* * * * * * * * * *
*  PolyBench 2.0  *
* * * * * * * * * *


* Available benchmarks:

::linear-algebra::
linear-algebra/kernels:
linear-algebra/kernels/2mm/2mm.c
linear-algebra/kernels/3mm/3mm.c
linear-algebra/kernels/atax/atax.c
linear-algebra/kernels/bicg/bicg.c
linear-algebra/kernels/cholesky/cholesky.c
linear-algebra/kernels/doitgen/doitgen.c
linear-algebra/kernels/gemm/gemm.c
linear-algebra/kernels/gemver/gemver.c
linear-algebra/kernels/gesummv/gesummv.c
linear-algebra/kernels/mvt/mvt.c
linear-algebra/kernels/symm/symm.c
linear-algebra/kernels/syr2k/syr2k.c
linear-algebra/kernels/syrk/syrk.c
linear-algebra/kernels/trisolv/trisolv.c
linear-algebra/kernels/trmm/trmm.c

linear-algebra/solvers:
linear-algebra/solvers/durbin/durbin.c
linear-algebra/solvers/dynprog/dynprog.c
linear-algebra/solvers/gramschmidt/gramschmidt.c
linear-algebra/solvers/lu/lu.c
linear-algebra/solvers/ludcmp/ludcmp.c

::datamining::
datamining/correlation/correlation.c
datamining/covariance/covariance.c

::image-processing::
image-processing/gauss-filter/gauss-filter.c
image-processing/reg_detect/reg_detect.c

::stencils::
stencils/adi/adi.c
stencils/fdtd-2d/fdtd-2d.c
stencils/fdtd-apml/fdtd-apml.c
stencils/jacobi-1d-imper/jacobi-1d-imper.c
stencils/jacobi-2d-imper/jacobi-2d-imper.c
stencils/seidel/seidel.c


* To compile a benchmark without any monitoring:

$> gcc -I utilities utilities/instrument.c linear-algebra/kernels/atax/atax.c



* To compile a benchmark with execution time reporting:

$> gcc -I utilities utilities/instrument.c linear-algebra/kernels/atax/atax.c -DPOLYBENCH_TIME


* Available options

They are all passed as macro definitions during compilation time. The
current list is:
- POLYBENCH_TIME: output execution time (gettimeofday) [default: off]
- POLYBENCH_NO_FLUSH_CACHE: don't flush the cache before calling the
  timer [default: flush the cache]
- POLYBENCH_LINUX_FIFO_SCHEDULER: use FIFO real-time scheduler for the
  kernel execution, the program must be run as root, under linux only,
  and compiled with -lc [default: off]
- POLYBENCH_CACHE_SIZE_KB: cache size to flush, in kB [default: 8192]
- POLYBENCH_TEST_MALLOC: use malloc instead of stack allocation [default: off]
- POLYBENCH_DUMP_ARRAYS: dump all live-out arrays on stderr [default: off]

- POLYBENCH_PAPI: use PAPI to get hardware counters as specified in
  utilities/papi_counters.list

- Nxx: override the default dataset values


* Collecting the execution time of all files

$> scripts/runall.sh <benchmark-category> <machine-name>

Note: the script runall must be run from the root directory of the
archive. The benchmark category is one of the following:
 - linear-algebra/kernels
 - linear-algebra/solvers
 - datamining
 - image-processing
 - stencils

Example:
$> scripts/runall.sh linear-algebra/kernels nehalem-gcc


** Specifying different dataset size for one file in particular

- Create a compiler.opts file under the directory of the benchmark
- Specify a set of options, e.g. -DX=1024 -DY=1024 (see gemm for an example)


** To specify another compiler

- Simplest is to edit scripts/runall.sh, and change the
  COMPILER_COMMAND variable. One can also export the adequate
  COMPILER_COMMAND variable in the shell environment.



* Author
Louis-Noel Pouchet <pouchet@cse.ohio-state.edu>

* Contributors
Uday Bondhugula <ubondhug@us.ibm.com>