Skip to content


Subversion checkout URL

You can clone with
Download ZIP
Fetching contributors…
Cannot retrieve contributors at this time
173 lines (138 sloc) 5.63 KB
Modified: 1999 11 07 dh.
This is the "to-do-list" - feel free to add stuff to it! Stuff that needs
to be done on the manual is include in the .tex files using \comment{} -
feel free to add stuff there too. Its better to add too much than to little!!
Emails (or just text) describing bugs, requests for new functionality, ect
can be found in the BUGemail directory. Fell free to add stuff there!
* gcc 2.95 gives error in operator new code which returns null
(status: fixed in 0.9.01b release)
* Is the stack implementation ok? Write stack section of manual.
(status: Tavi to do as part of Block Collection (BC) implementation)
* Cleanup #define in config.h
- flags or macros?
(status: both required. Closed.)
* Logging needs to be fixed
- separate application program and TPIE internal logging
(status: Tavi added APP as 4th level of logging: order is now
TPIE errors, TPIE warnings, APP messages, TPIE debug
messages. Default is to log APP-level and above, when
logging is turned on. Logging turned off by default.
-- V0.9.01b.
To discuss: turn on logging by default, log to a permanent
file with known name, other proposals?)
* Statistics
- currently only in BTE_mmb; should be in all BTE's
- should be accessible at AMI level
(Status: Dave to propose AMI/BTE cooperation mechanism for stats,
provide specs for stats collection in other BTEs, incl. BC)
* Support for parallel disks
- some code is probably present in merge_optimized.h and Rakesh should
have a new BTE
(Status: not in next release. Low priority.)
* Clear up confusion between AMI and BTE stream types
- make uniform naming and check that all AMI/BTE pairs works similarly
- r_only flag in AMI?
(Status: Dave to make proposal.)
* Clean up device stuff re scratch files in ami_single.h (and other places)
- what does it do?
(Status: to be determined.)
* Remove PERSISTENT_READ_ONCE code (depends on Linux hack)
(Status: to be determined)
* Why isn't substream constructor a real constructer?
- also; are BTE constructors ok? (separate names)
(Status: to be determined)
* Clean up files (did we remove all unused files?)
(Status: ongoing)
* Clean up bte_test.c program:
- automatic run with set of parameters?
- flag cleanup (READ_WRITE_TEST -> UFS_TEST)
- prefetching operations?
Clean up other test programs: Standard comment header in programs
explaining purpose and operation, readable output that indicates
explicitly whether test was successful.
(Status: Dave (and others?))
* Introduce better name/mechanism for register_new
* rename mm_register.h to mm_single.h (?)
* clean up code
(Status: New manual sections written. Changes to memory manager agreed
and complete.
Dave to make corresponding changes to library code. Laura to
make corresponding changes to test programs.)
* fix problem with constructor (bad path -> crash)
(The problem is that there is no way for a BTE constructor to tell
the AMI level that the file could not be opened. The AMI level
continues and eventually causes a severe error by trying to use the
unopened file.)
(Status: Dave to propose a solution for discussion.)
* AMI_merge_base should be AMI_merge_object
(Status: to be determined)
* Why so many AMI_MM_usage flags?
(Status: to be determined)
* Function objects in AMI_merge and AMI_sort (for improved efficiency)
(Status: Dave to report.)
* Cleanup/modification of ami_optimized_merge.h
- introduction of function objects should help
- add comparator class polymorph
- add forecasting stuff (I think there is code in there already)?
(Status: Dave to report. Related to function objects investigation.)
* Distribution needs to be written
- something probably available in kb_sort stuff
(Status: to be determined)
* Add distribution sweeping
- also add spatial join example (SSSJ)
(Status: to be determined)
* Add index structures (when BTE_bcc done)
- B-trees
- R-trees
- Priority queue
(Status: undergrad student to implement?)
* Overlap last merge in sorting with computation (save scan)
- can we maybe introduce a 'piping' mechanism?
(Status: to be determined)
* BTE_STREAM_STATUS used? - if not remove!
(Status: to be determined)
* BTE_ufs cleanup
- name should be BTE_stream_ufs instead of BTE_single_disk (for
- Should we remove aio option?
- fix references to MMB in UFS code
- rename mmap_stream_header
(Status: Dave to cleanup)
* Block collection (BTE_bcc)
- code for BTE exists in bcc_base.h and block_collection.h;
needs to be cleaned-up/tested
- code for AMI (typed view, ect) needs to be written
(Status: Tavi to make proposal), implement)
* Miscellaneous:
- Fix code that allows several BTEs to coexist
- Why is BTE_mmb not called BTE_mmp?
- madvice in BTE_mmb?
(Status: What are the problems here?)
- incomplete and missing descriptions in REFERENCE and
(Status: Dave to provide missing descriptions)
- stacks
- "function objects" (performance-wise equivalent to inlined
operate methods)
- in sort and merge
- in scan, permute, etc.?
- faster in-memory procedures
- (Tavi's quicksort mods in V0.9.01b)
- cache-efficient procedures (validate via cache simulations?)
- commit to 'optimized' versions of sort and merge
- profiling to detect hot spots
* reducing uncertainty and effort on releases
- scripts
- documentation / instructions
(Status: Rajiv wrote scripts and instructions. Dave adapted scripts
and instructions while preparing release 0.9.01b)
Jump to Line
Something went wrong with that request. Please try again.