[igt-dev] [PATCH i-g-t v3 2/3] tools/intel_gpu_top: Add generation info in header
Tvrtko Ursulin
tvrtko.ursulin at linux.intel.com
Wed Nov 18 10:04:30 UTC 2020
On 17/11/2020 16:34, Zbigniew Kempczyński wrote:
> In multi device world we may want to see generation of device we're
> tracking counters. Add generation number/codename to be more verbose.
>
> Signed-off-by: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>
> Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> Cc: Petri Latvala <petri.latvala at intel.com>
> ---
> lib/Makefile.am | 4 +++-
> lib/meson.build | 1 +
> tools/intel_gpu_top.c | 18 ++++++++++++++++--
> 3 files changed, 20 insertions(+), 3 deletions(-)
>
> diff --git a/lib/Makefile.am b/lib/Makefile.am
> index fecf34cd..c476eeab 100644
> --- a/lib/Makefile.am
> +++ b/lib/Makefile.am
> @@ -118,7 +118,9 @@ libigt_device_scan_la_SOURCES = \
> igt_list.c \
> igt_tools_stub.c \
> igt_device_scan.c \
> - igt_device_scan.h
> + igt_device_scan.h \
> + intel_device_info.c \
> + intel_chipset.h
>
> libigt_perf_la_SOURCES = \
> igt_perf.c \
> diff --git a/lib/meson.build b/lib/meson.build
> index 090a10cd..540facb2 100644
> --- a/lib/meson.build
> +++ b/lib/meson.build
> @@ -182,6 +182,7 @@ lib_igt_device_scan_build = static_library('igt_device_scan',
> ['igt_device_scan.c',
> 'igt_list.c',
> 'igt_tools_stub.c',
> + 'intel_device_info.c',
> ],
> dependencies : scan_dep,
> include_directories : inc)
> diff --git a/tools/intel_gpu_top.c b/tools/intel_gpu_top.c
> index 86de09aa..399416c3 100644
> --- a/tools/intel_gpu_top.c
> +++ b/tools/intel_gpu_top.c
> @@ -22,6 +22,7 @@
> */
>
> #include "igt_device_scan.h"
> +#include "intel_chipset.h"
>
> #include <stdio.h>
> #include <sys/types.h>
> @@ -1028,9 +1029,12 @@ static bool print_groups(struct cnt_group **groups)
>
> static int
> print_header(const struct igt_device_card *card,
> + const struct intel_device_info *info,
> struct engines *engines, double t,
> int lines, int con_w, int con_h, bool *consumed)
> {
> + static char *codename;
> + unsigned int gen = info ? ffs(info->gen) : 0;
> struct pmu_counter fake_pmu = {
> .present = true,
> .val.cur = 1,
> @@ -1106,8 +1110,15 @@ print_header(const struct igt_device_card *card,
> if (output_mode == INTERACTIVE) {
> printf("\033[H\033[J");
>
> + if (!codename) {
> + codename = info->codename ? strdup(info->codename) :
> + strdup("unknown");
> + codename[0] = toupper(codename[0]);
> + }
> +
> if (lines++ < con_h) {
> - printf("intel-gpu-top: %s - ", card->card);
> + printf("intel-gpu-top: %s (Gen%u) @ %s - ",
> + codename, gen, card->card);
Leaks codename and lets try and avoid code duplication. And print hex
string if unknown - so same comments as in previous patch.
> if (!engines->discrete)
> printf("%s/%s MHz; %s%% RC6; %s %s; %s irqs/s\n",
> freq_items[1].buf, freq_items[0].buf,
> @@ -1317,6 +1328,7 @@ int main(int argc, char **argv)
> bool list_device = false;
> char *pmu_device, *opt_device = NULL;
> struct igt_device_card card;
> + const struct intel_device_info *info = NULL;
>
> /* Parse options */
> while ((ch = getopt(argc, argv, "o:s:d:JLlh")) != -1) {
> @@ -1441,6 +1453,8 @@ int main(int argc, char **argv)
> ret = EXIT_SUCCESS;
>
> pmu_sample(engines);
> + if (card.pci_device)
> + info = intel_get_device_info(card.pci_device);
>
> while (!stop_top) {
> bool consumed = false;
> @@ -1463,7 +1477,7 @@ int main(int argc, char **argv)
> break;
>
> while (!consumed) {
> - lines = print_header(&card, engines,
> + lines = print_header(&card, info, engines,
> t, lines, con_w, con_h,
> &consumed);
>
>
I would really prefer print_header would not need to rebuild the pretty
string every time. And also that we try and limit adding new arguments
to it. So I have this idea:
Add the pretty string to struct engines and build it once on probe.
Probably change discover_engines to take the card and move the block
which figures out the pmu_device into the function. It can then build
the pretty string and store in engines->codename / engines->drm_card or
something.
Since I am bikeshedding a lot I am happy to take over the intel_gpu_top
patch at least. But if you feel these suggestions are solid also feel
free to carry on yourself.
Regards,
Tvrtko
More information about the igt-dev
mailing list