Refactor intel gpu discovery

This commit is contained in:
Daniel Hiltgen
2024-05-29 16:37:34 -07:00
parent 48702dd149
commit 4e2b7e181d
4 changed files with 304 additions and 169 deletions

View File

@@ -8,9 +8,13 @@ void oneapi_init(char *oneapi_lib_path, oneapi_init_resp_t *resp)
{
ze_result_t ret;
resp->err = NULL;
resp->oh.devices = NULL;
resp->oh.num_devices = NULL;
resp->oh.drivers = NULL;
resp->oh.num_drivers = 0;
const int buflen = 256;
char buf[buflen + 1];
int i;
int i, d, count;
struct lookup
{
char *s;
@@ -66,19 +70,65 @@ void oneapi_init(char *oneapi_lib_path, oneapi_init_resp_t *resp)
ret = (*resp->oh.zesInit)(0);
if (ret != ZE_RESULT_SUCCESS)
{
LOG(resp->oh.verbose, "zesInit err: %d\n", ret);
UNLOAD_LIBRARY(resp->oh.handle);
resp->oh.handle = NULL;
snprintf(buf, buflen, "oneapi vram init failure: %d", ret);
LOG(resp->oh.verbose, "zesInit err: %x\n", ret);
snprintf(buf, buflen, "oneapi vram init failure: %x", ret);
resp->err = strdup(buf);
oneapi_release(resp->oh);
return;
}
(*resp->oh.zesDriverGet)(&resp->num_devices, NULL);
count = 0;
ret = (*resp->oh.zesDriverGet)(&resp->oh.num_drivers, NULL);
if (ret != ZE_RESULT_SUCCESS)
{
LOG(resp->oh.verbose, "zesDriverGet err: %x\n", ret);
snprintf(buf, buflen, "unable to get driver count: %x", ret);
resp->err = strdup(buf);
oneapi_release(resp->oh);
return;
}
LOG(resp->oh.verbose, "oneapi driver count: %d\n", resp->oh.num_drivers);
resp->oh.drivers = malloc(resp->oh.num_drivers * sizeof(zes_driver_handle_t));
resp->oh.num_devices = malloc(resp->oh.num_drivers * sizeof(uint32_t));
memset(&resp->oh.num_devices[0], 0, resp->oh.num_drivers * sizeof(uint32_t));
resp->oh.devices = malloc(resp->oh.num_drivers * sizeof(zes_device_handle_t*));
ret = (*resp->oh.zesDriverGet)(&resp->oh.num_drivers, &resp->oh.drivers[0]);
if (ret != ZE_RESULT_SUCCESS)
{
LOG(resp->oh.verbose, "zesDriverGet err: %x\n", ret);
snprintf(buf, buflen, "unable to get driver count: %x", ret);
resp->err = strdup(buf);
oneapi_release(resp->oh);
return;
}
for (d = 0; d < resp->oh.num_drivers; d++) {
ret = (*resp->oh.zesDeviceGet)(resp->oh.drivers[d], &resp->oh.num_devices[d], NULL);
if (ret != ZE_RESULT_SUCCESS)
{
LOG(resp->oh.verbose, "zesDeviceGet err: %x\n", ret);
snprintf(buf, buflen, "unable to get device count: %x", ret);
resp->err = strdup(buf);
oneapi_release(resp->oh);
return;
}
resp->oh.devices[d] = malloc(resp->oh.num_devices[d] * sizeof(zes_device_handle_t));
ret = (*resp->oh.zesDeviceGet)(resp->oh.drivers[d], &resp->oh.num_devices[d], resp->oh.devices[d]);
if (ret != ZE_RESULT_SUCCESS)
{
LOG(resp->oh.verbose, "zesDeviceGet err: %x\n", ret);
snprintf(buf, buflen, "unable to get device count: %x", ret);
resp->err = strdup(buf);
oneapi_release(resp->oh);
return;
}
count += resp->oh.num_devices[d];
}
return;
}
void oneapi_check_vram(oneapi_handle_t h, mem_info_t *resp)
void oneapi_check_vram(oneapi_handle_t h, int driver, int device, mem_info_t *resp)
{
ze_result_t ret;
resp->err = NULL;
@@ -93,122 +143,135 @@ void oneapi_check_vram(oneapi_handle_t h, mem_info_t *resp)
resp->err = strdup("Level-Zero handle not initialized");
return;
}
uint32_t driversCount = 0;
ret = (*h.zesDriverGet)(&driversCount, NULL);
if (ret != ZE_RESULT_SUCCESS)
{
snprintf(buf, buflen, "unable to get driver count: %d", ret);
resp->err = strdup(buf);
if (driver > h.num_drivers || device > h.num_devices[driver]) {
resp->err = strdup("driver of device index out of bounds");
return;
}
LOG(h.verbose, "discovered %d Level-Zero drivers\n", driversCount);
zes_driver_handle_t *allDrivers =
malloc(driversCount * sizeof(zes_driver_handle_t));
(*h.zesDriverGet)(&driversCount, allDrivers);
resp->total = 0;
resp->free = 0;
for (d = 0; d < driversCount; d++)
zes_device_ext_properties_t ext_props;
ext_props.stype = ZES_STRUCTURE_TYPE_DEVICE_EXT_PROPERTIES;
ext_props.pNext = NULL;
zes_device_properties_t props;
props.stype = ZES_STRUCTURE_TYPE_DEVICE_PROPERTIES;
props.pNext = &ext_props;
ret = (*h.zesDeviceGetProperties)(h.devices[driver][device], &props);
if (ret != ZE_RESULT_SUCCESS)
{
uint32_t deviceCount = 0;
ret = (*h.zesDeviceGet)(allDrivers[d], &deviceCount, NULL);
snprintf(buf, buflen, "unable to get device properties: %d", ret);
resp->err = strdup(buf);
return;
}
snprintf(&resp->gpu_name[0], GPU_NAME_LEN, props.modelName);
// TODO this needs to map to ONEAPI_DEVICE_SELECTOR syntax
// (this is probably wrong...)
// TODO - the driver isn't included - what if there are multiple drivers?
snprintf(&resp->gpu_id[0], GPU_ID_LEN, "%d", device);
if (h.verbose)
{
// When in verbose mode, report more information about
// the card we discover.
LOG(h.verbose, "[%d:%d] oneAPI device name: %s\n", driver, device,
props.modelName);
LOG(h.verbose, "[%d:%d] oneAPI brand: %s\n", driver, device,
props.brandName);
LOG(h.verbose, "[%d:%d] oneAPI vendor: %s\n", driver, device,
props.vendorName);
LOG(h.verbose, "[%d:%d] oneAPI S/N: %s\n", driver, device,
props.serialNumber);
LOG(h.verbose, "[%d:%d] oneAPI board number: %s\n", driver, device,
props.boardNumber);
}
// TODO
// Compute Capability equivalent in resp->major, resp->minor, resp->patch
uint32_t memCount = 0;
ret = (*h.zesDeviceEnumMemoryModules)(h.devices[driver][device], &memCount, NULL);
if (ret != ZE_RESULT_SUCCESS)
{
snprintf(buf, buflen,
"unable to enumerate Level-Zero memory modules: %x", ret);
resp->err = strdup(buf);
return;
}
LOG(h.verbose, "discovered %d Level-Zero memory modules\n", memCount);
zes_mem_handle_t *mems = malloc(memCount * sizeof(zes_mem_handle_t));
(*h.zesDeviceEnumMemoryModules)(h.devices[driver][device], &memCount, mems);
for (m = 0; m < memCount; m++)
{
zes_mem_state_t state;
state.stype = ZES_STRUCTURE_TYPE_MEM_STATE;
state.pNext = NULL;
ret = (*h.zesMemoryGetState)(mems[m], &state);
if (ret != ZE_RESULT_SUCCESS)
{
snprintf(buf, buflen, "unable to get device count: %d", ret);
snprintf(buf, buflen, "unable to get memory state: %x", ret);
resp->err = strdup(buf);
free(allDrivers);
free(mems);
return;
}
LOG(h.verbose, "discovered %d Level-Zero devices\n", deviceCount);
zes_device_handle_t *devices =
malloc(deviceCount * sizeof(zes_device_handle_t));
(*h.zesDeviceGet)(allDrivers[d], &deviceCount, devices);
for (i = 0; i < deviceCount; i++)
{
zes_device_ext_properties_t ext_props;
ext_props.stype = ZES_STRUCTURE_TYPE_DEVICE_EXT_PROPERTIES;
ext_props.pNext = NULL;
zes_device_properties_t props;
props.stype = ZES_STRUCTURE_TYPE_DEVICE_PROPERTIES;
props.pNext = &ext_props;
ret = (*h.zesDeviceGetProperties)(devices[i], &props);
if (ret != ZE_RESULT_SUCCESS)
{
snprintf(buf, buflen, "unable to get device properties: %d", ret);
resp->err = strdup(buf);
free(allDrivers);
free(devices);
return;
}
if (h.verbose)
{
// When in verbose mode, report more information about
// the card we discover.
LOG(h.verbose, "[%d] oneAPI device name: %s\n", i,
props.modelName);
LOG(h.verbose, "[%d] oneAPI brand: %s\n", i,
props.brandName);
LOG(h.verbose, "[%d] oneAPI vendor: %s\n", i,
props.vendorName);
LOG(h.verbose, "[%d] oneAPI S/N: %s\n", i,
props.serialNumber);
LOG(h.verbose, "[%d] oneAPI board number: %s\n", i,
props.boardNumber);
}
uint32_t memCount = 0;
ret = (*h.zesDeviceEnumMemoryModules)(devices[i], &memCount, NULL);
if (ret != ZE_RESULT_SUCCESS)
{
snprintf(buf, buflen,
"unable to enumerate Level-Zero memory modules: %d", ret);
resp->err = strdup(buf);
free(allDrivers);
free(devices);
return;
}
LOG(h.verbose, "discovered %d Level-Zero memory modules\n", memCount);
zes_mem_handle_t *mems = malloc(memCount * sizeof(zes_mem_handle_t));
(*h.zesDeviceEnumMemoryModules)(devices[i], &memCount, mems);
for (m = 0; m < memCount; m++)
{
zes_mem_state_t state;
state.stype = ZES_STRUCTURE_TYPE_MEM_STATE;
state.pNext = NULL;
ret = (*h.zesMemoryGetState)(mems[m], &state);
if (ret != ZE_RESULT_SUCCESS)
{
snprintf(buf, buflen, "unable to get memory state: %d", ret);
resp->err = strdup(buf);
free(allDrivers);
free(devices);
free(mems);
return;
}
resp->total += state.size;
resp->free += state.free;
}
free(mems);
}
free(devices);
resp->total += state.size;
resp->free += state.free;
}
free(allDrivers);
free(mems);
}
void oneapi_release(oneapi_handle_t h)
{
int d;
LOG(h.verbose, "releasing oneapi library\n");
for (d = 0; d < h.num_drivers; d++)
{
if (h.devices != NULL && h.devices[d] != NULL)
{
free(h.devices[d]);
}
}
if (h.devices != NULL)
{
free(h.devices);
h.devices = NULL;
}
if (h.num_devices != NULL)
{
free(h.num_devices);
h.num_devices = NULL;
}
if (h.drivers != NULL)
{
free(h.drivers);
h.drivers = NULL;
}
h.num_drivers = 0;
UNLOAD_LIBRARY(h.handle);
h.handle = NULL;
}
int oneapi_get_device_count(oneapi_handle_t h, int driver)
{
if (h.handle == NULL || h.num_devices == NULL)
{
return 0;
}
if (driver > h.num_drivers)
{
return 0;
}
return (int)h.num_devices[driver];
}
#endif // __APPLE__