[Intel-gfx] [igt-dev] [PATCH igt] tests: Add a random load generator

[Chris Wilson]

Quoting Antonio Argenziano (2018-01-23 19:03:48)
> 
> 
> On 21/01/18 12:20, Chris Wilson wrote:
> > Apply a random load to one or all engines in order to apply stress to
> > RPS as it tries to constantly adjust the GPU frequency to meet the
> > changing workload.
> > 
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > ---
> >   tests/Makefile.sources |   1 +
> >   tests/gem_exec_load.c  | 124 +++++++++++++++++++++++++++++++++++++++++++++++++
> >   2 files changed, 125 insertions(+)
> >   create mode 100644 tests/gem_exec_load.c
> > 
> > diff --git a/tests/Makefile.sources b/tests/Makefile.sources
> > index 6d3857949..5c6242bca 100644
> > --- a/tests/Makefile.sources
> > +++ b/tests/Makefile.sources
> > @@ -83,6 +83,7 @@ TESTS_progs = \
> >       gem_exec_gttfill \
> >       gem_exec_latency \
> >       gem_exec_lut_handle \
> > +     gem_exec_load \
> >       gem_exec_nop \
> >       gem_exec_parallel \
> >       gem_exec_params \
> > diff --git a/tests/gem_exec_load.c b/tests/gem_exec_load.c
> > new file mode 100644
> > index 000000000..d3618f6c0
> > --- /dev/null
> > +++ b/tests/gem_exec_load.c
> > @@ -0,0 +1,124 @@
> > +/*
> > + * Copyright © 2018 Intel Corporation
> > + *
> > + * Permission is hereby granted, free of charge, to any person obtaining a
> > + * copy of this software and associated documentation files (the "Software"),
> > + * to deal in the Software without restriction, including without limitation
> > + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> > + * and/or sell copies of the Software, and to permit persons to whom the
> > + * Software is furnished to do so, subject to the following conditions:
> > + *
> > + * The above copyright notice and this permission notice (including the next
> > + * paragraph) shall be included in all copies or substantial portions of the
> > + * Software.
> > + *
> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
> > + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> > + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> > + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
> > + * IN THE SOFTWARE.
> > + */
> > +
> > +#include "igt.h"
> > +#include "igt_rand.h"
> > +#include "igt_sysfs.h"
> > +
> > +IGT_TEST_DESCRIPTION("Apply a random load to each engine");
> > +
> > +static inline uint32_t randmax(uint32_t *state, uint32_t ep_ro)
> > +{
> > +     return (uint64_t)hars_petruska_f54_1_random(state) * ep_ro >> 32;
> > +}
> > +
> > +static void one(int fd, unsigned int engine, int scale, unsigned int timeout)
> > +{
> > +     struct timespec tv = {};
> > +     uint32_t prng = engine;
> > +     igt_spin_t *spin[2] = {};
> > +     unsigned int max;
> > +     int sysfs;
> > +
> > +     sysfs = igt_sysfs_open(fd, NULL);
> > +
> > +     max = sysfs < 0 ? 1 : igt_sysfs_get_u32(sysfs, "gt_max_freq_mhz");
> > +
> > +     do {
> > +             /*
> > +              * For every second, we randomly assign a desire % busyness
> > +              * and the frequency with which to submit a batch, defining
> > +              * a pulse-width modulation (pwm).
> > +              */
> > +             unsigned int pwm = randmax(&prng, 1000);
> > +             unsigned int load = randmax(&prng, pwm / scale);
> > +             const unsigned int interval = 1e6 / pwm;
> > +             unsigned int cur =
> > +                     sysfs < 0 ? 1 : igt_sysfs_get_u32(sysfs, "gt_cur_freq_mhz");
> > +
> > +             while (pwm--) {
> > +                     if (randmax(&prng, pwm * cur) <= load * max) {
> > +                             spin[1] = __igt_spin_batch_new(fd, 0, engine, 0);
> > +                             load--;
> > +                     }
> > +                     igt_spin_batch_free(fd, spin[0]);
> > +                     spin[0] = spin[1];
> > +                     spin[1] = NULL;
> > +
> > +                     usleep(interval);
> > +             }
> > +     } while (igt_seconds_elapsed(&tv) < timeout);
> 
> Why not igt_until_timeout()?

I wasn't expecting it to be so simple and forgot about
igt_until_timeout. So no reason other than oversight.
-Chris