diff options
Diffstat (limited to 'tools/power/x86')
-rwxr-xr-x | tools/power/x86/intel_pstate_tracer/intel_pstate_tracer.py | 54 | ||||
-rw-r--r-- | tools/power/x86/turbostat/Makefile | 2 | ||||
-rw-r--r-- | tools/power/x86/turbostat/turbostat.8 | 26 | ||||
-rw-r--r-- | tools/power/x86/turbostat/turbostat.c | 908 | ||||
-rw-r--r-- | tools/power/x86/x86_energy_perf_policy/Makefile | 2 |
5 files changed, 706 insertions, 286 deletions
diff --git a/tools/power/x86/intel_pstate_tracer/intel_pstate_tracer.py b/tools/power/x86/intel_pstate_tracer/intel_pstate_tracer.py index 29f50d4cfea0..84e2b648e622 100755 --- a/tools/power/x86/intel_pstate_tracer/intel_pstate_tracer.py +++ b/tools/power/x86/intel_pstate_tracer/intel_pstate_tracer.py @@ -28,6 +28,7 @@ import subprocess import os import time import re +import signal import sys import getopt import Gnuplot @@ -78,11 +79,12 @@ def print_help(): print(' Or') print(' ./intel_pstate_tracer.py [--cpu cpus] ---trace_file <trace_file> --name <test_name>') print(' To generate trace file, parse and plot, use (sudo required):') - print(' sudo ./intel_pstate_tracer.py [-c cpus] -i <interval> -n <test_name>') + print(' sudo ./intel_pstate_tracer.py [-c cpus] -i <interval> -n <test_name> -m <kbytes>') print(' Or') - print(' sudo ./intel_pstate_tracer.py [--cpu cpus] --interval <interval> --name <test_name>') + print(' sudo ./intel_pstate_tracer.py [--cpu cpus] --interval <interval> --name <test_name> --memory <kbytes>') print(' Optional argument:') - print(' cpus: comma separated list of CPUs') + print(' cpus: comma separated list of CPUs') + print(' kbytes: Kilo bytes of memory per CPU to allocate to the trace buffer. Default: 10240') print(' Output:') print(' If not already present, creates a "results/test_name" folder in the current working directory with:') print(' cpu.csv - comma seperated values file with trace contents and some additional calculations.') @@ -379,7 +381,7 @@ def clear_trace_file(): f_handle.close() except: print('IO error clearing trace file ') - quit() + sys.exit(2) def enable_trace(): """ Enable trace """ @@ -389,7 +391,7 @@ def enable_trace(): , 'w').write("1") except: print('IO error enabling trace ') - quit() + sys.exit(2) def disable_trace(): """ Disable trace """ @@ -399,17 +401,17 @@ def disable_trace(): , 'w').write("0") except: print('IO error disabling trace ') - quit() + sys.exit(2) def set_trace_buffer_size(): """ Set trace buffer size """ try: - open('/sys/kernel/debug/tracing/buffer_size_kb' - , 'w').write("10240") + with open('/sys/kernel/debug/tracing/buffer_size_kb', 'w') as fp: + fp.write(memory) except: - print('IO error setting trace buffer size ') - quit() + print('IO error setting trace buffer size ') + sys.exit(2) def free_trace_buffer(): """ Free the trace buffer memory """ @@ -418,8 +420,8 @@ def free_trace_buffer(): open('/sys/kernel/debug/tracing/buffer_size_kb' , 'w').write("1") except: - print('IO error setting trace buffer size ') - quit() + print('IO error freeing trace buffer ') + sys.exit(2) def read_trace_data(filename): """ Read and parse trace data """ @@ -431,7 +433,7 @@ def read_trace_data(filename): data = open(filename, 'r').read() except: print('Error opening ', filename) - quit() + sys.exit(2) for line in data.splitlines(): search_obj = \ @@ -489,10 +491,22 @@ def read_trace_data(filename): # Now seperate the main overall csv file into per CPU csv files. split_csv() +def signal_handler(signal, frame): + print(' SIGINT: Forcing cleanup before exit.') + if interval: + disable_trace() + clear_trace_file() + # Free the memory + free_trace_buffer() + sys.exit(0) + +signal.signal(signal.SIGINT, signal_handler) + interval = "" filename = "" cpu_list = "" testname = "" +memory = "10240" graph_data_present = False; valid1 = False @@ -501,7 +515,7 @@ valid2 = False cpu_mask = zeros((MAX_CPUS,), dtype=int) try: - opts, args = getopt.getopt(sys.argv[1:],"ht:i:c:n:",["help","trace_file=","interval=","cpu=","name="]) + opts, args = getopt.getopt(sys.argv[1:],"ht:i:c:n:m:",["help","trace_file=","interval=","cpu=","name=","memory="]) except getopt.GetoptError: print_help() sys.exit(2) @@ -521,6 +535,8 @@ for opt, arg in opts: elif opt in ("-n", "--name"): valid2 = True testname = arg + elif opt in ("-m", "--memory"): + memory = arg if not (valid1 and valid2): print_help() @@ -569,6 +585,11 @@ current_max_cpu = 0 read_trace_data(filename) +clear_trace_file() +# Free the memory +if interval: + free_trace_buffer() + if graph_data_present == False: print('No valid data to plot') sys.exit(2) @@ -593,9 +614,4 @@ for root, dirs, files in os.walk('.'): for f in files: fix_ownership(f) -clear_trace_file() -# Free the memory -if interval: - free_trace_buffer() - os.chdir('../../') diff --git a/tools/power/x86/turbostat/Makefile b/tools/power/x86/turbostat/Makefile index a9bc914a8fe8..2ab25aa38263 100644 --- a/tools/power/x86/turbostat/Makefile +++ b/tools/power/x86/turbostat/Makefile @@ -25,4 +25,4 @@ install : turbostat install -d $(DESTDIR)$(PREFIX)/bin install $(BUILD_OUTPUT)/turbostat $(DESTDIR)$(PREFIX)/bin/turbostat install -d $(DESTDIR)$(PREFIX)/share/man/man8 - install turbostat.8 $(DESTDIR)$(PREFIX)/share/man/man8 + install -m 644 turbostat.8 $(DESTDIR)$(PREFIX)/share/man/man8 diff --git a/tools/power/x86/turbostat/turbostat.8 b/tools/power/x86/turbostat/turbostat.8 index ccf2a69365cc..ca9ef7017624 100644 --- a/tools/power/x86/turbostat/turbostat.8 +++ b/tools/power/x86/turbostat/turbostat.8 @@ -54,9 +54,12 @@ name as necessary to disambiguate it from others is necessary. Note that option .PP \fB--cpu cpu-set\fP limit output to system summary plus the specified cpu-set. If cpu-set is the string "core", then the system summary plus the first CPU in each core are printed -- eg. subsequent HT siblings are not printed. Or if cpu-set is the string "package", then the system summary plus the first CPU in each package is printed. Otherwise, the system summary plus the specified set of CPUs are printed. The cpu-set is ordered from low to high, comma delimited with ".." and "-" permitted to denote a range. eg. 1,2,8,14..17,21-44 .PP -\fB--hide column\fP do not show the specified columns. May be invoked multiple times, or with a comma-separated list of column names. Use "--hide sysfs" to hide the sysfs statistics columns as a group. +\fB--hide column\fP do not show the specified built-in columns. May be invoked multiple times, or with a comma-separated list of column names. Use "--hide sysfs" to hide the sysfs statistics columns as a group. .PP -\fB--show column\fP show only the specified columns. May be invoked multiple times, or with a comma-separated list of column names. Use "--show sysfs" to show the sysfs statistics columns as a group. +\fB--enable column\fP show the specified built-in columns, which are otherwise disabled, by default. Currently the only built-in counters disabled by default are "usec" and "Time_Of_Day_Seconds". +The column name "all" can be used to enable all disabled-by-default built-in counters. +.PP +\fB--show column\fP show only the specified built-in columns. May be invoked multiple times, or with a comma-separated list of column names. Use "--show sysfs" to show the sysfs statistics columns as a group. .PP \fB--Dump\fP displays the raw counter values. .PP @@ -64,6 +67,8 @@ name as necessary to disambiguate it from others is necessary. Note that option .PP \fB--interval seconds\fP overrides the default 5.0 second measurement interval. .PP +\fB--num_iterations num\fP number of the measurement iterations. +.PP \fB--out output_file\fP turbostat output is written to the specified output_file. The file is truncated if it already exists, and it is created if it does not exist. .PP @@ -86,6 +91,8 @@ displays the statistics gathered since it was forked. The system configuration dump (if --quiet is not used) is followed by statistics. The first row of the statistics labels the content of each column (below). The second row of statistics is the system summary line. The system summary line has a '-' in the columns for the Package, Core, and CPU. The contents of the system summary line depends on the type of column. Columns that count items (eg. IRQ) show the sum across all CPUs in the system. Columns that show a percentage show the average across all CPUs in the system. Columns that dump raw MSR values simply show 0 in the summary. After the system summary row, each row describes a specific Package/Core/CPU. Note that if the --cpu parameter is used to limit which specific CPUs are displayed, turbostat will still collect statistics for all CPUs in the system and will still show the system summary for all CPUs in the system. .SH COLUMN DESCRIPTIONS .nf +\fBusec\fP For each CPU, the number of microseconds elapsed during counter collection, including thread migration -- if any. This counter is disabled by default, and is enabled with "--enable usec", or --debug. On the summary row, usec refers to the total elapsed time to collect the counters on all cpus. +\fBTime_Of_Day_Seconds\fP For each CPU, the gettimeofday(2) value (seconds.subsec since Epoch) when the counters ending the measurement interval were collected. This column is disabled by default, and can be enabled with "--enable Time_Of_Day_Seconds" or "--debug". On the summary row, Time_Of_Day_Seconds refers to the timestamp following collection of counters on the last CPU. \fBCore\fP processor core number. Note that multiple CPUs per core indicate support for Intel(R) Hyper-Threading Technology (HT). \fBCPU\fP Linux CPU (logical processor) number. Yes, it is okay that on many systems the CPUs are not listed in numerical order -- for efficiency reasons, turbostat runs in topology order, so HT siblings appear together. \fBPackage\fP processor package number -- not present on systems with a single processor package. @@ -262,6 +269,21 @@ CPU PRF_CTRL .fi +.SH INPUT + +For interval-mode, turbostat will immediately end the current interval +when it sees a newline on standard input. +turbostat will then start the next interval. +Control-C will be send a SIGINT to turbostat, +which will immediately abort the program with no further processing. +.SH SIGNALS + +SIGINT will interrupt interval-mode. +The end-of-interval data will be collected and displayed before turbostat exits. + +SIGUSR1 will end current interval, +end-of-interval data will be collected and displayed before turbostat +starts a new interval. .SH NOTES .B "turbostat " diff --git a/tools/power/x86/turbostat/turbostat.c b/tools/power/x86/turbostat/turbostat.c index bd9c6b31a504..d6cff3070ebd 100644 --- a/tools/power/x86/turbostat/turbostat.c +++ b/tools/power/x86/turbostat/turbostat.c @@ -29,6 +29,7 @@ #include <sys/types.h> #include <sys/wait.h> #include <sys/stat.h> +#include <sys/select.h> #include <sys/resource.h> #include <fcntl.h> #include <signal.h> @@ -47,9 +48,13 @@ char *proc_stat = "/proc/stat"; FILE *outf; int *fd_percpu; +struct timeval interval_tv = {5, 0}; struct timespec interval_ts = {5, 0}; +struct timespec one_msec = {0, 1000000}; +unsigned int num_iterations; unsigned int debug; unsigned int quiet; +unsigned int shown; unsigned int sums_need_wide_columns; unsigned int rapl_joules; unsigned int summary_only; @@ -58,6 +63,7 @@ unsigned int dump_only; unsigned int do_snb_cstates; unsigned int do_knl_cstates; unsigned int do_slm_cstates; +unsigned int do_cnl_cstates; unsigned int use_c1_residency_msr; unsigned int has_aperf; unsigned int has_epb; @@ -80,6 +86,8 @@ unsigned int do_rapl; unsigned int do_dts; unsigned int do_ptm; unsigned long long gfx_cur_rc6_ms; +unsigned long long cpuidle_cur_cpu_lpi_us; +unsigned long long cpuidle_cur_sys_lpi_us; unsigned int gfx_cur_mhz; unsigned int tcc_activation_temp; unsigned int tcc_activation_temp_override; @@ -87,6 +95,7 @@ double rapl_power_units, rapl_time_units; double rapl_dram_energy_units, rapl_energy_units; double rapl_joule_counter_range; unsigned int do_core_perf_limit_reasons; +unsigned int has_automatic_cstate_conversion; unsigned int do_gfx_perf_limit_reasons; unsigned int do_ring_perf_limit_reasons; unsigned int crystal_hz; @@ -147,7 +156,9 @@ char *progname; #define CPU_SUBSET_MAXCPUS 1024 /* need to use before probe... */ cpu_set_t *cpu_present_set, *cpu_affinity_set, *cpu_subset; size_t cpu_present_setsize, cpu_affinity_setsize, cpu_subset_size; -#define MAX_ADDED_COUNTERS 16 +#define MAX_ADDED_COUNTERS 8 +#define MAX_ADDED_THREAD_COUNTERS 24 +#define BITMASK_SIZE 32 struct thread_data { struct timeval tv_begin; @@ -162,7 +173,7 @@ struct thread_data { unsigned int flags; #define CPU_IS_FIRST_THREAD_IN_CORE 0x2 #define CPU_IS_FIRST_CORE_IN_PACKAGE 0x4 - unsigned long long counter[MAX_ADDED_COUNTERS]; + unsigned long long counter[MAX_ADDED_THREAD_COUNTERS]; } *thread_even, *thread_odd; struct core_data { @@ -183,6 +194,8 @@ struct pkg_data { unsigned long long pc8; unsigned long long pc9; unsigned long long pc10; + unsigned long long cpu_lpi; + unsigned long long sys_lpi; unsigned long long pkg_wtd_core_c0; unsigned long long pkg_any_core_c0; unsigned long long pkg_any_gfxe_c0; @@ -203,12 +216,21 @@ struct pkg_data { #define ODD_COUNTERS thread_odd, core_odd, package_odd #define EVEN_COUNTERS thread_even, core_even, package_even -#define GET_THREAD(thread_base, thread_no, core_no, pkg_no) \ - (thread_base + (pkg_no) * topo.num_cores_per_pkg * \ - topo.num_threads_per_core + \ - (core_no) * topo.num_threads_per_core + (thread_no)) -#define GET_CORE(core_base, core_no, pkg_no) \ - (core_base + (pkg_no) * topo.num_cores_per_pkg + (core_no)) +#define GET_THREAD(thread_base, thread_no, core_no, node_no, pkg_no) \ + ((thread_base) + \ + ((pkg_no) * \ + topo.nodes_per_pkg * topo.cores_per_node * topo.threads_per_core) + \ + ((node_no) * topo.cores_per_node * topo.threads_per_core) + \ + ((core_no) * topo.threads_per_core) + \ + (thread_no)) + +#define GET_CORE(core_base, core_no, node_no, pkg_no) \ + ((core_base) + \ + ((pkg_no) * topo.nodes_per_pkg * topo.cores_per_node) + \ + ((node_no) * topo.cores_per_node) + \ + (core_no)) + + #define GET_PKG(pkg_base, pkg_no) (pkg_base + pkg_no) enum counter_scope {SCOPE_CPU, SCOPE_CORE, SCOPE_PACKAGE}; @@ -244,14 +266,25 @@ struct system_summary { struct pkg_data packages; } average; +struct cpu_topology { + int physical_package_id; + int logical_cpu_id; + int physical_node_id; + int logical_node_id; /* 0-based count within the package */ + int physical_core_id; + int thread_id; + cpu_set_t *put_ids; /* Processing Unit/Thread IDs */ +} *cpus; struct topo_params { int num_packages; int num_cpus; int num_cores; int max_cpu_num; - int num_cores_per_pkg; - int num_threads_per_core; + int max_node_num; + int nodes_per_pkg; + int cores_per_node; + int threads_per_core; } topo; struct timeval tv_even, tv_odd, tv_delta; @@ -273,27 +306,33 @@ int cpu_is_not_present(int cpu) int for_all_cpus(int (func)(struct thread_data *, struct core_data *, struct pkg_data *), struct thread_data *thread_base, struct core_data *core_base, struct pkg_data *pkg_base) { - int retval, pkg_no, core_no, thread_no; + int retval, pkg_no, core_no, thread_no, node_no; for (pkg_no = 0; pkg_no < topo.num_packages; ++pkg_no) { - for (core_no = 0; core_no < topo.num_cores_per_pkg; ++core_no) { - for (thread_no = 0; thread_no < - topo.num_threads_per_core; ++thread_no) { - struct thread_data *t; - struct core_data *c; - struct pkg_data *p; - - t = GET_THREAD(thread_base, thread_no, core_no, pkg_no); - - if (cpu_is_not_present(t->cpu_id)) - continue; - - c = GET_CORE(core_base, core_no, pkg_no); - p = GET_PKG(pkg_base, pkg_no); - - retval = func(t, c, p); - if (retval) - return retval; + for (core_no = 0; core_no < topo.cores_per_node; ++core_no) { + for (node_no = 0; node_no < topo.nodes_per_pkg; + node_no++) { + for (thread_no = 0; thread_no < + topo.threads_per_core; ++thread_no) { + struct thread_data *t; + struct core_data *c; + struct pkg_data *p; + + t = GET_THREAD(thread_base, thread_no, + core_no, node_no, + pkg_no); + + if (cpu_is_not_present(t->cpu_id)) + continue; + + c = GET_CORE(core_base, core_no, + node_no, pkg_no); + p = GET_PKG(pkg_base, pkg_no); + + retval = func(t, c, p); + if (retval) + return retval; + } } } } @@ -346,6 +385,8 @@ int get_msr(int cpu, off_t offset, unsigned long long *msr) * Thus, strings that are proper sub-sets must follow their more specific peers. */ struct msr_counter bic[] = { + { 0x0, "usec" }, + { 0x0, "Time_Of_Day_Seconds" }, { 0x0, "Package" }, { 0x0, "Avg_MHz" }, { 0x0, "Bzy_MHz" }, @@ -369,7 +410,9 @@ struct msr_counter bic[] = { { 0x0, "Pkg%pc7" }, { 0x0, "Pkg%pc8" }, { 0x0, "Pkg%pc9" }, - { 0x0, "Pkg%pc10" }, + { 0x0, "Pk%pc10" }, + { 0x0, "CPU%LPI" }, + { 0x0, "SYS%LPI" }, { 0x0, "PkgWatt" }, { 0x0, "CorWatt" }, { 0x0, "GFXWatt" }, @@ -389,62 +432,72 @@ struct msr_counter bic[] = { { 0x0, "Any%C0" }, { 0x0, "GFX%C0" }, { 0x0, "CPUGFX%" }, + { 0x0, "Node%" }, }; #define MAX_BIC (sizeof(bic) / sizeof(struct msr_counter)) -#define BIC_Package (1ULL << 0) -#define BIC_Avg_MHz (1ULL << 1) -#define BIC_Bzy_MHz (1ULL << 2) -#define BIC_TSC_MHz (1ULL << 3) -#define BIC_IRQ (1ULL << 4) -#define BIC_SMI (1ULL << 5) -#define BIC_Busy (1ULL << 6) -#define BIC_CPU_c1 (1ULL << 7) -#define BIC_CPU_c3 (1ULL << 8) -#define BIC_CPU_c6 (1ULL << 9) -#define BIC_CPU_c7 (1ULL << 10) -#define BIC_ThreadC (1ULL << 11) -#define BIC_CoreTmp (1ULL << 12) -#define BIC_CoreCnt (1ULL << 13) -#define BIC_PkgTmp (1ULL << 14) -#define BIC_GFX_rc6 (1ULL << 15) -#define BIC_GFXMHz (1ULL << 16) -#define BIC_Pkgpc2 (1ULL << 17) -#define BIC_Pkgpc3 (1ULL << 18) -#define BIC_Pkgpc6 (1ULL << 19) -#define BIC_Pkgpc7 (1ULL << 20) -#define BIC_Pkgpc8 (1ULL << 21) -#define BIC_Pkgpc9 (1ULL << 22) -#define BIC_Pkgpc10 (1ULL << 23) -#define BIC_PkgWatt (1ULL << 24) -#define BIC_CorWatt (1ULL << 25) -#define BIC_GFXWatt (1ULL << 26) -#define BIC_PkgCnt (1ULL << 27) -#define BIC_RAMWatt (1ULL << 28) -#define BIC_PKG__ (1ULL << 29) -#define BIC_RAM__ (1ULL << 30) -#define BIC_Pkg_J (1ULL << 31) -#define BIC_Cor_J (1ULL << 32) -#define BIC_GFX_J (1ULL << 33) -#define BIC_RAM_J (1ULL << 34) -#define BIC_Core (1ULL << 35) -#define BIC_CPU (1ULL << 36) -#define BIC_Mod_c6 (1ULL << 37) -#define BIC_sysfs (1ULL << 38) -#define BIC_Totl_c0 (1ULL << 39) -#define BIC_Any_c0 (1ULL << 40) -#define BIC_GFX_c0 (1ULL << 41) -#define BIC_CPUGFX (1ULL << 42) - -unsigned long long bic_enabled = 0xFFFFFFFFFFFFFFFFULL; -unsigned long long bic_present = BIC_sysfs; +#define BIC_USEC (1ULL << 0) +#define BIC_TOD (1ULL << 1) +#define BIC_Package (1ULL << 2) +#define BIC_Avg_MHz (1ULL << 3) +#define BIC_Bzy_MHz (1ULL << 4) +#define BIC_TSC_MHz (1ULL << 5) +#define BIC_IRQ (1ULL << 6) +#define BIC_SMI (1ULL << 7) +#define BIC_Busy (1ULL << 8) +#define BIC_CPU_c1 (1ULL << 9) +#define BIC_CPU_c3 (1ULL << 10) +#define BIC_CPU_c6 (1ULL << 11) +#define BIC_CPU_c7 (1ULL << 12) +#define BIC_ThreadC (1ULL << 13) +#define BIC_CoreTmp (1ULL << 14) +#define BIC_CoreCnt (1ULL << 15) +#define BIC_PkgTmp (1ULL << 16) +#define BIC_GFX_rc6 (1ULL << 17) +#define BIC_GFXMHz (1ULL << 18) +#define BIC_Pkgpc2 (1ULL << 19) +#define BIC_Pkgpc3 (1ULL << 20) +#define BIC_Pkgpc6 (1ULL << 21) +#define BIC_Pkgpc7 (1ULL << 22) +#define BIC_Pkgpc8 (1ULL << 23) +#define BIC_Pkgpc9 (1ULL << 24) +#define BIC_Pkgpc10 (1ULL << 25) +#define BIC_CPU_LPI (1ULL << 26) +#define BIC_SYS_LPI (1ULL << 27) +#define BIC_PkgWatt (1ULL << 26) +#define BIC_CorWatt (1ULL << 27) +#define BIC_GFXWatt (1ULL << 28) +#define BIC_PkgCnt (1ULL << 29) +#define BIC_RAMWatt (1ULL << 30) +#define BIC_PKG__ (1ULL << 31) +#define BIC_RAM__ (1ULL << 32) +#define BIC_Pkg_J (1ULL << 33) +#define BIC_Cor_J (1ULL << 34) +#define BIC_GFX_J (1ULL << 35) +#define BIC_RAM_J (1ULL << 36) +#define BIC_Core (1ULL << 37) +#define BIC_CPU (1ULL << 38) +#define BIC_Mod_c6 (1ULL << 39) +#define BIC_sysfs (1ULL << 40) +#define BIC_Totl_c0 (1ULL << 41) +#define BIC_Any_c0 (1ULL << 42) +#define BIC_GFX_c0 (1ULL << 43) +#define BIC_CPUGFX (1ULL << 44) +#define BIC_Node (1ULL << 45) + +#define BIC_DISABLED_BY_DEFAULT (BIC_USEC | BIC_TOD) + +unsigned long long bic_enabled = (0xFFFFFFFFFFFFFFFFULL & ~BIC_DISABLED_BY_DEFAULT); +unsigned long long bic_present = BIC_USEC | BIC_TOD | BIC_sysfs; #define DO_BIC(COUNTER_NAME) (bic_enabled & bic_present & COUNTER_NAME) +#define ENABLE_BIC(COUNTER_NAME) (bic_enabled |= COUNTER_NAME) #define BIC_PRESENT(COUNTER_BIT) (bic_present |= COUNTER_BIT) #define BIC_NOT_PRESENT(COUNTER_BIT) (bic_present &= ~COUNTER_BIT) + #define MAX_DEFERRED 16 char *deferred_skip_names[MAX_DEFERRED]; int deferred_skip_index; @@ -469,9 +522,10 @@ void help(void) "--cpu cpu-set limit output to summary plus cpu-set:\n" " {core | package | j,k,l..m,n-p }\n" "--quiet skip decoding system configuration header\n" - "--interval sec Override default 5-second measurement interval\n" + "--interval sec.subsec Override default 5-second measurement interval\n" "--help print this help message\n" "--list list column headers only\n" + "--num_iterations num number of the measurement iterations\n" "--out file create or truncate \"file\" for all output\n" "--version print version information\n" "\n" @@ -496,6 +550,9 @@ unsigned long long bic_lookup(char *name_list, enum show_hide_mode mode) if (comma) *comma = '\0'; + if (!strcmp(name_list, "all")) + return ~0; + for (i = 0; i < MAX_BIC; ++i) { if (!strcmp(name_list, bic[i].name)) { retval |= (1ULL << i); @@ -532,10 +589,14 @@ void print_header(char *delim) struct msr_counter *mp; int printed = 0; - if (debug) - outp += sprintf(outp, "usec %s", delim); + if (DO_BIC(BIC_USEC)) + outp += sprintf(outp, "%susec", (printed++ ? delim : "")); + if (DO_BIC(BIC_TOD)) + outp += sprintf(outp, "%sTime_Of_Day_Seconds", (printed++ ? delim : "")); if (DO_BIC(BIC_Package)) outp += sprintf(outp, "%sPackage", (printed++ ? delim : "")); + if (DO_BIC(BIC_Node)) + outp += sprintf(outp, "%sNode", (printed++ ? delim : "")); if (DO_BIC(BIC_Core)) outp += sprintf(outp, "%sCore", (printed++ ? delim : "")); if (DO_BIC(BIC_CPU)) @@ -576,7 +637,7 @@ void print_header(char *delim) if (DO_BIC(BIC_CPU_c1)) outp += sprintf(outp, "%sCPU%%c1", (printed++ ? delim : "")); - if (DO_BIC(BIC_CPU_c3) && !do_slm_cstates && !do_knl_cstates) + if (DO_BIC(BIC_CPU_c3) && !do_slm_cstates && !do_knl_cstates && !do_cnl_cstates) outp += sprintf(outp, "%sCPU%%c3", (printed++ ? delim : "")); if (DO_BIC(BIC_CPU_c6)) outp += sprintf(outp, "%sCPU%%c6", (printed++ ? delim : "")); @@ -635,6 +696,10 @@ void print_header(char *delim) outp += sprintf(outp, "%sPkg%%pc9", (printed++ ? delim : "")); if (DO_BIC(BIC_Pkgpc10)) outp += sprintf(outp, "%sPk%%pc10", (printed++ ? delim : "")); + if (DO_BIC(BIC_CPU_LPI)) + outp += sprintf(outp, "%sCPU%%LPI", (printed++ ? delim : "")); + if (DO_BIC(BIC_SYS_LPI)) + outp += sprintf(outp, "%sSYS%%LPI", (printed++ ? delim : "")); if (do_rapl && !rapl_joules) { if (DO_BIC(BIC_PkgWatt)) @@ -739,6 +804,9 @@ int dump_counters(struct thread_data *t, struct core_data *c, outp += sprintf(outp, "pc8: %016llX\n", p->pc8); outp += sprintf(outp, "pc9: %016llX\n", p->pc9); outp += sprintf(outp, "pc10: %016llX\n", p->pc10); + outp += sprintf(outp, "pc10: %016llX\n", p->pc10); + outp += sprintf(outp, "cpu_lpi: %016llX\n", p->cpu_lpi); + outp += sprintf(outp, "sys_lpi: %016llX\n", p->sys_lpi); outp += sprintf(outp, "Joules PKG: %0X\n", p->energy_pkg); outp += sprintf(outp, "Joules COR: %0X\n", p->energy_cores); outp += sprintf(outp, "Joules GFX: %0X\n", p->energy_gfx); @@ -786,7 +854,7 @@ int format_counters(struct thread_data *t, struct core_data *c, (cpu_subset && !CPU_ISSET_S(t->cpu_id, cpu_subset_size, cpu_subset))) return 0; - if (debug) { + if (DO_BIC(BIC_USEC)) { /* on each row, print how many usec each timestamp took to gather */ struct timeval tv; @@ -794,6 +862,10 @@ int format_counters(struct thread_data *t, struct core_data *c, outp += sprintf(outp, "%5ld\t", tv.tv_sec * 1000000 + tv.tv_usec); } + /* Time_Of_Day_Seconds: on each row, print sec.usec last timestamp taken */ + if (DO_BIC(BIC_TOD)) + outp += sprintf(outp, "%10ld.%06ld\t", t->tv_end.tv_sec, t->tv_end.tv_usec); + interval_float = tv_delta.tv_sec + tv_delta.tv_usec/1000000.0; tsc = t->tsc * tsc_tweak; @@ -802,6 +874,8 @@ int format_counters(struct thread_data *t, struct core_data *c, if (t == &average.threads) { if (DO_BIC(BIC_Package)) outp += sprintf(outp, "%s-", (printed++ ? delim : "")); + if (DO_BIC(BIC_Node)) + outp += sprintf(outp, "%s-", (printed++ ? delim : "")); if (DO_BIC(BIC_Core)) outp += sprintf(outp, "%s-", (printed++ ? delim : "")); if (DO_BIC(BIC_CPU)) @@ -813,6 +887,15 @@ int format_counters(struct thread_data *t, struct core_data *c, else outp += sprintf(outp, "%s-", (printed++ ? delim : "")); } + if (DO_BIC(BIC_Node)) { + if (t) + outp += sprintf(outp, "%s%d", + (printed++ ? delim : ""), + cpus[t->cpu_id].physical_node_id); + else + outp += sprintf(outp, "%s-", + (printed++ ? delim : "")); + } if (DO_BIC(BIC_Core)) { if (c) outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), c->core_id); @@ -882,7 +965,7 @@ int format_counters(struct thread_data *t, struct core_data *c, if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE)) goto done; - if (DO_BIC(BIC_CPU_c3) && !do_slm_cstates && !do_knl_cstates) + if (DO_BIC(BIC_CPU_c3) && !do_slm_cstates && !do_knl_cstates && !do_cnl_cstates) outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * c->c3/tsc); if (DO_BIC(BIC_CPU_c6)) outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * c->c6/tsc); @@ -959,6 +1042,11 @@ int format_counters(struct thread_data *t, struct core_data *c, if (DO_BIC(BIC_Pkgpc10)) outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc10/tsc); + if (DO_BIC(BIC_CPU_LPI)) + outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->cpu_lpi / 1000000.0 / interval_float); + if (DO_BIC(BIC_SYS_LPI)) + outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->sys_lpi / 1000000.0 / interval_float); + /* * If measurement interval exceeds minimum RAPL Joule Counter range, * indicate that results are suspect by printing "**" in fraction place. @@ -1006,7 +1094,8 @@ int format_counters(struct thread_data *t, struct core_data *c, } done: - outp += sprintf(outp, "\n"); + if (*(outp - 1) != '\n') + outp += sprintf(outp, "\n"); return 0; } @@ -1083,6 +1172,8 @@ delta_package(struct pkg_data *new, struct pkg_data *old) old->pc8 = new->pc8 - old->pc8; old->pc9 = new->pc9 - old->pc9; old->pc10 = new->pc10 - old->pc10; + old->cpu_lpi = new->cpu_lpi - old->cpu_lpi; + old->sys_lpi = new->sys_lpi - old->sys_lpi; old->pkg_temp_c = new->pkg_temp_c; /* flag an error when rc6 counter resets/wraps */ @@ -1140,6 +1231,15 @@ delta_thread(struct thread_data *new, struct thread_data *old, int i; struct msr_counter *mp; + /* + * the timestamps from start of measurement interval are in "old" + * the timestamp from end of measurement interval are in "new" + * over-write old w/ new so we can print end of interval values + */ + + old->tv_begin = new->tv_begin; + old->tv_end = new->tv_end; + old->tsc = new->tsc - old->tsc; /* check for TSC < 1 Mcycles over interval */ @@ -1228,6 +1328,11 @@ void clear_counters(struct thread_data *t, struct core_data *c, struct pkg_data int i; struct msr_counter *mp; + t->tv_begin.tv_sec = 0; + t->tv_begin.tv_usec = 0; + t->tv_end.tv_sec = 0; + t->tv_end.tv_usec = 0; + t->tsc = 0; t->aperf = 0; t->mperf = 0; @@ -1260,6 +1365,8 @@ void clear_counters(struct thread_data *t, struct core_data *c, struct pkg_data p->pc8 = 0; p->pc9 = 0; p->pc10 = 0; + p->cpu_lpi = 0; + p->sys_lpi = 0; p->energy_pkg = 0; p->energy_dram = 0; @@ -1286,6 +1393,13 @@ int sum_counters(struct thread_data *t, struct core_data *c, int i; struct msr_counter *mp; + /* remember first tv_begin */ + if (average.threads.tv_begin.tv_sec == 0) + average.threads.tv_begin = t->tv_begin; + + /* remember last tv_end */ + average.threads.tv_end = t->tv_end; + average.threads.tsc += t->tsc; average.threads.aperf += t->aperf; average.threads.mperf += t->mperf; @@ -1341,6 +1455,9 @@ int sum_counters(struct thread_data *t, struct core_data *c, average.packages.pc9 += p->pc9; average.packages.pc10 += p->pc10; + average.packages.cpu_lpi = p->cpu_lpi; + average.packages.sys_lpi = p->sys_lpi; + average.packages.energy_pkg += p->energy_pkg; average.packages.energy_dram += p->energy_dram; average.packages.energy_cores += p->energy_cores; @@ -1487,7 +1604,7 @@ int get_mp(int cpu, struct msr_counter *mp, unsigned long long *counterp) if (get_msr(cpu, mp->msr_num, counterp)) return -1; } else { - char path[128]; + char path[128 + PATH_BYTES]; if (mp->flags & SYSFS_PERCPU) { sprintf(path, "/sys/devices/system/cpu/cpu%d/%s", @@ -1603,7 +1720,7 @@ retry: if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE)) goto done; - if (DO_BIC(BIC_CPU_c3) && !do_slm_cstates && !do_knl_cstates) { + if (DO_BIC(BIC_CPU_c3) && !do_slm_cstates && !do_knl_cstates && !do_cnl_cstates) { if (get_msr(cpu, MSR_CORE_C3_RESIDENCY, &c->c3)) return -6; } @@ -1684,6 +1801,11 @@ retry: if (get_msr(cpu, MSR_PKG_C10_RESIDENCY, &p->pc10)) return -13; + if (DO_BIC(BIC_CPU_LPI)) + p->cpu_lpi = cpuidle_cur_cpu_lpi_us; + if (DO_BIC(BIC_SYS_LPI)) + p->sys_lpi = cpuidle_cur_sys_lpi_us; + if (do_rapl & RAPL_PKG) { if (get_msr(cpu, MSR_PKG_ENERGY_STATUS, &msr)) return -13; @@ -1769,7 +1891,7 @@ int slv_pkg_cstate_limits[16] = {PCL__0, PCL__1, PCLRSV, PCLRSV, PCL__4, PCLRSV, int amt_pkg_cstate_limits[16] = {PCLUNL, PCL__1, PCL__2, PCLRSV, PCLRSV, PCLRSV, PCL__6, PCL__7, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV}; int phi_pkg_cstate_limits[16] = {PCL__0, PCL__2, PCL_6N, PCL_6R, PCLRSV, PCLRSV, PCLRSV, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV}; int bxt_pkg_cstate_limits[16] = {PCL__0, PCL__2, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV}; -int skx_pkg_cstate_limits[16] = {PCL__0, PCL__2, PCL_6N, PCL_6R, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV}; +int skx_pkg_cstate_limits[16] = {PCL__0, PCL__2, PCL_6N, PCL_6R, PCLRSV, PCLRSV, PCLRSV, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV}; static void @@ -2071,12 +2193,9 @@ dump_nhm_cst_cfg(void) get_msr(base_cpu, MSR_PKG_CST_CONFIG_CONTROL, &msr); -#define SNB_C1_AUTO_UNDEMOTE (1UL << 27) -#define SNB_C3_AUTO_UNDEMOTE (1UL << 28) - fprintf(outf, "cpu%d: MSR_PKG_CST_CONFIG_CONTROL: 0x%08llx", base_cpu, msr); - fprintf(outf, " (%s%s%s%s%slocked: pkg-cstate-limit=%d: %s)\n", + fprintf(outf, " (%s%s%s%s%slocked, pkg-cstate-limit=%d (%s)", (msr & SNB_C3_AUTO_UNDEMOTE) ? "UNdemote-C3, " : "", (msr & SNB_C1_AUTO_UNDEMOTE) ? "UNdemote-C1, " : "", (msr & NHM_C3_AUTO_DEMOTE) ? "demote-C3, " : "", @@ -2084,6 +2203,15 @@ dump_nhm_cst_cfg(void) (msr & (1 << 15)) ? "" : "UN", (unsigned int)msr & 0xF, pkg_cstate_limit_strings[pkg_cstate_limit]); + +#define AUTOMATIC_CSTATE_CONVERSION (1UL << 16) + if (has_automatic_cstate_conversion) { + fprintf(outf, ", automatic c-state conversion=%s", + (msr & AUTOMATIC_CSTATE_CONVERSION) ? "on" : "off"); + } + + fprintf(outf, ")\n"); + return; } @@ -2184,6 +2312,8 @@ void free_fd_percpu(void) void free_all_buffers(void) { + int i; + CPU_FREE(cpu_present_set); cpu_present_set = NULL; cpu_present_setsize = 0; @@ -2216,6 +2346,12 @@ void free_all_buffers(void) free(irq_column_2_cpu); free(irqs_per_cpu); + + for (i = 0; i <= topo.max_cpu_num; ++i) { + if (cpus[i].put_ids) + CPU_FREE(cpus[i].put_ids); + } + free(cpus); } @@ -2240,44 +2376,6 @@ int parse_int_file(const char *fmt, ...) } /* - * get_cpu_position_in_core(cpu) - * return the position of the CPU among its HT siblings in the core - * return -1 if the sibling is not in list - */ -int get_cpu_position_in_core(int cpu) -{ - char path[64]; - FILE *filep; - int this_cpu; - char character; - int i; - - sprintf(path, - "/sys/devices/system/cpu/cpu%d/topology/thread_siblings_list", - cpu); - filep = fopen(path, "r"); - if (filep == NULL) { - perror(path); - exit(1); - } - - for (i = 0; i < topo.num_threads_per_core; i++) { - fscanf(filep, "%d", &this_cpu); - if (this_cpu == cpu) { - fclose(filep); - return i; - } - - /* Account for no separator after last thread*/ - if (i != (topo.num_threads_per_core - 1)) - fscanf(filep, "%c", &character); - } - - fclose(filep); - return -1; -} - -/* * cpu_is_first_core_in_package(cpu) * return 1 if given CPU is 1st core in package */ @@ -2296,35 +2394,115 @@ int get_core_id(int cpu) return parse_int_file("/sys/devices/system/cpu/cpu%d/topology/core_id", cpu); } -int get_num_ht_siblings(int cpu) +void set_node_data(void) { char path[80]; FILE *filep; - int sib1; - int matches = 0; - char character; - char str[100]; - char *ch; + int pkg, node, cpu; - sprintf(path, "/sys/devices/system/cpu/cpu%d/topology/thread_siblings_list", cpu); - filep = fopen_or_die(path, "r"); + struct pkg_node_info { + int count; + int min; + } *pni; - /* - * file format: - * A ',' separated or '-' separated set of numbers - * (eg 1-2 or 1,3,4,5) - */ - fscanf(filep, "%d%c\n", &sib1, &character); - fseek(filep, 0, SEEK_SET); - fgets(str, 100, filep); - ch = strchr(str, character); - while (ch != NULL) { - matches++; - ch = strchr(ch+1, character); + pni = calloc(topo.num_packages, sizeof(struct pkg_node_info)); + if (!pni) + err(1, "calloc pkg_node_count"); + + for (pkg = 0; pkg < topo.num_packages; pkg++) + pni[pkg].min = topo.num_cpus; + + for (node = 0; node <= topo.max_node_num; node++) { + /* find the "first" cpu in the node */ + sprintf(path, "/sys/bus/node/devices/node%d/cpulist", node); + filep = fopen(path, "r"); + if (!filep) + continue; + fscanf(filep, "%d", &cpu); + fclose(filep); + + pkg = cpus[cpu].physical_package_id; + pni[pkg].count++; + + if (node < pni[pkg].min) + pni[pkg].min = node; } + for (pkg = 0; pkg < topo.num_packages; pkg++) + if (pni[pkg].count > topo.nodes_per_pkg) + topo.nodes_per_pkg = pni[0].count; + + for (cpu = 0; cpu < topo.num_cpus; cpu++) { + pkg = cpus[cpu].physical_package_id; + node = cpus[cpu].physical_node_id; + cpus[cpu].logical_node_id = node - pni[pkg].min; + } |