static void fmt_ssha_init(struct fmt_main *pFmt)
{
char *kpc;
opencl_init("$JOHN/ssha_opencl_kernel.cl", gpu_id);
// create kernel to execute
crypt_kernel = clCreateKernel(program[gpu_id], "sha1_crypt_kernel", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel. Double-check kernel name?");
if( (kpc = getenv("LWS")) == NULL){
create_clobj(SSHA_NUM_KEYS);
find_best_workgroup();
release_clobj();
}else {
local_work_size = atoi(kpc);
}
if( (kpc = getenv("KPC")) == NULL){
max_keys_per_crypt = SSHA_NUM_KEYS;
create_clobj(SSHA_NUM_KEYS);
} else {
if (atoi(kpc) == 0){
//user chose to die of boredom
max_keys_per_crypt = SSHA_NUM_KEYS;
create_clobj(SSHA_NUM_KEYS);
find_best_kpc();
} else {
max_keys_per_crypt = atoi(kpc);
create_clobj(max_keys_per_crypt);
}
}
printf("Local work size (LWS) %d, Keys per crypt (KPC) %d\n",(int)local_work_size,max_keys_per_crypt);
pFmt->params.max_keys_per_crypt = max_keys_per_crypt;
}
static void fmt_ssha_init(struct fmt_main *pFmt)
{
char *temp;
opencl_init("$JOHN/ssha_kernel.cl", gpu_id, platform_id);
// create kernel to execute
crypt_kernel = clCreateKernel(program[gpu_id], "sha1_crypt_kernel", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel. Double-check kernel name?");
if ((temp = cfg_get_param(SECTION_OPTIONS, SUBSECTION_OPENCL,
LWS_CONFIG)))
local_work_size = atoi(temp);
if ((temp = getenv("LWS")))
local_work_size = atoi(temp);
if (!local_work_size) {
create_clobj(SSHA_NUM_KEYS);
find_best_workgroup();
release_clobj();
}
if ((temp = cfg_get_param(SECTION_OPTIONS, SUBSECTION_OPENCL,
KPC_CONFIG)))
max_keys_per_crypt = atoi(temp);
else
max_keys_per_crypt = SSHA_NUM_KEYS;
if ((temp = getenv("KPC")))
max_keys_per_crypt = atoi(temp);
if (max_keys_per_crypt) {
create_clobj(max_keys_per_crypt);
} else {
//user chose to die of boredom
max_keys_per_crypt = SSHA_NUM_KEYS;
create_clobj(SSHA_NUM_KEYS);
find_best_kpc();
}
printf("Local work size (LWS) %d, Keys per crypt (KPC) %d\n",(int)local_work_size,max_keys_per_crypt);
pFmt->params.max_keys_per_crypt = max_keys_per_crypt;
}
static void init(struct fmt_main *pFmt)
{
assert(sizeof(hccap_t) == HCCAP_SIZE);
inbuffer =
(wpapsk_password *) malloc(sizeof(wpapsk_password) *
MAX_KEYS_PER_CRYPT);
outbuffer =
(wpapsk_hash *) malloc(sizeof(wpapsk_hash) * MAX_KEYS_PER_CRYPT);
mic = (mic_t *) malloc(sizeof(mic_t) * MAX_KEYS_PER_CRYPT);
//listOpenCLdevices();
opencl_init("$JOHN/wpapsk_kernel.cl", gpu_id, platform_id);
/// Alocate memory
cl_int cl_error;
mem_in =
clCreateBuffer(context[gpu_id], CL_MEM_READ_ONLY, insize, NULL,
&cl_error);
HANDLE_CLERROR(cl_error, "Error alocating mem in");
mem_setting =
clCreateBuffer(context[gpu_id], CL_MEM_READ_ONLY, settingsize,
NULL, &cl_error);
HANDLE_CLERROR(cl_error, "Error alocating mem setting");
mem_out =
clCreateBuffer(context[gpu_id], CL_MEM_WRITE_ONLY, outsize, NULL,
&cl_error);
HANDLE_CLERROR(cl_error, "Error alocating mem out");
crypt_kernel = clCreateKernel(program[gpu_id], "wpapsk", &cl_error);
HANDLE_CLERROR(cl_error, "Error creating kernel");
clSetKernelArg(crypt_kernel, 0, sizeof(mem_in), &mem_in);
clSetKernelArg(crypt_kernel, 1, sizeof(mem_out), &mem_out);
clSetKernelArg(crypt_kernel, 2, sizeof(mem_setting), &mem_setting);
find_best_workgroup();
atexit(release_all);
}
static void init(struct fmt_main *pFmt)
{
opencl_init("$JOHN/cryptmd5_opencl_kernel.cl", gpu_id);
///Alocate memory on the GPU
cl_int cl_error;
mem_in =
clCreateBuffer(context[gpu_id], CL_MEM_READ_ONLY, insize, NULL,
&cl_error);
mem_salt =
clCreateBuffer(context[gpu_id], CL_MEM_READ_ONLY, saltsize, NULL,
&cl_error);
mem_out =
clCreateBuffer(context[gpu_id], CL_MEM_WRITE_ONLY, outsize, NULL,
&cl_error);
///Assign kernel parameters
crypt_kernel = clCreateKernel(program[gpu_id], KERNEL_NAME, &cl_error);
clSetKernelArg(crypt_kernel, 0, sizeof(mem_in), &mem_in);
clSetKernelArg(crypt_kernel, 1, sizeof(mem_out), &mem_out);
clSetKernelArg(crypt_kernel, 2, sizeof(mem_salt), &mem_salt);
find_best_workgroup();
//atexit(release_all);
}
/* ------- Initialization ------- */
static void init(struct fmt_main * self) {
char * tmp_value;
char * task = "$JOHN/sha256_kernel.cl";
opencl_init_dev(ocl_gpu_id, platform_id);
source_in_use = device_info[ocl_gpu_id];
if ((tmp_value = getenv("_TYPE")))
source_in_use = atoi(tmp_value);
opencl_build_kernel(task, ocl_gpu_id);
// create kernel(s) to execute
crypt_kernel = clCreateKernel(program[ocl_gpu_id], "kernel_crypt", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel. Double-check kernel name?");
cmp_kernel = clCreateKernel(program[ocl_gpu_id], "kernel_cmp", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel_cmp. Double-check kernel name?");
global_work_size = get_task_max_size();
local_work_size = 0;
if (source_in_use != device_info[ocl_gpu_id]) {
device_info[ocl_gpu_id] = source_in_use;
fprintf(stderr, "Selected runtime id %d, source (%s)\n", source_in_use, task);
}
if ((tmp_value = cfg_get_param(SECTION_OPTIONS,
SUBSECTION_OPENCL, LWS_CONFIG)))
local_work_size = atoi(tmp_value);
if ((tmp_value = getenv("LWS")))
local_work_size = atoi(tmp_value);
//Check if local_work_size is a valid number.
if (local_work_size > get_task_max_work_group_size()){
fprintf(stderr, "Error: invalid local work size (LWS). Max value allowed is: %zd\n" ,
get_task_max_work_group_size());
local_work_size = 0; //Force find a valid number.
}
self->params.max_keys_per_crypt = global_work_size;
if (!local_work_size) {
local_work_size = get_task_max_work_group_size();
create_clobj(global_work_size, self);
find_best_workgroup(self);
release_clobj();
}
if ((tmp_value = cfg_get_param(SECTION_OPTIONS,
SUBSECTION_OPENCL, GWS_CONFIG)))
global_work_size = atoi(tmp_value);
if ((tmp_value = getenv("GWS")))
global_work_size = atoi(tmp_value);
//Check if a valid multiple is used.
global_work_size = get_multiple(global_work_size, local_work_size);
if (global_work_size)
create_clobj(global_work_size, self);
else {
//user chose to die of boredom
global_work_size = get_task_max_size();
find_best_gws(self);
}
fprintf(stderr, "Local work size (LWS) %d, global work size (GWS) %zd\n",
(int) local_work_size, global_work_size);
self->params.max_keys_per_crypt = global_work_size;
}
size_t select_device(int jtrUniqDevNo, struct fmt_main *fmt) {
cl_int err;
const char *errMsg;
opencl_init_opt("$JOHN/kernels/pbkdf2_kernel.cl", jtrUniqDevNo, NULL);
globalObj[jtrUniqDevNo].krnl[0] = clCreateKernel(program[jtrUniqDevNo], "pbkdf2_preprocess", &err);
if (err) {
fprintf(stderr, "Create Kernel pbkdf2_preprocess FAILED\n");
return 0;
}
globalObj[jtrUniqDevNo].krnl[1] = clCreateKernel(program[jtrUniqDevNo], "pbkdf2_iter", &err);
if (err) {
fprintf(stderr, "Create Kernel pbkdf2_iter FAILED\n");
return 0;
}
globalObj[jtrUniqDevNo].krnl[2] = clCreateKernel(program[jtrUniqDevNo], "pbkdf2_postprocess", &err);
if (err) {
fprintf(stderr, "Create Kernel pbkdf2_postprocess FAILED\n");
return 0;
}
errMsg = "Create Buffer FAILED";
globalObj[jtrUniqDevNo].gpu_buffer.pass_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_READ_ONLY, 4 * MAX_KEYS_PER_CRYPT * sizeof(cl_uint), NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.pass_gpu == (cl_mem)0)
HANDLE_CLERROR(err,errMsg );
globalObj[jtrUniqDevNo].gpu_buffer.salt_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_READ_ONLY, (MAX_SALT_LENGTH / 2 + 1) * sizeof(cl_uint), NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.salt_gpu == (cl_mem)0)
HANDLE_CLERROR(err, errMsg);
globalObj[jtrUniqDevNo].gpu_buffer.hash_out_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_WRITE_ONLY, 4 * MAX_KEYS_PER_CRYPT * sizeof(cl_uint), NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.hash_out_gpu == (cl_mem)0)
HANDLE_CLERROR(err, errMsg);
globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_READ_WRITE, MAX_KEYS_PER_CRYPT * sizeof(temp_buf), NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu == (cl_mem)0)
HANDLE_CLERROR(err, errMsg);
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[0], 0, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.pass_gpu), "Set Kernel 0 Arg 0 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[0], 1, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.salt_gpu), "Set Kernel 0 Arg 1 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[0], 4, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu), "Set Kernel 0 Arg 4 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[1], 0, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu), "Set Kernel 1 Arg 0 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[2], 0, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu), "Set Kernel 2 Arg 0 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[2], 1, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.hash_out_gpu), "Set Kernel 2 Arg 1 :FAILED");
if (((!global_work_size) || ((!local_work_size) && global_work_size)) || (active_dev_ctr != 0))
find_best_workgroup(jtrUniqDevNo);
else {
size_t maxsize, maxsize2;
maxsize = get_kernel_preferred_work_group_size(jtrUniqDevNo, globalObj[jtrUniqDevNo].krnl[0]);
maxsize2 = get_kernel_preferred_work_group_size(jtrUniqDevNo, globalObj[jtrUniqDevNo].krnl[1]);
if (maxsize2 > maxsize)
maxsize = maxsize2;
maxsize2 = get_kernel_preferred_work_group_size(jtrUniqDevNo, globalObj[jtrUniqDevNo].krnl[2]);
if (maxsize2 > maxsize)
maxsize = maxsize2;
while (local_work_size > maxsize)
local_work_size /= 2;
if (options.verbosity > 3)
fprintf(stderr, "Local worksize (LWS) forced to %zu\n", local_work_size);
globalObj[jtrUniqDevNo].lws = local_work_size;
}
if ((!global_work_size) || (active_dev_ctr != 0))
find_best_gws(jtrUniqDevNo, fmt);
else {
if (options.verbosity > 3)
fprintf(stderr, "Global worksize (GWS) forced to %zu\n", global_work_size);
fmt -> params.max_keys_per_crypt = global_work_size;
fmt -> params.min_keys_per_crypt = max_lws();
}
active_dev_ctr++;
return globalObj[jtrUniqDevNo].lws;
}
/* ------- Initialization ------- */
static void init(struct fmt_main * self) {
char * tmp_value;
char * task = "$JOHN/cryptsha512_kernel_DEFAULT.cl";
uint64_t startTime, runtime;
opencl_init_dev(ocl_gpu_id, platform_id);
startTime = (unsigned long) time(NULL);
source_in_use = device_info[ocl_gpu_id];
if ((tmp_value = getenv("_TYPE")))
source_in_use = atoi(tmp_value);
if ((tmp_value = getenv("_FAST")))
fast_mode = TRUE;
if (use_local(source_in_use))
task = "$JOHN/cryptsha512_kernel_LOCAL.cl";
else if (gpu(source_in_use)) {
fprintf(stderr, "Building the kernel, this could take a while\n");
task = "$JOHN/cryptsha512_kernel_GPU.cl";
}
fflush(stdout);
opencl_build_kernel(task, ocl_gpu_id);
if ((runtime = (unsigned long) (time(NULL) - startTime)) > 2UL)
fprintf(stderr, "Elapsed time: %lu seconds\n", runtime);
fflush(stdout);
// create kernel(s) to execute
crypt_kernel = clCreateKernel(program[ocl_gpu_id], "kernel_crypt", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel. Double-check kernel name?");
if (gpu(source_in_use) || use_local(source_in_use)) {
prepare_kernel = clCreateKernel(program[ocl_gpu_id], "kernel_prepare", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel_prepare. Double-check kernel name?");
final_kernel = clCreateKernel(program[ocl_gpu_id], "kernel_final", &ret_code);
HANDLE_CLERROR(ret_code, "Error creating kernel_final. Double-check kernel name?");
}
global_work_size = get_task_max_size();
local_work_size = get_default_workgroup();
if (source_in_use != device_info[ocl_gpu_id])
fprintf(stderr, "Selected runtime id %d, source (%s)\n", source_in_use, task);
if ((tmp_value = cfg_get_param(SECTION_OPTIONS,
SUBSECTION_OPENCL, LWS_CONFIG)))
local_work_size = atoi(tmp_value);
if ((tmp_value = getenv("LWS")))
local_work_size = atoi(tmp_value);
//Check if local_work_size is a valid number.
if (local_work_size > get_task_max_work_group_size()){
local_work_size = 0; //Force find a valid number.
}
self->params.max_keys_per_crypt = global_work_size;
if (!local_work_size) {
local_work_size = get_task_max_work_group_size();
create_clobj(global_work_size, self);
find_best_workgroup(self);
release_clobj();
}
if ((tmp_value = cfg_get_param(SECTION_OPTIONS,
SUBSECTION_OPENCL, GWS_CONFIG)))
global_work_size = atoi(tmp_value);
if ((tmp_value = getenv("GWS")))
global_work_size = atoi(tmp_value);
//Check if a valid multiple is used.
global_work_size = get_multiple(global_work_size, local_work_size);
if (global_work_size)
create_clobj(global_work_size, self);
else {
//user chose to die of boredom
global_work_size = get_task_max_size();
find_best_gws(self);
}
fprintf(stderr, "Local work size (LWS) %d, global work size (GWS) %zd\n",
(int) local_work_size, global_work_size);
self->params.max_keys_per_crypt = global_work_size;
}
size_t select_device(int jtrUniqDevNo, struct fmt_main *fmt) {
cl_int err;
const char *errMsg;
size_t memAllocSz;
active_dev_ctr++;
opencl_init("$JOHN/kernels/pbkdf2_kernel.cl", jtrUniqDevNo, NULL);
globalObj[jtrUniqDevNo].krnl[0] = clCreateKernel(program[jtrUniqDevNo], "pbkdf2_preprocess_short", &err);
if (err) {
fprintf(stderr, "Create Kernel pbkdf2_preprocess_short FAILED\n");
return 0;
}
globalObj[jtrUniqDevNo].krnl[1] = clCreateKernel(program[jtrUniqDevNo], "pbkdf2_preprocess_long", &err);
if (err) {
fprintf(stderr, "Create Kernel pbkdf2_preprocess_long FAILED\n");
return 0;
}
globalObj[jtrUniqDevNo].krnl[2] = clCreateKernel(program[jtrUniqDevNo], "pbkdf2_iter", &err);
if (err) {
fprintf(stderr, "Create Kernel pbkdf2_iter FAILED\n");
return 0;
}
globalObj[jtrUniqDevNo].krnl[3] = clCreateKernel(program[jtrUniqDevNo], "pbkdf2_postprocess", &err);
if (err) {
fprintf(stderr, "Create Kernel pbkdf2_postprocess FAILED\n");
return 0;
}
errMsg = "Create Buffer FAILED";
memAllocSz = 4 * MAX_KEYS_PER_CRYPT * sizeof(cl_uint);
memAllocSz = memAllocSz < get_max_mem_alloc_size(jtrUniqDevNo) ? memAllocSz : get_max_mem_alloc_size(jtrUniqDevNo) / 4 * 4;
globalObj[jtrUniqDevNo].gpu_buffer.pass_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_READ_ONLY, memAllocSz, NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.pass_gpu == (cl_mem)0)
HANDLE_CLERROR(err,errMsg );
globalObj[jtrUniqDevNo].gpu_buffer.salt_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_READ_ONLY, (MAX_SALT_LENGTH / 2 + 1) * sizeof(cl_uint), NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.salt_gpu == (cl_mem)0)
HANDLE_CLERROR(err, errMsg);
globalObj[jtrUniqDevNo].gpu_buffer.hash_out_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_WRITE_ONLY, memAllocSz, NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.hash_out_gpu == (cl_mem)0)
HANDLE_CLERROR(err, errMsg);
memAllocSz = MAX_KEYS_PER_CRYPT * sizeof(temp_buf);
memAllocSz = memAllocSz < get_max_mem_alloc_size(jtrUniqDevNo) ? memAllocSz : get_max_mem_alloc_size(jtrUniqDevNo) / 4 * 4;
globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_READ_WRITE, memAllocSz, NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu == (cl_mem)0)
HANDLE_CLERROR(err, errMsg);
memAllocSz = 5 * MAX_KEYS_PER_CRYPT * sizeof(cl_uint);
memAllocSz = memAllocSz < get_max_mem_alloc_size(jtrUniqDevNo) ? memAllocSz : get_max_mem_alloc_size(jtrUniqDevNo) / 4 * 4;
globalObj[jtrUniqDevNo].gpu_buffer.hmac_sha1_gpu = clCreateBuffer(context[jtrUniqDevNo], CL_MEM_READ_WRITE, memAllocSz, NULL, &err);
if (globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu == (cl_mem)0)
HANDLE_CLERROR(err, errMsg);
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[0], 0, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.pass_gpu), "Set Kernel 0 Arg 0 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[0], 1, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.salt_gpu), "Set Kernel 0 Arg 1 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[0], 3, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu), "Set Kernel 0 Arg 3 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[1], 0, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.pass_gpu), "Set Kernel 1 Arg 0 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[1], 1, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu), "Set Kernel 1 Arg 1 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[1], 2, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.hmac_sha1_gpu), "Set Kernel 1 Arg 2 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[2], 0, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu), "Set Kernel 2 Arg 0 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[3], 0, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.temp_buf_gpu), "Set Kernel 3 Arg 0 :FAILED");
HANDLE_CLERROR(clSetKernelArg(globalObj[jtrUniqDevNo].krnl[3], 1, sizeof(cl_mem), &globalObj[jtrUniqDevNo].gpu_buffer.hash_out_gpu), "Set Kernel 3 Arg 1 :FAILED");
if (!local_work_size)
find_best_workgroup(jtrUniqDevNo, quick_bechmark(jtrUniqDevNo));
else {
size_t maxsize, maxsize2;
globalObj[jtrUniqDevNo].lws = local_work_size;
// Obey limits
HANDLE_CLERROR(clGetKernelWorkGroupInfo(globalObj[jtrUniqDevNo].krnl[0], devices[jtrUniqDevNo], CL_KERNEL_WORK_GROUP_SIZE, sizeof(maxsize), &maxsize, NULL), "Error querying max LWS");
HANDLE_CLERROR(clGetKernelWorkGroupInfo(globalObj[jtrUniqDevNo].krnl[1], devices[jtrUniqDevNo], CL_KERNEL_WORK_GROUP_SIZE, sizeof(maxsize2), &maxsize2, NULL), "Error querying max LWS");
if (maxsize2 > maxsize)
maxsize = maxsize2;
HANDLE_CLERROR(clGetKernelWorkGroupInfo(globalObj[jtrUniqDevNo].krnl[2], devices[jtrUniqDevNo], CL_KERNEL_WORK_GROUP_SIZE, sizeof(maxsize2), &maxsize2, NULL), "Error querying max LWS");
if (maxsize2 > maxsize)
maxsize = maxsize2;
HANDLE_CLERROR(clGetKernelWorkGroupInfo(globalObj[jtrUniqDevNo].krnl[3], devices[jtrUniqDevNo], CL_KERNEL_WORK_GROUP_SIZE, sizeof(maxsize2), &maxsize2, NULL), "Error querying max LWS");
if (maxsize2 > maxsize)
maxsize = maxsize2;
while (globalObj[jtrUniqDevNo].lws > maxsize)
globalObj[jtrUniqDevNo].lws /= 2;
if (options.verbosity > 3)
fprintf(stderr, "Local worksize (LWS) forced to "Zu"\n", globalObj[jtrUniqDevNo].lws);
globalObj[jtrUniqDevNo].exec_time_inv = 1;
}
if (!global_work_size)
find_best_gws(jtrUniqDevNo, fmt);
else {
if (options.verbosity > 3)
fprintf(stderr, "Global worksize (GWS) forced to "Zu"\n", global_work_size);
fmt -> params.max_keys_per_crypt = global_work_size;
fmt -> params.min_keys_per_crypt = max_lws();
}
return globalObj[jtrUniqDevNo].lws;
}
