Number of IO MPI tasks can be varied by selecting which
dimensions use parallel IO and which dimensions use Serial send to boss
I/O.
Thus can neck down from, say 1024 nodes = 4x4x8x8 to {1,8,32,64,128,256,1024} nodes
doing the I/O.
Interpolates nicely between ALL nodes write their data, a single boss per time-plane
in processor space [old UKQCD fortran code did this], and a single node doing all I/O.
Not sure I have the transfer sizes big enough and am not overly convinced fstream
is guaranteed to not give buffer inconsistencies unless I set streambuf size to zero.
Practically it has worked on 8 tasks, 2x1x2x2 writing /cloning NERSC configurations
on my MacOS + OpenMPI and Clang environment.
It is VERY easy to switch to pwrite at a later date, and also easy to send x-strips around from
each node in order to gather bigger chunks at the syscall level.
That would push us up to the circa 8x 18*4*8 == 4KB size write chunk, and by taking, say, x/y non
parallel we get to 16MB contiguous chunks written in multi 4KB transactions
per IOnode in 64^3 lattices for configuration I/O.
I suspect this is fine for system performance.
Test_serialisation has an example of *code* *free* object serialisation
to both ostream and to XML using macro magic.
Implementing TextReader/TextWriter, YAML, JSON etc.. should be trivial
and we can use configure time options to select the default "Reader" typedef.
Present done with
"using XMLPolicy::Reader"
to pick up the default serialisation strategy.
6000 matmuls CG unprec
2000 matmuls CG prec (4000 eo muls)
1050 matmuls PGCR on 16^3 x 32 x 8 m=.01
Substantial effort on timing and logging infrastructure
cut at Conjugate gradient. Also copied in Remez, Zolotarev, Chebyshev from
Mike Clark, Tony Kennedy and my BFM package respectively since we know we will
need these. I wanted the structure of
algorithms/approx
algorithms/iterative
etc.. to start taking shape.
