fixes for titan platform

Dieser Commit ist enthalten in:
Ben Sander
2016-02-26 05:25:30 -06:00
Ursprung 6e0ccdfb95
Commit ff66ef0779
5 geänderte Dateien mit 4 neuen und 12 gelöschten Zeilen
+1 -1
Datei anzeigen
@@ -43,7 +43,7 @@ THE SOFTWARE.
#include <stdlib.h>
#include <assert.h>
#ifdef __cplusplus
#if __cplusplus > 199711L
#include <thread>
#endif
+1 -5
Datei anzeigen
@@ -81,10 +81,8 @@ typedef struct hipDeviceProp_t {
int maxThreadsDim[3]; ///< Max number of threads in each dimension (XYZ) of a block.
int maxGridSize[3]; ///< Max grid dimensions (XYZ).
int clockRate; ///< Max clock frequency of the multiProcessors in khz.
#ifdef USE_ROCR_20
int memoryClockRate; ///< Max global memory clock frequency in khz.
int memoryBusWidth; ///< Global memory bus width in bits.
#endif
size_t totalConstMem; ///< Size of shared memory region (in bytes).
int major; ///< Major compute capability. On HCC, this is an approximation and features may differ from CUDA CC. See the arch feature flags for portable ways to query feature caps.
int minor; ///< Minor compute capability. On HCC, this is an approximation and features may differ from CUDA CC. See the arch feature flags for portable ways to query feature caps.
@@ -135,7 +133,7 @@ typedef struct hipPointerAttribute_t {
/*
* @brief hipError_t
asdasd* @brief hipError_t
* @enum
* @ingroup Enumerations
*/
@@ -175,10 +173,8 @@ typedef enum hipDeviceAttribute_t {
hipDeviceAttributeWarpSize, ///< Warp size in threads.
hipDeviceAttributeMaxRegistersPerBlock, ///< Maximum number of 32-bit registers available to a thread block. This number is shared by all thread blocks simultaneously resident on a multiprocessor.
hipDeviceAttributeClockRate, ///< Peak clock frequency in kilohertz.
#ifdef USE_ROCR_20
hipDeviceAttributeMemoryClockRate, ///< Peak memory clock frequency in kilohertz.
hipDeviceAttributeMemoryBusWidth, ///< Global memory bus width in bits.
#endif
hipDeviceAttributeMultiprocessorCount, ///< Number of multiprocessors on the device.
hipDeviceAttributeComputeMode, ///< Compute mode that device is currently in.
hipDeviceAttributeL2CacheSize, ///< Size of L2 cache in bytes. 0 if the device doesn't have L2 cache.
-2
Datei anzeigen
@@ -82,10 +82,8 @@ void printDeviceProp (int deviceId)
cout << setw(w1) << "maxThreadsPerMultiProcessor: " << props.maxThreadsPerMultiProcessor << endl;
cout << setw(w1) << "isMultiGpuBoard: " << props.isMultiGpuBoard << endl;
cout << setw(w1) << "clockRate: " << (float)props.clockRate / 1000.0 << " Mhz" << endl;
#ifdef USE_ROCR_20
cout << setw(w1) << "memoryClockRate: " << (float)props.memoryClockRate / 1000.0 << " Mhz" << endl;
cout << setw(w1) << "memoryBusWidth: " << props.memoryBusWidth << endl;
#endif
cout << setw(w1) << "clockInstructionRate: " << (float)props.clockInstructionRate / 1000.0 << " Mhz" << endl;
cout << setw(w1) << "totalGlobalMem: " << fixed << setprecision(2) << bytesToGB(props.totalGlobalMem) << " GB" << endl;
cout << setw(w1) << "maxSharedMemoryPerMultiProcessor: " << fixed << setprecision(2) << bytesToGB(props.maxSharedMemoryPerMultiProcessor) << " GB" << endl;
+2 -2
Datei anzeigen
@@ -792,7 +792,7 @@ hipError_t ihipDevice_t::getProperties(hipDeviceProp_t* prop)
// Group memory will not be paged out, so, the physical memory size is the total shared memory size, and also equal to the group region size.
prop->maxSharedMemoryPerMultiProcessor = prop->totalGlobalMem;
#ifdef USE_ROCR_20
#ifdef USE_ROCR_V2
// Get Max memory clock frequency
err = hsa_region_get_info(*am_region, (hsa_region_info_t)HSA_AMD_REGION_INFO_MAX_CLOCK_FREQUENCY, &prop->memoryClockRate);
DeviceErrorCheck(err);
@@ -1269,7 +1269,7 @@ hipError_t hipDeviceGetAttribute(int* pi, hipDeviceAttribute_t attr, int device)
*pi = prop->regsPerBlock; break;
case hipDeviceAttributeClockRate:
*pi = prop->clockRate; break;
#ifdef USE_ROCR_20
#ifdef USE_ROCR_V2
case hipDeviceAttributeMemoryClockRate:
*pi = prop->memoryClockRate; break;
case hipDeviceAttributeMemoryBusWidth:
-2
Datei anzeigen
@@ -67,10 +67,8 @@ int main(int argc, char *argv[])
CHECK(test_hipDeviceGetAttribute(deviceId, hipDeviceAttributeWarpSize, props.warpSize));
CHECK(test_hipDeviceGetAttribute(deviceId, hipDeviceAttributeMaxRegistersPerBlock, props.regsPerBlock));
CHECK(test_hipDeviceGetAttribute(deviceId, hipDeviceAttributeClockRate, props.clockRate));
#ifdef USE_ROCR_20
CHECK(test_hipDeviceGetAttribute(deviceId, hipDeviceAttributeMemoryClockRate, props.memoryClockRate));
CHECK(test_hipDeviceGetAttribute(deviceId, hipDeviceAttributeMemoryBusWidth, props.memoryBusWidth));
#endif
CHECK(test_hipDeviceGetAttribute(deviceId, hipDeviceAttributeMultiprocessorCount, props.multiProcessorCount));
CHECK(test_hipDeviceGetAttribute(deviceId, hipDeviceAttributeIsMultiGpuBoard, props.isMultiGpuBoard));
CHECK(test_hipDeviceGetAttribute(deviceId, hipDeviceAttributeComputeMode, props.computeMode));