Bugfix: wrong order on gpu
This commit is contained in:
@@ -33,6 +33,11 @@ public interface CUDA extends Library {
|
||||
|
||||
public int cuDeviceGetName(byte[] name, int len, int dev);
|
||||
|
||||
public int cuDeviceGet (IntByReference device, int ordinal);
|
||||
|
||||
public int cuDeviceGetAttribute (IntByReference pi, int attrib, int dev );
|
||||
|
||||
public int cuDeviceTotalMem_v2(LongByReference bytes, int dev);
|
||||
public int cuDeviceTotalMem(LongByReference bytes, int dev);
|
||||
|
||||
}
|
||||
|
||||
41
src/com/sheepit/client/hardware/gpu/CUDeviceAttribute.java
Normal file
41
src/com/sheepit/client/hardware/gpu/CUDeviceAttribute.java
Normal file
@@ -0,0 +1,41 @@
|
||||
/*
|
||||
* Copyright (C) 2018 Laurent CLOUET
|
||||
* Author Laurent CLOUET <laurent.clouet@nopnop.net>
|
||||
*
|
||||
* This program is free software; you can redistribute it and/or
|
||||
* modify it under the terms of the GNU General Public License
|
||||
* as published by the Free Software Foundation; version 2
|
||||
* of the License.
|
||||
*
|
||||
* This program is distributed in the hope that it will be useful,
|
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
* GNU General Public License for more details.
|
||||
*
|
||||
* You should have received a copy of the GNU General Public License
|
||||
* along with this program; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
|
||||
*/
|
||||
|
||||
package com.sheepit.client.hardware.gpu;
|
||||
|
||||
/**
|
||||
* CUDA Device properties. Taken directly from the online manual:
|
||||
* https://docs.nvidia.com/cuda/cuda-driver-api
|
||||
*/
|
||||
public class CUDeviceAttribute {
|
||||
/**
|
||||
* PCI bus ID of the device
|
||||
*/
|
||||
public static final int CU_DEVICE_ATTRIBUTE_PCI_BUS_ID = 33;
|
||||
|
||||
/**
|
||||
* PCI device ID of the device
|
||||
*/
|
||||
public static final int CU_DEVICE_ATTRIBUTE_PCI_DEVICE_ID = 34;
|
||||
|
||||
/**
|
||||
* PCI domain ID of the device
|
||||
*/
|
||||
public static final int CU_DEVICE_ATTRIBUTE_PCI_DOMAIN_ID = 50;
|
||||
}
|
||||
@@ -19,6 +19,7 @@
|
||||
|
||||
package com.sheepit.client.hardware.gpu;
|
||||
|
||||
import java.util.HashMap;
|
||||
import java.util.LinkedList;
|
||||
import java.util.List;
|
||||
|
||||
@@ -81,7 +82,23 @@ public class GPU {
|
||||
return false;
|
||||
}
|
||||
|
||||
HashMap<Integer, GPUDevice> devicesWithPciId = new HashMap<Integer, GPUDevice>(count.getValue());
|
||||
for (int num = 0; num < count.getValue(); num++) {
|
||||
IntByReference aDevice = new IntByReference();
|
||||
|
||||
result = cudalib.cuDeviceGet(aDevice, num);
|
||||
if (result != CUresult.CUDA_SUCCESS) {
|
||||
System.out.println("GPU::generate cuDeviceGet failed (ret: " + CUresult.stringFor(result) + ")");
|
||||
continue;
|
||||
}
|
||||
|
||||
IntByReference pciBusId = new IntByReference();
|
||||
result = cudalib.cuDeviceGetAttribute(pciBusId, CUDeviceAttribute.CU_DEVICE_ATTRIBUTE_PCI_BUS_ID, aDevice.getValue());
|
||||
if (result != CUresult.CUDA_SUCCESS) {
|
||||
System.out.println("GPU::generate cuDeviceGetAttribute for CU_DEVICE_ATTRIBUTE_PCI_BUS_ID failed (ret: " + CUresult.stringFor(result) + ")");
|
||||
continue;
|
||||
}
|
||||
|
||||
byte name[] = new byte[256];
|
||||
|
||||
result = cudalib.cuDeviceGetName(name, 256, num);
|
||||
@@ -104,8 +121,19 @@ public class GPU {
|
||||
return false;
|
||||
}
|
||||
|
||||
devices.add(new GPUDevice(new String(name).trim(), ram.getValue(), "CUDA_" + Integer.toString(num)));
|
||||
devicesWithPciId.put(pciBusId.getValue(), new GPUDevice(new String(name).trim(), ram.getValue(), "FAKE"));
|
||||
}
|
||||
|
||||
// generate proper cuda id
|
||||
// in theory a set to environment "CUDA_DEVICE_ORDER=PCI_BUS_ID" should be enough but it didn't work
|
||||
int i = 0;
|
||||
for (HashMap.Entry<Integer, GPUDevice> entry : devicesWithPciId.entrySet()){
|
||||
GPUDevice aDevice = entry.getValue();
|
||||
aDevice.setCudaName("CUDA_" + Integer.toString(i));
|
||||
devices.add(aDevice);
|
||||
i++;
|
||||
}
|
||||
|
||||
return true;
|
||||
}
|
||||
|
||||
|
||||
Reference in New Issue
Block a user