void sys_init(NppData nppd)
{
// Save NPP handles
_hNpp = nppd._nppHandle;
_hSci1 = nppd._scintillaMainHandle;
_hSci2 = nppd._scintillaSecondHandle;
_hHeap = ::GetProcessHeap();
// Allocate buffers
_cfgDir = (LPWSTR)sys_alloc(MAX_PATH * sizeof WCHAR);
_cfgFile = (LPWSTR)sys_alloc(MAX_PATH * sizeof WCHAR);
_glbFile = (LPWSTR)sys_alloc(MAX_PATH * sizeof WCHAR);
_ctxFile = (LPWSTR)sys_alloc(MAX_PATH * sizeof WCHAR);
_nppVersion = ::SendMessage(_hNpp, NPPM_GETNPPVERSION, 0, 0);
_winVersion = ::SendMessage(_hNpp, NPPM_GETWINDOWSVERSION, 0, 0);
// Get NPP's "plugins\Config" directory.
::SendMessage(_hNpp, NPPM_GETPLUGINSCONFIGDIR, MAX_PATH, (LPARAM)_ctxFile);
// Get SessionMgr config directory and create it if missing.
::StringCchCopyW(_cfgDir, MAX_PATH, _ctxFile);
pth::appendSlash(_cfgDir, MAX_PATH);
::StringCchCatW(_cfgDir, MAX_PATH, PLUGIN_DLL_NAME);
::StringCchCatW(_cfgDir, MAX_PATH, L"\\");
::CreateDirectoryW(_cfgDir, NULL);
// Get the global.xml file pathname and create it if missing.
::StringCchCopyW(_glbFile, MAX_PATH, _cfgDir);
::StringCchCatW(_glbFile, MAX_PATH, GLB_FILE_NAME);
pth::createFileIfMissing(_glbFile, GLB_DEFAULT_CONTENT);
// Get the settings.xml file pathname and load the configuration.
::StringCchCopyW(_cfgFile, MAX_PATH, _cfgDir);
::StringCchCatW(_cfgFile, MAX_PATH, CFG_FILE_NAME);
cfg::loadSettings();
// Create sessions directory if missing.
::CreateDirectoryW(cfg::getStr(kSessionDirectory), NULL);
// Create default session file if missing.
app_confirmDefaultSession();
// Get NPP's contextMenu.xml pathname (two directories up from "plugins\Config").
LPWSTR p = ::wcsstr(_ctxFile, L"plugins\\Config");
::StringCchCopyW(p, MAX_PATH, CTX_FILE_NAME);
// Backup existing config and session files.
if (cfg::getBool(kBackupOnStartup)) {
backupFiles();
}
}
STORAGE* storage_create(const STORAGE_DEVICE_DESCRIPTOR *device_descriptor, STORAGE_DRIVER_CB *driver_cb,
void *driver_param)
{
STORAGE* storage = (STORAGE*)sys_alloc(sizeof(STORAGE));
if (storage)
{
storage->rx_event = event_create();
storage->tx_event = event_create();
storage->device_descriptor = device_descriptor;
storage->driver_cb = driver_cb;
storage->driver_param = driver_param;
storage->media_descriptor = NULL;
storage->host_cb = NULL;
storage->host_io_cb = NULL;
storage->host_io_param = NULL;
storage->reading = false;
storage->writing = false;
storage->queue = NULL;
storage->block_size = 0;
}
else
error(ERROR_MEM_OUT_OF_SYSTEM_MEMORY, "STORAGE");
return storage;
}
static void
grow(equeue_t q) {
void * oarr = q->arr ;
int oalen = q->alen ;
int ohead = q->head ;
ofs_t i ;
q->head = 0 ;
q->alen = 2 * q->alen ;
q->arr = sys_alloc(q->size * q->alen) ;
memset(q->arr, 0, q->size * q->alen) ;
#ifndef MINIMIZE_CODE
if (q->alen * q->size > 1<<12) {
eprintf("QUEUE:expand:getting large:len=%d:%s\n", q->alen * q->size, q->debug) ;
}
#endif
/* This could be done with 2 block copies.
*/
for (i=0;i<q->len;i++) {
memcpy(buf_ofs(q->arr, i * q->size),
buf_ofs(oarr, ((ohead + i) & (oalen - 1)) * q->size),
q->size) ;
}
sys_free(oarr) ;
}
static void* do_alloc_large(size_t len)
{
node_t** ppn = &heap->index[MEM_SIZ_LARGE];
ALIGN4096(len);
void* mem = sys_alloc(NULL, len);
while (*ppn)
{
if ((*ppn)->next && ((*ppn)->next->addr > mem))
{
node_t* tmp = get_free_node();
tmp->next = (*ppn)->next;
tmp->addr = mem;
tmp->len = len;
tmp->stat = MEM_STAT_USED;
(*ppn)->next = tmp;
return mem;
}
ppn = &(*ppn)->next;
}
*ppn = get_free_node();
(*ppn)->addr = mem;
(*ppn)->len = len;
(*ppn)->next = NULL;
(*ppn)->stat = MEM_STAT_USED;
return mem;
}
static ErlDrvData
s1_start(ErlDrvPort port, char *args)
{
descriptor_t *desc;
int i = 0;
i = i;
edtk_debug("%s: starting, port = %ld, args = 0x%lx, %s", __FUNCTION__,
port, (unsigned long) args, args);
if ((desc = (descriptor_t *) sys_alloc(sizeof(descriptor_t))) == NULL) {
return ERL_DRV_ERROR_GENERAL;
}
memset(desc, 0, sizeof(descriptor_t));
desc->port = port;
desc->nextxid = 1;
for (i = 0; i < 32; i++) {
desc->valmap_fd[i] = -1;
}
/* TODO: Finish initializing descriptor members */
/* TODO: Take care of other port initialization tasks */
return (ErlDrvData) desc;
}
PRIVATESTUFF int
exp1_read(descriptor *d, char *buf, int buflen)
{
char cmd;
struct t_data *td = sys_alloc(sizeof(struct t_data));
assert(td != NULL);
assert(buflen >= 4);
cmd = *buf++;
td->cmd = cmd;
td->id = get_int32(buf);
buf += 4;
td->name = NULL;
fprintf(DEBUGIO, "exp1_read: cmd = %d, id = %ld\r\n", cmd, td->id);
switch (cmd) {
case EXP1_REQ_GETHOSTBYNAME:
/* buf = null-terminated hostname */
/* XXX use something other than strdup()? */
td->name = strdup(buf);
driver_async(d->port, KEY, invoke_gethostbyname, (void *) td,
free_t_data);
break;
}
/*
** XXX I guess we'll ignore everything else, let the caller hang.
*/
fprintf(DEBUGIO, "exp1_read: done\r\n");
return 0;
}
static ErlDrvData tun_start(ErlDrvPort port, char *args)
{
struct tun_state *state;
int fd;
int mode;
char dev_name[IFNAMSIZ];
state = (struct tun_state*) sys_alloc(sizeof(struct tun_state));
if (state == NULL) {
errno = ENOMEM; /* appropriate to set errno? */
return ERL_DRV_ERROR_ERRNO;
}
if (!parse_args(args, &mode, state->dev, IFNAMSIZ - 1)) {
return ERL_DRV_ERROR_BADARG;
}
fd = make_if(mode, state->dev);
if (fd < 0) {
return ERL_DRV_ERROR_GENERAL;
}
state->port = port;
state->fd = fd;
state->active = ACTIVE_FALSE;
set_input(state, 0);
return (ErlDrvData)state;
}
void ui_menucreate(ui_widget_t* widget)
{
ui_menudata_t* data=(ui_menudata_t*)sys_alloc(sizeof(ui_menudata_t));
memset(data,0,sizeof(ui_menudata_t));
data->selected=-1;
data->type=UI_MENU_VERTICAL;
widget->data=data;
}
mt_divider_t* mt_createdivider(unsigned int depth,const mt_vector3_t min,const mt_vector3_t max)
{
mt_divider_t* divider=sys_alloc(sizeof(mt_divider_t));
memset(divider,0,sizeof(mt_divider_t));
divider->root=sys_alloc(sizeof(mt_dividernode_t));
memset(divider->root,0,sizeof(mt_dividernode_t));
mt_vec3copy(divider->root->min,min);
mt_vec3copy(divider->root->max,max);
divider->depth=depth;
//Start subdivision
mt_dividenode(divider->root,depth);
return divider;
}
void heap_init(heap_t* heap)
{
assert(sizeof(void*) == 8);
assert(sizeof(page_t) == 4096);
memset(heap, 0, sizeof(heap_t));
heap->ppage = (page_t*)sys_alloc(NULL, sizeof(page_t));
heap->free_node = nodepage_init(heap->ppage);
}
equeue_t equeue_create(debug_t debug, int size) {
equeue_t q ;
q = record_create(equeue_t, q) ;
q->debug = debug ;
q->size = size ;
q->alen = 1<<5 ; /* must be a power of 2 */
q->head = 0 ;
q->len = 0 ;
q->arr = sys_alloc(size * q->alen) ;
memset(q->arr, 0, size * q->alen) ;
return q ;
}
PRIVATESTUFF long
exp1_start(int port, char *args)
{
descriptor *d;
fprintf(DEBUGIO, "exp1_start: starting, args = 0x%lx, %s\r\n", (unsigned long) args, args);
if ((d = (descriptor *) sys_alloc(sizeof(descriptor))) == NULL)
return -1;
d->fd = -1;
d->port = port;
d->flags = 0;
return (long) d;
}
extern void uart_enable(UART_CLASS port, UART_CB *cb, void *param, int priority)
{
if (port < UARTS_COUNT)
{
UART_HW* uart = (UART_HW*)sys_alloc(sizeof(UART_HW));
if (uart)
{
_uart_handlers[port] = uart;
uart->cb = cb;
uart->param = param;
uart->read_buf = NULL;
uart->write_buf = NULL;
uart->read_size = 0;
uart->write_size = 0;
//setup pins
#if defined(STM32F1)
if ((USART_TX_DISABLE_MASK & (1 << port)) == 0)
gpio_enable_afio(UART_TX_PINS[port], AFIO_MODE_PUSH_PULL);
if ((USART_RX_DISABLE_MASK & (1 << port)) == 0)
gpio_enable(UART_RX_PINS[port]);
#if (USART_REMAP_MASK)
if ((1 << port) & USART_REMAP_MASK)
afio_remap();
#endif //USART_REMAP_MASK
#elif defined(STM32F2)
if ((USART_TX_DISABLE_MASK & (1 << port)) == 0)
gpio_enable_afio(UART_TX_PINS[port], port < UART_4 ? AFIO_MODE_USART1_2_3 : AFIO_MODE_UART_4_5_USART_6);
if ((USART_RX_DISABLE_MASK & (1 << port)) == 0)
gpio_enable_afio(UART_RX_PINS[port], port < UART_4 ? AFIO_MODE_USART1_2_3 : AFIO_MODE_UART_4_5_USART_6);
#endif
//power up
if (port == UART_1 || port == UART_6)
RCC->APB2ENR |= RCC_UART[port];
else
RCC->APB1ENR |= RCC_UART[port];
//enable interrupts
NVIC_EnableIRQ(UART_IRQ_VECTORS[port]);
NVIC_SetPriority(UART_IRQ_VECTORS[port], priority);
USART[port]->CR1 |= USART_CR1_UE;
}
else
error_dev(ERROR_MEM_OUT_OF_SYSTEM_MEMORY, DEV_UART, port);
}
else
error_dev(ERROR_DEVICE_INDEX_OUT_OF_RANGE, DEV_UART, port);
}
static int ctl_reply(int rep, char *buf, int len, char **rbuf, int rsize)
{
char* ptr;
if ((len+1) > rsize) {
ptr = (char *)sys_alloc(len+1);
assert(ptr);
*rbuf = ptr;
}
else
ptr = *rbuf;
*ptr++ = rep;
memcpy(ptr, buf, len);
return len+1;
}
static void append_node()
{
page_t* pp = heap->ppage;
while (pp->next)
{
pp = pp->next;
}
pp->next = (page_t*)sys_alloc(NULL, sizeof(page_t));
node_t** ppn = &heap->free_node;
while (*ppn)
{
ppn = &(*ppn)->next;
}
*ppn = nodepage_init(pp->next);
}
/* Grow to a larger size array.
*/
void iq_grow(iq_t iq, seqno_t nlen) {
len_t olen = iq->alen ;
unsigned omask = iq->mask ;
unsigned nmask ;
item_t *na ;
seqno_t abs ;
iq_check(iq) ;
nlen = seqno_max(nlen, olen) ;
nlen = 1 << log2(2*nlen-1) ;
nlen = seqno_max(nlen, 16) ;
nmask = nlen - 1 ;
#if 0
eprintf("nlen=%d lo=%d maxi=%d nhi=%d pop=%d\n", nlen, iq->lo, maxi(iq), iq->lo + nlen,
iq_population(iq)) ;
#endif
na = sys_alloc(sizeof(na[0]) * nlen) ;
for (abs=iq->lo;abs<maxi(iq);abs++) {
seqno_t oi = abs & omask ;
seqno_t ni = abs & nmask ;
na[ni] = iq->arr[oi] ;
}
for (abs=maxi(iq);abs<iq->lo+nlen;abs++) {
seqno_t ni = abs & nmask ;
na[ni].msg = NULL ;
}
#if 0
for (abs=iq->lo;abs<iq->lo+nlen;abs++) {
seqno_t ni = abs & nmask ;
assert(na[ni].ctl == UNSET ||
na[ni].ctl == DATA) ;
}
#endif
sys_free(iq->arr) ;
iq->arr = na ;
iq->alen = nlen ;
iq->mask = nmask ;
iq_check(iq) ;
#if 0
eprintf("pop=%d nmask=0x%08x\n", iq_population(iq), iq->mask) ;
#endif
}
static void mt_dividenode(mt_dividernode_t* node,unsigned int depth)
{
const mt_vector3_t middle={(node->max[0]+node->min[0])/2,
(node->max[1]+node->min[1])/2,
(node->max[2]+node->min[2])/2};
mt_dividernode_t* child;
mt_vector3_t min,max;
int i,j,k;
node->colliders=0;
node->numcolliders=0;
node->maxcolliders=0;
//Do not subdivide more than needed
if(depth<=0)
{
node->children=0;
return;
}
node->children=sys_alloc(sizeof(mt_dividernode_t)*8);
//Subdivide this node between 8 children
for(i=0;i<2;i++)
{
min[0]=(i==0)?node->min[0]:middle[0];
max[0]=(i==0)?middle[0]:node->max[0];
for(j=0;j<2;j++)
{
min[1]=(j==0)?node->min[1]:middle[1];
max[1]=(j==0)?middle[1]:node->max[1];
for(k=0;k<2;k++)
{
min[2]=(k==0)?node->min[2]:middle[2];
max[2]=(k==0)?middle[2]:node->max[2];
child=&node->children[(i<<2)+(j<<1)+k];
mt_vec3copy(child->min,min);
mt_vec3copy(child->max,max);
mt_dividenode(child,depth-1);
}
}
}
}
static void expand_heap(size_t i)
{
void* mem = sys_alloc(NULL, 4096);
node_t** ppn = &heap->index[i];
while (*ppn)
{
if ((*ppn)->addr > mem)
{
break;
}
ppn = &(*ppn)->next;
}
node_t* tmp = get_free_node();
tmp->addr = mem;
tmp->len = 4096;
tmp->next = *ppn;
tmp->stat = MEM_STAT_FREE;
*ppn = tmp;
}
iq_t iq_create(debug_t debug, iq_free_t free) {
iq_t iq = record_create(iq_t, iq) ;
iq->lo = 0 ;
iq->hi = 0 ;
iq->alen = 1 ;
iq->read = 0 ;
iq->name = debug ;
iq->mask = 0 ;
iq->arr = sys_alloc(sizeof(iq->arr[0])) ;
iq->free = free ;
assert(iq->arr) ;
memset(iq->arr, 0, sizeof(iq->arr[0])) ;
iq->arr[0].msg = NULL ;
#if 0
#ifdef USE_GC
gc_tag(iq, debug) ;
gc_tag(iq->arr, debug) ;
#endif
#endif
return iq ;
}
e_texture_t* e_creategradient(e_device_t* device,const unsigned char* values,unsigned int numvalues)
{
e_texture_t* texture;
//We load a gradient that will be used for shading
if(e_pow2(numvalues)!=numvalues)
return 0;
texture=(e_texture_t*)sys_alloc(sizeof(e_texture_t));
texture->width=numvalues;
texture->height=1;
texture->bpp=8;
glGenTextures(1,&texture->id);
glBindTexture(GL_TEXTURE_1D,texture->id);
glTexParameteri(GL_TEXTURE_1D,GL_TEXTURE_MAG_FILTER,GL_NEAREST);
glTexParameteri(GL_TEXTURE_1D,GL_TEXTURE_MIN_FILTER,GL_NEAREST);
glTexImage1D(GL_TEXTURE_1D,0,GL_RGBA,numvalues,0,GL_LUMINANCE,GL_UNSIGNED_BYTE,values);
glTexParameteri(GL_TEXTURE_1D,GL_TEXTURE_WRAP_S,GL_CLAMP);
return texture;
}
static inline QUEUE* svc_queue_create(unsigned int block_size, unsigned int blocks_count, unsigned int align)
{
QUEUE* queue = NULL;
int i;
//first, try to allocate space for queue data. In thread's current mempool
unsigned int align_offset = sizeof(DLIST);
if (align > align_offset)
align_offset = align;
void* mem_block = malloc_aligned(blocks_count * (block_size + align_offset), align_offset);
if (mem_block)
{
queue = sys_alloc(sizeof(QUEUE));
if (queue != NULL)
{
queue->align_offset = align_offset;
queue->mem_block = mem_block;
queue->pull_waiters = NULL;
queue->push_waiters = NULL;
DO_MAGIC(queue, MAGIC_QUEUE);
//set all as free
queue->free_blocks = NULL;
queue->filled_blocks = NULL;
for (i = 0; i < blocks_count; ++i)
dlist_add_tail(&queue->free_blocks, (DLIST*)((unsigned int)mem_block + i * (block_size + align_offset)));
}
else
{
free(mem_block);
fatal_error(ERROR_MEM_OUT_OF_SYSTEM_MEMORY, QUEUE_NAME);
}
}
else
error(ERROR_MEM_OUT_OF_HEAP, svc_thread_name(svc_thread_get_current()));
return queue;
}
// ===========================================================================
// ===========================================================================
void fft_myrmics_cfft2_row(rid_t r_single_row, float **a, float *w, int n,
int tile_size, int blk_size) {
int i, j, k;
int off;
float *pa;
//printf("%d: row begin\r\n", sys_get_worker_id());
// tmp buffer in order to hold the data of a single row
pa = sys_alloc(2 * n * sizeof(float), r_single_row);
for (i = 0; i < blk_size; i++) {
off = 2 * i * blk_size;
// merge data from all blocks "a" into a single table "pa"
for (j = 0; j < tile_size; j++) {
for (k = 0; k < 2*blk_size; k++) {
pa[j*2*blk_size+k] = a[j][k+off];
}
}
fft_myrmics_cfft2(n, pa, pa, w);
// write results back from "pa" to block "a"
for (j = 0; j < tile_size; j++) {
for (k = 0; k < 2*blk_size; k++) {
a[j][k+off] = pa[j*2*blk_size+k];
}
}
}
sys_free(pa);
printf("%d: row end\r\n", sys_get_worker_id());
}
static void s1_outputv(ErlDrvData drv_data, ErlIOVec *ev)
{
descriptor_t *desc = (descriptor_t *) drv_data;
unsigned char cmd;
int p = 0, q = 1;
callstate_t *c = NULL;
int do_async_call = default_async_calls;
unsigned long binlen;
int index;
void *tmp = NULL;
binlen = binlen;
index = index;
tmp = tmp;
if (desc == NULL || ev == NULL || ev->size < 1) {
edtk_debug("%s: bad arg(s)", __FUNCTION__);
return;
}
if (! EV_GET_CHAR(ev, &cmd, &p, &q)) {
edtk_debug("%s: empty command", __FUNCTION__);
reply_error(desc, EINVAL);
}
if ((c = sys_alloc(sizeof(callstate_t))) == NULL) {
reply_error(desc, ENOMEM);
return;
}
c->cmd = cmd;
c->key = NULL;
c->free = sys_free;
c->xid = 0; /* XXX unused right now */
c->o.__expect = 1; /* Default is that expectation is always met */
edtk_debug("%s: my threadid = %lx, cmd = %d", __FUNCTION__, pthread_self(), cmd);
switch (cmd) {
case S1_DEBUG:
EV_GET_UINT32(ev, &edtk_debug_flag, &p, &q);
reply_ok_num(desc, edtk_debug_flag); /* Immediate reply */
sys_free(c);
c = NULL;
break;
case S1_NULL:
c->invoke = invoke_s1_null;
break;
case S1_OPEN:
c->invoke = invoke_s1_open;
EV_GET_UINT32(ev, &binlen, &p, &q);
c->i.filename = (char *) EV_GETPOS(ev, p, q);
if (edtk_ev_forward_N(ev, binlen, &p, &q, 1) < 0) {
goto error;
}
EV_GET_UINT32(ev, &c->i.flags, &p, &q);
break;
case S1_GETFD:
c->invoke = invoke_s1_getfd;
EV_GET_UINT32(ev, &index, &p, &q);
if (desc->valmap_fd[index] == -1) {
goto error;
} else {
edtk_debug("%s: valmap fd index %d = 0x%lx", __FUNCTION__, index, desc->valmap_fd[index]);
c->i.fd = desc->valmap_fd[index];
c->i.__valmap_fd_index = index;
}
break;
case S1_SENDFD:
c->invoke = invoke_s1_sendfd;
EV_GET_UINT32(ev, &c->i.unixdom_fd, &p, &q);
EV_GET_UINT32(ev, &c->i.fd_to_be_sent, &p, &q);
break;
case S1_RECEIVEFD:
c->invoke = invoke_s1_receivefd;
EV_GET_UINT32(ev, &c->i.unixdom_fd, &p, &q);
break;
case S1_CLOSE:
c->invoke = invoke_s1_close;
EV_GET_UINT32(ev, &index, &p, &q);
if (index == -1 && 0 == 1) {
edtk_debug("%s: valmap fd index %d = default value 0x%lx", __FUNCTION__, index, -1);
c->i.fd = -1;
c->i.__valmap_fd_index = -1;
} else if (desc->valmap_fd[index] == -1) {
goto error;
} else {
edtk_debug("%s: valmap fd index %d = 0x%lx", __FUNCTION__, index, desc->valmap_fd[index]);
c->i.fd = desc->valmap_fd[index];
c->i.__valmap_fd_index = index;
}
break;
}
if (c != NULL) {
if (do_async_call) {
driver_async(desc->port, c->key, c->invoke, c, c->free);
} else {
/*
** Execute the bottom half right away, then send the result.
*/
(*(c->invoke))((void *) c);
s1_ready_async((ErlDrvData) desc, (ErlDrvThreadData) c);
/*
** c is already freed for us by s1_ready_async()
*/
}
}
return;
error:
if (c != NULL) {
sys_free(c);
}
reply_error_atom(desc, am_badarg);
}
void* malloc(int size){
return sys_alloc(size);
}
// ===========================================================================
// fft_myrmics() Myrmics version of FFT
// A n x n input table is partitioned into
// tile_size x tile_size sub-blocks where each
// block has size (n/tile_size) x (n/tile_size)
// ===========================================================================
// * INPUTS
// int n Number of rows and columns of the input table
// int tile_size Number of tile rows and columns in the 2D layout
// ===========================================================================
void fft_myrmics(int tile_size, int n) {
rid_t r;
rid_t *ra_row;
rid_t *rb_row;
int blk_size;
int blk_size_sq;
int i, j, k, l;
float ***a;
float ***a_copy;
float ***b;
float ***b_copy;
float **w;
unsigned int time_start;
void *args[4];
unsigned int deps[4];
// Sanity checks
if (n & (n - 1)) {
printf("N must be a power of 2\r\n");
return;
}
if (tile_size & (tile_size - 1)) {
printf("Tile size must be a power of 2\r\n");
return;
}
blk_size = n/tile_size;
blk_size_sq = blk_size * blk_size;
// Create holding region
r = sys_ralloc(0, 99); // highest level
a = sys_alloc(tile_size * sizeof(float **), r);
b = sys_alloc(tile_size * sizeof(float **), r);
w = sys_alloc(tile_size * sizeof(float *), r);
// Initialize tiles and regions
ra_row = sys_alloc(tile_size * sizeof(rid_t), r);
rb_row = sys_alloc(tile_size * sizeof(rid_t), r);
for (i = 0; i < tile_size; i++) {
// Create a region for each row of tiles, for buffers a and b
ra_row[i] = sys_ralloc(r, 0);
rb_row[i] = sys_ralloc(r, 0);
// Allocate pointers for each tile in the row
a[i] = sys_alloc(tile_size * sizeof(float *), ra_row[i]);
b[i] = sys_alloc(tile_size * sizeof(float *), rb_row[i]);
// Allocate tiles in the row
sys_balloc(2* blk_size_sq * sizeof(float), ra_row[i], tile_size,
(void *) a[i]);
sys_balloc(2* blk_size_sq * sizeof(float), rb_row[i], tile_size,
(void *) b[i]);
// Allocate w array
w[i] = sys_alloc(2 * n * sizeof(float), r);
// Initialize tables with some values
for (j = 0; j < tile_size; j++) {
for (k = 0; k < blk_size; k++) {
for (l = 0; l < 2 * blk_size; l++) {
a[i][j][k * 2 * blk_size + l] = 0.01F;
}
}
}
for (j = 0; j < 2 * n; j++) {
w[i][j] = 0.3F;
}
}
// Copy the pointers of the a and b tables, because we'll need them in
// fft_myrmics_Xpose() to spawn the tasks. We need to do this, because in
// fft_myrmics_cfft2_phase() we delegate all r_row[*] to children tasks:
// a[*][*] and b[*][*] are allocated in these regions, so we don't have
// access there anymore from the master task.
a_copy = sys_alloc(tile_size * sizeof(float **), 0);
b_copy = sys_alloc(tile_size * sizeof(float **), 0);
sys_balloc(tile_size * sizeof(float *), 0, tile_size, (void *) a_copy);
sys_balloc(tile_size * sizeof(float *), 0, tile_size, (void *) b_copy);
for (i = 0; i < tile_size; i++) {
for (j = 0; j < tile_size; j++) {
a_copy[i][j] = a[i][j];
b_copy[i][j] = b[i][j];
}
}
// Starting FFT
printf("FFT 2D-block of %d x %d starting split into %d x %d tiles \r\n",
n, n, tile_size, tile_size);
// Start time
time_start = ar_free_timer_get_ticks();
// Run first phase on buffer a
fft_myrmics_cfft2_phase(ra_row, a, w, n, tile_size, blk_size);
// Transpose a->b
fft_myrmics_Xpose(a_copy, rb_row, b, tile_size, blk_size);
// Run second phase on buffer b
fft_myrmics_cfft2_phase(rb_row, b, w, n, tile_size, blk_size);
// Transpose b->a
fft_myrmics_Xpose(b_copy, ra_row, a, tile_size, blk_size);
// Stop time
args[0] = (void *) r;
deps[0] = SYS_TYPE_REGION_ARG | SYS_TYPE_INOUT_ARG;
args[1] = (void *) time_start;
deps[1] = SYS_TYPE_BYVALUE_ARG;
sys_spawn(3, args, deps, 2); // fft_myrmics_time()
// Checksum
args[0] = (void *) r;
deps[0] = SYS_TYPE_REGION_ARG | SYS_TYPE_INOUT_ARG;
args[1] = (void *) a;
deps[1] = SYS_TYPE_BYVALUE_ARG;
args[2] = (void *) tile_size;
deps[2] = SYS_TYPE_BYVALUE_ARG;
args[3] = (void *) blk_size;
deps[3] = SYS_TYPE_BYVALUE_ARG;
sys_spawn(4, args, deps, 4); // fft_myrmics_checksum()
printf("%d: spawns done\r\n", sys_get_worker_id());
}
static e_texture_t* e_loadjpg(e_device_t* device,int mipmaps,fs_handle_t file)
{
struct jpeg_decompress_struct cinfo;
unsigned char* data=0;
unsigned int width,height;
unsigned int stride;
e_jpgerror_t error;
e_jpgsource_t* src=0;
e_texture_t* texture;
cinfo.err=jpeg_std_error(&error.mgr);
error.mgr.error_exit=e_jpgerror;
if(setjmp(error.buffer))
{
//Destroy memory
src=(e_jpgsource_t*)cinfo.src;
jpeg_destroy_decompress(&cinfo);
if(src)
{
sys_free(src->buffer);
sys_free(src);
}
if(data)
sys_free(data);
return 0;
}
cinfo.src=0;
jpeg_create_decompress(&cinfo);
cinfo.src=(struct jpeg_source_mgr*)sys_alloc(sizeof(e_jpgsource_t));
//Init data source
src=(e_jpgsource_t*)cinfo.src;
src->mgr.init_source=e_jpgnop;
src->mgr.fill_input_buffer=e_jpgfillbuffer;
src->mgr.skip_input_data=e_jpgskipdata;
src->mgr.resync_to_restart=jpeg_resync_to_restart;
src->mgr.term_source=e_jpgnop;
src->mgr.bytes_in_buffer=0;
src->mgr.next_input_byte=0;
src->file=file;
src->size=256;
src->buffer=sys_alloc(sizeof(unsigned char)*src->size);
jpeg_read_header(&cinfo,TRUE);
jpeg_start_decompress(&cinfo);
//Do not accept textures with non-power-of-two dimensions
if(cinfo.output_width!=e_pow2(cinfo.output_width)||
cinfo.output_height!=e_pow2(cinfo.output_height))
longjmp(error.buffer,1);
//Force RGB output
cinfo.out_color_space=JCS_RGB;
width=cinfo.output_width;
height=cinfo.output_height;
stride=sizeof(unsigned char)*3*cinfo.output_width;
data=(unsigned char*)sys_alloc(stride*cinfo.output_height);
//Read each scanline
while(cinfo.output_scanline<cinfo.output_height)
{
unsigned char* ptr=data+stride*(cinfo.output_height-(cinfo.output_scanline+1));
unsigned char** row=&ptr;
jpeg_read_scanlines(&cinfo,row,1);
}
jpeg_finish_decompress(&cinfo);
jpeg_destroy_decompress(&cinfo);
sys_free(src->buffer);
sys_free(src);
texture=(e_texture_t*)sys_alloc(sizeof(e_texture_t));
texture->width=width;
texture->height=height;
texture->bpp=24;
glGenTextures(1,&texture->id);
glBindTexture(GL_TEXTURE_2D,texture->id);
glTexParameteri(GL_TEXTURE_2D,GL_TEXTURE_MAG_FILTER,GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D,GL_TEXTURE_MIN_FILTER,GL_LINEAR);
if(mipmaps)
gluBuild2DMipmaps(GL_TEXTURE_2D,3,texture->width,texture->height,GL_RGB,GL_UNSIGNED_BYTE,data);
else
glTexImage2D(GL_TEXTURE_2D,0,GL_RGB,texture->width,texture->height,0,GL_RGB,GL_UNSIGNED_BYTE,data);
sys_free(data);
return texture;
}
static e_texture_t* e_loadpng(e_device_t* device,int mipmaps,fs_handle_t file)
{
unsigned int width,height;
unsigned int rowbytes;
unsigned char* data;
unsigned char** rows;
unsigned int i;
int bpp,type;
int format;
png_structp png;
png_infop info;
e_texture_t* texture;
if(!e_checkpng(file))
return 0;
png=png_create_read_struct(PNG_LIBPNG_VER_STRING,NULL,(png_error_ptr)0,(png_error_ptr)0);
if(!png)
return 0;
info=png_create_info_struct(png);
if(!info)
{
png_destroy_read_struct(&png,0,0);
return 0;
}
//Needed by libpng
if(setjmp(png_jmpbuf(png)))
{
png_destroy_read_struct(&png,&info,0);
if(rows)
sys_free(rows);
if(data)
sys_free(data);
return 0;
}
//Read up to image data
png_set_read_fn(png,(png_voidp)file,e_readpng);
png_set_sig_bytes(png,8);
png_read_info(png,info);
png_get_IHDR(png,info,&width,&height,&bpp,&type,0,0,0);
if(width!=e_pow2(width)||
height!=e_pow2(height))
longjmp(png_jmpbuf(png),1);
//Convert this image to fit the required format
if(type==PNG_COLOR_TYPE_PALETTE)
png_set_palette_to_rgb(png);
if(type==PNG_COLOR_TYPE_GRAY||
type==PNG_COLOR_TYPE_GRAY_ALPHA)
png_set_gray_to_rgb(png);
if(type==PNG_COLOR_TYPE_GRAY||bpp<8)
png_set_expand(png);
if(png_get_valid(png,info,PNG_INFO_tRNS))
png_set_tRNS_to_alpha(png);
png_read_update_info(png,info);
rowbytes=png_get_rowbytes(png,info);
//Allocate memory to store the image
data=(unsigned char*)sys_alloc(rowbytes*height);
if(!data)
longjmp(png_jmpbuf(png),1);
rows=(unsigned char**)sys_alloc(sizeof(unsigned char*)*height);
if(!rows)
longjmp(png_jmpbuf(png),1);
for(i=0;i<height;i++)
rows[i]=data+(height-(i+1))*rowbytes;
//Read png
png_read_image(png,rows);
png_read_end(png,0);
sys_free(rows);
texture=(e_texture_t*)sys_alloc(sizeof(e_texture_t));
texture->width=width;
texture->height=height;
glGenTextures(1,&texture->id);
glBindTexture(GL_TEXTURE_2D,texture->id);
texture->bpp=(type&PNG_COLOR_MASK_ALPHA)?32:24;
format=(type&PNG_COLOR_MASK_ALPHA)?GL_RGBA:GL_RGB;
glTexParameteri(GL_TEXTURE_2D,GL_TEXTURE_MAG_FILTER,GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D,GL_TEXTURE_MIN_FILTER,GL_LINEAR);
if(mipmaps)
gluBuild2DMipmaps(GL_TEXTURE_2D,(type&PNG_COLOR_MASK_ALPHA)?4:3,texture->width,texture->height,format,GL_UNSIGNED_BYTE,data);
else
glTexImage2D(GL_TEXTURE_2D,0,format,texture->width,texture->height,0,format,GL_UNSIGNED_BYTE,data);
sys_free(data);
return texture;
}
void* myrealloc(void* addr, size_t len)
{
size_t i1 = ind(len);
size_t i2 = MEM_SIZ_LARGE;
node_t* prev = NULL;
node_t* pn = NULL;
if (addr_find(addr, &prev, &pn, &i2))
{
if (i2 != MEM_SIZ_LARGE)
{
if (i1 == i2)
{
return addr;
}
else
{
void* ret = myalloc(len);
memcpy(ret, addr, (pn->len < len) ? pn->len : len);
myfree(addr);
return ret;
}
}
else
{
if (i1 == i2)
{
ALIGN4096(len);
if (len <= pn->len)
{
return pn->addr;
}
else
{
size_t delta = len - pn->len;
void* p = sys_try_alloc((void*)((size_t)pn->addr + pn->len), delta);
if (p)
{
pn->len = len;
return pn->addr;
}
else
{
void* src = pn->addr;
pn->addr = sys_alloc(NULL, len);
memcpy(pn->addr, src, pn->len);
sys_free(src, pn->len);
pn->len = len;
return pn->addr;
}
}
}
else
{
void* ret = myalloc(len);
memcpy(ret, addr, len);
myfree(addr);
return ret;
}
}
}
return NULL;
}