klish/lubheap

/**
\defgroup lubheap "Little Useful Bits Heap" Library
@{
\section Overview
This library can be linked with executables to override the standard memory
management provided by the system libraries. This may provide a
performance and debugging advantage.

The implementation uses the \ref lub_heap component to control system memory
and hence provides all the features available within that memory manager.
(including leak detection)

\section VxWorks
Replaces the memPartLib.o, memLib.o and memShow.o found in the
libos.a library archive.

\subsection memlib memLib.o functions

calloc, cfree, memFindMax, memInit, memLibInit, memOptionsSet,
memPartFindMax, memPartOptionsSet, memPartRealloc, memalign, realloc, valloc


\subsection mempartlib memPartLib.o functions

free, malloc, memAddToPool, memPartAddToPool, memPartAlignedAlloc,
memPartAlloc, memPartBlockIsValid, memPartCreate, memPartFree,
memPartInit, memPartLibInit


\subsection memshow memShow.o functions

memShow, memPartInfoGet, memParShow


\section posix POSIX
Replaces the dynamic memory functionality found in the libc.a
library archive.

This implementation relies on the sbrk() function to provide memory
chunks to feed the standard heap.

\subsection Functions

calloc, cfree, free, malloc, memalign, realloc, valloc



\section Performance
Running the unittests with this library provides the following comparisons on
some test machines:

\section x86_64-unknown-linux-gnu

Native memory management (./test/mallocTest)

lubheap memory management (./test/lubMallocTest)

*** 100 in 0 milliseconds *** 200 in 0 milliseconds *** 300 in 0 milliseconds *** 400 in 0 milliseconds *** 500 in 1 milliseconds *** 1000 in 3 milliseconds *** 2000 in 7 milliseconds *** 3000 in 12 milliseconds *** 4000 in 16 milliseconds *** 5000 in 21 milliseconds *** 10000 in 48 milliseconds *** 15000 in 79 milliseconds *** 20000 in 143 milliseconds *** 25000 in 200 milliseconds *** 30000 in 253 milliseconds *** 35000 in 314 milliseconds *** 40000 in 378 milliseconds *** 45000 in 460 milliseconds *** 50000 in 561 milliseconds *** 60000 in 790 milliseconds *** 70000 in 1036 milliseconds *** 80000 in 1320 milliseconds *** 90000 in 1653 milliseconds *** 100000 in 2037 milliseconds

*** 100 in 0 milliseconds *** 200 in 0 milliseconds *** 300 in 0 milliseconds *** 400 in 1 milliseconds *** 500 in 1 milliseconds *** 1000 in 3 milliseconds *** 2000 in 6 milliseconds *** 3000 in 9 milliseconds *** 4000 in 11 milliseconds *** 5000 in 13 milliseconds *** 10000 in 37 milliseconds *** 15000 in 49 milliseconds *** 20000 in 61 milliseconds *** 25000 in 75 milliseconds *** 30000 in 89 milliseconds *** 35000 in 99 milliseconds *** 40000 in 113 milliseconds *** 45000 in 125 milliseconds *** 50000 in 138 milliseconds *** 60000 in 177 milliseconds *** 70000 in 203 milliseconds *** 80000 in 228 milliseconds *** 90000 in 254 milliseconds *** 100000 in 277 milliseconds

\section i686-pc-linux-gnu

Native memory management (./test/mallocTest)

lubheap memory management (./test/lubMallocTest)

*** 100 in 0 milliseconds *** 200 in 0 milliseconds *** 300 in 0 milliseconds *** 400 in 1 milliseconds *** 500 in 1 milliseconds *** 1000 in 3 milliseconds *** 2000 in 7 milliseconds *** 3000 in 12 milliseconds *** 4000 in 16 milliseconds *** 5000 in 21 milliseconds *** 10000 in 46 milliseconds *** 15000 in 76 milliseconds *** 20000 in 111 milliseconds *** 25000 in 151 milliseconds *** 30000 in 199 milliseconds *** 35000 in 256 milliseconds *** 40000 in 318 milliseconds *** 45000 in 393 milliseconds *** 50000 in 472 milliseconds *** 60000 in 658 milliseconds *** 70000 in 874 milliseconds *** 80000 in 1121 milliseconds *** 90000 in 1411 milliseconds *** 100000 in 1745 milliseconds

*** 100 in 0 milliseconds *** 200 in 0 milliseconds *** 300 in 1 milliseconds *** 400 in 1 milliseconds *** 500 in 1 milliseconds *** 1000 in 4 milliseconds *** 2000 in 7 milliseconds *** 3000 in 11 milliseconds *** 4000 in 13 milliseconds *** 5000 in 15 milliseconds *** 10000 in 42 milliseconds *** 15000 in 57 milliseconds *** 20000 in 72 milliseconds *** 25000 in 89 milliseconds *** 30000 in 105 milliseconds *** 35000 in 121 milliseconds *** 40000 in 136 milliseconds *** 45000 in 154 milliseconds *** 50000 in 169 milliseconds *** 60000 in 213 milliseconds *** 70000 in 246 milliseconds *** 80000 in 278 milliseconds *** 90000 in 311 milliseconds *** 100000 in 341 milliseconds

\section i686-pc-cygwin
For some reason the native memory management under cygwin is quicker. My suspcion is that there is an overhead in the
sbrk() call via the cygwin sub-system which slows the lubheap implementation...

Native memory management (./test/mallocTest)

lubheap memory management (./test/lubMallocTest)

*** 100 in 0 milliseconds *** 200 in 0 milliseconds *** 300 in 0 milliseconds *** 400 in 0 milliseconds *** 500 in 0 milliseconds *** 1000 in 0 milliseconds *** 2000 in 0 milliseconds *** 3000 in 16 milliseconds *** 4000 in 15 milliseconds *** 5000 in 16 milliseconds *** 10000 in 31 milliseconds *** 15000 in 47 milliseconds *** 20000 in 63 milliseconds *** 25000 in 93 milliseconds *** 30000 in 125 milliseconds *** 35000 in 204 milliseconds *** 40000 in 203 milliseconds *** 45000 in 218 milliseconds *** 50000 in 266 milliseconds *** 60000 in 328 milliseconds *** 70000 in 360 milliseconds *** 80000 in 406 milliseconds *** 90000 in 500 milliseconds *** 100000 in 515 milliseconds

*** 100 in 0 milliseconds *** 200 in 0 milliseconds *** 300 in 0 milliseconds *** 400 in 0 milliseconds *** 500 in 15 milliseconds *** 1000 in 0 milliseconds *** 2000 in 16 milliseconds *** 3000 in 31 milliseconds *** 4000 in 31 milliseconds *** 5000 in 47 milliseconds *** 10000 in 94 milliseconds *** 15000 in 156 milliseconds *** 20000 in 188 milliseconds *** 25000 in 250 milliseconds *** 30000 in 281 milliseconds *** 35000 in 375 milliseconds *** 40000 in 375 milliseconds *** 45000 in 453 milliseconds *** 50000 in 484 milliseconds *** 60000 in 625 milliseconds *** 70000 in 672 milliseconds *** 80000 in 813 milliseconds *** 90000 in 875 milliseconds *** 100000 in 984 milliseconds

@}
*/