From: "Hogander, Jouni" <jouni.hogander@intel.com>
To: "igt-dev@lists.freedesktop.org" <igt-dev@lists.freedesktop.org>
Cc: "Brost, Matthew" <matthew.brost@intel.com>,
"Naladala, Ramanaidu" <ramanaidu.naladala@intel.com>,
"Nerlige Ramappa, Umesh" <umesh.nerlige.ramappa@intel.com>,
"De Marchi, Lucas" <lucas.demarchi@intel.com>
Subject: Re: [PATCH i-g-t v2 1/4] tests/intel/xe_drm_fdinfo: Implement virtual/parallel exec queues
Date: Tue, 10 Sep 2024 13:18:09 +0000 [thread overview]
Message-ID: <adf16abfaaad7d2c2b53e240c2c7e77d7bf1733d.camel@intel.com> (raw)
In-Reply-To: <20240910131248.334521-2-jouni.hogander@intel.com>
Hello,
I made mistake in sending patches. Please ignore this and
"lib/intel_batchbuffer: Introduce
intel_bb_create_with_context_in_region" sent by me.
Sorry for inconvenience,
Jouni Högander
On Tue, 2024-09-10 at 16:12 +0300, Jouni Högander wrote:
> From: Lucas De Marchi <lucas.demarchi@intel.com>
>
> Implement a similar function to utilization_single(), but also taking
> virtual/parallel into account. I chose some different variable names
> to make it more obvious what exactly it is testing and integrated
> with the xe_gt_fill_engines_by_class() function recently added.
> A possible refactor in the future is to make the other tests use
> this function and remove utilization_single().
>
> Based on previous patch by Umesh.
>
> Cc: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
> Reviewed-by: Matthew Brost <matthew.brost@intel.com>
> Link:
> https://lore.kernel.org/r/20240904225746.2857448-3-lucas.demarchi@intel.com
> Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
> ---
> tests/intel/xe_drm_fdinfo.c | 124
> +++++++++++++++++++++++++++++++++++-
> 1 file changed, 123 insertions(+), 1 deletion(-)
>
> diff --git a/tests/intel/xe_drm_fdinfo.c
> b/tests/intel/xe_drm_fdinfo.c
> index 8acb95040..747b6155c 100644
> --- a/tests/intel/xe_drm_fdinfo.c
> +++ b/tests/intel/xe_drm_fdinfo.c
> @@ -12,6 +12,8 @@
> #include "xe/xe_ioctl.h"
> #include "xe/xe_query.h"
> #include "xe/xe_spin.h"
> +#include "xe/xe_util.h"
> +
> /**
> * TEST: xe drm fdinfo
> * Description: Read and verify drm client memory consumption and
> engine utilization using fdinfo
> @@ -67,6 +69,8 @@ IGT_TEST_DESCRIPTION("Read and verify drm client
> memory consumption and engine u
> #define TEST_BUSY (1 << 0)
> #define TEST_TRAILING_IDLE (1 << 1)
> #define TEST_ISOLATION (1 << 2)
> +#define TEST_VIRTUAL (1 << 3)
> +#define TEST_PARALLEL (1 << 4)
>
> enum expected_load {
> EXPECTED_LOAD_IDLE,
> @@ -715,10 +719,102 @@ utilization_all_full_load(int fd)
>
> xe_vm_destroy(fd, vm);
> }
> +
> +/**
> + * SUBTEST: %s-utilization-single-idle
> + * Description: Check that each engine shows no load
> + *
> + * SUBTEST: %s-utilization-single-full-load
> + * Description: Check that each engine shows full load
> + *
> + * SUBTEST: %s-utilization-single-full-load-isolation
> + * Description: Check that each engine load does not spill over to
> other drm clients
> + *
> + * arg[1]:
> + *
> + * @virtual: virtual
> + * @parallel: parallel
> + */
> +static void
> +utilization_multi(int fd, int gt, int class, unsigned int flags)
> +{
> + struct pceu_cycles pceu[2][DRM_XE_ENGINE_CLASS_COMPUTE + 1];
> + struct pceu_cycles pceu_spill[2][DRM_XE_ENGINE_CLASS_COMPUTE
> + 1];
> + struct drm_xe_engine_class_instance
> eci[XE_MAX_ENGINE_INSTANCE];
> + struct spin_ctx *ctx = NULL;
> + enum expected_load expected_load;
> + int fd_spill, num_placements;
> + uint32_t vm;
> + bool virtual = flags & TEST_VIRTUAL;
> + bool parallel = flags & TEST_PARALLEL;
> + uint16_t width;
> +
> + igt_assert(virtual ^ parallel);
> +
> + num_placements = xe_gt_fill_engines_by_class(fd, gt, class,
> eci);
> + if (num_placements < 2)
> + return;
> +
> + if (parallel) {
> + width = num_placements;
> + num_placements = 1;
> + } else {
> + width = 1;
> + }
> +
> + if (flags & TEST_ISOLATION)
> + fd_spill = drm_reopen_driver(fd);
> +
> + vm = xe_vm_create(fd, 0, 0);
> + if (flags & TEST_BUSY) {
> + ctx = spin_ctx_init(fd, eci, vm, width,
> num_placements);
> + spin_sync_start(fd, ctx);
> + }
> +
> + read_engine_cycles(fd, pceu[0]);
> + if (flags & TEST_ISOLATION)
> + read_engine_cycles(fd_spill, pceu_spill[0]);
> +
> + usleep(batch_duration_usec);
> + if (flags & TEST_TRAILING_IDLE)
> + spin_sync_end(fd, ctx);
> +
> + read_engine_cycles(fd, pceu[1]);
> + if (flags & TEST_ISOLATION)
> + read_engine_cycles(fd_spill, pceu_spill[1]);
> +
> + expected_load = flags & TEST_BUSY ?
> + EXPECTED_LOAD_FULL : EXPECTED_LOAD_IDLE;
> + check_results(pceu[0], pceu[1], class, width, expected_load);
> +
> + if (flags & TEST_ISOLATION) {
> + /*
> + * Load from one client shouldn't spill on another,
> + * so check for idle
> + */
> + check_results(pceu_spill[0], pceu_spill[1], class,
> width,
> + EXPECTED_LOAD_IDLE);
> + close(fd_spill);
> + }
> +
> + spin_sync_end(fd, ctx);
> + spin_ctx_destroy(fd, ctx);
> +
> + xe_vm_destroy(fd, vm);
> +}
> +
> igt_main
> {
> + const struct section {
> + const char *name;
> + unsigned int flags;
> + } sections[] = {
> + { .name = "virtual", .flags = TEST_VIRTUAL },
> + { .name = "parallel", .flags = TEST_PARALLEL },
> + { }
> + };
> struct drm_xe_engine_class_instance *hwe;
> - int xe;
> + int xe, gt, class;
>
> igt_fixture {
> struct drm_client_fdinfo info = { };
> @@ -775,6 +871,32 @@ igt_main
> igt_subtest("utilization-all-full-load")
> utilization_all_full_load(xe);
>
> +
> + for (const struct section *s = sections; s->name; s++) {
> + igt_subtest_f("%s-utilization-single-idle", s->name)
> + xe_for_each_gt(xe, gt)
> + xe_for_each_engine_class(class)
> + utilization_multi(xe, gt,
> class, s->flags);
> +
> + igt_subtest_f("%s-utilization-single-full-load", s-
> >name)
> + xe_for_each_gt(xe, gt)
> + xe_for_each_engine_class(class)
> + utilization_multi(xe, gt,
> class,
> + s->flags |
> + TEST_BUSY |
> +
> TEST_TRAILING_IDLE);
> +
> + igt_subtest_f("%s-utilization-single-full-load-
> isolation",
> + s->name)
> + xe_for_each_gt(xe, gt)
> + xe_for_each_engine_class(class)
> + utilization_multi(xe, gt,
> class,
> + s->flags |
> + TEST_BUSY |
> +
> TEST_TRAILING_IDLE |
> +
> TEST_ISOLATION);
> + }
> +
> igt_fixture {
> drm_close_driver(xe);
> }
next prev parent reply other threads:[~2024-09-10 13:18 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-09-10 13:12 [PATCH i-g-t v2 0/4] PSR testing improvement by checking sink status Jouni Högander
2024-09-10 13:12 ` [PATCH i-g-t v2 1/4] tests/intel/xe_drm_fdinfo: Implement virtual/parallel exec queues Jouni Högander
2024-09-10 13:18 ` Hogander, Jouni [this message]
2024-09-10 13:12 ` [PATCH i-g-t v2 2/4] lib/intel_batchbuffer: Introduce intel_bb_create_with_context_in_region Jouni Högander
2024-09-10 13:12 ` [PATCH i-g-t v2 3/4] lib/igt_psr: Add mechanism to check sink status as well Jouni Högander
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=adf16abfaaad7d2c2b53e240c2c7e77d7bf1733d.camel@intel.com \
--to=jouni.hogander@intel.com \
--cc=igt-dev@lists.freedesktop.org \
--cc=lucas.demarchi@intel.com \
--cc=matthew.brost@intel.com \
--cc=ramanaidu.naladala@intel.com \
--cc=umesh.nerlige.ramappa@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox