Running tests#

Basic usage#

The simplest way to run the cuPyNumeric test suite is to use the test.py test driver script.

./test.py --color --debug

This will start test execution for the default CPU stage, as well as enabling additional debug output:

################################################################################
###
### Test Suite Configuration
###
### * Feature stages       : cpus
### * Test files per stage : 135
### * TestSystem description   : 6 cpus / 2 gpus
###
################################################################################

################################################################################
### Entering stage: CPU (with 2 workers)
################################################################################

[PASS] (CPU) examples/benchmark.py
[PASS] (CPU) examples/black_scholes.py
[PASS] (CPU) examples/cg.py
[PASS] (CPU) examples/cholesky.py
[PASS] (CPU) examples/einsum.py

After all the tests have finished a stage summary will display, including details of any test failures, as well as an overall summary for all stages combined:

[PASS] (CPU) tests/integration/test_view.py
[PASS] (CPU) tests/integration/test_vstack.py
[PASS] (CPU) tests/integration/test_where.py
[PASS] (CPU) tests/integration/test_window.py
                                CPU: Passed 135 of 135 tests (100.0%) in 196.08s

################################################################################
###
### Exiting stage: CPU
###
### * Results      : 135 / 135 files passed (100.0%)
### * Elapsed time : 0:03:16.075726
###
################################################################################

    ----------------------------------------------------------------------------

################################################################################
###
### Overall summary
###
### * CPU   : 135 / 135 passed in 196.08s
###
### All tests: Passed 135 of 135 tests (100.0%) in 196.08s
###
################################################################################

Stage configuration#

There are four possible test stages that can be run:

  • cpus runs the test stage using CPUs

  • cuda runs the test stage using GPUs

  • openmp runs the test stage using OpenMP

  • eager runs the test stage using plain Numpy

The stages are specified via the --use command line parameter, and may be combined in a comma separated list to run multiple stages, for example:

legate test.py --use=cpus,eager,cuda,openmp

There are a number of options to control the runtime configuration for different stages:

--cpus CPUS

Number of CPUs per node to use

--gpus GPUS

Number of GPUs per node to use

--fbmem FBMEM

GPU framebuffer memory (MB)

--omps OMPS

Number of OpenMP processors per node to use

--ompthreads THREADS

Number of threads per OpenMP processor

--utility UTILITY

Number of cores to reserve for runtime services

There are also options to get more verbose or color-coded terminal output:

--color

Whether to use color terminal output (if colorama is installed)

-v, --verbose

Display verbose output. Use -vv for even more output (test stdout)

--dry-run

Print the test plan but don’t run anything

--debug

Print out the commands that are to be executed

for full details see the output of test.py --help.