From: Ben Widawsky <ben@bwidawsk.net>
To: "O'Rourke, Tom" <Tom.O'Rourke@intel.com>
Cc: Intel GFX <intel-gfx@lists.freedesktop.org>,
"Widawsky, Benjamin" <benjamin.widawsky@intel.com>
Subject: Re: [PATCH] [v2] intel_frequency: A tool to manipulate Intel GPU frequency
Date: Tue, 13 Jan 2015 14:36:56 -0800 [thread overview]
Message-ID: <20150113223655.GA26112@bwidawsk.net> (raw)
In-Reply-To: <411E5DC12A51ED4CB1159E14310E532B9EB6C5CB@FMSMSX119.amr.corp.intel.com>
On Tue, Jan 13, 2015 at 09:19:04PM +0000, O'Rourke, Tom wrote:
> >Sent: Sunday, January 11, 2015 7:48 PM
> >To: Widawsky, Benjamin
> >Cc: Intel GFX
> >Subject: Re: [Intel-gfx] [PATCH] [v2] intel_frequency: A tool to manipulate Intel
> >GPU frequency
> >
> >On Sun, Jan 11, 2015 at 07:35:21PM -0800, Ben Widawsky wrote:
> >> WARNING: very minimally tested
> >>
> >> In general you should not need this tool. It's primary purpose is for
> >> benchmarking, and for debugging performance issues.
> >
> >I noticed the "it's" vs "its" on v1, but forgot to fix it. IT'S fixed locally
> >though.
> >
> >>
> >> For many kernel releases now sysfs has supported reading and writing the GPU
> >> frequency. Therefore, this tool provides no new functionality. What it does
> >> provide is an easy to package (for distros) tool that handles the most common
> >> scenarios.
> [TOR:] This is a nice tool.
> I am concerned that this tool may be confusing RP1 frequency with RPe (Efficient) frequency. On many platforms, these are not the same thing.
> Thanks,
> Tom O'Rourke
>
Any platform other than BYT/CHV?
>> + case 'e': /* efficient */
>> + if (IS_VALLEYVIEW(devid) || IS_CHERRYVIEW(devid)) {
>> + /* the LP parts have special efficient >> frequencies
>*/
>> + fprintf(stderr,
>> + "FIXME: Warning efficient frequency >information is incorrect.\n");
>> + exit(EXIT_FAILURE);
>> + }
> >>
> >> v2:
> >> Get rid of -f from the usage message (Jordan)
> >> Add space before [-s (Jordan)
> >> Add a -c/--custom example (Jordan)
> >> Add a setting for resetting to hardware default (Ken)
> >> Replicate examples in commit message in the source code. (me)
> >>
> >> Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
> >> Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
> >> Cc: Kenneth Graunke <kenneth@whitecape.org>
> >>
> >> Here are some sample usages:
> >> $ sudo intel_frequency --get=cur,min,max,eff
> >> cur: 200 MHz
> >> min: 200 MHz
> >> RP1: 200 MHz
> >> max: 1200 MHz
> >>
> >> $ sudo intel_frequency -g
> >> cur: 200 MHz
> >> min: 200 MHz
> >> RP1: 200 MHz
> >> max: 1200 MHz
> >>
> >> $ sudo intel_frequency -geff
> >> RP1: 200 MHz
> >>
> >> $ sudo intel_frequency --set min=300
> >> $ sudo intel_frequency --get min
> >> cur: 300 MHz
> >> min: 300 MHz
> >> RP1: 200 MHz
> >> max: 1200 MHz
> >>
> >> $ sudo intel_frequency --custom max=900
> >> $ sudo intel_frequency --get max
> >> cur: 300 MHz
> >> min: 300 MHz
> >> RP1: 200 MHz
> >> max: 900 MHz
> >>
> >> $ sudo intel_frequency --max
> >> $ sudo intel_frequency -g
> >> cur: 1200 MHz
> >> min: 1200 MHz
> >> RP1: 200 MHz
> >> max: 1200 MHz
> >>
> >> $ sudo intel_frequency -e
> >> $ sudo intel_frequency -g
> >> cur: 200 MHz
> >> min: 200 MHz
> >> RP1: 200 MHz
> >> max: 200 MHz
> >>
> >> $ sudo intel_frequency --max
> >> $ sudo intel_frequency -g
> >> cur: 1200 MHz
> >> min: 1200 MHz
> >> RP1: 200 MHz
> >> max: 1200 MHz
> >>
> >> $ sudo intel_frequency --min
> >> $ sudo intel_frequency -g
> >> cur: 200 MHz
> >> min: 200 MHz
> >> RP1: 200 MHz
> >> max: 200 MHz
> >> ---
> >> tools/Makefile.sources | 1 +
> >> tools/intel_frequency.c | 363
> >++++++++++++++++++++++++++++++++++++++++++++++++
> >> 2 files changed, 364 insertions(+)
> >> create mode 100644 tools/intel_frequency.c
> >>
> >> diff --git a/tools/Makefile.sources b/tools/Makefile.sources
> >> index b85a6b8..2bea389 100644
> >> --- a/tools/Makefile.sources
> >> +++ b/tools/Makefile.sources
> >> @@ -14,6 +14,7 @@ bin_PROGRAMS = \
> >> intel_dump_decode \
> >> intel_error_decode \
> >> intel_forcewaked \
> >> + intel_frequency \
> >> intel_framebuffer_dump \
> >> intel_gpu_time \
> >> intel_gpu_top \
> >> diff --git a/tools/intel_frequency.c b/tools/intel_frequency.c
> >> new file mode 100644
> >> index 0000000..59f3814
> >> --- /dev/null
> >> +++ b/tools/intel_frequency.c
> >> @@ -0,0 +1,363 @@
> >> +/*
> >> + * Copyright © 2015 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.
> >> + *
> >> + * Example:
> >> + * Get all frequencies:
> >> + * intel_frequency --get=cur,min,max,eff
> >> + *
> >> + * Same as above:
> >> + * intel_frequency -g
> >> + *
> >> + * Get the efficient frequency:
> >> + * intel_frequency -geff
> >> + *
> >> + * Lock the GPU frequency to 300MHz:
> >> + * intel_frequency --set min=300
> >> + *
> >> + * Set the maximum frequency to 900MHz:
> >> + * intel_frequency --custom max=900
> >> + *
> >> + * Lock the GPU frequency to its maximum frequency:
> >> + * intel_frequency --max
> >> + *
> >> + * Lock the GPU frequency to its most efficient frequency:
> >> + * intel_frequency -e
> >> + *
> >> + * Lock The GPU frequency to its minimum frequency:
> >> + * intel_frequency --min
> >> + *
> >> + * Reset the GPU to hardware defaults
> >> + * intel_frequency -d
> >> + */
> >> +
> >> +#define _GNU_SOURCE
> >> +#include <assert.h>
> >> +#include <getopt.h>
> >> +#include <stdio.h>
> >> +#include <time.h>
> >> +#include <unistd.h>
> >> +
> >> +#include "drmtest.h"
> >> +#include "intel_chipset.h"
> >> +
> >> +static int device, devid;
> >> +
> >> +enum {
> >> + CUR=0,
> >> + MIN,
> >> + EFF,
> >> + MAX,
> >> + RP0,
> >> + RPn
> >> +};
> >> +
> >> +struct freq_info {
> >> + const char *name;
> >> + const char *mode;
> >> + FILE *filp;
> >> + char *path;
> >> +};
> >> +
> >> +static struct freq_info info[] = {
> >> + { "cur", "r" },
> >> + { "min", "rb+" },
> >> + { "RP1", "r" },
> >> + { "max", "rb+" },
> >> + { "RP0", "r" },
> >> + { "RPn", "r" }
> >> +};
> >> +
> >> +static char *
> >> +get_sysfs_path(const char *which)
> >> +{
> >> + static const char fmt[] = "/sys/class/drm/card%1d/gt_%3s_freq_mhz";
> >> + char *path;
> >> + int ret;
> >> +
> >> +#define STATIC_STRLEN(string) (sizeof(string) / sizeof(string [0]))
> >> + ret = asprintf(&path, fmt, device, which);
> >> + assert(ret == (STATIC_STRLEN(fmt) - 3));
> >> +#undef STATIC_STRLEN
> >> +
> >> + return path;
> >> +}
> >> +
> >> +static void
> >> +initialize_freq_info(struct freq_info *freq_info)
> >> +{
> >> + if (freq_info->filp)
> >> + return;
> >> +
> >> + freq_info->path = get_sysfs_path(freq_info->name);
> >> + assert(freq_info->path);
> >> + freq_info->filp = fopen(freq_info->path, freq_info->mode);
> >> + assert(freq_info->filp);
> >> +}
> >> +
> >> +static void wait_freq_settle(void)
> >> +{
> >> + struct timespec ts;
> >> +
> >> + /* FIXME: Lazy sleep without check. */
> >> + ts.tv_sec = 0;
> >> + ts.tv_nsec = 20000;
> >> + clock_nanosleep(CLOCK_MONOTONIC, 0, &ts, NULL);
> >> +}
> >> +
> >> +static void set_frequency(struct freq_info *freq_info, int val)
> >> +{
> >> + initialize_freq_info(freq_info);
> >> + rewind(freq_info->filp);
> >> + assert(fprintf(freq_info->filp, "%d", val) > 0);
> >> +
> >> + wait_freq_settle();
> >> +}
> >> +
> >> +static int get_frequency(struct freq_info *freq_info)
> >> +{
> >> + int val;
> >> +
> >> + initialize_freq_info(freq_info);
> >> + rewind(freq_info->filp);
> >> + assert(fscanf(freq_info->filp, "%d", &val)==1);
> >> +
> >> + return val;
> >> +}
> >> +
> >> +static void
> >> +usage(const char *prog)
> >> +{
> >> + printf("Usage: %s [-e] [--min | --max] [-g (min|max|efficient)] [-s
> >frequency_mhz]\n\n", prog);
> >> + printf("%s A program to manipulate Intel GPU frequencies.\n\n", prog);
> >> + printf("Options: \n");
> >> + printf(" -e Lock frequency to the most efficient
> >frequency\n");
> >> + printf(" -g, --get= Get the frequency (optional arg:
> >\"cur\"|\"min\"|\"max\"|\"eff\")\n");
> >> + printf(" -s, --set Lock frequency to an absolute value (MHz)\n");
> >> + printf(" -c, --custom Set a min, or max frequency \"min=X |
> >max=Y\"\n");
> >> + printf(" -m --max Lock frequency to max frequency\n");
> >> + printf(" -i --min Lock frequency to min (never a good idea, DEBUG
> >ONLY)\n");
> >> + printf(" -d --defaults Return the system to hardware defaults\n");
> >> + printf("Examples:\n");
> >> + printf("\tintel_frequency -gmin,cur Get the current and minimum
> >frequency\n");
> >> + printf("\tintel_frequency -s 400 Lock frequency to 400Mhz\n");
> >> + printf("\tintel_frequency -c max=750 Set the max frequency to
> >750MHz\n");
> >> + exit(EXIT_FAILURE);
> >> +}
> >> +
> >> +/* Returns read or write operation */
> >> +static bool
> >> +parse(int argc, char *argv[], bool *act_upon, int *new_freq)
> >> +{
> >> + int c, tmp;
> >> + bool write = false;
> >> +
> >> + char *token[] = {
> >> + (char *)info[CUR].name,
> >> + (char *)info[MIN].name,
> >> + (char *)"eff",
> >> + (char *)info[MAX].name
> >> + };
> >> +
> >> + /* No args means -g" */
> >> + if (argc == 1) {
> >> + for (c = 0; c < ARRAY_SIZE(act_upon); c++)
> >> + act_upon[c] = true;
> >> + goto done;
> >> + }
> >> + while (1) {
> >> + int option_index = 0;
> >> + static struct option long_options[] = {
> >> + { "get", optional_argument, NULL, 'g' },
> >> + { "set", required_argument, NULL, 's' },
> >> + { "custom", required_argument, NULL, 'c'},
> >> + { "min", no_argument, NULL, 'i' },
> >> + { "max", no_argument, NULL, 'm' },
> >> + { "defaults", no_argument, NULL, 'd' },
> >> + { "help", no_argument, NULL, 'h' },
> >> + { NULL, 0, NULL, 0}
> >> + };
> >> +
> >> + c = getopt_long(argc, argv, "eg::s:c:midh", long_options,
> >&option_index);
> >> + if (c == -1)
> >> + break;
> >> +
> >> + switch (c) {
> >> + case 'g':
> >> + if (write == true)
> >> + fprintf(stderr, "Read and write operations not
> >support simultaneously.\n");
> >> +
> >> + if (optarg) {
> >> + char *value, *subopts = optarg;
> >> + int x;
> >> + while (*subopts != '\0') {
> >> + x = getsubopt(&subopts, token,
> >&value);
> >> + if (x == -1) {
> >> + fprintf(stderr, "Unrecognized
> >option (%s)\n", value);
> >> + break;
> >> + } else
> >> + act_upon[x] = true;
> >> + }
> >> + } else {
> >> + int i;
> >> + for (i = 0; i < ARRAY_SIZE(act_upon); i++)
> >> + act_upon[i] = true;
> >> + }
> >> + break;
> >> + case 's':
> >> + if (!optarg)
> >> + usage(argv[0]);
> >> +
> >> + if (write == true) {
> >> + fprintf(stderr, "Only one write may be specified
> >at a time\n");
> >> + exit(EXIT_FAILURE);
> >> + }
> >> +
> >> + write = true;
> >> + act_upon[MIN] = true;
> >> + act_upon[MAX] = true;
> >> + sscanf(optarg, "%d", &new_freq[MAX]);
> >> + new_freq[MIN] = new_freq[MAX];
> >> + break;
> >> + case 'c':
> >> + if (!optarg)
> >> + usage(argv[0]);
> >> +
> >> + if (write == true) {
> >> + fprintf(stderr, "Only one write may be specified
> >at a time\n");
> >> + exit(EXIT_FAILURE);
> >> + }
> >> +
> >> + write = true;
> >> +
> >> + if (!strncmp("min=", optarg, 4)) {
> >> + act_upon[MIN] = true;
> >> + sscanf(optarg+4, "%d", &new_freq[MIN]);
> >> + } else if (!strncmp("max=", optarg, 4)) {
> >> + act_upon[MAX] = true;
> >> + sscanf(optarg+4, "%d", &new_freq[MAX]);
> >> + } else {
> >> + fprintf(stderr, "Selected unmodifiable
> >frequency\n");
> >> + exit(EXIT_FAILURE);
> >> + }
> >> + break;
> >> + case 'e': /* efficient */
> >> + if (IS_VALLEYVIEW(devid) || IS_CHERRYVIEW(devid)) {
> >> + /* the LP parts have special efficient frequencies
> >*/
> >> + fprintf(stderr,
> >> + "FIXME: Warning efficient frequency
> >information is incorrect.\n");
> >> + exit(EXIT_FAILURE);
> >> + }
> >> + tmp = get_frequency(&info[EFF]);
> >> + new_freq[MIN] = tmp;
> >> + new_freq[MAX] = tmp;
> >> + act_upon[MIN] = true;
> >> + act_upon[MAX] = true;
> >> + write = true;
> >> + break;
> >> + case 'i': /* mIn */
> >> + tmp = get_frequency(&info[RPn]);
> >> + new_freq[MIN] = tmp;
> >> + new_freq[MAX] = tmp;
> >> + act_upon[MIN] = true;
> >> + act_upon[MAX] = true;
> >> + write = true;
> >> + break;
> >> + case 'm': /* max */
> >> + tmp = get_frequency(&info[RP0]);
> >> + new_freq[MIN] = tmp;
> >> + new_freq[MAX] = tmp;
> >> + act_upon[MIN] = true;
> >> + act_upon[MAX] = true;
> >> + write = true;
> >> + break;
> >> + case 'd': /* defaults */
> >> + new_freq[MIN] = get_frequency(&info[RPn]);
> >> + new_freq[MAX] = get_frequency(&info[RP0]);
> >> + act_upon[MIN] = true;
> >> + act_upon[MAX] = true;
> >> + write = true;
> >> + break;
> >> + case 'h':
> >> + default:
> >> + usage(argv[0]);
> >> + }
> >> + }
> >> +
> >> +done:
> >> + return write;
> >> +}
> >> +
> >> +int main(int argc, char *argv[])
> >> +{
> >> +
> >> + bool write, fail, targets[MAX+1] = {false};
> >> + int i, try = 1, set_freq[MAX+1] = {0};
> >> +
> >> + devid = intel_get_drm_devid(drm_open_any());
> >> + device = drm_get_card();
> >> +
> >> + write = parse(argc, argv, targets, set_freq);
> >> + fail = write;
> >> +
> >> + /* If we've previously locked the frequency, we need to make sure to set
> >things
> >> + * in the correct order, or else the operation will fail (ie. min = max =
> >200,
> >> + * and we set min to 300, we fail because it would try to set min >
> >> + * max). This can be accomplished be going either forward or reverse
> >> + * through the loop. MIN is always before MAX.
> >> + *
> >> + * XXX: Since only min and max are at play, the super lazy way is to do
> >this
> >> + * 3 times and if we still fail after 3, it's for real.
> >> + */
> >> +again:
> >> + if (try > 2) {
> >> + fprintf(stderr, "Did not achieve desired freq.\n");
> >> + exit(EXIT_FAILURE);
> >> + }
> >> + for (i = 0; i < ARRAY_SIZE(targets); i++) {
> >> + if (targets[i] == false)
> >> + continue;
> >> +
> >> + if (write) {
> >> + set_frequency(&info[i], set_freq[i]);
> >> + if (get_frequency(&info[i]) != set_freq[i])
> >> + fail = true;
> >> + else
> >> + fail = false;
> >> + } else {
> >> + printf("%s: %d MHz\n", info[i].name,
> >get_frequency(&info[i]));
> >> + }
> >> + }
> >> +
> >> + if (fail) {
> >> + try++;
> >> + goto again;
> >> + }
> >> +
> >> + for (i = 0; i < ARRAY_SIZE(targets); i++) {
> >> + if (info[i].filp) {
> >> + fclose(info[i].filp);
> >> + free(info[i].path);
> >> + }
> >> + }
> >> +
> >> + return EXIT_SUCCESS;
> >> +}
> >> --
> >> 2.2.1
> >>
> >_______________________________________________
> >Intel-gfx mailing list
> >Intel-gfx@lists.freedesktop.org
> >http://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
next prev parent reply other threads:[~2015-01-13 22:37 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-01-11 4:19 [PATCH 1/2] tools/Makefile: Alphabetize the list Ben Widawsky
2015-01-11 4:19 ` [PATCH 2/2] intel_frequency: A tool to manipulate Intel GPU frequency Ben Widawsky
2015-01-11 4:36 ` Ben Widawsky
2015-01-11 17:54 ` Jordan Justen
2015-01-12 3:35 ` [PATCH] [v2] " Ben Widawsky
2015-01-12 3:47 ` Ben Widawsky
2015-01-13 21:19 ` O'Rourke, Tom
2015-01-13 22:36 ` Ben Widawsky [this message]
2015-01-14 17:42 ` O'Rourke, Tom
2015-01-15 0:51 ` Ben Widawsky
2015-01-15 14:00 ` Dave Gordon
2015-01-15 17:24 ` Dave Gordon
2015-01-15 21:15 ` Ben Widawsky
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=20150113223655.GA26112@bwidawsk.net \
--to=ben@bwidawsk.net \
--cc=Tom.O'Rourke@intel.com \
--cc=benjamin.widawsky@intel.com \
--cc=intel-gfx@lists.freedesktop.org \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.