Added wrappers err_fputc and err_fputs to catch failures in fput and fputs.
Macros err_putchar and err_puts call the new wrappers and can be used in
place of putchar and puts.
To avoid having to make millions of function calls when printing out
sequences, the code to print them in bwa_print_sam1 using putchar has
been replaced by a new version in bwa_print_seq that puts the sequence
into a buffer and then outputs the lot with err_fwrite. In testing, the
new code was slightly faster than the old version, with the added benefit
that it will stop promptly if IO problems are detected.
Added a new utils.c wrapper err_gzclose and changed gzclose calls to use it.
Put in some more err_fflush calls before files being written are closed.
Made err_fflush call fsync. This is useful for remote filesystems where
errors may not be reported on fflush or fclose as problems at the server
end may only be detected after they have returned. If bwa is being used
only to write to local filesystems, calling fsync is not really necessary.
To disable it, comment out #define FSYNC_ON_FLUSH in utils.c.
While we can compute smems very efficiently, there is still a long way to get
the alignment. On simulated data, this smem-based algorithm is 4X faster than
bwasw and twice as fast as bowtie2, but the accuracy is far lower than bwasw
and even lower than bowtie2 in the high-mapQ range. I am kind of sure that if
we continue to increase the mapping accuracy, the speed will approach to bwasw,
if not slower.
Smem-based mapping algorithm is still interesting, but given that I am short of
time, I will not explore it further.
Heng Li
Rob Davies
Heng Li