[i-g-t,8/8] gputop: Basic vendor agnostic GPU top tool

Commit Message

Tvrtko Ursulin Jan. 31, 2023, 11:32 a.m. UTC

From: Tvrtko Ursulin <[e-mail]>

Rudimentary vendor agnostic example of how lib_igt_drm_clients can be used
to display a sorted by card and usage list of processes using GPUs.

Borrows a bit of code from intel_gpu_top but for now omits the fancy
features like interactive functionality, card selection, client
aggregation, sort modes, JSON output  and pretty engine names. Also no
support for global GPU or system metrics.

On the other hand it shows clients from all DRM cards which
intel_gpu_top does not do.

Signed-off-by: Tvrtko Ursulin <[e-mail]>
Cc: Rob Clark <[e-mail]>
Cc: Christian König <[e-mail]>
Acked-by: Christian König <[e-mail]>
---
 tools/gputop.c    | 260 ++++++++++++++++++++++++++++++++++++++++++++++
 tools/meson.build |   5 +
 2 files changed, 265 insertions(+)
 create mode 100644 tools/gputop.c

Comments

Kamil Konieczny Feb. 3, 2023, 4:42 p.m. UTC | #1

Hi Tvrtko,

On 2023-01-31 at 11:32:37 +0000, Tvrtko Ursulin wrote:
> From: Tvrtko Ursulin <[e-mail]>
> 
> Rudimentary vendor agnostic example of how lib_igt_drm_clients can be used
> to display a sorted by card and usage list of processes using GPUs.
> 
> Borrows a bit of code from intel_gpu_top but for now omits the fancy
> features like interactive functionality, card selection, client
> aggregation, sort modes, JSON output  and pretty engine names. Also no
> support for global GPU or system metrics.
> 
> On the other hand it shows clients from all DRM cards which
> intel_gpu_top does not do.
> 
> Signed-off-by: Tvrtko Ursulin <[e-mail]>
> Cc: Rob Clark <[e-mail]>
> Cc: Christian König <[e-mail]>
> Acked-by: Christian König <[e-mail]>

I run it with:
# ./gputop

but it do not work on my Skylake card, I see no output,
kernel 5.19.0-29-generic, ubuntu 22.10

# ./lsgpu
card0                    Intel Skylake (Gen9)              drm:/dev/dri/card0
└─renderD128                                               drm:/dev/dri/renderD128

Please add some options like debug, version, debug with high
verbose level, help. It seems like q or Q do not exit.

Regards,
Kamil

> ---
>  tools/gputop.c    | 260 ++++++++++++++++++++++++++++++++++++++++++++++
>  tools/meson.build |   5 +
>  2 files changed, 265 insertions(+)
>  create mode 100644 tools/gputop.c
> 
> diff --git a/tools/gputop.c b/tools/gputop.c
> new file mode 100644
> index 000000000000..d259cac1ab17
> --- /dev/null
> +++ b/tools/gputop.c
> @@ -0,0 +1,260 @@
> +// SPDX-License-Identifier: MIT
> +/*
> + * Copyright © 2022 Intel Corporation
> + */
> +
> +#include <assert.h>
> +#include <ctype.h>
> +#include <dirent.h>
> +#include <errno.h>
> +#include <fcntl.h>
> +#include <inttypes.h>
> +#include <limits.h>
> +#include <locale.h>
> +#include <math.h>
> +#include <poll.h>
> +#include <signal.h>
> +#include <stdint.h>
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <string.h>
> +#include <sys/ioctl.h>
> +#include <sys/stat.h>
> +#include <sys/types.h>
> +#include <unistd.h>
> +#include <termios.h>
> +#include <sys/sysmacros.h>
> +#include <stdbool.h>
> +
> +#include "igt_drm_clients.h"
> +#include "igt_drm_fdinfo.h"
> +
> +static const char *bars[] = { " ", "▏", "▎", "▍", "▌", "▋", "▊", "▉", "█" };
> +
> +static void n_spaces(const unsigned int n)
> +{
> +	unsigned int i;
> +
> +	for (i = 0; i < n; i++)
> +		putchar(' ');
> +}
> +
> +static void print_percentage_bar(double percent, int max_len)
> +{
> +	int bar_len, i, len = max_len - 2;
> +	const int w = 8;
> +
> +	assert(max_len > 0);
> +
> +	bar_len = ceil(w * percent * len / 100.0);
> +	if (bar_len > w * len)
> +		bar_len = w * len;
> +
> +	putchar('|');
> +
> +	for (i = bar_len; i >= w; i -= w)
> +		printf("%s", bars[w]);
> +	if (i)
> +		printf("%s", bars[i]);
> +
> +	len -= (bar_len + (w - 1)) / w;
> +	n_spaces(len);
> +
> +	putchar('|');
> +}
> +
> +static int
> +print_client_header(struct igt_drm_client *c, int lines, int con_w, int con_h,
> +		    int *engine_w)
> +{
> +	const char *pidname = "    PID               NAME ";
> +	int ret, len = strlen(pidname);
> +
> +	if (lines++ >= con_h || len >= con_w)
> +		return lines;
> +	printf("\033[7m");
> +	ret = printf("DRM minor %u", c->drm_minor);
> +	n_spaces(con_w - ret);
> +
> +	if (lines++ >= con_h)
> +		return lines;
> +	printf("\n%s", pidname);
> +
> +	if (c->engines->num_engines) {
> +		unsigned int i;
> +		int width;
> +
> +		*engine_w = width = (con_w - len) / c->engines->num_engines;
> +
> +		for (i = 0; i <= c->engines->max_engine_id; i++) {
> +			const char *name = c->engines->names[i];
> +			int name_len = strlen(name);
> +			int pad = (width - name_len) / 2;
> +			int spaces = width - pad - name_len;
> +
> +			if (!name)
> +				continue;
> +
> +			if (pad < 0 || spaces < 0)
> +				continue;
> +
> +			n_spaces(pad);
> +			printf("%s", name);
> +			n_spaces(spaces);
> +			len += pad + name_len + spaces;
> +		}
> +	}
> +
> +	n_spaces(con_w - len);
> +	printf("\033[0m\n");
> +
> +	return lines;
> +}
> +
> +
> +static bool
> +newheader(const struct igt_drm_client *c, const struct igt_drm_client *pc)
> +{
> +	return !pc || c->drm_minor != pc->drm_minor;
> +}
> +
> +static int
> +print_client(struct igt_drm_client *c, struct igt_drm_client **prevc,
> +	     double t, int lines, int con_w, int con_h,
> +	     unsigned int period_us, int *engine_w)
> +{
> +	unsigned int i;
> +
> +	/* Filter out idle clients. */
> +	if (!c->total_runtime || c->samples < 2)
> +		return lines;
> +
> +	/* Print header when moving to a different DRM card. */
> +	if (newheader(c, *prevc)) {
> +		lines = print_client_header(c, lines, con_w, con_h, engine_w);
> +		if (lines >= con_h)
> +			return lines;
> +	}
> +
> +	*prevc = c;
> +
> +	printf("%8u %17s ", c->pid, c->print_name);
> +	lines++;
> +
> +	for (i = 0; c->samples > 1 && i <= c->engines->max_engine_id; i++) {
> +		double pct;
> +
> +		if (!c->engines->capacity[i])
> +			continue;
> +
> +		pct = (double)c->val[i] / period_us / 1e3 * 100 /
> +		      c->engines->capacity[i];
> +
> +		/*
> +		 * Guard against fluctuations between our scanning period and
> +		 * GPU times as exported by the kernel in fdinfo.
> +		 */
> +		if (pct > 100.0)
> +			pct = 100.0;
> +
> +		print_percentage_bar(pct, *engine_w);
> +	}
> +
> +	putchar('\n');
> +
> +	return lines;
> +}
> +
> +static int
> +__client_id_cmp(const struct igt_drm_client *a,
> +		const struct igt_drm_client *b)
> +{
> +	if (a->id > b->id)
> +		return 1;
> +	else if (a->id < b->id)
> +		return -1;
> +	else
> +		return 0;
> +}
> +
> +static int client_cmp(const void *_a, const void *_b, void *unused)
> +{
> +	const struct igt_drm_client *a = _a;
> +	const struct igt_drm_client *b = _b;
> +	long val_a, val_b;
> +
> +	/* DRM cards into consecutive buckets first. */
> +	val_a = a->drm_minor;
> +	val_b = b->drm_minor;
> +	if (val_a > val_b)
> +		return 1;
> +	else if (val_b > val_a)
> +		return -1;
> +
> +	/*
> +	 * Within buckets sort by last sampling period aggregated runtime, with
> +	 * client id as a tie-breaker.
> +	 */
> +	val_a = a->last_runtime;
> +	val_b = b->last_runtime;
> +	if (val_a == val_b)
> +		return __client_id_cmp(a, b);
> +	else if (val_b > val_a)
> +		return 1;
> +	else
> +		return -1;
> +
> +}
> +
> +int main(int argc, char **argv)
> +{
> +	unsigned int period_us = 2e6;
> +	struct igt_drm_clients *clients = NULL;
> +	int con_w = -1, con_h = -1;
> +
> +	clients = igt_drm_clients_init(NULL);
> +	if (!clients)
> +		exit(1);
> +
> +	igt_drm_clients_scan(clients, NULL, NULL, 0);
> +
> +	for (;;) {
> +		struct igt_drm_client *c, *prevc = NULL;
> +		int i, engine_w = 0, lines = 0;
> +		struct winsize ws;
> +
> +		if (ioctl(0, TIOCGWINSZ, &ws) != -1) {
> +			con_w = ws.ws_col;
> +			con_h = ws.ws_row;
> +			if (con_w == 0 && con_h == 0) {
> +				/* Serial console. */
> +				con_w = 80;
> +				con_h = 24;
> +			}
> +		}
> +
> +		igt_drm_clients_scan(clients, NULL, NULL, 0);
> +		igt_drm_clients_sort(clients, client_cmp);
> +
> +		printf("\033[H\033[J");
> +
> +		igt_for_each_drm_client(clients, c, i) {
> +			assert(c->status != IGT_DRM_CLIENT_PROBE);
> +			if (c->status != IGT_DRM_CLIENT_ALIVE)
> +				break; /* Active clients are first in the array. */
> +
> +			lines = print_client(c, &prevc, (double)period_us / 1e6,
> +					     lines, con_w, con_h, period_us,
> +					     &engine_w);
> +			if (lines >= con_h)
> +				break;
> +		}
> +
> +		if (lines++ < con_h)
> +			printf("\n");
> +
> +		usleep(period_us);
> +	}
> +
> +	return 0;
> +}
> diff --git a/tools/meson.build b/tools/meson.build
> index c6194fd15daa..0a3973dee90d 100644
> --- a/tools/meson.build
> +++ b/tools/meson.build
> @@ -65,6 +65,11 @@ if libudev.found()
>  		   install : true)
>  endif
>  
> +executable('gputop', 'gputop.c',
> +           install : true,
> +           install_rpath : bindir_rpathdir,
> +           dependencies : [lib_igt_drm_clients,lib_igt_drm_fdinfo,math])
> +
>  intel_l3_parity_src = [ 'intel_l3_parity.c', 'intel_l3_udev_listener.c' ]
>  executable('intel_l3_parity', sources : intel_l3_parity_src,
>  	   dependencies : tool_deps,
> -- 
> 2.34.1
>

Tvrtko Ursulin Feb. 6, 2023, 9:19 a.m. UTC | #2

On 03/02/2023 16:42, Kamil Konieczny wrote:
> Hi Tvrtko,
> 
> On 2023-01-31 at 11:32:37 +0000, Tvrtko Ursulin wrote:
>> From: Tvrtko Ursulin <[e-mail]>
>>
>> Rudimentary vendor agnostic example of how lib_igt_drm_clients can be used
>> to display a sorted by card and usage list of processes using GPUs.
>>
>> Borrows a bit of code from intel_gpu_top but for now omits the fancy
>> features like interactive functionality, card selection, client
>> aggregation, sort modes, JSON output  and pretty engine names. Also no
>> support for global GPU or system metrics.
>>
>> On the other hand it shows clients from all DRM cards which
>> intel_gpu_top does not do.
>>
>> Signed-off-by: Tvrtko Ursulin <[e-mail]>
>> Cc: Rob Clark <[e-mail]>
>> Cc: Christian König <[e-mail]>
>> Acked-by: Christian König <[e-mail]>
> 
> I run it with:
> # ./gputop
> 
> but it do not work on my Skylake card, I see no output,
> kernel 5.19.0-29-generic, ubuntu 22.10

Odd, 5.19 should have the support. Intel_gpu_top works - it is showing 
the individual clients?

> 
> # ./lsgpu
> card0                    Intel Skylake (Gen9)              drm:/dev/dri/card0
> └─renderD128                                               drm:/dev/dri/renderD128
> 
> Please add some options like debug, version, debug with high
> verbose level, help. It seems like q or Q do not exit.

As the cover letter hints I was only set out to demonstrate an extremely 
rudimentary vendor agnostic tool. To quote the cover letter more - "..It 
also makes no effort to provide sorting modes, well any interactivity, 
or any pretty names for GPUs or engines..". I have no scope presently to 
make it better or nicer.

The tool however can serve as a starting point and people had reported 
it working as-is with a few other drivers, AMD, msm and most recently I 
believe etnaviv. So perhaps a pool of people to further improve it will 
be found there in the future.

In summary I think it's worth reviewing so that the common code gets 
extracted from intel_gpu_top into respective libraries. After that I was 
hoping other people start contributing further improvements.

Regards,

Tvrtko

Kamil Konieczny Feb. 6, 2023, 2:04 p.m. UTC | #3

Hi Tvrtko,

On 2023-02-06 at 09:19:02 +0000, Tvrtko Ursulin wrote:
> 
> On 03/02/2023 16:42, Kamil Konieczny wrote:
> > Hi Tvrtko,
> > 
> > On 2023-01-31 at 11:32:37 +0000, Tvrtko Ursulin wrote:
> > > From: Tvrtko Ursulin <[e-mail]>
> > > 
> > > Rudimentary vendor agnostic example of how lib_igt_drm_clients can be used
> > > to display a sorted by card and usage list of processes using GPUs.
> > > 
> > > Borrows a bit of code from intel_gpu_top but for now omits the fancy
> > > features like interactive functionality, card selection, client
> > > aggregation, sort modes, JSON output  and pretty engine names. Also no
> > > support for global GPU or system metrics.
> > > 
> > > On the other hand it shows clients from all DRM cards which
> > > intel_gpu_top does not do.
> > > 
> > > Signed-off-by: Tvrtko Ursulin <[e-mail]>
> > > Cc: Rob Clark <[e-mail]>
> > > Cc: Christian König <[e-mail]>
> > > Acked-by: Christian König <[e-mail]>
> > 
> > I run it with:
> > # ./gputop
> > 
> > but it do not work on my Skylake card, I see no output,
> > kernel 5.19.0-29-generic, ubuntu 22.10
> 
> Odd, 5.19 should have the support. Intel_gpu_top works - it is showing the
> individual clients?

There is no display nor any gpu app running, when I run some igt tests
then it does sometimes prints some activity. Intel_gpu_top works and
shows headers and zero activity so it mislead me.

> 
> > 
> > # ./lsgpu
> > card0                    Intel Skylake (Gen9)              drm:/dev/dri/card0
> > └─renderD128                                               drm:/dev/dri/renderD128
> > 
> > Please add some options like debug, version, debug with high
> > verbose level, help. It seems like q or Q do not exit.
> 
> As the cover letter hints I was only set out to demonstrate an extremely
> rudimentary vendor agnostic tool. To quote the cover letter more - "..It
> also makes no effort to provide sorting modes, well any interactivity, or
> any pretty names for GPUs or engines..". I have no scope presently to make
> it better or nicer.
> 
> The tool however can serve as a starting point and people had reported it
> working as-is with a few other drivers, AMD, msm and most recently I believe
> etnaviv. So perhaps a pool of people to further improve it will be found
> there in the future.
> 
> In summary I think it's worth reviewing so that the common code gets
> extracted from intel_gpu_top into respective libraries. After that I was
> hoping other people start contributing further improvements.
> 
> Regards,
> 
> Tvrtko

I agree it is good starting point, it may have something like overall
gpu activity, imho something which shows gpu % usage (mem % and power %)
for all GPUs present.

Regards,
Kamil

Rob Clark April 5, 2023, 5:57 p.m. UTC | #4

On Tue, Jan 31, 2023 at 3:33 AM Tvrtko Ursulin
<[e-mail]> wrote:
>
> From: Tvrtko Ursulin <[e-mail]>
>
> Rudimentary vendor agnostic example of how lib_igt_drm_clients can be used
> to display a sorted by card and usage list of processes using GPUs.
>
> Borrows a bit of code from intel_gpu_top but for now omits the fancy
> features like interactive functionality, card selection, client
> aggregation, sort modes, JSON output  and pretty engine names. Also no
> support for global GPU or system metrics.
>
> On the other hand it shows clients from all DRM cards which
> intel_gpu_top does not do.
>
> Signed-off-by: Tvrtko Ursulin <[e-mail]>
> Cc: Rob Clark <[e-mail]>
> Cc: Christian König <[e-mail]>
> Acked-by: Christian König <[e-mail]>

Reviewed-by: Rob Clark <[e-mail]>

> ---
>  tools/gputop.c    | 260 ++++++++++++++++++++++++++++++++++++++++++++++
>  tools/meson.build |   5 +
>  2 files changed, 265 insertions(+)
>  create mode 100644 tools/gputop.c
>
> diff --git a/tools/gputop.c b/tools/gputop.c
> new file mode 100644
> index 000000000000..d259cac1ab17
> --- /dev/null
> +++ b/tools/gputop.c
> @@ -0,0 +1,260 @@
> +// SPDX-License-Identifier: MIT
> +/*
> + * Copyright © 2022 Intel Corporation
> + */
> +
> +#include <assert.h>
> +#include <ctype.h>
> +#include <dirent.h>
> +#include <errno.h>
> +#include <fcntl.h>
> +#include <inttypes.h>
> +#include <limits.h>
> +#include <locale.h>
> +#include <math.h>
> +#include <poll.h>
> +#include <signal.h>
> +#include <stdint.h>
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <string.h>
> +#include <sys/ioctl.h>
> +#include <sys/stat.h>
> +#include <sys/types.h>
> +#include <unistd.h>
> +#include <termios.h>
> +#include <sys/sysmacros.h>
> +#include <stdbool.h>
> +
> +#include "igt_drm_clients.h"
> +#include "igt_drm_fdinfo.h"
> +
> +static const char *bars[] = { " ", "▏", "▎", "▍", "▌", "▋", "▊", "▉", "█" };
> +
> +static void n_spaces(const unsigned int n)
> +{
> +       unsigned int i;
> +
> +       for (i = 0; i < n; i++)
> +               putchar(' ');
> +}
> +
> +static void print_percentage_bar(double percent, int max_len)
> +{
> +       int bar_len, i, len = max_len - 2;
> +       const int w = 8;
> +
> +       assert(max_len > 0);
> +
> +       bar_len = ceil(w * percent * len / 100.0);
> +       if (bar_len > w * len)
> +               bar_len = w * len;
> +
> +       putchar('|');
> +
> +       for (i = bar_len; i >= w; i -= w)
> +               printf("%s", bars[w]);
> +       if (i)
> +               printf("%s", bars[i]);
> +
> +       len -= (bar_len + (w - 1)) / w;
> +       n_spaces(len);
> +
> +       putchar('|');
> +}
> +
> +static int
> +print_client_header(struct igt_drm_client *c, int lines, int con_w, int con_h,
> +                   int *engine_w)
> +{
> +       const char *pidname = "    PID               NAME ";
> +       int ret, len = strlen(pidname);
> +
> +       if (lines++ >= con_h || len >= con_w)
> +               return lines;
> +       printf("\033[7m");
> +       ret = printf("DRM minor %u", c->drm_minor);
> +       n_spaces(con_w - ret);
> +
> +       if (lines++ >= con_h)
> +               return lines;
> +       printf("\n%s", pidname);
> +
> +       if (c->engines->num_engines) {
> +               unsigned int i;
> +               int width;
> +
> +               *engine_w = width = (con_w - len) / c->engines->num_engines;
> +
> +               for (i = 0; i <= c->engines->max_engine_id; i++) {
> +                       const char *name = c->engines->names[i];
> +                       int name_len = strlen(name);
> +                       int pad = (width - name_len) / 2;
> +                       int spaces = width - pad - name_len;
> +
> +                       if (!name)
> +                               continue;
> +
> +                       if (pad < 0 || spaces < 0)
> +                               continue;
> +
> +                       n_spaces(pad);
> +                       printf("%s", name);
> +                       n_spaces(spaces);
> +                       len += pad + name_len + spaces;
> +               }
> +       }
> +
> +       n_spaces(con_w - len);
> +       printf("\033[0m\n");
> +
> +       return lines;
> +}
> +
> +
> +static bool
> +newheader(const struct igt_drm_client *c, const struct igt_drm_client *pc)
> +{
> +       return !pc || c->drm_minor != pc->drm_minor;
> +}
> +
> +static int
> +print_client(struct igt_drm_client *c, struct igt_drm_client **prevc,
> +            double t, int lines, int con_w, int con_h,
> +            unsigned int period_us, int *engine_w)
> +{
> +       unsigned int i;
> +
> +       /* Filter out idle clients. */
> +       if (!c->total_runtime || c->samples < 2)
> +               return lines;
> +
> +       /* Print header when moving to a different DRM card. */
> +       if (newheader(c, *prevc)) {
> +               lines = print_client_header(c, lines, con_w, con_h, engine_w);
> +               if (lines >= con_h)
> +                       return lines;
> +       }
> +
> +       *prevc = c;
> +
> +       printf("%8u %17s ", c->pid, c->print_name);
> +       lines++;
> +
> +       for (i = 0; c->samples > 1 && i <= c->engines->max_engine_id; i++) {
> +               double pct;
> +
> +               if (!c->engines->capacity[i])
> +                       continue;
> +
> +               pct = (double)c->val[i] / period_us / 1e3 * 100 /
> +                     c->engines->capacity[i];
> +
> +               /*
> +                * Guard against fluctuations between our scanning period and
> +                * GPU times as exported by the kernel in fdinfo.
> +                */
> +               if (pct > 100.0)
> +                       pct = 100.0;
> +
> +               print_percentage_bar(pct, *engine_w);
> +       }
> +
> +       putchar('\n');
> +
> +       return lines;
> +}
> +
> +static int
> +__client_id_cmp(const struct igt_drm_client *a,
> +               const struct igt_drm_client *b)
> +{
> +       if (a->id > b->id)
> +               return 1;
> +       else if (a->id < b->id)
> +               return -1;
> +       else
> +               return 0;
> +}
> +
> +static int client_cmp(const void *_a, const void *_b, void *unused)
> +{
> +       const struct igt_drm_client *a = _a;
> +       const struct igt_drm_client *b = _b;
> +       long val_a, val_b;
> +
> +       /* DRM cards into consecutive buckets first. */
> +       val_a = a->drm_minor;
> +       val_b = b->drm_minor;
> +       if (val_a > val_b)
> +               return 1;
> +       else if (val_b > val_a)
> +               return -1;
> +
> +       /*
> +        * Within buckets sort by last sampling period aggregated runtime, with
> +        * client id as a tie-breaker.
> +        */
> +       val_a = a->last_runtime;
> +       val_b = b->last_runtime;
> +       if (val_a == val_b)
> +               return __client_id_cmp(a, b);
> +       else if (val_b > val_a)
> +               return 1;
> +       else
> +               return -1;
> +
> +}
> +
> +int main(int argc, char **argv)
> +{
> +       unsigned int period_us = 2e6;
> +       struct igt_drm_clients *clients = NULL;
> +       int con_w = -1, con_h = -1;
> +
> +       clients = igt_drm_clients_init(NULL);
> +       if (!clients)
> +               exit(1);
> +
> +       igt_drm_clients_scan(clients, NULL, NULL, 0);
> +
> +       for (;;) {
> +               struct igt_drm_client *c, *prevc = NULL;
> +               int i, engine_w = 0, lines = 0;
> +               struct winsize ws;
> +
> +               if (ioctl(0, TIOCGWINSZ, &ws) != -1) {
> +                       con_w = ws.ws_col;
> +                       con_h = ws.ws_row;
> +                       if (con_w == 0 && con_h == 0) {
> +                               /* Serial console. */
> +                               con_w = 80;
> +                               con_h = 24;
> +                       }
> +               }
> +
> +               igt_drm_clients_scan(clients, NULL, NULL, 0);
> +               igt_drm_clients_sort(clients, client_cmp);
> +
> +               printf("\033[H\033[J");
> +
> +               igt_for_each_drm_client(clients, c, i) {
> +                       assert(c->status != IGT_DRM_CLIENT_PROBE);
> +                       if (c->status != IGT_DRM_CLIENT_ALIVE)
> +                               break; /* Active clients are first in the array. */
> +
> +                       lines = print_client(c, &prevc, (double)period_us / 1e6,
> +                                            lines, con_w, con_h, period_us,
> +                                            &engine_w);
> +                       if (lines >= con_h)
> +                               break;
> +               }
> +
> +               if (lines++ < con_h)
> +                       printf("\n");
> +
> +               usleep(period_us);
> +       }
> +
> +       return 0;
> +}
> diff --git a/tools/meson.build b/tools/meson.build
> index c6194fd15daa..0a3973dee90d 100644
> --- a/tools/meson.build
> +++ b/tools/meson.build
> @@ -65,6 +65,11 @@ if libudev.found()
>                    install : true)
>  endif
>
> +executable('gputop', 'gputop.c',
> +           install : true,
> +           install_rpath : bindir_rpathdir,
> +           dependencies : [lib_igt_drm_clients,lib_igt_drm_fdinfo,math])
> +
>  intel_l3_parity_src = [ 'intel_l3_parity.c', 'intel_l3_udev_listener.c' ]
>  executable('intel_l3_parity', sources : intel_l3_parity_src,
>            dependencies : tool_deps,
> --
> 2.34.1
>

Tvrtko Ursulin April 6, 2023, 11:08 a.m. UTC | #5

On 05/04/2023 18:57, Rob Clark wrote:
> On Tue, Jan 31, 2023 at 3:33 AM Tvrtko Ursulin
> <[e-mail]> wrote:
>>
>> From: Tvrtko Ursulin <[e-mail]>
>>
>> Rudimentary vendor agnostic example of how lib_igt_drm_clients can be used
>> to display a sorted by card and usage list of processes using GPUs.
>>
>> Borrows a bit of code from intel_gpu_top but for now omits the fancy
>> features like interactive functionality, card selection, client
>> aggregation, sort modes, JSON output  and pretty engine names. Also no
>> support for global GPU or system metrics.
>>
>> On the other hand it shows clients from all DRM cards which
>> intel_gpu_top does not do.
>>
>> Signed-off-by: Tvrtko Ursulin <[e-mail]>
>> Cc: Rob Clark <[e-mail]>
>> Cc: Christian König <[e-mail]>
>> Acked-by: Christian König <[e-mail]>
> 
> Reviewed-by: Rob Clark <[e-mail]>

Presumably for 8/8 only?

The rest of the series does not apply any more by now. I need to rebase..

Regards,

Tvrtko

> 
>> ---
>>   tools/gputop.c    | 260 ++++++++++++++++++++++++++++++++++++++++++++++
>>   tools/meson.build |   5 +
>>   2 files changed, 265 insertions(+)
>>   create mode 100644 tools/gputop.c
>>
>> diff --git a/tools/gputop.c b/tools/gputop.c
>> new file mode 100644
>> index 000000000000..d259cac1ab17
>> --- /dev/null
>> +++ b/tools/gputop.c
>> @@ -0,0 +1,260 @@
>> +// SPDX-License-Identifier: MIT
>> +/*
>> + * Copyright © 2022 Intel Corporation
>> + */
>> +
>> +#include <assert.h>
>> +#include <ctype.h>
>> +#include <dirent.h>
>> +#include <errno.h>
>> +#include <fcntl.h>
>> +#include <inttypes.h>
>> +#include <limits.h>
>> +#include <locale.h>
>> +#include <math.h>
>> +#include <poll.h>
>> +#include <signal.h>
>> +#include <stdint.h>
>> +#include <stdio.h>
>> +#include <stdlib.h>
>> +#include <string.h>
>> +#include <sys/ioctl.h>
>> +#include <sys/stat.h>
>> +#include <sys/types.h>
>> +#include <unistd.h>
>> +#include <termios.h>
>> +#include <sys/sysmacros.h>
>> +#include <stdbool.h>
>> +
>> +#include "igt_drm_clients.h"
>> +#include "igt_drm_fdinfo.h"
>> +
>> +static const char *bars[] = { " ", "▏", "▎", "▍", "▌", "▋", "▊", "▉", "█" };
>> +
>> +static void n_spaces(const unsigned int n)
>> +{
>> +       unsigned int i;
>> +
>> +       for (i = 0; i < n; i++)
>> +               putchar(' ');
>> +}
>> +
>> +static void print_percentage_bar(double percent, int max_len)
>> +{
>> +       int bar_len, i, len = max_len - 2;
>> +       const int w = 8;
>> +
>> +       assert(max_len > 0);
>> +
>> +       bar_len = ceil(w * percent * len / 100.0);
>> +       if (bar_len > w * len)
>> +               bar_len = w * len;
>> +
>> +       putchar('|');
>> +
>> +       for (i = bar_len; i >= w; i -= w)
>> +               printf("%s", bars[w]);
>> +       if (i)
>> +               printf("%s", bars[i]);
>> +
>> +       len -= (bar_len + (w - 1)) / w;
>> +       n_spaces(len);
>> +
>> +       putchar('|');
>> +}
>> +
>> +static int
>> +print_client_header(struct igt_drm_client *c, int lines, int con_w, int con_h,
>> +                   int *engine_w)
>> +{
>> +       const char *pidname = "    PID               NAME ";
>> +       int ret, len = strlen(pidname);
>> +
>> +       if (lines++ >= con_h || len >= con_w)
>> +               return lines;
>> +       printf("\033[7m");
>> +       ret = printf("DRM minor %u", c->drm_minor);
>> +       n_spaces(con_w - ret);
>> +
>> +       if (lines++ >= con_h)
>> +               return lines;
>> +       printf("\n%s", pidname);
>> +
>> +       if (c->engines->num_engines) {
>> +               unsigned int i;
>> +               int width;
>> +
>> +               *engine_w = width = (con_w - len) / c->engines->num_engines;
>> +
>> +               for (i = 0; i <= c->engines->max_engine_id; i++) {
>> +                       const char *name = c->engines->names[i];
>> +                       int name_len = strlen(name);
>> +                       int pad = (width - name_len) / 2;
>> +                       int spaces = width - pad - name_len;
>> +
>> +                       if (!name)
>> +                               continue;
>> +
>> +                       if (pad < 0 || spaces < 0)
>> +                               continue;
>> +
>> +                       n_spaces(pad);
>> +                       printf("%s", name);
>> +                       n_spaces(spaces);
>> +                       len += pad + name_len + spaces;
>> +               }
>> +       }
>> +
>> +       n_spaces(con_w - len);
>> +       printf("\033[0m\n");
>> +
>> +       return lines;
>> +}
>> +
>> +
>> +static bool
>> +newheader(const struct igt_drm_client *c, const struct igt_drm_client *pc)
>> +{
>> +       return !pc || c->drm_minor != pc->drm_minor;
>> +}
>> +
>> +static int
>> +print_client(struct igt_drm_client *c, struct igt_drm_client **prevc,
>> +            double t, int lines, int con_w, int con_h,
>> +            unsigned int period_us, int *engine_w)
>> +{
>> +       unsigned int i;
>> +
>> +       /* Filter out idle clients. */
>> +       if (!c->total_runtime || c->samples < 2)
>> +               return lines;
>> +
>> +       /* Print header when moving to a different DRM card. */
>> +       if (newheader(c, *prevc)) {
>> +               lines = print_client_header(c, lines, con_w, con_h, engine_w);
>> +               if (lines >= con_h)
>> +                       return lines;
>> +       }
>> +
>> +       *prevc = c;
>> +
>> +       printf("%8u %17s ", c->pid, c->print_name);
>> +       lines++;
>> +
>> +       for (i = 0; c->samples > 1 && i <= c->engines->max_engine_id; i++) {
>> +               double pct;
>> +
>> +               if (!c->engines->capacity[i])
>> +                       continue;
>> +
>> +               pct = (double)c->val[i] / period_us / 1e3 * 100 /
>> +                     c->engines->capacity[i];
>> +
>> +               /*
>> +                * Guard against fluctuations between our scanning period and
>> +                * GPU times as exported by the kernel in fdinfo.
>> +                */
>> +               if (pct > 100.0)
>> +                       pct = 100.0;
>> +
>> +               print_percentage_bar(pct, *engine_w);
>> +       }
>> +
>> +       putchar('\n');
>> +
>> +       return lines;
>> +}
>> +
>> +static int
>> +__client_id_cmp(const struct igt_drm_client *a,
>> +               const struct igt_drm_client *b)
>> +{
>> +       if (a->id > b->id)
>> +               return 1;
>> +       else if (a->id < b->id)
>> +               return -1;
>> +       else
>> +               return 0;
>> +}
>> +
>> +static int client_cmp(const void *_a, const void *_b, void *unused)
>> +{
>> +       const struct igt_drm_client *a = _a;
>> +       const struct igt_drm_client *b = _b;
>> +       long val_a, val_b;
>> +
>> +       /* DRM cards into consecutive buckets first. */
>> +       val_a = a->drm_minor;
>> +       val_b = b->drm_minor;
>> +       if (val_a > val_b)
>> +               return 1;
>> +       else if (val_b > val_a)
>> +               return -1;
>> +
>> +       /*
>> +        * Within buckets sort by last sampling period aggregated runtime, with
>> +        * client id as a tie-breaker.
>> +        */
>> +       val_a = a->last_runtime;
>> +       val_b = b->last_runtime;
>> +       if (val_a == val_b)
>> +               return __client_id_cmp(a, b);
>> +       else if (val_b > val_a)
>> +               return 1;
>> +       else
>> +               return -1;
>> +
>> +}
>> +
>> +int main(int argc, char **argv)
>> +{
>> +       unsigned int period_us = 2e6;
>> +       struct igt_drm_clients *clients = NULL;
>> +       int con_w = -1, con_h = -1;
>> +
>> +       clients = igt_drm_clients_init(NULL);
>> +       if (!clients)
>> +               exit(1);
>> +
>> +       igt_drm_clients_scan(clients, NULL, NULL, 0);
>> +
>> +       for (;;) {
>> +               struct igt_drm_client *c, *prevc = NULL;
>> +               int i, engine_w = 0, lines = 0;
>> +               struct winsize ws;
>> +
>> +               if (ioctl(0, TIOCGWINSZ, &ws) != -1) {
>> +                       con_w = ws.ws_col;
>> +                       con_h = ws.ws_row;
>> +                       if (con_w == 0 && con_h == 0) {
>> +                               /* Serial console. */
>> +                               con_w = 80;
>> +                               con_h = 24;
>> +                       }
>> +               }
>> +
>> +               igt_drm_clients_scan(clients, NULL, NULL, 0);
>> +               igt_drm_clients_sort(clients, client_cmp);
>> +
>> +               printf("\033[H\033[J");
>> +
>> +               igt_for_each_drm_client(clients, c, i) {
>> +                       assert(c->status != IGT_DRM_CLIENT_PROBE);
>> +                       if (c->status != IGT_DRM_CLIENT_ALIVE)
>> +                               break; /* Active clients are first in the array. */
>> +
>> +                       lines = print_client(c, &prevc, (double)period_us / 1e6,
>> +                                            lines, con_w, con_h, period_us,
>> +                                            &engine_w);
>> +                       if (lines >= con_h)
>> +                               break;
>> +               }
>> +
>> +               if (lines++ < con_h)
>> +                       printf("\n");
>> +
>> +               usleep(period_us);
>> +       }
>> +
>> +       return 0;
>> +}
>> diff --git a/tools/meson.build b/tools/meson.build
>> index c6194fd15daa..0a3973dee90d 100644
>> --- a/tools/meson.build
>> +++ b/tools/meson.build
>> @@ -65,6 +65,11 @@ if libudev.found()
>>                     install : true)
>>   endif
>>
>> +executable('gputop', 'gputop.c',
>> +           install : true,
>> +           install_rpath : bindir_rpathdir,
>> +           dependencies : [lib_igt_drm_clients,lib_igt_drm_fdinfo,math])
>> +
>>   intel_l3_parity_src = [ 'intel_l3_parity.c', 'intel_l3_udev_listener.c' ]
>>   executable('intel_l3_parity', sources : intel_l3_parity_src,
>>             dependencies : tool_deps,
>> --
>> 2.34.1
>>

Rob Clark April 6, 2023, 2:21 p.m. UTC | #6

On Thu, Apr 6, 2023 at 4:08 AM Tvrtko Ursulin
<[e-mail]> wrote:
>
>
> On 05/04/2023 18:57, Rob Clark wrote:
> > On Tue, Jan 31, 2023 at 3:33 AM Tvrtko Ursulin
> > <[e-mail]> wrote:
> >>
> >> From: Tvrtko Ursulin <[e-mail]>
> >>
> >> Rudimentary vendor agnostic example of how lib_igt_drm_clients can be used
> >> to display a sorted by card and usage list of processes using GPUs.
> >>
> >> Borrows a bit of code from intel_gpu_top but for now omits the fancy
> >> features like interactive functionality, card selection, client
> >> aggregation, sort modes, JSON output  and pretty engine names. Also no
> >> support for global GPU or system metrics.
> >>
> >> On the other hand it shows clients from all DRM cards which
> >> intel_gpu_top does not do.
> >>
> >> Signed-off-by: Tvrtko Ursulin <[e-mail]>
> >> Cc: Rob Clark <[e-mail]>
> >> Cc: Christian König <[e-mail]>
> >> Acked-by: Christian König <[e-mail]>
> >
> > Reviewed-by: Rob Clark <[e-mail]>
>
> Presumably for 8/8 only?
>
> The rest of the series does not apply any more by now. I need to rebase..

I didn't look closely at the rest of the series (was kinda assuming
that was mostly just moving things around).. but I see you rebased it
so I can take a look.

BR,
-R

> Regards,
>
> Tvrtko
>
> >
> >> ---
> >>   tools/gputop.c    | 260 ++++++++++++++++++++++++++++++++++++++++++++++
> >>   tools/meson.build |   5 +
> >>   2 files changed, 265 insertions(+)
> >>   create mode 100644 tools/gputop.c
> >>
> >> diff --git a/tools/gputop.c b/tools/gputop.c
> >> new file mode 100644
> >> index 000000000000..d259cac1ab17
> >> --- /dev/null
> >> +++ b/tools/gputop.c
> >> @@ -0,0 +1,260 @@
> >> +// SPDX-License-Identifier: MIT
> >> +/*
> >> + * Copyright © 2022 Intel Corporation
> >> + */
> >> +
> >> +#include <assert.h>
> >> +#include <ctype.h>
> >> +#include <dirent.h>
> >> +#include <errno.h>
> >> +#include <fcntl.h>
> >> +#include <inttypes.h>
> >> +#include <limits.h>
> >> +#include <locale.h>
> >> +#include <math.h>
> >> +#include <poll.h>
> >> +#include <signal.h>
> >> +#include <stdint.h>
> >> +#include <stdio.h>
> >> +#include <stdlib.h>
> >> +#include <string.h>
> >> +#include <sys/ioctl.h>
> >> +#include <sys/stat.h>
> >> +#include <sys/types.h>
> >> +#include <unistd.h>
> >> +#include <termios.h>
> >> +#include <sys/sysmacros.h>
> >> +#include <stdbool.h>
> >> +
> >> +#include "igt_drm_clients.h"
> >> +#include "igt_drm_fdinfo.h"
> >> +
> >> +static const char *bars[] = { " ", "▏", "▎", "▍", "▌", "▋", "▊", "▉", "█" };
> >> +
> >> +static void n_spaces(const unsigned int n)
> >> +{
> >> +       unsigned int i;
> >> +
> >> +       for (i = 0; i < n; i++)
> >> +               putchar(' ');
> >> +}
> >> +
> >> +static void print_percentage_bar(double percent, int max_len)
> >> +{
> >> +       int bar_len, i, len = max_len - 2;
> >> +       const int w = 8;
> >> +
> >> +       assert(max_len > 0);
> >> +
> >> +       bar_len = ceil(w * percent * len / 100.0);
> >> +       if (bar_len > w * len)
> >> +               bar_len = w * len;
> >> +
> >> +       putchar('|');
> >> +
> >> +       for (i = bar_len; i >= w; i -= w)
> >> +               printf("%s", bars[w]);
> >> +       if (i)
> >> +               printf("%s", bars[i]);
> >> +
> >> +       len -= (bar_len + (w - 1)) / w;
> >> +       n_spaces(len);
> >> +
> >> +       putchar('|');
> >> +}
> >> +
> >> +static int
> >> +print_client_header(struct igt_drm_client *c, int lines, int con_w, int con_h,
> >> +                   int *engine_w)
> >> +{
> >> +       const char *pidname = "    PID               NAME ";
> >> +       int ret, len = strlen(pidname);
> >> +
> >> +       if (lines++ >= con_h || len >= con_w)
> >> +               return lines;
> >> +       printf("\033[7m");
> >> +       ret = printf("DRM minor %u", c->drm_minor);
> >> +       n_spaces(con_w - ret);
> >> +
> >> +       if (lines++ >= con_h)
> >> +               return lines;
> >> +       printf("\n%s", pidname);
> >> +
> >> +       if (c->engines->num_engines) {
> >> +               unsigned int i;
> >> +               int width;
> >> +
> >> +               *engine_w = width = (con_w - len) / c->engines->num_engines;
> >> +
> >> +               for (i = 0; i <= c->engines->max_engine_id; i++) {
> >> +                       const char *name = c->engines->names[i];
> >> +                       int name_len = strlen(name);
> >> +                       int pad = (width - name_len) / 2;
> >> +                       int spaces = width - pad - name_len;
> >> +
> >> +                       if (!name)
> >> +                               continue;
> >> +
> >> +                       if (pad < 0 || spaces < 0)
> >> +                               continue;
> >> +
> >> +                       n_spaces(pad);
> >> +                       printf("%s", name);
> >> +                       n_spaces(spaces);
> >> +                       len += pad + name_len + spaces;
> >> +               }
> >> +       }
> >> +
> >> +       n_spaces(con_w - len);
> >> +       printf("\033[0m\n");
> >> +
> >> +       return lines;
> >> +}
> >> +
> >> +
> >> +static bool
> >> +newheader(const struct igt_drm_client *c, const struct igt_drm_client *pc)
> >> +{
> >> +       return !pc || c->drm_minor != pc->drm_minor;
> >> +}
> >> +
> >> +static int
> >> +print_client(struct igt_drm_client *c, struct igt_drm_client **prevc,
> >> +            double t, int lines, int con_w, int con_h,
> >> +            unsigned int period_us, int *engine_w)
> >> +{
> >> +       unsigned int i;
> >> +
> >> +       /* Filter out idle clients. */
> >> +       if (!c->total_runtime || c->samples < 2)
> >> +               return lines;
> >> +
> >> +       /* Print header when moving to a different DRM card. */
> >> +       if (newheader(c, *prevc)) {
> >> +               lines = print_client_header(c, lines, con_w, con_h, engine_w);
> >> +               if (lines >= con_h)
> >> +                       return lines;
> >> +       }
> >> +
> >> +       *prevc = c;
> >> +
> >> +       printf("%8u %17s ", c->pid, c->print_name);
> >> +       lines++;
> >> +
> >> +       for (i = 0; c->samples > 1 && i <= c->engines->max_engine_id; i++) {
> >> +               double pct;
> >> +
> >> +               if (!c->engines->capacity[i])
> >> +                       continue;
> >> +
> >> +               pct = (double)c->val[i] / period_us / 1e3 * 100 /
> >> +                     c->engines->capacity[i];
> >> +
> >> +               /*
> >> +                * Guard against fluctuations between our scanning period and
> >> +                * GPU times as exported by the kernel in fdinfo.
> >> +                */
> >> +               if (pct > 100.0)
> >> +                       pct = 100.0;
> >> +
> >> +               print_percentage_bar(pct, *engine_w);
> >> +       }
> >> +
> >> +       putchar('\n');
> >> +
> >> +       return lines;
> >> +}
> >> +
> >> +static int
> >> +__client_id_cmp(const struct igt_drm_client *a,
> >> +               const struct igt_drm_client *b)
> >> +{
> >> +       if (a->id > b->id)
> >> +               return 1;
> >> +       else if (a->id < b->id)
> >> +               return -1;
> >> +       else
> >> +               return 0;
> >> +}
> >> +
> >> +static int client_cmp(const void *_a, const void *_b, void *unused)
> >> +{
> >> +       const struct igt_drm_client *a = _a;
> >> +       const struct igt_drm_client *b = _b;
> >> +       long val_a, val_b;
> >> +
> >> +       /* DRM cards into consecutive buckets first. */
> >> +       val_a = a->drm_minor;
> >> +       val_b = b->drm_minor;
> >> +       if (val_a > val_b)
> >> +               return 1;
> >> +       else if (val_b > val_a)
> >> +               return -1;
> >> +
> >> +       /*
> >> +        * Within buckets sort by last sampling period aggregated runtime, with
> >> +        * client id as a tie-breaker.
> >> +        */
> >> +       val_a = a->last_runtime;
> >> +       val_b = b->last_runtime;
> >> +       if (val_a == val_b)
> >> +               return __client_id_cmp(a, b);
> >> +       else if (val_b > val_a)
> >> +               return 1;
> >> +       else
> >> +               return -1;
> >> +
> >> +}
> >> +
> >> +int main(int argc, char **argv)
> >> +{
> >> +       unsigned int period_us = 2e6;
> >> +       struct igt_drm_clients *clients = NULL;
> >> +       int con_w = -1, con_h = -1;
> >> +
> >> +       clients = igt_drm_clients_init(NULL);
> >> +       if (!clients)
> >> +               exit(1);
> >> +
> >> +       igt_drm_clients_scan(clients, NULL, NULL, 0);
> >> +
> >> +       for (;;) {
> >> +               struct igt_drm_client *c, *prevc = NULL;
> >> +               int i, engine_w = 0, lines = 0;
> >> +               struct winsize ws;
> >> +
> >> +               if (ioctl(0, TIOCGWINSZ, &ws) != -1) {
> >> +                       con_w = ws.ws_col;
> >> +                       con_h = ws.ws_row;
> >> +                       if (con_w == 0 && con_h == 0) {
> >> +                               /* Serial console. */
> >> +                               con_w = 80;
> >> +                               con_h = 24;
> >> +                       }
> >> +               }
> >> +
> >> +               igt_drm_clients_scan(clients, NULL, NULL, 0);
> >> +               igt_drm_clients_sort(clients, client_cmp);
> >> +
> >> +               printf("\033[H\033[J");
> >> +
> >> +               igt_for_each_drm_client(clients, c, i) {
> >> +                       assert(c->status != IGT_DRM_CLIENT_PROBE);
> >> +                       if (c->status != IGT_DRM_CLIENT_ALIVE)
> >> +                               break; /* Active clients are first in the array. */
> >> +
> >> +                       lines = print_client(c, &prevc, (double)period_us / 1e6,
> >> +                                            lines, con_w, con_h, period_us,
> >> +                                            &engine_w);
> >> +                       if (lines >= con_h)
> >> +                               break;
> >> +               }
> >> +
> >> +               if (lines++ < con_h)
> >> +                       printf("\n");
> >> +
> >> +               usleep(period_us);
> >> +       }
> >> +
> >> +       return 0;
> >> +}
> >> diff --git a/tools/meson.build b/tools/meson.build
> >> index c6194fd15daa..0a3973dee90d 100644
> >> --- a/tools/meson.build
> >> +++ b/tools/meson.build
> >> @@ -65,6 +65,11 @@ if libudev.found()
> >>                     install : true)
> >>   endif
> >>
> >> +executable('gputop', 'gputop.c',
> >> +           install : true,
> >> +           install_rpath : bindir_rpathdir,
> >> +           dependencies : [lib_igt_drm_clients,lib_igt_drm_fdinfo,math])
> >> +
> >>   intel_l3_parity_src = [ 'intel_l3_parity.c', 'intel_l3_udev_listener.c' ]
> >>   executable('intel_l3_parity', sources : intel_l3_parity_src,
> >>             dependencies : tool_deps,
> >> --
> >> 2.34.1
> >>

Tvrtko Ursulin April 6, 2023, 2:31 p.m. UTC | #7

On 06/04/2023 15:21, Rob Clark wrote:
> On Thu, Apr 6, 2023 at 4:08 AM Tvrtko Ursulin
> <[e-mail]> wrote:
>>
>>
>> On 05/04/2023 18:57, Rob Clark wrote:
>>> On Tue, Jan 31, 2023 at 3:33 AM Tvrtko Ursulin
>>> <[e-mail]> wrote:
>>>>
>>>> From: Tvrtko Ursulin <[e-mail]>
>>>>
>>>> Rudimentary vendor agnostic example of how lib_igt_drm_clients can be used
>>>> to display a sorted by card and usage list of processes using GPUs.
>>>>
>>>> Borrows a bit of code from intel_gpu_top but for now omits the fancy
>>>> features like interactive functionality, card selection, client
>>>> aggregation, sort modes, JSON output  and pretty engine names. Also no
>>>> support for global GPU or system metrics.
>>>>
>>>> On the other hand it shows clients from all DRM cards which
>>>> intel_gpu_top does not do.
>>>>
>>>> Signed-off-by: Tvrtko Ursulin <[e-mail]>
>>>> Cc: Rob Clark <[e-mail]>
>>>> Cc: Christian König <[e-mail]>
>>>> Acked-by: Christian König <[e-mail]>
>>>
>>> Reviewed-by: Rob Clark <[e-mail]>
>>
>> Presumably for 8/8 only?
>>
>> The rest of the series does not apply any more by now. I need to rebase..
> 
> I didn't look closely at the rest of the series (was kinda assuming
> that was mostly just moving things around).. but I see you rebased it
> so I can take a look.

There's a lot in there - first patch is extracting some code into a 
library, with the corresponding renames, but then there are six patches 
of tweaks and feature additions which finally make gputop possible.

Hopefully you can penetrate the concepts. It was all at least Valgrind 
clean back in the day I first did it.

Regards,

Tvrtko

Rob Clark May 12, 2023, 2:18 p.m. UTC | #8

On Thu, Apr 6, 2023 at 7:33 AM Tvrtko Ursulin
<[e-mail]> wrote:
>
>
> On 06/04/2023 15:21, Rob Clark wrote:
> > On Thu, Apr 6, 2023 at 4:08 AM Tvrtko Ursulin
> > <[e-mail]> wrote:
> >>
> >>
> >> On 05/04/2023 18:57, Rob Clark wrote:
> >>> On Tue, Jan 31, 2023 at 3:33 AM Tvrtko Ursulin
> >>> <[e-mail]> wrote:
> >>>>
> >>>> From: Tvrtko Ursulin <[e-mail]>
> >>>>
> >>>> Rudimentary vendor agnostic example of how lib_igt_drm_clients can be used
> >>>> to display a sorted by card and usage list of processes using GPUs.
> >>>>
> >>>> Borrows a bit of code from intel_gpu_top but for now omits the fancy
> >>>> features like interactive functionality, card selection, client
> >>>> aggregation, sort modes, JSON output  and pretty engine names. Also no
> >>>> support for global GPU or system metrics.
> >>>>
> >>>> On the other hand it shows clients from all DRM cards which
> >>>> intel_gpu_top does not do.
> >>>>
> >>>> Signed-off-by: Tvrtko Ursulin <[e-mail]>
> >>>> Cc: Rob Clark <[e-mail]>
> >>>> Cc: Christian König <[e-mail]>
> >>>> Acked-by: Christian König <[e-mail]>
> >>>
> >>> Reviewed-by: Rob Clark <[e-mail]>
> >>
> >> Presumably for 8/8 only?
> >>
> >> The rest of the series does not apply any more by now. I need to rebase..
> >
> > I didn't look closely at the rest of the series (was kinda assuming
> > that was mostly just moving things around).. but I see you rebased it
> > so I can take a look.
>
> There's a lot in there - first patch is extracting some code into a
> library, with the corresponding renames, but then there are six patches
> of tweaks and feature additions which finally make gputop possible.
>
> Hopefully you can penetrate the concepts. It was all at least Valgrind
> clean back in the day I first did it.
>

by now I've read (and rebased locally) the series, and even added a
couple things on top.. so r-b for the series, we should get this
landed

BR,
-R

Tvrtko Ursulin May 15, 2023, 11:10 a.m. UTC | #9

On 12/05/2023 15:18, Rob Clark wrote:
> On Thu, Apr 6, 2023 at 7:33 AM Tvrtko Ursulin
> <[e-mail]> wrote:
>>
>>
>> On 06/04/2023 15:21, Rob Clark wrote:
>>> On Thu, Apr 6, 2023 at 4:08 AM Tvrtko Ursulin
>>> <[e-mail]> wrote:
>>>>
>>>>
>>>> On 05/04/2023 18:57, Rob Clark wrote:
>>>>> On Tue, Jan 31, 2023 at 3:33 AM Tvrtko Ursulin
>>>>> <[e-mail]> wrote:
>>>>>>
>>>>>> From: Tvrtko Ursulin <[e-mail]>
>>>>>>
>>>>>> Rudimentary vendor agnostic example of how lib_igt_drm_clients can be used
>>>>>> to display a sorted by card and usage list of processes using GPUs.
>>>>>>
>>>>>> Borrows a bit of code from intel_gpu_top but for now omits the fancy
>>>>>> features like interactive functionality, card selection, client
>>>>>> aggregation, sort modes, JSON output  and pretty engine names. Also no
>>>>>> support for global GPU or system metrics.
>>>>>>
>>>>>> On the other hand it shows clients from all DRM cards which
>>>>>> intel_gpu_top does not do.
>>>>>>
>>>>>> Signed-off-by: Tvrtko Ursulin <[e-mail]>
>>>>>> Cc: Rob Clark <[e-mail]>
>>>>>> Cc: Christian König <[e-mail]>
>>>>>> Acked-by: Christian König <[e-mail]>
>>>>>
>>>>> Reviewed-by: Rob Clark <[e-mail]>
>>>>
>>>> Presumably for 8/8 only?
>>>>
>>>> The rest of the series does not apply any more by now. I need to rebase..
>>>
>>> I didn't look closely at the rest of the series (was kinda assuming
>>> that was mostly just moving things around).. but I see you rebased it
>>> so I can take a look.
>>
>> There's a lot in there - first patch is extracting some code into a
>> library, with the corresponding renames, but then there are six patches
>> of tweaks and feature additions which finally make gputop possible.
>>
>> Hopefully you can penetrate the concepts. It was all at least Valgrind
>> clean back in the day I first did it.
>>
> 
> by now I've read (and rebased locally) the series, and even added a
> couple things on top.. so r-b for the series, we should get this
> landed

Thanks for the reviews, pushed!

Lets see if the future holds for this small tool.

Regards,

Tvrtko

Patch

diff --git a/tools/gputop.c b/tools/gputop.c
new file mode 100644
index 000000000000..d259cac1ab17
--- /dev/null
+++ b/tools/gputop.c
@@ -0,0 +1,260 @@ 
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright © 2022 Intel Corporation
+ */
+
+#include <assert.h>
+#include <ctype.h>
+#include <dirent.h>
+#include <errno.h>
+#include <fcntl.h>
+#include <inttypes.h>
+#include <limits.h>
+#include <locale.h>
+#include <math.h>
+#include <poll.h>
+#include <signal.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/ioctl.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include <termios.h>
+#include <sys/sysmacros.h>
+#include <stdbool.h>
+
+#include "igt_drm_clients.h"
+#include "igt_drm_fdinfo.h"
+
+static const char *bars[] = { " ", "▏", "▎", "▍", "▌", "▋", "▊", "▉", "█" };
+
+static void n_spaces(const unsigned int n)
+{
+	unsigned int i;
+
+	for (i = 0; i < n; i++)
+		putchar(' ');
+}
+
+static void print_percentage_bar(double percent, int max_len)
+{
+	int bar_len, i, len = max_len - 2;
+	const int w = 8;
+
+	assert(max_len > 0);
+
+	bar_len = ceil(w * percent * len / 100.0);
+	if (bar_len > w * len)
+		bar_len = w * len;
+
+	putchar('|');
+
+	for (i = bar_len; i >= w; i -= w)
+		printf("%s", bars[w]);
+	if (i)
+		printf("%s", bars[i]);
+
+	len -= (bar_len + (w - 1)) / w;
+	n_spaces(len);
+
+	putchar('|');
+}
+
+static int
+print_client_header(struct igt_drm_client *c, int lines, int con_w, int con_h,
+		    int *engine_w)
+{
+	const char *pidname = "    PID               NAME ";
+	int ret, len = strlen(pidname);
+
+	if (lines++ >= con_h || len >= con_w)
+		return lines;
+	printf("\033[7m");
+	ret = printf("DRM minor %u", c->drm_minor);
+	n_spaces(con_w - ret);
+
+	if (lines++ >= con_h)
+		return lines;
+	printf("\n%s", pidname);
+
+	if (c->engines->num_engines) {
+		unsigned int i;
+		int width;
+
+		*engine_w = width = (con_w - len) / c->engines->num_engines;
+
+		for (i = 0; i <= c->engines->max_engine_id; i++) {
+			const char *name = c->engines->names[i];
+			int name_len = strlen(name);
+			int pad = (width - name_len) / 2;
+			int spaces = width - pad - name_len;
+
+			if (!name)
+				continue;
+
+			if (pad < 0 || spaces < 0)
+				continue;
+
+			n_spaces(pad);
+			printf("%s", name);
+			n_spaces(spaces);
+			len += pad + name_len + spaces;
+		}
+	}
+
+	n_spaces(con_w - len);
+	printf("\033[0m\n");
+
+	return lines;
+}
+
+
+static bool
+newheader(const struct igt_drm_client *c, const struct igt_drm_client *pc)
+{
+	return !pc || c->drm_minor != pc->drm_minor;
+}
+
+static int
+print_client(struct igt_drm_client *c, struct igt_drm_client **prevc,
+	     double t, int lines, int con_w, int con_h,
+	     unsigned int period_us, int *engine_w)
+{
+	unsigned int i;
+
+	/* Filter out idle clients. */
+	if (!c->total_runtime || c->samples < 2)
+		return lines;
+
+	/* Print header when moving to a different DRM card. */
+	if (newheader(c, *prevc)) {
+		lines = print_client_header(c, lines, con_w, con_h, engine_w);
+		if (lines >= con_h)
+			return lines;
+	}
+
+	*prevc = c;
+
+	printf("%8u %17s ", c->pid, c->print_name);
+	lines++;
+
+	for (i = 0; c->samples > 1 && i <= c->engines->max_engine_id; i++) {
+		double pct;
+
+		if (!c->engines->capacity[i])
+			continue;
+
+		pct = (double)c->val[i] / period_us / 1e3 * 100 /
+		      c->engines->capacity[i];
+
+		/*
+		 * Guard against fluctuations between our scanning period and
+		 * GPU times as exported by the kernel in fdinfo.
+		 */
+		if (pct > 100.0)
+			pct = 100.0;
+
+		print_percentage_bar(pct, *engine_w);
+	}
+
+	putchar('\n');
+
+	return lines;
+}
+
+static int
+__client_id_cmp(const struct igt_drm_client *a,
+		const struct igt_drm_client *b)
+{
+	if (a->id > b->id)
+		return 1;
+	else if (a->id < b->id)
+		return -1;
+	else
+		return 0;
+}
+
+static int client_cmp(const void *_a, const void *_b, void *unused)
+{
+	const struct igt_drm_client *a = _a;
+	const struct igt_drm_client *b = _b;
+	long val_a, val_b;
+
+	/* DRM cards into consecutive buckets first. */
+	val_a = a->drm_minor;
+	val_b = b->drm_minor;
+	if (val_a > val_b)
+		return 1;
+	else if (val_b > val_a)
+		return -1;
+
+	/*
+	 * Within buckets sort by last sampling period aggregated runtime, with
+	 * client id as a tie-breaker.
+	 */
+	val_a = a->last_runtime;
+	val_b = b->last_runtime;
+	if (val_a == val_b)
+		return __client_id_cmp(a, b);
+	else if (val_b > val_a)
+		return 1;
+	else
+		return -1;
+
+}
+
+int main(int argc, char **argv)
+{
+	unsigned int period_us = 2e6;
+	struct igt_drm_clients *clients = NULL;
+	int con_w = -1, con_h = -1;
+
+	clients = igt_drm_clients_init(NULL);
+	if (!clients)
+		exit(1);
+
+	igt_drm_clients_scan(clients, NULL, NULL, 0);
+
+	for (;;) {
+		struct igt_drm_client *c, *prevc = NULL;
+		int i, engine_w = 0, lines = 0;
+		struct winsize ws;
+
+		if (ioctl(0, TIOCGWINSZ, &ws) != -1) {
+			con_w = ws.ws_col;
+			con_h = ws.ws_row;
+			if (con_w == 0 && con_h == 0) {
+				/* Serial console. */
+				con_w = 80;
+				con_h = 24;
+			}
+		}
+
+		igt_drm_clients_scan(clients, NULL, NULL, 0);
+		igt_drm_clients_sort(clients, client_cmp);
+
+		printf("\033[H\033[J");
+
+		igt_for_each_drm_client(clients, c, i) {
+			assert(c->status != IGT_DRM_CLIENT_PROBE);
+			if (c->status != IGT_DRM_CLIENT_ALIVE)
+				break; /* Active clients are first in the array. */
+
+			lines = print_client(c, &prevc, (double)period_us / 1e6,
+					     lines, con_w, con_h, period_us,
+					     &engine_w);
+			if (lines >= con_h)
+				break;
+		}
+
+		if (lines++ < con_h)
+			printf("\n");
+
+		usleep(period_us);
+	}
+
+	return 0;
+}
diff --git a/tools/meson.build b/tools/meson.build
index c6194fd15daa..0a3973dee90d 100644
--- a/tools/meson.build
+++ b/tools/meson.build
@@ -65,6 +65,11 @@  if libudev.found()
 		   install : true)
 endif
 
+executable('gputop', 'gputop.c',
+           install : true,
+           install_rpath : bindir_rpathdir,
+           dependencies : [lib_igt_drm_clients,lib_igt_drm_fdinfo,math])
+
 intel_l3_parity_src = [ 'intel_l3_parity.c', 'intel_l3_udev_listener.c' ]
 executable('intel_l3_parity', sources : intel_l3_parity_src,
 	   dependencies : tool_deps,