Corrected clean local data partition indexing

Signed-off-by: Maisam Arif <Maisam.Arif@amd.com> Change-Id: Ib0eeb065f160fccd3c3f4a2d13f0869af01a74ae
ROCm · Oct 9, 2024 · 9578815 · 9578815
1 parent b2088a6
commit 9578815
Show file tree

Hide file tree

Showing 4 changed files with 9 additions and 11 deletions.
diff --git a/amdsmi_cli/amdsmi_parser.py b/amdsmi_cli/amdsmi_parser.py
@@ -1014,7 +1014,7 @@ def _add_set_value_parser(self, subparsers, func):
         set_power_cap_help = "Set power capacity limit"
         set_soc_pstate_help = "Set the GPU soc pstate policy using policy id\n"
         set_xgmi_plpd_help = "Set the GPU XGMI per-link power down policy using policy id\n"
-        set_process_isolation_help = "Enable or disable the GPU process isolation: 0 for disable and 1 for enable.\n"
+        set_process_isolation_help = "Enable or disable the GPU process isolation on a per partition basis: 0 for disable and 1 for enable.\n"
 
         # Help text for CPU set options
         set_cpu_pwr_limit_help = "Set power limit for the given socket. Input parameter is power limit value."
@@ -1104,7 +1104,7 @@ def _add_reset_parser(self, subparsers, func):
         reset_compute_help = "Reset compute partitions on the specified GPU"
         reset_memory_help = "Reset memory partitions on the specified GPU"
         reset_power_cap_help = "Reset power capacity limit to max capable"
-        reset_gpu_clean_local_data_help = "Clean up local data in LDS/GPRs"
+        reset_gpu_clean_local_data_help = "Clean up local data in LDS/GPRs on a per partition basis"
 
         # Create reset subparser
         reset_parser = subparsers.add_parser('reset', help=reset_help, description=reset_subcommand_help)

diff --git a/rocm_smi/include/rocm_smi/rocm_smi.h b/rocm_smi/include/rocm_smi/rocm_smi.h
@@ -3461,12 +3461,9 @@ rsmi_status_t rsmi_dev_process_isolation_set(uint32_t dv_ind,
  *
  *  @param[in] dv_ind a device index
  *
- *  @param[in] sclean the clean flag. Only 1 will take effect and other number
- *  are reserved for future usage.
- *
  *  @return ::RSMI_STATUS_SUCCESS is returned upon successful call, non-zero on fail
  */
-rsmi_status_t rsmi_dev_gpu_run_cleaner_shader(uint32_t dv_ind, uint32_t sclean);
+rsmi_status_t rsmi_dev_gpu_run_cleaner_shader(uint32_t dv_ind);
 
 /** @} */  // end of PerfCont
 

diff --git a/rocm_smi/src/rocm_smi.cc b/rocm_smi/src/rocm_smi.cc
@@ -2122,8 +2122,7 @@ rsmi_status_t rsmi_dev_process_isolation_set(uint32_t dv_ind,
   CATCH
 }
 
-rsmi_status_t rsmi_dev_gpu_run_cleaner_shader(uint32_t dv_ind,
-    uint32_t sclean) {
+rsmi_status_t rsmi_dev_gpu_run_cleaner_shader(uint32_t dv_ind) {
   rsmi_status_t ret;
 
   TRY
@@ -2134,7 +2133,10 @@ rsmi_status_t rsmi_dev_gpu_run_cleaner_shader(uint32_t dv_ind,
   DEVICE_MUTEX
   GET_DEV_FROM_INDX
 
-  std::string value = std::to_string(sclean);
+  // To reset you need to provide the partition id
+  // echo "0" | sudo tee  /sys/class/drm/cardX/device/run_cleaner_shader
+  int partition_id = dev->get_partition_id();
+  std::string value = std::to_string(partition_id);
   int ret = dev->writeDevInfo(amd::smi::kDevShaderClean , value);
   return amd::smi::ErrnoToRsmiStatus(ret);
 

diff --git a/src/amd_smi/amd_smi.cc b/src/amd_smi/amd_smi.cc
@@ -1462,8 +1462,7 @@ amdsmi_status_t amdsmi_set_gpu_process_isolation(amdsmi_processor_handle process
 amdsmi_status_t amdsmi_clean_gpu_local_data(amdsmi_processor_handle processor_handle) {
     AMDSMI_CHECK_INIT();
 
-    return rsmi_wrapper(rsmi_dev_gpu_run_cleaner_shader, processor_handle,
-                    1);
+    return rsmi_wrapper(rsmi_dev_gpu_run_cleaner_shader, processor_handle);
 }
 
 amdsmi_status_t