/* initialize the simulator */
void
sim_init(void)
{
sim_num_refs = 0;
/* allocate and initialize register file */
regs_init(®s);
/* allocate and initialize memory space */
mem = mem_create("mem");
mem_init(mem);
}
void FrmEmuCreate(FrmEmu *self)
{
/* Parent */
EmuCreate(asEmu(self), "Fermi");
/* Initialize */
self->grids = list_create();
self->pending_grids = list_create();
self->running_grids = list_create();
self->finished_grids = list_create();
self->global_mem = mem_create();
self->global_mem->safe = 0;
self->global_mem_top = 0;
self->total_global_mem_size = 1 << 31; /* 2GB */
self->free_global_mem_size = 1 << 31; /* 2GB */
self->const_mem = mem_create();
self->const_mem->safe = 0;
/* Virtual functions */
asObject(self)->Dump = FrmEmuDump;
asEmu(self)->DumpSummary = FrmEmuDumpSummary;
asEmu(self)->Run = FrmEmuRun;
}
void X86ContextCreate(X86Context *self, X86Emu *emu)
{
/* Baseline initialization */
X86ContextDoCreate(self, emu);
/* Loader */
self->loader = x86_loader_create();
/* Memory */
self->address_space_index = emu->address_space_index++;
self->mem = mem_create();
self->spec_mem = spec_mem_create(self->mem);
/* Signal handlers and file descriptor table */
self->signal_handler_table = x86_signal_handler_table_create();
self->file_desc_table = x86_file_desc_table_create();
}
void SIEmuCreate(SIEmu *self)
{
/* Parent */
EmuCreate(asEmu(self), "SouthernIslands");
/* Initialize */
self->video_mem = mem_create();
self->video_mem->safe = 0;
self->video_mem_top = 0;
self->waiting_work_groups = list_create();
self->running_work_groups = list_create();
/* Set global memory to video memory by default */
self->global_mem = self->video_mem;
/* Virtual functions */
asObject(self)->Dump = SIEmuDump;
asEmu(self)->DumpSummary = SIEmuDumpSummary;
asEmu(self)->Run = SIEmuRun;
}
/* load program into simulated state */
void
sim_load_prog(char *fname, /* program to load */
int argc, char **argv, /* program arguments */
char **envp) /* program environment */
{
/* load program text and data, set up environment, memory, and regs */
ld_load_prog(fname, argc, argv, envp, ®s, mem, TRUE);
#ifdef TARGET_ALPHA
/* pre-decode text segment */
{
unsigned i, num_insn = (ld_text_size + 3) / 4;
fprintf(stderr, "** pre-decoding %u insts...", num_insn);
/* allocate decoded text space */
dec = mem_create("dec");
for (i=0; i < num_insn; i++)
{
enum md_opcode op;
md_inst_t inst;
md_addr_t PC;
/* compute PC */
PC = ld_text_base + i * sizeof(md_inst_t);
/* get instruction from memory */
MD_FETCH_INST(inst, mem, PC);
/* decode the instruction */
MD_SET_OPCODE(op, inst);
/* insert into decoded opcode space */
MEM_WRITE_WORD(dec, PC << 1, (word_t)op);
MEM_WRITE_WORD(dec, (PC << 1)+sizeof(word_t), inst);
}
fprintf(stderr, "done\n");
}
#endif /* TARGET_ALPHA */
}
/* initialize the simulator */
void
sim_init(void)
{
/* allocate and initialize register file */
regs_init(®s);
/* allocate and initialize memory space */
mem = mem_create("mem");
mem_init(mem);
// Initiate empty stage
EMPTY_DE_BUF.inst.a = NOP;
EMPTY_FD_BUF.inst.a = NOP;
buf.fd.inst.a = NOP;
buf.de.inst.a = NOP;
buf.em.inst.a = NOP;
buf.mw.inst.a = NOP;
/* Initiate log */
log_file = fopen("trace.txt", "w+");
}
/* initialize the simulator */
void
sim_init(void)
{
/* allocate and initialize register file */
regs_init(®s);
/* allocate and initialize memory space */
mem = mem_create("mem");
mem_init(mem);
/* initialize stage latches*/
/* IF/ID */
fd.inst.a = NOP;
fd.inst.b = 0;
fd.PC = 0;
fd.NPC = 0;
/* ID/EX */
de.inst.a = NOP;
de.inst.b = 0;
de.PC = 0;
// de.latched = 0;
/* EX/MEM */
em.inst.a = NOP;
em.inst.b = 0;
em.PC = 0;
em.oprand.out1 = DNA;
em.oprand.out2 = DNA;
/* MEM/WB */
mw.inst.a = NOP;
mw.inst.b = 0;
mw.PC = 0;
mw.oprand.out1 = DNA;
mw.oprand.out2 = DNA;
}
void init_openstack_host(){
if(g_openstack_host_network_id != NULL){
mem_destroy(g_openstack_host_network_id);
}
g_openstack_host_network_id = mem_create(sizeof(openstack_network), OPENSTACK_NETWORK_MAX_NUM);
if(g_openstack_host_subnet_id != NULL){
mem_destroy(g_openstack_host_subnet_id);
}
g_openstack_host_subnet_id = mem_create(sizeof(openstack_subnet), OPENSTACK_SUBNET_MAX_NUM);
if(g_openstack_host_port_id != NULL){
mem_destroy(g_openstack_host_port_id);
}
g_openstack_host_port_id = mem_create(sizeof(openstack_port), OPENSTACK_PORT_MAX_NUM);
if(g_openstack_host_node_id != NULL){
mem_destroy(g_openstack_host_node_id);
}
g_openstack_host_node_id = mem_create(sizeof(openstack_node), OPENSTACK_NODE_MAX_NUM);
if (g_openstack_security_group_id != NULL) {
mem_destroy(g_openstack_security_group_id);
}
g_openstack_security_group_id = mem_create(sizeof(openstack_security), OPENSTACK_SECURITY_GROUP_MAX_NUM);
if (g_openstack_host_security_id != NULL) {
mem_destroy(g_openstack_host_security_id);
}
g_openstack_host_security_id = mem_create(sizeof(openstack_node), OPENSTACK_HOST_SECURITY_MAX_NUM);
if (g_openstack_security_rule_id != NULL) {
mem_destroy(g_openstack_security_rule_id);
}
g_openstack_security_rule_id = mem_create(sizeof(openstack_security_rule), OPENSTACK_SECURITY_RULE_MAX_NUM);
g_openstack_host_network_list = NULL;
g_openstack_host_subnet_list = NULL;
g_openstack_host_port_list = NULL;
return;
};
magnify ()
{
struct pixrect *prr, *prw, *prc;
RCOLORS * rcolors;
WCOLORS * wcolors;
WCOLORS * ccolors;
int x1, x2, y1, y2;
int last_x1, last_x2;
int last_y1, last_y2;
int x, y;
int i, j, k, l;
prr = mem_create ( 40, 40, 8 );
prw = mem_create ( 200, 200, 8 );
prc = mem_create ( 200, 200, 8 );
rcolors = (RCOLORS *)mpr_d(prr)->md_image;
wcolors = (WCOLORS *)mpr_d(prw)->md_image;
ccolors = (WCOLORS *)mpr_d(prw)->md_image;
for ( i = 0; i < 200; i++ )
for ( j = 0; j < 200; j++ )
{
wcolors[i][j] = ccolors[i][j] = 8;
}
for ( i = (1<<3); i < (1<<3)+8; i++ ) mapcolor ( i, 0, 0, 0 );
for ( i = (2<<3); i < (2<<3)+8; i++ ) mapcolor ( i, 255, 0, 0 );
for ( i = (3<<3); i < (3<<3)+8; i++ ) mapcolor ( i, 0, 255, 0 );
for ( i = (4<<3); i < (4<<3)+8; i++ ) mapcolor ( i, 255, 255, 0 );
for ( i = (5<<3); i < (5<<3)+8; i++ ) mapcolor ( i, 0, 0, 255 );
for ( i = (6<<3); i < (6<<3)+8; i++ ) mapcolor ( i, 255, 0, 255 );
for ( i = (7<<3); i < (7<<3)+8; i++ ) mapcolor ( i, 0, 255, 255 );
for ( i = (8<<3); i < (8<<3)+8; i++ ) mapcolor ( i, 255, 255, 255 );
pw_putcolormap ( pw, 0, 64, red, green, blue );
writemask ( 7 << 3 );
last_x1 = last_x2 = 0;
last_y1 = last_y2 = 0;
for ( ;; )
{
/*
for ( dev = 0; dev == 0; )
{
*/
notify_dispatch ();
x = dev_x;
y = dev_y;
if ( x < 20 || x > (width-20) ) continue;
if ( y < 20 || y > (height-20) ) continue;
x1 = x - 20; x2 = x + 20;
y1 = y - 20; y2 = y + 20;
if ( x1 == last_x1 && y1 == last_y1 ) continue;
color = 0;
recti ( last_x1, last_y1, last_x1+40, last_y1+40 );
color = 16;
recti ( x1, y1, x2, y2 );
last_x1 = x1; last_y1 = y1;
/*
}
*/
if ( dev == MS_LEFT )
break;
pw_write ( pw, last_x2, last_y2, 200, 200, PIX_SRC, prc, 0, 0 );
pw_read ( prr, 0, 0, 40, 40, PIX_SRC, pw, x-20, height-y-20 );
recti ( x1, y1, x2, y2 );
for ( i = 0; i < 40; i++ )
{
for ( j = 0; j < 40; j++ )
{
for ( k = 1; k <= 3; k++ )
for ( l = 0; l <= 4; l++ )
{
wcolors[i*5+k][j*5+l] = ((rcolors[i][j]&7)+1) << 3;
}
}
}
x1 = x + 60;
if ( x1 > (width-200) )
x1 = x - 260;
x2 = x1 + 199;
y1 = y - 100;
if ( y1 < 0 )
y1 = 0;
if ( y1 > (height-200) )
y1 = height-200;
y2 = y1 + 199;
pw_write ( pw, x1, height - y2, 200, 200, PIX_SRC, prw, 0, 0 );
color = 16;
recti ( x1, y1, x2, y2 );
last_x2 = x1; last_y2 = height - y2;
}
pr_destroy ( prr );
pr_destroy ( prw );
pr_destroy ( prc );
clear ();
writemask ( -1 );
}
struct si_work_group_t *si_work_group_create(unsigned int work_group_id,
struct si_ndrange_t *ndrange)
{
struct si_bin_enc_user_element_t *user_elements;
struct si_wavefront_t *wavefront;
struct si_work_group_t *work_group;
struct si_work_item_t *work_item;
int i;
int lid;
int lidx, lidy, lidz;
int tid;
int user_element_count;
int wavefront_id;
int wavefront_offset;
int work_item_id;
int work_item_gidx_start;
int work_item_gidy_start;
int work_item_gidz_start;
/* Number of in work-items in work-group */
unsigned int work_items_per_group = ndrange->local_size3[0] *
ndrange->local_size3[1] * ndrange->local_size3[2];
assert(work_items_per_group > 0);
/* Number of wavefronts in work-group */
unsigned int wavefronts_per_group = (work_items_per_group +
(si_emu_wavefront_size - 1)) / si_emu_wavefront_size;
assert(wavefronts_per_group > 0);
/* Initialize */
work_group = xcalloc(1, sizeof(struct si_work_group_t));
work_group->id = work_group_id;
work_group->ndrange = ndrange;
/* Create LDS */
work_group->lds_module = mem_create();
work_group->lds_module->safe = 0;
/* Allocate pointers for work-items (will actually be created when
* wavefronts are created) */
work_group->work_items = xcalloc(si_emu_wavefront_size *
wavefronts_per_group, sizeof(void *));
work_group->wavefronts = xcalloc(wavefronts_per_group, sizeof(void *));
work_group->wavefront_count = wavefronts_per_group;
/* Allocate wavefronts and work-items */
SI_FOREACH_WAVEFRONT_IN_WORK_GROUP(work_group, wavefront_id)
{
work_group->wavefronts[wavefront_id] = si_wavefront_create(
work_group->id * wavefronts_per_group + wavefront_id,
work_group);
wavefront = work_group->wavefronts[wavefront_id];
SI_FOREACH_WORK_ITEM_IN_WAVEFRONT(wavefront, work_item_id)
{
wavefront_offset = wavefront_id * si_emu_wavefront_size;
work_group->work_items[wavefront_offset+work_item_id] =
wavefront->work_items[work_item_id];
work_group->work_items[wavefront_offset+work_item_id]->
work_group = work_group;
}
/* *** */
static int bam(const char *scriptfile, const char **targets, int num_targets)
{
struct CONTEXT context;
int build_error = 0;
int setup_error = 0;
int report_done = 0;
/* build time */
time_t starttime = time(0x0);
/* create the cache and tmp directory */
file_createdir(".bam");
/* zero out and create memory heap, graph */
memset(&context, 0, sizeof(struct CONTEXT));
context.graphheap = mem_create();
context.deferredheap = mem_create();
context.graph = node_graph_create(context.graphheap);
context.exit_on_error = option_abort_on_error;
context.buildtime = timestamp();
/* create lua context */
/* HACK: Store the context pointer as the userdata pointer to the allocator to make
sure that we have fast access to it. This makes the context_get_pointer call very fast */
context.lua = lua_newstate(lua_alloctor_malloc, &context);
/* install panic function */
lua_atpanic(context.lua, lf_panicfunc);
/* load cache (thread?) */
if(option_no_cache == 0)
{
/* create a hash of all the external variables that can cause the
script to generate different results */
hash_t cache_hash = 0;
char hashstr[64];
int i;
for(i = 0; i < option_num_scriptargs; i++)
cache_hash = string_hash_add(cache_hash, option_scriptargs[i]);
string_hash_tostr(cache_hash, hashstr);
sprintf(cache_filename, ".bam/%s", hashstr);
event_begin(0, "cache load", cache_filename);
context.cache = cache_load(cache_filename);
event_end(0, "cache load", NULL);
}
/* do the setup */
setup_error = bam_setup(&context, scriptfile, targets, num_targets);
/* done with the loopup heap */
mem_destroy(context.deferredheap);
/* close the lua state */
lua_close(context.lua);
/* do actions if we don't have any errors */
if(!setup_error)
{
event_begin(0, "prepare", NULL);
build_error = context_build_prepare(&context);
event_end(0, "prepare", NULL);
if(!build_error)
{
event_begin(0, "prioritize", NULL);
build_error = context_build_prioritize(&context);
event_end(0, "prioritize", NULL);
}
if(!build_error)
{
if(option_debug_nodes) /* debug dump all nodes */
node_debug_dump(context.graph);
else if(option_debug_nodes_detailed) /* debug dump all nodes detailed */
node_debug_dump_detailed(context.graph);
else if(option_debug_jobs) /* debug dump all jobs */
node_debug_dump_jobs(context.graph);
else if(option_debug_joblist) /* debug dumps the joblist */
context_dump_joblist(&context);
else if(option_debug_dot) /* debug dump all nodes as dot */
node_debug_dump_dot(context.graph, context.target);
else if(option_debug_jobs_dot) /* debug dump all jobs as dot */
node_debug_dump_jobs_dot(context.graph, context.target);
else if(option_dry)
{
}
else
{
/* run build or clean */
if(option_clean)
{
event_begin(0, "clean", NULL);
build_error = context_build_clean(&context);
event_end(0, "end", NULL);
}
else
{
event_begin(0, "build", NULL);
build_error = context_build_make(&context);
event_end(0, "build", NULL);
report_done = 1;
}
}
}
}
/* save cache (thread?) */
if(option_no_cache == 0 && setup_error == 0)
{
event_begin(0, "cache save", cache_filename);
cache_save(cache_filename, context.graph);
event_end(0, "cache save", NULL);
}
/* clean up */
mem_destroy(context.graphheap);
free(context.joblist);
cache_free(context.cache);
/* print final report and return */
if(setup_error)
{
/* no error message on setup error, it reports fine itself */
return setup_error;
}
else if(build_error)
printf("%s: error: a build step failed\n", session.name);
else if(report_done)
{
if(context.num_jobs == 0)
printf("%s: targets are up to date already\n", session.name);
else
{
time_t s = time(0x0) - starttime;
if(s <= 1)
printf("%s: done\n", session.name);
else
printf("%s: done (%d:%.2d)\n", session.name, (int)(s/60), (int)(s%60));
}
}
return build_error;
}
/// Create and Initialize a memory pool.
/// \param[in] pool_def memory pool definition referenced with \ref osPool.
/// \return memory pool ID for reference by other functions or NULL in case of error.
/// \note MUST REMAIN UNCHANGED: \b osPoolCreate shall be consistent in every CMSIS-RTOS.
osPoolId osPoolCreate (const osPoolDef_t *pool_def)
{
return mem_create(pool_def->pool_sz, pool_def->item_sz);
}
struct si_work_group_t *si_work_group_create(unsigned int work_group_id,
struct si_ndrange_t *ndrange)
{
struct si_bin_enc_user_element_t *user_elements;
struct si_wavefront_t *wavefront;
struct si_work_group_t *work_group;
struct si_work_item_t *work_item;
/*MIAOW Start */
struct opencl_si_kernel_t *kernel = ndrange->kernel;
unsigned long long mask = 0x00000000;
/*MIAOW Stop */
int i;
int lid;
int lidx, lidy, lidz;
int tid;
int user_element_count;
int wavefront_id;
int wavefront_offset;
int work_item_id;
int work_item_gidx_start;
int work_item_gidy_start;
int work_item_gidz_start;
/*MIAOW start*/
//UNIT TEST
char unit_test_input_buf[150000];
char *tok = NULL;
char *config_read_result = NULL;
char vreg_str[64][2500];
char sreg_str[2500];
int num_of_threads = 0;
FILE* unit_test_config = fopen("unit_test_config.txt", "r");
if (unit_test_config != 0)
{
ndrange->local_size3[2] = 1;
ndrange->local_size3[1] = 1;
ndrange->global_size3[2] = 1;
ndrange->global_size3[1] = 1;
config_read_result = fgets(unit_test_input_buf, 150000, unit_test_config);
if(config_read_result != NULL)
{
tok = strtok(unit_test_input_buf, ";"); //WG count
ndrange->group_count = atoi(tok);
tok = strtok(NULL, ";"); //total number of threads
num_of_threads = atoi(tok);
ndrange->global_size = atoi(tok);
ndrange->global_size3[0] = atoi(tok);
ndrange->local_size3[0] = atoi(tok);
ndrange->local_size = atoi(tok);
}
}
//WorkGroup count and thread count
//fprintf(config,"%d;%d;\n", ndrange->group_count, ndrange->global_size);
#ifdef MIAOW_DEBUG
fflush(config);
#endif
/*MIAOW stop*/
/* Number of in work-items in work-group */
unsigned int work_items_per_group = ndrange->local_size3[0] *
ndrange->local_size3[1] * ndrange->local_size3[2];
assert(work_items_per_group > 0);
/* Number of wavefronts in work-group */
unsigned int wavefronts_per_group = (work_items_per_group +
(si_emu_wavefront_size - 1)) / si_emu_wavefront_size;
assert(wavefronts_per_group > 0);
/* Initialize */
work_group = xcalloc(1, sizeof(struct si_work_group_t));
work_group->id = work_group_id;
work_group->ndrange = ndrange;
/* Create LDS */
work_group->lds_module = mem_create();
work_group->lds_module->safe = 0;
/* Allocate pointers for work-items (will actually be created when
* wavefronts are created) */
work_group->work_items = xcalloc(si_emu_wavefront_size *
wavefronts_per_group, sizeof(void *));
work_group->wavefronts = xcalloc(wavefronts_per_group, sizeof(void *));
work_group->wavefront_count = wavefronts_per_group;
/* Allocate wavefronts and work-items */
SI_FOREACH_WAVEFRONT_IN_WORK_GROUP(work_group, wavefront_id)
{
work_group->wavefronts[wavefront_id] = si_wavefront_create(
work_group->id * wavefronts_per_group + wavefront_id,
work_group);
wavefront = work_group->wavefronts[wavefront_id];
SI_FOREACH_WORK_ITEM_IN_WAVEFRONT(wavefront, work_item_id)
{
wavefront_offset = wavefront_id * si_emu_wavefront_size;
work_group->work_items[wavefront_offset+work_item_id] =
wavefront->work_items[work_item_id];
work_group->work_items[wavefront_offset+work_item_id]->
work_group = work_group;
}