AI/rocm-systems
2
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix breaking changes introduced with CPU support

Browse Source 
Changes introduced in 3bdca8b8b6
broke RDC if it was compiled without ESMI support, or if esmi driver is
not loaded when RDC is being used.

Change-Id: Id54e1e9002d2e3cf09240081149eed84178700af
Signed-off-by: Galantsev, Dmitrii <dmitrii.galantsev@amd.com>
This commit is contained in:
Galantsev, Dmitrii authored and Galantsev, Dmitrii committed 2025-04-07 14:41:46 -05:00
1 parent 3bdca8b8b6
commit 0aeceefcb3
2 files changed
+22 -10
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

rdc_libs/rdc/src/RdcEmbeddedHandler.cc
+19 -10
View File
@@ -50,11 +50,18 @@ class smi_initializer {
// Make sure smi will not be initialized multiple times
amdsmi_shut_down();
amdsmi_status_t ret;
uint64_t init_flag_;
//initialize CPU and GPU instances
init_flag_ = AMDSMI_INIT_AMD_GPUS | AMDSMI_INIT_AMD_CPUS;
uint64_t init_flag_ = AMDSMI_INIT_AMD_GPUS;
#ifdef ENABLE_ESMI_LIB
init_flag_ = init_flag_ | AMDSMI_INIT_AMD_CPUS;
#endif
ret = amdsmi_init(init_flag_);
if (init_flag_ & AMDSMI_INIT_AMD_CPUS) {
RDC_LOG(RDC_ERROR,
"Failed to initalize amdsmi with CPUs and GPUs enabled.. trying GPUs only.");
ret = amdsmi_init(AMDSMI_INIT_AMD_GPUS);
}
if (ret != AMDSMI_STATUS_SUCCESS) {
RDC_LOG(RDC_ERROR, "SMI FAILED with" << ret);
throw amd::rdc::RdcException(RDC_ST_FAIL_LOAD_MODULE, "SMI initialize fail");
}
}
@@ -198,14 +205,15 @@ rdc_status_t RdcEmbeddedHandler::rdc_device_get_all(uint32_t gpu_index_list[RDC_
}
// Discovery API
rdc_status_t RdcEmbeddedHandler::rdc_device_get_all_cpu(uint32_t cpu_index_list[RDC_MAX_NUM_DEVICES],
uint32_t* count) {
rdc_status_t RdcEmbeddedHandler::rdc_device_get_all_cpu(
uint32_t cpu_index_list[RDC_MAX_NUM_DEVICES], uint32_t* count) {
if (!count) {
return RDC_ST_BAD_PARAMETER;
}
rdc_field_value device_count;
rdc_status_t status = metric_fetcher_->fetch_smi_cpu_field(0, RDC_FI_DEV_CPU_COUNT, &device_count);
rdc_status_t status =
metric_fetcher_->fetch_smi_cpu_field(0, RDC_FI_DEV_CPU_COUNT, &device_count);
if (status != RDC_ST_OK) {
std::cout << "rdc_device_get_all_cpu failed to get cpu count";
return status;
@@ -216,9 +224,9 @@ rdc_status_t RdcEmbeddedHandler::rdc_device_get_all_cpu(uint32_t cpu_index_list[
for (uint32_t i = 0; i < *count; i++) {
cpu_index_list[i] = i;
}
return RDC_ST_OK;
}
rdc_status_t RdcEmbeddedHandler::rdc_device_get_attributes(uint32_t gpu_index,
rdc_device_attributes_t* p_rdc_attr) {
if (!p_rdc_attr) {
@@ -230,14 +238,15 @@ rdc_status_t RdcEmbeddedHandler::rdc_device_get_attributes(uint32_t gpu_index,
return status;
}
rdc_status_t RdcEmbeddedHandler::rdc_device_get_cpu_attributes(uint32_t cpu_index,
rdc_device_attributes_t* p_rdc_attr) {
rdc_status_t RdcEmbeddedHandler::rdc_device_get_cpu_attributes(
uint32_t cpu_index, rdc_device_attributes_t* p_rdc_attr) {
if (!p_rdc_attr) {
return RDC_ST_BAD_PARAMETER;
}
rdc_field_value device_name;
rdc_status_t status = metric_fetcher_->fetch_smi_cpu_field(cpu_index, RDC_FI_DEV_CPU_MODEL, &device_name);
rdc_status_t status =
metric_fetcher_->fetch_smi_cpu_field(cpu_index, RDC_FI_DEV_CPU_MODEL, &device_name);
strncpy_with_null(p_rdc_attr->device_name, device_name.value.str, RDC_MAX_STR_LENGTH);
return status;