SWDEV-344135 - Enable stream ordered memory and mempool tests
- Implement CUDA mappings for stream ordered memory allocator and memory pool APIs Change-Id: I2434118ff043527ec7c3389cd5175e1e21d032bf
Этот коммит содержится в:
Rakesh Roy
коммит произвёл
Rakesh Roy
Rakesh Roy
родитель
b875936c5d
Коммит
a44d54f279
@@ -449,10 +449,21 @@ typedef struct cudaPos hipPos;
|
||||
|
||||
typedef cudaMemPool_t hipMemPool_t;
|
||||
typedef enum cudaMemPoolAttr hipMemPoolAttr;
|
||||
#define hipMemPoolReuseFollowEventDependencies cudaMemPoolReuseFollowEventDependencies
|
||||
#define hipMemPoolReuseAllowOpportunistic cudaMemPoolReuseAllowOpportunistic
|
||||
#define hipMemPoolReuseAllowInternalDependencies cudaMemPoolReuseAllowInternalDependencies
|
||||
#define hipMemPoolAttrReleaseThreshold cudaMemPoolAttrReleaseThreshold
|
||||
#define hipMemPoolAttrReservedMemCurrent cudaMemPoolAttrReservedMemCurrent
|
||||
#define hipMemPoolAttrReservedMemHigh cudaMemPoolAttrReservedMemHigh
|
||||
#define hipMemPoolAttrUsedMemCurrent cudaMemPoolAttrUsedMemCurrent
|
||||
#define hipMemPoolAttrUsedMemHigh cudaMemPoolAttrUsedMemHigh
|
||||
typedef struct cudaMemLocation hipMemLocation;
|
||||
typedef struct cudaMemPoolProps hipMemPoolProps;
|
||||
typedef struct cudaMemAccessDesc hipMemAccessDesc;
|
||||
typedef enum cudaMemAccessFlags hipMemAccessFlags;
|
||||
#define hipMemAccessFlagsProtNone cudaMemAccessFlagsProtNone
|
||||
#define hipMemAccessFlagsProtRead cudaMemAccessFlagsProtRead
|
||||
#define hipMemAccessFlagsProtReadWrite cudaMemAccessFlagsProtReadWrite
|
||||
typedef enum cudaMemAllocationHandleType hipMemAllocationHandleType;
|
||||
typedef struct cudaMemPoolPtrExportData hipMemPoolPtrExportData;
|
||||
|
||||
@@ -1151,24 +1162,59 @@ typedef enum cudaStreamUpdateCaptureDependenciesFlags hipStreamUpdateCaptureDepe
|
||||
#define hipStreamSetCaptureDependencies cudaStreamSetCaptureDependencies
|
||||
#endif
|
||||
#if CUDA_VERSION >= CUDA_10020
|
||||
typedef struct CUmemAllocationProp_st hipMemAllocationProp;
|
||||
#define hipMemAllocationGranularityMinimum CU_MEM_ALLOC_GRANULARITY_MINIMUM
|
||||
#define hipMemAllocationGranularityRecommended CU_MEM_ALLOC_GRANULARITY_RECOMMENDED
|
||||
typedef enum CUmemAllocationGranularity_flags_enum hipMemAllocationGranularity_flags;
|
||||
//typedef struct CUmemLocation_st hipMemLocation;
|
||||
typedef enum CUmemLocationType_enum hipMemLocationType;
|
||||
#define hipMemLocationTypeInvalid CU_MEM_LOCATION_TYPE_INVALID
|
||||
#define hipMemLocationTypeDevice CU_MEM_LOCATION_TYPE_DEVICE
|
||||
//typedef enum CUmemAllocationHandleType_enum hipMemAllocationHandleType;
|
||||
#define hipMemHandleTypeNone CU_MEM_HANDLE_TYPE_NONE
|
||||
#define hipMemHandleTypePosixFileDescriptor CU_MEM_HANDLE_TYPE_POSIX_FILE_DESCRIPTOR
|
||||
#define hipMemHandleTypeWin32 CU_MEM_HANDLE_TYPE_WIN32
|
||||
#define hipMemHandleTypeWin32Kmt CU_MEM_HANDLE_TYPE_WIN32_KMT
|
||||
typedef enum CUmemAllocationType_enum hipMemAllocationType;
|
||||
#define hipMemAllocationTypeInvalid CU_MEM_ALLOCATION_TYPE_INVALID
|
||||
#define hipMemAllocationTypePinned CU_MEM_ALLOCATION_TYPE_PINNED
|
||||
#define hipMemAllocationTypeMax CU_MEM_ALLOCATION_TYPE_MAX
|
||||
typedef enum cudaMemLocationType hipMemLocationType;
|
||||
#define hipMemLocationTypeInvalid cudaMemLocationTypeInvalid
|
||||
#define hipMemLocationTypeDevice cudaMemLocationTypeDevice
|
||||
#define hipMemHandleTypeNone cudaMemHandleTypeNone
|
||||
#define hipMemHandleTypePosixFileDescriptor cudaMemHandleTypePosixFileDescriptor
|
||||
#define hipMemHandleTypeWin32 cudaMemHandleTypeWin32
|
||||
#define hipMemHandleTypeWin32Kmt cudaMemHandleTypeWin32Kmt
|
||||
typedef enum cudaMemAllocationType hipMemAllocationType;
|
||||
#define hipMemAllocationTypeInvalid cudaMemAllocationTypeInvalid
|
||||
#define hipMemAllocationTypePinned cudaMemAllocationTypePinned
|
||||
#define hipMemAllocationTypeMax cudaMemAllocationTypeMax
|
||||
#define hipMemGenericAllocationHandle_t CUmemGenericAllocationHandle
|
||||
// Explicitely declaring hipMemAllocationProp based on CUmemAllocationProp but using CUDA runtime members instead
|
||||
// Because hipMemAllocationType, hipMemAllocationHandleType & hipMemLocation are defined using CUDA runtime data types & also used by hipMemPoolProps
|
||||
// Currently there doesn't exist CUDA inbuilt runtime structure corresponding to CUmemAllocationProp
|
||||
// Need to update this structure accordingly if CUDA updates CUmemAllocationProp
|
||||
typedef struct hipMemAllocationProp {
|
||||
/** Memory allocation type */
|
||||
hipMemAllocationType type;
|
||||
/** Requested handle type */
|
||||
hipMemAllocationHandleType requestedHandleTypes;
|
||||
/** Location of allocation */
|
||||
hipMemLocation location;
|
||||
/**
|
||||
* Windows-specific POBJECT_ATTRIBUTES required when
|
||||
* ::CU_MEM_HANDLE_TYPE_WIN32 is specified. This object atributes structure
|
||||
* includes security attributes that define
|
||||
* the scope of which exported allocations may be tranferred to other
|
||||
* processes. In all other cases, this field is required to be zero.
|
||||
*/
|
||||
void *win32HandleMetaData;
|
||||
struct {
|
||||
/**
|
||||
* Allocation hint for requesting compressible memory.
|
||||
* On devices that support Compute Data Compression, compressible
|
||||
* memory can be used to accelerate accesses to data with unstructured
|
||||
* sparsity and other compressible data patterns. Applications are
|
||||
* expected to query allocation property of the handle obtained with
|
||||
* ::cuMemCreate using ::cuMemGetAllocationPropertiesFromHandle to
|
||||
* validate if the obtained allocation is compressible or not. Note that
|
||||
* compressed memory may not be mappable on all devices.
|
||||
*/
|
||||
unsigned char compressionType;
|
||||
/** RDMA capable */
|
||||
unsigned char gpuDirectRDMACapable;
|
||||
/** Bitmask indicating intended usage for this allocation */
|
||||
unsigned short usage;
|
||||
unsigned char reserved[4];
|
||||
} allocFlags;
|
||||
} hipMemAllocationProp;
|
||||
#endif
|
||||
/**
|
||||
* Stream CallBack struct
|
||||
@@ -1877,16 +1923,35 @@ inline static hipError_t hipDeviceGetAttribute(int* pi, hipDeviceAttribute_t att
|
||||
return hipCUDAErrorTohipError(cerror);
|
||||
}
|
||||
#if CUDA_VERSION >= CUDA_10020
|
||||
inline static CUmemAllocationProp hipMemAllocationPropToCUmemAllocationProp(const hipMemAllocationProp* prop) {
|
||||
CUmemAllocationProp cuProp;
|
||||
cuProp.type = (CUmemAllocationType)prop->type;
|
||||
cuProp.requestedHandleTypes = (CUmemAllocationHandleType)prop->requestedHandleTypes;
|
||||
cuProp.location.type = (CUmemLocationType)prop->location.type;
|
||||
cuProp.location.id = prop->location.id;
|
||||
cuProp.win32HandleMetaData = prop->win32HandleMetaData;
|
||||
cuProp.allocFlags.compressionType = prop->allocFlags.compressionType;
|
||||
cuProp.allocFlags.gpuDirectRDMACapable = prop->allocFlags.gpuDirectRDMACapable;
|
||||
cuProp.allocFlags.usage = prop->allocFlags.usage;
|
||||
cuProp.allocFlags.reserved[0] = prop->allocFlags.reserved[0];
|
||||
cuProp.allocFlags.reserved[1] = prop->allocFlags.reserved[1];
|
||||
cuProp.allocFlags.reserved[2] = prop->allocFlags.reserved[2];
|
||||
cuProp.allocFlags.reserved[3] = prop->allocFlags.reserved[3];
|
||||
return cuProp;
|
||||
}
|
||||
|
||||
inline static hipError_t hipMemGetAllocationGranularity(size_t* granularity,
|
||||
const hipMemAllocationProp* prop,
|
||||
hipMemAllocationGranularity_flags option) {
|
||||
return hipCUResultTohipError(cuMemGetAllocationGranularity(granularity, prop, option));
|
||||
CUmemAllocationProp cuProp = hipMemAllocationPropToCUmemAllocationProp(prop);
|
||||
return hipCUResultTohipError(cuMemGetAllocationGranularity(granularity, &cuProp, option));
|
||||
}
|
||||
inline static hipError_t hipMemCreate(hipMemGenericAllocationHandle_t* handle,
|
||||
size_t size,
|
||||
const hipMemAllocationProp* prop,
|
||||
unsigned long long flags) {
|
||||
return hipCUResultTohipError(cuMemCreate(handle, size, prop, flags));
|
||||
CUmemAllocationProp cuProp = hipMemAllocationPropToCUmemAllocationProp(prop);
|
||||
return hipCUResultTohipError(cuMemCreate(handle, size, &cuProp, flags));
|
||||
}
|
||||
inline static hipError_t hipMemRelease(hipMemGenericAllocationHandle_t handle) {
|
||||
return hipCUResultTohipError(cuMemRelease(handle));
|
||||
|
||||
Ссылка в новой задаче
Block a user