data->activeset;
long *nb_SV = data->nb_SV;
double *lp_rhs = data->lp_rhs;
FILE *fp = data->logfile_ptr;
pthread_mutex_unlock(&thread_data_mutex); // Release thread_data for next thread
// Local variables
int do_eval;
char yesno;
long long return_status = -1;
// Prepare the cache
struct TrainingCache cache;
cache.chunk_size = chunk_size;
LLW_alloc_memory(&cache, model->Q, model->nb_data, chunk_size);
cache.kc = kernelcache;
cache.activeset = activeset;
cache.lp_rhs = lp_rhs;
double **delta = matrix(chunk_size, model->Q);
double previous_ratio = 0.0;
double improvement = 1.0;
double theta_opt;
int jump = false;
if(accuracy == 0)
do_eval = 0;
else
do_eval = 1;
/*
Prepare parallel gradient computations:
- the gradient vector is split into NUMTHREADS_GRAD parts (along i)
- each part is updated by a different thread
*/
// max number of threads for gradient updates is nprocs
pthread_t *grad_threads = (pthread_t *)malloc(sizeof(pthread_t) * nprocs);
// start with 1 thread (main load on kernel evaluations)
int numthreads_grad = 1;
void *status;
int rc;
long k;
struct ThreadGradient_data *grad_data = (struct ThreadGradient_data *)malloc(sizeof(struct ThreadGradient_data) * nprocs);
// Disable parallel gradient computation for small data sets
int parallel_gradient_update = 1;
if(model->nb_data < 5000 || nprocs == 1)
parallel_gradient_update = 0;
if(parallel_gradient_update) {
for(k=0;k<nprocs;k++) {
grad_data[k].gradient = gradient;
grad_data[k].H_alpha = H_alpha;
grad_data[k].cache = &cache;
grad_data[k].model = model;
}
grad_data[0].start_i = 1;
grad_data[0].end_i = model->nb_data / numthreads_grad;
for(k=1;k<numthreads_grad-1;k++) {
grad_data[k].start_i = grad_data[k-1].end_i + 1;
grad_data[k].end_i = grad_data[k].start_i + model->nb_data / numthreads_grad -1;
}
if(numthreads_grad>1) {
grad_data[numthreads_grad-1].start_i = grad_data[numthreads_grad-2].end_i + 1;
grad_data[numthreads_grad-1].end_i = model->nb_data;
}
}
#ifdef _WIN32
// Init POOL
TP_WORK ** work;
if(parallel_gradient_update) {
work = malloc(sizeof(TP_WORK *) * nprocs);
for(k=0;k<nprocs;k++)
work[k] = CreateThreadpoolWork(LLW_update_gradient_thread2, (void *) &grad_data[k], NULL);
}
#endif
// Switch to nprocs/4 threads for gradient update when 25% of the kernel matrix is cached
int percentage_step = 1;
long percentage = model->nb_data / 4;
int next_numthreads_grad = nprocs/4;
if(next_numthreads_grad == 0)
next_numthreads_grad = 1;
// Main loop
int thread_stop = 0;
do {
if((TRAIN_SMALL_STEP < TRAIN_STEP) && (model->iter%TRAIN_SMALL_STEP) == 0) {
printf(".");
fflush(stdout);
}
// Select a random chunk of data to optimize
select_random_chunk(&cache,model);
// Compute the kernel submatrix for this chunk
compute_K(&cache,model);
// Enter Critical Section (using and modifying the model)
pthread_mutex_lock(&(model->mutex));
jump = LLW_solve_lp(gradient, &cache, model);
if(jump == false)
jump = LLW_check_opt_sol(gradient,&cache,model);
if(jump == false) {
LLW_compute_delta(delta,&cache,model);
theta_opt = LLW_compute_theta_opt(delta, &cache, model);
if (theta_opt > 0.0) {
*nb_SV += LLW_compute_new_alpha(theta_opt,&cache,model);
if(parallel_gradient_update) {
// Update gradient in parallel
for(k=0;k<numthreads_grad;k++) {
#ifdef _WIN32
SubmitThreadpoolWork(work[k]);
#else
rc = pthread_create(&grad_threads[k], NULL, LLW_update_gradient_thread, (void *) &grad_data[k]);
#endif
}
// Wait for gradient computations to terminate
for(k=0;k<numthreads_grad;k++) {
#ifdef _WIN32
WaitForThreadpoolWorkCallbacks(work[k], FALSE);
#else
rc = pthread_join(grad_threads[k],&status);
#endif
}
}
else {
// old-style non-threaded gradient update (for small data sets)
LLW_update_gradient(gradient,H_alpha, &cache,model);
}
}
}
if((do_eval && (model->iter%TRAIN_STEP) == 0) || EVAL || STOP || (do_eval && model->ratio >= accuracy) )
{
if(fp != NULL)
fprintf(fp,"%ld ",model->iter);
if(EVAL)
printf("\n\n*** Evaluating the model at iteration %ld...\n",model->iter);
// Evaluate how far we are in the optimization
// (prints more info if interrutped by user)
previous_ratio = model->ratio;
model->ratio = MSVM_eval(best_primal_upper_bound, gradient, H_alpha, NULL, model, EVAL, fp);
print_training_info(*nb_SV, model);
improvement = model->ratio - previous_ratio;
if(EVAL) // if interrupted by user (otherwise let the ratio decide if we go on training)
{
printf("\n *** Do you want to continue training ([y]/n)? ");
yesno = getchar();
if(yesno=='n') {
STOP = 1;
}
EVAL = 0; // reset interruption trigger
}
}
// Release kernel submatrix in cache
release_K(&cache);
// Check if a sufficient % of the kernel matrix is cached
if( parallel_gradient_update && cache.kc->max_idx >= percentage ) {
// and switch thread to compute gradient upates instead of kernel rows if it is
thread_stop = switch_thread(nprocs, &numthreads_grad, &next_numthreads_grad, &percentage, &percentage_step, grad_data, thread_id, model->nb_data);
// (threads are actually stopped to leave the CPUs
// to other threads that will compute gradient updates)
}
model->iter++;
// Release mutex: End of critical section
pthread_mutex_unlock(&(model->mutex));
} while(model->iter <= MSVM_TRAIN_MAXIT && (!do_eval || (model->ratio < accuracy && improvement != 0.0)) && !STOP && !thread_stop);
// Release mutex: End of critical section (see below)
pthread_mutex_unlock(&(model->mutex));
#ifdef _WIN32
if(parallel_gradient_update){
for(k=0;k<numthreads_grad;k++)
CloseThreadpoolWork(work[k]);
}
#endif
// compute return_status
if(do_eval && (model->ratio >= accuracy || improvement==0.0))
return_status = 0; // optimum reached or no more improvement.
// Free memory
LLW_free_memory(&cache);
free(delta[1]);free(delta);
free(grad_threads);
free(grad_data);
pthread_exit((void*)return_status);
}
void OnDestroy( void ){
CloseThreadpoolWork( m_pWorkItem );
}
int TestPoolWork(int argc, char* argv[])
{
int index;
PTP_POOL pool;
PTP_WORK work;
PTP_CLEANUP_GROUP cleanupGroup;
TP_CALLBACK_ENVIRON environment;
printf("Global Thread Pool\n");
work = CreateThreadpoolWork((PTP_WORK_CALLBACK) test_WorkCallback, "world", NULL);
if (!work)
{
printf("CreateThreadpoolWork failure\n");
return -1;
}
/**
* You can post a work object one or more times (up to MAXULONG) without waiting for prior callbacks to complete.
* The callbacks will execute in parallel. To improve efficiency, the thread pool may throttle the threads.
*/
for (index = 0; index < 10; index++)
SubmitThreadpoolWork(work);
WaitForThreadpoolWorkCallbacks(work, FALSE);
CloseThreadpoolWork(work);
printf("Private Thread Pool\n");
if (!(pool = CreateThreadpool(NULL)))
{
printf("CreateThreadpool failure\n");
return -1;
}
if (!SetThreadpoolThreadMinimum(pool, 4))
{
printf("SetThreadpoolThreadMinimum failure\n");
return -1;
}
SetThreadpoolThreadMaximum(pool, 8);
InitializeThreadpoolEnvironment(&environment);
SetThreadpoolCallbackPool(&environment, pool);
cleanupGroup = CreateThreadpoolCleanupGroup();
if (!cleanupGroup)
{
printf("CreateThreadpoolCleanupGroup failure\n");
return -1;
}
SetThreadpoolCallbackCleanupGroup(&environment, cleanupGroup, NULL);
work = CreateThreadpoolWork((PTP_WORK_CALLBACK) test_WorkCallback, "world", &environment);
if (!work)
{
printf("CreateThreadpoolWork failure\n");
return -1;
}
for (index = 0; index < 10; index++)
SubmitThreadpoolWork(work);
WaitForThreadpoolWorkCallbacks(work, FALSE);
CloseThreadpoolCleanupGroupMembers(cleanupGroup, TRUE, NULL);
CloseThreadpoolCleanupGroup(cleanupGroup);
DestroyThreadpoolEnvironment(&environment);
/**
* See Remarks at https://msdn.microsoft.com/en-us/library/windows/desktop/ms682043(v=vs.85).aspx
* If there is a cleanup group associated with the work object,
* it is not necessary to call CloseThreadpoolWork !
* calling the CloseThreadpoolCleanupGroupMembers function releases the work, wait,
* and timer objects associated with the cleanup group.
*/
/* CloseThreadpoolWork(work); // this would segfault, see comment above. */
CloseThreadpool(pool);
return 0;
}
static BOOL rfx_process_message_tileset(RFX_CONTEXT* context, RFX_MESSAGE* message, wStream* s, UINT16* pExpecedBlockType)
{
BOOL rc;
int i, close_cnt;
int pos;
BYTE quant;
RFX_TILE* tile;
UINT32* quants;
UINT16 subtype;
UINT32 blockLen;
UINT32 blockType;
UINT32 tilesDataSize;
PTP_WORK* work_objects = NULL;
RFX_TILE_PROCESS_WORK_PARAM* params = NULL;
void *pmem;
if (*pExpecedBlockType != WBT_EXTENSION)
{
WLog_ERR(TAG, "%s: message unexpeced", __FUNCTION__);
return FALSE;
}
*pExpecedBlockType = WBT_FRAME_END;
if (Stream_GetRemainingLength(s) < 14)
{
WLog_ERR(TAG, "RfxMessageTileSet packet too small");
return FALSE;
}
Stream_Read_UINT16(s, subtype); /* subtype (2 bytes) must be set to CBT_TILESET (0xCAC2) */
if (subtype != CBT_TILESET)
{
WLog_ERR(TAG, "invalid subtype, expected CBT_TILESET.");
return FALSE;
}
Stream_Seek_UINT16(s); /* idx (2 bytes), must be set to 0x0000 */
Stream_Seek_UINT16(s); /* properties (2 bytes) */
Stream_Read_UINT8(s, context->numQuant); /* numQuant (1 byte) */
Stream_Seek_UINT8(s); /* tileSize (1 byte), must be set to 0x40 */
if (context->numQuant < 1)
{
WLog_ERR(TAG, "no quantization value.");
return FALSE;
}
Stream_Read_UINT16(s, message->numTiles); /* numTiles (2 bytes) */
if (message->numTiles < 1)
{
WLog_ERR(TAG, "no tiles.");
return FALSE;
}
Stream_Read_UINT32(s, tilesDataSize); /* tilesDataSize (4 bytes) */
if (!(pmem = realloc((void*) context->quants, context->numQuant * 10 * sizeof(UINT32))))
return FALSE;
quants = context->quants = (UINT32*) pmem;
/* quantVals */
if (Stream_GetRemainingLength(s) < (size_t) (context->numQuant * 5))
{
WLog_ERR(TAG, "RfxMessageTileSet packet too small for num_quants=%d", context->numQuant);
return FALSE;
}
for (i = 0; i < context->numQuant; i++)
{
/* RFX_CODEC_QUANT */
Stream_Read_UINT8(s, quant);
*quants++ = (quant & 0x0F);
*quants++ = (quant >> 4);
Stream_Read_UINT8(s, quant);
*quants++ = (quant & 0x0F);
*quants++ = (quant >> 4);
Stream_Read_UINT8(s, quant);
*quants++ = (quant & 0x0F);
*quants++ = (quant >> 4);
Stream_Read_UINT8(s, quant);
*quants++ = (quant & 0x0F);
*quants++ = (quant >> 4);
Stream_Read_UINT8(s, quant);
*quants++ = (quant & 0x0F);
*quants++ = (quant >> 4);
WLog_Print(context->priv->log, WLOG_DEBUG, "quant %d (%d %d %d %d %d %d %d %d %d %d).",
i, context->quants[i * 10], context->quants[i * 10 + 1],
context->quants[i * 10 + 2], context->quants[i * 10 + 3],
context->quants[i * 10 + 4], context->quants[i * 10 + 5],
context->quants[i * 10 + 6], context->quants[i * 10 + 7],
context->quants[i * 10 + 8], context->quants[i * 10 + 9]);
}
if (!(message->tiles = (RFX_TILE**) calloc(message->numTiles, sizeof(RFX_TILE*))))
{
message->numTiles = 0;
return FALSE;
}
if (context->priv->UseThreads)
{
work_objects = (PTP_WORK*) calloc(message->numTiles, sizeof(PTP_WORK));
params = (RFX_TILE_PROCESS_WORK_PARAM*) calloc(message->numTiles, sizeof(RFX_TILE_PROCESS_WORK_PARAM));
if (!work_objects)
{
free(params);
return FALSE;
}
if (!params)
{
free(work_objects);
return FALSE;
}
}
/* tiles */
close_cnt = 0;
rc = TRUE;
for (i = 0; i < message->numTiles; i++)
{
if (!(tile = (RFX_TILE*) ObjectPool_Take(context->priv->TilePool)))
{
WLog_ERR(TAG, "RfxMessageTileSet failed to get tile from object pool");
rc = FALSE;
break;
}
message->tiles[i] = tile;
/* RFX_TILE */
if (Stream_GetRemainingLength(s) < 6)
{
WLog_ERR(TAG, "RfxMessageTileSet packet too small to read tile %d/%d", i, message->numTiles);
rc = FALSE;
break;
}
Stream_Read_UINT16(s, blockType); /* blockType (2 bytes), must be set to CBT_TILE (0xCAC3) */
Stream_Read_UINT32(s, blockLen); /* blockLen (4 bytes) */
if (Stream_GetRemainingLength(s) < blockLen - 6)
{
WLog_ERR(TAG, "RfxMessageTileSet not enough bytes to read tile %d/%d with blocklen=%d",
i, message->numTiles, blockLen);
rc = FALSE;
break;
}
pos = Stream_GetPosition(s) - 6 + blockLen;
if (blockType != CBT_TILE)
{
WLog_ERR(TAG, "unknown block type 0x%X, expected CBT_TILE (0xCAC3).", blockType);
rc = FALSE;
break;
}
Stream_Read_UINT8(s, tile->quantIdxY); /* quantIdxY (1 byte) */
Stream_Read_UINT8(s, tile->quantIdxCb); /* quantIdxCb (1 byte) */
Stream_Read_UINT8(s, tile->quantIdxCr); /* quantIdxCr (1 byte) */
Stream_Read_UINT16(s, tile->xIdx); /* xIdx (2 bytes) */
Stream_Read_UINT16(s, tile->yIdx); /* yIdx (2 bytes) */
Stream_Read_UINT16(s, tile->YLen); /* YLen (2 bytes) */
Stream_Read_UINT16(s, tile->CbLen); /* CbLen (2 bytes) */
Stream_Read_UINT16(s, tile->CrLen); /* CrLen (2 bytes) */
Stream_GetPointer(s, tile->YData);
Stream_Seek(s, tile->YLen);
Stream_GetPointer(s, tile->CbData);
Stream_Seek(s, tile->CbLen);
Stream_GetPointer(s, tile->CrData);
Stream_Seek(s, tile->CrLen);
tile->x = tile->xIdx * 64;
tile->y = tile->yIdx * 64;
if (context->priv->UseThreads)
{
assert(params);
params[i].context = context;
params[i].tile = message->tiles[i];
if (!(work_objects[i] = CreateThreadpoolWork((PTP_WORK_CALLBACK) rfx_process_message_tile_work_callback,
(void*) ¶ms[i], &context->priv->ThreadPoolEnv)))
{
WLog_ERR(TAG, "CreateThreadpoolWork failed.");
rc = FALSE;
break;
}
SubmitThreadpoolWork(work_objects[i]);
close_cnt = i + 1;
}
else
{
rfx_decode_rgb(context, tile, tile->data, 64 * 4);
}
Stream_SetPosition(s, pos);
}
if (context->priv->UseThreads)
{
for (i = 0; i < close_cnt; i++)
{
WaitForThreadpoolWorkCallbacks(work_objects[i], FALSE);
CloseThreadpoolWork(work_objects[i]);
}
free(work_objects);
free(params);
}
for (i = 0; i < message->numTiles; i++)
{
if (!(tile = message->tiles[i]))
continue;
tile->YLen = tile->CbLen = tile->CrLen = 0;
tile->YData = tile->CbData = tile->CrData = NULL;
}
return rc;
}
RFX_MESSAGE* rfx_encode_message(RFX_CONTEXT* context, const RFX_RECT* rects, int numRects,
BYTE* data, int width, int height, int scanline)
{
int i, maxNbTiles, maxTilesX, maxTilesY;
int xIdx, yIdx, regionNbRects;
int gridRelX, gridRelY, ax, ay, bytesPerPixel;
RFX_TILE* tile;
RFX_RECT* rfxRect;
RFX_MESSAGE* message = NULL;
PTP_WORK* workObject = NULL;
RFX_TILE_COMPOSE_WORK_PARAM *workParam = NULL;
BOOL success = FALSE;
REGION16 rectsRegion, tilesRegion;
RECTANGLE_16 currentTileRect;
const RECTANGLE_16 *regionRect;
const RECTANGLE_16 *extents;
assert(data);
assert(rects);
assert(numRects > 0);
assert(width > 0);
assert(height > 0);
assert(scanline > 0);
if (!(message = (RFX_MESSAGE*)calloc(1, sizeof(RFX_MESSAGE))))
return NULL;
region16_init(&tilesRegion);
region16_init(&rectsRegion);
if (context->state == RFX_STATE_SEND_HEADERS)
rfx_update_context_properties(context);
message->frameIdx = context->frameIdx++;
if (!context->numQuant)
{
if (!(context->quants = (UINT32*) malloc(sizeof(rfx_default_quantization_values))))
goto skip_encoding_loop;
CopyMemory(context->quants, &rfx_default_quantization_values, sizeof(rfx_default_quantization_values));
context->numQuant = 1;
context->quantIdxY = 0;
context->quantIdxCb = 0;
context->quantIdxCr = 0;
}
message->numQuant = context->numQuant;
message->quantVals = context->quants;
bytesPerPixel = (context->bits_per_pixel / 8);
if (!computeRegion(rects, numRects, &rectsRegion, width, height))
goto skip_encoding_loop;
extents = region16_extents(&rectsRegion);
assert(extents->right - extents->left > 0);
assert(extents->bottom - extents->top > 0);
maxTilesX = 1 + TILE_NO(extents->right - 1) - TILE_NO(extents->left);
maxTilesY = 1 + TILE_NO(extents->bottom - 1) - TILE_NO(extents->top);
maxNbTiles = maxTilesX * maxTilesY;
if (!(message->tiles = calloc(maxNbTiles, sizeof(RFX_TILE*))))
goto skip_encoding_loop;
if (!setupWorkers(context, maxNbTiles))
goto skip_encoding_loop;
if (context->priv->UseThreads)
{
workObject = context->priv->workObjects;
workParam = context->priv->tileWorkParams;
}
regionRect = region16_rects(&rectsRegion, ®ionNbRects);
if (!(message->rects = calloc(regionNbRects, sizeof(RFX_RECT))))
goto skip_encoding_loop;
message->numRects = regionNbRects;
for (i = 0, rfxRect = message->rects; i < regionNbRects; i++, regionRect++, rfxRect++)
{
int startTileX = regionRect->left / 64;
int endTileX = (regionRect->right - 1) / 64;
int startTileY = regionRect->top / 64;
int endTileY = (regionRect->bottom - 1) / 64;
rfxRect->x = regionRect->left;
rfxRect->y = regionRect->top;
rfxRect->width = (regionRect->right - regionRect->left);
rfxRect->height = (regionRect->bottom - regionRect->top);
for (yIdx = startTileY, gridRelY = startTileY * 64; yIdx <= endTileY; yIdx++, gridRelY += 64 )
{
int tileHeight = 64;
if ((yIdx == endTileY) && (gridRelY + 64 > height))
tileHeight = height - gridRelY;
currentTileRect.top = gridRelY;
currentTileRect.bottom = gridRelY + tileHeight;
for (xIdx = startTileX, gridRelX = startTileX * 64; xIdx <= endTileX; xIdx++, gridRelX += 64)
{
int tileWidth = 64;
if ((xIdx == endTileX) && (gridRelX + 64 > width))
tileWidth = width - gridRelX;
currentTileRect.left = gridRelX;
currentTileRect.right = gridRelX + tileWidth;
/* checks if this tile is already treated */
if (region16_intersects_rect(&tilesRegion, ¤tTileRect))
continue;
if (!(tile = (RFX_TILE*) ObjectPool_Take(context->priv->TilePool)))
goto skip_encoding_loop;
tile->xIdx = xIdx;
tile->yIdx = yIdx;
tile->x = gridRelX;
tile->y = gridRelY;
tile->scanline = scanline;
tile->width = tileWidth;
tile->height = tileHeight;
ax = gridRelX;
ay = gridRelY;
if (tile->data && tile->allocated)
{
free(tile->data);
tile->allocated = FALSE;
}
tile->data = &data[(ay * scanline) + (ax * bytesPerPixel)];
tile->quantIdxY = context->quantIdxY;
tile->quantIdxCb = context->quantIdxCb;
tile->quantIdxCr = context->quantIdxCr;
tile->YLen = tile->CbLen = tile->CrLen = 0;
if (!(tile->YCbCrData = (BYTE *)BufferPool_Take(context->priv->BufferPool, -1)))
goto skip_encoding_loop;
tile->YData = (BYTE*) &(tile->YCbCrData[((8192 + 32) * 0) + 16]);
tile->CbData = (BYTE*) &(tile->YCbCrData[((8192 + 32) * 1) + 16]);
tile->CrData = (BYTE*) &(tile->YCbCrData[((8192 + 32) * 2) + 16]);
message->tiles[message->numTiles] = tile;
message->numTiles++;
if (context->priv->UseThreads)
{
workParam->context = context;
workParam->tile = tile;
if (!(*workObject = CreateThreadpoolWork(
(PTP_WORK_CALLBACK)rfx_compose_message_tile_work_callback,
(void*) workParam,
&context->priv->ThreadPoolEnv)))
{
goto skip_encoding_loop;
}
SubmitThreadpoolWork(*workObject);
workObject++;
workParam++;
}
else
{
rfx_encode_rgb(context, tile);
}
if (!region16_union_rect(&tilesRegion, &tilesRegion, ¤tTileRect))
goto skip_encoding_loop;
} /* xIdx */
} /* yIdx */
} /* rects */
success = TRUE;
skip_encoding_loop:
if (success && message->numTiles != maxNbTiles)
{
void* pmem = realloc((void*) message->tiles, sizeof(RFX_TILE*) * message->numTiles);
if (pmem)
message->tiles = (RFX_TILE**) pmem;
else
success = FALSE;
}
/* when using threads ensure all computations are done */
message->tilesDataSize = 0;
workObject = context->priv->workObjects;
for (i = 0; i < message->numTiles; i++)
{
tile = message->tiles[i];
if (context->priv->UseThreads)
{
if (*workObject)
{
WaitForThreadpoolWorkCallbacks(*workObject, FALSE);
CloseThreadpoolWork(*workObject);
}
workObject++;
}
message->tilesDataSize += rfx_tile_length(tile);
}
region16_uninit(&tilesRegion);
region16_uninit(&rectsRegion);
if (success)
return message;
WLog_ERR(TAG, "%s: failed", __FUNCTION__);
message->freeRects = TRUE;
rfx_message_free(context, message);
return NULL;
}
int TestPoolWork(int argc, char* argv[])
{
int index;
PTP_POOL pool;
PTP_WORK work;
PTP_CLEANUP_GROUP cleanupGroup;
TP_CALLBACK_ENVIRON environment;
printf("Global Thread Pool\n");
work = CreateThreadpoolWork((PTP_WORK_CALLBACK) test_WorkCallback, "world", NULL);
if (!work)
{
printf("CreateThreadpoolWork failure\n");
return -1;
}
/**
* You can post a work object one or more times (up to MAXULONG) without waiting for prior callbacks to complete.
* The callbacks will execute in parallel. To improve efficiency, the thread pool may throttle the threads.
*/
for (index = 0; index < 10; index++)
SubmitThreadpoolWork(work);
WaitForThreadpoolWorkCallbacks(work, FALSE);
CloseThreadpoolWork(work);
printf("Private Thread Pool\n");
pool = CreateThreadpool(NULL);
SetThreadpoolThreadMinimum(pool, 4);
SetThreadpoolThreadMaximum(pool, 8);
InitializeThreadpoolEnvironment(&environment);
SetThreadpoolCallbackPool(&environment, pool);
cleanupGroup = CreateThreadpoolCleanupGroup();
if (!cleanupGroup)
{
printf("CreateThreadpoolCleanupGroup failure\n");
return -1;
}
SetThreadpoolCallbackCleanupGroup(&environment, cleanupGroup, NULL);
work = CreateThreadpoolWork((PTP_WORK_CALLBACK) test_WorkCallback, "world", &environment);
if (!work)
{
printf("CreateThreadpoolWork failure\n");
return -1;
}
for (index = 0; index < 10; index++)
SubmitThreadpoolWork(work);
WaitForThreadpoolWorkCallbacks(work, FALSE);
CloseThreadpoolCleanupGroupMembers(cleanupGroup, TRUE, NULL);
CloseThreadpoolCleanupGroup(cleanupGroup);
DestroyThreadpoolEnvironment(&environment);
CloseThreadpoolWork(work);
CloseThreadpool(pool);
return 0;
}
int main (int argc, char * argv[])
{
DWORD nchar = 0, nword = 0, nline = 0;
PTP_WORK *pWorkObjects;
WORK_OBJECT_ARG ** pWorkObjArgsArray, *pObjectArg;
TP_CALLBACK_ENVIRON cbe; // Callback environment
int nThread, iThrd;
if (!WindowsVersionOK (6, 0))
ReportError ("This program requires Windows NT 6.0 or greater", 1, TRUE);
if (argc < 2) {
printf ("Usage: wcMT_vtp filename ... filename\n");
return 1;
}
/* Create a worker thread for each file on the command line */
nThread = (DWORD)argc - 1;
pWorkObjects = malloc (nThread * sizeof(PTP_WORK));
if (pWorkObjects != NULL)
pWorkObjArgsArray = malloc (nThread * sizeof(WORK_OBJECT_ARG *));
if (pWorkObjects == NULL || pWorkObjArgsArray == NULL)
ReportError ("Cannot allocate working memory for worke item or argument array.", 2, TRUE);
InitializeThreadpoolEnvironment (&cbe);
/* Create a work object argument for each file on the command line.
First put the file names in the thread arguments. */
for (iThrd = 0; iThrd < nThread; iThrd++) {
pObjectArg = (pWorkObjArgsArray[iThrd] = _aligned_malloc (sizeof(WORK_OBJECT_ARG), CACHE_LINE_SIZE));
if (NULL == pObjectArg)
ReportError ("Cannot allocate memory for a thread argument structure.", 3, TRUE);
pObjectArg->filename = argv[iThrd+1];
pObjectArg->kword = pObjectArg->kchar = pObjectArg->kline = 0;
pWorkObjects[iThrd] = CreateThreadpoolWork (wcfunc, pObjectArg, &cbe);
if (pWorkObjects[iThrd] == NULL)
ReportError ("Cannot create consumer thread", 4, TRUE);
SubmitThreadpoolWork (pWorkObjects[iThrd]);
}
/* Worker objects are all submitted. Wait for them */
/* to complete and accumulate the results */
for (iThrd = 0; iThrd < nThread; iThrd++) {
/* Wait for the thread pool work item to complete */
WaitForThreadpoolWorkCallbacks (pWorkObjects[iThrd], FALSE);
CloseThreadpoolWork(pWorkObjects[iThrd]);
}
free (pWorkObjects);
/* Accumulate the results */
for (iThrd = 0; iThrd < argc - 1; iThrd++) {
pObjectArg = pWorkObjArgsArray[iThrd];
nchar += pObjectArg->kchar;
nword += pObjectArg->kword;
nline += pObjectArg->kline;
printf ("%10d %9d %9d %s\n", pObjectArg->kline,
pObjectArg->kword, pObjectArg->kchar,
pObjectArg->filename);
}
free (pWorkObjArgsArray);
printf ("%10d %9d %9d \n", nline, nword, nchar);
return 0;
}
static BOOL rfx_process_message_tileset(RFX_CONTEXT* context, RFX_MESSAGE* message, wStream* s)
{
int i;
int pos;
BYTE quant;
UINT32* quants;
UINT16 subtype;
UINT32 blockLen;
UINT32 blockType;
UINT32 tilesDataSize;
PTP_WORK* work_objects = NULL;
RFX_TILE_WORK_PARAM* params = NULL;
if (Stream_GetRemainingLength(s) < 14)
{
DEBUG_WARN("RfxMessageTileSet packet too small");
return FALSE;
}
Stream_Read_UINT16(s, subtype); /* subtype (2 bytes) must be set to CBT_TILESET (0xCAC2) */
if (subtype != CBT_TILESET)
{
DEBUG_WARN("invalid subtype, expected CBT_TILESET.");
return FALSE;
}
Stream_Seek_UINT16(s); /* idx (2 bytes), must be set to 0x0000 */
Stream_Seek_UINT16(s); /* properties (2 bytes) */
Stream_Read_UINT8(s, context->num_quants); /* numQuant (1 byte) */
Stream_Seek_UINT8(s); /* tileSize (1 byte), must be set to 0x40 */
if (context->num_quants < 1)
{
DEBUG_WARN("no quantization value.");
return TRUE;
}
Stream_Read_UINT16(s, message->num_tiles); /* numTiles (2 bytes) */
if (message->num_tiles < 1)
{
DEBUG_WARN("no tiles.");
return TRUE;
}
Stream_Read_UINT32(s, tilesDataSize); /* tilesDataSize (4 bytes) */
if (context->quants != NULL)
context->quants = (UINT32*) realloc((void*) context->quants, context->num_quants * 10 * sizeof(UINT32));
else
context->quants = (UINT32*) malloc(context->num_quants * 10 * sizeof(UINT32));
quants = context->quants;
/* quantVals */
if (Stream_GetRemainingLength(s) < context->num_quants * 5)
{
DEBUG_WARN("RfxMessageTileSet packet too small for num_quants=%d", context->num_quants);
return FALSE;
}
for (i = 0; i < context->num_quants; i++)
{
/* RFX_CODEC_QUANT */
Stream_Read_UINT8(s, quant);
*quants++ = (quant & 0x0F);
*quants++ = (quant >> 4);
Stream_Read_UINT8(s, quant);
*quants++ = (quant & 0x0F);
*quants++ = (quant >> 4);
Stream_Read_UINT8(s, quant);
*quants++ = (quant & 0x0F);
*quants++ = (quant >> 4);
Stream_Read_UINT8(s, quant);
*quants++ = (quant & 0x0F);
*quants++ = (quant >> 4);
Stream_Read_UINT8(s, quant);
*quants++ = (quant & 0x0F);
*quants++ = (quant >> 4);
DEBUG_RFX("quant %d (%d %d %d %d %d %d %d %d %d %d).",
i, context->quants[i * 10], context->quants[i * 10 + 1],
context->quants[i * 10 + 2], context->quants[i * 10 + 3],
context->quants[i * 10 + 4], context->quants[i * 10 + 5],
context->quants[i * 10 + 6], context->quants[i * 10 + 7],
context->quants[i * 10 + 8], context->quants[i * 10 + 9]);
}
message->tiles = (RFX_TILE**) malloc(sizeof(RFX_TILE*) * message->num_tiles);
ZeroMemory(message->tiles, sizeof(RFX_TILE*) * message->num_tiles);
if (context->priv->UseThreads)
{
work_objects = (PTP_WORK*) malloc(sizeof(PTP_WORK) * message->num_tiles);
params = (RFX_TILE_WORK_PARAM*) malloc(sizeof(RFX_TILE_WORK_PARAM) * message->num_tiles);
}
/* tiles */
for (i = 0; i < message->num_tiles; i++)
{
/* RFX_TILE */
if (Stream_GetRemainingLength(s) < 6)
{
DEBUG_WARN("RfxMessageTileSet packet too small to read tile %d/%d", i, message->num_tiles);
return FALSE;
}
Stream_Read_UINT16(s, blockType); /* blockType (2 bytes), must be set to CBT_TILE (0xCAC3) */
Stream_Read_UINT32(s, blockLen); /* blockLen (4 bytes) */
if (Stream_GetRemainingLength(s) < blockLen - 6)
{
DEBUG_WARN("RfxMessageTileSet not enough bytes to read tile %d/%d with blocklen=%d", i, message->num_tiles, blockLen);
return FALSE;
}
pos = Stream_GetPosition(s) - 6 + blockLen;
if (blockType != CBT_TILE)
{
DEBUG_WARN("unknown block type 0x%X, expected CBT_TILE (0xCAC3).", blockType);
break;
}
message->tiles[i] = rfx_tile_pool_take(context);
if (context->priv->UseThreads)
{
params[i].context = context;
params[i].tile = message->tiles[i];
CopyMemory(&(params[i].s), s, sizeof(wStream));
work_objects[i] = CreateThreadpoolWork((PTP_WORK_CALLBACK) rfx_process_message_tile_work_callback,
(void*) ¶ms[i], &context->priv->ThreadPoolEnv);
SubmitThreadpoolWork(work_objects[i]);
}
else
{
rfx_process_message_tile(context, message->tiles[i], s);
}
Stream_SetPosition(s, pos);
}
if (context->priv->UseThreads)
{
for (i = 0; i < message->num_tiles; i++)
{
WaitForThreadpoolWorkCallbacks(work_objects[i], FALSE);
CloseThreadpoolWork(work_objects[i]);
}
free(work_objects);
free(params);
}
return TRUE;
}
