Page tree

 

JAVA

FORTRAN

C++

C

 

Link

H5P_SET_CHUNK_CACHE

Sets the raw data chunk cache parameters

Procedure:

H5P_SET_CHUNK_CACHE ( dapl_id, rdcc_nslots, rdcc_nbytes, rdcc_w0 )

Signature:

herr_t H5Pset_chunk_cache(
            hid_t dapl_id,
            size_t rdcc_nslots,
            size_t rdcc_nbytes,
            double rdcc_w0
        )
  

Fortran90 Interface: h5pset_chunk_cache_f
    
SUBROUTINE h5pset_chunk_cache_f(dapl_id, rdcc_nslots, rdcc_nbytes, rdcc_w0, hdferr)
  IMPLICIT NONE
  INTEGER(HID_T), INTENT(IN) :: dapl_id      ! Dataset access property list 
                                             ! identifier.
  INTEGER(SIZE_T), INTENT(IN) :: rdcc_nslots ! The number of chunk slots in the 
                                             ! raw data chunk cache for this 
                                             ! dataset.
  INTEGER(SIZE_T), INTENT(IN) :: rdcc_nbytes ! The total size of the raw data 
                                             ! chunk cache for this dataset.
  REAL, INTENT(IN) :: rdcc_w0                ! The chunk preemption policy for 
                                             ! this dataset.
  INTEGER, INTENT(OUT) :: hdferr             ! Error code
                                             ! 0 on success and -1 on failure
END SUBROUTINE h5pset_chunk_cache_f
    

Parameters:
hid_t dapl_idIN: Dataset access property list identifier
size_t  rdcc_nslotsIN:The number of chunk slots in the raw data chunk cache for this dataset. Increasing this value reduces the number of cache collisions, but slightly increases the memory used. Due to the hashing strategy, this value should ideally be a prime number. As a rule of thumb, this value should be at least 10 times the number of chunks that can fit in rdcc_nbytes bytes. For maximum performance, this value should be set approximately 100 times that number of chunks.

The default value is 521. If the value passed is H5D_CHUNK_CACHE_NSLOTS_DEFAULT, then the property will not be set on dapl_id and the parameter will come from the file access property list used to open the file.

size_t rdcc_nbytesIN: The total size of the raw data chunk cache for this dataset. In most cases increasing this number will improve performance, as long as you have enough free memory.

The default size is 1 MB. If the value passed is H5D_CHUNK_CACHE_NBYTES_DEFAULT, then the property will not be set on dapl_id and the parameter will come from the file access property list.

double rdcc_w0IN: The chunk preemption policy for this dataset. This must be between 0 and 1 inclusive and indicates the weighting according to which chunks which have been fully read or written are penalized when determining which chunks to flush from cache. A value of 0 means fully read or written chunks are treated no differently than other chunks (the preemption is strictly LRU) while a value of 1 means fully read or written chunks are always preempted before other chunks. If your application only reads or writes data once, this can be safely set to 1. Otherwise, this should be set lower, depending on how often you re-read or re-write the same data.

The default value is 0.75. If the value passed is H5D_CHUNK_CACHE_W0_DEFAULT, then the property will not be set on dapl_id and the parameter will come from the file access property list.

Motivation:
H5P_SET_CHUNK_CACHE is used to adjust the chunk cache parameters on a per-dataset basis, as opposed to a global setting for the file using H5P_SET_CACHE. The optimum chunk cache parameters may vary widely with different data layout and access patterns, so for optimal performance they must be set individually for each dataset. It may also be beneficial to reduce the size of the chunk cache for datasets whose performance is not important in order to save memory space.

Description:

H5P_SET_CHUNK_CACHE sets the number of elements, the total number of bytes, and the preemption policy value in the raw data chunk cache on a dataset access property list. After calling this function, the values set in the property list will override the values in the file's file access property list.

The raw data chunk cache inserts chunks into the cache by first computing a hash value using the address of a chunk, then using that hash value as the chunk's index into the table of cached chunks. The size of this hash table, i.e., and the number of possible hash values, is determined by the rdcc_nslots parameter. If a different chunk in the cache has the same hash value, this causes a collision, which reduces efficiency. If inserting the chunk into cache would cause the cache to be too big, then the cache is pruned according to the rdcc_w0 parameter.

Returns:

Returns a non-negative value if successful; otherwise returns a negative value.

Example Usage:

The following code sets the chunk cache to use a hash table with 12421 elements and a maximum size of 16 MB, while using the preemption policy specified for the entire file:

 

                H5Pset_chunk_cache(dapl_id, 12421, 16*1024*1024, H5D_CHUNK_CACHE_W0_DEFAULT);

 

Usage Notes:

The chunk cache size is a property for accessing a dataset and is not stored with a dataset or a file. To guarantee the same chunk cache settings each time the dataset is opened, call H5D_OPEN with a dataset access property list where the chunk cache size is set by calling H5P_SET_CHUNK_CACHE for that property list. The property list can be used for multiple accesses in the same application.

For files where the same chunk cache size will be appropriate for all or most datasets, H5P_SET_CACHE can be called with a file access property list to set the chunk cache size for accessing all datasets in the file.

Both methods can be used in combination, in which case the chunk cache size set by H5P_SET_CACHE will apply except for specific datasets where H5D_OPEN is called with dataset property list with the chunk cache size set by H5P_SET_CHUNK_CACHE .

In the absence of any cache settings, H5D_OPEN will by default create a 1 MB chunk cache for the opened dataset. If this size happens to be appropriate, no call will be needed to either function to set the chunk cache size.

It is also possible that a change in access pattern for later access to a dataset will change the appropriate chunk cache size. For more detailed information on chunks, chunk caching, and performance pitfalls, see “Chunking in HDF5”.

History:
Release    Change
1.8.3C function introduced in this release.

--- Last Modified: January 02, 2018 | 03:23 PM