diff options
author | Quincey Koziol <koziol@hdfgroup.org> | 2002-07-24 18:56:48 (GMT) |
---|---|---|
committer | Quincey Koziol <koziol@hdfgroup.org> | 2002-07-24 18:56:48 (GMT) |
commit | 40df66ebd027351e48eea8b7499490c9654d3943 (patch) | |
tree | 9dd8b9170bec792e840e2b350ca48153cd45ce88 /src/H5Shyper.c | |
parent | c968525b29ce8cee8e842f6a7a1bbd78acc6b63f (diff) | |
download | hdf5-40df66ebd027351e48eea8b7499490c9654d3943.zip hdf5-40df66ebd027351e48eea8b7499490c9654d3943.tar.gz hdf5-40df66ebd027351e48eea8b7499490c9654d3943.tar.bz2 |
[svn-r5834] Purpose:
Large code cleanup/re-write
Description:
This is phase 1 of the data I/O re-architecture, with the following changes:
- Changed the selection drivers to not actually do any I/O, they
only generate the sequences of offset/length pairs needed for
the I/O (or memory access, in the case of iterating or filling
a selection in a memory buffer)
- Wrote more abstract I/O routines which get the sequence of offset/
length pairs for each selection and access perform the I/O or
memory access.
Benefits of this change include:
- Removed ~3400 lines of quite redundant code, with corresponding
reduction in the size of library binary.
- Any selection can now directly access memory when performing I/O,
if no type conversions are required, instead of just "regular"
hyperslab and 'all' selections, which speeds up I/O.
- Sped up I/O for hyperslab selections which have contiguous lower
dimensions by "flattening" them out into lesser dimensional objects
for the I/O.
No file format or API changes were necessary for this change.
The next phase will be to create a "selection driver" for each type of
selection, allowing each type of selection to directly call certain
methods that only apply to that type of selection, instead of passing
through dozens of functions which have switch statements to call the
appropriate method for each selection type. This will also reduce
the amount of code in the library and speed things up a bit more.
Phase 3 will involve generating an MPI datatype for all types of selections,
instead of only "regular" hyperslab and 'all' selections. This will
allow collective parallel I/O for all I/O operations which don't
require type conversions. It will also open up the door for allowing
collective I/O on datasets which require type conversion.
Phase 4 will involve changing the access pattern to deal with chunked
datasets in a more optimal way (in serial).
Phase 5 will deal with accessing chunked datasets more optimally for
collective parallel I/O operations.
Platforms tested:
FreeBSD 4.6 (sleipnir) w/ parallel & C++ and IRIX64 6.5 (modi4) w/parallel
Diffstat (limited to 'src/H5Shyper.c')
-rw-r--r-- | src/H5Shyper.c | 5847 |
1 files changed, 1180 insertions, 4667 deletions
diff --git a/src/H5Shyper.c b/src/H5Shyper.c index 0e30a73..04cecf6 100644 --- a/src/H5Shyper.c +++ b/src/H5Shyper.c @@ -10,17 +10,12 @@ #define H5S_PACKAGE /*suppress error about including H5Spkg */ -#include "H5private.h" -#include "H5Dprivate.h" -#include "H5Eprivate.h" -#include "H5Fprivate.h" -#include "H5FLprivate.h" /*Free Lists */ -#include "H5Iprivate.h" -#include "H5MMprivate.h" -#include "H5Pprivate.h" /* Property Lists */ -#include "H5Spkg.h" -#include "H5Tprivate.h" /* Datatypes */ -#include "H5Vprivate.h" +#include "H5private.h" /* Generic Functions */ +#include "H5Eprivate.h" /* Error handling */ +#include "H5FLprivate.h" /* Free Lists */ +#include "H5Iprivate.h" /* ID Functions */ +#include "H5Spkg.h" /* Dataspace functions */ +#include "H5Vprivate.h" /* Vector functions */ /* Interface initialization */ #define PABLO_MASK H5Shyper_mask @@ -28,58 +23,14 @@ static int interface_initialize_g = 0; /* Local datatypes */ -/* Parameter block for H5S_hyper_select_iter_mem */ -typedef struct { - hid_t dt; - size_t elem_size; - const H5S_t *space; - H5S_sel_iter_t *iter; - void *src; - H5D_operator_t op; - void * op_data; -} H5S_hyper_iter_info_t; /* Static function prototypes */ -static herr_t H5S_hyper_init (const H5S_t *space, size_t elmt_size, H5S_sel_iter_t *iter); -static hsize_t H5S_hyper_favail (const H5S_t UNUSED *space, - const H5S_sel_iter_t *sel_iter, hsize_t max); -static hsize_t H5S_hyper_fgath (H5F_t *f, const struct H5O_layout_t *layout, - H5P_genplist_t *dc_plist, size_t elmt_size, - const H5S_t *file_space, H5S_sel_iter_t *file_iter, - hsize_t nelmts, hid_t dxpl_id, void *_buf/*out*/); -static herr_t H5S_hyper_fscat (H5F_t *f, const struct H5O_layout_t *layout, - H5P_genplist_t *dc_plist, size_t elmt_size, - const H5S_t *file_space, H5S_sel_iter_t *file_iter, - hsize_t nelmts, hid_t dxpl_id, const void *_buf); -static hsize_t H5S_hyper_mgath (const void *_buf, size_t elmt_size, - const H5S_t *mem_space, H5S_sel_iter_t *mem_iter, - hsize_t nelmts, void *_tconv_buf/*out*/); -static herr_t H5S_hyper_mscat (const void *_tconv_buf, size_t elmt_size, - const H5S_t *mem_space, H5S_sel_iter_t *mem_iter, - hsize_t nelmts, void *_buf/*out*/); static herr_t H5S_hyper_free_span_info (H5S_hyper_span_info_t *span_info); static herr_t H5S_hyper_free_span (H5S_hyper_span_t *span); static H5S_hyper_span_info_t *H5S_hyper_copy_span (H5S_hyper_span_info_t *spans); static herr_t H5S_hyper_span_scratch (H5S_hyper_span_info_t *spans, void *scr_value); static herr_t H5S_hyper_span_precompute (H5S_hyper_span_info_t *spans, size_t elmt_size); -const H5S_fconv_t H5S_HYPER_FCONV[1] = {{ - "hslab", /*name */ - H5S_SEL_HYPERSLABS, /*selection type */ - H5S_hyper_init, /*initialize */ - H5S_hyper_favail, /*available */ - H5S_hyper_fgath, /*gather */ - H5S_hyper_fscat, /*scatter */ -}}; - -const H5S_mconv_t H5S_HYPER_MCONV[1] = {{ - "hslab", /*name */ - H5S_SEL_HYPERSLABS, /*selection type */ - H5S_hyper_init, /*initialize */ - H5S_hyper_mgath, /*gather */ - H5S_hyper_mscat, /*scatter */ -}}; - /* Declare a free list to manage the H5S_hyper_span_t struct */ H5FL_DEFINE_STATIC(H5S_hyper_span_t); @@ -89,8 +40,8 @@ H5FL_ARR_DEFINE_STATIC(H5S_hyper_span_t,H5S_MAX_RANK); /* Declare a free list to manage the H5S_hyper_span_info_t struct */ H5FL_DEFINE_STATIC(H5S_hyper_span_info_t); -/* Declare a free list to manage arrays of size_t */ -H5FL_ARR_DEFINE_STATIC(size_t,-1); +/* Declare external the free list for hssize_t arrays */ +H5FL_ARR_EXTERN(hssize_t); /* Declare a free list to manage arrays of hsize_t */ H5FL_ARR_DEFINE_STATIC(hsize_t,-1); @@ -148,11 +99,14 @@ H5S_hyper_print_spans(const struct H5S_hyper_span_info_t *span_lst) * *------------------------------------------------------------------------- */ -static herr_t +herr_t H5S_hyper_init (const H5S_t *space, size_t elmt_size, H5S_sel_iter_t *sel_iter) { + unsigned cont_dim; /* Maximum contiguous dimension */ + const H5S_hyper_dim_t *tdiminfo; /* Temporary pointer to diminfo information */ H5S_hyper_span_info_t *spans; /* Pointer to hyperslab span info node */ - unsigned u; /* Index variable */ + unsigned u; /* Index variable */ + int i; /* Index variable */ FUNC_ENTER_NOAPI(H5S_hyper_init, FAIL); @@ -163,14 +117,87 @@ H5S_hyper_init (const H5S_t *space, size_t elmt_size, H5S_sel_iter_t *sel_iter) /* Initialize the number of points to iterate over */ sel_iter->hyp.elmt_left=space->select.num_elem; + sel_iter->hyp.iter_rank=0; + + /* Set the temporary pointer to the dimension information */ + tdiminfo=space->select.sel_info.hslab.diminfo; /* Check for the special case of just one H5Sselect_hyperslab call made */ - if(space->select.sel_info.hslab.diminfo!=NULL) { + if(tdiminfo!=NULL) { /* Initialize the information needed for regular hyperslab I/O */ - /* Allocate the position & initialize to initial location */ - sel_iter->hyp.off = H5FL_ARR_ALLOC(hsize_t,space->extent.u.simple.rank,0); - for(u=0; u<space->extent.u.simple.rank; u++) - sel_iter->hyp.off[u]=space->select.sel_info.hslab.diminfo[u].start; + const hsize_t *mem_size; /* Temporary pointer to dataspace extent's dimension sizes */ + hsize_t acc; /* Accumulator for "flattened" dimension's sizes */ + + /* Set the temporary pointer to the dataspace extent's dimension sizes */ + mem_size=space->extent.u.simple.size; + + /* + * For a regular hyperslab to be contiguous up to some dimension, it + * must have only one block (i.e. count==1 in all dimensions up to that + * dimension) and the block size must be the same as the dataspace's + * extent in that dimension and all dimensions up to that dimension. + */ + + /* Initialize the number of contiguous dimensions to be the same as the dataspace's rank */ + cont_dim=space->extent.u.simple.rank; + + /* Check for a "contiguous" block */ + for(u=space->extent.u.simple.rank-1; u>0; u--) { + if(tdiminfo[u].count==1 && tdiminfo[u].block==mem_size[u]) + cont_dim=u; + } /* end for */ + + /* Check if the regular selection can be "flattened" */ + if(cont_dim<space->extent.u.simple.rank) { + /* Set the iterator's rank to the contiguous dimensions */ + sel_iter->hyp.iter_rank=cont_dim; + + /* Allocate the position & initialize to initial location */ + sel_iter->hyp.off = H5FL_ARR_ALLOC(hsize_t,cont_dim,0); + sel_iter->hyp.diminfo = H5FL_ARR_ALLOC(H5S_hyper_dim_t,cont_dim,0); + sel_iter->hyp.size = H5FL_ARR_ALLOC(hsize_t,cont_dim,0); + sel_iter->hyp.sel_off = H5FL_ARR_ALLOC(hssize_t,cont_dim,0); + + /* "Flatten" dataspace extent and selection information */ + for(i=space->extent.u.simple.rank-1, acc=1; i>=0; i--) { + if(tdiminfo[i].block==mem_size[i] && i>0) { + assert(tdiminfo[i].start==0); + acc *= mem_size[i]; + } /* end if */ + else { + if((unsigned)i==(cont_dim-1)) { + sel_iter->hyp.diminfo[i].start = tdiminfo[i].start*acc; + /* Special case for stride==1 regular selections */ + if(tdiminfo[i].stride==1) + sel_iter->hyp.diminfo[i].stride = 1; + else + sel_iter->hyp.diminfo[i].stride = tdiminfo[i].stride*acc; + sel_iter->hyp.diminfo[i].count = tdiminfo[i].count; + sel_iter->hyp.diminfo[i].block = tdiminfo[i].block*acc; + sel_iter->hyp.size[i] = mem_size[i]*acc; + sel_iter->hyp.sel_off[i] = space->select.offset[i]*acc; + } /* end if */ + else { + sel_iter->hyp.diminfo[i].start = tdiminfo[i].start; + sel_iter->hyp.diminfo[i].stride = tdiminfo[i].stride; + sel_iter->hyp.diminfo[i].count = tdiminfo[i].count; + sel_iter->hyp.diminfo[i].block = tdiminfo[i].block; + sel_iter->hyp.size[i] = mem_size[i]; + sel_iter->hyp.sel_off[i] = space->select.offset[i]; + } /* end else */ + } /* end if */ + } /* end for */ + + /* Initialize "flattened" iterator offset to initial location and dataspace extent and selection information to correct values */ + for(u=0; u<cont_dim; u++) + sel_iter->hyp.off[u]=sel_iter->hyp.diminfo[u].start; + } /* end if */ + else { + /* Allocate the position & initialize to initial location */ + sel_iter->hyp.off = H5FL_ARR_ALLOC(hsize_t,space->extent.u.simple.rank,0); + for(u=0; u<space->extent.u.simple.rank; u++) + sel_iter->hyp.off[u]=tdiminfo[u].start; + } /* end else */ } /* end if */ else { /* Initialize the information needed for non-regular hyperslab I/O */ @@ -231,10 +258,23 @@ H5S_hyper_sel_iter_release (H5S_sel_iter_t *sel_iter) /* Check args */ assert (sel_iter); - /* Release the array of offsets/positions */ + /* Release the common array of offsets/positions */ if(sel_iter->hyp.off!=NULL) H5FL_ARR_FREE(hsize_t,sel_iter->hyp.off); +/* Release the information needed for "flattened" regular hyperslab I/O */ + /* Free the "flattened" dataspace extent */ + if(sel_iter->hyp.size!=NULL) + H5FL_ARR_FREE(hsize_t,sel_iter->hyp.size); + + /* Free the "flattened" regular hyperslab selection */ + if(sel_iter->hyp.diminfo!=NULL) + H5FL_ARR_FREE(H5S_hyper_dim_t,sel_iter->hyp.diminfo); + + /* Free the "flattened" selection offset */ + if(sel_iter->hyp.sel_off!=NULL) + H5FL_ARR_FREE(hssize_t,sel_iter->hyp.sel_off); + /* Release the information needed for non-regular hyperslab I/O */ /* Free the copy of the selections span tree */ if(sel_iter->hyp.spans!=NULL) @@ -262,7 +302,7 @@ H5S_hyper_sel_iter_release (H5S_sel_iter_t *sel_iter) * *------------------------------------------------------------------------- */ -static hsize_t +hsize_t H5S_hyper_favail (const H5S_t * UNUSED space, const H5S_sel_iter_t *sel_iter, hsize_t max) { @@ -296,8 +336,9 @@ H5S_hyper_favail (const H5S_t * UNUSED space, *------------------------------------------------------------------------- */ static int -H5S_hyper_iter_next (const H5S_t *file_space, H5S_sel_iter_t *file_iter) +H5S_hyper_iter_next (const H5S_t *space, H5S_sel_iter_t *iter) { + const H5S_hyper_dim_t *tdiminfo; /* Temporary pointer to diminfo information */ hsize_t iter_offset[H5O_LAYOUT_NDIMS]; hsize_t iter_count[H5O_LAYOUT_NDIMS]; int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ @@ -307,19 +348,33 @@ H5S_hyper_iter_next (const H5S_t *file_space, H5S_sel_iter_t *file_iter) FUNC_ENTER_NOINIT(H5S_hyper_iter_next); - /* Set some useful rank information */ - fast_dim=file_space->extent.u.simple.rank-1; - ndims=file_space->extent.u.simple.rank; + /* Check if this is a "flattened" regular hyperslab selection */ + if(iter->hyp.iter_rank!=0 && iter->hyp.iter_rank<space->extent.u.simple.rank) { + /* Set the aliases for a few important dimension ranks */ + ndims=iter->hyp.iter_rank; + fast_dim=ndims-1; + + /* Set the local copy of the diminfo pointer */ + tdiminfo=iter->hyp.diminfo; + } /* end if */ + else { + /* Set the aliases for a few important dimension ranks */ + ndims=space->extent.u.simple.rank; + fast_dim=ndims-1; + + /* Set the local copy of the diminfo pointer */ + tdiminfo=space->select.sel_info.hslab.diminfo; + } /* end else */ /* Calculate the offset and block count for each dimension */ for(i=0; i<ndims; i++) { - if(file_space->select.sel_info.hslab.diminfo[i].stride==1) { - iter_offset[i]=file_iter->hyp.off[i]-file_space->select.sel_info.hslab.diminfo[i].start; + if(tdiminfo[i].stride==1) { + iter_offset[i]=iter->hyp.off[i]-tdiminfo[i].start; iter_count[i]=0; } /* end if */ else { - iter_offset[i]=(file_iter->hyp.off[i]-file_space->select.sel_info.hslab.diminfo[i].start)%file_space->select.sel_info.hslab.diminfo[i].stride; - iter_count[i]=(file_iter->hyp.off[i]-file_space->select.sel_info.hslab.diminfo[i].start)/file_space->select.sel_info.hslab.diminfo[i].stride; + iter_offset[i]=(iter->hyp.off[i]-tdiminfo[i].start)%tdiminfo[i].stride; + iter_count[i]=(iter->hyp.off[i]-tdiminfo[i].start)/tdiminfo[i].stride; } /* end else */ } /* end for */ @@ -332,7 +387,7 @@ H5S_hyper_iter_next (const H5S_t *file_space, H5S_sel_iter_t *file_iter) iter_count[temp_dim]++; /* If this block is still in the range of blocks to output for the dimension, break out of loop */ - if(iter_count[temp_dim]<file_space->select.sel_info.hslab.diminfo[temp_dim].count) + if(iter_count[temp_dim]<tdiminfo[temp_dim].count) break; else iter_count[temp_dim]=0; /* reset back to the beginning of the line */ @@ -342,7 +397,7 @@ H5S_hyper_iter_next (const H5S_t *file_space, H5S_sel_iter_t *file_iter) iter_offset[temp_dim]++; /* If this block is still in the range of blocks to output for the dimension, break out of loop */ - if(iter_offset[temp_dim]<file_space->select.sel_info.hslab.diminfo[temp_dim].block) + if(iter_offset[temp_dim]<tdiminfo[temp_dim].block) break; else { /* Move to the next block in the current dimension */ @@ -350,7 +405,7 @@ H5S_hyper_iter_next (const H5S_t *file_space, H5S_sel_iter_t *file_iter) iter_count[temp_dim]++; /* If this block is still in the range of blocks to output for the dimension, break out of loop */ - if(iter_count[temp_dim]<file_space->select.sel_info.hslab.diminfo[temp_dim].count) + if(iter_count[temp_dim]<tdiminfo[temp_dim].count) break; else iter_count[temp_dim]=0; /* reset back to the beginning of the line */ @@ -363,3735 +418,12 @@ H5S_hyper_iter_next (const H5S_t *file_space, H5S_sel_iter_t *file_iter) /* Translate current iter_offset and iter_count into iterator position */ for(i=0; i<ndims; i++) - file_iter->hyp.off[i]=file_space->select.sel_info.hslab.diminfo[i].start+(file_space->select.sel_info.hslab.diminfo[i].stride*iter_count[i])+iter_offset[i]; + iter->hyp.off[i]=tdiminfo[i].start+(tdiminfo[i].stride*iter_count[i])+iter_offset[i]; FUNC_LEAVE (SUCCEED); } /* H5S_hyper_iter_next() */ -/*------------------------------------------------------------------------- - * Function: H5S_hyper_fread - * - * Purpose: Performs an optimized gather from the file, based on a hyperslab - * span tree. - * - * Return: Success: Number of elements copied. - * Failure: 0 - * - * Programmer: Quincey Koziol - * Friday, September 8, 2000 - * - * Modifications: - * - *------------------------------------------------------------------------- - */ -static hsize_t -H5S_hyper_fread (H5F_t *f, const struct H5O_layout_t *layout, - H5P_genplist_t *dc_plist, size_t elmt_size, - const H5S_t *space, H5S_sel_iter_t *iter, - hsize_t nelem, hid_t dxpl_id, void *_buf/*out*/) -{ - uint8_t *dst=(uint8_t *)_buf; /* Alias for pointer arithmetic */ - H5S_hyper_span_t **ispan; /* Iterator's hyperslab span nodes */ - H5S_hyper_span_t *curr_span; /* Current hyperslab span node */ - hsize_t slab[H5O_LAYOUT_NDIMS]; /* Cumulative size of each dimension in bytes */ - hsize_t acc; /* Accumulator for computing cumulative sizes */ - hssize_t *abs_arr; /* Absolute hyperslab span position */ - hssize_t *off_arr; /* Offset within the dataspace extent */ - int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ - int curr_dim; /* Current dimension being operated on */ - int ndims; /* Number of dimensions of dataset */ - hsize_t loc_off; /* Element offset in the dataspace */ - size_t span_size=0; /* Number of bytes in current span to actually process */ - size_t io_bytes_left; /* Number of bytes left to process */ - int i; /* Index variable */ - size_t *seq_len_arr=NULL; /* Array of sequence lengths */ - hsize_t *buf_off_arr=NULL; /* Array of dataset offsets */ - size_t last_io_bytes_left=0; /* Last I/O bytes left before readv() called */ - size_t nseq=0; /* Number of sequence/offsets stored in the arrays */ - size_t vector_size; /* Value for vector size */ - H5P_genplist_t *plist; /* Property list */ - hssize_t ret_value=FAIL; - - FUNC_ENTER_NOINIT(H5S_hyper_fread); - - /* Check args */ - assert(f); - assert(layout); - assert(elmt_size>0); - assert(space); - assert(iter); - assert(nelem>0); - assert(dst); - - /* Set the rank of the fastest changing dimension */ - ndims=space->extent.u.simple.rank; - fast_dim=(ndims-1); - - /* Get the pointers to the current span info and span nodes */ - curr_span=iter->hyp.span[fast_dim]; - abs_arr=iter->hyp.off; - off_arr=space->select.offset; - ispan=iter->hyp.span; - - /* Set the amount of elements to perform I/O on, etc. */ - H5_ASSIGN_OVERFLOW(io_bytes_left,(nelem*elmt_size),hsize_t,size_t); - - /* Compute the cumulative size of dataspace dimensions */ - for(i=fast_dim, acc=elmt_size; i>=0; i--) { - slab[i]=acc; - acc*=space->extent.u.simple.size[i]; - } /* end for */ - - /* Set the offset of the first element iterated on */ - for(i=0, loc_off=0; i<ndims; i++) - /* Compute the sequential element offset */ - loc_off+=(abs_arr[i]+space->select.offset[i])*slab[i]; - - /* Get the hyperslab vector size */ - if(TRUE!=H5P_isa_class(dxpl_id,H5P_DATASET_XFER) || NULL == (plist = H5I_object(dxpl_id))) - HGOTO_ERROR(H5E_ARGS, H5E_BADTYPE, FAIL, "not a file access property list"); - if (H5P_get(plist,H5D_XFER_HYPER_VECTOR_SIZE_NAME,&vector_size)<0) - HGOTO_ERROR(H5E_PLIST, H5E_CANTGET, 0, "unable to get value"); - - /* Allocate the vector I/O arrays */ - if((seq_len_arr = H5FL_ARR_ALLOC(size_t,vector_size,0))==NULL) - HGOTO_ERROR(H5E_RESOURCE, H5E_NOSPACE, 0, "can't allocate vector I/O array"); - if((buf_off_arr = H5FL_ARR_ALLOC(hsize_t,vector_size,0))==NULL) - HGOTO_ERROR(H5E_RESOURCE, H5E_NOSPACE, 0, "can't allocate vector I/O array"); - - /* Range check against number of elements left in selection */ - assert(io_bytes_left<=(iter->hyp.elmt_left*elmt_size)); - - /* Take care of any partial spans leftover from previous I/Os */ - if(abs_arr[fast_dim]!=curr_span->low) { - - /* Finish the span in the fastest changing dimension */ - - /* Compute the number of bytes to attempt in this span */ - H5_ASSIGN_OVERFLOW(span_size,((curr_span->high-abs_arr[fast_dim])+1)*elmt_size,hsize_t,size_t); - - /* Check number of bytes against upper bounds allowed */ - if(span_size>io_bytes_left) - span_size=io_bytes_left; - - if (H5F_seq_read(f, dxpl_id, layout, dc_plist, space, - elmt_size, span_size, loc_off, dst/*out*/)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_READERROR, 0, "read error"); - - /* Increment offset in destination */ - dst+=span_size; - - /* Decrement I/O left to perform */ - io_bytes_left-=span_size; - - /* Check if we are done */ - if(io_bytes_left>0) { - /* Move to next span in fastest changing dimension */ - curr_span=curr_span->next; - - if(curr_span!=NULL) { - /* Move location offset of destination */ - loc_off+=(curr_span->low-abs_arr[fast_dim])*elmt_size; - - /* Move iterator for fastest changing dimension */ - abs_arr[fast_dim]=curr_span->low; - } /* end if */ - } /* end if */ - else { - abs_arr[fast_dim]+=span_size/elmt_size; - - /* Check if we are still within the span */ - if(abs_arr[fast_dim]<=curr_span->high) { - iter->hyp.span[fast_dim]=curr_span; - - goto partial_done; /* finished with partial span */ - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[fast_dim]=curr_span->low; - iter->hyp.span[fast_dim]=curr_span; - - goto partial_done; /* finished with partial span */ - } /* end if */ - } /* end else */ - } /* end else */ - - /* Adjust iterator pointers */ - - if(curr_span==NULL) { -/* Same as code in main loop */ - /* Start at the next fastest dim */ - curr_dim=fast_dim-1; - - /* Work back up through the dimensions */ - while(curr_dim>=0) { - /* Reset the current span */ - curr_span=iter->hyp.span[curr_dim]; - - /* Increment absolute position */ - abs_arr[curr_dim]++; - - /* Check if we are still within the span */ - if(abs_arr[curr_dim]<=curr_span->high) { - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[curr_dim]=curr_span->low; - - break; - } /* end if */ - else { - /* If we finished the span list in this dimension, decrement the dimension worked on and loop again */ - curr_dim--; - } /* end else */ - } /* end else */ - } /* end while */ - - /* Check if we are finished with the spans in the tree */ - if(curr_dim<0) { - /* We had better be done with I/O or bad things are going to happen... */ - assert(io_bytes_left==0); - - goto partial_done; /* finished with partial span */ - } /* end if */ - else { - /* Reset the span in the current dimension */ - ispan[curr_dim]=curr_span; - - /* Walk back down the iterator positions, reseting them */ - while(curr_dim<fast_dim) { - assert(curr_span); - assert(curr_span->down); - assert(curr_span->down->head); - - /* Set the new span_info & span for this dimension */ - iter->hyp.span[curr_dim+1]=curr_span->down->head; - - /* Advance span down the tree */ - curr_span=curr_span->down->head; - - /* Reset the absolute offset for the dim */ - abs_arr[curr_dim+1]=curr_span->low; - - /* Increment current dimension */ - curr_dim++; - } /* end while */ - - /* Verify that the curr_span points to the fastest dim */ - assert(curr_span==iter->hyp.span[fast_dim]); - } /* end else */ - - /* Reset the buffer offset */ - for(i=0, loc_off=0; i<ndims; i++) - loc_off+=(abs_arr[i]+off_arr[i])*slab[i]; - } /* end if */ - } /* end if */ - -partial_done: /* Yes, goto's are evil, so sue me... :-) */ - /* Get the number of bytes left to process currently */ - last_io_bytes_left=io_bytes_left; - - /* Perform the I/O on the elements, based on the position of the iterator */ - while(io_bytes_left>0) { - /* Adjust location offset of destination to compensate for initial increment below */ - loc_off-=curr_span->pstride; - - /* Loop over all the spans in the fastest changing dimension */ - while(curr_span!=NULL) { - /* Move location offset of destination */ - loc_off+=curr_span->pstride; - - /* Compute the number of elements to attempt in this span */ - H5_ASSIGN_OVERFLOW(span_size,curr_span->nelem,hsize_t,size_t); - - /* Check number of elements against upper bounds allowed */ - if(span_size>=io_bytes_left) { - /* Trim the number of bytes to output */ - span_size=io_bytes_left; - io_bytes_left=0; - -/* COMMON */ - /* Store the I/O information for the span */ - seq_len_arr[nseq]=span_size; - buf_off_arr[nseq]=loc_off; - - /* Increment the number of sequences in arrays */ - nseq++; - - /* If the sequence & offset arrays are full, read them in */ - if(nseq>=vector_size) { - /* Read in the sequences */ - if (H5F_seq_readv(f, dxpl_id, layout, dc_plist, space, - elmt_size, nseq, seq_len_arr, buf_off_arr, dst/*out*/)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_READERROR, 0, "read error"); - - /* Increment the offset of the destination buffer */ - dst+=(last_io_bytes_left-io_bytes_left); - - /* Keep around the current number of I/O bytes left */ - last_io_bytes_left=io_bytes_left; - nseq=0; - } /* end else */ -/* end COMMON */ - - /* Break out now, we are finished with I/O */ - break; - } /* end if */ - else { - /* Decrement I/O left to perform */ - io_bytes_left-=span_size; - -/* COMMON */ - /* Store the I/O information for the span */ - seq_len_arr[nseq]=span_size; - buf_off_arr[nseq]=loc_off; - - /* Increment the number of sequences in arrays */ - nseq++; - - /* If the sequence & offset arrays are full, read them in */ - if(nseq>=vector_size) { - /* Read in the sequences */ - if (H5F_seq_readv(f, dxpl_id, layout, dc_plist, space, - elmt_size, nseq, seq_len_arr, buf_off_arr, dst/*out*/)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_READERROR, 0, "read error"); - - /* Increment the offset of the destination buffer */ - dst+=(last_io_bytes_left-io_bytes_left); - - /* Keep around the current number of I/O bytes left */ - last_io_bytes_left=io_bytes_left; - nseq=0; - } /* end else */ -/* end COMMON */ - } /* end else */ - - /* Move to next span in fastest changing dimension */ - curr_span=curr_span->next; - } /* end while */ - - /* Check if we are done */ - if(io_bytes_left==0) { - abs_arr[fast_dim]=curr_span->low+(span_size/elmt_size); - - /* Check if we are still within the span */ - if(abs_arr[fast_dim]<=curr_span->high) { - iter->hyp.span[fast_dim]=curr_span; - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[fast_dim]=curr_span->low; - iter->hyp.span[fast_dim]=curr_span; - break; - } /* end if */ - } /* end else */ - } /* end if */ - - /* Adjust iterator pointers */ - - /* Start at the next fastest dim */ - curr_dim=fast_dim-1; - - /* Work back up through the dimensions */ - while(curr_dim>=0) { - /* Reset the current span */ - curr_span=iter->hyp.span[curr_dim]; - - /* Increment absolute position */ - abs_arr[curr_dim]++; - - /* Check if we are still within the span */ - if(abs_arr[curr_dim]<=curr_span->high) { - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[curr_dim]=curr_span->low; - - break; - } /* end if */ - else { - /* If we finished the span list in this dimension, decrement the dimension worked on and loop again */ - curr_dim--; - } /* end else */ - } /* end else */ - } /* end while */ - - /* Check if we are finished with the spans in the tree */ - if(curr_dim<0) { - /* We had better be done with I/O or bad things are going to happen... */ - assert(io_bytes_left==0); - break; - } /* end if */ - else { - /* Reset the span in the current dimension */ - ispan[curr_dim]=curr_span; - - /* Walk back down the iterator positions, reseting them */ - while(curr_dim<fast_dim) { - assert(curr_span); - assert(curr_span->down); - assert(curr_span->down->head); - - /* Set the new span for the next dimension down */ - iter->hyp.span[curr_dim+1]=curr_span->down->head; - - /* Advance span down the tree */ - curr_span=curr_span->down->head; - - /* Reset the absolute offset for the dim */ - abs_arr[curr_dim+1]=curr_span->low; - - /* Increment current dimension */ - curr_dim++; - } /* end while */ - - /* Verify that the curr_span points to the fastest dim */ - assert(curr_span==iter->hyp.span[fast_dim]); - } /* end else */ - - /* Reset the buffer offset */ - for(i=0, loc_off=0; i<ndims; i++) - loc_off+=(abs_arr[i]+off_arr[i])*slab[i]; - } /* end while */ - - /* Check for any stored sequences which need to be flushed */ - if(nseq>0) { - /* Read in the sequence */ - if (H5F_seq_readv(f, dxpl_id, layout, dc_plist, space, - elmt_size, nseq, seq_len_arr, buf_off_arr, dst/*out*/)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_READERROR, 0, "read error"); - } /* end if */ - - /* Increment amount of I/O performed */ - iter->hyp.elmt_left-=nelem; - - /* Success! */ - ret_value=nelem; - -done: - if(seq_len_arr!=NULL) - H5FL_ARR_FREE(size_t,seq_len_arr); - if(buf_off_arr!=NULL) - H5FL_ARR_FREE(hsize_t,buf_off_arr); - - FUNC_LEAVE (ret_value); -} /* H5S_hyper_fread() */ - - -/*------------------------------------------------------------------------- - * Function: H5S_hyper_fread_opt - * - * Purpose: Performs an optimized gather from the file, based on a regular - * hyperslab (i.e. one which was generated from just one call to - * H5Sselect_hyperslab). - * - * Return: Success: Number of elements copied. - * Failure: 0 - * - * Programmer: Quincey Koziol - * Friday, September 8, 2000 - * - * Modifications: - * - *------------------------------------------------------------------------- - */ -static hsize_t -H5S_hyper_fread_opt (H5F_t *f, const struct H5O_layout_t *layout, - H5P_genplist_t *dc_plist, size_t elmt_size, - const H5S_t *file_space, H5S_sel_iter_t *file_iter, - hsize_t nelmts, hid_t dxpl_id, void *_buf/*out*/) -{ - size_t *seq_len_arr=NULL; /* Array of sequence lengths */ - hsize_t *buf_off_arr=NULL; /* Array of dataset offsets */ - size_t nseq=0; /* Number of sequence/offsets stored in the arrays */ - size_t tot_buf_size=0; /* Total number of bytes in buffer */ - - hssize_t offset[H5O_LAYOUT_NDIMS]; /* Offset on disk */ - hsize_t slab[H5O_LAYOUT_NDIMS]; /* Hyperslab size */ - hssize_t wrap[H5O_LAYOUT_NDIMS]; /* Bytes to wrap around at the end of a row */ - hsize_t skip[H5O_LAYOUT_NDIMS]; /* Bytes to skip between blocks */ - hsize_t tmp_count[H5O_LAYOUT_NDIMS]; /* Temporary block count */ - hsize_t tmp_block[H5O_LAYOUT_NDIMS]; /* Temporary block offset */ - uint8_t *buf=(uint8_t *)_buf; /* Alias for pointer arithmetic */ - const H5S_hyper_dim_t *tdiminfo; /* Local pointer to diminfo information */ - hssize_t fast_dim_start, /* Local copies of fastest changing dimension info */ - fast_dim_offset; - hsize_t fast_dim_stride, /* Local copies of fastest changing dimension info */ - fast_dim_block, - fast_dim_buf_off; - size_t fast_dim_count; - size_t tot_blk_count; /* Total number of blocks left to output */ - size_t act_blk_count; /* Actual number of blocks to output */ - int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ - int temp_dim; /* Temporary rank holder */ - hsize_t acc; /* Accumulator */ - hsize_t buf_off; /* Current buffer offset for copying memory */ - int i; /* Counters */ - unsigned u; /* Counters */ - int ndims; /* Number of dimensions of dataset */ - size_t actual_read; /* The actual number of elements to read in */ - size_t actual_bytes; /* The actual number of bytes to copy */ - size_t io_left; /* The number of elements left in I/O operation */ - size_t tot_seq; /* The number of sequences filled */ - hsize_t *buf_off_arr_p; /* Pointer into the buffer offset array */ - size_t seq_count; /* Temporary count of sequences left to process */ -#ifndef NO_DUFFS_DEVICE - size_t duffs_index; /* Counting index for Duff's device */ -#endif /* NO_DUFFS_DEVICE */ - size_t vector_size; /* Value for vector size */ - H5P_genplist_t *plist; /* Property list */ - hsize_t ret_value=0; /* Return value */ - - FUNC_ENTER_NOINIT(H5S_hyper_fread_opt); - - /* Get the hyperslab vector size */ - if(TRUE!=H5P_isa_class(dxpl_id,H5P_DATASET_XFER) || NULL == (plist = H5I_object(dxpl_id))) - HGOTO_ERROR(H5E_ARGS, H5E_BADTYPE, 0, "not a file access property list"); - if (H5P_get(plist,H5D_XFER_HYPER_VECTOR_SIZE_NAME,&vector_size)<0) - HGOTO_ERROR(H5E_PLIST, H5E_CANTGET, 0, "unable to get value"); - - /* Allocate the vector I/O arrays */ - if((seq_len_arr = H5FL_ARR_ALLOC(size_t,vector_size,0))==NULL) - HGOTO_ERROR(H5E_RESOURCE, H5E_NOSPACE, 0, "can't allocate vector I/O array"); - if((buf_off_arr = H5FL_ARR_ALLOC(hsize_t,vector_size,0))==NULL) - HGOTO_ERROR(H5E_RESOURCE, H5E_NOSPACE, 0, "can't allocate vector I/O array"); - - /* Set the rank of the fastest changing dimension */ - fast_dim=file_space->extent.u.simple.rank-1; - ndims=file_space->extent.u.simple.rank; - - /* initialize row sizes for each dimension */ - for(i=(ndims-1),acc=1; i>=0; i--) { - slab[i]=acc*elmt_size; - acc*=file_space->extent.u.simple.size[i]; - } /* end for */ - - /* Set the number of elements left for I/O */ - H5_ASSIGN_OVERFLOW(io_left,nelmts,hsize_t,size_t); - - /* Check if we stopped in the middle of a sequence of elements */ - if((file_iter->hyp.off[fast_dim]-file_space->select.sel_info.hslab.diminfo[fast_dim].start)%file_space->select.sel_info.hslab.diminfo[fast_dim].stride!=0 || - ((file_iter->hyp.off[fast_dim]!=file_space->select.sel_info.hslab.diminfo[fast_dim].start) && file_space->select.sel_info.hslab.diminfo[fast_dim].stride==1)) { - hsize_t leftover; /* The number of elements left over from the last sequence */ - - /* Calculate the number of elements left in the sequence */ - if(file_space->select.sel_info.hslab.diminfo[fast_dim].stride==1) - leftover=file_space->select.sel_info.hslab.diminfo[fast_dim].block-(file_iter->hyp.off[fast_dim]-file_space->select.sel_info.hslab.diminfo[fast_dim].start); - else - leftover=file_space->select.sel_info.hslab.diminfo[fast_dim].block-((file_iter->hyp.off[fast_dim]-file_space->select.sel_info.hslab.diminfo[fast_dim].start)%file_space->select.sel_info.hslab.diminfo[fast_dim].stride); - - /* Make certain that we don't read too many */ - H5_CHECK_OVERFLOW(leftover,hsize_t,size_t); - actual_read=MIN((size_t)leftover,io_left); - actual_bytes=actual_read*elmt_size; - - /* Copy the location of the point to get */ - HDmemcpy(offset, file_iter->hyp.off,ndims*sizeof(hssize_t)); - offset[ndims] = 0; - - /* Add in the selection offset */ - for(i=0; i<ndims; i++) - offset[i] += file_space->select.offset[i]; - - /* Compute the initial buffer offset */ - for(i=0,buf_off=0; i<ndims; i++) - buf_off+=offset[i]*slab[i]; - - /* Read in the rest of the sequence */ - if (H5F_seq_read(f, dxpl_id, layout, dc_plist, file_space, - elmt_size, actual_bytes, buf_off, buf/*out*/)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_READERROR, 0, "read error"); - - /* Increment the offset of the buffer */ - buf+=actual_bytes; - - /* Decrement the amount left to read */ - io_left-=actual_read; - - /* Advance the point iterator */ - /* If we had enough buffer space to read in the rest of the sequence - * in the fastest changing dimension, move the iterator offset to - * the beginning of the next block to read. Otherwise, just advance - * the iterator in the fastest changing dimension. - */ - if(actual_read==leftover) { - /* Move iterator offset to beginning of next sequence in the fastest changing dimension */ - H5S_hyper_iter_next(file_space,file_iter); - } /* end if */ - else { - file_iter->hyp.off[fast_dim]+=actual_read; /* whole sequence not read in, just advance fastest dimension offset */ - } /* end if */ - } /* end if */ - - /* Now that we've cleared the "remainder" of the previous fastest dimension - * sequence, we must be at the beginning of a sequence, so use the fancy - * algorithm to compute the offsets and run through as many as possible, - * until the buffer fills up. - */ - if(io_left>0) { /* Just in case the "remainder" above filled the buffer */ - /* Compute the arrays to perform I/O on */ - /* Copy the location of the point to get */ - HDmemcpy(offset, file_iter->hyp.off,ndims*sizeof(hssize_t)); - offset[ndims] = 0; - - /* Add in the selection offset */ - for(i=0; i<ndims; i++) - offset[i] += file_space->select.offset[i]; - - /* Compute the current "counts" for this location */ - for(i=0; i<ndims; i++) { - if(file_space->select.sel_info.hslab.diminfo[i].stride==1) { - tmp_count[i] = 0; - tmp_block[i] = file_iter->hyp.off[i]-file_space->select.sel_info.hslab.diminfo[i].start; - } /* end if */ - else { - tmp_count[i] = (file_iter->hyp.off[i]-file_space->select.sel_info.hslab.diminfo[i].start)/file_space->select.sel_info.hslab.diminfo[i].stride; - tmp_block[i] = (file_iter->hyp.off[i]-file_space->select.sel_info.hslab.diminfo[i].start)%file_space->select.sel_info.hslab.diminfo[i].stride; - } /* end else */ - } /* end for */ - - /* Compute the initial buffer offset */ - for(i=0,buf_off=0; i<ndims; i++) - buf_off+=offset[i]*slab[i]; - - /* Set the number of elements to read each time */ - H5_ASSIGN_OVERFLOW(actual_read,file_space->select.sel_info.hslab.diminfo[fast_dim].block,hsize_t,size_t); - - /* Set the number of actual bytes */ - actual_bytes=actual_read*elmt_size; - - /* Set the local copy of the diminfo pointer */ - tdiminfo=file_space->select.sel_info.hslab.diminfo; - - /* Set local copies of information for the fastest changing dimension */ - fast_dim_start=tdiminfo[fast_dim].start; - fast_dim_stride=tdiminfo[fast_dim].stride; - fast_dim_block=tdiminfo[fast_dim].block; - fast_dim_buf_off=slab[fast_dim]*fast_dim_stride; - fast_dim_offset=fast_dim_start+file_space->select.offset[fast_dim]; - - /* Compute the number of blocks which would fit into the buffer */ - H5_ASSIGN_OVERFLOW(tot_blk_count,(io_left/fast_dim_block),hsize_t,size_t); - - /* Compute the amount to wrap at the end of each row */ - for(i=0; i<ndims; i++) - wrap[i]=(file_space->extent.u.simple.size[i]-(tdiminfo[i].stride*tdiminfo[i].count))*slab[i]; - - /* Compute the amount to skip between blocks */ - for(i=0; i<ndims; i++) - skip[i]=(tdiminfo[i].stride-tdiminfo[i].block)*slab[i]; - - /* Fill the sequence length array (since they will all be the same for optimized hyperslabs) */ - for(u=0; u<vector_size; u++) - seq_len_arr[u]=actual_bytes; - - /* Read in data until an entire sequence can't be read in any longer */ - while(io_left>0) { - /* Reset copy of number of blocks in fastest dimension */ - H5_ASSIGN_OVERFLOW(fast_dim_count,tdiminfo[fast_dim].count-tmp_count[fast_dim],hsize_t,size_t); - - /* Check if this entire row will fit into buffer */ - if(fast_dim_count<=tot_blk_count) { - - /* Entire row of blocks fits into buffer */ - act_blk_count=fast_dim_count; - - /* Loop over all the blocks in the fastest changing dimension */ - while(fast_dim_count>0) { - /* Gather the sequence */ - - /* Compute the number of sequences to fill */ - tot_seq=MIN(vector_size-nseq,fast_dim_count); - - /* Get a copy of the number of sequences to fill */ - seq_count=tot_seq; - - /* Set the pointer to the correct starting array element */ - buf_off_arr_p=&buf_off_arr[nseq]; - -#ifdef NO_DUFFS_DEVICE - /* Fill up the buffer, or finish up the blocks in this dimension */ - while(seq_count>0) { - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - seq_count--; - } /* end while */ -#else /* NO_DUFFS_DEVICE */ - duffs_index = (seq_count + 7) / 8; - switch (seq_count % 8) { - case 0: - do - { - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 7: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 6: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 5: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 4: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 3: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 2: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 1: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - } while (--duffs_index > 0); - } /* end switch */ -#endif /* NO_DUFFS_DEVICE */ - - /* Increment number of array elements used */ - nseq+=tot_seq; - - /* Increment the total number of bytes contained in arrays */ - tot_buf_size += tot_seq*actual_bytes; - - /* Decrement number of blocks left */ - fast_dim_count -= tot_seq; - - /* If the sequence & offset arrays are full, read them in */ - if(nseq>=vector_size) { - /* Read in the sequences */ - if (H5F_seq_readv(f, dxpl_id, layout, dc_plist, file_space, - elmt_size, nseq, seq_len_arr, buf_off_arr, buf/*out*/)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_READERROR, 0, "read error"); - - /* Increment the offset of the destination buffer */ - buf+=tot_buf_size; - - /* Reset the number of bytes & sequences */ - tot_buf_size=0; - nseq=0; - } /* end else */ - } /* end while */ - - /* Decrement number of elements left */ - io_left -= actual_read*act_blk_count; - - /* Decrement number of blocks left */ - tot_blk_count -= act_blk_count; - - /* Increment information to reflect block just processed */ - offset[fast_dim]=fast_dim_offset; /* reset the offset in the fastest dimension */ - tmp_count[fast_dim]=0; - - /* Increment offset in destination buffer */ - buf_off += wrap[fast_dim]; - } /* end if */ - else { - - /* Entire row of blocks doesn't fit into buffer */ - act_blk_count=tot_blk_count; - - /* Reduce number of blocks to output */ - fast_dim_count=tot_blk_count; - - /* Loop over all the blocks in the fastest changing dimension */ - while(fast_dim_count>0) { - /* Gather the sequence */ - - /* Compute the number of sequences to fill */ - tot_seq=MIN(vector_size-nseq,fast_dim_count); - - /* Get a copy of the number of sequences to fill */ - seq_count=tot_seq; - - /* Set the pointer to the correct starting array element */ - buf_off_arr_p=&buf_off_arr[nseq]; - - /* Fill up the buffer, or finish up the blocks in this dimension */ - while(seq_count>0) { - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - seq_count--; - } /* end while */ - - /* Increment number of array elements used */ - nseq+=tot_seq; - - /* Increment the total number of bytes contained in arrays */ - tot_buf_size += tot_seq*actual_bytes; - - /* Decrement number of blocks left */ - fast_dim_count -= tot_seq; - - /* If the sequence & offset arrays are full, read them in */ - if(nseq>=vector_size) { - /* Read in the sequences */ - if (H5F_seq_readv(f, dxpl_id, layout, dc_plist, file_space, - elmt_size, nseq, seq_len_arr, buf_off_arr, buf/*out*/)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_READERROR, 0, "read error"); - - /* Increment the offset of the destination buffer */ - buf+=tot_buf_size; - - /* Reset the number of bytes & sequences */ - tot_buf_size=0; - nseq=0; - } /* end else */ - } /* end while */ - - /* Decrement number of elements left */ - io_left -= actual_read*act_blk_count; - - /* Decrement number of blocks left */ - tot_blk_count -= act_blk_count; - - /* Increment information to reflect block just processed */ - offset[fast_dim]+=(fast_dim_stride*act_blk_count); /* reset the offset in the fastest dimension */ - tmp_count[fast_dim]+=act_blk_count; - - /* Handle any leftover, partial blocks in this row */ - if(io_left>0) { - actual_read=io_left; - actual_bytes=actual_read*elmt_size; - - /* Gather the sequence */ - - /* Store of length & offset */ - seq_len_arr[nseq]=actual_bytes; - buf_off_arr[nseq]=buf_off; - - /* Increment the total number of bytes contained in arrays */ - tot_buf_size += actual_bytes; - - /* Increment the number of sequences in arrays */ - nseq++; - - /* If the sequence & offset arrays are full, read them in */ - if(nseq>=vector_size) { - /* Read in the sequences */ - if (H5F_seq_readv(f, dxpl_id, layout, dc_plist, file_space, - elmt_size, nseq, seq_len_arr, buf_off_arr, buf/*out*/)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_READERROR, 0, "read error"); - - /* Increment the offset of the destination buffer */ - buf+=tot_buf_size; - - /* Reset the number of bytes & sequences */ - tot_buf_size=0; - nseq=0; - } /* end else */ - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - /* Decrement the number of elements left */ - io_left -= actual_read; - - /* Increment buffer correctly */ - offset[fast_dim]+=actual_read; - } /* end if */ - - /* don't bother checking slower dimensions */ - assert(tot_blk_count==0); - assert(io_left==0); - break; - } /* end else */ - - /* Increment the offset and count for the other dimensions */ - temp_dim=fast_dim-1; - while(temp_dim>=0) { - /* Move to the next row in the curent dimension */ - offset[temp_dim]++; - tmp_block[temp_dim]++; - - /* If this block is still in the range of blocks to output for the dimension, break out of loop */ - if(tmp_block[temp_dim]<tdiminfo[temp_dim].block) - break; - else { - /* Move to the next block in the current dimension */ - offset[temp_dim]+=(tdiminfo[temp_dim].stride-tdiminfo[temp_dim].block); - buf_off += skip[temp_dim]; - tmp_block[temp_dim]=0; - tmp_count[temp_dim]++; - - /* If this block is still in the range of blocks to output for the dimension, break out of loop */ - if(tmp_count[temp_dim]<tdiminfo[temp_dim].count) - break; - else { - offset[temp_dim]=tdiminfo[temp_dim].start+file_space->select.offset[temp_dim]; - buf_off += wrap[temp_dim]; - tmp_count[temp_dim]=0; /* reset back to the beginning of the line */ - tmp_block[temp_dim]=0; - } /* end else */ - } /* end else */ - - /* Decrement dimension count */ - temp_dim--; - } /* end while */ - } /* end while */ - - /* Check for any stored sequences which need to be flushed */ - if(nseq>0) { - /* Read in the sequence */ - if (H5F_seq_readv(f, dxpl_id, layout, dc_plist, file_space, - elmt_size, nseq, seq_len_arr, buf_off_arr, buf/*out*/)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_READERROR, 0, "read error"); - } /* end if */ - - /* Subtract out the selection offset */ - for(i=0; i<ndims; i++) - offset[i] -= file_space->select.offset[i]; - - /* Update the iterator with the location we stopped */ - HDmemcpy(file_iter->hyp.off, offset, ndims*sizeof(hssize_t)); - } /* end if */ - - /* Decrement the number of elements left in selection */ - file_iter->hyp.elmt_left -= (nelmts-io_left); - - /* Set the return value */ - ret_value= (nelmts-io_left); - -done: - if(seq_len_arr!=NULL) - H5FL_ARR_FREE(size_t,seq_len_arr); - if(buf_off_arr!=NULL) - H5FL_ARR_FREE(hsize_t,buf_off_arr); - - FUNC_LEAVE (ret_value); -} /* H5S_hyper_fread_opt() */ - - -/*------------------------------------------------------------------------- - * Function: H5S_hyper_fgath - * - * Purpose: Gathers data points from file F and accumulates them in the - * type conversion buffer BUF. The LAYOUT argument describes - * how the data is stored on disk and EFL describes how the data - * is organized in external files. ELMT_SIZE is the size in - * bytes of a datum which this function treats as opaque. - * FILE_SPACE describes the data space of the dataset on disk - * and the elements that have been selected for reading (via - * hyperslab, etc). This function will copy at most NELMTS elements. - * - * Return: Success: Number of elements copied. - * - * Failure: 0 - * - * Programmer: Quincey Koziol - * Tuesday, June 16, 1998 - * - * Modifications: - * Robb Matzke, 1999-08-03 - * The data transfer properties are passed by ID since that's - * what the virtual file layer needs. - *------------------------------------------------------------------------- - */ -static hsize_t -H5S_hyper_fgath (H5F_t *f, const struct H5O_layout_t *layout, - H5P_genplist_t *dc_plist, size_t elmt_size, - const H5S_t *file_space, H5S_sel_iter_t *file_iter, - hsize_t nelmts, hid_t dxpl_id, void *_buf/*out*/) -{ - hsize_t num_read=0; /* number of elements read into buffer */ - herr_t ret_value=SUCCEED; - - FUNC_ENTER_NOAPI(H5S_hyper_fgath, 0); - - /* Check args */ - assert (f); - assert (layout); - assert (elmt_size>0); - assert (file_space); - assert (file_iter); - assert (nelmts>0); - assert (_buf); - - /* Check for the special case of just one H5Sselect_hyperslab call made */ - if(file_space->select.sel_info.hslab.diminfo!=NULL) { - /* Use optimized call to read in regular hyperslab */ - num_read=H5S_hyper_fread_opt(f,layout,dc_plist,elmt_size,file_space,file_iter,nelmts,dxpl_id,_buf); - } /* end if */ - else { - /* Perform generic hyperslab operation */ - num_read=H5S_hyper_fread(f,layout,dc_plist,elmt_size,file_space,file_iter,nelmts,dxpl_id,_buf); - } /* end else */ - - FUNC_LEAVE (ret_value==SUCCEED ? num_read : 0); -} /* H5S_hyper_fgath() */ - - -/*------------------------------------------------------------------------- - * Function: H5S_hyper_fwrite - * - * Purpose: Performs an optimized scatter to the file, based on a hyperslab - * span selection. - * - * Return: Success: Number of elements copied. - * Failure: 0 - * - * Programmer: Quincey Koziol - * Tuesday, September 12, 2000 - * - * Modifications: - * - *------------------------------------------------------------------------- - */ -static hsize_t -H5S_hyper_fwrite (H5F_t *f, const struct H5O_layout_t *layout, - H5P_genplist_t *dc_plist, size_t elmt_size, - const H5S_t *space, H5S_sel_iter_t *iter, - hsize_t nelem, hid_t dxpl_id, const void *_buf) -{ - const uint8_t *src=(const uint8_t *)_buf; /* Alias for pointer arithmetic */ - H5S_hyper_span_t **ispan; /* Iterator's hyperslab span nodes */ - H5S_hyper_span_t *curr_span; /* Current hyperslab span node */ - hsize_t slab[H5O_LAYOUT_NDIMS]; /* Cumulative size of each dimension in bytes */ - hsize_t acc; /* Accumulator for computing cumulative sizes */ - hssize_t *abs_arr; /* Absolute hyperslab span position */ - hssize_t *off_arr; /* Offset within the dataspace extent */ - int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ - int curr_dim; /* Current dimension being operated on */ - int ndims; /* Number of dimensions of dataset */ - hsize_t loc_off; /* Element offset in the dataspace */ - size_t span_size=0; /* Number of bytes in current span to actually process */ - size_t io_bytes_left; /* Number of bytes left to process */ - int i; /* Index variable */ - size_t *seq_len_arr=NULL; /* Array of sequence lengths */ - hsize_t *buf_off_arr=NULL; /* Array of dataset offsets */ - size_t last_io_bytes_left=0; /* Last I/O bytes left before readv() called */ - size_t nseq=0; /* Number of sequence/offsets stored in the arrays */ - size_t vector_size; /* Value for vector size */ - H5P_genplist_t *plist; /* Property list */ - hssize_t ret_value=FAIL; - - FUNC_ENTER_NOINIT(H5S_hyper_fwrite); - - /* Check args */ - assert(f); - assert(layout); - assert(elmt_size>0); - assert(space); - assert(iter); - assert(nelem>0); - assert(src); - - /* Set the rank of the fastest changing dimension */ - ndims=space->extent.u.simple.rank; - fast_dim=(ndims-1); - - /* Get the pointers to the current span info and span nodes */ - curr_span=iter->hyp.span[fast_dim]; - abs_arr=iter->hyp.off; - off_arr=space->select.offset; - ispan=iter->hyp.span; - - /* Set the amount of elements to perform I/O on, etc. */ - H5_ASSIGN_OVERFLOW(io_bytes_left,(nelem*elmt_size),hsize_t,size_t); - - /* Compute the cumulative size of dataspace dimensions */ - for(i=fast_dim, acc=elmt_size; i>=0; i--) { - slab[i]=acc; - acc*=space->extent.u.simple.size[i]; - } /* end for */ - - /* Set the offset of the first element iterated on */ - for(i=0, loc_off=0; i<ndims; i++) - /* Compute the sequential element offset */ - loc_off+=(abs_arr[i]+space->select.offset[i])*slab[i]; - - /* Range check against number of elements left in selection */ - assert(io_bytes_left<=(iter->hyp.elmt_left*elmt_size)); - - /* Get the hyperslab vector size */ - if(TRUE!=H5P_isa_class(dxpl_id,H5P_DATASET_XFER) || NULL == (plist = H5I_object(dxpl_id))) - HGOTO_ERROR(H5E_ARGS, H5E_BADTYPE, 0, "not a file access property list"); - if (H5P_get(plist,H5D_XFER_HYPER_VECTOR_SIZE_NAME,&vector_size)<0) - HGOTO_ERROR(H5E_PLIST, H5E_CANTGET, 0, "unable to get value"); - - /* Allocate the vector I/O arrays */ - if((seq_len_arr = H5FL_ARR_ALLOC(size_t,vector_size,0))==NULL) - HGOTO_ERROR(H5E_RESOURCE, H5E_NOSPACE, 0, "can't allocate vector I/O array"); - if((buf_off_arr = H5FL_ARR_ALLOC(hsize_t,vector_size,0))==NULL) - HGOTO_ERROR(H5E_RESOURCE, H5E_NOSPACE, 0, "can't allocate vector I/O array"); - - /* Take care of any partial spans leftover from previous I/Os */ - if(abs_arr[fast_dim]!=curr_span->low) { - - /* Finish the span in the fastest changing dimension */ - - /* Compute the number of bytes to attempt in this span */ - H5_ASSIGN_OVERFLOW(span_size,((curr_span->high-abs_arr[fast_dim])+1)*elmt_size,hsize_t,size_t); - - /* Check number of bytes against upper bounds allowed */ - if(span_size>io_bytes_left) - span_size=io_bytes_left; - - if (H5F_seq_write(f, dxpl_id, layout, dc_plist, space, - elmt_size, span_size, loc_off, src)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_WRITEERROR, 0, "write error"); - - /* Increment offset in destination */ - src+=span_size; - - /* Decrement I/O left to perform */ - io_bytes_left-=span_size; - - /* Check if we are done */ - if(io_bytes_left>0) { - /* Move to next span in fastest changing dimension */ - curr_span=curr_span->next; - - if(curr_span!=NULL) { - /* Move location offset of destination */ - loc_off+=(curr_span->low-abs_arr[fast_dim])*elmt_size; - - /* Move iterator for fastest changing dimension */ - abs_arr[fast_dim]=curr_span->low; - } /* end if */ - } /* end if */ - else { - abs_arr[fast_dim]+=span_size/elmt_size; - - /* Check if we are still within the span */ - if(abs_arr[fast_dim]<=curr_span->high) { - iter->hyp.span[fast_dim]=curr_span; - - goto partial_done; /* finished with partial span */ - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[fast_dim]=curr_span->low; - iter->hyp.span[fast_dim]=curr_span; - - goto partial_done; /* finished with partial span */ - } /* end if */ - } /* end else */ - } /* end else */ - - /* Adjust iterator pointers */ - - if(curr_span==NULL) { -/* Same as code in main loop */ - /* Start at the next fastest dim */ - curr_dim=fast_dim-1; - - /* Work back up through the dimensions */ - while(curr_dim>=0) { - /* Reset the current span */ - curr_span=iter->hyp.span[curr_dim]; - - /* Increment absolute position */ - abs_arr[curr_dim]++; - - /* Check if we are still within the span */ - if(abs_arr[curr_dim]<=curr_span->high) { - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[curr_dim]=curr_span->low; - - break; - } /* end if */ - else { - /* If we finished the span list in this dimension, decrement the dimension worked on and loop again */ - curr_dim--; - } /* end else */ - } /* end else */ - } /* end while */ - - /* Check if we are finished with the spans in the tree */ - if(curr_dim<0) { - /* We had better be done with I/O or bad things are going to happen... */ - assert(io_bytes_left==0); - - goto partial_done; /* finished with partial span */ - } /* end if */ - else { - /* Reset the span in the current dimension */ - ispan[curr_dim]=curr_span; - - /* Walk back down the iterator positions, reseting them */ - while(curr_dim<fast_dim) { - assert(curr_span); - assert(curr_span->down); - assert(curr_span->down->head); - - /* Set the new span for this dimension */ - iter->hyp.span[curr_dim+1]=curr_span->down->head; - - /* Advance span down the tree */ - curr_span=curr_span->down->head; - - /* Reset the absolute offset for the dim */ - abs_arr[curr_dim+1]=curr_span->low; - - /* Increment current dimension */ - curr_dim++; - } /* end while */ - - /* Verify that the curr_span points to the fastest dim */ - assert(curr_span==iter->hyp.span[fast_dim]); - } /* end else */ - - /* Reset the buffer offset */ - for(i=0, loc_off=0; i<ndims; i++) - loc_off+=(abs_arr[i]+off_arr[i])*slab[i]; - } /* end if */ - } /* end if */ - -partial_done: /* Yes, goto's are evil, so sue me... :-) */ - /* Get the number of bytes left to process currently */ - last_io_bytes_left=io_bytes_left; - - /* Perform the I/O on the elements, based on the position of the iterator */ - while(io_bytes_left>0) { - /* Adjust location offset of destination to compensate for initial increment below */ - loc_off-=curr_span->pstride; - - /* Loop over all the spans in the fastest changing dimension */ - while(curr_span!=NULL) { - /* Move location offset of destination */ - loc_off+=curr_span->pstride; - - /* Compute the number of elements to attempt in this span */ - H5_ASSIGN_OVERFLOW(span_size,curr_span->nelem,hsize_t,size_t); - - /* Check number of elements against upper bounds allowed */ - if(span_size>=io_bytes_left) { - /* Trim the number of bytes to output */ - span_size=io_bytes_left; - io_bytes_left=0; - -/* COMMON */ - /* Store the I/O information for the span */ - seq_len_arr[nseq]=span_size; - buf_off_arr[nseq]=loc_off; - - /* Increment the number of sequences in arrays */ - nseq++; - - /* If the sequence & offset arrays are full, read them in */ - if(nseq>=vector_size) { - /* Write out the sequences */ - if (H5F_seq_writev(f, dxpl_id, layout, dc_plist, space, - elmt_size, nseq, seq_len_arr, buf_off_arr, src)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_WRITEERROR, 0, "write error"); - - /* Increment the offset of the destination buffer */ - src+=(last_io_bytes_left-io_bytes_left); - - /* Keep around the current number of I/O bytes left */ - last_io_bytes_left=io_bytes_left; - nseq=0; - } /* end else */ -/* end COMMON */ - - /* Break out now, we are finished with I/O */ - break; - } /* end if */ - else { - /* Decrement I/O left to perform */ - io_bytes_left-=span_size; - -/* COMMON */ - /* Store the I/O information for the span */ - seq_len_arr[nseq]=span_size; - buf_off_arr[nseq]=loc_off; - - /* Increment the number of sequences in arrays */ - nseq++; - - /* If the sequence & offset arrays are full, read them in */ - if(nseq>=vector_size) { - /* Write out the sequences */ - if (H5F_seq_writev(f, dxpl_id, layout, dc_plist, space, - elmt_size, nseq, seq_len_arr, buf_off_arr, src)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_WRITEERROR, 0, "write error"); - - /* Increment the offset of the destination buffer */ - src+=(last_io_bytes_left-io_bytes_left); - - /* Keep around the current number of I/O bytes left */ - last_io_bytes_left=io_bytes_left; - nseq=0; - } /* end else */ -/* end COMMON */ - } /* end else */ - - /* Move to next span in fastest changing dimension */ - curr_span=curr_span->next; - } /* end while */ - - /* Check if we are done */ - if(io_bytes_left==0) { - abs_arr[fast_dim]=curr_span->low+(span_size/elmt_size); - - /* Check if we are still within the span */ - if(abs_arr[fast_dim]<=curr_span->high) { - iter->hyp.span[fast_dim]=curr_span; - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[fast_dim]=curr_span->low; - iter->hyp.span[fast_dim]=curr_span; - break; - } /* end if */ - } /* end else */ - } /* end if */ - - /* Adjust iterator pointers */ - - /* Start at the next fastest dim */ - curr_dim=fast_dim-1; - - /* Work back up through the dimensions */ - while(curr_dim>=0) { - /* Reset the current span */ - curr_span=iter->hyp.span[curr_dim]; - - /* Increment absolute position */ - abs_arr[curr_dim]++; - - /* Check if we are still within the span */ - if(abs_arr[curr_dim]<=curr_span->high) { - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[curr_dim]=curr_span->low; - - break; - } /* end if */ - else { - /* If we finished the span list in this dimension, decrement the dimension worked on and loop again */ - curr_dim--; - } /* end else */ - } /* end else */ - } /* end while */ - - /* Check if we are finished with the spans in the tree */ - if(curr_dim<0) { - /* We had better be done with I/O or bad things are going to happen... */ - assert(io_bytes_left==0); - break; - } /* end if */ - else { - /* Reset the span in the current dimension */ - ispan[curr_dim]=curr_span; - - /* Walk back down the iterator positions, reseting them */ - while(curr_dim<fast_dim) { - assert(curr_span); - assert(curr_span->down); - assert(curr_span->down->head); - - /* Set the new span for the next dimension down */ - iter->hyp.span[curr_dim+1]=curr_span->down->head; - - /* Advance span down the tree */ - curr_span=curr_span->down->head; - - /* Reset the absolute offset for the dim */ - abs_arr[curr_dim+1]=curr_span->low; - - /* Increment current dimension */ - curr_dim++; - } /* end while */ - - /* Verify that the curr_span points to the fastest dim */ - assert(curr_span==iter->hyp.span[fast_dim]); - } /* end else */ - - /* Reset the buffer offset */ - for(i=0, loc_off=0; i<ndims; i++) - loc_off+=(abs_arr[i]+off_arr[i])*slab[i]; - } /* end while */ - - /* Check for any stored sequences which need to be flushed */ - if(nseq>0) { - /* Write out the sequence */ - if (H5F_seq_writev(f, dxpl_id, layout, dc_plist, space, - elmt_size, nseq, seq_len_arr, buf_off_arr, src)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_WRITEERROR, 0, "write error"); - } /* end if */ - - /* Increment amount of I/O performed */ - iter->hyp.elmt_left-=nelem; - - /* Success! */ - ret_value=nelem; - -done: - if(seq_len_arr!=NULL) - H5FL_ARR_FREE(size_t,seq_len_arr); - if(buf_off_arr!=NULL) - H5FL_ARR_FREE(hsize_t,buf_off_arr); - - FUNC_LEAVE (ret_value); -} /* H5S_hyper_fwrite() */ - - -/*------------------------------------------------------------------------- - * Function: H5S_hyper_fwrite_opt - * - * Purpose: Performs an optimized scatter to the file, based on a regular - * hyperslab (i.e. one which was generated from just one call to - * H5Sselect_hyperslab). - * - * Return: Success: Number of elements copied. - * Failure: 0 - * - * Programmer: Quincey Koziol - * Friday, July 6, 2001 - * - * Modifications: - * - *------------------------------------------------------------------------- - */ -static hsize_t -H5S_hyper_fwrite_opt (H5F_t *f, const struct H5O_layout_t *layout, - H5P_genplist_t *dc_plist, size_t elmt_size, - const H5S_t *file_space, H5S_sel_iter_t *file_iter, - hsize_t nelmts, hid_t dxpl_id, const void *_buf) -{ - size_t *seq_len_arr=NULL; /* Array of sequence lengths */ - hsize_t *buf_off_arr=NULL; /* Array of dataset offsets */ - size_t nseq=0; /* Number of sequence/offsets stored in the arrays */ - size_t tot_buf_size=0; /* Total number of bytes in buffer */ - - hssize_t offset[H5O_LAYOUT_NDIMS]; /* Offset on disk */ - hsize_t slab[H5O_LAYOUT_NDIMS]; /* Hyperslab size */ - hssize_t wrap[H5O_LAYOUT_NDIMS]; /* Bytes to wrap around at the end of a row */ - hsize_t skip[H5O_LAYOUT_NDIMS]; /* Bytes to skip between blocks */ - hsize_t tmp_count[H5O_LAYOUT_NDIMS]; /* Temporary block count */ - hsize_t tmp_block[H5O_LAYOUT_NDIMS]; /* Temporary block offset */ - const uint8_t *buf=_buf; /* Alias for pointer arithmetic */ - const H5S_hyper_dim_t *tdiminfo; /* Local pointer to diminfo information */ - hssize_t fast_dim_start, /* Local copies of fastest changing dimension info */ - fast_dim_offset; - hsize_t fast_dim_stride, /* Local copies of fastest changing dimension info */ - fast_dim_block, - fast_dim_buf_off; - size_t fast_dim_count; - size_t tot_blk_count; /* Total number of blocks left to output */ - size_t act_blk_count; /* Actual number of blocks to output */ - int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ - int temp_dim; /* Temporary rank holder */ - hsize_t acc; /* Accumulator */ - hsize_t buf_off; /* Current buffer offset for copying memory */ - int i; /* Counters */ - unsigned u; /* Counters */ - int ndims; /* Number of dimensions of dataset */ - size_t actual_write; /* The actual number of elements to write out */ - size_t actual_bytes; /* The actual number of bytes to copy */ - size_t io_left; /* The number of elements left in I/O operation */ - size_t tot_seq; /* The number of sequences filled */ - hsize_t *buf_off_arr_p; /* Pointer into the buffer offset array */ - size_t seq_count; /* Temporary count of sequences left to process */ -#ifndef NO_DUFFS_DEVICE - size_t duffs_index; /* Counting index for Duff's device */ -#endif /* NO_DUFFS_DEVICE */ - size_t vector_size; /* Value for vector size */ - H5P_genplist_t *plist; /* Property list */ - hsize_t ret_value=0; /* Return value */ - - FUNC_ENTER_NOINIT(H5S_hyper_fwrite_opt); - - /* Get the hyperslab vector size */ - if(TRUE!=H5P_isa_class(dxpl_id,H5P_DATASET_XFER) || NULL == (plist = H5I_object(dxpl_id))) - HGOTO_ERROR(H5E_ARGS, H5E_BADTYPE, 0, "not a file access property list"); - if (H5P_get(plist,H5D_XFER_HYPER_VECTOR_SIZE_NAME,&vector_size)<0) - HGOTO_ERROR(H5E_PLIST, H5E_CANTGET, 0, "unable to get value"); - - /* Allocate the vector I/O arrays */ - if((seq_len_arr = H5FL_ARR_ALLOC(size_t,vector_size,0))==NULL) - HGOTO_ERROR(H5E_RESOURCE, H5E_NOSPACE, 0, "can't allocate vector I/O array"); - if((buf_off_arr = H5FL_ARR_ALLOC(hsize_t,vector_size,0))==NULL) - HGOTO_ERROR(H5E_RESOURCE, H5E_NOSPACE, 0, "can't allocate vector I/O array"); - - /* Set the rank of the fastest changing dimension */ - fast_dim=file_space->extent.u.simple.rank-1; - ndims=file_space->extent.u.simple.rank; - - /* initialize row sizes for each dimension */ - for(i=(ndims-1),acc=1; i>=0; i--) { - slab[i]=acc*elmt_size; - acc*=file_space->extent.u.simple.size[i]; - } /* end for */ - - /* Set the number of elements left for I/O */ - H5_ASSIGN_OVERFLOW(io_left,nelmts,hsize_t,size_t); - - /* Check if we stopped in the middle of a sequence of elements */ - if((file_iter->hyp.off[fast_dim]-file_space->select.sel_info.hslab.diminfo[fast_dim].start)%file_space->select.sel_info.hslab.diminfo[fast_dim].stride!=0 || - ((file_iter->hyp.off[fast_dim]!=file_space->select.sel_info.hslab.diminfo[fast_dim].start) && file_space->select.sel_info.hslab.diminfo[fast_dim].stride==1)) { - hsize_t leftover; /* The number of elements left over from the last sequence */ - - /* Calculate the number of elements left in the sequence */ - if(file_space->select.sel_info.hslab.diminfo[fast_dim].stride==1) - leftover=file_space->select.sel_info.hslab.diminfo[fast_dim].block-(file_iter->hyp.off[fast_dim]-file_space->select.sel_info.hslab.diminfo[fast_dim].start); - else - leftover=file_space->select.sel_info.hslab.diminfo[fast_dim].block-((file_iter->hyp.off[fast_dim]-file_space->select.sel_info.hslab.diminfo[fast_dim].start)%file_space->select.sel_info.hslab.diminfo[fast_dim].stride); - - /* Make certain that we don't write too many */ - H5_CHECK_OVERFLOW(leftover,hsize_t,size_t); - actual_write=MIN((size_t)leftover,io_left); - actual_bytes=actual_write*elmt_size; - - /* Copy the location of the point to get */ - HDmemcpy(offset, file_iter->hyp.off,ndims*sizeof(hssize_t)); - offset[ndims] = 0; - - /* Add in the selection offset */ - for(i=0; i<ndims; i++) - offset[i] += file_space->select.offset[i]; - - /* Compute the initial buffer offset */ - for(i=0,buf_off=0; i<ndims; i++) - buf_off+=offset[i]*slab[i]; - - /* Write out the rest of the sequence */ - if (H5F_seq_write(f, dxpl_id, layout, dc_plist, file_space, - elmt_size, actual_bytes, buf_off, buf)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_WRITEERROR, 0, "write error"); - - /* Increment the offset of the buffer */ - buf+=actual_bytes; - - /* Decrement the amount left to write */ - io_left-=actual_write; - - /* Advance the point iterator */ - /* If we had enough buffer space to write out the rest of the sequence - * in the fastest changing dimension, move the iterator offset to - * the beginning of the next block to write. Otherwise, just advance - * the iterator in the fastest changing dimension. - */ - if(actual_write==leftover) { - /* Move iterator offset to beginning of next sequence in the fastest changing dimension */ - H5S_hyper_iter_next(file_space,file_iter); - } /* end if */ - else { - file_iter->hyp.off[fast_dim]+=actual_write; /* whole sequence not written out, just advance fastest dimension offset */ - } /* end if */ - } /* end if */ - - /* Now that we've cleared the "remainder" of the previous fastest dimension - * sequence, we must be at the beginning of a sequence, so use the fancy - * algorithm to compute the offsets and run through as many as possible, - * until the buffer runs dry. - */ - if(io_left>0) { /* Just in case the "remainder" above emptied the buffer */ - /* Compute the arrays to perform I/O on */ - /* Copy the location of the point to get */ - HDmemcpy(offset, file_iter->hyp.off,ndims*sizeof(hssize_t)); - offset[ndims] = 0; - - /* Add in the selection offset */ - for(i=0; i<ndims; i++) - offset[i] += file_space->select.offset[i]; - - /* Compute the current "counts" for this location */ - for(i=0; i<ndims; i++) { - if(file_space->select.sel_info.hslab.diminfo[i].stride==1) { - tmp_count[i] = 0; - tmp_block[i] = file_iter->hyp.off[i]-file_space->select.sel_info.hslab.diminfo[i].start; - } /* end if */ - else { - tmp_count[i] = (file_iter->hyp.off[i]-file_space->select.sel_info.hslab.diminfo[i].start)/file_space->select.sel_info.hslab.diminfo[i].stride; - tmp_block[i] = (file_iter->hyp.off[i]-file_space->select.sel_info.hslab.diminfo[i].start)%file_space->select.sel_info.hslab.diminfo[i].stride; - } /* end else */ - } /* end for */ - - /* Compute the initial buffer offset */ - for(i=0,buf_off=0; i<ndims; i++) - buf_off+=offset[i]*slab[i]; - - /* Set the number of elements to write each time */ - H5_ASSIGN_OVERFLOW(actual_write,file_space->select.sel_info.hslab.diminfo[fast_dim].block,hsize_t,size_t); - - /* Set the number of actual bytes */ - actual_bytes=actual_write*elmt_size; - - /* Set the local copy of the diminfo pointer */ - tdiminfo=file_space->select.sel_info.hslab.diminfo; - - /* Set local copies of information for the fastest changing dimension */ - fast_dim_start=tdiminfo[fast_dim].start; - fast_dim_stride=tdiminfo[fast_dim].stride; - fast_dim_block=tdiminfo[fast_dim].block; - fast_dim_buf_off=slab[fast_dim]*fast_dim_stride; - fast_dim_offset=fast_dim_start+file_space->select.offset[fast_dim]; - - /* Compute the number of blocks which would fit into the buffer */ - H5_ASSIGN_OVERFLOW(tot_blk_count,(io_left/fast_dim_block),hsize_t,size_t); - - /* Compute the amount to wrap at the end of each row */ - for(i=0; i<ndims; i++) - wrap[i]=(file_space->extent.u.simple.size[i]-(tdiminfo[i].stride*tdiminfo[i].count))*slab[i]; - - /* Compute the amount to skip between blocks */ - for(i=0; i<ndims; i++) - skip[i]=(tdiminfo[i].stride-tdiminfo[i].block)*slab[i]; - - /* Fill the sequence length array (since they will all be the same for optimized hyperslabs) */ - for(u=0; u<vector_size; u++) - seq_len_arr[u]=actual_bytes; - - /* Write out data until an entire sequence can't be written any longer */ - while(io_left>0) { - /* Reset copy of number of blocks in fastest dimension */ - H5_ASSIGN_OVERFLOW(fast_dim_count,tdiminfo[fast_dim].count-tmp_count[fast_dim],hsize_t,size_t); - - /* Check if this entire row will fit into buffer */ - if(fast_dim_count<=tot_blk_count) { - - /* Entire row of blocks fits into buffer */ - act_blk_count=fast_dim_count; - - /* Loop over all the blocks in the fastest changing dimension */ - while(fast_dim_count>0) { - /* Gather the sequence */ - - /* Compute the number of sequences to fill */ - tot_seq=MIN(vector_size-nseq,fast_dim_count); - - /* Get a copy of the number of sequences to fill */ - seq_count=tot_seq; - - /* Set the pointer to the correct starting array element */ - buf_off_arr_p=&buf_off_arr[nseq]; - -#ifdef NO_DUFFS_DEVICE - /* Fill up the buffer, or finish up the blocks in this dimension */ - while(seq_count>0) { - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - seq_count--; - } /* end while */ -#else /* NO_DUFFS_DEVICE */ - duffs_index = (seq_count + 7) / 8; - switch (seq_count % 8) { - case 0: - do - { - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 7: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 6: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 5: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 4: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 3: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 2: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - case 1: - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - } while (--duffs_index > 0); - } /* end switch */ -#endif /* NO_DUFFS_DEVICE */ - - /* Increment number of array elements used */ - nseq+=tot_seq; - - /* Increment the total number of bytes contained in arrays */ - tot_buf_size += tot_seq*actual_bytes; - - /* Decrement number of blocks left */ - fast_dim_count -= tot_seq; - - /* If the sequence & offset arrays are full, write them out */ - if(nseq>=vector_size) { - /* Write out the sequences */ - if (H5F_seq_writev(f, dxpl_id, layout, dc_plist, file_space, - elmt_size, nseq, seq_len_arr, buf_off_arr, buf)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_WRITEERROR, 0, "write error"); - - /* Increment the offset of the destination buffer */ - buf+=tot_buf_size; - - /* Reset the number of bytes & sequences */ - tot_buf_size=0; - nseq=0; - } /* end else */ - } /* end while */ - - /* Decrement number of elements left */ - io_left -= actual_write*act_blk_count; - - /* Decrement number of blocks left */ - tot_blk_count -= act_blk_count; - - /* Increment information to reflect block just processed */ - offset[fast_dim]=fast_dim_offset; /* reset the offset in the fastest dimension */ - tmp_count[fast_dim]=0; - - /* Increment offset in destination buffer */ - buf_off += wrap[fast_dim]; - } /* end if */ - else { - - /* Entire row of blocks doesn't fit into buffer */ - act_blk_count=tot_blk_count; - - /* Reduce number of blocks to output */ - fast_dim_count=tot_blk_count; - - /* Loop over all the blocks in the fastest changing dimension */ - while(fast_dim_count>0) { - /* Gather the sequence */ - - /* Compute the number of sequences to fill */ - tot_seq=MIN(vector_size-nseq,fast_dim_count); - - /* Get a copy of the number of sequences to fill */ - seq_count=tot_seq; - - /* Set the pointer to the correct starting array element */ - buf_off_arr_p=&buf_off_arr[nseq]; - - /* Fill up the buffer, or finish up the blocks in this dimension */ - while(seq_count>0) { - /* Store of length & offset */ - /* seq_len_arr[nseq] already has the correct value */ - *buf_off_arr_p++=buf_off; - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - seq_count--; - } /* end while */ - - /* Increment number of array elements used */ - nseq+=tot_seq; - - /* Increment the total number of bytes contained in arrays */ - tot_buf_size += tot_seq*actual_bytes; - - /* Decrement number of blocks left */ - fast_dim_count -= tot_seq; - - /* If the sequence & offset arrays are full, write them out */ - if(nseq>=vector_size) { - /* Write out the sequences */ - if (H5F_seq_writev(f, dxpl_id, layout, dc_plist, file_space, - elmt_size, nseq, seq_len_arr, buf_off_arr, buf)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_WRITEERROR, 0, "write error"); - - /* Increment the offset of the destination buffer */ - buf+=tot_buf_size; - - /* Reset the number of bytes & sequences */ - tot_buf_size=0; - nseq=0; - } /* end else */ - } /* end while */ - - /* Decrement number of elements left */ - io_left -= actual_write*act_blk_count; - - /* Decrement number of blocks left */ - tot_blk_count -= act_blk_count; - - /* Increment information to reflect block just processed */ - offset[fast_dim]+=(fast_dim_stride*act_blk_count); /* reset the offset in the fastest dimension */ - tmp_count[fast_dim]+=act_blk_count; - - /* Handle any leftover, partial blocks in this row */ - if(io_left>0) { - actual_write=io_left; - actual_bytes=actual_write*elmt_size; - - /* Gather the sequence */ - - /* Store of length & offset */ - seq_len_arr[nseq]=actual_bytes; - buf_off_arr[nseq]=buf_off; - - /* Increment the total number of bytes contained in arrays */ - tot_buf_size += actual_bytes; - - /* Increment the number of sequences in arrays */ - nseq++; - - /* If the sequence & offset arrays are full, write them out */ - if(nseq>=vector_size) { - /* Write out the sequences */ - if (H5F_seq_writev(f, dxpl_id, layout, dc_plist, file_space, - elmt_size, nseq, seq_len_arr, buf_off_arr, buf)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_WRITEERROR, 0, "write error"); - - /* Increment the offset of the destination buffer */ - buf+=tot_buf_size; - - /* Reset the number of bytes & sequences */ - tot_buf_size=0; - nseq=0; - } /* end else */ - - /* Increment the source offset */ - buf_off+=fast_dim_buf_off; - - /* Decrement the number of elements left */ - io_left -= actual_write; - - /* Increment buffer correctly */ - offset[fast_dim]+=actual_write; - } /* end if */ - - /* don't bother checking slower dimensions */ - assert(tot_blk_count==0); - assert(io_left==0); - break; - } /* end else */ - - /* Increment the offset and count for the other dimensions */ - temp_dim=fast_dim-1; - while(temp_dim>=0) { - /* Move to the next row in the curent dimension */ - offset[temp_dim]++; - tmp_block[temp_dim]++; - - /* If this block is still in the range of blocks to output for the dimension, break out of loop */ - if(tmp_block[temp_dim]<tdiminfo[temp_dim].block) - break; - else { - /* Move to the next block in the current dimension */ - offset[temp_dim]+=(tdiminfo[temp_dim].stride-tdiminfo[temp_dim].block); - buf_off += skip[temp_dim]; - tmp_block[temp_dim]=0; - tmp_count[temp_dim]++; - - /* If this block is still in the range of blocks to output for the dimension, break out of loop */ - if(tmp_count[temp_dim]<tdiminfo[temp_dim].count) - break; - else { - offset[temp_dim]=tdiminfo[temp_dim].start+file_space->select.offset[temp_dim]; - buf_off += wrap[temp_dim]; - tmp_count[temp_dim]=0; /* reset back to the beginning of the line */ - tmp_block[temp_dim]=0; - } /* end else */ - } /* end else */ - - /* Decrement dimension count */ - temp_dim--; - } /* end while */ - } /* end while */ - - /* Check for any stored sequences which need to be flushed */ - if(nseq>0) { - /* Write out the sequence */ - if (H5F_seq_writev(f, dxpl_id, layout, dc_plist, file_space, - elmt_size, nseq, seq_len_arr, buf_off_arr, buf)<0) - HGOTO_ERROR(H5E_DATASPACE, H5E_WRITEERROR, 0, "write error"); - } /* end if */ - - /* Subtract out the selection offset */ - for(i=0; i<ndims; i++) - offset[i] -= file_space->select.offset[i]; - - /* Update the iterator with the location we stopped */ - HDmemcpy(file_iter->hyp.off, offset, ndims*sizeof(hssize_t)); - } /* end if */ - - /* Decrement the number of elements left in selection */ - file_iter->hyp.elmt_left -= (nelmts-io_left); - - ret_value= (nelmts-io_left); - -done: - if(seq_len_arr!=NULL) - H5FL_ARR_FREE(size_t,seq_len_arr); - if(buf_off_arr!=NULL) - H5FL_ARR_FREE(hsize_t,buf_off_arr); - - FUNC_LEAVE (ret_value); -} /* H5S_hyper_fwrite_opt() */ - - -/*------------------------------------------------------------------------- - * Function: H5S_hyper_fscat - * - * Purpose: Scatters dataset elements from the type conversion buffer BUF - * to the file F where the data points are arranged according to - * the file data space FILE_SPACE and stored according to - * LAYOUT and EFL. Each element is ELMT_SIZE bytes. - * The caller is requesting that NELMTS elements are copied. - * - * Return: Non-negative on success/Negative on failure - * - * Programmer: Quincey Koziol - * Tuesday, June 16, 1998 - * - * Modifications: - * Robb Matzke, 1999-08-03 - * The data transfer properties are passed by ID since that's - * what the virtual file layer needs. - *------------------------------------------------------------------------- - */ -static herr_t -H5S_hyper_fscat (H5F_t *f, const struct H5O_layout_t *layout, - H5P_genplist_t *dc_plist, size_t elmt_size, - const H5S_t *file_space, H5S_sel_iter_t *file_iter, - hsize_t nelmts, hid_t dxpl_id, const void *_buf) -{ - hsize_t num_written=0; /* number of elements read into buffer */ - herr_t ret_value=SUCCEED; - - FUNC_ENTER_NOAPI(H5S_hyper_fscat, 0); - - /* Check args */ - assert (f); - assert (layout); - assert (elmt_size>0); - assert (file_space); - assert (file_iter); - assert (nelmts>0); - assert (_buf); - - /* Check for the special case of just one H5Sselect_hyperslab call made */ - if(file_space->select.sel_info.hslab.diminfo!=NULL) { - /* Use optimized call to write out regular hyperslab */ - num_written=H5S_hyper_fwrite_opt(f,layout,dc_plist,elmt_size,file_space,file_iter,nelmts,dxpl_id,_buf); - } /* end if */ - else { - /* Perform generic hyperslab operation */ - num_written=H5S_hyper_fwrite(f,layout,dc_plist,elmt_size,file_space,file_iter,nelmts,dxpl_id,_buf); - } /* end else */ - - FUNC_LEAVE (ret_value==FAIL ? ret_value : (num_written >0) ? SUCCEED : FAIL); -} /* H5S_hyper_fscat() */ - - -/*------------------------------------------------------------------------- - * Function: H5S_hyper_mread - * - * Purpose: Performs an optimized gather from a memory buffer, based on a - * hyperslab span selection. - * - * Return: Success: Number of elements copied. - * Failure: 0 - * - * Programmer: Quincey Koziol - * Tuesday, September 12, 2000 - * - * Modifications: - * - *------------------------------------------------------------------------- - */ -static hssize_t -H5S_hyper_mread (const void *_buf, size_t elmt_size, const H5S_t *space, - H5S_sel_iter_t *iter, hsize_t nelem, void *_tconv_buf/*out*/) -{ - const uint8_t *src=(const uint8_t *)_buf; /* Alias for pointer arithmetic */ - const uint8_t *tmp_src; /* Alias for pointer arithmetic */ - uint8_t *dst=(uint8_t *)_tconv_buf; /* Alias for pointer arithmetic */ - H5S_hyper_span_t **ispan; /* Iterator's hyperslab span nodes */ - H5S_hyper_span_t *curr_span; /* Current hyperslab span node */ - hsize_t slab[H5O_LAYOUT_NDIMS]; /* Cumulative size of each dimension in bytes */ - hsize_t acc; /* Accumulator for computing cumulative sizes */ - hssize_t *abs_arr; /* Absolute hyperslab span position */ - hssize_t *off_arr; /* Offset within the dataspace extent */ - int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ - int curr_dim; /* Current dimension being operated on */ - int ndims; /* Number of dimensions of dataset */ - size_t span_size=0; /* Number of bytes in current span to actually process */ - size_t io_bytes_left; /* Number of bytes left to process */ - int i; /* Index variable */ - hssize_t ret_value=FAIL; - - FUNC_ENTER_NOINIT(H5S_hyper_mread); - - /* Check args */ - assert(src); - assert(elmt_size>0); - assert(space); - assert(iter); - assert(nelem>0); - assert(dst); - - /* Set the rank of the fastest changing dimension */ - ndims=space->extent.u.simple.rank; - fast_dim=(ndims-1); - - /* Get the pointers to the current span info and span nodes */ - curr_span=iter->hyp.span[fast_dim]; - abs_arr=iter->hyp.off; - off_arr=space->select.offset; - ispan=iter->hyp.span; - - /* Set the amount of elements to perform I/O on, etc. */ - H5_ASSIGN_OVERFLOW(io_bytes_left,nelem*elmt_size,hsize_t,size_t); - - /* Compute the cumulative size of dataspace dimensions */ - for(i=fast_dim, acc=elmt_size; i>=0; i--) { - slab[i]=acc; - acc*=space->extent.u.simple.size[i]; - } /* end for */ - - /* Set the offset of the first element iterated on */ - for(i=0, tmp_src=src; i<ndims; i++) - /* Compute the sequential element offset */ - tmp_src+=(abs_arr[i]+space->select.offset[i])*slab[i]; - - /* Range check against number of elements left in selection */ - assert(io_bytes_left<=(iter->hyp.elmt_left*elmt_size)); - - /* Take care of any partial spans leftover from previous I/Os */ - if(abs_arr[fast_dim]!=curr_span->low) { - - /* Finish the span in the fastest changing dimension */ - - /* Compute the number of bytes to attempt in this span */ - H5_ASSIGN_OVERFLOW(span_size,((curr_span->high-abs_arr[fast_dim])+1)*elmt_size,hsize_t,size_t); - - /* Check number of elements against upper bounds allowed */ - if(span_size>io_bytes_left) - span_size=io_bytes_left; - - HDmemcpy(dst,tmp_src,span_size); - - /* Increment offset in destination */ - dst+=span_size; - - /* Decrement I/O left to perform */ - io_bytes_left-=span_size; - - /* Check if we are done */ - if(io_bytes_left>0) { - /* Move to next span in fastest changing dimension */ - curr_span=curr_span->next; - - if(curr_span!=NULL) { - /* Move location offset of destination */ - tmp_src+=(curr_span->low-abs_arr[fast_dim])*elmt_size; - - /* Move iterator for fastest changing dimension */ - abs_arr[fast_dim]=curr_span->low; - } /* end if */ - } /* end if */ - else { - abs_arr[fast_dim]+=span_size/elmt_size; - - /* Check if we are still within the span */ - if(abs_arr[fast_dim]<=curr_span->high) { - iter->hyp.span[fast_dim]=curr_span; - - goto partial_done; /* finished with partial span */ - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[fast_dim]=curr_span->low; - iter->hyp.span[fast_dim]=curr_span; - - goto partial_done; /* finished with partial span */ - } /* end if */ - } /* end else */ - } /* end else */ - - /* Adjust iterator pointers */ - - if(curr_span==NULL) { -/* Same as code in main loop */ - /* Start at the next fastest dim */ - curr_dim=fast_dim-1; - - /* Work back up through the dimensions */ - while(curr_dim>=0) { - /* Reset the current span */ - curr_span=iter->hyp.span[curr_dim]; - - /* Increment absolute position */ - abs_arr[curr_dim]++; - - /* Check if we are still within the span */ - if(abs_arr[curr_dim]<=curr_span->high) { - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[curr_dim]=curr_span->low; - - break; - } /* end if */ - else { - /* If we finished the span list in this dimension, decrement the dimension worked on and loop again */ - curr_dim--; - } /* end else */ - } /* end else */ - } /* end while */ - - /* Check if we are finished with the spans in the tree */ - if(curr_dim<0) { - /* We had better be done with I/O or bad things are going to happen... */ - assert(io_bytes_left==0); - - goto partial_done; /* finished with partial span */ - } /* end if */ - else { - /* Reset the span in the current dimension */ - ispan[curr_dim]=curr_span; - - /* Walk back down the iterator positions, reseting them */ - while(curr_dim<fast_dim) { - assert(curr_span); - assert(curr_span->down); - assert(curr_span->down->head); - - /* Set the new span for this dimension */ - iter->hyp.span[curr_dim+1]=curr_span->down->head; - - /* Advance span down the tree */ - curr_span=curr_span->down->head; - - /* Reset the absolute offset for the dim */ - abs_arr[curr_dim+1]=curr_span->low; - - /* Increment current dimension */ - curr_dim++; - } /* end while */ - - /* Verify that the curr_span points to the fastest dim */ - assert(curr_span==iter->hyp.span[fast_dim]); - } /* end else */ - - /* Reset the buffer offset */ - for(i=0, tmp_src=src; i<ndims; i++) - tmp_src+=(abs_arr[i]+off_arr[i])*slab[i]; - } /* end if */ - } /* end if */ - -partial_done: /* Yes, goto's are evil, so sue me... :-) */ - - /* Perform the I/O on the elements, based on the position of the iterator */ - while(io_bytes_left>0) { - /* Adjust buffer offset of source to compensate for initial increment below */ - tmp_src-=curr_span->pstride; - - /* Loop over all the spans in the fastest changing dimension */ - while(curr_span!=NULL) { - /* Move buffer offset of source */ - tmp_src+=curr_span->pstride; - - /* Compute the number of elements to attempt in this span */ - H5_ASSIGN_OVERFLOW(span_size,curr_span->nelem,hsize_t,size_t); - - /* Check number of elements against upper bounds allowed */ - if(span_size>=io_bytes_left) { - /* Trim the number of bytes to output */ - span_size=io_bytes_left; - io_bytes_left=0; - -/* COMMON */ - /* "Read" the data from the source buffer */ - HDmemcpy(dst,tmp_src,span_size); - - /* Increment offset in destination */ - dst+=span_size; -/* end COMMON */ - - /* Break out now, we are finished with I/O */ - break; - } /* end if */ - else { - /* Decrement I/O left to perform */ - io_bytes_left-=span_size; - -/* COMMON */ - /* "Read" the data from the source buffer */ - HDmemcpy(dst,tmp_src,span_size); - - /* Increment offset in destination */ - dst+=span_size; -/* end COMMON */ - } /* end else */ - - /* Move to next span in fastest changing dimension */ - curr_span=curr_span->next; - } /* end while */ - - /* Check if we are done */ - if(io_bytes_left==0) { - abs_arr[fast_dim]=curr_span->low+(span_size/elmt_size); - - /* Check if we are still within the span */ - if(abs_arr[fast_dim]<=curr_span->high) { - iter->hyp.span[fast_dim]=curr_span; - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[fast_dim]=curr_span->low; - iter->hyp.span[fast_dim]=curr_span; - break; - } /* end if */ - } /* end else */ - } /* end if */ - - /* Adjust iterator pointers */ - - /* Start at the next fastest dim */ - curr_dim=fast_dim-1; - - /* Work back up through the dimensions */ - while(curr_dim>=0) { - /* Reset the current span */ - curr_span=iter->hyp.span[curr_dim]; - - /* Increment absolute position */ - abs_arr[curr_dim]++; - - /* Check if we are still within the span */ - if(abs_arr[curr_dim]<=curr_span->high) { - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[curr_dim]=curr_span->low; - - break; - } /* end if */ - else { - /* If we finished the span list in this dimension, decrement the dimension worked on and loop again */ - curr_dim--; - } /* end else */ - } /* end else */ - } /* end while */ - - /* Check if we are finished with the spans in the tree */ - if(curr_dim<0) { - /* We had better be done with I/O or bad things are going to happen... */ - assert(io_bytes_left==0); - break; - } /* end if */ - else { - /* Reset the span in the current dimension */ - ispan[curr_dim]=curr_span; - - /* Walk back down the iterator positions, reseting them */ - while(curr_dim<fast_dim) { - assert(curr_span); - assert(curr_span->down); - assert(curr_span->down->head); - - /* Set the new span for the next dimension down */ - iter->hyp.span[curr_dim+1]=curr_span->down->head; - - /* Advance span down the tree */ - curr_span=curr_span->down->head; - - /* Reset the absolute offset for the dim */ - abs_arr[curr_dim+1]=curr_span->low; - - /* Increment current dimension */ - curr_dim++; - } /* end while */ - - /* Verify that the curr_span points to the fastest dim */ - assert(curr_span==iter->hyp.span[fast_dim]); - } /* end else */ - - /* Reset the buffer offset */ - for(i=0, tmp_src=src; i<ndims; i++) - tmp_src+=(abs_arr[i]+off_arr[i])*slab[i]; - } /* end while */ - - /* Increment amount of I/O performed */ - iter->hyp.elmt_left-=nelem; - - /* Success! */ - ret_value=nelem; - -#ifdef LATER -done: -#endif /* LATER */ - FUNC_LEAVE (ret_value); -} /* end H5S_hyper_mread() */ - - -/*------------------------------------------------------------------------- - * Function: H5S_hyper_mread_opt - * - * Purpose: Performs an optimized gather from a memory buffer, based on a - * regular hyperslab (i.e. one which was generated from just one call to - * H5Sselect_hyperslab). - * - * Return: Success: Number of elements copied. - * Failure: 0 - * - * Programmer: Quincey Koziol - * Tuesday, September 12, 2000 - * - * Modifications: - * - *------------------------------------------------------------------------- - */ -static hsize_t -H5S_hyper_mread_opt (const void *_buf, size_t elmt_size, - const H5S_t *mem_space, H5S_sel_iter_t *mem_iter, - hsize_t nelmts, void *_tconv_buf/*out*/) -{ - hsize_t mem_size[H5O_LAYOUT_NDIMS]; /* Size of the source buffer */ - hsize_t slab[H5O_LAYOUT_NDIMS]; /* Hyperslab size */ - hssize_t wrap[H5O_LAYOUT_NDIMS]; /* Bytes to wrap around at the end of a row */ - hsize_t skip[H5O_LAYOUT_NDIMS]; /* Bytes to skip between blocks */ - hssize_t offset[H5O_LAYOUT_NDIMS]; /* Offset on disk */ - hsize_t tmp_count[H5O_LAYOUT_NDIMS]; /* Temporary block count */ - hsize_t tmp_block[H5O_LAYOUT_NDIMS]; /* Temporary block offset */ - const uint8_t *src=(const uint8_t *)_buf; /* Alias for pointer arithmetic */ - uint8_t *dst=(uint8_t *)_tconv_buf; /* Alias for pointer arithmetic */ - const H5S_hyper_dim_t *tdiminfo; /* Temporary pointer to diminfo information */ - hssize_t fast_dim_start, /* Local copies of fastest changing dimension info */ - fast_dim_offset; - hsize_t fast_dim_stride, /* Local copies of fastest changing dimension info */ - fast_dim_block; - size_t fast_dim_count; - size_t tot_blk_count; /* Total number of blocks left to output */ - size_t act_blk_count; /* Actual number of blocks to output */ - size_t fast_dim_buf_off; /* Local copy of amount to move fastest dimension buffer offset */ - int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ - int temp_dim; /* Temporary rank holder */ - hsize_t acc; /* Accumulator */ - int i; /* Counters */ - int ndims; /* Number of dimensions of dataset */ - size_t actual_read; /* The actual number of elements to read in */ - size_t actual_bytes; /* The actual number of bytes to copy */ - size_t io_left; /* The number of elements left in I/O operation */ -#ifndef NO_DUFFS_DEVICE - size_t duffs_index; /* Counting index for Duff's device */ -#endif /* NO_DUFFS_DEVICE */ - - FUNC_ENTER_NOINIT(H5S_hyper_mread_opt); - - /* Set the aliases for a few important dimension ranks */ - fast_dim=mem_space->extent.u.simple.rank-1; - ndims=mem_space->extent.u.simple.rank; - - /* Set up the size of the memory space */ - HDmemcpy(mem_size, mem_space->extent.u.simple.size,mem_space->extent.u.simple.rank*sizeof(hsize_t)); - mem_size[mem_space->extent.u.simple.rank]=elmt_size; - - /* initialize row sizes for each dimension */ - for(i=(ndims-1),acc=1; i>=0; i--) { - slab[i]=acc*elmt_size; - acc*=mem_size[i]; - } /* end for */ - - /* Set the number of elements left for I/O */ - H5_ASSIGN_OVERFLOW(io_left,nelmts,hsize_t,size_t); - - /* Check if we stopped in the middle of a sequence of elements */ - if((mem_iter->hyp.off[fast_dim]-mem_space->select.sel_info.hslab.diminfo[fast_dim].start)%mem_space->select.sel_info.hslab.diminfo[fast_dim].stride!=0 || - ((mem_iter->hyp.off[fast_dim]!=mem_space->select.sel_info.hslab.diminfo[fast_dim].start) && mem_space->select.sel_info.hslab.diminfo[fast_dim].stride==1)) { - hsize_t leftover; /* The number of elements left over from the last sequence */ - - /* Calculate the number of elements left in the sequence */ - if(mem_space->select.sel_info.hslab.diminfo[fast_dim].stride==1) - leftover=mem_space->select.sel_info.hslab.diminfo[fast_dim].block-(mem_iter->hyp.off[fast_dim]-mem_space->select.sel_info.hslab.diminfo[fast_dim].start); - else - leftover=mem_space->select.sel_info.hslab.diminfo[fast_dim].block-((mem_iter->hyp.off[fast_dim]-mem_space->select.sel_info.hslab.diminfo[fast_dim].start)%mem_space->select.sel_info.hslab.diminfo[fast_dim].stride); - - /* Make certain that we don't write too many */ - H5_CHECK_OVERFLOW(leftover,hsize_t,size_t); - actual_read=MIN((size_t)leftover,io_left); - actual_bytes=actual_read*elmt_size; - - /* Copy the location of the point to get */ - HDmemcpy(offset, mem_iter->hyp.off,ndims*sizeof(hssize_t)); - offset[ndims] = 0; - - /* Add in the selection offset */ - for(i=0; i<ndims; i++) - offset[i] += mem_space->select.offset[i]; - - /* Compute the initial buffer offset */ - for(i=0,src=_buf; i<ndims; i++) - src+=offset[i]*slab[i]; - - /* Scatter out the rest of the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - dst+=actual_bytes; - - /* Decrement the number of elements written out */ - io_left -= actual_read; - - /* Advance the point iterator */ - /* If we had enough buffer space to read in the rest of the sequence - * in the fastest changing dimension, move the iterator offset to - * the beginning of the next block to read. Otherwise, just advance - * the iterator in the fastest changing dimension. - */ - if(actual_read==leftover) { - /* Move iterator offset to beginning of next sequence in the fastest changing dimension */ - H5S_hyper_iter_next(mem_space,mem_iter); - } /* end if */ - else { - mem_iter->hyp.off[fast_dim]+=actual_read; /* whole sequence not written out, just advance fastest dimension offset */ - } /* end if */ - } /* end if */ - - /* Now that we've cleared the "remainder" of the previous fastest dimension - * sequence, we must be at the beginning of a sequence, so use the fancy - * algorithm to compute the offsets and run through as many as possible, - * until the buffer fills up. - */ - if(io_left>0) { /* Just in case the "remainder" above filled the buffer */ - /* Compute the arrays to perform I/O on */ - /* Copy the location of the point to get */ - HDmemcpy(offset, mem_iter->hyp.off,ndims*sizeof(hssize_t)); - offset[ndims] = 0; - - /* Add in the selection offset */ - for(i=0; i<ndims; i++) - offset[i] += mem_space->select.offset[i]; - - /* Compute the current "counts" for this location */ - for(i=0; i<ndims; i++) { - if(mem_space->select.sel_info.hslab.diminfo[i].stride==1) { - tmp_count[i] = 0; - tmp_block[i] = mem_iter->hyp.off[i]-mem_space->select.sel_info.hslab.diminfo[i].start; - } /* end if */ - else { - tmp_count[i] = (mem_iter->hyp.off[i]-mem_space->select.sel_info.hslab.diminfo[i].start)/mem_space->select.sel_info.hslab.diminfo[i].stride; - tmp_block[i] = (mem_iter->hyp.off[i]-mem_space->select.sel_info.hslab.diminfo[i].start)%mem_space->select.sel_info.hslab.diminfo[i].stride; - } /* end else */ - } /* end for */ - - /* Compute the initial buffer offset */ - for(i=0,src=_buf; i<ndims; i++) - src+=offset[i]*slab[i]; - - /* Set the number of elements to write each time */ - H5_ASSIGN_OVERFLOW(actual_read,mem_space->select.sel_info.hslab.diminfo[fast_dim].block,hsize_t,size_t); - - /* Set the number of actual bytes */ - actual_bytes=actual_read*elmt_size; - - /* Set the local copy of the diminfo pointer */ - tdiminfo=mem_space->select.sel_info.hslab.diminfo; - - /* Set local copies of information for the fastest changing dimension */ - fast_dim_start=tdiminfo[fast_dim].start; - fast_dim_stride=tdiminfo[fast_dim].stride; - fast_dim_block=tdiminfo[fast_dim].block; - H5_ASSIGN_OVERFLOW(fast_dim_buf_off,(slab[fast_dim]*fast_dim_stride),hsize_t,size_t); - fast_dim_offset=fast_dim_start+mem_space->select.offset[fast_dim]; - - /* Compute the number of blocks which would fit into the buffer */ - H5_ASSIGN_OVERFLOW(tot_blk_count,(io_left/fast_dim_block),hsize_t,size_t); - - /* Compute the amount to wrap at the end of each row */ - for(i=0; i<ndims; i++) - wrap[i]=(mem_size[i]-(tdiminfo[i].stride*tdiminfo[i].count))*slab[i]; - - /* Compute the amount to skip between blocks */ - for(i=0; i<ndims; i++) - skip[i]=(tdiminfo[i].stride-tdiminfo[i].block)*slab[i]; - - /* Read in data until an entire sequence can't be written out any longer */ - while(io_left>0) { - /* Reset copy of number of blocks in fastest dimension */ - H5_ASSIGN_OVERFLOW(fast_dim_count,tdiminfo[fast_dim].count-tmp_count[fast_dim],hsize_t,size_t); - - /* Check if this entire row will fit into buffer */ - if(fast_dim_count<=tot_blk_count) { - - /* Entire row of blocks fits into buffer */ - act_blk_count=fast_dim_count; - -#ifdef NO_DUFFS_DEVICE - /* Loop over all the blocks in the fastest changing dimension */ - while(fast_dim_count>0) { - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - dst+=actual_bytes; - - /* Increment information to reflect block just processed */ - src+=fast_dim_buf_off; - - /* Decrement number of blocks */ - fast_dim_count--; - } /* end while */ -#else /* NO_DUFFS_DEVICE */ - duffs_index = (fast_dim_count + 7) / 8; - switch (fast_dim_count % 8) { - case 0: - do - { - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - dst+=actual_bytes; - - /* Increment information to reflect block just processed */ - src+=fast_dim_buf_off; - - case 7: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - dst+=actual_bytes; - - /* Increment information to reflect block just processed */ - src+=fast_dim_buf_off; - - case 6: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - dst+=actual_bytes; - - /* Increment information to reflect block just processed */ - src+=fast_dim_buf_off; - - case 5: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - dst+=actual_bytes; - - /* Increment information to reflect block just processed */ - src+=fast_dim_buf_off; - - case 4: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - dst+=actual_bytes; - - /* Increment information to reflect block just processed */ - src+=fast_dim_buf_off; - - case 3: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - dst+=actual_bytes; - - /* Increment information to reflect block just processed */ - src+=fast_dim_buf_off; - - case 2: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - dst+=actual_bytes; - - /* Increment information to reflect block just processed */ - src+=fast_dim_buf_off; - - case 1: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - dst+=actual_bytes; - - /* Increment information to reflect block just processed */ - src+=fast_dim_buf_off; - - } while (--duffs_index > 0); - } /* end switch */ -#endif /* NO_DUFFS_DEVICE */ - - /* Decrement number of elements left */ - io_left -= actual_read*act_blk_count; - - /* Decrement number of blocks left */ - tot_blk_count -= act_blk_count; - - /* Increment information to reflect block just processed */ - offset[fast_dim]=fast_dim_offset; /* reset the offset in the fastest dimension */ - tmp_count[fast_dim]=0; - - /* Increment offset in destination buffer */ - src += wrap[fast_dim]; - } /* end if */ - else { - - /* Entire row of blocks doesn't fit into buffer */ - act_blk_count=tot_blk_count; - - /* Reduce number of blocks to output */ - fast_dim_count=tot_blk_count; - - /* Loop over all the blocks in the fastest changing dimension */ - while(fast_dim_count>0) { - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - dst+=actual_bytes; - - /* Increment information to reflect block just processed */ - src+=fast_dim_buf_off; - - /* Decrement number of blocks */ - fast_dim_count--; - } /* end while */ - - /* Decrement number of elements left */ - io_left -= actual_read*act_blk_count; - - /* Decrement number of blocks left */ - tot_blk_count -= act_blk_count; - - /* Increment information to reflect block just processed */ - offset[fast_dim]+=(fast_dim_stride*act_blk_count); /* reset the offset in the fastest dimension */ - tmp_count[fast_dim]+=act_blk_count; - - /* Handle any leftover, partial blocks in this row */ - if(io_left>0) { - actual_read=io_left; - actual_bytes=actual_read*elmt_size; - - /* Scatter out the rest of the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - dst+=actual_bytes; - - /* Decrement the number of elements left */ - io_left -= actual_read; - - /* Increment buffer correctly */ - offset[fast_dim]+=actual_read; - } /* end if */ - - /* don't bother checking slower dimensions */ - assert(tot_blk_count==0); - assert(io_left==0); - break; - } /* end else */ - - /* Increment the offset and count for the other dimensions */ - temp_dim=fast_dim-1; - while(temp_dim>=0) { - /* Move to the next row in the curent dimension */ - offset[temp_dim]++; - tmp_block[temp_dim]++; - - /* If this block is still in the range of blocks to output for the dimension, break out of loop */ - if(tmp_block[temp_dim]<tdiminfo[temp_dim].block) - break; - else { - /* Move to the next block in the current dimension */ - offset[temp_dim]+=(tdiminfo[temp_dim].stride-tdiminfo[temp_dim].block); - src += skip[temp_dim]; - tmp_block[temp_dim]=0; - tmp_count[temp_dim]++; - - /* If this block is still in the range of blocks to output for the dimension, break out of loop */ - if(tmp_count[temp_dim]<tdiminfo[temp_dim].count) - break; - else { - offset[temp_dim]=tdiminfo[temp_dim].start+mem_space->select.offset[temp_dim]; - src += wrap[temp_dim]; - tmp_count[temp_dim]=0; /* reset back to the beginning of the line */ - tmp_block[temp_dim]=0; - } /* end else */ - } /* end else */ - - /* Decrement dimension count */ - temp_dim--; - } /* end while */ - } /* end while */ - - /* Subtract out the selection offset */ - for(i=0; i<ndims; i++) - offset[i] -= mem_space->select.offset[i]; - - /* Update the iterator with the location we stopped */ - HDmemcpy(mem_iter->hyp.off, offset, ndims*sizeof(hssize_t)); - } /* end if */ - - /* Decrement the number of elements left in selection */ - mem_iter->hyp.elmt_left-=(nelmts-io_left); - - FUNC_LEAVE (nelmts-io_left); -} /* end H5S_hyper_mread_opt() */ - - -/*------------------------------------------------------------------------- - * Function: H5S_hyper_mgath - * - * Purpose: Gathers dataset elements from application memory BUF and - * copies them into the data type conversion buffer TCONV_BUF. - * Each element is ELMT_SIZE bytes and arranged in application - * memory according to MEM_SPACE. - * The caller is requesting that at most NELMTS be gathered. - * - * Return: Success: Number of elements copied. - * - * Failure: 0 - * - * Programmer: Quincey Koziol - * Tuesday, June 16, 1998 - * - * Modifications: - * - *------------------------------------------------------------------------- - */ -static hsize_t -H5S_hyper_mgath (const void *_buf, size_t elmt_size, - const H5S_t *mem_space, H5S_sel_iter_t *mem_iter, - hsize_t nelmts, void *_tconv_buf/*out*/) -{ - hsize_t num_read; /* number of elements read into buffer */ - - FUNC_ENTER_NOAPI(H5S_hyper_mgath, 0); - - /* Check args */ - assert (elmt_size>0); - assert (mem_space); - assert (mem_iter); - assert (nelmts>0); - assert (_buf); - assert (_tconv_buf); - - /* Check for the special case of just one H5Sselect_hyperslab call made */ - if(mem_space->select.sel_info.hslab.diminfo!=NULL) { - /* Use optimized call to read in regular hyperslab */ - num_read=H5S_hyper_mread_opt(_buf,elmt_size,mem_space,mem_iter,nelmts,_tconv_buf); - } /* end if */ - else { - /* Perform generic hyperslab operation */ - num_read=H5S_hyper_mread(_buf,elmt_size,mem_space,mem_iter,nelmts,_tconv_buf); - } /* end else */ - - FUNC_LEAVE (num_read); -} /* H5S_hyper_mgath() */ - - -/*------------------------------------------------------------------------- - * Function: H5S_hyper_mwrite - * - * Purpose: Performs an optimized gather from a memory buffer, based on a - * hyperslab span selection. - * - * Return: Success: Number of elements copied. - * Failure: 0 - * - * Programmer: Quincey Koziol - * Tuesday, September 12, 2000 - * - * Modifications: - * - *------------------------------------------------------------------------- - */ -static hssize_t -H5S_hyper_mwrite (const void *_tconv_buf, size_t elmt_size, const H5S_t *space, - H5S_sel_iter_t *iter, hsize_t nelem, void *_buf/*out*/) -{ - const uint8_t *src=(const uint8_t *)_tconv_buf; /* Alias for pointer arithmetic */ - uint8_t *dst=(uint8_t *)_buf; /* Alias for pointer arithmetic */ - uint8_t *tmp_dst; /* Alias for pointer arithmetic */ - H5S_hyper_span_t **ispan; /* Iterator's hyperslab span nodes */ - H5S_hyper_span_t *curr_span; /* Current hyperslab span node */ - hsize_t slab[H5O_LAYOUT_NDIMS]; /* Cumulative size of each dimension in bytes */ - hsize_t acc; /* Accumulator for computing cumulative sizes */ - hssize_t *abs_arr; /* Absolute hyperslab span position */ - hssize_t *off_arr; /* Offset within the dataspace extent */ - int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ - int curr_dim; /* Current dimension being operated on */ - int ndims; /* Number of dimensions of dataset */ - size_t span_size=0; /* Number of bytes in current span to actually process */ - size_t io_bytes_left; /* Number of bytes left to process */ - int i; /* Index variable */ - hssize_t ret_value=FAIL; - - FUNC_ENTER_NOINIT(H5S_hyper_mwrite); - - /* Check args */ - assert(src); - assert(elmt_size>0); - assert(space); - assert(iter); - assert(nelem>0); - assert(dst); - - /* Set the rank of the fastest changing dimension */ - ndims=space->extent.u.simple.rank; - fast_dim=(ndims-1); - - /* Get the pointers to the current span info and span nodes */ - curr_span=iter->hyp.span[fast_dim]; - abs_arr=iter->hyp.off; - off_arr=space->select.offset; - ispan=iter->hyp.span; - - /* Set the amount of elements to perform I/O on, etc. */ - H5_ASSIGN_OVERFLOW(io_bytes_left,nelem*elmt_size,hsize_t,size_t); - - /* Compute the cumulative size of dataspace dimensions */ - for(i=fast_dim, acc=elmt_size; i>=0; i--) { - slab[i]=acc; - acc*=space->extent.u.simple.size[i]; - } /* end for */ - - /* Set the offset of the first element iterated on */ - for(i=0, tmp_dst=dst; i<ndims; i++) - /* Compute the sequential element offset */ - tmp_dst+=(abs_arr[i]+space->select.offset[i])*slab[i]; - - /* Range check against number of elements left in selection */ - assert(io_bytes_left<=(iter->hyp.elmt_left*elmt_size)); - - /* Take care of any partial spans leftover from previous I/Os */ - if(abs_arr[fast_dim]!=curr_span->low) { - - /* Finish the span in the fastest changing dimension */ - - /* Compute the number of bytes to attempt in this span */ - H5_ASSIGN_OVERFLOW(span_size,((curr_span->high-abs_arr[fast_dim])+1)*elmt_size,hsize_t,size_t); - - /* Check number of elements against upper bounds allowed */ - if(span_size>io_bytes_left) - span_size=io_bytes_left; - - HDmemcpy(tmp_dst,src,span_size); - - /* Increment offset in destination */ - src+=span_size; - - /* Decrement I/O left to perform */ - io_bytes_left-=span_size; - - /* Check if we are done */ - if(io_bytes_left>0) { - /* Move to next span in fastest changing dimension */ - curr_span=curr_span->next; - - if(curr_span!=NULL) { - /* Move location offset of destination */ - tmp_dst+=(curr_span->low-abs_arr[fast_dim])*elmt_size; - - /* Move iterator for fastest changing dimension */ - abs_arr[fast_dim]=curr_span->low; - } /* end if */ - } /* end if */ - else { - abs_arr[fast_dim]+=span_size/elmt_size; - - /* Check if we are still within the span */ - if(abs_arr[fast_dim]<=curr_span->high) { - iter->hyp.span[fast_dim]=curr_span; - - goto partial_done; /* finished with partial span */ - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[fast_dim]=curr_span->low; - iter->hyp.span[fast_dim]=curr_span; - - goto partial_done; /* finished with partial span */ - } /* end if */ - } /* end else */ - } /* end else */ - - /* Adjust iterator pointers */ - - if(curr_span==NULL) { -/* Same as code in main loop */ - /* Start at the next fastest dim */ - curr_dim=fast_dim-1; - - /* Work back up through the dimensions */ - while(curr_dim>=0) { - /* Reset the current span */ - curr_span=iter->hyp.span[curr_dim]; - - /* Increment absolute position */ - abs_arr[curr_dim]++; - - /* Check if we are still within the span */ - if(abs_arr[curr_dim]<=curr_span->high) { - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[curr_dim]=curr_span->low; - - break; - } /* end if */ - else { - /* If we finished the span list in this dimension, decrement the dimension worked on and loop again */ - curr_dim--; - } /* end else */ - } /* end else */ - } /* end while */ - - /* Check if we are finished with the spans in the tree */ - if(curr_dim<0) { - /* We had better be done with I/O or bad things are going to happen... */ - assert(io_bytes_left==0); - - goto partial_done; /* finished with partial span */ - } /* end if */ - else { - /* Reset the span in the current dimension */ - ispan[curr_dim]=curr_span; - - /* Walk back down the iterator positions, reseting them */ - while(curr_dim<fast_dim) { - assert(curr_span); - assert(curr_span->down); - assert(curr_span->down->head); - - /* Set the new span_info & span for this dimension */ - iter->hyp.span[curr_dim+1]=curr_span->down->head; - - /* Advance span down the tree */ - curr_span=curr_span->down->head; - - /* Reset the absolute offset for the dim */ - abs_arr[curr_dim+1]=curr_span->low; - - /* Increment current dimension */ - curr_dim++; - } /* end while */ - - /* Verify that the curr_span points to the fastest dim */ - assert(curr_span==iter->hyp.span[fast_dim]); - } /* end else */ - - /* Reset the buffer offset */ - for(i=0, tmp_dst=dst; i<ndims; i++) - tmp_dst+=(abs_arr[i]+off_arr[i])*slab[i]; - } /* end if */ - } /* end if */ - -partial_done: /* Yes, goto's are evil, so sue me... :-) */ - - /* Perform the I/O on the elements, based on the position of the iterator */ - while(io_bytes_left>0) { - /* Adjust buffer offset of destination to compensate for initial increment below */ - tmp_dst-=(size_t)curr_span->pstride; - - /* Loop over all the spans in the fastest changing dimension */ - while(curr_span!=NULL) { - /* Move buffer offset of destination */ - tmp_dst+=(size_t)curr_span->pstride; - - /* Compute the number of elements to attempt in this span */ - H5_ASSIGN_OVERFLOW(span_size,curr_span->nelem,hsize_t,size_t); - - /* Check number of elements against upper bounds allowed */ - if(span_size>=io_bytes_left) { - /* Trim the number of bytes to output */ - span_size=io_bytes_left; - io_bytes_left=0; - -/* COMMON */ - /* "Write" the data into the destination buffer */ - HDmemcpy(tmp_dst,src,span_size); - - /* Increment offset in destination */ - src+=span_size; -/* end COMMON */ - - /* Break out now, we are finished with I/O */ - break; - } /* end if */ - else { - /* Decrement I/O left to perform */ - io_bytes_left-=span_size; - -/* COMMON */ - /* "Write" the data into the destination buffer */ - HDmemcpy(tmp_dst,src,span_size); - - /* Increment offset in destination */ - src+=span_size; -/* end COMMON */ - } /* end else */ - - /* Move to next span in fastest changing dimension */ - curr_span=curr_span->next; - } /* end while */ - - /* Check if we are done */ - if(io_bytes_left==0) { - abs_arr[fast_dim]=curr_span->low+(span_size/elmt_size); - - /* Check if we are still within the span */ - if(abs_arr[fast_dim]<=curr_span->high) { - iter->hyp.span[fast_dim]=curr_span; - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[fast_dim]=curr_span->low; - iter->hyp.span[fast_dim]=curr_span; - break; - } /* end if */ - } /* end else */ - } /* end if */ - - /* Adjust iterator pointers */ - - /* Start at the next fastest dim */ - curr_dim=fast_dim-1; - - /* Work back up through the dimensions */ - while(curr_dim>=0) { - /* Reset the current span */ - curr_span=iter->hyp.span[curr_dim]; - - /* Increment absolute position */ - abs_arr[curr_dim]++; - - /* Check if we are still within the span */ - if(abs_arr[curr_dim]<=curr_span->high) { - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset absolute position */ - abs_arr[curr_dim]=curr_span->low; - - break; - } /* end if */ - else { - /* If we finished the span list in this dimension, decrement the dimension worked on and loop again */ - curr_dim--; - } /* end else */ - } /* end else */ - } /* end while */ - - /* Check if we are finished with the spans in the tree */ - if(curr_dim<0) { - /* We had better be done with I/O or bad things are going to happen... */ - assert(io_bytes_left==0); - break; - } /* end if */ - else { - /* Reset the span in the current dimension */ - ispan[curr_dim]=curr_span; - - /* Walk back down the iterator positions, reseting them */ - while(curr_dim<fast_dim) { - assert(curr_span); - assert(curr_span->down); - assert(curr_span->down->head); - - /* Set the new span_info & span for the next dimension down */ - iter->hyp.span[curr_dim+1]=curr_span->down->head; - - /* Advance span down the tree */ - curr_span=curr_span->down->head; - - /* Reset the absolute offset for the dim */ - abs_arr[curr_dim+1]=curr_span->low; - - /* Increment current dimension */ - curr_dim++; - } /* end while */ - - /* Verify that the curr_span points to the fastest dim */ - assert(curr_span==iter->hyp.span[fast_dim]); - } /* end else */ - - /* Reset the buffer offset */ - for(i=0, tmp_dst=dst; i<ndims; i++) - tmp_dst+=(abs_arr[i]+off_arr[i])*slab[i]; - } /* end while */ - - /* Increment amount of I/O performed */ - iter->hyp.elmt_left-=nelem; - - /* Success! */ - ret_value=nelem; - -#ifdef LATER -done: -#endif /* LATER */ - FUNC_LEAVE (ret_value); -} /* end H5S_hyper_mwrite() */ - - -/*------------------------------------------------------------------------- - * Function: H5S_hyper_mwrite_opt - * - * Purpose: Performs an optimized scatter to a memory buffer, based on a - * regular hyperslab (i.e. one which was generated from just one call to - * H5Sselect_hyperslab). - * - * Return: Success: Number of elements copied. - * Failure: 0 - * - * Programmer: Quincey Koziol - * Tuesday, September 12, 2000 - * - * Modifications: - * - *------------------------------------------------------------------------- - */ -static hsize_t -H5S_hyper_mwrite_opt (const void *_tconv_buf, size_t elmt_size, - const H5S_t *mem_space, H5S_sel_iter_t *mem_iter, - hsize_t nelmts, void *_buf/*out*/) -{ - hsize_t mem_size[H5O_LAYOUT_NDIMS]; /* Size of the source buffer */ - hsize_t slab[H5O_LAYOUT_NDIMS]; /* Hyperslab size */ - hssize_t wrap[H5O_LAYOUT_NDIMS]; /* Bytes to wrap around at the end of a row */ - hsize_t skip[H5O_LAYOUT_NDIMS]; /* Bytes to skip between blocks */ - hssize_t offset[H5O_LAYOUT_NDIMS]; /* Offset on disk */ - hsize_t tmp_count[H5O_LAYOUT_NDIMS]; /* Temporary block count */ - hsize_t tmp_block[H5O_LAYOUT_NDIMS]; /* Temporary block offset */ - const uint8_t *src=(const uint8_t *)_tconv_buf; /* Alias for pointer arithmetic */ - uint8_t *dst=(uint8_t *)_buf; /* Alias for pointer arithmetic */ - const H5S_hyper_dim_t *tdiminfo; /* Temporary pointer to diminfo information */ - hssize_t fast_dim_start, /* Local copies of fastest changing dimension info */ - fast_dim_offset; - hsize_t fast_dim_stride, /* Local copies of fastest changing dimension info */ - fast_dim_block; - size_t fast_dim_count; - size_t tot_blk_count; /* Total number of blocks left to output */ - size_t act_blk_count; /* Actual number of blocks to output */ - size_t fast_dim_buf_off; /* Local copy of amount to move fastest dimension buffer offset */ - int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ - int temp_dim; /* Temporary rank holder */ - hsize_t acc; /* Accumulator */ - int i; /* Counters */ - int ndims; /* Number of dimensions of dataset */ - size_t actual_write; /* The actual number of elements to read in */ - size_t actual_bytes; /* The actual number of bytes to copy */ - size_t io_left; /* The number of elements left in I/O operation */ -#ifndef NO_DUFFS_DEVICE - size_t duffs_index; /* Counting index for Duff's device */ -#endif /* NO_DUFFS_DEVICE */ - - FUNC_ENTER_NOINIT(H5S_hyper_mwrite_opt); - - /* Set the aliases for a few important dimension ranks */ - fast_dim=mem_space->extent.u.simple.rank-1; - ndims=mem_space->extent.u.simple.rank; - - /* Set up the size of the memory space */ - HDmemcpy(mem_size, mem_space->extent.u.simple.size,mem_space->extent.u.simple.rank*sizeof(hsize_t)); - mem_size[mem_space->extent.u.simple.rank]=elmt_size; - - /* initialize row sizes for each dimension */ - for(i=(ndims-1),acc=1; i>=0; i--) { - slab[i]=acc*elmt_size; - acc*=mem_size[i]; - } /* end for */ - - /* Set the number of elements left for I/O */ - H5_ASSIGN_OVERFLOW(io_left,nelmts,hsize_t,size_t); - - /* Check if we stopped in the middle of a sequence of elements */ - if((mem_iter->hyp.off[fast_dim]-mem_space->select.sel_info.hslab.diminfo[fast_dim].start)%mem_space->select.sel_info.hslab.diminfo[fast_dim].stride!=0 || - ((mem_iter->hyp.off[fast_dim]!=mem_space->select.sel_info.hslab.diminfo[fast_dim].start) && mem_space->select.sel_info.hslab.diminfo[fast_dim].stride==1)) { - hsize_t leftover; /* The number of elements left over from the last sequence */ - - /* Calculate the number of elements left in the sequence */ - if(mem_space->select.sel_info.hslab.diminfo[fast_dim].stride==1) - leftover=mem_space->select.sel_info.hslab.diminfo[fast_dim].block-(mem_iter->hyp.off[fast_dim]-mem_space->select.sel_info.hslab.diminfo[fast_dim].start); - else - leftover=mem_space->select.sel_info.hslab.diminfo[fast_dim].block-((mem_iter->hyp.off[fast_dim]-mem_space->select.sel_info.hslab.diminfo[fast_dim].start)%mem_space->select.sel_info.hslab.diminfo[fast_dim].stride); - - /* Make certain that we don't write too many */ - H5_CHECK_OVERFLOW(leftover,hsize_t,size_t); - actual_write=MIN((size_t)leftover,io_left); - actual_bytes=actual_write*elmt_size; - - /* Copy the location of the point to get */ - HDmemcpy(offset, mem_iter->hyp.off,ndims*sizeof(hssize_t)); - offset[ndims] = 0; - - /* Add in the selection offset */ - for(i=0; i<ndims; i++) - offset[i] += mem_space->select.offset[i]; - - /* Compute the initial buffer offset */ - for(i=0,dst=(unsigned char *)_buf; i<ndims; i++) - dst+=offset[i]*slab[i]; - - /* Scatter out the rest of the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - src+=actual_bytes; - - /* Decrement the number of elements written out */ - io_left -= actual_write; - - /* Advance the point iterator */ - /* If we had enough buffer space to write out the rest of the sequence - * in the fastest changing dimension, move the iterator offset to - * the beginning of the next block to write. Otherwise, just advance - * the iterator in the fastest changing dimension. - */ - if(actual_write==leftover) { - /* Move iterator offset to beginning of next sequence in the fastest changing dimension */ - H5S_hyper_iter_next(mem_space,mem_iter); - } /* end if */ - else { - mem_iter->hyp.off[fast_dim]+=actual_write; /* whole sequence not written out, just advance fastest dimension offset */ - } /* end if */ - } /* end if */ - - /* Now that we've cleared the "remainder" of the previous fastest dimension - * sequence, we must be at the beginning of a sequence, so use the fancy - * algorithm to compute the offsets and run through as many as possible, - * until the buffer fills up. - */ - if(io_left>0) { /* Just in case the "remainder" above filled the buffer */ - /* Compute the arrays to perform I/O on */ - /* Copy the location of the point to get */ - HDmemcpy(offset, mem_iter->hyp.off,ndims*sizeof(hssize_t)); - offset[ndims] = 0; - - /* Add in the selection offset */ - for(i=0; i<ndims; i++) - offset[i] += mem_space->select.offset[i]; - - /* Compute the current "counts" for this location */ - for(i=0; i<ndims; i++) { - if(mem_space->select.sel_info.hslab.diminfo[i].stride==1) { - tmp_count[i] = 0; - tmp_block[i] = mem_iter->hyp.off[i]-mem_space->select.sel_info.hslab.diminfo[i].start; - } /* end if */ - else { - tmp_count[i] = (mem_iter->hyp.off[i]-mem_space->select.sel_info.hslab.diminfo[i].start)/mem_space->select.sel_info.hslab.diminfo[i].stride; - tmp_block[i] = (mem_iter->hyp.off[i]-mem_space->select.sel_info.hslab.diminfo[i].start)%mem_space->select.sel_info.hslab.diminfo[i].stride; - } /* end else */ - } /* end for */ - - /* Compute the initial buffer offset */ - for(i=0,dst=(unsigned char *)_buf; i<ndims; i++) - dst+=offset[i]*slab[i]; - - /* Set the number of elements to write each time */ - H5_ASSIGN_OVERFLOW(actual_write,mem_space->select.sel_info.hslab.diminfo[fast_dim].block,hsize_t,size_t); - - /* Set the number of actual bytes */ - actual_bytes=actual_write*elmt_size; - - /* Set the local copy of the diminfo pointer */ - tdiminfo=mem_space->select.sel_info.hslab.diminfo; - - /* Set local copies of information for the fastest changing dimension */ - fast_dim_start=tdiminfo[fast_dim].start; - fast_dim_stride=tdiminfo[fast_dim].stride; - fast_dim_block=tdiminfo[fast_dim].block; - H5_ASSIGN_OVERFLOW(fast_dim_buf_off,slab[fast_dim]*fast_dim_stride,hsize_t,size_t); - fast_dim_offset=fast_dim_start+mem_space->select.offset[fast_dim]; - - /* Compute the number of blocks which would fit into the buffer */ - H5_ASSIGN_OVERFLOW(tot_blk_count,(io_left/fast_dim_block),hsize_t,size_t); - - /* Compute the amount to wrap at the end of each row */ - for(i=0; i<ndims; i++) - wrap[i]=(mem_size[i]-(tdiminfo[i].stride*tdiminfo[i].count))*slab[i]; - - /* Compute the amount to skip between blocks */ - for(i=0; i<ndims; i++) - skip[i]=(tdiminfo[i].stride-tdiminfo[i].block)*slab[i]; - - /* Read in data until an entire sequence can't be written out any longer */ - while(io_left>0) { - /* Reset copy of number of blocks in fastest dimension */ - H5_ASSIGN_OVERFLOW(fast_dim_count,tdiminfo[fast_dim].count-tmp_count[fast_dim],hsize_t,size_t); - - /* Check if this entire row will fit into buffer */ - if(fast_dim_count<=tot_blk_count) { - - /* Entire row of blocks fits into buffer */ - act_blk_count=fast_dim_count; - -#ifdef NO_DUFFS_DEVICE - /* Loop over all the blocks in the fastest changing dimension */ - while(fast_dim_count>0) { - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - src+=actual_bytes; - - /* Increment information to reflect block just processed */ - dst+=fast_dim_buf_off; - - /* Decrement number of blocks */ - fast_dim_count--; - } /* end while */ -#else /* NO_DUFFS_DEVICE */ - duffs_index = (fast_dim_count + 7) / 8; - switch (fast_dim_count % 8) { - case 0: - do - { - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - src+=actual_bytes; - - /* Increment information to reflect block just processed */ - dst+=fast_dim_buf_off; - - case 7: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - src+=actual_bytes; - - /* Increment information to reflect block just processed */ - dst+=fast_dim_buf_off; - - case 6: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - src+=actual_bytes; - - /* Increment information to reflect block just processed */ - dst+=fast_dim_buf_off; - - case 5: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - src+=actual_bytes; - - /* Increment information to reflect block just processed */ - dst+=fast_dim_buf_off; - - case 4: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - src+=actual_bytes; - - /* Increment information to reflect block just processed */ - dst+=fast_dim_buf_off; - - case 3: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - src+=actual_bytes; - - /* Increment information to reflect block just processed */ - dst+=fast_dim_buf_off; - - case 2: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - src+=actual_bytes; - - /* Increment information to reflect block just processed */ - dst+=fast_dim_buf_off; - - case 1: - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - src+=actual_bytes; - - /* Increment information to reflect block just processed */ - dst+=fast_dim_buf_off; - - } while (--duffs_index > 0); - } /* end switch */ -#endif /* NO_DUFFS_DEVICE */ - - /* Decrement number of elements left */ - io_left -= actual_write*act_blk_count; - - /* Decrement number of blocks left */ - tot_blk_count -= act_blk_count; - - /* Increment information to reflect block just processed */ - offset[fast_dim]=fast_dim_offset; /* reset the offset in the fastest dimension */ - tmp_count[fast_dim]=0; - - /* Increment offset in destination buffer */ - dst += wrap[fast_dim]; - } /* end if */ - else { - - /* Entire row of blocks doesn't fit into buffer */ - act_blk_count=tot_blk_count; - - /* Reduce number of blocks to output */ - fast_dim_count=tot_blk_count; - - /* Loop over all the blocks in the fastest changing dimension */ - while(fast_dim_count>0) { - /* Scatter out the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - src+=actual_bytes; - - /* Increment information to reflect block just processed */ - dst+=fast_dim_buf_off; - - /* Decrement number of blocks */ - fast_dim_count--; - } /* end while */ - - /* Decrement number of elements left */ - io_left -= actual_write*act_blk_count; - - /* Decrement number of blocks left */ - tot_blk_count -= act_blk_count; - - /* Increment information to reflect block just processed */ - offset[fast_dim]+=(fast_dim_stride*act_blk_count); /* reset the offset in the fastest dimension */ - tmp_count[fast_dim]+=act_blk_count; - - /* Handle any leftover, partial blocks in this row */ - if(io_left>0) { - actual_write=io_left; - actual_bytes=actual_write*elmt_size; - - /* Scatter out the rest of the sequence */ - HDmemcpy(dst,src,actual_bytes); - - /* Increment the offset of the buffer */ - src+=actual_bytes; - - /* Decrement the number of elements left */ - io_left -= actual_write; - - /* Increment buffer correctly */ - offset[fast_dim]+=actual_write; - } /* end if */ - - /* don't bother checking slower dimensions */ - assert(tot_blk_count==0); - assert(io_left==0); - break; - } /* end else */ - - /* Increment the offset and count for the other dimensions */ - temp_dim=fast_dim-1; - while(temp_dim>=0) { - /* Move to the next row in the curent dimension */ - offset[temp_dim]++; - tmp_block[temp_dim]++; - - /* If this block is still in the range of blocks to output for the dimension, break out of loop */ - if(tmp_block[temp_dim]<tdiminfo[temp_dim].block) - break; - else { - /* Move to the next block in the current dimension */ - offset[temp_dim]+=(tdiminfo[temp_dim].stride-tdiminfo[temp_dim].block); - dst += skip[temp_dim]; - tmp_block[temp_dim]=0; - tmp_count[temp_dim]++; - - /* If this block is still in the range of blocks to output for the dimension, break out of loop */ - if(tmp_count[temp_dim]<tdiminfo[temp_dim].count) - break; - else { - offset[temp_dim]=tdiminfo[temp_dim].start+mem_space->select.offset[temp_dim]; - dst += wrap[temp_dim]; - tmp_count[temp_dim]=0; /* reset back to the beginning of the line */ - tmp_block[temp_dim]=0; - } /* end else */ - } /* end else */ - - /* Decrement dimension count */ - temp_dim--; - } /* end while */ - } /* end while */ - - /* Subtract out the selection offset */ - for(i=0; i<ndims; i++) - offset[i] -= mem_space->select.offset[i]; - - /* Update the iterator with the location we stopped */ - HDmemcpy(mem_iter->hyp.off, offset, ndims*sizeof(hssize_t)); - } /* end if */ - - /* Decrement the number of elements left in selection */ - mem_iter->hyp.elmt_left-=(nelmts-io_left); - - FUNC_LEAVE (nelmts-io_left); -} /* end H5S_hyper_mwrite_opt() */ - - -/*------------------------------------------------------------------------- - * Function: H5S_hyper_mscat - * - * Purpose: Scatters NELMTS data points from the type conversion buffer - * TCONV_BUF to the application buffer BUF. Each element is - * ELMT_SIZE bytes and they are organized in application memory - * according to MEM_SPACE. - * - * Return: Non-negative on success/Negative on failure - * - * Programmer: Quincey Koziol - * Wednesday, June 17, 1998 - * - * Modifications: - * - *------------------------------------------------------------------------- - */ -static herr_t -H5S_hyper_mscat (const void *_tconv_buf, size_t elmt_size, - const H5S_t *mem_space, H5S_sel_iter_t *mem_iter, - hsize_t nelmts, void *_buf/*out*/) -{ - hsize_t num_written; /* number of elements written into buffer */ - - FUNC_ENTER_NOAPI(H5S_hyper_mscat, 0); - - /* Check args */ - assert (elmt_size>0); - assert (mem_space); - assert (mem_iter); - assert (nelmts>0); - assert (_buf); - assert (_tconv_buf); - - /* Check for the special case of just one H5Sselect_hyperslab call made */ - if(mem_space->select.sel_info.hslab.diminfo!=NULL) { - /* Use optimized call to write out regular hyperslab */ - num_written=H5S_hyper_mwrite_opt(_tconv_buf,elmt_size,mem_space,mem_iter,nelmts,_buf); - } /* end if */ - else { - /* Perform generic hyperslab operation */ - num_written=H5S_hyper_mwrite(_tconv_buf,elmt_size,mem_space,mem_iter,nelmts,_buf); - } /* end else */ - - FUNC_LEAVE (num_written>0 ? SUCCEED : FAIL); -} /* H5S_hyper_mscat() */ - - /*-------------------------------------------------------------------------- NAME H5S_hyper_npoints @@ -4874,7 +1206,7 @@ H5S_hyper_select_valid (const H5S_t *space) Count the number of blocks in a span tree USAGE hssize_t H5S_hyper_span_nblocks(spans) - const H5S_hyper_span_info_t *spans; IN: Hyperslan span tree to count elements of + const H5S_hyper_span_info_t *spans; IN: Hyperslab span tree to count elements of RETURNS Number of blocks in span tree on success; negative on failure DESCRIPTION @@ -5840,475 +2172,6 @@ H5S_hyper_select_regular(const H5S_t *space) /*-------------------------------------------------------------------------- NAME - H5S_hyper_select_iterate_mem_gen - PURPOSE - Internal routine to iterate over the elements of a span tree hyperslab selection - USAGE - herr_t H5S_hyper_select_iterate_mem_gen(iter_info) - H5S_hyper_iter_info_t *iter_info; IN/OUT: Block of iteration parameters to pass into recursive calls - RETURNS - Non-negative on success, negative on failure - DESCRIPTION - Iterates over the elements in a hyperslab span tree selection, calling a - user's callback routine for each element. - GLOBAL VARIABLES - COMMENTS, BUGS, ASSUMPTIONS - EXAMPLES - REVISION LOG ---------------------------------------------------------------------------*/ -static herr_t -H5S_hyper_select_iterate_mem_gen(H5S_hyper_iter_info_t *iter_info) -{ - const H5S_t *space; /* Dataspace operating with */ - H5S_sel_iter_t *iter; /* Selection iterator */ - H5S_hyper_span_t *curr_span; /* Current hyperslab span node */ - hsize_t slab[H5O_LAYOUT_NDIMS]; /* Cumulative size of each dimension in bytes */ - hsize_t acc; /* Accumulator for computing cumulative sizes */ - hssize_t off_arr[H5O_LAYOUT_NDIMS]; /* Current hyperslab span position */ - hssize_t coord_arr[H5O_LAYOUT_NDIMS]; /* Current coordinate position */ - int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ - int curr_dim; /* Current dimension being operated on */ - int ndims; /* Number of dimensions of dataset */ - hsize_t span_io; /* Number of elements in current span to actually process */ - herr_t user_ret=0; /* User's return value */ - uint8_t *loc; /* Current element location pointer */ - hsize_t loc_off; /* Element offset in the dataspace */ - int i; /* Index variable */ - unsigned u; /* Index variable */ - herr_t ret_value=FAIL; - - FUNC_ENTER_NOINIT(H5S_hyper_select_iterate_mem_gen); - - /* Check args */ - assert(iter_info); - - /* Retrieve some information from the interation info */ - space=iter_info->space; - iter=iter_info->iter; - - /* Set the rank of the fastest changing dimension */ - ndims=space->extent.u.simple.rank; - fast_dim=(ndims-1); - - /* Get the pointers to the current span info and span nodes */ - curr_span=iter->hyp.span[fast_dim]; - - /* Compute the cumulative size of dataspace dimensions */ - for(i=fast_dim, acc=iter_info->elem_size; i>=0; i--) { - slab[i]=acc; - acc*=space->extent.u.simple.size[i]; - } /* end for */ - - /* Set the offset of the first element iterated on */ - for(i=0, loc_off=0; i<ndims; i++) { - /* Set the location */ - off_arr[i]=iter->hyp.span[i]->low; - coord_arr[i]=off_arr[i]+space->select.offset[i]; - - /* Compute the sequential element offset */ - loc_off+=coord_arr[i]*slab[i]; - } /* end for */ - - /* Perform the I/O on the elements, based on the position of the iterator */ - user_ret=0; - while(curr_span!=NULL && user_ret==0) { - /* Compute the number of elements to attempt in this span */ - span_io=(curr_span->high-curr_span->low)+1; - - /* Iterate through all the span elements */ - for(u=0, loc=(uint8_t *)iter_info->src+loc_off; u<span_io && user_ret==0; u++) { - /* Call the user's callback routine */ - user_ret=(*(iter_info->op))(loc,iter_info->dt,(hsize_t)ndims,coord_arr,iter_info->op_data); - - /* Increment the element location */ - off_arr[fast_dim]++; - coord_arr[fast_dim]++; - - /* Increment the buffer offset */ - loc+=slab[fast_dim]; - } /* end for */ - - /* Get out now for user return values not zero */ - if(user_ret!=0) - break; - - /* Adjust iterator pointers */ - - /* Advance span in fastest dimension */ - curr_span=curr_span->next; - - /* See if we are still in the fastest changing dimension */ - if(curr_span!=NULL) { - /* Move the buffer offset */ - loc_off+=(span_io+(curr_span->low-off_arr[fast_dim]))*iter_info->elem_size; - - /* Move the element location */ - off_arr[fast_dim]=curr_span->low; - coord_arr[fast_dim]=off_arr[fast_dim]+space->select.offset[fast_dim]; - } /* end if */ - /* We walked off the spans for the fastest dimension, work our way back up */ - else { - /* Start at the fastest dim */ - curr_dim=fast_dim-1; - - /* Get the pointer to the correct dimension */ - curr_span=iter->hyp.span[curr_dim]; - - /* Work back up through the dimensions */ - while(curr_dim>=0) { - /* Increment position in span */ - off_arr[curr_dim]++; - coord_arr[curr_dim]++; - - /* Check if we are still within the span */ - if(off_arr[curr_dim]<=curr_span->high) { - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset the offset for the dim */ - off_arr[curr_dim]=curr_span->low; - coord_arr[curr_dim]=off_arr[curr_dim]+space->select.offset[curr_dim]; - - break; - } /* end if */ - else { - /* If we finished the span list in this dimension, decrement the dimension worked on and loop again */ - curr_dim--; - - /* Reset the curr_span to the next dim */ - if(curr_dim>=0) - curr_span=iter->hyp.span[curr_dim]; - } /* end else */ - } /* end else */ - } /* end while */ - - /* Check if we are finished with the spans in the tree */ - if(curr_dim<0) { - /* We had better be done with I/O or bad things are going to happen... */ - break; - } /* end if */ - else { - /* Reset the span in the current dimension */ - iter->hyp.span[curr_dim]=curr_span; - - /* Walk back down the iterator positions, reseting them */ - while(curr_dim<fast_dim) { - assert(curr_span); - assert(curr_span->down); - assert(curr_span->down->head); - - /* Set the new span for this dimension */ - iter->hyp.span[curr_dim+1]=curr_span->down->head; - - /* Advance span down the tree */ - curr_span=curr_span->down->head; - - /* Reset the offset for the dim */ - off_arr[curr_dim+1]=curr_span->low; - coord_arr[curr_dim+1]=off_arr[curr_dim+1]+space->select.offset[curr_dim+1]; - - /* Increment current dimension */ - curr_dim++; - } /* end while */ - - /* Verify that the curr_span points to the fastest dim */ - assert(curr_span==iter->hyp.span[fast_dim]); - - /* Verify that the offset is correct for the fastest dim */ - assert(off_arr[fast_dim]==curr_span->low); - } /* end else */ - - /* Reset the buffer offset */ - for(i=0, loc_off=0; i<ndims; i++) - loc_off+=coord_arr[i]*slab[i]; - } /* end else */ - } /* end while */ - - /* Success! */ - ret_value=(user_ret==0 ? SUCCEED : user_ret); - -#ifdef LATER -done: -#endif /* LATER */ - FUNC_LEAVE (ret_value); -} /* end H5S_hyper_select_iterate_mem_gen() */ - - -/*-------------------------------------------------------------------------- - NAME - H5S_hyper_select_iterate_mem_opt - PURPOSE - Iterate over the data points in a regular hyperslab selection, calling a - user's function for each element. - USAGE - herr_t H5S_hyper_select_iterate_mem_opt(buf, type_id, space, op, operator_data) - H5S_sel_iter_t *iter; IN/OUT: Selection iterator - void *buf; IN/OUT: Buffer containing elements to iterate over - hid_t type_id; IN: Datatype ID of BUF array. - H5S_t *space; IN: Dataspace object containing selection to iterate over - H5D_operator_t op; IN: Function pointer to the routine to be - called for each element in BUF iterated over. - void *op_data; IN/OUT: Pointer to any user-defined data associated - with the operation. - RETURNS - Returns the return value of the last operator if it was non-zero, or zero - if all elements were processed. Otherwise returns a negative value. - DESCRIPTION - Iterates over the selected elements in a memory buffer, calling the user's - callback function for each element. The selection in the dataspace is - modified so that any elements already iterated over are removed from the - selection if the iteration is interrupted (by the H5D_operator_t function - returning non-zero) in the "middle" of the iteration and may be re-started - by the user where it left off. - - NOTE: Until "subtracting" elements from a selection is implemented, - the selection is not modified. - GLOBAL VARIABLES - COMMENTS, BUGS, ASSUMPTIONS - EXAMPLES - REVISION LOG ---------------------------------------------------------------------------*/ -static herr_t -H5S_hyper_select_iterate_mem_opt(H5S_sel_iter_t UNUSED *iter, void *buf, hid_t type_id, H5S_t *space, H5D_operator_t op, - void *op_data) -{ - H5S_hyper_dim_t *diminfo; /* Alias for dataspace's diminfo information */ - hsize_t tmp_count[H5O_LAYOUT_NDIMS]; /* Temporary hyperslab counts */ - hsize_t tmp_block[H5O_LAYOUT_NDIMS]; /* Temporary hyperslab blocks */ - hssize_t offset[H5O_LAYOUT_NDIMS]; /* Offset of element in dataspace */ - hsize_t slab[H5O_LAYOUT_NDIMS]; /* Size of objects in buffer */ - size_t elem_size; /* Size of data element in buffer */ - hssize_t temp_off; /* Offset in a given dimension */ - uint8_t *loc; /* Current element location */ - int i; /* Counter */ - unsigned u; /* Counter */ - int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ - int temp_dim; /* Temporary rank holder */ - unsigned ndims; /* Rank of the dataspace */ - H5T_t *dt; /* Datatype structure */ - herr_t user_ret=0; /* User's return value */ - - FUNC_ENTER_NOINIT(H5S_hyper_select_iterate_mem_opt); - - /* Set some convienence values */ - ndims=space->extent.u.simple.rank; - fast_dim=ndims-1; - diminfo=space->select.sel_info.hslab.diminfo; - - /* Get the data element size */ - if (NULL==(dt=H5I_object(type_id))) - HRETURN_ERROR(H5E_ARGS, H5E_BADTYPE, FAIL, "not an valid base datatype"); - elem_size=H5T_get_size(dt); - - /* Elements in the fastest dimension are 'elem_size' */ - slab[ndims-1]=elem_size; - - /* If we have two or more dimensions, build the other dimension's element sizes */ - if(ndims>=2) { - /* Build the table of next-dimension down 'element' sizes */ - for(i=ndims-2; i>=0; i--) - slab[i]=slab[i+1]*space->extent.u.simple.size[i+1]; - } /* end if */ - - /* Build the tables of count & block sizes as well as the initial offset */ - for(u=0; u<ndims; u++) { - tmp_count[u]=diminfo[u].count; - tmp_block[u]=diminfo[u].block; - offset[u]=(diminfo[u].start+space->select.offset[u]); - } /* end for */ - - /* Initialize the starting location */ - for(loc=buf,u=0; u<ndims; u++) - loc+=offset[u]*slab[u]; - - /* Go iterate over the hyperslabs */ - while(user_ret==0) { - /* Iterate over the blocks in the fastest dimension */ - while(tmp_count[fast_dim]>0 && user_ret==0) { - - /* Iterate over the elements in the fastest dimension */ - while(tmp_block[fast_dim]>0 && user_ret==0) { - user_ret=(*op)(loc,type_id,(hsize_t)ndims,offset,op_data); - - /* Increment the buffer location */ - loc+=slab[fast_dim]; - - /* Increment the offset in the dataspace */ - offset[fast_dim]++; - - /* Decrement the sequence count */ - tmp_block[fast_dim]--; - } /* end while */ - - /* Reset the sequence count */ - tmp_block[fast_dim]=diminfo[fast_dim].block; - - /* Move the location to the next sequence to start */ - loc+=(diminfo[fast_dim].stride-diminfo[fast_dim].block)*slab[fast_dim]; - - /* Move the offset to the next sequence to start */ - offset[fast_dim]+=(diminfo[fast_dim].stride-diminfo[fast_dim].block); - - /* Decrement the block count */ - tmp_count[fast_dim]--; - } /* end while */ - - /* Check for getting out of iterator, we're done in the 1-D case */ - if(ndims==1) - goto done; /* Yes, an evil goto.. :-) -QAK */ - - /* Work on other dimensions if necessary */ - if(fast_dim>0 && user_ret==0) { - /* Reset the sequence and block counts */ - tmp_block[fast_dim]=diminfo[fast_dim].block; - tmp_count[fast_dim]=diminfo[fast_dim].count; - - /* Bubble up the decrement to the slower changing dimensions */ - temp_dim=fast_dim-1; - while(temp_dim>=0) { - /* Decrement the sequence count in this dimension */ - tmp_block[temp_dim]--; - - /* Check if we are still in the sequence */ - if(tmp_block[temp_dim]>0) - break; - - /* Reset the sequence count in this dimension */ - tmp_block[temp_dim]=diminfo[temp_dim].block; - - /* Decrement the block count */ - tmp_count[temp_dim]--; - - /* Check if we have more blocks left */ - if(tmp_count[temp_dim]>0) - break; - - /* Check for getting out of iterator */ - if(temp_dim==0) - goto done; /* Yes, an evil goto.. :-) -QAK */ - - /* Reset the block count in this dimension */ - tmp_count[temp_dim]=diminfo[temp_dim].count; - - /* Wrapped a dimension, go up to next dimension */ - temp_dim--; - } /* end while */ - } /* end if */ - - /* Re-compute buffer location & offset array */ - for(loc=buf,u=0; u<ndims; u++) { - temp_off=(diminfo[u].start+space->select.offset[u]) - +diminfo[u].stride*(diminfo[u].count-tmp_count[u]) - +(diminfo[u].block-tmp_block[u]); - loc+=temp_off*slab[u]; - offset[u]=temp_off; - } /* end for */ - } /* end while */ - -done: - FUNC_LEAVE (user_ret); -} /* end H5S_hyper_select_iterate_mem_opt() */ - - -/*-------------------------------------------------------------------------- - NAME - H5S_hyper_select_iterate - PURPOSE - Iterate over a hyperslab selection, calling a user's function for each - element. - USAGE - herr_t H5S_hyper_select_iterate(buf, type_id, space, op, operator_data) - void *buf; IN/OUT: Buffer containing elements to iterate over - hid_t type_id; IN: Datatype ID of BUF array. - H5S_t *space; IN: Dataspace object containing selection to iterate over - H5D_operator_t op; IN: Function pointer to the routine to be - called for each element in BUF iterated over. - void *operator_data; IN/OUT: Pointer to any user-defined data - associated with the operation. - RETURNS - Returns the return value of the last operator if it was non-zero, or zero - if all elements were processed. Otherwise returns a negative value. - DESCRIPTION - Iterates over the selected elements in a memory buffer, calling the user's - callback function for each element. The selection in the dataspace is - modified so that any elements already iterated over are removed from the - selection if the iteration is interrupted (by the H5D_operator_t function - returning non-zero) in the "middle" of the iteration and may be re-started - by the user where it left off. - - NOTE: Until "subtracting" elements from a selection is implemented, - the selection is not modified. - GLOBAL VARIABLES - COMMENTS, BUGS, ASSUMPTIONS - EXAMPLES - REVISION LOG ---------------------------------------------------------------------------*/ -herr_t -H5S_hyper_select_iterate(void *buf, hid_t type_id, H5S_t *space, H5D_operator_t op, - void *operator_data) -{ - H5S_hyper_iter_info_t iter_info; /* Block of parameters to pass into recursive calls */ - H5S_sel_iter_t iter; /* selection iteration info*/ - size_t elmt_size; /* Datatype size */ - H5T_t *dt; /* Datatype structure */ - herr_t ret_value=FAIL; /* return value */ - - FUNC_ENTER_NOAPI(H5S_hyper_select_iterate, FAIL); - - assert(buf); - assert(space); - assert(op); - assert(H5I_DATATYPE == H5I_get_type(type_id)); - - /* Initialize iterator */ - HDmemset(&iter,0,sizeof(H5S_sel_iter_t)); - - /* Get the datatype size */ - if (NULL==(dt=H5I_object(type_id))) - HGOTO_ERROR(H5E_ARGS, H5E_BADTYPE, FAIL, "not an valid base datatype"); - elmt_size=H5T_get_size(dt); - - /* Construct iterator for hyperslab selection */ - if (H5S_hyper_init(space, elmt_size, &iter)<0) - HGOTO_ERROR (H5E_DATASPACE, H5E_CANTINIT, FAIL, "unable to initialize selection information"); - - /* Check for the special case of just one H5Sselect_hyperslab call made */ - if(space->select.sel_info.hslab.diminfo!=NULL) { - /* Use optimized call to iterate over regular hyperslab */ - ret_value=H5S_hyper_select_iterate_mem_opt(&iter,buf,type_id,space,op,operator_data); - } - else { - /* Initialize parameter block for recursive calls */ - iter_info.dt=type_id; - iter_info.elem_size=elmt_size; - iter_info.space=space; - iter_info.iter=&iter; - iter_info.src=buf; - - /* Copy the location of the region in the file */ - iter_info.op=op; - iter_info.op_data=operator_data; - - /* Call the recursive iterator routine */ - ret_value=H5S_hyper_select_iterate_mem_gen(&iter_info); - } /* end else */ - - /* Release selection iterator */ - H5S_sel_iter_release(space,&iter); - -done: - FUNC_LEAVE (ret_value); -} /* H5S_hyper_select_iterate() */ - - -/*-------------------------------------------------------------------------- - NAME H5S_hyper_release PURPOSE Release hyperslab selection information for a dataspace @@ -6364,402 +2227,6 @@ done: /*-------------------------------------------------------------------------- NAME - H5S_hyper_select_fill_gen - PURPOSE - Fill a hyperslab selection in memory with a value - USAGE - herr_t H5S_hyper_select_fill_gen(fill,fill_size,space,buf) - const void *fill; IN: Pointer to fill value to use - size_t fill_size; IN: Size of elements in memory buffer & size of - fill value - H5S_t *space; IN: Dataspace describing memory buffer & - containing selection to use. - void *buf; IN/OUT: Memory buffer to fill selection in - RETURNS - Non-negative on success/Negative on failure. - DESCRIPTION - Use the selection in the dataspace to fill elements in a memory buffer. - GLOBAL VARIABLES - COMMENTS, BUGS, ASSUMPTIONS - The memory buffer elements are assumed to have the same datatype as the - fill value being placed into them. - EXAMPLES - REVISION LOG ---------------------------------------------------------------------------*/ -static herr_t -H5S_hyper_select_fill_gen(const void *fill, size_t fill_size, const H5S_t *space, void *buf) -{ - H5S_hyper_span_info_t *spans=NULL; /* Pointer to copy of the span tree */ - H5S_hyper_span_info_t *tmp_spans; /* Temporary pointer to a span tree */ - H5S_hyper_span_t *span[H5O_LAYOUT_NDIMS]; /* Array of pointers to span nodes */ - H5S_hyper_span_t *curr_span; /* Current hyperslab span node */ - hsize_t slab[H5O_LAYOUT_NDIMS]; /* Cumulative size of each dimension in bytes */ - hssize_t abs_arr[H5O_LAYOUT_NDIMS]; /* Absolute hyperslab span position */ - hssize_t *off_arr; /* Offset within the dataspace extent */ - hsize_t acc; /* Accumulator for computing cumulative sizes */ - int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ - int curr_dim; /* Current dimension being operated on */ - int ndims; /* Number of dimensions of dataset */ - hsize_t span_io; /* Number of elements in current span to actually process */ - hsize_t num_elem; /* Number of elements in the selection */ - uint8_t *loc; /* Current element location pointer */ - int i; /* Index variable */ - unsigned u; /* Index variable */ - herr_t ret_value=SUCCEED; /* Return value */ - - FUNC_ENTER_NOINIT(H5S_hyper_select_fill_gen); - - /* Set the rank of the fastest changing dimension */ - ndims=space->extent.u.simple.rank; - fast_dim=(ndims-1); - - /* Set the pointer to the selection offset in the dataspace */ - off_arr=space->select.offset; - - /* Make a copy of the span tree to iterate over */ - spans=H5S_hyper_copy_span(space->select.sel_info.hslab.span_lst); - - /* Set the nelem & pstride values according to the element size */ - H5S_hyper_span_precompute(spans,fill_size); - - /* Build arrays of span heads & offsets in each dimension */ - for(u=0, tmp_spans=spans; u<space->extent.u.simple.rank; u++) { - /* Set the pointers to the initial span in each dimension */ - assert(tmp_spans); - assert(tmp_spans->head); - - /* Set the pointer to the first span in the list for this node */ - span[u] = tmp_spans->head; - - /* Set the initial offset to low bound of span */ - abs_arr[u]=span[u]->low; - - /* Get the pointer to the next level down */ - tmp_spans=tmp_spans->head->down; - } /* end for */ - - /* Compute sizes of "slab" in each dimension */ - for(i=fast_dim, acc=fill_size; i>=0; i--) { - slab[i]=acc; - acc*=space->extent.u.simple.size[i]; - } /* end for */ - - /* Set the offset of the first element iterated on */ - for(i=0, loc=buf; i<ndims; i++) - /* Compute the sequential element offset */ - loc+=(abs_arr[i]+off_arr[i])*slab[i]; - - /* Get the number of elements in selection */ - num_elem=space->select.num_elem; - - /* Get the pointer to the current span nodes */ - curr_span=span[fast_dim]; - - /* Go iterate over the hyperslabs */ - while(num_elem>0) { - /* Loop through the fastest dim's spans */ - while(curr_span!=NULL) { - /* Compute the number of elements to attempt in this span */ - span_io=(curr_span->high-curr_span->low)+1; - - /* Double check that things haven't gotten out of sync */ - assert(num_elem>0); - - /* Fill the elements in the current block */ - H5_CHECK_OVERFLOW(span_io,hsize_t,size_t); - H5V_array_fill(loc, fill, fill_size, (size_t)span_io); - - /* Increment the buffer offset */ - loc+=curr_span->pstride; - - /* Decrement the number of elements left */ - num_elem-=span_io; - - /* Advance span in fastest dimension */ - curr_span=curr_span->next; - } /* end while */ - - /* Check if there are more elements left */ - if(num_elem>0) { - /* Recursively advance to next offset (not necessarily span) in next dimension up */ - - /* Start at the fastest dim */ - curr_dim=fast_dim-1; - - /* Get the pointer to the correct dimension */ - curr_span=span[curr_dim]; - - /* Work back up through the dimensions */ - while(curr_dim>=0) { - /* Increment position in span */ - abs_arr[curr_dim]++; - - /* Check if we are still within the span */ - if(abs_arr[curr_dim]<=curr_span->high) { - break; - } /* end if */ - /* If we walked off that span, advance to the next span */ - else { - /* Advance span in this dimension */ - curr_span=curr_span->next; - - /* Check if we have a valid span in this dimension still */ - if(curr_span!=NULL) { - /* Reset the offset for the dim */ - abs_arr[curr_dim]=curr_span->low; - - break; - } /* end if */ - else { - /* If we finished the span list in this dimension, decrement the dimension worked on and loop again */ - curr_dim--; - - /* Reset the curr_span to the next dim */ - if(curr_dim>=0) - curr_span=span[curr_dim]; - } /* end else */ - } /* end else */ - } /* end while */ - - /* Reset the span in the current dimension */ - span[curr_dim]=curr_span; - - /* Walk back down the iterator positions, reseting them */ - while(curr_dim<fast_dim) { - assert(curr_span); - assert(curr_span->down); - assert(curr_span->down->head); - - /* Set the new span for this dimension */ - span[curr_dim+1]=curr_span->down->head; - - /* Advance span down the tree */ - curr_span=curr_span->down->head; - - /* Reset the offset for the dim */ - abs_arr[curr_dim+1]=curr_span->low; - - /* Increment current dimension */ - curr_dim++; - } /* end while */ - - /* Verify that the curr_span points to the fastest dim */ - assert(curr_span==span[fast_dim]); - - /* Verify that the offset is correct for the fastest dim */ - assert(abs_arr[fast_dim]==curr_span->low); - - /* Recompute offset in buffer */ - for(i=0, loc=buf; i<ndims; i++) - loc+=(abs_arr[i]+off_arr[i])*slab[i]; - } /* end if */ - } /* end while */ - -#ifdef LATER -done: -#endif /* LATER */ - /* Free the copy of the selections span tree */ - if(spans!=NULL) - H5S_hyper_free_span_info(spans); - - FUNC_LEAVE (ret_value); -} /* end H5S_hyper_select_fill_gen() */ - - -/*-------------------------------------------------------------------------- - NAME - H5S_hyper_select_fill_opt - PURPOSE - Fill a hyperslab selection in memory with a value - USAGE - herr_t H5S_hyper_select_fill_opt(fill,fill_size,space,buf) - const void *fill; IN: Pointer to fill value to use - size_t fill_size; IN: Size of elements in memory buffer & size of - fill value - H5S_t *space; IN: Dataspace describing memory buffer & - containing selection to use. - void *buf; IN/OUT: Memory buffer to fill selection in - RETURNS - Non-negative on success/Negative on failure. - DESCRIPTION - Use the selection in the dataspace to fill elements in a memory buffer. - GLOBAL VARIABLES - COMMENTS, BUGS, ASSUMPTIONS - The memory buffer elements are assumed to have the same datatype as the - fill value being placed into them. - EXAMPLES - REVISION LOG ---------------------------------------------------------------------------*/ -static herr_t -H5S_hyper_select_fill_opt(const void *fill, size_t fill_size, const H5S_t *space, void *buf) -{ - H5S_hyper_dim_t *diminfo; /* Alias for dataspace's diminfo information */ - hsize_t tmp_count[H5O_LAYOUT_NDIMS]; /* Temporary hyperslab counts */ - hsize_t tmp_block[H5O_LAYOUT_NDIMS]; /* Temporary hyperslab blocks */ - hsize_t slab[H5O_LAYOUT_NDIMS]; /* Size of objects in buffer */ - hsize_t acc; /* Size accumulator */ - hsize_t num_elem; /* Number of elements in the selection */ - hsize_t fast_elem; /* Block size in the fastest dimension */ - hsize_t fast_stride; /* Stride in the fastest dimension */ - hssize_t temp_off; /* Offset in a given dimension */ - uint8_t *loc; /* Current element location */ - int i; /* Counter */ - unsigned u; /* Counter */ - int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ - int temp_dim; /* Temporary rank holder */ - unsigned ndims; /* Rank of the dataspace */ - herr_t ret_value=SUCCEED; /* Return value */ - - FUNC_ENTER_NOINIT(H5S_hyper_select_fill_opt); - - /* Set some convienence values */ - ndims=space->extent.u.simple.rank; - fast_dim=ndims-1; - diminfo=space->select.sel_info.hslab.diminfo; - - /* Build the table of next-dimension down 'element' sizes */ - for(i=fast_dim, acc=fill_size; i>=0; i--) { - slab[i]=acc; - acc*=space->extent.u.simple.size[i]; - } /* end for */ - - /* Build the tables of count & block sizes as well as the initial starting location */ - for(u=0, loc=buf; u<ndims; u++) { - tmp_count[u]=diminfo[u].count; - tmp_block[u]=diminfo[u].block; - loc+=(diminfo[u].start+space->select.offset[u])*slab[u]; - } /* end for */ - - /* Get the number of elements in selection */ - num_elem=space->select.num_elem; - - /* Get the number of elements in the fastest dimension of the selection */ - fast_elem=tmp_block[fast_dim]; - fast_stride=diminfo[fast_dim].stride; - - /* Go iterate over the hyperslabs */ - while(num_elem>0) { - /* Iterate over the blocks in the fastest dimension */ - while(tmp_count[fast_dim]>0) { - /* Double check that things haven't gotten out of sync */ - assert(num_elem>0); - - /* Fill the elements in the current block */ - H5_CHECK_OVERFLOW(fast_elem,hsize_t,size_t); - H5V_array_fill(loc, fill, fill_size, (size_t)fast_elem); - - /* Advance the pointer in the memory buffer */ - loc+=(fast_stride*fill_size); - - /* Decrement the block count */ - tmp_count[fast_dim]--; - - /* Decrement the number of elements to process */ - num_elem-=fast_elem; - } /* end while */ - - /* Work on other dimensions if necessary */ - if(num_elem>0) { - /* Reset the sequence and block counts */ - tmp_count[fast_dim]=diminfo[fast_dim].count; - - /* Bubble up the decrement to the slower changing dimensions */ - temp_dim=fast_dim-1; - while(temp_dim>=0) { - /* Decrement the sequence count in this dimension */ - tmp_block[temp_dim]--; - - /* Check if we are still in the sequence */ - if(tmp_block[temp_dim]>0) - break; - - /* Reset the sequence count in this dimension */ - tmp_block[temp_dim]=diminfo[temp_dim].block; - - /* Decrement the block count */ - tmp_count[temp_dim]--; - - /* Check if we have more blocks left */ - if(tmp_count[temp_dim]>0) - break; - - /* Reset the block count in this dimension */ - tmp_count[temp_dim]=diminfo[temp_dim].count; - - /* Wrapped a dimension, go up to next dimension */ - temp_dim--; - } /* end while */ - - /* Re-compute buffer location */ - for(loc=buf,u=0; u<ndims; u++) { - temp_off=(diminfo[u].start+space->select.offset[u]) - +diminfo[u].stride*(diminfo[u].count-tmp_count[u]) - +(diminfo[u].block-tmp_block[u]); - loc+=temp_off*slab[u]; - } /* end for */ - } /* end if */ - } /* end while */ - -#ifdef LATER -done: -#endif /* LATER */ - FUNC_LEAVE (ret_value); -} /* end H5S_hyper_select_fill_opt() */ - - -/*-------------------------------------------------------------------------- - NAME - H5S_hyper_select_fill - PURPOSE - Fill a hyperslab selection in memory with a value - USAGE - herr_t H5S_hyper_select_fill(fill,fill_size,space,buf) - const void *fill; IN: Pointer to fill value to use - size_t fill_size; IN: Size of elements in memory buffer & size of - fill value - H5S_t *space; IN: Dataspace describing memory buffer & - containing selection to use. - void *buf; IN/OUT: Memory buffer to fill selection in - RETURNS - Non-negative on success/Negative on failure. - DESCRIPTION - Use the selection in the dataspace to fill elements in a memory buffer. - GLOBAL VARIABLES - COMMENTS, BUGS, ASSUMPTIONS - The memory buffer elements are assumed to have the same datatype as the - fill value being placed into them. - EXAMPLES - REVISION LOG ---------------------------------------------------------------------------*/ -herr_t -H5S_hyper_select_fill(const void *fill, size_t fill_size, const H5S_t *space, void *buf) -{ - herr_t ret_value=SUCCEED; /* return value */ - - FUNC_ENTER_NOAPI(H5S_hyper_select_fill, FAIL); - - /* Check args */ - assert(fill); - assert(fill_size>0); - assert(space); - assert(buf); - - /* Fill the selection in the memory buffer */ - - /* Check for the special case of just one H5Sselect_hyperslab call made */ - if(space->select.sel_info.hslab.diminfo!=NULL) - /* Use optimized call to fill regular hyperslab */ - ret_value=H5S_hyper_select_fill_opt(fill, fill_size, space, buf); - else - /* Call the general fill routine */ - ret_value=H5S_hyper_select_fill_gen(fill, fill_size, space, buf); - - FUNC_LEAVE (ret_value); -} /* H5S_hyper_select_fill() */ - - -/*-------------------------------------------------------------------------- - NAME H5S_hyper_recover_span PURPOSE Recover a generated span, if appropriate @@ -9263,3 +4730,1049 @@ done: FUNC_LEAVE (ret_value); } /* end H5Sselect_select() */ #endif /* NEW_HYPERSLAB_API */ /* Works */ + + +/*-------------------------------------------------------------------------- + NAME + H5S_hyper_select_get_seq_list_gen + PURPOSE + Create a list of offsets & lengths for a selection + USAGE + herr_t H5S_hyper_select_get_file_list_gen(space,iter,maxseq,nseq,off,len) + H5S_t *space; IN: Dataspace containing selection to use. + H5S_sel_iter_t *iter; IN/OUT: Selection iterator describing last + position of interest in selection. + size_t elem_size; IN: Size of an element + size_t maxseq; IN: Maximum number of sequences to generate + size_t maxbytes; IN: Maximum number of bytes to include in the + generated sequences + size_t *nseq; OUT: Actual number of sequences generated + size_t *nbytes; OUT: Actual number of bytes in sequences generated + hsize_t *off; OUT: Array of offsets + size_t *len; OUT: Array of lengths + RETURNS + Non-negative on success/Negative on failure. + DESCRIPTION + Use the selection in the dataspace to generate a list of byte offsets and + lengths for the region(s) selected. Start/Restart from the position in the + ITER parameter. The number of sequences generated is limited by the MAXSEQ + parameter and the number of sequences actually generated is stored in the + NSEQ parameter. + GLOBAL VARIABLES + COMMENTS, BUGS, ASSUMPTIONS + EXAMPLES + REVISION LOG +--------------------------------------------------------------------------*/ +static herr_t +H5S_hyper_select_get_seq_list_gen(const H5S_t *space,H5S_sel_iter_t *iter, + size_t elem_size, size_t maxseq, size_t maxbytes, size_t *nseq, size_t *nbytes, + hsize_t *off, size_t *len) +{ + H5S_hyper_span_t *curr_span; /* Current hyperslab span node */ + H5S_hyper_span_t **ispan; /* Iterator's hyperslab span nodes */ + hsize_t slab[H5O_LAYOUT_NDIMS]; /* Cumulative size of each dimension in bytes */ + hsize_t acc; /* Accumulator for computing cumulative sizes */ + hsize_t loc_off; /* Element offset in the dataspace */ + hssize_t *abs_arr; /* Absolute hyperslab span position */ + hssize_t *off_arr; /* Offset within the dataspace extent */ + size_t span_size=0; /* Number of bytes in current span to actually process */ + size_t nelem; /* Number of elements left to process */ + size_t io_bytes_left; /* Number of bytes left to process */ + size_t start_io_bytes_left; /* Initial number of bytes left to process */ + size_t curr_seq=0; /* Number of sequence/offsets stored in the arrays */ + int ndims; /* Number of dimensions of dataset */ + int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ + int curr_dim; /* Current dimension being operated on */ + int i; /* Index variable */ + herr_t ret_value=SUCCEED; /* return value */ + + FUNC_ENTER_NOINIT (H5S_hyper_select_get_seq_list_gen); + + /* Check args */ + assert(space); + assert(iter); + assert(elem_size>0); + assert(maxseq>0); + assert(maxbytes>0); + assert(nseq); + assert(nbytes); + assert(off); + assert(len); + + /* "round" off the maxbytes allowed to a multiple of the element size */ + maxbytes=(maxbytes/elem_size)*elem_size; + + /* Set the rank of the fastest changing dimension */ + ndims=space->extent.u.simple.rank; + fast_dim=(ndims-1); + + /* Get the pointers to the current span info and span nodes */ + curr_span=iter->hyp.span[fast_dim]; + abs_arr=iter->hyp.off; + off_arr=space->select.offset; + ispan=iter->hyp.span; + + /* Set the amount of elements to perform I/O on, etc. */ + start_io_bytes_left=io_bytes_left=MIN(maxbytes,(iter->hyp.elmt_left*elem_size)); + nelem=io_bytes_left/elem_size; + + /* Compute the cumulative size of dataspace dimensions */ + for(i=fast_dim, acc=elem_size; i>=0; i--) { + slab[i]=acc; + acc*=space->extent.u.simple.size[i]; + } /* end for */ + + /* Set the offset of the first element iterated on */ + for(i=0, loc_off=0; i<ndims; i++) + /* Compute the sequential element offset */ + loc_off+=(abs_arr[i]+off_arr[i])*slab[i]; + + /* Range check against number of elements left in selection */ + assert(io_bytes_left<=(iter->hyp.elmt_left*elem_size)); + + /* Take care of any partial spans leftover from previous I/Os */ + if(abs_arr[fast_dim]!=curr_span->low) { + + /* Finish the span in the fastest changing dimension */ + + /* Compute the number of bytes to attempt in this span */ + H5_ASSIGN_OVERFLOW(span_size,((curr_span->high-abs_arr[fast_dim])+1)*elem_size,hsize_t,size_t); + + /* Check number of bytes against upper bounds allowed */ + if(span_size>io_bytes_left) + span_size=io_bytes_left; + + /* Add the partial span to the list of sequences */ + off[curr_seq]=loc_off; + len[curr_seq]=span_size; + + /* Increment sequence count */ + curr_seq++; + + /* Decrement I/O left to perform */ + io_bytes_left-=span_size; + + /* Advance the hyperslab iterator */ + /* Check if we are done */ + if(io_bytes_left>0) { + /* Move to next span in fastest changing dimension */ + curr_span=curr_span->next; + + if(curr_span!=NULL) { + /* Move location offset of destination */ + loc_off+=(curr_span->low-abs_arr[fast_dim])*elem_size; + + /* Move iterator for fastest changing dimension */ + abs_arr[fast_dim]=curr_span->low; + } /* end if */ + } /* end if */ + else { + abs_arr[fast_dim]+=span_size/elem_size; + + /* Check if we are still within the span */ + if(abs_arr[fast_dim]<=curr_span->high) { + iter->hyp.span[fast_dim]=curr_span; + + goto partial_done; /* finished with partial span */ + } /* end if */ + /* If we walked off that span, advance to the next span */ + else { + /* Advance span in this dimension */ + curr_span=curr_span->next; + + /* Check if we have a valid span in this dimension still */ + if(curr_span!=NULL) { + /* Reset absolute position */ + abs_arr[fast_dim]=curr_span->low; + iter->hyp.span[fast_dim]=curr_span; + + goto partial_done; /* finished with partial span */ + } /* end if */ + } /* end else */ + } /* end else */ + + /* Adjust iterator pointers */ + + if(curr_span==NULL) { +/* Same as code in main loop */ + /* Start at the next fastest dim */ + curr_dim=fast_dim-1; + + /* Work back up through the dimensions */ + while(curr_dim>=0) { + /* Reset the current span */ + curr_span=iter->hyp.span[curr_dim]; + + /* Increment absolute position */ + abs_arr[curr_dim]++; + + /* Check if we are still within the span */ + if(abs_arr[curr_dim]<=curr_span->high) { + break; + } /* end if */ + /* If we walked off that span, advance to the next span */ + else { + /* Advance span in this dimension */ + curr_span=curr_span->next; + + /* Check if we have a valid span in this dimension still */ + if(curr_span!=NULL) { + /* Reset the span in the current dimension */ + ispan[curr_dim]=curr_span; + + /* Reset absolute position */ + abs_arr[curr_dim]=curr_span->low; + + break; + } /* end if */ + else { + /* If we finished the span list in this dimension, decrement the dimension worked on and loop again */ + curr_dim--; + } /* end else */ + } /* end else */ + } /* end while */ + + /* Check if we are finished with the spans in the tree */ + if(curr_dim<0) { + /* We had better be done with I/O or bad things are going to happen... */ + assert(io_bytes_left==0); + + goto partial_done; /* finished with partial span */ + } /* end if */ + else { + /* Walk back down the iterator positions, reseting them */ + while(curr_dim<fast_dim) { + assert(curr_span); + assert(curr_span->down); + assert(curr_span->down->head); + + /* Increment current dimension */ + curr_dim++; + + /* Set the new span_info & span for this dimension */ + iter->hyp.span[curr_dim]=curr_span->down->head; + + /* Advance span down the tree */ + curr_span=curr_span->down->head; + + /* Reset the absolute offset for the dim */ + abs_arr[curr_dim]=curr_span->low; + } /* end while */ + + /* Verify that the curr_span points to the fastest dim */ + assert(curr_span==iter->hyp.span[fast_dim]); + } /* end else */ + + /* Reset the buffer offset */ + for(i=0, loc_off=0; i<ndims; i++) + loc_off+=(abs_arr[i]+off_arr[i])*slab[i]; + } /* end if */ + } /* end if */ + +partial_done: /* Yes, goto's are evil, so sue me... :-) */ + + /* Perform the I/O on the elements, based on the position of the iterator */ + while(io_bytes_left>0 && curr_seq<maxseq) { + /* Adjust location offset of destination to compensate for initial increment below */ + loc_off-=curr_span->pstride; + + /* Loop over all the spans in the fastest changing dimension */ + while(curr_span!=NULL) { + /* Move location offset of destination */ + loc_off+=curr_span->pstride; + + /* Compute the number of elements to attempt in this span */ + H5_ASSIGN_OVERFLOW(span_size,curr_span->nelem,hsize_t,size_t); + + /* Check number of elements against upper bounds allowed */ + if(span_size>=io_bytes_left) { + /* Trim the number of bytes to output */ + span_size=io_bytes_left; + io_bytes_left=0; + +/* COMMON */ + /* Store the I/O information for the span */ + off[curr_seq]=loc_off; + len[curr_seq]=span_size; + + /* Increment the number of sequences in arrays */ + curr_seq++; + + /* If the sequence & offset arrays are full, do what? */ + if(curr_seq>=maxseq) { + /* Break out now, we are finished with sequences */ + break; + + } /* end else */ +/* end COMMON */ + + /* Break out now, we are finished with I/O */ + break; + } /* end if */ + else { + /* Decrement I/O left to perform */ + io_bytes_left-=span_size; + +/* COMMON */ + /* Store the I/O information for the span */ + off[curr_seq]=loc_off; + len[curr_seq]=span_size; + + /* Increment the number of sequences in arrays */ + curr_seq++; + + /* If the sequence & offset arrays are full, do what? */ + if(curr_seq>=maxseq) { + /* Break out now, we are finished with sequences */ + break; + } /* end else */ +/* end COMMON */ + } /* end else */ + + /* Move to next span in fastest changing dimension */ + curr_span=curr_span->next; + } /* end while */ + + /* Check if we are done */ + if(io_bytes_left==0 || curr_seq>=maxseq) { + abs_arr[fast_dim]=curr_span->low+(span_size/elem_size); + + /* Check if we are still within the span */ + if(abs_arr[fast_dim]<=curr_span->high) { + iter->hyp.span[fast_dim]=curr_span; + break; + } /* end if */ + /* If we walked off that span, advance to the next span */ + else { + /* Advance span in this dimension */ + curr_span=curr_span->next; + + /* Check if we have a valid span in this dimension still */ + if(curr_span!=NULL) { + /* Reset absolute position */ + abs_arr[fast_dim]=curr_span->low; + iter->hyp.span[fast_dim]=curr_span; + break; + } /* end if */ + } /* end else */ + } /* end if */ + + /* Adjust iterator pointers */ + + /* Start at the next fastest dim */ + curr_dim=fast_dim-1; + + /* Work back up through the dimensions */ + while(curr_dim>=0) { + /* Reset the current span */ + curr_span=iter->hyp.span[curr_dim]; + + /* Increment absolute position */ + abs_arr[curr_dim]++; + + /* Check if we are still within the span */ + if(abs_arr[curr_dim]<=curr_span->high) { + break; + } /* end if */ + /* If we walked off that span, advance to the next span */ + else { + /* Advance span in this dimension */ + curr_span=curr_span->next; + + /* Check if we have a valid span in this dimension still */ + if(curr_span!=NULL) { + /* Reset the span in the current dimension */ + ispan[curr_dim]=curr_span; + + /* Reset absolute position */ + abs_arr[curr_dim]=curr_span->low; + + break; + } /* end if */ + else { + /* If we finished the span list in this dimension, decrement the dimension worked on and loop again */ + curr_dim--; + } /* end else */ + } /* end else */ + } /* end while */ + + /* Check if we are finished with the spans in the tree */ + if(curr_dim<0) { + /* We had better be done with I/O or bad things are going to happen... */ + assert(io_bytes_left==0); + break; + } /* end if */ + else { + /* Walk back down the iterator positions, reseting them */ + while(curr_dim<fast_dim) { + assert(curr_span); + assert(curr_span->down); + assert(curr_span->down->head); + + /* Increment current dimension to the next dimension down */ + curr_dim++; + + /* Set the new span for the next dimension down */ + iter->hyp.span[curr_dim]=curr_span->down->head; + + /* Advance span down the tree */ + curr_span=curr_span->down->head; + + /* Reset the absolute offset for the dim */ + abs_arr[curr_dim]=curr_span->low; + } /* end while */ + + /* Verify that the curr_span points to the fastest dim */ + assert(curr_span==iter->hyp.span[fast_dim]); + } /* end else */ + + /* Reset the buffer offset */ + for(i=0, loc_off=0; i<ndims; i++) + loc_off+=(abs_arr[i]+off_arr[i])*slab[i]; + } /* end while */ + + /* Decrement number of elements left in iterator */ + iter->hyp.elmt_left-=(nelem-(io_bytes_left/elem_size)); + + /* Set the number of sequences generated */ + *nseq=curr_seq; + + /* Set the number of bytes used */ + *nbytes=(start_io_bytes_left-io_bytes_left); + +#ifdef LATER +done: +#endif /* LATER */ + FUNC_LEAVE (ret_value); +} /* end H5S_hyper_select_get_seq_list_gen() */ + + +/*-------------------------------------------------------------------------- + NAME + H5S_hyper_select_get_seq_list_opt + PURPOSE + Create a list of offsets & lengths for a selection + USAGE + herr_t H5S_hyper_select_get_file_list_opt(space,iter,maxseq,nseq,off,len) + H5S_t *space; IN: Dataspace containing selection to use. + H5S_sel_iter_t *iter; IN/OUT: Selection iterator describing last + position of interest in selection. + size_t elem_size; IN: Size of an element + size_t maxseq; IN: Maximum number of sequences to generate + size_t maxbytes; IN: Maximum number of bytes to include in the + generated sequences + size_t *nseq; OUT: Actual number of sequences generated + size_t *nbytes; OUT: Actual number of bytes in sequences generated + hsize_t *off; OUT: Array of offsets + size_t *len; OUT: Array of lengths + RETURNS + Non-negative on success/Negative on failure. + DESCRIPTION + Use the selection in the dataspace to generate a list of byte offsets and + lengths for the region(s) selected. Start/Restart from the position in the + ITER parameter. The number of sequences generated is limited by the MAXSEQ + parameter and the number of sequences actually generated is stored in the + NSEQ parameter. + GLOBAL VARIABLES + COMMENTS, BUGS, ASSUMPTIONS + EXAMPLES + REVISION LOG +--------------------------------------------------------------------------*/ +static herr_t +H5S_hyper_select_get_seq_list_opt(const H5S_t *space,H5S_sel_iter_t *iter, + size_t elmt_size, size_t maxseq, size_t maxbytes, size_t *nseq, size_t *nbytes, + hsize_t *off, size_t *len) +{ + hsize_t mem_size[H5O_LAYOUT_NDIMS]; /* Size of the source buffer */ + hsize_t slab[H5O_LAYOUT_NDIMS]; /* Hyperslab size */ + hssize_t *sel_off; /* Selection offset in dataspace */ + hssize_t offset[H5O_LAYOUT_NDIMS]; /* Coordinate offset in dataspace */ + hsize_t tmp_count[H5O_LAYOUT_NDIMS];/* Temporary block count */ + hsize_t tmp_block[H5O_LAYOUT_NDIMS];/* Temporary block offset */ + hssize_t wrap[H5O_LAYOUT_NDIMS]; /* Bytes to wrap around at the end of a row */ + hsize_t skip[H5O_LAYOUT_NDIMS]; /* Bytes to skip between blocks */ + const H5S_hyper_dim_t *tdiminfo; /* Temporary pointer to diminfo information */ + hssize_t fast_dim_start, /* Local copies of fastest changing dimension info */ + fast_dim_offset; + hsize_t fast_dim_stride, /* Local copies of fastest changing dimension info */ + fast_dim_block; + size_t fast_dim_buf_off; /* Local copy of amount to move fastest dimension buffer offset */ + size_t fast_dim_count; /* Number of blocks left in fastest changing dimension */ + size_t tot_blk_count; /* Total number of blocks left to output */ + size_t act_blk_count; /* Actual number of blocks to output */ + size_t total_rows; /* Total number of entire rows to output */ + size_t curr_rows; /* Current number of entire rows to output */ + int fast_dim; /* Rank of the fastest changing dimension for the dataspace */ + int temp_dim; /* Temporary rank holder */ + int ndims; /* Number of dimensions of dataset */ + hsize_t acc; /* Accumulator */ + hsize_t loc; /* Coordinate offset */ + int i; /* Local index variable */ + size_t curr_seq=0; /* Current sequence being operated on */ + size_t actual_elem; /* The actual number of elements to count */ + size_t actual_bytes;/* The actual number of bytes to copy */ + size_t nelmts; /* Starting number of elements */ + size_t io_left; /* The number of elements left in I/O operation */ + size_t start_io_left; /* The initial number of elements left in I/O operation */ +#ifndef NO_DUFFS_DEVICE + size_t duffs_index; /* Counting index for Duff's device */ +#endif /* NO_DUFFS_DEVICE */ + herr_t ret_value=SUCCEED; /* return value */ + + FUNC_ENTER_NOINIT (H5S_hyper_select_get_seq_list_opt); + + /* Check args */ + assert(space); + assert(iter); + assert(elmt_size>0); + assert(maxseq>0); + assert(maxbytes>0); + assert(nseq); + assert(nbytes); + assert(off); + assert(len); + + /* Check if this is a "flattened" regular hyperslab selection */ + if(iter->hyp.iter_rank!=0 && iter->hyp.iter_rank<space->extent.u.simple.rank) { + /* Set the aliases for a few important dimension ranks */ + ndims=iter->hyp.iter_rank; + fast_dim=ndims-1; + + /* Set the local copy of the diminfo pointer */ + tdiminfo=iter->hyp.diminfo; + + /* Set the local copy of the selection offset */ + sel_off=iter->hyp.sel_off; + + /* Set up the size of the memory space */ + HDmemcpy(mem_size, iter->hyp.size, ndims*sizeof(hsize_t)); + } /* end if */ + else { + /* Set the aliases for a few important dimension ranks */ + ndims=space->extent.u.simple.rank; + fast_dim=ndims-1; + + /* Set the local copy of the diminfo pointer */ + tdiminfo=space->select.sel_info.hslab.diminfo; + + /* Set the local copy of the selection offset */ + sel_off=space->select.offset; + + /* Set up the size of the memory space */ + HDmemcpy(mem_size, space->extent.u.simple.size, ndims*sizeof(hsize_t)); + } /* end else */ + mem_size[ndims]=elmt_size; + + /* initialize row sizes for each dimension */ + for(i=(ndims-1),acc=1; i>=0; i--) { + slab[i]=acc*elmt_size; + acc*=mem_size[i]; + } /* end for */ + + /* Get the number of elements left in the selection */ + H5_ASSIGN_OVERFLOW(io_left,iter->hyp.elmt_left,hsize_t,size_t); + + /* Calculate the number of elements to sequence through */ + start_io_left=io_left=MIN(io_left,(maxbytes/elmt_size)); + + /* Check if we stopped in the middle of a sequence of elements */ + if((iter->hyp.off[fast_dim]-tdiminfo[fast_dim].start)%tdiminfo[fast_dim].stride!=0 || + ((iter->hyp.off[fast_dim]!=tdiminfo[fast_dim].start) && tdiminfo[fast_dim].stride==1)) { + size_t leftover; /* The number of elements left over from the last sequence */ + + /* Calculate the number of elements left in the sequence */ + if(tdiminfo[fast_dim].stride==1) + leftover=tdiminfo[fast_dim].block-(iter->hyp.off[fast_dim]-tdiminfo[fast_dim].start); + else + leftover=tdiminfo[fast_dim].block-((iter->hyp.off[fast_dim]-tdiminfo[fast_dim].start)%tdiminfo[fast_dim].stride); + + /* Make certain that we don't write too many */ + actual_elem=MIN(leftover,io_left); + + /* Compute the initial buffer offset */ + for(i=0,loc=0; i<ndims; i++) + loc+=(iter->hyp.off[i]+sel_off[i])*slab[i]; + + /* Add a new sequence */ + off[curr_seq]=loc; + H5_ASSIGN_OVERFLOW(len[curr_seq],actual_elem*elmt_size,hsize_t,size_t); + + /* Increment sequence count */ + curr_seq++; + + /* Decrement the number of elements left */ + io_left -= actual_elem; + + /* Advance the hyperslab iterator */ + + /* If we had enough room to count the rest of the sequence + * in the fastest changing dimension, move the iterator offset to + * the beginning of the next block to write. Otherwise, just advance + * the iterator in the fastest changing dimension. + */ + if(actual_elem==leftover) { + /* Move iterator offset to beginning of next sequence in the fastest changing dimension */ + H5S_hyper_iter_next(space,iter); + } /* end if */ + else { + iter->hyp.off[fast_dim]+=actual_elem; /* whole sequence not written out, just advance fastest dimension offset */ + } /* end else */ + + /* Decrement the number of elements left in selection */ + iter->hyp.elmt_left-=actual_elem; + } /* end if */ + + /* Now that we've cleared the "remainder" of the previous fastest dimension + * sequence, we must be at the beginning of a sequence, so use the fancy + * algorithm to compute the offsets and run through as many as possible, + * until the buffer fills up. + */ + if(io_left>0 && curr_seq<maxseq) { /* Just in case the "remainder" above filled the buffer */ + /* Keep the number of elements we started with */ + nelmts=io_left; + + /* Compute the arrays to perform I/O on */ + /* Copy the location of the point to get */ + HDmemcpy(offset, iter->hyp.off,ndims*sizeof(hssize_t)); + + /* Add in the selection offset */ + for(i=0; i<ndims; i++) + offset[i] += sel_off[i]; + + /* Compute the current "counts" for this location */ + for(i=0; i<ndims; i++) { + if(tdiminfo[i].stride==1) { + tmp_count[i] = 0; + tmp_block[i] = iter->hyp.off[i]-tdiminfo[i].start; + } /* end if */ + else { + tmp_count[i] = (iter->hyp.off[i]-tdiminfo[i].start)/tdiminfo[i].stride; + tmp_block[i] = (iter->hyp.off[i]-tdiminfo[i].start)%tdiminfo[i].stride; + } /* end else */ + } /* end for */ + + /* Compute the initial buffer offset */ + for(i=0,loc=0; i<ndims; i++) + loc+=offset[i]*slab[i]; + + /* Set the number of elements to write each time */ + H5_ASSIGN_OVERFLOW(actual_elem,tdiminfo[fast_dim].block,hsize_t,size_t); + + /* Set the number of actual bytes */ + actual_bytes=actual_elem*elmt_size; + + /* Set local copies of information for the fastest changing dimension */ + fast_dim_start=tdiminfo[fast_dim].start; + fast_dim_stride=tdiminfo[fast_dim].stride; + fast_dim_block=tdiminfo[fast_dim].block; + H5_ASSIGN_OVERFLOW(fast_dim_buf_off,slab[fast_dim]*fast_dim_stride,hsize_t,size_t); + fast_dim_offset=fast_dim_start+sel_off[fast_dim]; + + /* Compute the number of blocks which would fit into the buffer */ + tot_blk_count=io_left/fast_dim_block; + + /* Don't go over the maximum number of sequences allowed */ + tot_blk_count=MIN(tot_blk_count,(maxseq-curr_seq)); + + /* Compute the amount to wrap at the end of each row */ + for(i=0; i<ndims; i++) + wrap[i]=(mem_size[i]-(tdiminfo[i].stride*tdiminfo[i].count))*slab[i]; + + /* Compute the amount to skip between blocks */ + for(i=0; i<ndims; i++) + skip[i]=(tdiminfo[i].stride-tdiminfo[i].block)*slab[i]; + + /* Check if there is a partial row left (with full blocks) */ + if(tmp_count[fast_dim]>0) { + /* Get number of blocks in fastest dimension */ + H5_ASSIGN_OVERFLOW(fast_dim_count,tdiminfo[fast_dim].count-tmp_count[fast_dim],hsize_t,size_t); + + /* Make certain this entire row will fit into buffer */ + fast_dim_count=MIN(fast_dim_count,tot_blk_count); + + /* Number of blocks to sequence over */ + act_blk_count=fast_dim_count; + + /* Loop over all the blocks in the fastest changing dimension */ + while(fast_dim_count>0) { + /* Store the sequence information */ + off[curr_seq]=loc; + len[curr_seq]=actual_bytes; + + /* Increment sequence count */ + curr_seq++; + + /* Increment information to reflect block just processed */ + loc+=fast_dim_buf_off; + + /* Decrement number of blocks */ + fast_dim_count--; + } /* end while */ + + /* Decrement number of elements left */ + io_left -= actual_elem*act_blk_count; + + /* Decrement number of blocks left */ + tot_blk_count -= act_blk_count; + + /* Increment information to reflect block just processed */ + tmp_count[fast_dim]+=act_blk_count; + + /* Check if we finished the entire row of blocks */ + if(tmp_count[fast_dim]>=tdiminfo[fast_dim].count) { + /* Increment offset in destination buffer */ + loc += wrap[fast_dim]; + + /* Increment information to reflect block just processed */ + offset[fast_dim]=fast_dim_offset; /* reset the offset in the fastest dimension */ + tmp_count[fast_dim]=0; + + /* Increment the offset and count for the other dimensions */ + temp_dim=fast_dim-1; + while(temp_dim>=0) { + /* Move to the next row in the curent dimension */ + offset[temp_dim]++; + tmp_block[temp_dim]++; + + /* If this block is still in the range of blocks to output for the dimension, break out of loop */ + if(tmp_block[temp_dim]<tdiminfo[temp_dim].block) + break; + else { + /* Move to the next block in the current dimension */ + offset[temp_dim]+=(tdiminfo[temp_dim].stride-tdiminfo[temp_dim].block); + loc += skip[temp_dim]; + tmp_block[temp_dim]=0; + tmp_count[temp_dim]++; + + /* If this block is still in the range of blocks to output for the dimension, break out of loop */ + if(tmp_count[temp_dim]<tdiminfo[temp_dim].count) + break; + else { + offset[temp_dim]=tdiminfo[temp_dim].start+sel_off[temp_dim]; + loc += wrap[temp_dim]; + tmp_count[temp_dim]=0; /* reset back to the beginning of the line */ + tmp_block[temp_dim]=0; + } /* end else */ + } /* end else */ + + /* Decrement dimension count */ + temp_dim--; + } /* end while */ + } /* end if */ + else { + /* Update the offset in the fastest dimension */ + offset[fast_dim]+=(fast_dim_stride*act_blk_count); + } /* end else */ + } /* end if */ + + /* Compute the number of entire rows to read in */ + curr_rows=total_rows=tot_blk_count/tdiminfo[fast_dim].count; + + /* Reset copy of number of blocks in fastest dimension */ + H5_ASSIGN_OVERFLOW(fast_dim_count,tdiminfo[fast_dim].count,hsize_t,size_t); + + /* Read in data until an entire sequence can't be written out any longer */ + while(curr_rows>0) { +#ifdef NO_DUFFS_DEVICE + /* Loop over all the blocks in the fastest changing dimension */ + while(fast_dim_count>0) { + /* Store the sequence information */ + off[curr_seq]=loc; + len[curr_seq]=actual_bytes; + + /* Increment sequence count */ + curr_seq++; + + /* Increment information to reflect block just processed */ + loc+=fast_dim_buf_off; + + /* Decrement number of blocks */ + fast_dim_count--; + } /* end while */ +#else /* NO_DUFFS_DEVICE */ + duffs_index = (fast_dim_count + 7) / 8; + switch (fast_dim_count % 8) { + case 0: + do + { + /* Store the sequence information */ + off[curr_seq]=loc; + len[curr_seq]=actual_bytes; + + /* Increment sequence count */ + curr_seq++; + + /* Increment information to reflect block just processed */ + loc+=fast_dim_buf_off; + + case 7: + /* Store the sequence information */ + off[curr_seq]=loc; + len[curr_seq]=actual_bytes; + + /* Increment sequence count */ + curr_seq++; + + /* Increment information to reflect block just processed */ + loc+=fast_dim_buf_off; + + case 6: + /* Store the sequence information */ + off[curr_seq]=loc; + len[curr_seq]=actual_bytes; + + /* Increment sequence count */ + curr_seq++; + + /* Increment information to reflect block just processed */ + loc+=fast_dim_buf_off; + + case 5: + /* Store the sequence information */ + off[curr_seq]=loc; + len[curr_seq]=actual_bytes; + + /* Increment sequence count */ + curr_seq++; + + /* Increment information to reflect block just processed */ + loc+=fast_dim_buf_off; + + case 4: + /* Store the sequence information */ + off[curr_seq]=loc; + len[curr_seq]=actual_bytes; + + /* Increment sequence count */ + curr_seq++; + + /* Increment information to reflect block just processed */ + loc+=fast_dim_buf_off; + + case 3: + /* Store the sequence information */ + off[curr_seq]=loc; + len[curr_seq]=actual_bytes; + + /* Increment sequence count */ + curr_seq++; + + /* Increment information to reflect block just processed */ + loc+=fast_dim_buf_off; + + case 2: + /* Store the sequence information */ + off[curr_seq]=loc; + len[curr_seq]=actual_bytes; + + /* Increment sequence count */ + curr_seq++; + + /* Increment information to reflect block just processed */ + loc+=fast_dim_buf_off; + + case 1: + /* Store the sequence information */ + off[curr_seq]=loc; + len[curr_seq]=actual_bytes; + + /* Increment sequence count */ + curr_seq++; + + /* Increment information to reflect block just processed */ + loc+=fast_dim_buf_off; + + } while (--duffs_index > 0); + } /* end switch */ +#endif /* NO_DUFFS_DEVICE */ + + /* Increment offset in destination buffer */ + loc += wrap[fast_dim]; + + /* Increment the offset and count for the other dimensions */ + temp_dim=fast_dim-1; + while(temp_dim>=0) { + /* Move to the next row in the curent dimension */ + offset[temp_dim]++; + tmp_block[temp_dim]++; + + /* If this block is still in the range of blocks to output for the dimension, break out of loop */ + if(tmp_block[temp_dim]<tdiminfo[temp_dim].block) + break; + else { + /* Move to the next block in the current dimension */ + offset[temp_dim]+=(tdiminfo[temp_dim].stride-tdiminfo[temp_dim].block); + loc += skip[temp_dim]; + tmp_block[temp_dim]=0; + tmp_count[temp_dim]++; + + /* If this block is still in the range of blocks to output for the dimension, break out of loop */ + if(tmp_count[temp_dim]<tdiminfo[temp_dim].count) + break; + else { + offset[temp_dim]=tdiminfo[temp_dim].start+sel_off[temp_dim]; + loc += wrap[temp_dim]; + tmp_count[temp_dim]=0; /* reset back to the beginning of the line */ + tmp_block[temp_dim]=0; + } /* end else */ + } /* end else */ + + /* Decrement dimension count */ + temp_dim--; + } /* end while */ + + /* Decrement the number of rows left */ + curr_rows--; + } /* end while */ + + /* Adjust the number of blocks & elements left to transfer */ + + /* Decrement number of elements left */ + io_left -= actual_elem*(total_rows*tdiminfo[fast_dim].count); + + /* Decrement number of blocks left */ + tot_blk_count -= (total_rows*tdiminfo[fast_dim].count); + + /* Read in partial row of blocks */ + if(io_left>0 && curr_seq<maxseq) { + /* Get remaining number of blocks left to output */ + fast_dim_count=tot_blk_count; + + /* Loop over all the blocks in the fastest changing dimension */ + while(fast_dim_count>0) { + /* Store the sequence information */ + off[curr_seq]=loc; + len[curr_seq]=actual_bytes; + + /* Increment sequence count */ + curr_seq++; + + /* Increment information to reflect block just processed */ + loc+=fast_dim_buf_off; + + /* Decrement number of blocks */ + fast_dim_count--; + } /* end while */ + + /* Decrement number of elements left */ + io_left -= actual_elem*tot_blk_count; + + /* Increment information to reflect block just processed */ + offset[fast_dim]+=(fast_dim_stride*tot_blk_count); /* move the offset in the fastest dimension */ + + /* Handle any leftover, partial blocks in this row */ + if(io_left>0 && curr_seq<maxseq) { + actual_elem=io_left; + actual_bytes=actual_elem*elmt_size; + + /* Store the sequence information */ + off[curr_seq]=loc; + len[curr_seq]=actual_bytes; + + /* Increment sequence count */ + curr_seq++; + + /* Decrement the number of elements left */ + io_left -= actual_elem; + + /* Increment buffer correctly */ + offset[fast_dim]+=actual_elem; + } /* end if */ + + /* don't bother checking slower dimensions */ + assert(io_left==0 || curr_seq==maxseq); + } /* end if */ + + + /* Update the iterator */ + + /* Subtract out the selection offset */ + for(i=0; i<ndims; i++) + offset[i] -= sel_off[i]; + + /* Update the iterator with the location we stopped */ + HDmemcpy(iter->hyp.off, offset, ndims*sizeof(hssize_t)); + + /* Decrement the number of elements left in selection */ + iter->hyp.elmt_left-=(nelmts-io_left); + } /* end if */ + + /* Set the number of sequences generated */ + *nseq=curr_seq; + + /* Set the number of bytes used */ + *nbytes=(start_io_left-io_left)*elmt_size; + +#ifdef LATER +done: +#endif /* LATER */ + FUNC_LEAVE (ret_value); +} /* end H5S_hyper_select_get_seq_list_opt() */ + + +/*-------------------------------------------------------------------------- + NAME + H5S_hyper_select_get_seq_list + PURPOSE + Create a list of offsets & lengths for a selection + USAGE + herr_t H5S_hyper_select_get_seq_list(flags,space,iter,elem_size,maxseq,maxbytes,nseq,nbytes,off,len) + unsigned flags; IN: Flags for extra information about operation + H5S_t *space; IN: Dataspace containing selection to use. + H5S_sel_iter_t *iter; IN/OUT: Selection iterator describing last + position of interest in selection. + size_t elem_size; IN: Size of an element + size_t maxseq; IN: Maximum number of sequences to generate + size_t maxbytes; IN: Maximum number of bytes to include in the + generated sequences + size_t *nseq; OUT: Actual number of sequences generated + size_t *nbytes; OUT: Actual number of bytes in sequences generated + hsize_t *off; OUT: Array of offsets + size_t *len; OUT: Array of lengths + RETURNS + Non-negative on success/Negative on failure. + DESCRIPTION + Use the selection in the dataspace to generate a list of byte offsets and + lengths for the region(s) selected. Start/Restart from the position in the + ITER parameter. The number of sequences generated is limited by the MAXSEQ + parameter and the number of sequences actually generated is stored in the + NSEQ parameter. + GLOBAL VARIABLES + COMMENTS, BUGS, ASSUMPTIONS + EXAMPLES + REVISION LOG +--------------------------------------------------------------------------*/ +herr_t +H5S_hyper_select_get_seq_list(unsigned UNUSED flags, const H5S_t *space,H5S_sel_iter_t *iter, + size_t elem_size, size_t maxseq, size_t maxbytes, size_t *nseq, size_t *nbytes, + hsize_t *off, size_t *len) +{ + herr_t ret_value=FAIL; /* return value */ + + FUNC_ENTER_NOAPI (H5S_hyper_select_get_seq_list, FAIL); + + /* Check args */ + assert(space); + assert(iter); + assert(elem_size>0); + assert(maxseq>0); + assert(maxbytes>0); + assert(nseq); + assert(nbytes); + assert(off); + assert(len); + + /* Check for the special case of just one H5Sselect_hyperslab call made */ + if(space->select.sel_info.hslab.diminfo!=NULL) + /* Use optimized call to generate sequence list */ + ret_value=H5S_hyper_select_get_seq_list_opt(space,iter,elem_size,maxseq,maxbytes,nseq,nbytes,off,len); + else + /* Call the general sequence generator routine */ + ret_value=H5S_hyper_select_get_seq_list_gen(space,iter,elem_size,maxseq,maxbytes,nseq,nbytes,off,len); + +#ifdef LATER +done: +#endif /* LATER */ + FUNC_LEAVE (ret_value); +} /* end H5S_hyper_select_get_seq_list() */ + |