Commit | Line | Data |
---|---|---|
a61127c2 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
103a8fea LB |
2 | /* |
3 | * turbostat -- show CPU frequency and C-state residency | |
34041551 | 4 | * on modern Intel and AMD processors. |
103a8fea | 5 | * |
144b44b1 | 6 | * Copyright (c) 2013 Intel Corporation. |
103a8fea | 7 | * Len Brown <len.brown@intel.com> |
103a8fea LB |
8 | */ |
9 | ||
88c3281f | 10 | #define _GNU_SOURCE |
b731f311 | 11 | #include MSRHEADER |
869ce69e | 12 | #include INTEL_FAMILY_HEADER |
95aebc44 | 13 | #include <stdarg.h> |
103a8fea | 14 | #include <stdio.h> |
b2c95d90 | 15 | #include <err.h> |
103a8fea LB |
16 | #include <unistd.h> |
17 | #include <sys/types.h> | |
18 | #include <sys/wait.h> | |
19 | #include <sys/stat.h> | |
b9ad8ee0 | 20 | #include <sys/select.h> |
103a8fea LB |
21 | #include <sys/resource.h> |
22 | #include <fcntl.h> | |
23 | #include <signal.h> | |
24 | #include <sys/time.h> | |
25 | #include <stdlib.h> | |
d8af6f5f | 26 | #include <getopt.h> |
103a8fea LB |
27 | #include <dirent.h> |
28 | #include <string.h> | |
29 | #include <ctype.h> | |
88c3281f | 30 | #include <sched.h> |
2a0609c0 | 31 | #include <time.h> |
2b92865e | 32 | #include <cpuid.h> |
fcaa681c | 33 | #include <sys/capability.h> |
98481e79 | 34 | #include <errno.h> |
9392bd98 | 35 | #include <math.h> |
2af4f9b8 LB |
36 | #include <linux/perf_event.h> |
37 | #include <asm/unistd.h> | |
103a8fea | 38 | |
103a8fea | 39 | char *proc_stat = "/proc/stat"; |
b7d8c148 | 40 | FILE *outf; |
36229897 | 41 | int *fd_percpu; |
2af4f9b8 | 42 | int *fd_instr_count_percpu; |
b9ad8ee0 | 43 | struct timeval interval_tv = {5, 0}; |
47936f94 | 44 | struct timespec interval_ts = {5, 0}; |
023fe0ac | 45 | unsigned int num_iterations; |
d8af6f5f | 46 | unsigned int debug; |
96e47158 | 47 | unsigned int quiet; |
3f44a5c6 | 48 | unsigned int shown; |
0de6c0df | 49 | unsigned int sums_need_wide_columns; |
d8af6f5f LB |
50 | unsigned int rapl_joules; |
51 | unsigned int summary_only; | |
c8ade361 | 52 | unsigned int list_header_only; |
d8af6f5f | 53 | unsigned int dump_only; |
103a8fea | 54 | unsigned int do_snb_cstates; |
fb5d4327 | 55 | unsigned int do_knl_cstates; |
144b44b1 LB |
56 | unsigned int do_slm_cstates; |
57 | unsigned int use_c1_residency_msr; | |
103a8fea | 58 | unsigned int has_aperf; |
889facbe | 59 | unsigned int has_epb; |
5a63426e LB |
60 | unsigned int do_irtl_snb; |
61 | unsigned int do_irtl_hsw; | |
fc04cc67 | 62 | unsigned int units = 1000000; /* MHz etc */ |
103a8fea | 63 | unsigned int genuine_intel; |
34041551 | 64 | unsigned int authentic_amd; |
c1c10cc7 | 65 | unsigned int hygon_genuine; |
34041551 | 66 | unsigned int max_level, max_extended_level; |
103a8fea | 67 | unsigned int has_invariant_tsc; |
d7899447 | 68 | unsigned int do_nhm_platform_info; |
cf4cbe53 | 69 | unsigned int no_MSR_MISC_PWR_MGMT; |
b2b34dfe | 70 | unsigned int aperf_mperf_multiplier = 1; |
103a8fea | 71 | double bclk; |
a2b7b749 | 72 | double base_hz; |
21ed5574 | 73 | unsigned int has_base_hz; |
a2b7b749 | 74 | double tsc_tweak = 1.0; |
c98d5d94 LB |
75 | unsigned int show_pkg_only; |
76 | unsigned int show_core_only; | |
77 | char *output_buffer, *outp; | |
889facbe LB |
78 | unsigned int do_rapl; |
79 | unsigned int do_dts; | |
80 | unsigned int do_ptm; | |
2af4f9b8 | 81 | unsigned int do_ipc; |
fdf676e5 | 82 | unsigned long long gfx_cur_rc6_ms; |
be0e54c4 LB |
83 | unsigned long long cpuidle_cur_cpu_lpi_us; |
84 | unsigned long long cpuidle_cur_sys_lpi_us; | |
27d47356 | 85 | unsigned int gfx_cur_mhz; |
b4b91569 | 86 | unsigned int gfx_act_mhz; |
889facbe LB |
87 | unsigned int tcc_activation_temp; |
88 | unsigned int tcc_activation_temp_override; | |
40ee8e3b AS |
89 | double rapl_power_units, rapl_time_units; |
90 | double rapl_dram_energy_units, rapl_energy_units; | |
889facbe | 91 | double rapl_joule_counter_range; |
3a9a941d | 92 | unsigned int do_core_perf_limit_reasons; |
ac980e13 | 93 | unsigned int has_automatic_cstate_conversion; |
3a9a941d LB |
94 | unsigned int do_gfx_perf_limit_reasons; |
95 | unsigned int do_ring_perf_limit_reasons; | |
8a5bdf41 LB |
96 | unsigned int crystal_hz; |
97 | unsigned long long tsc_hz; | |
7ce7d5de | 98 | int base_cpu; |
21ed5574 | 99 | double discover_bclk(unsigned int family, unsigned int model); |
7f5c258e LB |
100 | unsigned int has_hwp; /* IA32_PM_ENABLE, IA32_HWP_CAPABILITIES */ |
101 | /* IA32_HWP_REQUEST, IA32_HWP_STATUS */ | |
102 | unsigned int has_hwp_notify; /* IA32_HWP_INTERRUPT */ | |
103 | unsigned int has_hwp_activity_window; /* IA32_HWP_REQUEST[bits 41:32] */ | |
104 | unsigned int has_hwp_epp; /* IA32_HWP_REQUEST[bits 31:24] */ | |
105 | unsigned int has_hwp_pkg; /* IA32_HWP_REQUEST_PKG */ | |
33148d67 | 106 | unsigned int has_misc_feature_control; |
4c2122d4 | 107 | unsigned int first_counter_read = 1; |
c026c236 | 108 | int ignore_stdin; |
889facbe | 109 | |
e6f9bb3c LB |
110 | #define RAPL_PKG (1 << 0) |
111 | /* 0x610 MSR_PKG_POWER_LIMIT */ | |
112 | /* 0x611 MSR_PKG_ENERGY_STATUS */ | |
113 | #define RAPL_PKG_PERF_STATUS (1 << 1) | |
114 | /* 0x613 MSR_PKG_PERF_STATUS */ | |
115 | #define RAPL_PKG_POWER_INFO (1 << 2) | |
116 | /* 0x614 MSR_PKG_POWER_INFO */ | |
117 | ||
118 | #define RAPL_DRAM (1 << 3) | |
119 | /* 0x618 MSR_DRAM_POWER_LIMIT */ | |
120 | /* 0x619 MSR_DRAM_ENERGY_STATUS */ | |
e6f9bb3c LB |
121 | #define RAPL_DRAM_PERF_STATUS (1 << 4) |
122 | /* 0x61b MSR_DRAM_PERF_STATUS */ | |
0b2bb692 LB |
123 | #define RAPL_DRAM_POWER_INFO (1 << 5) |
124 | /* 0x61c MSR_DRAM_POWER_INFO */ | |
e6f9bb3c | 125 | |
9148494c | 126 | #define RAPL_CORES_POWER_LIMIT (1 << 6) |
e6f9bb3c | 127 | /* 0x638 MSR_PP0_POWER_LIMIT */ |
0b2bb692 | 128 | #define RAPL_CORE_POLICY (1 << 7) |
e6f9bb3c LB |
129 | /* 0x63a MSR_PP0_POLICY */ |
130 | ||
0b2bb692 | 131 | #define RAPL_GFX (1 << 8) |
e6f9bb3c LB |
132 | /* 0x640 MSR_PP1_POWER_LIMIT */ |
133 | /* 0x641 MSR_PP1_ENERGY_STATUS */ | |
134 | /* 0x642 MSR_PP1_POLICY */ | |
9148494c JP |
135 | |
136 | #define RAPL_CORES_ENERGY_STATUS (1 << 9) | |
137 | /* 0x639 MSR_PP0_ENERGY_STATUS */ | |
9392bd98 CW |
138 | #define RAPL_PER_CORE_ENERGY (1 << 10) |
139 | /* Indicates cores energy collection is per-core, | |
140 | * not per-package. */ | |
141 | #define RAPL_AMD_F17H (1 << 11) | |
142 | /* 0xc0010299 MSR_RAPL_PWR_UNIT */ | |
143 | /* 0xc001029a MSR_CORE_ENERGY_STAT */ | |
144 | /* 0xc001029b MSR_PKG_ENERGY_STAT */ | |
9148494c | 145 | #define RAPL_CORES (RAPL_CORES_ENERGY_STATUS | RAPL_CORES_POWER_LIMIT) |
889facbe LB |
146 | #define TJMAX_DEFAULT 100 |
147 | ||
9392bd98 CW |
148 | /* MSRs that are not yet in the kernel-provided header. */ |
149 | #define MSR_RAPL_PWR_UNIT 0xc0010299 | |
150 | #define MSR_CORE_ENERGY_STAT 0xc001029a | |
151 | #define MSR_PKG_ENERGY_STAT 0xc001029b | |
152 | ||
889facbe | 153 | #define MAX(a, b) ((a) > (b) ? (a) : (b)) |
103a8fea | 154 | |
388e9c81 LB |
155 | /* |
156 | * buffer size used by sscanf() for added column names | |
157 | * Usually truncated to 7 characters, but also handles 18 columns for raw 64-bit counters | |
158 | */ | |
159 | #define NAME_BYTES 20 | |
495c7654 | 160 | #define PATH_BYTES 128 |
388e9c81 | 161 | |
103a8fea LB |
162 | int backwards_count; |
163 | char *progname; | |
103a8fea | 164 | |
1ef7d21a LB |
165 | #define CPU_SUBSET_MAXCPUS 1024 /* need to use before probe... */ |
166 | cpu_set_t *cpu_present_set, *cpu_affinity_set, *cpu_subset; | |
167 | size_t cpu_present_setsize, cpu_affinity_setsize, cpu_subset_size; | |
0748eaf0 LB |
168 | #define MAX_ADDED_COUNTERS 8 |
169 | #define MAX_ADDED_THREAD_COUNTERS 24 | |
0e2d8f05 | 170 | #define BITMASK_SIZE 32 |
c98d5d94 LB |
171 | |
172 | struct thread_data { | |
f4fdf2b4 LB |
173 | struct timeval tv_begin; |
174 | struct timeval tv_end; | |
d4794f25 | 175 | struct timeval tv_delta; |
c98d5d94 LB |
176 | unsigned long long tsc; |
177 | unsigned long long aperf; | |
178 | unsigned long long mperf; | |
144b44b1 | 179 | unsigned long long c1; |
2af4f9b8 | 180 | unsigned long long instr_count; |
0de6c0df | 181 | unsigned long long irq_count; |
1ed51011 | 182 | unsigned int smi_count; |
c98d5d94 | 183 | unsigned int cpu_id; |
4c2122d4 LB |
184 | unsigned int apic_id; |
185 | unsigned int x2apic_id; | |
c98d5d94 LB |
186 | unsigned int flags; |
187 | #define CPU_IS_FIRST_THREAD_IN_CORE 0x2 | |
188 | #define CPU_IS_FIRST_CORE_IN_PACKAGE 0x4 | |
0748eaf0 | 189 | unsigned long long counter[MAX_ADDED_THREAD_COUNTERS]; |
c98d5d94 LB |
190 | } *thread_even, *thread_odd; |
191 | ||
192 | struct core_data { | |
193 | unsigned long long c3; | |
194 | unsigned long long c6; | |
195 | unsigned long long c7; | |
0539ba11 | 196 | unsigned long long mc6_us; /* duplicate as per-core for now, even though per module */ |
889facbe | 197 | unsigned int core_temp_c; |
9392bd98 | 198 | unsigned int core_energy; /* MSR_CORE_ENERGY_STAT */ |
c98d5d94 | 199 | unsigned int core_id; |
678a3bd1 | 200 | unsigned long long counter[MAX_ADDED_COUNTERS]; |
c98d5d94 LB |
201 | } *core_even, *core_odd; |
202 | ||
203 | struct pkg_data { | |
204 | unsigned long long pc2; | |
205 | unsigned long long pc3; | |
206 | unsigned long long pc6; | |
207 | unsigned long long pc7; | |
ca58710f KCA |
208 | unsigned long long pc8; |
209 | unsigned long long pc9; | |
210 | unsigned long long pc10; | |
be0e54c4 LB |
211 | unsigned long long cpu_lpi; |
212 | unsigned long long sys_lpi; | |
0b2bb692 LB |
213 | unsigned long long pkg_wtd_core_c0; |
214 | unsigned long long pkg_any_core_c0; | |
215 | unsigned long long pkg_any_gfxe_c0; | |
216 | unsigned long long pkg_both_core_gfxe_c0; | |
9185e988 | 217 | long long gfx_rc6_ms; |
27d47356 | 218 | unsigned int gfx_mhz; |
b4b91569 | 219 | unsigned int gfx_act_mhz; |
c98d5d94 | 220 | unsigned int package_id; |
7c2ccc50 CY |
221 | unsigned long long energy_pkg; /* MSR_PKG_ENERGY_STATUS */ |
222 | unsigned long long energy_dram; /* MSR_DRAM_ENERGY_STATUS */ | |
223 | unsigned long long energy_cores; /* MSR_PP0_ENERGY_STATUS */ | |
224 | unsigned long long energy_gfx; /* MSR_PP1_ENERGY_STATUS */ | |
225 | unsigned long long rapl_pkg_perf_status; /* MSR_PKG_PERF_STATUS */ | |
226 | unsigned long long rapl_dram_perf_status; /* MSR_DRAM_PERF_STATUS */ | |
889facbe | 227 | unsigned int pkg_temp_c; |
678a3bd1 | 228 | unsigned long long counter[MAX_ADDED_COUNTERS]; |
c98d5d94 LB |
229 | } *package_even, *package_odd; |
230 | ||
231 | #define ODD_COUNTERS thread_odd, core_odd, package_odd | |
232 | #define EVEN_COUNTERS thread_even, core_even, package_even | |
233 | ||
40f5cfe7 PB |
234 | #define GET_THREAD(thread_base, thread_no, core_no, node_no, pkg_no) \ |
235 | ((thread_base) + \ | |
236 | ((pkg_no) * \ | |
237 | topo.nodes_per_pkg * topo.cores_per_node * topo.threads_per_core) + \ | |
238 | ((node_no) * topo.cores_per_node * topo.threads_per_core) + \ | |
239 | ((core_no) * topo.threads_per_core) + \ | |
240 | (thread_no)) | |
241 | ||
242 | #define GET_CORE(core_base, core_no, node_no, pkg_no) \ | |
243 | ((core_base) + \ | |
244 | ((pkg_no) * topo.nodes_per_pkg * topo.cores_per_node) + \ | |
245 | ((node_no) * topo.cores_per_node) + \ | |
246 | (core_no)) | |
247 | ||
248 | ||
c98d5d94 LB |
249 | #define GET_PKG(pkg_base, pkg_no) (pkg_base + pkg_no) |
250 | ||
388e9c81 | 251 | enum counter_scope {SCOPE_CPU, SCOPE_CORE, SCOPE_PACKAGE}; |
41618e63 | 252 | enum counter_type {COUNTER_ITEMS, COUNTER_CYCLES, COUNTER_SECONDS, COUNTER_USEC}; |
388e9c81 LB |
253 | enum counter_format {FORMAT_RAW, FORMAT_DELTA, FORMAT_PERCENT}; |
254 | ||
255 | struct msr_counter { | |
256 | unsigned int msr_num; | |
257 | char name[NAME_BYTES]; | |
495c7654 | 258 | char path[PATH_BYTES]; |
388e9c81 LB |
259 | unsigned int width; |
260 | enum counter_type type; | |
261 | enum counter_format format; | |
262 | struct msr_counter *next; | |
812db3f7 LB |
263 | unsigned int flags; |
264 | #define FLAGS_HIDE (1 << 0) | |
265 | #define FLAGS_SHOW (1 << 1) | |
41618e63 | 266 | #define SYSFS_PERCPU (1 << 1) |
388e9c81 LB |
267 | }; |
268 | ||
87e15da9 CY |
269 | /* |
270 | * The accumulated sum of MSR is defined as a monotonic | |
271 | * increasing MSR, it will be accumulated periodically, | |
272 | * despite its register's bit width. | |
273 | */ | |
274 | enum { | |
275 | IDX_PKG_ENERGY, | |
276 | IDX_DRAM_ENERGY, | |
277 | IDX_PP0_ENERGY, | |
278 | IDX_PP1_ENERGY, | |
279 | IDX_PKG_PERF, | |
280 | IDX_DRAM_PERF, | |
281 | IDX_COUNT, | |
282 | }; | |
283 | ||
284 | int get_msr_sum(int cpu, off_t offset, unsigned long long *msr); | |
285 | ||
286 | struct msr_sum_array { | |
287 | /* get_msr_sum() = sum + (get_msr() - last) */ | |
288 | struct { | |
289 | /*The accumulated MSR value is updated by the timer*/ | |
290 | unsigned long long sum; | |
291 | /*The MSR footprint recorded in last timer*/ | |
292 | unsigned long long last; | |
293 | } entries[IDX_COUNT]; | |
294 | }; | |
295 | ||
296 | /* The percpu MSR sum array.*/ | |
297 | struct msr_sum_array *per_cpu_msr_sum; | |
298 | ||
299 | int idx_to_offset(int idx) | |
300 | { | |
301 | int offset; | |
302 | ||
303 | switch (idx) { | |
304 | case IDX_PKG_ENERGY: | |
305 | offset = MSR_PKG_ENERGY_STATUS; | |
306 | break; | |
307 | case IDX_DRAM_ENERGY: | |
308 | offset = MSR_DRAM_ENERGY_STATUS; | |
309 | break; | |
310 | case IDX_PP0_ENERGY: | |
311 | offset = MSR_PP0_ENERGY_STATUS; | |
312 | break; | |
313 | case IDX_PP1_ENERGY: | |
314 | offset = MSR_PP1_ENERGY_STATUS; | |
315 | break; | |
316 | case IDX_PKG_PERF: | |
317 | offset = MSR_PKG_PERF_STATUS; | |
318 | break; | |
319 | case IDX_DRAM_PERF: | |
320 | offset = MSR_DRAM_PERF_STATUS; | |
321 | break; | |
322 | default: | |
323 | offset = -1; | |
324 | } | |
325 | return offset; | |
326 | } | |
327 | ||
328 | int offset_to_idx(int offset) | |
329 | { | |
330 | int idx; | |
331 | ||
332 | switch (offset) { | |
333 | case MSR_PKG_ENERGY_STATUS: | |
334 | idx = IDX_PKG_ENERGY; | |
335 | break; | |
336 | case MSR_DRAM_ENERGY_STATUS: | |
337 | idx = IDX_DRAM_ENERGY; | |
338 | break; | |
339 | case MSR_PP0_ENERGY_STATUS: | |
340 | idx = IDX_PP0_ENERGY; | |
341 | break; | |
342 | case MSR_PP1_ENERGY_STATUS: | |
343 | idx = IDX_PP1_ENERGY; | |
344 | break; | |
345 | case MSR_PKG_PERF_STATUS: | |
346 | idx = IDX_PKG_PERF; | |
347 | break; | |
348 | case MSR_DRAM_PERF_STATUS: | |
349 | idx = IDX_DRAM_PERF; | |
350 | break; | |
351 | default: | |
352 | idx = -1; | |
353 | } | |
354 | return idx; | |
355 | } | |
356 | ||
357 | int idx_valid(int idx) | |
358 | { | |
359 | switch (idx) { | |
360 | case IDX_PKG_ENERGY: | |
361 | return do_rapl & RAPL_PKG; | |
362 | case IDX_DRAM_ENERGY: | |
363 | return do_rapl & RAPL_DRAM; | |
364 | case IDX_PP0_ENERGY: | |
365 | return do_rapl & RAPL_CORES_ENERGY_STATUS; | |
366 | case IDX_PP1_ENERGY: | |
367 | return do_rapl & RAPL_GFX; | |
368 | case IDX_PKG_PERF: | |
369 | return do_rapl & RAPL_PKG_PERF_STATUS; | |
370 | case IDX_DRAM_PERF: | |
371 | return do_rapl & RAPL_DRAM_PERF_STATUS; | |
372 | default: | |
373 | return 0; | |
374 | } | |
375 | } | |
388e9c81 | 376 | struct sys_counters { |
678a3bd1 LB |
377 | unsigned int added_thread_counters; |
378 | unsigned int added_core_counters; | |
379 | unsigned int added_package_counters; | |
388e9c81 LB |
380 | struct msr_counter *tp; |
381 | struct msr_counter *cp; | |
382 | struct msr_counter *pp; | |
383 | } sys; | |
384 | ||
c98d5d94 LB |
385 | struct system_summary { |
386 | struct thread_data threads; | |
387 | struct core_data cores; | |
388 | struct pkg_data packages; | |
388e9c81 | 389 | } average; |
c98d5d94 | 390 | |
0e2d8f05 LB |
391 | struct cpu_topology { |
392 | int physical_package_id; | |
6de68fe1 | 393 | int die_id; |
0e2d8f05 | 394 | int logical_cpu_id; |
ef605741 PB |
395 | int physical_node_id; |
396 | int logical_node_id; /* 0-based count within the package */ | |
0e2d8f05 | 397 | int physical_core_id; |
8cb48b32 | 398 | int thread_id; |
0e2d8f05 LB |
399 | cpu_set_t *put_ids; /* Processing Unit/Thread IDs */ |
400 | } *cpus; | |
c98d5d94 LB |
401 | |
402 | struct topo_params { | |
403 | int num_packages; | |
6de68fe1 | 404 | int num_die; |
c98d5d94 LB |
405 | int num_cpus; |
406 | int num_cores; | |
407 | int max_cpu_num; | |
ef605741 | 408 | int max_node_num; |
70a9c6e8 PB |
409 | int nodes_per_pkg; |
410 | int cores_per_node; | |
411 | int threads_per_core; | |
c98d5d94 LB |
412 | } topo; |
413 | ||
414 | struct timeval tv_even, tv_odd, tv_delta; | |
415 | ||
562a2d37 LB |
416 | int *irq_column_2_cpu; /* /proc/interrupts column numbers */ |
417 | int *irqs_per_cpu; /* indexed by cpu_num */ | |
418 | ||
c98d5d94 LB |
419 | void setup_all_buffers(void); |
420 | ||
1f81c5ef LB |
421 | char *sys_lpi_file; |
422 | char *sys_lpi_file_sysfs = "/sys/devices/system/cpu/cpuidle/low_power_idle_system_residency_us"; | |
423 | char *sys_lpi_file_debugfs = "/sys/kernel/debug/pmc_core/slp_s0_residency_usec"; | |
424 | ||
c98d5d94 | 425 | int cpu_is_not_present(int cpu) |
d15cf7c1 | 426 | { |
c98d5d94 | 427 | return !CPU_ISSET_S(cpu, cpu_present_setsize, cpu_present_set); |
d15cf7c1 | 428 | } |
88c3281f | 429 | /* |
c98d5d94 LB |
430 | * run func(thread, core, package) in topology order |
431 | * skip non-present cpus | |
88c3281f | 432 | */ |
c98d5d94 LB |
433 | |
434 | int for_all_cpus(int (func)(struct thread_data *, struct core_data *, struct pkg_data *), | |
435 | struct thread_data *thread_base, struct core_data *core_base, struct pkg_data *pkg_base) | |
88c3281f | 436 | { |
40f5cfe7 | 437 | int retval, pkg_no, core_no, thread_no, node_no; |
d15cf7c1 | 438 | |
c98d5d94 | 439 | for (pkg_no = 0; pkg_no < topo.num_packages; ++pkg_no) { |
df2f677d LB |
440 | for (node_no = 0; node_no < topo.nodes_per_pkg; node_no++) { |
441 | for (core_no = 0; core_no < topo.cores_per_node; ++core_no) { | |
40f5cfe7 PB |
442 | for (thread_no = 0; thread_no < |
443 | topo.threads_per_core; ++thread_no) { | |
444 | struct thread_data *t; | |
445 | struct core_data *c; | |
446 | struct pkg_data *p; | |
447 | ||
448 | t = GET_THREAD(thread_base, thread_no, | |
449 | core_no, node_no, | |
450 | pkg_no); | |
451 | ||
452 | if (cpu_is_not_present(t->cpu_id)) | |
453 | continue; | |
454 | ||
455 | c = GET_CORE(core_base, core_no, | |
456 | node_no, pkg_no); | |
457 | p = GET_PKG(pkg_base, pkg_no); | |
458 | ||
459 | retval = func(t, c, p); | |
460 | if (retval) | |
461 | return retval; | |
462 | } | |
c98d5d94 LB |
463 | } |
464 | } | |
465 | } | |
466 | return 0; | |
88c3281f LB |
467 | } |
468 | ||
469 | int cpu_migrate(int cpu) | |
470 | { | |
c98d5d94 LB |
471 | CPU_ZERO_S(cpu_affinity_setsize, cpu_affinity_set); |
472 | CPU_SET_S(cpu, cpu_affinity_setsize, cpu_affinity_set); | |
473 | if (sched_setaffinity(0, cpu_affinity_setsize, cpu_affinity_set) == -1) | |
88c3281f LB |
474 | return -1; |
475 | else | |
476 | return 0; | |
477 | } | |
36229897 | 478 | int get_msr_fd(int cpu) |
103a8fea | 479 | { |
103a8fea LB |
480 | char pathname[32]; |
481 | int fd; | |
482 | ||
36229897 LB |
483 | fd = fd_percpu[cpu]; |
484 | ||
485 | if (fd) | |
486 | return fd; | |
487 | ||
103a8fea LB |
488 | sprintf(pathname, "/dev/cpu/%d/msr", cpu); |
489 | fd = open(pathname, O_RDONLY); | |
15aaa346 | 490 | if (fd < 0) |
98481e79 | 491 | err(-1, "%s open failed, try chown or chmod +r /dev/cpu/*/msr, or run as root", pathname); |
103a8fea | 492 | |
36229897 LB |
493 | fd_percpu[cpu] = fd; |
494 | ||
495 | return fd; | |
496 | } | |
497 | ||
2af4f9b8 LB |
498 | static long perf_event_open(struct perf_event_attr *hw_event, pid_t pid, int cpu, int group_fd, unsigned long flags) |
499 | { | |
500 | return syscall(__NR_perf_event_open, hw_event, pid, cpu, group_fd, flags); | |
501 | } | |
502 | ||
503 | static int perf_instr_count_open(int cpu_num) | |
504 | { | |
505 | struct perf_event_attr pea; | |
506 | int fd; | |
507 | ||
508 | memset(&pea, 0, sizeof(struct perf_event_attr)); | |
509 | pea.type = PERF_TYPE_HARDWARE; | |
510 | pea.size = sizeof(struct perf_event_attr); | |
511 | pea.config = PERF_COUNT_HW_INSTRUCTIONS; | |
512 | ||
513 | /* counter for cpu_num, including user + kernel and all processes */ | |
514 | fd = perf_event_open(&pea, -1, cpu_num, -1, 0); | |
515 | if (fd == -1) | |
516 | err(-1, "cpu%d: perf instruction counter\n", cpu_num); | |
517 | ||
518 | return fd; | |
519 | } | |
520 | ||
521 | int get_instr_count_fd(int cpu) | |
522 | { | |
523 | if (fd_instr_count_percpu[cpu]) | |
524 | return fd_instr_count_percpu[cpu]; | |
525 | ||
526 | fd_instr_count_percpu[cpu] = perf_instr_count_open(cpu); | |
527 | ||
528 | return fd_instr_count_percpu[cpu]; | |
529 | } | |
530 | ||
36229897 LB |
531 | int get_msr(int cpu, off_t offset, unsigned long long *msr) |
532 | { | |
533 | ssize_t retval; | |
534 | ||
535 | retval = pread(get_msr_fd(cpu), msr, sizeof(*msr), offset); | |
15aaa346 | 536 | |
98481e79 | 537 | if (retval != sizeof *msr) |
cf4cbe53 | 538 | err(-1, "cpu%d: msr offset 0x%llx read failed", cpu, (unsigned long long)offset); |
15aaa346 LB |
539 | |
540 | return 0; | |
103a8fea LB |
541 | } |
542 | ||
fc04cc67 | 543 | /* |
bdd5ae3a LB |
544 | * This list matches the column headers, except |
545 | * 1. built-in only, the sysfs counters are not here -- we learn of those at run-time | |
546 | * 2. Core and CPU are moved to the end, we can't have strings that contain them | |
547 | * matching on them for --show and --hide. | |
fc04cc67 | 548 | */ |
812db3f7 | 549 | struct msr_counter bic[] = { |
3f44a5c6 LB |
550 | { 0x0, "usec" }, |
551 | { 0x0, "Time_Of_Day_Seconds" }, | |
812db3f7 | 552 | { 0x0, "Package" }, |
bdd5ae3a | 553 | { 0x0, "Node" }, |
812db3f7 | 554 | { 0x0, "Avg_MHz" }, |
bdd5ae3a | 555 | { 0x0, "Busy%" }, |
812db3f7 LB |
556 | { 0x0, "Bzy_MHz" }, |
557 | { 0x0, "TSC_MHz" }, | |
558 | { 0x0, "IRQ" }, | |
495c7654 | 559 | { 0x0, "SMI", "", 32, 0, FORMAT_DELTA, NULL}, |
bdd5ae3a | 560 | { 0x0, "sysfs" }, |
812db3f7 LB |
561 | { 0x0, "CPU%c1" }, |
562 | { 0x0, "CPU%c3" }, | |
563 | { 0x0, "CPU%c6" }, | |
564 | { 0x0, "CPU%c7" }, | |
565 | { 0x0, "ThreadC" }, | |
566 | { 0x0, "CoreTmp" }, | |
567 | { 0x0, "CoreCnt" }, | |
568 | { 0x0, "PkgTmp" }, | |
569 | { 0x0, "GFX%rc6" }, | |
570 | { 0x0, "GFXMHz" }, | |
571 | { 0x0, "Pkg%pc2" }, | |
572 | { 0x0, "Pkg%pc3" }, | |
573 | { 0x0, "Pkg%pc6" }, | |
574 | { 0x0, "Pkg%pc7" }, | |
0f47c08d LB |
575 | { 0x0, "Pkg%pc8" }, |
576 | { 0x0, "Pkg%pc9" }, | |
4bd1f8f2 | 577 | { 0x0, "Pk%pc10" }, |
be0e54c4 LB |
578 | { 0x0, "CPU%LPI" }, |
579 | { 0x0, "SYS%LPI" }, | |
812db3f7 LB |
580 | { 0x0, "PkgWatt" }, |
581 | { 0x0, "CorWatt" }, | |
582 | { 0x0, "GFXWatt" }, | |
583 | { 0x0, "PkgCnt" }, | |
584 | { 0x0, "RAMWatt" }, | |
585 | { 0x0, "PKG_%" }, | |
586 | { 0x0, "RAM_%" }, | |
587 | { 0x0, "Pkg_J" }, | |
588 | { 0x0, "Cor_J" }, | |
589 | { 0x0, "GFX_J" }, | |
590 | { 0x0, "RAM_J" }, | |
0539ba11 | 591 | { 0x0, "Mod%c6" }, |
a99d8730 LB |
592 | { 0x0, "Totl%C0" }, |
593 | { 0x0, "Any%C0" }, | |
594 | { 0x0, "GFX%C0" }, | |
595 | { 0x0, "CPUGFX%" }, | |
bdd5ae3a LB |
596 | { 0x0, "Core" }, |
597 | { 0x0, "CPU" }, | |
4c2122d4 LB |
598 | { 0x0, "APIC" }, |
599 | { 0x0, "X2APIC" }, | |
6de68fe1 | 600 | { 0x0, "Die" }, |
b4b91569 | 601 | { 0x0, "GFXAMHz" }, |
2af4f9b8 | 602 | { 0x0, "IPC" }, |
812db3f7 LB |
603 | }; |
604 | ||
605 | #define MAX_BIC (sizeof(bic) / sizeof(struct msr_counter)) | |
3f44a5c6 LB |
606 | #define BIC_USEC (1ULL << 0) |
607 | #define BIC_TOD (1ULL << 1) | |
608 | #define BIC_Package (1ULL << 2) | |
bdd5ae3a LB |
609 | #define BIC_Node (1ULL << 3) |
610 | #define BIC_Avg_MHz (1ULL << 4) | |
611 | #define BIC_Busy (1ULL << 5) | |
612 | #define BIC_Bzy_MHz (1ULL << 6) | |
613 | #define BIC_TSC_MHz (1ULL << 7) | |
614 | #define BIC_IRQ (1ULL << 8) | |
615 | #define BIC_SMI (1ULL << 9) | |
616 | #define BIC_sysfs (1ULL << 10) | |
617 | #define BIC_CPU_c1 (1ULL << 11) | |
618 | #define BIC_CPU_c3 (1ULL << 12) | |
619 | #define BIC_CPU_c6 (1ULL << 13) | |
620 | #define BIC_CPU_c7 (1ULL << 14) | |
621 | #define BIC_ThreadC (1ULL << 15) | |
622 | #define BIC_CoreTmp (1ULL << 16) | |
623 | #define BIC_CoreCnt (1ULL << 17) | |
624 | #define BIC_PkgTmp (1ULL << 18) | |
625 | #define BIC_GFX_rc6 (1ULL << 19) | |
626 | #define BIC_GFXMHz (1ULL << 20) | |
627 | #define BIC_Pkgpc2 (1ULL << 21) | |
628 | #define BIC_Pkgpc3 (1ULL << 22) | |
629 | #define BIC_Pkgpc6 (1ULL << 23) | |
630 | #define BIC_Pkgpc7 (1ULL << 24) | |
631 | #define BIC_Pkgpc8 (1ULL << 25) | |
632 | #define BIC_Pkgpc9 (1ULL << 26) | |
633 | #define BIC_Pkgpc10 (1ULL << 27) | |
634 | #define BIC_CPU_LPI (1ULL << 28) | |
635 | #define BIC_SYS_LPI (1ULL << 29) | |
636 | #define BIC_PkgWatt (1ULL << 30) | |
637 | #define BIC_CorWatt (1ULL << 31) | |
638 | #define BIC_GFXWatt (1ULL << 32) | |
639 | #define BIC_PkgCnt (1ULL << 33) | |
640 | #define BIC_RAMWatt (1ULL << 34) | |
641 | #define BIC_PKG__ (1ULL << 35) | |
642 | #define BIC_RAM__ (1ULL << 36) | |
643 | #define BIC_Pkg_J (1ULL << 37) | |
644 | #define BIC_Cor_J (1ULL << 38) | |
645 | #define BIC_GFX_J (1ULL << 39) | |
646 | #define BIC_RAM_J (1ULL << 40) | |
647 | #define BIC_Mod_c6 (1ULL << 41) | |
648 | #define BIC_Totl_c0 (1ULL << 42) | |
649 | #define BIC_Any_c0 (1ULL << 43) | |
650 | #define BIC_GFX_c0 (1ULL << 44) | |
651 | #define BIC_CPUGFX (1ULL << 45) | |
652 | #define BIC_Core (1ULL << 46) | |
653 | #define BIC_CPU (1ULL << 47) | |
4c2122d4 LB |
654 | #define BIC_APIC (1ULL << 48) |
655 | #define BIC_X2APIC (1ULL << 49) | |
6de68fe1 | 656 | #define BIC_Die (1ULL << 50) |
b4b91569 | 657 | #define BIC_GFXACTMHz (1ULL << 51) |
2af4f9b8 | 658 | #define BIC_IPC (1ULL << 52) |
3f44a5c6 | 659 | |
4c2122d4 | 660 | #define BIC_DISABLED_BY_DEFAULT (BIC_USEC | BIC_TOD | BIC_APIC | BIC_X2APIC) |
3f44a5c6 LB |
661 | |
662 | unsigned long long bic_enabled = (0xFFFFFFFFFFFFFFFFULL & ~BIC_DISABLED_BY_DEFAULT); | |
4c2122d4 | 663 | unsigned long long bic_present = BIC_USEC | BIC_TOD | BIC_sysfs | BIC_APIC | BIC_X2APIC; |
812db3f7 LB |
664 | |
665 | #define DO_BIC(COUNTER_NAME) (bic_enabled & bic_present & COUNTER_NAME) | |
1e9042b9 | 666 | #define DO_BIC_READ(COUNTER_NAME) (bic_present & COUNTER_NAME) |
3f44a5c6 | 667 | #define ENABLE_BIC(COUNTER_NAME) (bic_enabled |= COUNTER_NAME) |
812db3f7 | 668 | #define BIC_PRESENT(COUNTER_BIT) (bic_present |= COUNTER_BIT) |
0f47c08d | 669 | #define BIC_NOT_PRESENT(COUNTER_BIT) (bic_present &= ~COUNTER_BIT) |
2af4f9b8 | 670 | #define BIC_IS_ENABLED(COUNTER_BIT) (bic_enabled & COUNTER_BIT) |
812db3f7 | 671 | |
3f44a5c6 | 672 | |
dd778a5e LB |
673 | #define MAX_DEFERRED 16 |
674 | char *deferred_skip_names[MAX_DEFERRED]; | |
675 | int deferred_skip_index; | |
676 | ||
677 | /* | |
678 | * HIDE_LIST - hide this list of counters, show the rest [default] | |
679 | * SHOW_LIST - show this list of counters, hide the rest | |
680 | */ | |
681 | enum show_hide_mode { SHOW_LIST, HIDE_LIST } global_show_hide_mode = HIDE_LIST; | |
682 | ||
683 | void help(void) | |
684 | { | |
685 | fprintf(outf, | |
686 | "Usage: turbostat [OPTIONS][(--interval seconds) | COMMAND ...]\n" | |
687 | "\n" | |
688 | "Turbostat forks the specified COMMAND and prints statistics\n" | |
689 | "when COMMAND completes.\n" | |
690 | "If no COMMAND is specified, turbostat wakes every 5-seconds\n" | |
691 | "to print statistics, until interrupted.\n" | |
cc481650 NC |
692 | " -a, --add add a counter\n" |
693 | " eg. --add msr0x10,u64,cpu,delta,MY_TSC\n" | |
694 | " -c, --cpu cpu-set limit output to summary plus cpu-set:\n" | |
695 | " {core | package | j,k,l..m,n-p }\n" | |
9ce80578 NC |
696 | " -d, --debug displays usec, Time_Of_Day_Seconds and more debugging\n" |
697 | " -D, --Dump displays the raw counter values\n" | |
698 | " -e, --enable [all | column]\n" | |
699 | " shows all or the specified disabled column\n" | |
700 | " -H, --hide [column|column,column,...]\n" | |
701 | " hide the specified column(s)\n" | |
cc481650 NC |
702 | " -i, --interval sec.subsec\n" |
703 | " Override default 5-second measurement interval\n" | |
9ce80578 | 704 | " -J, --Joules displays energy in Joules instead of Watts\n" |
cc481650 NC |
705 | " -l, --list list column headers only\n" |
706 | " -n, --num_iterations num\n" | |
707 | " number of the measurement iterations\n" | |
708 | " -o, --out file\n" | |
709 | " create or truncate \"file\" for all output\n" | |
710 | " -q, --quiet skip decoding system configuration header\n" | |
9ce80578 NC |
711 | " -s, --show [column|column,column,...]\n" |
712 | " show only the specified column(s)\n" | |
713 | " -S, --Summary\n" | |
714 | " limits output to 1-line system summary per interval\n" | |
715 | " -T, --TCC temperature\n" | |
716 | " sets the Thermal Control Circuit temperature in\n" | |
717 | " degrees Celsius\n" | |
cc481650 NC |
718 | " -h, --help print this help message\n" |
719 | " -v, --version print version information\n" | |
dd778a5e LB |
720 | "\n" |
721 | "For more help, run \"man turbostat\"\n"); | |
722 | } | |
723 | ||
812db3f7 LB |
724 | /* |
725 | * bic_lookup | |
726 | * for all the strings in comma separate name_list, | |
727 | * set the approprate bit in return value. | |
728 | */ | |
dd778a5e | 729 | unsigned long long bic_lookup(char *name_list, enum show_hide_mode mode) |
812db3f7 LB |
730 | { |
731 | int i; | |
732 | unsigned long long retval = 0; | |
733 | ||
734 | while (name_list) { | |
735 | char *comma; | |
736 | ||
737 | comma = strchr(name_list, ','); | |
738 | ||
739 | if (comma) | |
740 | *comma = '\0'; | |
741 | ||
3f44a5c6 LB |
742 | if (!strcmp(name_list, "all")) |
743 | return ~0; | |
744 | ||
812db3f7 LB |
745 | for (i = 0; i < MAX_BIC; ++i) { |
746 | if (!strcmp(name_list, bic[i].name)) { | |
747 | retval |= (1ULL << i); | |
748 | break; | |
749 | } | |
750 | } | |
751 | if (i == MAX_BIC) { | |
dd778a5e LB |
752 | if (mode == SHOW_LIST) { |
753 | fprintf(stderr, "Invalid counter name: %s\n", name_list); | |
754 | exit(-1); | |
755 | } | |
756 | deferred_skip_names[deferred_skip_index++] = name_list; | |
757 | if (debug) | |
758 | fprintf(stderr, "deferred \"%s\"\n", name_list); | |
759 | if (deferred_skip_index >= MAX_DEFERRED) { | |
760 | fprintf(stderr, "More than max %d un-recognized --skip options '%s'\n", | |
761 | MAX_DEFERRED, name_list); | |
762 | help(); | |
763 | exit(1); | |
764 | } | |
812db3f7 LB |
765 | } |
766 | ||
767 | name_list = comma; | |
768 | if (name_list) | |
769 | name_list++; | |
770 | ||
771 | } | |
772 | return retval; | |
773 | } | |
fc04cc67 | 774 | |
dd778a5e | 775 | |
c8ade361 | 776 | void print_header(char *delim) |
103a8fea | 777 | { |
388e9c81 | 778 | struct msr_counter *mp; |
6168c2e0 | 779 | int printed = 0; |
388e9c81 | 780 | |
3f44a5c6 LB |
781 | if (DO_BIC(BIC_USEC)) |
782 | outp += sprintf(outp, "%susec", (printed++ ? delim : "")); | |
783 | if (DO_BIC(BIC_TOD)) | |
784 | outp += sprintf(outp, "%sTime_Of_Day_Seconds", (printed++ ? delim : "")); | |
812db3f7 | 785 | if (DO_BIC(BIC_Package)) |
6168c2e0 | 786 | outp += sprintf(outp, "%sPackage", (printed++ ? delim : "")); |
6de68fe1 LB |
787 | if (DO_BIC(BIC_Die)) |
788 | outp += sprintf(outp, "%sDie", (printed++ ? delim : "")); | |
01235041 PB |
789 | if (DO_BIC(BIC_Node)) |
790 | outp += sprintf(outp, "%sNode", (printed++ ? delim : "")); | |
812db3f7 | 791 | if (DO_BIC(BIC_Core)) |
6168c2e0 | 792 | outp += sprintf(outp, "%sCore", (printed++ ? delim : "")); |
812db3f7 | 793 | if (DO_BIC(BIC_CPU)) |
6168c2e0 | 794 | outp += sprintf(outp, "%sCPU", (printed++ ? delim : "")); |
4c2122d4 LB |
795 | if (DO_BIC(BIC_APIC)) |
796 | outp += sprintf(outp, "%sAPIC", (printed++ ? delim : "")); | |
797 | if (DO_BIC(BIC_X2APIC)) | |
798 | outp += sprintf(outp, "%sX2APIC", (printed++ ? delim : "")); | |
812db3f7 | 799 | if (DO_BIC(BIC_Avg_MHz)) |
6168c2e0 | 800 | outp += sprintf(outp, "%sAvg_MHz", (printed++ ? delim : "")); |
812db3f7 | 801 | if (DO_BIC(BIC_Busy)) |
6168c2e0 | 802 | outp += sprintf(outp, "%sBusy%%", (printed++ ? delim : "")); |
812db3f7 | 803 | if (DO_BIC(BIC_Bzy_MHz)) |
6168c2e0 | 804 | outp += sprintf(outp, "%sBzy_MHz", (printed++ ? delim : "")); |
812db3f7 | 805 | if (DO_BIC(BIC_TSC_MHz)) |
6168c2e0 | 806 | outp += sprintf(outp, "%sTSC_MHz", (printed++ ? delim : "")); |
1cc21f7b | 807 | |
2af4f9b8 LB |
808 | if (DO_BIC(BIC_IPC)) |
809 | outp += sprintf(outp, "%sIPC", (printed++ ? delim : "")); | |
810 | ||
0de6c0df LB |
811 | if (DO_BIC(BIC_IRQ)) { |
812 | if (sums_need_wide_columns) | |
6168c2e0 | 813 | outp += sprintf(outp, "%s IRQ", (printed++ ? delim : "")); |
0de6c0df | 814 | else |
6168c2e0 | 815 | outp += sprintf(outp, "%sIRQ", (printed++ ? delim : "")); |
0de6c0df LB |
816 | } |
817 | ||
812db3f7 | 818 | if (DO_BIC(BIC_SMI)) |
6168c2e0 | 819 | outp += sprintf(outp, "%sSMI", (printed++ ? delim : "")); |
1cc21f7b | 820 | |
388e9c81 | 821 | for (mp = sys.tp; mp; mp = mp->next) { |
dd778a5e | 822 | |
388e9c81 LB |
823 | if (mp->format == FORMAT_RAW) { |
824 | if (mp->width == 64) | |
dd778a5e | 825 | outp += sprintf(outp, "%s%18.18s", (printed++ ? delim : ""), mp->name); |
388e9c81 | 826 | else |
dd778a5e | 827 | outp += sprintf(outp, "%s%10.10s", (printed++ ? delim : ""), mp->name); |
388e9c81 | 828 | } else { |
0de6c0df | 829 | if ((mp->type == COUNTER_ITEMS) && sums_need_wide_columns) |
dd778a5e | 830 | outp += sprintf(outp, "%s%8s", (printed++ ? delim : ""), mp->name); |
0de6c0df | 831 | else |
dd778a5e | 832 | outp += sprintf(outp, "%s%s", (printed++ ? delim : ""), mp->name); |
388e9c81 LB |
833 | } |
834 | } | |
835 | ||
41618e63 | 836 | if (DO_BIC(BIC_CPU_c1)) |
6168c2e0 | 837 | outp += sprintf(outp, "%sCPU%%c1", (printed++ ? delim : "")); |
562855ee | 838 | if (DO_BIC(BIC_CPU_c3)) |
6168c2e0 | 839 | outp += sprintf(outp, "%sCPU%%c3", (printed++ ? delim : "")); |
812db3f7 | 840 | if (DO_BIC(BIC_CPU_c6)) |
6168c2e0 | 841 | outp += sprintf(outp, "%sCPU%%c6", (printed++ ? delim : "")); |
812db3f7 | 842 | if (DO_BIC(BIC_CPU_c7)) |
6168c2e0 | 843 | outp += sprintf(outp, "%sCPU%%c7", (printed++ ? delim : "")); |
678a3bd1 | 844 | |
0539ba11 | 845 | if (DO_BIC(BIC_Mod_c6)) |
6168c2e0 | 846 | outp += sprintf(outp, "%sMod%%c6", (printed++ ? delim : "")); |
678a3bd1 | 847 | |
812db3f7 | 848 | if (DO_BIC(BIC_CoreTmp)) |
6168c2e0 | 849 | outp += sprintf(outp, "%sCoreTmp", (printed++ ? delim : "")); |
388e9c81 | 850 | |
9392bd98 CW |
851 | if (do_rapl && !rapl_joules) { |
852 | if (DO_BIC(BIC_CorWatt) && (do_rapl & RAPL_PER_CORE_ENERGY)) | |
853 | outp += sprintf(outp, "%sCorWatt", (printed++ ? delim : "")); | |
854 | } else if (do_rapl && rapl_joules) { | |
855 | if (DO_BIC(BIC_Cor_J) && (do_rapl & RAPL_PER_CORE_ENERGY)) | |
856 | outp += sprintf(outp, "%sCor_J", (printed++ ? delim : "")); | |
857 | } | |
858 | ||
388e9c81 LB |
859 | for (mp = sys.cp; mp; mp = mp->next) { |
860 | if (mp->format == FORMAT_RAW) { | |
861 | if (mp->width == 64) | |
c8ade361 | 862 | outp += sprintf(outp, "%s%18.18s", delim, mp->name); |
388e9c81 | 863 | else |
c8ade361 | 864 | outp += sprintf(outp, "%s%10.10s", delim, mp->name); |
388e9c81 | 865 | } else { |
0de6c0df LB |
866 | if ((mp->type == COUNTER_ITEMS) && sums_need_wide_columns) |
867 | outp += sprintf(outp, "%s%8s", delim, mp->name); | |
868 | else | |
869 | outp += sprintf(outp, "%s%s", delim, mp->name); | |
388e9c81 LB |
870 | } |
871 | } | |
872 | ||
812db3f7 | 873 | if (DO_BIC(BIC_PkgTmp)) |
6168c2e0 | 874 | outp += sprintf(outp, "%sPkgTmp", (printed++ ? delim : "")); |
889facbe | 875 | |
812db3f7 | 876 | if (DO_BIC(BIC_GFX_rc6)) |
6168c2e0 | 877 | outp += sprintf(outp, "%sGFX%%rc6", (printed++ ? delim : "")); |
fdf676e5 | 878 | |
812db3f7 | 879 | if (DO_BIC(BIC_GFXMHz)) |
6168c2e0 | 880 | outp += sprintf(outp, "%sGFXMHz", (printed++ ? delim : "")); |
27d47356 | 881 | |
b4b91569 RA |
882 | if (DO_BIC(BIC_GFXACTMHz)) |
883 | outp += sprintf(outp, "%sGFXAMHz", (printed++ ? delim : "")); | |
884 | ||
a99d8730 | 885 | if (DO_BIC(BIC_Totl_c0)) |
6168c2e0 | 886 | outp += sprintf(outp, "%sTotl%%C0", (printed++ ? delim : "")); |
a99d8730 | 887 | if (DO_BIC(BIC_Any_c0)) |
6168c2e0 | 888 | outp += sprintf(outp, "%sAny%%C0", (printed++ ? delim : "")); |
a99d8730 | 889 | if (DO_BIC(BIC_GFX_c0)) |
6168c2e0 | 890 | outp += sprintf(outp, "%sGFX%%C0", (printed++ ? delim : "")); |
a99d8730 | 891 | if (DO_BIC(BIC_CPUGFX)) |
6168c2e0 | 892 | outp += sprintf(outp, "%sCPUGFX%%", (printed++ ? delim : "")); |
0b2bb692 | 893 | |
0f47c08d | 894 | if (DO_BIC(BIC_Pkgpc2)) |
6168c2e0 | 895 | outp += sprintf(outp, "%sPkg%%pc2", (printed++ ? delim : "")); |
0f47c08d | 896 | if (DO_BIC(BIC_Pkgpc3)) |
6168c2e0 | 897 | outp += sprintf(outp, "%sPkg%%pc3", (printed++ ? delim : "")); |
0f47c08d | 898 | if (DO_BIC(BIC_Pkgpc6)) |
6168c2e0 | 899 | outp += sprintf(outp, "%sPkg%%pc6", (printed++ ? delim : "")); |
0f47c08d | 900 | if (DO_BIC(BIC_Pkgpc7)) |
6168c2e0 | 901 | outp += sprintf(outp, "%sPkg%%pc7", (printed++ ? delim : "")); |
0f47c08d | 902 | if (DO_BIC(BIC_Pkgpc8)) |
6168c2e0 | 903 | outp += sprintf(outp, "%sPkg%%pc8", (printed++ ? delim : "")); |
0f47c08d | 904 | if (DO_BIC(BIC_Pkgpc9)) |
6168c2e0 | 905 | outp += sprintf(outp, "%sPkg%%pc9", (printed++ ? delim : "")); |
0f47c08d | 906 | if (DO_BIC(BIC_Pkgpc10)) |
6168c2e0 | 907 | outp += sprintf(outp, "%sPk%%pc10", (printed++ ? delim : "")); |
be0e54c4 LB |
908 | if (DO_BIC(BIC_CPU_LPI)) |
909 | outp += sprintf(outp, "%sCPU%%LPI", (printed++ ? delim : "")); | |
910 | if (DO_BIC(BIC_SYS_LPI)) | |
911 | outp += sprintf(outp, "%sSYS%%LPI", (printed++ ? delim : "")); | |
103a8fea | 912 | |
5c56be9a | 913 | if (do_rapl && !rapl_joules) { |
812db3f7 | 914 | if (DO_BIC(BIC_PkgWatt)) |
6168c2e0 | 915 | outp += sprintf(outp, "%sPkgWatt", (printed++ ? delim : "")); |
9392bd98 | 916 | if (DO_BIC(BIC_CorWatt) && !(do_rapl & RAPL_PER_CORE_ENERGY)) |
6168c2e0 | 917 | outp += sprintf(outp, "%sCorWatt", (printed++ ? delim : "")); |
812db3f7 | 918 | if (DO_BIC(BIC_GFXWatt)) |
6168c2e0 | 919 | outp += sprintf(outp, "%sGFXWatt", (printed++ ? delim : "")); |
812db3f7 | 920 | if (DO_BIC(BIC_RAMWatt)) |
6168c2e0 | 921 | outp += sprintf(outp, "%sRAMWatt", (printed++ ? delim : "")); |
812db3f7 | 922 | if (DO_BIC(BIC_PKG__)) |
6168c2e0 | 923 | outp += sprintf(outp, "%sPKG_%%", (printed++ ? delim : "")); |
812db3f7 | 924 | if (DO_BIC(BIC_RAM__)) |
6168c2e0 | 925 | outp += sprintf(outp, "%sRAM_%%", (printed++ ? delim : "")); |
d7899447 | 926 | } else if (do_rapl && rapl_joules) { |
812db3f7 | 927 | if (DO_BIC(BIC_Pkg_J)) |
6168c2e0 | 928 | outp += sprintf(outp, "%sPkg_J", (printed++ ? delim : "")); |
9392bd98 | 929 | if (DO_BIC(BIC_Cor_J) && !(do_rapl & RAPL_PER_CORE_ENERGY)) |
6168c2e0 | 930 | outp += sprintf(outp, "%sCor_J", (printed++ ? delim : "")); |
812db3f7 | 931 | if (DO_BIC(BIC_GFX_J)) |
6168c2e0 | 932 | outp += sprintf(outp, "%sGFX_J", (printed++ ? delim : "")); |
812db3f7 | 933 | if (DO_BIC(BIC_RAM_J)) |
6168c2e0 | 934 | outp += sprintf(outp, "%sRAM_J", (printed++ ? delim : "")); |
812db3f7 | 935 | if (DO_BIC(BIC_PKG__)) |
6168c2e0 | 936 | outp += sprintf(outp, "%sPKG_%%", (printed++ ? delim : "")); |
812db3f7 | 937 | if (DO_BIC(BIC_RAM__)) |
6168c2e0 | 938 | outp += sprintf(outp, "%sRAM_%%", (printed++ ? delim : "")); |
5c56be9a | 939 | } |
388e9c81 LB |
940 | for (mp = sys.pp; mp; mp = mp->next) { |
941 | if (mp->format == FORMAT_RAW) { | |
942 | if (mp->width == 64) | |
c8ade361 | 943 | outp += sprintf(outp, "%s%18.18s", delim, mp->name); |
388e9c81 | 944 | else |
c8ade361 | 945 | outp += sprintf(outp, "%s%10.10s", delim, mp->name); |
388e9c81 | 946 | } else { |
0de6c0df LB |
947 | if ((mp->type == COUNTER_ITEMS) && sums_need_wide_columns) |
948 | outp += sprintf(outp, "%s%8s", delim, mp->name); | |
949 | else | |
950 | outp += sprintf(outp, "%s%s", delim, mp->name); | |
388e9c81 LB |
951 | } |
952 | } | |
953 | ||
c98d5d94 | 954 | outp += sprintf(outp, "\n"); |
103a8fea LB |
955 | } |
956 | ||
c98d5d94 LB |
957 | int dump_counters(struct thread_data *t, struct core_data *c, |
958 | struct pkg_data *p) | |
103a8fea | 959 | { |
388e9c81 LB |
960 | int i; |
961 | struct msr_counter *mp; | |
962 | ||
3b4d5c7f | 963 | outp += sprintf(outp, "t %p, c %p, p %p\n", t, c, p); |
c98d5d94 LB |
964 | |
965 | if (t) { | |
3b4d5c7f AS |
966 | outp += sprintf(outp, "CPU: %d flags 0x%x\n", |
967 | t->cpu_id, t->flags); | |
968 | outp += sprintf(outp, "TSC: %016llX\n", t->tsc); | |
969 | outp += sprintf(outp, "aperf: %016llX\n", t->aperf); | |
970 | outp += sprintf(outp, "mperf: %016llX\n", t->mperf); | |
971 | outp += sprintf(outp, "c1: %016llX\n", t->c1); | |
6886fee4 | 972 | |
2af4f9b8 LB |
973 | if (DO_BIC(BIC_IPC)) |
974 | outp += sprintf(outp, "IPC: %lld\n", t->instr_count); | |
975 | ||
812db3f7 | 976 | if (DO_BIC(BIC_IRQ)) |
0de6c0df | 977 | outp += sprintf(outp, "IRQ: %lld\n", t->irq_count); |
812db3f7 | 978 | if (DO_BIC(BIC_SMI)) |
218f0e8d | 979 | outp += sprintf(outp, "SMI: %d\n", t->smi_count); |
388e9c81 LB |
980 | |
981 | for (i = 0, mp = sys.tp; mp; i++, mp = mp->next) { | |
982 | outp += sprintf(outp, "tADDED [%d] msr0x%x: %08llX\n", | |
983 | i, mp->msr_num, t->counter[i]); | |
984 | } | |
c98d5d94 | 985 | } |
103a8fea | 986 | |
c98d5d94 | 987 | if (c) { |
3b4d5c7f AS |
988 | outp += sprintf(outp, "core: %d\n", c->core_id); |
989 | outp += sprintf(outp, "c3: %016llX\n", c->c3); | |
990 | outp += sprintf(outp, "c6: %016llX\n", c->c6); | |
991 | outp += sprintf(outp, "c7: %016llX\n", c->c7); | |
992 | outp += sprintf(outp, "DTS: %dC\n", c->core_temp_c); | |
9392bd98 | 993 | outp += sprintf(outp, "Joules: %0X\n", c->core_energy); |
388e9c81 LB |
994 | |
995 | for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) { | |
996 | outp += sprintf(outp, "cADDED [%d] msr0x%x: %08llX\n", | |
997 | i, mp->msr_num, c->counter[i]); | |
998 | } | |
0539ba11 | 999 | outp += sprintf(outp, "mc6_us: %016llX\n", c->mc6_us); |
c98d5d94 | 1000 | } |
103a8fea | 1001 | |
c98d5d94 | 1002 | if (p) { |
3b4d5c7f | 1003 | outp += sprintf(outp, "package: %d\n", p->package_id); |
0b2bb692 LB |
1004 | |
1005 | outp += sprintf(outp, "Weighted cores: %016llX\n", p->pkg_wtd_core_c0); | |
1006 | outp += sprintf(outp, "Any cores: %016llX\n", p->pkg_any_core_c0); | |
1007 | outp += sprintf(outp, "Any GFX: %016llX\n", p->pkg_any_gfxe_c0); | |
1008 | outp += sprintf(outp, "CPU + GFX: %016llX\n", p->pkg_both_core_gfxe_c0); | |
1009 | ||
3b4d5c7f | 1010 | outp += sprintf(outp, "pc2: %016llX\n", p->pc2); |
0f47c08d | 1011 | if (DO_BIC(BIC_Pkgpc3)) |
ee7e38e3 | 1012 | outp += sprintf(outp, "pc3: %016llX\n", p->pc3); |
0f47c08d | 1013 | if (DO_BIC(BIC_Pkgpc6)) |
ee7e38e3 | 1014 | outp += sprintf(outp, "pc6: %016llX\n", p->pc6); |
0f47c08d | 1015 | if (DO_BIC(BIC_Pkgpc7)) |
ee7e38e3 | 1016 | outp += sprintf(outp, "pc7: %016llX\n", p->pc7); |
3b4d5c7f AS |
1017 | outp += sprintf(outp, "pc8: %016llX\n", p->pc8); |
1018 | outp += sprintf(outp, "pc9: %016llX\n", p->pc9); | |
1019 | outp += sprintf(outp, "pc10: %016llX\n", p->pc10); | |
be0e54c4 LB |
1020 | outp += sprintf(outp, "cpu_lpi: %016llX\n", p->cpu_lpi); |
1021 | outp += sprintf(outp, "sys_lpi: %016llX\n", p->sys_lpi); | |
7c2ccc50 CY |
1022 | outp += sprintf(outp, "Joules PKG: %0llX\n", p->energy_pkg); |
1023 | outp += sprintf(outp, "Joules COR: %0llX\n", p->energy_cores); | |
1024 | outp += sprintf(outp, "Joules GFX: %0llX\n", p->energy_gfx); | |
1025 | outp += sprintf(outp, "Joules RAM: %0llX\n", p->energy_dram); | |
1026 | outp += sprintf(outp, "Throttle PKG: %0llX\n", | |
3b4d5c7f | 1027 | p->rapl_pkg_perf_status); |
7c2ccc50 | 1028 | outp += sprintf(outp, "Throttle RAM: %0llX\n", |
3b4d5c7f AS |
1029 | p->rapl_dram_perf_status); |
1030 | outp += sprintf(outp, "PTM: %dC\n", p->pkg_temp_c); | |
388e9c81 LB |
1031 | |
1032 | for (i = 0, mp = sys.pp; mp; i++, mp = mp->next) { | |
1033 | outp += sprintf(outp, "pADDED [%d] msr0x%x: %08llX\n", | |
1034 | i, mp->msr_num, p->counter[i]); | |
1035 | } | |
c98d5d94 | 1036 | } |
3b4d5c7f AS |
1037 | |
1038 | outp += sprintf(outp, "\n"); | |
1039 | ||
c98d5d94 | 1040 | return 0; |
103a8fea LB |
1041 | } |
1042 | ||
e23da037 LB |
1043 | /* |
1044 | * column formatting convention & formats | |
e23da037 | 1045 | */ |
c98d5d94 LB |
1046 | int format_counters(struct thread_data *t, struct core_data *c, |
1047 | struct pkg_data *p) | |
103a8fea | 1048 | { |
008d396e | 1049 | double interval_float, tsc; |
fc04cc67 | 1050 | char *fmt8; |
388e9c81 LB |
1051 | int i; |
1052 | struct msr_counter *mp; | |
6168c2e0 LB |
1053 | char *delim = "\t"; |
1054 | int printed = 0; | |
103a8fea | 1055 | |
c98d5d94 LB |
1056 | /* if showing only 1st thread in core and this isn't one, bail out */ |
1057 | if (show_core_only && !(t->flags & CPU_IS_FIRST_THREAD_IN_CORE)) | |
1058 | return 0; | |
1059 | ||
1060 | /* if showing only 1st thread in pkg and this isn't one, bail out */ | |
1061 | if (show_pkg_only && !(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE)) | |
1062 | return 0; | |
1063 | ||
1ef7d21a LB |
1064 | /*if not summary line and --cpu is used */ |
1065 | if ((t != &average.threads) && | |
1066 | (cpu_subset && !CPU_ISSET_S(t->cpu_id, cpu_subset_size, cpu_subset))) | |
1067 | return 0; | |
1068 | ||
3f44a5c6 | 1069 | if (DO_BIC(BIC_USEC)) { |
f4fdf2b4 LB |
1070 | /* on each row, print how many usec each timestamp took to gather */ |
1071 | struct timeval tv; | |
1072 | ||
1073 | timersub(&t->tv_end, &t->tv_begin, &tv); | |
1074 | outp += sprintf(outp, "%5ld\t", tv.tv_sec * 1000000 + tv.tv_usec); | |
1075 | } | |
1076 | ||
3f44a5c6 LB |
1077 | /* Time_Of_Day_Seconds: on each row, print sec.usec last timestamp taken */ |
1078 | if (DO_BIC(BIC_TOD)) | |
1079 | outp += sprintf(outp, "%10ld.%06ld\t", t->tv_end.tv_sec, t->tv_end.tv_usec); | |
1080 | ||
d4794f25 | 1081 | interval_float = t->tv_delta.tv_sec + t->tv_delta.tv_usec/1000000.0; |
103a8fea | 1082 | |
008d396e LB |
1083 | tsc = t->tsc * tsc_tweak; |
1084 | ||
c98d5d94 LB |
1085 | /* topo columns, print blanks on 1st (average) line */ |
1086 | if (t == &average.threads) { | |
812db3f7 | 1087 | if (DO_BIC(BIC_Package)) |
6168c2e0 | 1088 | outp += sprintf(outp, "%s-", (printed++ ? delim : "")); |
6de68fe1 LB |
1089 | if (DO_BIC(BIC_Die)) |
1090 | outp += sprintf(outp, "%s-", (printed++ ? delim : "")); | |
01235041 PB |
1091 | if (DO_BIC(BIC_Node)) |
1092 | outp += sprintf(outp, "%s-", (printed++ ? delim : "")); | |
812db3f7 | 1093 | if (DO_BIC(BIC_Core)) |
6168c2e0 | 1094 | outp += sprintf(outp, "%s-", (printed++ ? delim : "")); |
812db3f7 | 1095 | if (DO_BIC(BIC_CPU)) |
6168c2e0 | 1096 | outp += sprintf(outp, "%s-", (printed++ ? delim : "")); |
4c2122d4 LB |
1097 | if (DO_BIC(BIC_APIC)) |
1098 | outp += sprintf(outp, "%s-", (printed++ ? delim : "")); | |
1099 | if (DO_BIC(BIC_X2APIC)) | |
1100 | outp += sprintf(outp, "%s-", (printed++ ? delim : "")); | |
103a8fea | 1101 | } else { |
812db3f7 | 1102 | if (DO_BIC(BIC_Package)) { |
c98d5d94 | 1103 | if (p) |
6168c2e0 | 1104 | outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), p->package_id); |
c98d5d94 | 1105 | else |
6168c2e0 | 1106 | outp += sprintf(outp, "%s-", (printed++ ? delim : "")); |
c98d5d94 | 1107 | } |
6de68fe1 LB |
1108 | if (DO_BIC(BIC_Die)) { |
1109 | if (c) | |
1110 | outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), cpus[t->cpu_id].die_id); | |
1111 | else | |
1112 | outp += sprintf(outp, "%s-", (printed++ ? delim : "")); | |
1113 | } | |
01235041 PB |
1114 | if (DO_BIC(BIC_Node)) { |
1115 | if (t) | |
1116 | outp += sprintf(outp, "%s%d", | |
1117 | (printed++ ? delim : ""), | |
1118 | cpus[t->cpu_id].physical_node_id); | |
1119 | else | |
1120 | outp += sprintf(outp, "%s-", | |
1121 | (printed++ ? delim : "")); | |
1122 | } | |
812db3f7 | 1123 | if (DO_BIC(BIC_Core)) { |
c98d5d94 | 1124 | if (c) |
6168c2e0 | 1125 | outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), c->core_id); |
c98d5d94 | 1126 | else |
6168c2e0 | 1127 | outp += sprintf(outp, "%s-", (printed++ ? delim : "")); |
c98d5d94 | 1128 | } |
812db3f7 | 1129 | if (DO_BIC(BIC_CPU)) |
6168c2e0 | 1130 | outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), t->cpu_id); |
4c2122d4 LB |
1131 | if (DO_BIC(BIC_APIC)) |
1132 | outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), t->apic_id); | |
1133 | if (DO_BIC(BIC_X2APIC)) | |
1134 | outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), t->x2apic_id); | |
103a8fea | 1135 | } |
fc04cc67 | 1136 | |
812db3f7 | 1137 | if (DO_BIC(BIC_Avg_MHz)) |
6168c2e0 | 1138 | outp += sprintf(outp, "%s%.0f", (printed++ ? delim : ""), |
fc04cc67 LB |
1139 | 1.0 / units * t->aperf / interval_float); |
1140 | ||
812db3f7 | 1141 | if (DO_BIC(BIC_Busy)) |
6168c2e0 | 1142 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * t->mperf/tsc); |
103a8fea | 1143 | |
812db3f7 | 1144 | if (DO_BIC(BIC_Bzy_MHz)) { |
21ed5574 | 1145 | if (has_base_hz) |
6168c2e0 | 1146 | outp += sprintf(outp, "%s%.0f", (printed++ ? delim : ""), base_hz / units * t->aperf / t->mperf); |
21ed5574 | 1147 | else |
6168c2e0 | 1148 | outp += sprintf(outp, "%s%.0f", (printed++ ? delim : ""), |
008d396e | 1149 | tsc / units * t->aperf / t->mperf / interval_float); |
21ed5574 | 1150 | } |
103a8fea | 1151 | |
812db3f7 | 1152 | if (DO_BIC(BIC_TSC_MHz)) |
6168c2e0 | 1153 | outp += sprintf(outp, "%s%.0f", (printed++ ? delim : ""), 1.0 * t->tsc/units/interval_float); |
103a8fea | 1154 | |
2af4f9b8 LB |
1155 | if (DO_BIC(BIC_IPC)) |
1156 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 1.0 * t->instr_count / t->aperf); | |
1157 | ||
562a2d37 | 1158 | /* IRQ */ |
0de6c0df LB |
1159 | if (DO_BIC(BIC_IRQ)) { |
1160 | if (sums_need_wide_columns) | |
6168c2e0 | 1161 | outp += sprintf(outp, "%s%8lld", (printed++ ? delim : ""), t->irq_count); |
0de6c0df | 1162 | else |
6168c2e0 | 1163 | outp += sprintf(outp, "%s%lld", (printed++ ? delim : ""), t->irq_count); |
0de6c0df | 1164 | } |
562a2d37 | 1165 | |
1cc21f7b | 1166 | /* SMI */ |
812db3f7 | 1167 | if (DO_BIC(BIC_SMI)) |
6168c2e0 | 1168 | outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), t->smi_count); |
1cc21f7b | 1169 | |
678a3bd1 | 1170 | /* Added counters */ |
388e9c81 LB |
1171 | for (i = 0, mp = sys.tp; mp; i++, mp = mp->next) { |
1172 | if (mp->format == FORMAT_RAW) { | |
1173 | if (mp->width == 32) | |
5f3aea57 | 1174 | outp += sprintf(outp, "%s0x%08x", (printed++ ? delim : ""), (unsigned int) t->counter[i]); |
388e9c81 | 1175 | else |
6168c2e0 | 1176 | outp += sprintf(outp, "%s0x%016llx", (printed++ ? delim : ""), t->counter[i]); |
388e9c81 | 1177 | } else if (mp->format == FORMAT_DELTA) { |
0de6c0df | 1178 | if ((mp->type == COUNTER_ITEMS) && sums_need_wide_columns) |
6168c2e0 | 1179 | outp += sprintf(outp, "%s%8lld", (printed++ ? delim : ""), t->counter[i]); |
0de6c0df | 1180 | else |
6168c2e0 | 1181 | outp += sprintf(outp, "%s%lld", (printed++ ? delim : ""), t->counter[i]); |
388e9c81 | 1182 | } else if (mp->format == FORMAT_PERCENT) { |
41618e63 | 1183 | if (mp->type == COUNTER_USEC) |
6168c2e0 | 1184 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), t->counter[i]/interval_float/10000); |
41618e63 | 1185 | else |
6168c2e0 | 1186 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * t->counter[i]/tsc); |
388e9c81 LB |
1187 | } |
1188 | } | |
1189 | ||
41618e63 LB |
1190 | /* C1 */ |
1191 | if (DO_BIC(BIC_CPU_c1)) | |
6168c2e0 | 1192 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * t->c1/tsc); |
41618e63 LB |
1193 | |
1194 | ||
678a3bd1 LB |
1195 | /* print per-core data only for 1st thread in core */ |
1196 | if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE)) | |
1197 | goto done; | |
1198 | ||
562855ee | 1199 | if (DO_BIC(BIC_CPU_c3)) |
6168c2e0 | 1200 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * c->c3/tsc); |
812db3f7 | 1201 | if (DO_BIC(BIC_CPU_c6)) |
6168c2e0 | 1202 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * c->c6/tsc); |
812db3f7 | 1203 | if (DO_BIC(BIC_CPU_c7)) |
6168c2e0 | 1204 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * c->c7/tsc); |
678a3bd1 | 1205 | |
0539ba11 LB |
1206 | /* Mod%c6 */ |
1207 | if (DO_BIC(BIC_Mod_c6)) | |
6168c2e0 | 1208 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * c->mc6_us / tsc); |
0539ba11 | 1209 | |
812db3f7 | 1210 | if (DO_BIC(BIC_CoreTmp)) |
6168c2e0 | 1211 | outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), c->core_temp_c); |
889facbe | 1212 | |
388e9c81 LB |
1213 | for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) { |
1214 | if (mp->format == FORMAT_RAW) { | |
1215 | if (mp->width == 32) | |
5f3aea57 | 1216 | outp += sprintf(outp, "%s0x%08x", (printed++ ? delim : ""), (unsigned int) c->counter[i]); |
388e9c81 | 1217 | else |
6168c2e0 | 1218 | outp += sprintf(outp, "%s0x%016llx", (printed++ ? delim : ""), c->counter[i]); |
388e9c81 | 1219 | } else if (mp->format == FORMAT_DELTA) { |
0de6c0df | 1220 | if ((mp->type == COUNTER_ITEMS) && sums_need_wide_columns) |
6168c2e0 | 1221 | outp += sprintf(outp, "%s%8lld", (printed++ ? delim : ""), c->counter[i]); |
0de6c0df | 1222 | else |
6168c2e0 | 1223 | outp += sprintf(outp, "%s%lld", (printed++ ? delim : ""), c->counter[i]); |
388e9c81 | 1224 | } else if (mp->format == FORMAT_PERCENT) { |
6168c2e0 | 1225 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * c->counter[i]/tsc); |
388e9c81 LB |
1226 | } |
1227 | } | |
1228 | ||
9972d5d8 | 1229 | fmt8 = "%s%.2f"; |
9392bd98 CW |
1230 | |
1231 | if (DO_BIC(BIC_CorWatt) && (do_rapl & RAPL_PER_CORE_ENERGY)) | |
1232 | outp += sprintf(outp, fmt8, (printed++ ? delim : ""), c->core_energy * rapl_energy_units / interval_float); | |
1233 | if (DO_BIC(BIC_Cor_J) && (do_rapl & RAPL_PER_CORE_ENERGY)) | |
1234 | outp += sprintf(outp, fmt8, (printed++ ? delim : ""), c->core_energy * rapl_energy_units); | |
1235 | ||
c98d5d94 LB |
1236 | /* print per-package data only for 1st core in package */ |
1237 | if (!(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE)) | |
1238 | goto done; | |
1239 | ||
0b2bb692 | 1240 | /* PkgTmp */ |
812db3f7 | 1241 | if (DO_BIC(BIC_PkgTmp)) |
6168c2e0 | 1242 | outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), p->pkg_temp_c); |
889facbe | 1243 | |
fdf676e5 | 1244 | /* GFXrc6 */ |
812db3f7 | 1245 | if (DO_BIC(BIC_GFX_rc6)) { |
ba3dec99 | 1246 | if (p->gfx_rc6_ms == -1) { /* detect GFX counter reset */ |
6168c2e0 | 1247 | outp += sprintf(outp, "%s**.**", (printed++ ? delim : "")); |
9185e988 | 1248 | } else { |
6168c2e0 | 1249 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), |
9185e988 LB |
1250 | p->gfx_rc6_ms / 10.0 / interval_float); |
1251 | } | |
1252 | } | |
fdf676e5 | 1253 | |
27d47356 | 1254 | /* GFXMHz */ |
812db3f7 | 1255 | if (DO_BIC(BIC_GFXMHz)) |
6168c2e0 | 1256 | outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), p->gfx_mhz); |
27d47356 | 1257 | |
b4b91569 RA |
1258 | /* GFXACTMHz */ |
1259 | if (DO_BIC(BIC_GFXACTMHz)) | |
1260 | outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), p->gfx_act_mhz); | |
1261 | ||
0b2bb692 | 1262 | /* Totl%C0, Any%C0 GFX%C0 CPUGFX% */ |
a99d8730 | 1263 | if (DO_BIC(BIC_Totl_c0)) |
6168c2e0 | 1264 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pkg_wtd_core_c0/tsc); |
a99d8730 | 1265 | if (DO_BIC(BIC_Any_c0)) |
6168c2e0 | 1266 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pkg_any_core_c0/tsc); |
a99d8730 | 1267 | if (DO_BIC(BIC_GFX_c0)) |
6168c2e0 | 1268 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pkg_any_gfxe_c0/tsc); |
a99d8730 | 1269 | if (DO_BIC(BIC_CPUGFX)) |
6168c2e0 | 1270 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pkg_both_core_gfxe_c0/tsc); |
0b2bb692 | 1271 | |
0f47c08d | 1272 | if (DO_BIC(BIC_Pkgpc2)) |
6168c2e0 | 1273 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc2/tsc); |
0f47c08d | 1274 | if (DO_BIC(BIC_Pkgpc3)) |
6168c2e0 | 1275 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc3/tsc); |
0f47c08d | 1276 | if (DO_BIC(BIC_Pkgpc6)) |
6168c2e0 | 1277 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc6/tsc); |
0f47c08d | 1278 | if (DO_BIC(BIC_Pkgpc7)) |
6168c2e0 | 1279 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc7/tsc); |
0f47c08d | 1280 | if (DO_BIC(BIC_Pkgpc8)) |
6168c2e0 | 1281 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc8/tsc); |
0f47c08d | 1282 | if (DO_BIC(BIC_Pkgpc9)) |
6168c2e0 | 1283 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc9/tsc); |
0f47c08d | 1284 | if (DO_BIC(BIC_Pkgpc10)) |
6168c2e0 | 1285 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc10/tsc); |
889facbe | 1286 | |
be0e54c4 LB |
1287 | if (DO_BIC(BIC_CPU_LPI)) |
1288 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->cpu_lpi / 1000000.0 / interval_float); | |
1289 | if (DO_BIC(BIC_SYS_LPI)) | |
1290 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->sys_lpi / 1000000.0 / interval_float); | |
1291 | ||
812db3f7 | 1292 | if (DO_BIC(BIC_PkgWatt)) |
6168c2e0 | 1293 | outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_pkg * rapl_energy_units / interval_float); |
9392bd98 | 1294 | if (DO_BIC(BIC_CorWatt) && !(do_rapl & RAPL_PER_CORE_ENERGY)) |
6168c2e0 | 1295 | outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_cores * rapl_energy_units / interval_float); |
812db3f7 | 1296 | if (DO_BIC(BIC_GFXWatt)) |
6168c2e0 | 1297 | outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_gfx * rapl_energy_units / interval_float); |
812db3f7 | 1298 | if (DO_BIC(BIC_RAMWatt)) |
6168c2e0 | 1299 | outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_dram * rapl_dram_energy_units / interval_float); |
812db3f7 | 1300 | if (DO_BIC(BIC_Pkg_J)) |
6168c2e0 | 1301 | outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_pkg * rapl_energy_units); |
9392bd98 | 1302 | if (DO_BIC(BIC_Cor_J) && !(do_rapl & RAPL_PER_CORE_ENERGY)) |
6168c2e0 | 1303 | outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_cores * rapl_energy_units); |
812db3f7 | 1304 | if (DO_BIC(BIC_GFX_J)) |
6168c2e0 | 1305 | outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_gfx * rapl_energy_units); |
812db3f7 | 1306 | if (DO_BIC(BIC_RAM_J)) |
6168c2e0 | 1307 | outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_dram * rapl_dram_energy_units); |
812db3f7 | 1308 | if (DO_BIC(BIC_PKG__)) |
6168c2e0 | 1309 | outp += sprintf(outp, fmt8, (printed++ ? delim : ""), 100.0 * p->rapl_pkg_perf_status * rapl_time_units / interval_float); |
812db3f7 | 1310 | if (DO_BIC(BIC_RAM__)) |
6168c2e0 | 1311 | outp += sprintf(outp, fmt8, (printed++ ? delim : ""), 100.0 * p->rapl_dram_perf_status * rapl_time_units / interval_float); |
812db3f7 | 1312 | |
388e9c81 LB |
1313 | for (i = 0, mp = sys.pp; mp; i++, mp = mp->next) { |
1314 | if (mp->format == FORMAT_RAW) { | |
1315 | if (mp->width == 32) | |
5f3aea57 | 1316 | outp += sprintf(outp, "%s0x%08x", (printed++ ? delim : ""), (unsigned int) p->counter[i]); |
388e9c81 | 1317 | else |
6168c2e0 | 1318 | outp += sprintf(outp, "%s0x%016llx", (printed++ ? delim : ""), p->counter[i]); |
388e9c81 | 1319 | } else if (mp->format == FORMAT_DELTA) { |
0de6c0df | 1320 | if ((mp->type == COUNTER_ITEMS) && sums_need_wide_columns) |
6168c2e0 | 1321 | outp += sprintf(outp, "%s%8lld", (printed++ ? delim : ""), p->counter[i]); |
0de6c0df | 1322 | else |
6168c2e0 | 1323 | outp += sprintf(outp, "%s%lld", (printed++ ? delim : ""), p->counter[i]); |
388e9c81 | 1324 | } else if (mp->format == FORMAT_PERCENT) { |
6168c2e0 | 1325 | outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->counter[i]/tsc); |
388e9c81 LB |
1326 | } |
1327 | } | |
1328 | ||
c98d5d94 | 1329 | done: |
94d6ab4b LB |
1330 | if (*(outp - 1) != '\n') |
1331 | outp += sprintf(outp, "\n"); | |
c98d5d94 LB |
1332 | |
1333 | return 0; | |
103a8fea LB |
1334 | } |
1335 | ||
b7d8c148 | 1336 | void flush_output_stdout(void) |
c98d5d94 | 1337 | { |
b7d8c148 LB |
1338 | FILE *filep; |
1339 | ||
1340 | if (outf == stderr) | |
1341 | filep = stdout; | |
1342 | else | |
1343 | filep = outf; | |
1344 | ||
1345 | fputs(output_buffer, filep); | |
1346 | fflush(filep); | |
1347 | ||
c98d5d94 LB |
1348 | outp = output_buffer; |
1349 | } | |
b7d8c148 | 1350 | void flush_output_stderr(void) |
c98d5d94 | 1351 | { |
b7d8c148 LB |
1352 | fputs(output_buffer, outf); |
1353 | fflush(outf); | |
c98d5d94 LB |
1354 | outp = output_buffer; |
1355 | } | |
1356 | void format_all_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p) | |
103a8fea | 1357 | { |
e23da037 | 1358 | static int printed; |
103a8fea | 1359 | |
e23da037 | 1360 | if (!printed || !summary_only) |
c8ade361 | 1361 | print_header("\t"); |
103a8fea | 1362 | |
9d83601a | 1363 | format_counters(&average.threads, &average.cores, &average.packages); |
103a8fea | 1364 | |
e23da037 LB |
1365 | printed = 1; |
1366 | ||
1367 | if (summary_only) | |
1368 | return; | |
1369 | ||
c98d5d94 | 1370 | for_all_cpus(format_counters, t, c, p); |
103a8fea LB |
1371 | } |
1372 | ||
889facbe | 1373 | #define DELTA_WRAP32(new, old) \ |
7c2ccc50 | 1374 | old = ((((unsigned long long)new << 32) - ((unsigned long long)old << 32)) >> 32); |
889facbe | 1375 | |
ba3dec99 | 1376 | int |
c98d5d94 LB |
1377 | delta_package(struct pkg_data *new, struct pkg_data *old) |
1378 | { | |
388e9c81 LB |
1379 | int i; |
1380 | struct msr_counter *mp; | |
0b2bb692 | 1381 | |
a99d8730 LB |
1382 | |
1383 | if (DO_BIC(BIC_Totl_c0)) | |
0b2bb692 | 1384 | old->pkg_wtd_core_c0 = new->pkg_wtd_core_c0 - old->pkg_wtd_core_c0; |
a99d8730 | 1385 | if (DO_BIC(BIC_Any_c0)) |
0b2bb692 | 1386 | old->pkg_any_core_c0 = new->pkg_any_core_c0 - old->pkg_any_core_c0; |
a99d8730 | 1387 | if (DO_BIC(BIC_GFX_c0)) |
0b2bb692 | 1388 | old->pkg_any_gfxe_c0 = new->pkg_any_gfxe_c0 - old->pkg_any_gfxe_c0; |
a99d8730 | 1389 | if (DO_BIC(BIC_CPUGFX)) |
0b2bb692 | 1390 | old->pkg_both_core_gfxe_c0 = new->pkg_both_core_gfxe_c0 - old->pkg_both_core_gfxe_c0; |
a99d8730 | 1391 | |
c98d5d94 | 1392 | old->pc2 = new->pc2 - old->pc2; |
0f47c08d | 1393 | if (DO_BIC(BIC_Pkgpc3)) |
ee7e38e3 | 1394 | old->pc3 = new->pc3 - old->pc3; |
0f47c08d | 1395 | if (DO_BIC(BIC_Pkgpc6)) |
ee7e38e3 | 1396 | old->pc6 = new->pc6 - old->pc6; |
0f47c08d | 1397 | if (DO_BIC(BIC_Pkgpc7)) |
ee7e38e3 | 1398 | old->pc7 = new->pc7 - old->pc7; |
ca58710f KCA |
1399 | old->pc8 = new->pc8 - old->pc8; |
1400 | old->pc9 = new->pc9 - old->pc9; | |
1401 | old->pc10 = new->pc10 - old->pc10; | |
be0e54c4 LB |
1402 | old->cpu_lpi = new->cpu_lpi - old->cpu_lpi; |
1403 | old->sys_lpi = new->sys_lpi - old->sys_lpi; | |
889facbe LB |
1404 | old->pkg_temp_c = new->pkg_temp_c; |
1405 | ||
9185e988 LB |
1406 | /* flag an error when rc6 counter resets/wraps */ |
1407 | if (old->gfx_rc6_ms > new->gfx_rc6_ms) | |
1408 | old->gfx_rc6_ms = -1; | |
1409 | else | |
1410 | old->gfx_rc6_ms = new->gfx_rc6_ms - old->gfx_rc6_ms; | |
1411 | ||
27d47356 | 1412 | old->gfx_mhz = new->gfx_mhz; |
b4b91569 | 1413 | old->gfx_act_mhz = new->gfx_act_mhz; |
27d47356 | 1414 | |
87e15da9 CY |
1415 | old->energy_pkg = new->energy_pkg - old->energy_pkg; |
1416 | old->energy_cores = new->energy_cores - old->energy_cores; | |
1417 | old->energy_gfx = new->energy_gfx - old->energy_gfx; | |
1418 | old->energy_dram = new->energy_dram - old->energy_dram; | |
1419 | old->rapl_pkg_perf_status = new->rapl_pkg_perf_status - old->rapl_pkg_perf_status; | |
1420 | old->rapl_dram_perf_status = new->rapl_dram_perf_status - old->rapl_dram_perf_status; | |
ba3dec99 | 1421 | |
388e9c81 LB |
1422 | for (i = 0, mp = sys.pp; mp; i++, mp = mp->next) { |
1423 | if (mp->format == FORMAT_RAW) | |
1424 | old->counter[i] = new->counter[i]; | |
1425 | else | |
1426 | old->counter[i] = new->counter[i] - old->counter[i]; | |
1427 | } | |
1428 | ||
ba3dec99 | 1429 | return 0; |
c98d5d94 | 1430 | } |
103a8fea | 1431 | |
c98d5d94 LB |
1432 | void |
1433 | delta_core(struct core_data *new, struct core_data *old) | |
103a8fea | 1434 | { |
388e9c81 LB |
1435 | int i; |
1436 | struct msr_counter *mp; | |
1437 | ||
c98d5d94 LB |
1438 | old->c3 = new->c3 - old->c3; |
1439 | old->c6 = new->c6 - old->c6; | |
1440 | old->c7 = new->c7 - old->c7; | |
889facbe | 1441 | old->core_temp_c = new->core_temp_c; |
0539ba11 | 1442 | old->mc6_us = new->mc6_us - old->mc6_us; |
388e9c81 | 1443 | |
9392bd98 CW |
1444 | DELTA_WRAP32(new->core_energy, old->core_energy); |
1445 | ||
388e9c81 LB |
1446 | for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) { |
1447 | if (mp->format == FORMAT_RAW) | |
1448 | old->counter[i] = new->counter[i]; | |
1449 | else | |
1450 | old->counter[i] = new->counter[i] - old->counter[i]; | |
1451 | } | |
c98d5d94 | 1452 | } |
103a8fea | 1453 | |
1e9042b9 SP |
1454 | int soft_c1_residency_display(int bic) |
1455 | { | |
1456 | if (!DO_BIC(BIC_CPU_c1) || use_c1_residency_msr) | |
1457 | return 0; | |
1458 | ||
1459 | return DO_BIC_READ(bic); | |
1460 | } | |
1461 | ||
c3ae331d LB |
1462 | /* |
1463 | * old = new - old | |
1464 | */ | |
ba3dec99 | 1465 | int |
c98d5d94 LB |
1466 | delta_thread(struct thread_data *new, struct thread_data *old, |
1467 | struct core_data *core_delta) | |
1468 | { | |
388e9c81 LB |
1469 | int i; |
1470 | struct msr_counter *mp; | |
1471 | ||
4c2122d4 LB |
1472 | /* we run cpuid just the 1st time, copy the results */ |
1473 | if (DO_BIC(BIC_APIC)) | |
1474 | new->apic_id = old->apic_id; | |
1475 | if (DO_BIC(BIC_X2APIC)) | |
1476 | new->x2apic_id = old->x2apic_id; | |
1477 | ||
3f44a5c6 LB |
1478 | /* |
1479 | * the timestamps from start of measurement interval are in "old" | |
1480 | * the timestamp from end of measurement interval are in "new" | |
1481 | * over-write old w/ new so we can print end of interval values | |
1482 | */ | |
1483 | ||
d4794f25 | 1484 | timersub(&new->tv_begin, &old->tv_begin, &old->tv_delta); |
3f44a5c6 LB |
1485 | old->tv_begin = new->tv_begin; |
1486 | old->tv_end = new->tv_end; | |
1487 | ||
c98d5d94 LB |
1488 | old->tsc = new->tsc - old->tsc; |
1489 | ||
1490 | /* check for TSC < 1 Mcycles over interval */ | |
b2c95d90 JT |
1491 | if (old->tsc < (1000 * 1000)) |
1492 | errx(-3, "Insanely slow TSC rate, TSC stops in idle?\n" | |
1493 | "You can disable all c-states by booting with \"idle=poll\"\n" | |
1494 | "or just the deep ones with \"processor.max_cstate=1\""); | |
103a8fea | 1495 | |
c98d5d94 | 1496 | old->c1 = new->c1 - old->c1; |
103a8fea | 1497 | |
1e9042b9 SP |
1498 | if (DO_BIC(BIC_Avg_MHz) || DO_BIC(BIC_Busy) || DO_BIC(BIC_Bzy_MHz) || |
1499 | soft_c1_residency_display(BIC_Avg_MHz)) { | |
a729617c LB |
1500 | if ((new->aperf > old->aperf) && (new->mperf > old->mperf)) { |
1501 | old->aperf = new->aperf - old->aperf; | |
1502 | old->mperf = new->mperf - old->mperf; | |
1503 | } else { | |
ba3dec99 | 1504 | return -1; |
103a8fea | 1505 | } |
c98d5d94 | 1506 | } |
103a8fea | 1507 | |
103a8fea | 1508 | |
144b44b1 LB |
1509 | if (use_c1_residency_msr) { |
1510 | /* | |
1511 | * Some models have a dedicated C1 residency MSR, | |
1512 | * which should be more accurate than the derivation below. | |
1513 | */ | |
1514 | } else { | |
1515 | /* | |
1516 | * As counter collection is not atomic, | |
1517 | * it is possible for mperf's non-halted cycles + idle states | |
1518 | * to exceed TSC's all cycles: show c1 = 0% in that case. | |
1519 | */ | |
95149369 | 1520 | if ((old->mperf + core_delta->c3 + core_delta->c6 + core_delta->c7) > (old->tsc * tsc_tweak)) |
144b44b1 LB |
1521 | old->c1 = 0; |
1522 | else { | |
1523 | /* normal case, derive c1 */ | |
008d396e | 1524 | old->c1 = (old->tsc * tsc_tweak) - old->mperf - core_delta->c3 |
c98d5d94 | 1525 | - core_delta->c6 - core_delta->c7; |
144b44b1 | 1526 | } |
c98d5d94 | 1527 | } |
c3ae331d | 1528 | |
c98d5d94 | 1529 | if (old->mperf == 0) { |
b7d8c148 LB |
1530 | if (debug > 1) |
1531 | fprintf(outf, "cpu%d MPERF 0!\n", old->cpu_id); | |
c98d5d94 | 1532 | old->mperf = 1; /* divide by 0 protection */ |
103a8fea | 1533 | } |
c98d5d94 | 1534 | |
2af4f9b8 LB |
1535 | if (DO_BIC(BIC_IPC)) |
1536 | old->instr_count = new->instr_count - old->instr_count; | |
1537 | ||
812db3f7 | 1538 | if (DO_BIC(BIC_IRQ)) |
562a2d37 LB |
1539 | old->irq_count = new->irq_count - old->irq_count; |
1540 | ||
812db3f7 | 1541 | if (DO_BIC(BIC_SMI)) |
1ed51011 | 1542 | old->smi_count = new->smi_count - old->smi_count; |
ba3dec99 | 1543 | |
388e9c81 LB |
1544 | for (i = 0, mp = sys.tp; mp; i++, mp = mp->next) { |
1545 | if (mp->format == FORMAT_RAW) | |
1546 | old->counter[i] = new->counter[i]; | |
1547 | else | |
1548 | old->counter[i] = new->counter[i] - old->counter[i]; | |
1549 | } | |
ba3dec99 | 1550 | return 0; |
c98d5d94 LB |
1551 | } |
1552 | ||
1553 | int delta_cpu(struct thread_data *t, struct core_data *c, | |
1554 | struct pkg_data *p, struct thread_data *t2, | |
1555 | struct core_data *c2, struct pkg_data *p2) | |
1556 | { | |
ba3dec99 LB |
1557 | int retval = 0; |
1558 | ||
c98d5d94 LB |
1559 | /* calculate core delta only for 1st thread in core */ |
1560 | if (t->flags & CPU_IS_FIRST_THREAD_IN_CORE) | |
1561 | delta_core(c, c2); | |
1562 | ||
1563 | /* always calculate thread delta */ | |
ba3dec99 LB |
1564 | retval = delta_thread(t, t2, c2); /* c2 is core delta */ |
1565 | if (retval) | |
1566 | return retval; | |
c98d5d94 LB |
1567 | |
1568 | /* calculate package delta only for 1st core in package */ | |
1569 | if (t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE) | |
ba3dec99 | 1570 | retval = delta_package(p, p2); |
c98d5d94 | 1571 | |
ba3dec99 | 1572 | return retval; |
103a8fea LB |
1573 | } |
1574 | ||
c98d5d94 LB |
1575 | void clear_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p) |
1576 | { | |
388e9c81 LB |
1577 | int i; |
1578 | struct msr_counter *mp; | |
1579 | ||
3f44a5c6 LB |
1580 | t->tv_begin.tv_sec = 0; |
1581 | t->tv_begin.tv_usec = 0; | |
1582 | t->tv_end.tv_sec = 0; | |
1583 | t->tv_end.tv_usec = 0; | |
d4794f25 YG |
1584 | t->tv_delta.tv_sec = 0; |
1585 | t->tv_delta.tv_usec = 0; | |
3f44a5c6 | 1586 | |
c98d5d94 LB |
1587 | t->tsc = 0; |
1588 | t->aperf = 0; | |
1589 | t->mperf = 0; | |
1590 | t->c1 = 0; | |
1591 | ||
2af4f9b8 LB |
1592 | t->instr_count = 0; |
1593 | ||
562a2d37 LB |
1594 | t->irq_count = 0; |
1595 | t->smi_count = 0; | |
1596 | ||
c98d5d94 LB |
1597 | /* tells format_counters to dump all fields from this set */ |
1598 | t->flags = CPU_IS_FIRST_THREAD_IN_CORE | CPU_IS_FIRST_CORE_IN_PACKAGE; | |
1599 | ||
1600 | c->c3 = 0; | |
1601 | c->c6 = 0; | |
1602 | c->c7 = 0; | |
0539ba11 | 1603 | c->mc6_us = 0; |
889facbe | 1604 | c->core_temp_c = 0; |
9392bd98 | 1605 | c->core_energy = 0; |
c98d5d94 | 1606 | |
0b2bb692 LB |
1607 | p->pkg_wtd_core_c0 = 0; |
1608 | p->pkg_any_core_c0 = 0; | |
1609 | p->pkg_any_gfxe_c0 = 0; | |
1610 | p->pkg_both_core_gfxe_c0 = 0; | |
1611 | ||
c98d5d94 | 1612 | p->pc2 = 0; |
0f47c08d | 1613 | if (DO_BIC(BIC_Pkgpc3)) |
ee7e38e3 | 1614 | p->pc3 = 0; |
0f47c08d | 1615 | if (DO_BIC(BIC_Pkgpc6)) |
ee7e38e3 | 1616 | p->pc6 = 0; |
0f47c08d | 1617 | if (DO_BIC(BIC_Pkgpc7)) |
ee7e38e3 | 1618 | p->pc7 = 0; |
ca58710f KCA |
1619 | p->pc8 = 0; |
1620 | p->pc9 = 0; | |
1621 | p->pc10 = 0; | |
be0e54c4 LB |
1622 | p->cpu_lpi = 0; |
1623 | p->sys_lpi = 0; | |
889facbe LB |
1624 | |
1625 | p->energy_pkg = 0; | |
1626 | p->energy_dram = 0; | |
1627 | p->energy_cores = 0; | |
1628 | p->energy_gfx = 0; | |
1629 | p->rapl_pkg_perf_status = 0; | |
1630 | p->rapl_dram_perf_status = 0; | |
1631 | p->pkg_temp_c = 0; | |
27d47356 | 1632 | |
fdf676e5 | 1633 | p->gfx_rc6_ms = 0; |
27d47356 | 1634 | p->gfx_mhz = 0; |
b4b91569 | 1635 | p->gfx_act_mhz = 0; |
388e9c81 LB |
1636 | for (i = 0, mp = sys.tp; mp; i++, mp = mp->next) |
1637 | t->counter[i] = 0; | |
1638 | ||
1639 | for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) | |
1640 | c->counter[i] = 0; | |
1641 | ||
1642 | for (i = 0, mp = sys.pp; mp; i++, mp = mp->next) | |
1643 | p->counter[i] = 0; | |
c98d5d94 LB |
1644 | } |
1645 | int sum_counters(struct thread_data *t, struct core_data *c, | |
1646 | struct pkg_data *p) | |
103a8fea | 1647 | { |
388e9c81 LB |
1648 | int i; |
1649 | struct msr_counter *mp; | |
1650 | ||
4c2122d4 LB |
1651 | /* copy un-changing apic_id's */ |
1652 | if (DO_BIC(BIC_APIC)) | |
1653 | average.threads.apic_id = t->apic_id; | |
1654 | if (DO_BIC(BIC_X2APIC)) | |
1655 | average.threads.x2apic_id = t->x2apic_id; | |
1656 | ||
3f44a5c6 LB |
1657 | /* remember first tv_begin */ |
1658 | if (average.threads.tv_begin.tv_sec == 0) | |
1659 | average.threads.tv_begin = t->tv_begin; | |
1660 | ||
1661 | /* remember last tv_end */ | |
1662 | average.threads.tv_end = t->tv_end; | |
1663 | ||
c98d5d94 LB |
1664 | average.threads.tsc += t->tsc; |
1665 | average.threads.aperf += t->aperf; | |
1666 | average.threads.mperf += t->mperf; | |
1667 | average.threads.c1 += t->c1; | |
103a8fea | 1668 | |
2af4f9b8 LB |
1669 | average.threads.instr_count += t->instr_count; |
1670 | ||
562a2d37 LB |
1671 | average.threads.irq_count += t->irq_count; |
1672 | average.threads.smi_count += t->smi_count; | |
1673 | ||
388e9c81 LB |
1674 | for (i = 0, mp = sys.tp; mp; i++, mp = mp->next) { |
1675 | if (mp->format == FORMAT_RAW) | |
1676 | continue; | |
1677 | average.threads.counter[i] += t->counter[i]; | |
1678 | } | |
1679 | ||
c98d5d94 LB |
1680 | /* sum per-core values only for 1st thread in core */ |
1681 | if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE)) | |
1682 | return 0; | |
103a8fea | 1683 | |
c98d5d94 LB |
1684 | average.cores.c3 += c->c3; |
1685 | average.cores.c6 += c->c6; | |
1686 | average.cores.c7 += c->c7; | |
0539ba11 | 1687 | average.cores.mc6_us += c->mc6_us; |
c98d5d94 | 1688 | |
889facbe LB |
1689 | average.cores.core_temp_c = MAX(average.cores.core_temp_c, c->core_temp_c); |
1690 | ||
9392bd98 CW |
1691 | average.cores.core_energy += c->core_energy; |
1692 | ||
388e9c81 LB |
1693 | for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) { |
1694 | if (mp->format == FORMAT_RAW) | |
1695 | continue; | |
1696 | average.cores.counter[i] += c->counter[i]; | |
1697 | } | |
1698 | ||
c98d5d94 LB |
1699 | /* sum per-pkg values only for 1st core in pkg */ |
1700 | if (!(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE)) | |
1701 | return 0; | |
1702 | ||
a99d8730 | 1703 | if (DO_BIC(BIC_Totl_c0)) |
0b2bb692 | 1704 | average.packages.pkg_wtd_core_c0 += p->pkg_wtd_core_c0; |
a99d8730 | 1705 | if (DO_BIC(BIC_Any_c0)) |
0b2bb692 | 1706 | average.packages.pkg_any_core_c0 += p->pkg_any_core_c0; |
a99d8730 | 1707 | if (DO_BIC(BIC_GFX_c0)) |
0b2bb692 | 1708 | average.packages.pkg_any_gfxe_c0 += p->pkg_any_gfxe_c0; |
a99d8730 | 1709 | if (DO_BIC(BIC_CPUGFX)) |
0b2bb692 | 1710 | average.packages.pkg_both_core_gfxe_c0 += p->pkg_both_core_gfxe_c0; |
0b2bb692 | 1711 | |
c98d5d94 | 1712 | average.packages.pc2 += p->pc2; |
0f47c08d | 1713 | if (DO_BIC(BIC_Pkgpc3)) |
ee7e38e3 | 1714 | average.packages.pc3 += p->pc3; |
0f47c08d | 1715 | if (DO_BIC(BIC_Pkgpc6)) |
ee7e38e3 | 1716 | average.packages.pc6 += p->pc6; |
0f47c08d | 1717 | if (DO_BIC(BIC_Pkgpc7)) |
ee7e38e3 | 1718 | average.packages.pc7 += p->pc7; |
ca58710f KCA |
1719 | average.packages.pc8 += p->pc8; |
1720 | average.packages.pc9 += p->pc9; | |
1721 | average.packages.pc10 += p->pc10; | |
c98d5d94 | 1722 | |
be0e54c4 LB |
1723 | average.packages.cpu_lpi = p->cpu_lpi; |
1724 | average.packages.sys_lpi = p->sys_lpi; | |
1725 | ||
889facbe LB |
1726 | average.packages.energy_pkg += p->energy_pkg; |
1727 | average.packages.energy_dram += p->energy_dram; | |
1728 | average.packages.energy_cores += p->energy_cores; | |
1729 | average.packages.energy_gfx += p->energy_gfx; | |
1730 | ||
fdf676e5 | 1731 | average.packages.gfx_rc6_ms = p->gfx_rc6_ms; |
27d47356 | 1732 | average.packages.gfx_mhz = p->gfx_mhz; |
b4b91569 | 1733 | average.packages.gfx_act_mhz = p->gfx_act_mhz; |
27d47356 | 1734 | |
889facbe LB |
1735 | average.packages.pkg_temp_c = MAX(average.packages.pkg_temp_c, p->pkg_temp_c); |
1736 | ||
1737 | average.packages.rapl_pkg_perf_status += p->rapl_pkg_perf_status; | |
1738 | average.packages.rapl_dram_perf_status += p->rapl_dram_perf_status; | |
388e9c81 LB |
1739 | |
1740 | for (i = 0, mp = sys.pp; mp; i++, mp = mp->next) { | |
1741 | if (mp->format == FORMAT_RAW) | |
1742 | continue; | |
1743 | average.packages.counter[i] += p->counter[i]; | |
1744 | } | |
c98d5d94 LB |
1745 | return 0; |
1746 | } | |
1747 | /* | |
1748 | * sum the counters for all cpus in the system | |
1749 | * compute the weighted average | |
1750 | */ | |
1751 | void compute_average(struct thread_data *t, struct core_data *c, | |
1752 | struct pkg_data *p) | |
1753 | { | |
388e9c81 LB |
1754 | int i; |
1755 | struct msr_counter *mp; | |
1756 | ||
c98d5d94 LB |
1757 | clear_counters(&average.threads, &average.cores, &average.packages); |
1758 | ||
1759 | for_all_cpus(sum_counters, t, c, p); | |
1760 | ||
d4794f25 YG |
1761 | /* Use the global time delta for the average. */ |
1762 | average.threads.tv_delta = tv_delta; | |
1763 | ||
c98d5d94 LB |
1764 | average.threads.tsc /= topo.num_cpus; |
1765 | average.threads.aperf /= topo.num_cpus; | |
1766 | average.threads.mperf /= topo.num_cpus; | |
2af4f9b8 | 1767 | average.threads.instr_count /= topo.num_cpus; |
c98d5d94 LB |
1768 | average.threads.c1 /= topo.num_cpus; |
1769 | ||
0de6c0df LB |
1770 | if (average.threads.irq_count > 9999999) |
1771 | sums_need_wide_columns = 1; | |
1772 | ||
c98d5d94 LB |
1773 | average.cores.c3 /= topo.num_cores; |
1774 | average.cores.c6 /= topo.num_cores; | |
1775 | average.cores.c7 /= topo.num_cores; | |
0539ba11 | 1776 | average.cores.mc6_us /= topo.num_cores; |
c98d5d94 | 1777 | |
a99d8730 | 1778 | if (DO_BIC(BIC_Totl_c0)) |
0b2bb692 | 1779 | average.packages.pkg_wtd_core_c0 /= topo.num_packages; |
a99d8730 | 1780 | if (DO_BIC(BIC_Any_c0)) |
0b2bb692 | 1781 | average.packages.pkg_any_core_c0 /= topo.num_packages; |
a99d8730 | 1782 | if (DO_BIC(BIC_GFX_c0)) |
0b2bb692 | 1783 | average.packages.pkg_any_gfxe_c0 /= topo.num_packages; |
a99d8730 | 1784 | if (DO_BIC(BIC_CPUGFX)) |
0b2bb692 | 1785 | average.packages.pkg_both_core_gfxe_c0 /= topo.num_packages; |
0b2bb692 | 1786 | |
c98d5d94 | 1787 | average.packages.pc2 /= topo.num_packages; |
0f47c08d | 1788 | if (DO_BIC(BIC_Pkgpc3)) |
ee7e38e3 | 1789 | average.packages.pc3 /= topo.num_packages; |
0f47c08d | 1790 | if (DO_BIC(BIC_Pkgpc6)) |
ee7e38e3 | 1791 | average.packages.pc6 /= topo.num_packages; |
0f47c08d | 1792 | if (DO_BIC(BIC_Pkgpc7)) |
ee7e38e3 | 1793 | average.packages.pc7 /= topo.num_packages; |
ca58710f KCA |
1794 | |
1795 | average.packages.pc8 /= topo.num_packages; | |
1796 | average.packages.pc9 /= topo.num_packages; | |
1797 | average.packages.pc10 /= topo.num_packages; | |
388e9c81 LB |
1798 | |
1799 | for (i = 0, mp = sys.tp; mp; i++, mp = mp->next) { | |
1800 | if (mp->format == FORMAT_RAW) | |
1801 | continue; | |
0de6c0df LB |
1802 | if (mp->type == COUNTER_ITEMS) { |
1803 | if (average.threads.counter[i] > 9999999) | |
1804 | sums_need_wide_columns = 1; | |
41618e63 | 1805 | continue; |
0de6c0df | 1806 | } |
388e9c81 LB |
1807 | average.threads.counter[i] /= topo.num_cpus; |
1808 | } | |
1809 | for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) { | |
1810 | if (mp->format == FORMAT_RAW) | |
1811 | continue; | |
0de6c0df LB |
1812 | if (mp->type == COUNTER_ITEMS) { |
1813 | if (average.cores.counter[i] > 9999999) | |
1814 | sums_need_wide_columns = 1; | |
1815 | } | |
388e9c81 LB |
1816 | average.cores.counter[i] /= topo.num_cores; |
1817 | } | |
1818 | for (i = 0, mp = sys.pp; mp; i++, mp = mp->next) { | |
1819 | if (mp->format == FORMAT_RAW) | |
1820 | continue; | |
0de6c0df LB |
1821 | if (mp->type == COUNTER_ITEMS) { |
1822 | if (average.packages.counter[i] > 9999999) | |
1823 | sums_need_wide_columns = 1; | |
1824 | } | |
388e9c81 LB |
1825 | average.packages.counter[i] /= topo.num_packages; |
1826 | } | |
103a8fea LB |
1827 | } |
1828 | ||
c98d5d94 | 1829 | static unsigned long long rdtsc(void) |
103a8fea | 1830 | { |
c98d5d94 | 1831 | unsigned int low, high; |
15aaa346 | 1832 | |
c98d5d94 | 1833 | asm volatile("rdtsc" : "=a" (low), "=d" (high)); |
15aaa346 | 1834 | |
c98d5d94 LB |
1835 | return low | ((unsigned long long)high) << 32; |
1836 | } | |
15aaa346 | 1837 | |
495c7654 LB |
1838 | /* |
1839 | * Open a file, and exit on failure | |
1840 | */ | |
1841 | FILE *fopen_or_die(const char *path, const char *mode) | |
1842 | { | |
1843 | FILE *filep = fopen(path, mode); | |
1844 | ||
1845 | if (!filep) | |
1846 | err(1, "%s: open failed", path); | |
1847 | return filep; | |
1848 | } | |
1849 | /* | |
1850 | * snapshot_sysfs_counter() | |
1851 | * | |
1852 | * return snapshot of given counter | |
1853 | */ | |
1854 | unsigned long long snapshot_sysfs_counter(char *path) | |
1855 | { | |
1856 | FILE *fp; | |
1857 | int retval; | |
1858 | unsigned long long counter; | |
1859 | ||
1860 | fp = fopen_or_die(path, "r"); | |
1861 | ||
1862 | retval = fscanf(fp, "%lld", &counter); | |
1863 | if (retval != 1) | |
1864 | err(1, "snapshot_sysfs_counter(%s)", path); | |
1865 | ||
1866 | fclose(fp); | |
1867 | ||
1868 | return counter; | |
1869 | } | |
1870 | ||
1871 | int get_mp(int cpu, struct msr_counter *mp, unsigned long long *counterp) | |
1872 | { | |
1873 | if (mp->msr_num != 0) { | |
1874 | if (get_msr(cpu, mp->msr_num, counterp)) | |
1875 | return -1; | |
1876 | } else { | |
46c27978 | 1877 | char path[128 + PATH_BYTES]; |
41618e63 LB |
1878 | |
1879 | if (mp->flags & SYSFS_PERCPU) { | |
1880 | sprintf(path, "/sys/devices/system/cpu/cpu%d/%s", | |
1881 | cpu, mp->path); | |
1882 | ||
1883 | *counterp = snapshot_sysfs_counter(path); | |
1884 | } else { | |
1885 | *counterp = snapshot_sysfs_counter(mp->path); | |
1886 | } | |
495c7654 LB |
1887 | } |
1888 | ||
1889 | return 0; | |
1890 | } | |
1891 | ||
6d6501d9 BP |
1892 | int get_epb(int cpu) |
1893 | { | |
1894 | char path[128 + PATH_BYTES]; | |
7f1b11ba | 1895 | unsigned long long msr; |
6d6501d9 BP |
1896 | int ret, epb = -1; |
1897 | FILE *fp; | |
1898 | ||
1899 | sprintf(path, "/sys/devices/system/cpu/cpu%d/power/energy_perf_bias", cpu); | |
1900 | ||
7f1b11ba BP |
1901 | fp = fopen(path, "r"); |
1902 | if (!fp) | |
1903 | goto msr_fallback; | |
6d6501d9 BP |
1904 | |
1905 | ret = fscanf(fp, "%d", &epb); | |
1906 | if (ret != 1) | |
1907 | err(1, "%s(%s)", __func__, path); | |
1908 | ||
1909 | fclose(fp); | |
1910 | ||
1911 | return epb; | |
7f1b11ba BP |
1912 | |
1913 | msr_fallback: | |
1914 | get_msr(cpu, MSR_IA32_ENERGY_PERF_BIAS, &msr); | |
1915 | ||
1916 | return msr & 0xf; | |
6d6501d9 BP |
1917 | } |
1918 | ||
4c2122d4 LB |
1919 | void get_apic_id(struct thread_data *t) |
1920 | { | |
34041551 | 1921 | unsigned int eax, ebx, ecx, edx; |
4c2122d4 | 1922 | |
34041551 LB |
1923 | if (DO_BIC(BIC_APIC)) { |
1924 | eax = ebx = ecx = edx = 0; | |
1925 | __cpuid(1, eax, ebx, ecx, edx); | |
4c2122d4 | 1926 | |
34041551 LB |
1927 | t->apic_id = (ebx >> 24) & 0xff; |
1928 | } | |
1929 | ||
1930 | if (!DO_BIC(BIC_X2APIC)) | |
4c2122d4 LB |
1931 | return; |
1932 | ||
c1c10cc7 | 1933 | if (authentic_amd || hygon_genuine) { |
34041551 | 1934 | unsigned int topology_extensions; |
4c2122d4 | 1935 | |
34041551 LB |
1936 | if (max_extended_level < 0x8000001e) |
1937 | return; | |
4c2122d4 | 1938 | |
34041551 LB |
1939 | eax = ebx = ecx = edx = 0; |
1940 | __cpuid(0x80000001, eax, ebx, ecx, edx); | |
1941 | topology_extensions = ecx & (1 << 22); | |
1942 | ||
1943 | if (topology_extensions == 0) | |
1944 | return; | |
1945 | ||
1946 | eax = ebx = ecx = edx = 0; | |
1947 | __cpuid(0x8000001e, eax, ebx, ecx, edx); | |
1948 | ||
1949 | t->x2apic_id = eax; | |
4c2122d4 | 1950 | return; |
34041551 | 1951 | } |
4c2122d4 | 1952 | |
34041551 LB |
1953 | if (!genuine_intel) |
1954 | return; | |
1955 | ||
1956 | if (max_level < 0xb) | |
4c2122d4 LB |
1957 | return; |
1958 | ||
1959 | ecx = 0; | |
1960 | __cpuid(0xb, eax, ebx, ecx, edx); | |
1961 | t->x2apic_id = edx; | |
1962 | ||
34041551 LB |
1963 | if (debug && (t->apic_id != (t->x2apic_id & 0xff))) |
1964 | fprintf(outf, "cpu%d: BIOS BUG: apic 0x%x x2apic 0x%x\n", | |
1965 | t->cpu_id, t->apic_id, t->x2apic_id); | |
4c2122d4 LB |
1966 | } |
1967 | ||
c98d5d94 LB |
1968 | /* |
1969 | * get_counters(...) | |
1970 | * migrate to cpu | |
1971 | * acquire and record local counters for that cpu | |
1972 | */ | |
1973 | int get_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p) | |
1974 | { | |
1975 | int cpu = t->cpu_id; | |
889facbe | 1976 | unsigned long long msr; |
0102b067 | 1977 | int aperf_mperf_retry_count = 0; |
388e9c81 LB |
1978 | struct msr_counter *mp; |
1979 | int i; | |
88c3281f | 1980 | |
e52966c0 | 1981 | if (cpu_migrate(cpu)) { |
3d7772ea | 1982 | fprintf(outf, "get_counters: Could not migrate to CPU %d\n", cpu); |
c98d5d94 | 1983 | return -1; |
e52966c0 | 1984 | } |
15aaa346 | 1985 | |
d4794f25 YG |
1986 | gettimeofday(&t->tv_begin, (struct timezone *)NULL); |
1987 | ||
4c2122d4 LB |
1988 | if (first_counter_read) |
1989 | get_apic_id(t); | |
0102b067 | 1990 | retry: |
c98d5d94 LB |
1991 | t->tsc = rdtsc(); /* we are running on local CPU of interest */ |
1992 | ||
1e9042b9 SP |
1993 | if (DO_BIC(BIC_Avg_MHz) || DO_BIC(BIC_Busy) || DO_BIC(BIC_Bzy_MHz) || |
1994 | soft_c1_residency_display(BIC_Avg_MHz)) { | |
0102b067 LB |
1995 | unsigned long long tsc_before, tsc_between, tsc_after, aperf_time, mperf_time; |
1996 | ||
1997 | /* | |
1998 | * The TSC, APERF and MPERF must be read together for | |
1999 | * APERF/MPERF and MPERF/TSC to give accurate results. | |
2000 | * | |
2001 | * Unfortunately, APERF and MPERF are read by | |
2002 | * individual system call, so delays may occur | |
2003 | * between them. If the time to read them | |
2004 | * varies by a large amount, we re-read them. | |
2005 | */ | |
2006 | ||
2007 | /* | |
2008 | * This initial dummy APERF read has been seen to | |
2009 | * reduce jitter in the subsequent reads. | |
2010 | */ | |
2011 | ||
2012 | if (get_msr(cpu, MSR_IA32_APERF, &t->aperf)) | |
2013 | return -3; | |
2014 | ||
2015 | t->tsc = rdtsc(); /* re-read close to APERF */ | |
2016 | ||
2017 | tsc_before = t->tsc; | |
2018 | ||
9c63a650 | 2019 | if (get_msr(cpu, MSR_IA32_APERF, &t->aperf)) |
c98d5d94 | 2020 | return -3; |
0102b067 LB |
2021 | |
2022 | tsc_between = rdtsc(); | |
2023 | ||
9c63a650 | 2024 | if (get_msr(cpu, MSR_IA32_MPERF, &t->mperf)) |
c98d5d94 | 2025 | return -4; |
0102b067 LB |
2026 | |
2027 | tsc_after = rdtsc(); | |
2028 | ||
2029 | aperf_time = tsc_between - tsc_before; | |
2030 | mperf_time = tsc_after - tsc_between; | |
2031 | ||
2032 | /* | |
2033 | * If the system call latency to read APERF and MPERF | |
2034 | * differ by more than 2x, then try again. | |
2035 | */ | |
2036 | if ((aperf_time > (2 * mperf_time)) || (mperf_time > (2 * aperf_time))) { | |
2037 | aperf_mperf_retry_count++; | |
2038 | if (aperf_mperf_retry_count < 5) | |
2039 | goto retry; | |
2040 | else | |
2041 | warnx("cpu%d jitter %lld %lld", | |
2042 | cpu, aperf_time, mperf_time); | |
2043 | } | |
2044 | aperf_mperf_retry_count = 0; | |
2045 | ||
b2b34dfe HC |
2046 | t->aperf = t->aperf * aperf_mperf_multiplier; |
2047 | t->mperf = t->mperf * aperf_mperf_multiplier; | |
c98d5d94 LB |
2048 | } |
2049 | ||
2af4f9b8 LB |
2050 | if (DO_BIC(BIC_IPC)) |
2051 | if (read(get_instr_count_fd(cpu), &t->instr_count, sizeof(long long)) != sizeof(long long)) | |
2052 | return -4; | |
2053 | ||
812db3f7 | 2054 | if (DO_BIC(BIC_IRQ)) |
562a2d37 | 2055 | t->irq_count = irqs_per_cpu[cpu]; |
812db3f7 | 2056 | if (DO_BIC(BIC_SMI)) { |
1ed51011 LB |
2057 | if (get_msr(cpu, MSR_SMI_COUNT, &msr)) |
2058 | return -5; | |
2059 | t->smi_count = msr & 0xFFFFFFFF; | |
2060 | } | |
0539ba11 | 2061 | if (DO_BIC(BIC_CPU_c1) && use_c1_residency_msr) { |
144b44b1 LB |
2062 | if (get_msr(cpu, MSR_CORE_C1_RES, &t->c1)) |
2063 | return -6; | |
2064 | } | |
2065 | ||
388e9c81 | 2066 | for (i = 0, mp = sys.tp; mp; i++, mp = mp->next) { |
495c7654 | 2067 | if (get_mp(cpu, mp, &t->counter[i])) |
388e9c81 LB |
2068 | return -10; |
2069 | } | |
2070 | ||
c98d5d94 LB |
2071 | /* collect core counters only for 1st thread in core */ |
2072 | if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE)) | |
f4fdf2b4 | 2073 | goto done; |
c98d5d94 | 2074 | |
1e9042b9 | 2075 | if (DO_BIC(BIC_CPU_c3) || soft_c1_residency_display(BIC_CPU_c3)) { |
c98d5d94 LB |
2076 | if (get_msr(cpu, MSR_CORE_C3_RESIDENCY, &c->c3)) |
2077 | return -6; | |
144b44b1 LB |
2078 | } |
2079 | ||
1e9042b9 | 2080 | if ((DO_BIC(BIC_CPU_c6) || soft_c1_residency_display(BIC_CPU_c6)) && !do_knl_cstates) { |
c98d5d94 LB |
2081 | if (get_msr(cpu, MSR_CORE_C6_RESIDENCY, &c->c6)) |
2082 | return -7; | |
1e9042b9 | 2083 | } else if (do_knl_cstates || soft_c1_residency_display(BIC_CPU_c6)) { |
fb5d4327 DC |
2084 | if (get_msr(cpu, MSR_KNL_CORE_C6_RESIDENCY, &c->c6)) |
2085 | return -7; | |
c98d5d94 LB |
2086 | } |
2087 | ||
1e9042b9 | 2088 | if (DO_BIC(BIC_CPU_c7) || soft_c1_residency_display(BIC_CPU_c7)) |
c98d5d94 LB |
2089 | if (get_msr(cpu, MSR_CORE_C7_RESIDENCY, &c->c7)) |
2090 | return -8; | |
2091 | ||
0539ba11 LB |
2092 | if (DO_BIC(BIC_Mod_c6)) |
2093 | if (get_msr(cpu, MSR_MODULE_C6_RES_MS, &c->mc6_us)) | |
2094 | return -8; | |
2095 | ||
812db3f7 | 2096 | if (DO_BIC(BIC_CoreTmp)) { |
889facbe LB |
2097 | if (get_msr(cpu, MSR_IA32_THERM_STATUS, &msr)) |
2098 | return -9; | |
2099 | c->core_temp_c = tcc_activation_temp - ((msr >> 16) & 0x7F); | |
2100 | } | |
2101 | ||
9392bd98 CW |
2102 | if (do_rapl & RAPL_AMD_F17H) { |
2103 | if (get_msr(cpu, MSR_CORE_ENERGY_STAT, &msr)) | |
2104 | return -14; | |
2105 | c->core_energy = msr & 0xFFFFFFFF; | |
2106 | } | |
2107 | ||
388e9c81 | 2108 | for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) { |
495c7654 | 2109 | if (get_mp(cpu, mp, &c->counter[i])) |
388e9c81 LB |
2110 | return -10; |
2111 | } | |
889facbe | 2112 | |
c98d5d94 LB |
2113 | /* collect package counters only for 1st core in package */ |
2114 | if (!(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE)) | |
f4fdf2b4 | 2115 | goto done; |
c98d5d94 | 2116 | |
a99d8730 | 2117 | if (DO_BIC(BIC_Totl_c0)) { |
0b2bb692 LB |
2118 | if (get_msr(cpu, MSR_PKG_WEIGHTED_CORE_C0_RES, &p->pkg_wtd_core_c0)) |
2119 | return -10; | |
a99d8730 LB |
2120 | } |
2121 | if (DO_BIC(BIC_Any_c0)) { | |
0b2bb692 LB |
2122 | if (get_msr(cpu, MSR_PKG_ANY_CORE_C0_RES, &p->pkg_any_core_c0)) |
2123 | return -11; | |
a99d8730 LB |
2124 | } |
2125 | if (DO_BIC(BIC_GFX_c0)) { | |
0b2bb692 LB |
2126 | if (get_msr(cpu, MSR_PKG_ANY_GFXE_C0_RES, &p->pkg_any_gfxe_c0)) |
2127 | return -12; | |
a99d8730 LB |
2128 | } |
2129 | if (DO_BIC(BIC_CPUGFX)) { | |
0b2bb692 LB |
2130 | if (get_msr(cpu, MSR_PKG_BOTH_CORE_GFXE_C0_RES, &p->pkg_both_core_gfxe_c0)) |
2131 | return -13; | |
2132 | } | |
0f47c08d | 2133 | if (DO_BIC(BIC_Pkgpc3)) |
c98d5d94 LB |
2134 | if (get_msr(cpu, MSR_PKG_C3_RESIDENCY, &p->pc3)) |
2135 | return -9; | |
0f47c08d | 2136 | if (DO_BIC(BIC_Pkgpc6)) { |
0539ba11 LB |
2137 | if (do_slm_cstates) { |
2138 | if (get_msr(cpu, MSR_ATOM_PKG_C6_RESIDENCY, &p->pc6)) | |
2139 | return -10; | |
2140 | } else { | |
2141 | if (get_msr(cpu, MSR_PKG_C6_RESIDENCY, &p->pc6)) | |
2142 | return -10; | |
2143 | } | |
2144 | } | |
2145 | ||
0f47c08d | 2146 | if (DO_BIC(BIC_Pkgpc2)) |
c98d5d94 LB |
2147 | if (get_msr(cpu, MSR_PKG_C2_RESIDENCY, &p->pc2)) |
2148 | return -11; | |
0f47c08d | 2149 | if (DO_BIC(BIC_Pkgpc7)) |
c98d5d94 LB |
2150 | if (get_msr(cpu, MSR_PKG_C7_RESIDENCY, &p->pc7)) |
2151 | return -12; | |
0f47c08d | 2152 | if (DO_BIC(BIC_Pkgpc8)) |
ca58710f KCA |
2153 | if (get_msr(cpu, MSR_PKG_C8_RESIDENCY, &p->pc8)) |
2154 | return -13; | |
0f47c08d | 2155 | if (DO_BIC(BIC_Pkgpc9)) |
ca58710f KCA |
2156 | if (get_msr(cpu, MSR_PKG_C9_RESIDENCY, &p->pc9)) |
2157 | return -13; | |
0f47c08d | 2158 | if (DO_BIC(BIC_Pkgpc10)) |
ca58710f KCA |
2159 | if (get_msr(cpu, MSR_PKG_C10_RESIDENCY, &p->pc10)) |
2160 | return -13; | |
0f47c08d | 2161 | |
be0e54c4 LB |
2162 | if (DO_BIC(BIC_CPU_LPI)) |
2163 | p->cpu_lpi = cpuidle_cur_cpu_lpi_us; | |
2164 | if (DO_BIC(BIC_SYS_LPI)) | |
2165 | p->sys_lpi = cpuidle_cur_sys_lpi_us; | |
2166 | ||
889facbe | 2167 | if (do_rapl & RAPL_PKG) { |
9972d5d8 | 2168 | if (get_msr_sum(cpu, MSR_PKG_ENERGY_STATUS, &msr)) |
889facbe | 2169 | return -13; |
9972d5d8 | 2170 | p->energy_pkg = msr; |
889facbe | 2171 | } |
9148494c | 2172 | if (do_rapl & RAPL_CORES_ENERGY_STATUS) { |
9972d5d8 | 2173 | if (get_msr_sum(cpu, MSR_PP0_ENERGY_STATUS, &msr)) |
889facbe | 2174 | return -14; |
9972d5d8 | 2175 | p->energy_cores = msr; |
889facbe LB |
2176 | } |
2177 | if (do_rapl & RAPL_DRAM) { | |
9972d5d8 | 2178 | if (get_msr_sum(cpu, MSR_DRAM_ENERGY_STATUS, &msr)) |
889facbe | 2179 | return -15; |
9972d5d8 | 2180 | p->energy_dram = msr; |
889facbe LB |
2181 | } |
2182 | if (do_rapl & RAPL_GFX) { | |
9972d5d8 | 2183 | if (get_msr_sum(cpu, MSR_PP1_ENERGY_STATUS, &msr)) |
889facbe | 2184 | return -16; |
9972d5d8 | 2185 | p->energy_gfx = msr; |
889facbe LB |
2186 | } |
2187 | if (do_rapl & RAPL_PKG_PERF_STATUS) { | |
9972d5d8 | 2188 | if (get_msr_sum(cpu, MSR_PKG_PERF_STATUS, &msr)) |
889facbe | 2189 | return -16; |
9972d5d8 | 2190 | p->rapl_pkg_perf_status = msr; |
889facbe LB |
2191 | } |
2192 | if (do_rapl & RAPL_DRAM_PERF_STATUS) { | |
9972d5d8 | 2193 | if (get_msr_sum(cpu, MSR_DRAM_PERF_STATUS, &msr)) |
889facbe | 2194 | return -16; |
9972d5d8 | 2195 | p->rapl_dram_perf_status = msr; |
889facbe | 2196 | } |
3316f99a | 2197 | if (do_rapl & RAPL_AMD_F17H) { |
9972d5d8 | 2198 | if (get_msr_sum(cpu, MSR_PKG_ENERGY_STAT, &msr)) |
3316f99a | 2199 | return -13; |
9972d5d8 | 2200 | p->energy_pkg = msr; |
3316f99a | 2201 | } |
812db3f7 | 2202 | if (DO_BIC(BIC_PkgTmp)) { |
889facbe LB |
2203 | if (get_msr(cpu, MSR_IA32_PACKAGE_THERM_STATUS, &msr)) |
2204 | return -17; | |
2205 | p->pkg_temp_c = tcc_activation_temp - ((msr >> 16) & 0x7F); | |
2206 | } | |
fdf676e5 | 2207 | |
812db3f7 | 2208 | if (DO_BIC(BIC_GFX_rc6)) |
fdf676e5 LB |
2209 | p->gfx_rc6_ms = gfx_cur_rc6_ms; |
2210 | ||
812db3f7 | 2211 | if (DO_BIC(BIC_GFXMHz)) |
27d47356 LB |
2212 | p->gfx_mhz = gfx_cur_mhz; |
2213 | ||
b4b91569 RA |
2214 | if (DO_BIC(BIC_GFXACTMHz)) |
2215 | p->gfx_act_mhz = gfx_act_mhz; | |
2216 | ||
388e9c81 | 2217 | for (i = 0, mp = sys.pp; mp; i++, mp = mp->next) { |
495c7654 | 2218 | if (get_mp(cpu, mp, &p->counter[i])) |
388e9c81 LB |
2219 | return -10; |
2220 | } | |
f4fdf2b4 LB |
2221 | done: |
2222 | gettimeofday(&t->tv_end, (struct timezone *)NULL); | |
388e9c81 | 2223 | |
15aaa346 | 2224 | return 0; |
103a8fea LB |
2225 | } |
2226 | ||
ee7e38e3 LB |
2227 | /* |
2228 | * MSR_PKG_CST_CONFIG_CONTROL decoding for pkg_cstate_limit: | |
2229 | * If you change the values, note they are used both in comparisons | |
2230 | * (>= PCL__7) and to index pkg_cstate_limit_strings[]. | |
2231 | */ | |
2232 | ||
2233 | #define PCLUKN 0 /* Unknown */ | |
2234 | #define PCLRSV 1 /* Reserved */ | |
2235 | #define PCL__0 2 /* PC0 */ | |
2236 | #define PCL__1 3 /* PC1 */ | |
2237 | #define PCL__2 4 /* PC2 */ | |
2238 | #define PCL__3 5 /* PC3 */ | |
2239 | #define PCL__4 6 /* PC4 */ | |
2240 | #define PCL__6 7 /* PC6 */ | |
2241 | #define PCL_6N 8 /* PC6 No Retention */ | |
2242 | #define PCL_6R 9 /* PC6 Retention */ | |
2243 | #define PCL__7 10 /* PC7 */ | |
2244 | #define PCL_7S 11 /* PC7 Shrink */ | |
0b2bb692 LB |
2245 | #define PCL__8 12 /* PC8 */ |
2246 | #define PCL__9 13 /* PC9 */ | |
445640a5 LB |
2247 | #define PCL_10 14 /* PC10 */ |
2248 | #define PCLUNL 15 /* Unlimited */ | |
ee7e38e3 LB |
2249 | |
2250 | int pkg_cstate_limit = PCLUKN; | |
2251 | char *pkg_cstate_limit_strings[] = { "reserved", "unknown", "pc0", "pc1", "pc2", | |
445640a5 | 2252 | "pc3", "pc4", "pc6", "pc6n", "pc6r", "pc7", "pc7s", "pc8", "pc9", "pc10", "unlimited"}; |
ee7e38e3 | 2253 | |
e9257f5f LB |
2254 | int nhm_pkg_cstate_limits[16] = {PCL__0, PCL__1, PCL__3, PCL__6, PCL__7, PCLRSV, PCLRSV, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV}; |
2255 | int snb_pkg_cstate_limits[16] = {PCL__0, PCL__2, PCL_6N, PCL_6R, PCL__7, PCL_7S, PCLRSV, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV}; | |
2256 | int hsw_pkg_cstate_limits[16] = {PCL__0, PCL__2, PCL__3, PCL__6, PCL__7, PCL_7S, PCL__8, PCL__9, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV}; | |
0539ba11 | 2257 | int slv_pkg_cstate_limits[16] = {PCL__0, PCL__1, PCLRSV, PCLRSV, PCL__4, PCLRSV, PCL__6, PCL__7, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCL__6, PCL__7}; |
f2642888 | 2258 | 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}; |
e9257f5f | 2259 | 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}; |
445640a5 | 2260 | int glm_pkg_cstate_limits[16] = {PCLUNL, PCL__1, PCL__3, PCL__6, PCL__7, PCL_7S, PCL__8, PCL__9, PCL_10, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV}; |
2085e124 | 2261 | 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}; |
ee7e38e3 | 2262 | |
a2b7b749 LB |
2263 | |
2264 | static void | |
2265 | calculate_tsc_tweak() | |
2266 | { | |
a2b7b749 LB |
2267 | tsc_tweak = base_hz / tsc_hz; |
2268 | } | |
2269 | ||
fcd17211 LB |
2270 | static void |
2271 | dump_nhm_platform_info(void) | |
103a8fea LB |
2272 | { |
2273 | unsigned long long msr; | |
2274 | unsigned int ratio; | |
2275 | ||
ec0adc53 | 2276 | get_msr(base_cpu, MSR_PLATFORM_INFO, &msr); |
103a8fea | 2277 | |
b7d8c148 | 2278 | fprintf(outf, "cpu%d: MSR_PLATFORM_INFO: 0x%08llx\n", base_cpu, msr); |
6574a5d5 | 2279 | |
103a8fea | 2280 | ratio = (msr >> 40) & 0xFF; |
710f273b | 2281 | fprintf(outf, "%d * %.1f = %.1f MHz max efficiency frequency\n", |
103a8fea LB |
2282 | ratio, bclk, ratio * bclk); |
2283 | ||
2284 | ratio = (msr >> 8) & 0xFF; | |
710f273b | 2285 | fprintf(outf, "%d * %.1f = %.1f MHz base frequency\n", |
103a8fea LB |
2286 | ratio, bclk, ratio * bclk); |
2287 | ||
7ce7d5de | 2288 | get_msr(base_cpu, MSR_IA32_POWER_CTL, &msr); |
b7d8c148 | 2289 | fprintf(outf, "cpu%d: MSR_IA32_POWER_CTL: 0x%08llx (C1E auto-promotion: %sabled)\n", |
bfae2052 | 2290 | base_cpu, msr, msr & 0x2 ? "EN" : "DIS"); |
67920418 | 2291 | |
fcd17211 LB |
2292 | return; |
2293 | } | |
2294 | ||
2295 | static void | |
2296 | dump_hsw_turbo_ratio_limits(void) | |
2297 | { | |
2298 | unsigned long long msr; | |
2299 | unsigned int ratio; | |
2300 | ||
7ce7d5de | 2301 | get_msr(base_cpu, MSR_TURBO_RATIO_LIMIT2, &msr); |
fcd17211 | 2302 | |
b7d8c148 | 2303 | fprintf(outf, "cpu%d: MSR_TURBO_RATIO_LIMIT2: 0x%08llx\n", base_cpu, msr); |
fcd17211 LB |
2304 | |
2305 | ratio = (msr >> 8) & 0xFF; | |
2306 | if (ratio) | |
710f273b | 2307 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 18 active cores\n", |
fcd17211 LB |
2308 | ratio, bclk, ratio * bclk); |
2309 | ||
2310 | ratio = (msr >> 0) & 0xFF; | |
2311 | if (ratio) | |
710f273b | 2312 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 17 active cores\n", |
fcd17211 LB |
2313 | ratio, bclk, ratio * bclk); |
2314 | return; | |
2315 | } | |
2316 | ||
2317 | static void | |
2318 | dump_ivt_turbo_ratio_limits(void) | |
2319 | { | |
2320 | unsigned long long msr; | |
2321 | unsigned int ratio; | |
6574a5d5 | 2322 | |
7ce7d5de | 2323 | get_msr(base_cpu, MSR_TURBO_RATIO_LIMIT1, &msr); |
6574a5d5 | 2324 | |
b7d8c148 | 2325 | fprintf(outf, "cpu%d: MSR_TURBO_RATIO_LIMIT1: 0x%08llx\n", base_cpu, msr); |
6574a5d5 LB |
2326 | |
2327 | ratio = (msr >> 56) & 0xFF; | |
2328 | if (ratio) | |
710f273b | 2329 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 16 active cores\n", |
6574a5d5 LB |
2330 | ratio, bclk, ratio * bclk); |
2331 | ||
2332 | ratio = (msr >> 48) & 0xFF; | |
2333 | if (ratio) | |
710f273b | 2334 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 15 active cores\n", |
6574a5d5 LB |
2335 | ratio, bclk, ratio * bclk); |
2336 | ||
2337 | ratio = (msr >> 40) & 0xFF; | |
2338 | if (ratio) | |
710f273b | 2339 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 14 active cores\n", |
6574a5d5 LB |
2340 | ratio, bclk, ratio * bclk); |
2341 | ||
2342 | ratio = (msr >> 32) & 0xFF; | |
2343 | if (ratio) | |
710f273b | 2344 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 13 active cores\n", |
6574a5d5 LB |
2345 | ratio, bclk, ratio * bclk); |
2346 | ||
2347 | ratio = (msr >> 24) & 0xFF; | |
2348 | if (ratio) | |
710f273b | 2349 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 12 active cores\n", |
6574a5d5 LB |
2350 | ratio, bclk, ratio * bclk); |
2351 | ||
2352 | ratio = (msr >> 16) & 0xFF; | |
2353 | if (ratio) | |
710f273b | 2354 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 11 active cores\n", |
6574a5d5 LB |
2355 | ratio, bclk, ratio * bclk); |
2356 | ||
2357 | ratio = (msr >> 8) & 0xFF; | |
2358 | if (ratio) | |
710f273b | 2359 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 10 active cores\n", |
6574a5d5 LB |
2360 | ratio, bclk, ratio * bclk); |
2361 | ||
2362 | ratio = (msr >> 0) & 0xFF; | |
2363 | if (ratio) | |
710f273b | 2364 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 9 active cores\n", |
6574a5d5 | 2365 | ratio, bclk, ratio * bclk); |
fcd17211 LB |
2366 | return; |
2367 | } | |
31e07522 LB |
2368 | int has_turbo_ratio_group_limits(int family, int model) |
2369 | { | |
2370 | ||
2371 | if (!genuine_intel) | |
2372 | return 0; | |
2373 | ||
2374 | switch (model) { | |
2375 | case INTEL_FAM6_ATOM_GOLDMONT: | |
2376 | case INTEL_FAM6_SKYLAKE_X: | |
5ebb34ed | 2377 | case INTEL_FAM6_ATOM_GOLDMONT_D: |
20de0dab | 2378 | case INTEL_FAM6_ATOM_TREMONT_D: |
31e07522 LB |
2379 | return 1; |
2380 | } | |
2381 | return 0; | |
2382 | } | |
6574a5d5 | 2383 | |
fcd17211 | 2384 | static void |
31e07522 | 2385 | dump_turbo_ratio_limits(int family, int model) |
fcd17211 | 2386 | { |
31e07522 LB |
2387 | unsigned long long msr, core_counts; |
2388 | unsigned int ratio, group_size; | |
103a8fea | 2389 | |
7ce7d5de | 2390 | get_msr(base_cpu, MSR_TURBO_RATIO_LIMIT, &msr); |
b7d8c148 | 2391 | fprintf(outf, "cpu%d: MSR_TURBO_RATIO_LIMIT: 0x%08llx\n", base_cpu, msr); |
6574a5d5 | 2392 | |
31e07522 LB |
2393 | if (has_turbo_ratio_group_limits(family, model)) { |
2394 | get_msr(base_cpu, MSR_TURBO_RATIO_LIMIT1, &core_counts); | |
2395 | fprintf(outf, "cpu%d: MSR_TURBO_RATIO_LIMIT1: 0x%08llx\n", base_cpu, core_counts); | |
2396 | } else { | |
2397 | core_counts = 0x0807060504030201; | |
2398 | } | |
2399 | ||
6574a5d5 | 2400 | ratio = (msr >> 56) & 0xFF; |
31e07522 | 2401 | group_size = (core_counts >> 56) & 0xFF; |
6574a5d5 | 2402 | if (ratio) |
31e07522 LB |
2403 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo %d active cores\n", |
2404 | ratio, bclk, ratio * bclk, group_size); | |
6574a5d5 LB |
2405 | |
2406 | ratio = (msr >> 48) & 0xFF; | |
31e07522 | 2407 | group_size = (core_counts >> 48) & 0xFF; |
6574a5d5 | 2408 | if (ratio) |
31e07522 LB |
2409 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo %d active cores\n", |
2410 | ratio, bclk, ratio * bclk, group_size); | |
6574a5d5 LB |
2411 | |
2412 | ratio = (msr >> 40) & 0xFF; | |
31e07522 | 2413 | group_size = (core_counts >> 40) & 0xFF; |
6574a5d5 | 2414 | if (ratio) |
31e07522 LB |
2415 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo %d active cores\n", |
2416 | ratio, bclk, ratio * bclk, group_size); | |
6574a5d5 LB |
2417 | |
2418 | ratio = (msr >> 32) & 0xFF; | |
31e07522 | 2419 | group_size = (core_counts >> 32) & 0xFF; |
6574a5d5 | 2420 | if (ratio) |
31e07522 LB |
2421 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo %d active cores\n", |
2422 | ratio, bclk, ratio * bclk, group_size); | |
6574a5d5 | 2423 | |
103a8fea | 2424 | ratio = (msr >> 24) & 0xFF; |
31e07522 | 2425 | group_size = (core_counts >> 24) & 0xFF; |
103a8fea | 2426 | if (ratio) |
31e07522 LB |
2427 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo %d active cores\n", |
2428 | ratio, bclk, ratio * bclk, group_size); | |
103a8fea LB |
2429 | |
2430 | ratio = (msr >> 16) & 0xFF; | |
31e07522 | 2431 | group_size = (core_counts >> 16) & 0xFF; |
103a8fea | 2432 | if (ratio) |
31e07522 LB |
2433 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo %d active cores\n", |
2434 | ratio, bclk, ratio * bclk, group_size); | |
103a8fea LB |
2435 | |
2436 | ratio = (msr >> 8) & 0xFF; | |
31e07522 | 2437 | group_size = (core_counts >> 8) & 0xFF; |
103a8fea | 2438 | if (ratio) |
31e07522 LB |
2439 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo %d active cores\n", |
2440 | ratio, bclk, ratio * bclk, group_size); | |
103a8fea LB |
2441 | |
2442 | ratio = (msr >> 0) & 0xFF; | |
31e07522 | 2443 | group_size = (core_counts >> 0) & 0xFF; |
103a8fea | 2444 | if (ratio) |
31e07522 LB |
2445 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo %d active cores\n", |
2446 | ratio, bclk, ratio * bclk, group_size); | |
fcd17211 LB |
2447 | return; |
2448 | } | |
3a9a941d | 2449 | |
0f7887c4 LB |
2450 | static void |
2451 | dump_atom_turbo_ratio_limits(void) | |
2452 | { | |
2453 | unsigned long long msr; | |
2454 | unsigned int ratio; | |
2455 | ||
2456 | get_msr(base_cpu, MSR_ATOM_CORE_RATIOS, &msr); | |
2457 | fprintf(outf, "cpu%d: MSR_ATOM_CORE_RATIOS: 0x%08llx\n", base_cpu, msr & 0xFFFFFFFF); | |
2458 | ||
2459 | ratio = (msr >> 0) & 0x3F; | |
2460 | if (ratio) | |
2461 | fprintf(outf, "%d * %.1f = %.1f MHz minimum operating frequency\n", | |
2462 | ratio, bclk, ratio * bclk); | |
2463 | ||
2464 | ratio = (msr >> 8) & 0x3F; | |
2465 | if (ratio) | |
2466 | fprintf(outf, "%d * %.1f = %.1f MHz low frequency mode (LFM)\n", | |
2467 | ratio, bclk, ratio * bclk); | |
2468 | ||
2469 | ratio = (msr >> 16) & 0x3F; | |
2470 | if (ratio) | |
2471 | fprintf(outf, "%d * %.1f = %.1f MHz base frequency\n", | |
2472 | ratio, bclk, ratio * bclk); | |
2473 | ||
2474 | get_msr(base_cpu, MSR_ATOM_CORE_TURBO_RATIOS, &msr); | |
2475 | fprintf(outf, "cpu%d: MSR_ATOM_CORE_TURBO_RATIOS: 0x%08llx\n", base_cpu, msr & 0xFFFFFFFF); | |
2476 | ||
2477 | ratio = (msr >> 24) & 0x3F; | |
2478 | if (ratio) | |
2479 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 4 active cores\n", | |
2480 | ratio, bclk, ratio * bclk); | |
2481 | ||
2482 | ratio = (msr >> 16) & 0x3F; | |
2483 | if (ratio) | |
2484 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 3 active cores\n", | |
2485 | ratio, bclk, ratio * bclk); | |
2486 | ||
2487 | ratio = (msr >> 8) & 0x3F; | |
2488 | if (ratio) | |
2489 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 2 active cores\n", | |
2490 | ratio, bclk, ratio * bclk); | |
2491 | ||
2492 | ratio = (msr >> 0) & 0x3F; | |
2493 | if (ratio) | |
2494 | fprintf(outf, "%d * %.1f = %.1f MHz max turbo 1 active core\n", | |
2495 | ratio, bclk, ratio * bclk); | |
2496 | } | |
2497 | ||
fb5d4327 DC |
2498 | static void |
2499 | dump_knl_turbo_ratio_limits(void) | |
2500 | { | |
cbf97aba HC |
2501 | const unsigned int buckets_no = 7; |
2502 | ||
fb5d4327 | 2503 | unsigned long long msr; |
cbf97aba HC |
2504 | int delta_cores, delta_ratio; |
2505 | int i, b_nr; | |
2506 | unsigned int cores[buckets_no]; | |
2507 | unsigned int ratio[buckets_no]; | |
fb5d4327 | 2508 | |
ebf5926a | 2509 | get_msr(base_cpu, MSR_TURBO_RATIO_LIMIT, &msr); |
fb5d4327 | 2510 | |
b7d8c148 | 2511 | fprintf(outf, "cpu%d: MSR_TURBO_RATIO_LIMIT: 0x%08llx\n", |
bfae2052 | 2512 | base_cpu, msr); |
fb5d4327 DC |
2513 | |
2514 | /** | |
2515 | * Turbo encoding in KNL is as follows: | |
cbf97aba HC |
2516 | * [0] -- Reserved |
2517 | * [7:1] -- Base value of number of active cores of bucket 1. | |
fb5d4327 DC |
2518 | * [15:8] -- Base value of freq ratio of bucket 1. |
2519 | * [20:16] -- +ve delta of number of active cores of bucket 2. | |
2520 | * i.e. active cores of bucket 2 = | |
2521 | * active cores of bucket 1 + delta | |
2522 | * [23:21] -- Negative delta of freq ratio of bucket 2. | |
2523 | * i.e. freq ratio of bucket 2 = | |
2524 | * freq ratio of bucket 1 - delta | |
2525 | * [28:24]-- +ve delta of number of active cores of bucket 3. | |
2526 | * [31:29]-- -ve delta of freq ratio of bucket 3. | |
2527 | * [36:32]-- +ve delta of number of active cores of bucket 4. | |
2528 | * [39:37]-- -ve delta of freq ratio of bucket 4. | |
2529 | * [44:40]-- +ve delta of number of active cores of bucket 5. | |
2530 | * [47:45]-- -ve delta of freq ratio of bucket 5. | |
2531 | * [52:48]-- +ve delta of number of active cores of bucket 6. | |
2532 | * [55:53]-- -ve delta of freq ratio of bucket 6. | |
2533 | * [60:56]-- +ve delta of number of active cores of bucket 7. | |
2534 | * [63:61]-- -ve delta of freq ratio of bucket 7. | |
2535 | */ | |
cbf97aba HC |
2536 | |
2537 | b_nr = 0; | |
2538 | cores[b_nr] = (msr & 0xFF) >> 1; | |
2539 | ratio[b_nr] = (msr >> 8) & 0xFF; | |
2540 | ||
2541 | for (i = 16; i < 64; i += 8) { | |
fb5d4327 | 2542 | delta_cores = (msr >> i) & 0x1F; |
cbf97aba HC |
2543 | delta_ratio = (msr >> (i + 5)) & 0x7; |
2544 | ||
2545 | cores[b_nr + 1] = cores[b_nr] + delta_cores; | |
2546 | ratio[b_nr + 1] = ratio[b_nr] - delta_ratio; | |
2547 | b_nr++; | |
fb5d4327 | 2548 | } |
cbf97aba HC |
2549 | |
2550 | for (i = buckets_no - 1; i >= 0; i--) | |
2551 | if (i > 0 ? ratio[i] != ratio[i - 1] : 1) | |
b7d8c148 | 2552 | fprintf(outf, |
710f273b | 2553 | "%d * %.1f = %.1f MHz max turbo %d active cores\n", |
cbf97aba | 2554 | ratio[i], bclk, ratio[i] * bclk, cores[i]); |
fb5d4327 DC |
2555 | } |
2556 | ||
fcd17211 LB |
2557 | static void |
2558 | dump_nhm_cst_cfg(void) | |
2559 | { | |
2560 | unsigned long long msr; | |
2561 | ||
1df2e55a | 2562 | get_msr(base_cpu, MSR_PKG_CST_CONFIG_CONTROL, &msr); |
fcd17211 | 2563 | |
1df2e55a | 2564 | fprintf(outf, "cpu%d: MSR_PKG_CST_CONFIG_CONTROL: 0x%08llx", base_cpu, msr); |
fcd17211 | 2565 | |
3e8b62bf | 2566 | fprintf(outf, " (%s%s%s%s%slocked, pkg-cstate-limit=%d (%s)", |
fcd17211 LB |
2567 | (msr & SNB_C3_AUTO_UNDEMOTE) ? "UNdemote-C3, " : "", |
2568 | (msr & SNB_C1_AUTO_UNDEMOTE) ? "UNdemote-C1, " : "", | |
2569 | (msr & NHM_C3_AUTO_DEMOTE) ? "demote-C3, " : "", | |
2570 | (msr & NHM_C1_AUTO_DEMOTE) ? "demote-C1, " : "", | |
2571 | (msr & (1 << 15)) ? "" : "UN", | |
6c34f160 | 2572 | (unsigned int)msr & 0xF, |
fcd17211 | 2573 | pkg_cstate_limit_strings[pkg_cstate_limit]); |
ac980e13 AB |
2574 | |
2575 | #define AUTOMATIC_CSTATE_CONVERSION (1UL << 16) | |
2576 | if (has_automatic_cstate_conversion) { | |
2577 | fprintf(outf, ", automatic c-state conversion=%s", | |
2578 | (msr & AUTOMATIC_CSTATE_CONVERSION) ? "on" : "off"); | |
2579 | } | |
2580 | ||
2581 | fprintf(outf, ")\n"); | |
2582 | ||
fcd17211 | 2583 | return; |
103a8fea LB |
2584 | } |
2585 | ||
6fb3143b LB |
2586 | static void |
2587 | dump_config_tdp(void) | |
2588 | { | |
2589 | unsigned long long msr; | |
2590 | ||
2591 | get_msr(base_cpu, MSR_CONFIG_TDP_NOMINAL, &msr); | |
b7d8c148 | 2592 | fprintf(outf, "cpu%d: MSR_CONFIG_TDP_NOMINAL: 0x%08llx", base_cpu, msr); |
685b535b | 2593 | fprintf(outf, " (base_ratio=%d)\n", (unsigned int)msr & 0xFF); |
6fb3143b LB |
2594 | |
2595 | get_msr(base_cpu, MSR_CONFIG_TDP_LEVEL_1, &msr); | |
b7d8c148 | 2596 | fprintf(outf, "cpu%d: MSR_CONFIG_TDP_LEVEL_1: 0x%08llx (", base_cpu, msr); |
6fb3143b | 2597 | if (msr) { |
685b535b CY |
2598 | fprintf(outf, "PKG_MIN_PWR_LVL1=%d ", (unsigned int)(msr >> 48) & 0x7FFF); |
2599 | fprintf(outf, "PKG_MAX_PWR_LVL1=%d ", (unsigned int)(msr >> 32) & 0x7FFF); | |
2600 | fprintf(outf, "LVL1_RATIO=%d ", (unsigned int)(msr >> 16) & 0xFF); | |
2601 | fprintf(outf, "PKG_TDP_LVL1=%d", (unsigned int)(msr) & 0x7FFF); | |
6fb3143b | 2602 | } |
b7d8c148 | 2603 | fprintf(outf, ")\n"); |
6fb3143b LB |
2604 | |
2605 | get_msr(base_cpu, MSR_CONFIG_TDP_LEVEL_2, &msr); | |
b7d8c148 | 2606 | fprintf(outf, "cpu%d: MSR_CONFIG_TDP_LEVEL_2: 0x%08llx (", base_cpu, msr); |
6fb3143b | 2607 | if (msr) { |
685b535b CY |
2608 | fprintf(outf, "PKG_MIN_PWR_LVL2=%d ", (unsigned int)(msr >> 48) & 0x7FFF); |
2609 | fprintf(outf, "PKG_MAX_PWR_LVL2=%d ", (unsigned int)(msr >> 32) & 0x7FFF); | |
2610 | fprintf(outf, "LVL2_RATIO=%d ", (unsigned int)(msr >> 16) & 0xFF); | |
2611 | fprintf(outf, "PKG_TDP_LVL2=%d", (unsigned int)(msr) & 0x7FFF); | |
6fb3143b | 2612 | } |
b7d8c148 | 2613 | fprintf(outf, ")\n"); |
6fb3143b LB |
2614 | |
2615 | get_msr(base_cpu, MSR_CONFIG_TDP_CONTROL, &msr); | |
b7d8c148 | 2616 | fprintf(outf, "cpu%d: MSR_CONFIG_TDP_CONTROL: 0x%08llx (", base_cpu, msr); |
6fb3143b | 2617 | if ((msr) & 0x3) |
b7d8c148 LB |
2618 | fprintf(outf, "TDP_LEVEL=%d ", (unsigned int)(msr) & 0x3); |
2619 | fprintf(outf, " lock=%d", (unsigned int)(msr >> 31) & 1); | |
2620 | fprintf(outf, ")\n"); | |
36229897 | 2621 | |
6fb3143b | 2622 | get_msr(base_cpu, MSR_TURBO_ACTIVATION_RATIO, &msr); |
b7d8c148 | 2623 | fprintf(outf, "cpu%d: MSR_TURBO_ACTIVATION_RATIO: 0x%08llx (", base_cpu, msr); |
685b535b | 2624 | fprintf(outf, "MAX_NON_TURBO_RATIO=%d", (unsigned int)(msr) & 0xFF); |
b7d8c148 LB |
2625 | fprintf(outf, " lock=%d", (unsigned int)(msr >> 31) & 1); |
2626 | fprintf(outf, ")\n"); | |
6fb3143b | 2627 | } |
5a63426e LB |
2628 | |
2629 | unsigned int irtl_time_units[] = {1, 32, 1024, 32768, 1048576, 33554432, 0, 0 }; | |
2630 | ||
2631 | void print_irtl(void) | |
2632 | { | |
2633 | unsigned long long msr; | |
2634 | ||
2635 | get_msr(base_cpu, MSR_PKGC3_IRTL, &msr); | |
2636 | fprintf(outf, "cpu%d: MSR_PKGC3_IRTL: 0x%08llx (", base_cpu, msr); | |
2637 | fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT", | |
2638 | (msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]); | |
2639 | ||
2640 | get_msr(base_cpu, MSR_PKGC6_IRTL, &msr); | |
2641 | fprintf(outf, "cpu%d: MSR_PKGC6_IRTL: 0x%08llx (", base_cpu, msr); | |
2642 | fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT", | |
2643 | (msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]); | |
2644 | ||
2645 | get_msr(base_cpu, MSR_PKGC7_IRTL, &msr); | |
2646 | fprintf(outf, "cpu%d: MSR_PKGC7_IRTL: 0x%08llx (", base_cpu, msr); | |
2647 | fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT", | |
2648 | (msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]); | |
2649 | ||
2650 | if (!do_irtl_hsw) | |
2651 | return; | |
2652 | ||
2653 | get_msr(base_cpu, MSR_PKGC8_IRTL, &msr); | |
2654 | fprintf(outf, "cpu%d: MSR_PKGC8_IRTL: 0x%08llx (", base_cpu, msr); | |
2655 | fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT", | |
2656 | (msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]); | |
2657 | ||
2658 | get_msr(base_cpu, MSR_PKGC9_IRTL, &msr); | |
2659 | fprintf(outf, "cpu%d: MSR_PKGC9_IRTL: 0x%08llx (", base_cpu, msr); | |
2660 | fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT", | |
2661 | (msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]); | |
2662 | ||
2663 | get_msr(base_cpu, MSR_PKGC10_IRTL, &msr); | |
2664 | fprintf(outf, "cpu%d: MSR_PKGC10_IRTL: 0x%08llx (", base_cpu, msr); | |
2665 | fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT", | |
2666 | (msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]); | |
2667 | ||
2668 | } | |
36229897 LB |
2669 | void free_fd_percpu(void) |
2670 | { | |
2671 | int i; | |
2672 | ||
01a67adf | 2673 | for (i = 0; i < topo.max_cpu_num + 1; ++i) { |
36229897 LB |
2674 | if (fd_percpu[i] != 0) |
2675 | close(fd_percpu[i]); | |
2676 | } | |
2677 | ||
2678 | free(fd_percpu); | |
6fb3143b LB |
2679 | } |
2680 | ||
c98d5d94 | 2681 | void free_all_buffers(void) |
103a8fea | 2682 | { |
0e2d8f05 LB |
2683 | int i; |
2684 | ||
c98d5d94 LB |
2685 | CPU_FREE(cpu_present_set); |
2686 | cpu_present_set = NULL; | |
36229897 | 2687 | cpu_present_setsize = 0; |
103a8fea | 2688 | |
c98d5d94 LB |
2689 | CPU_FREE(cpu_affinity_set); |
2690 | cpu_affinity_set = NULL; | |
2691 | cpu_affinity_setsize = 0; | |
103a8fea | 2692 | |
c98d5d94 LB |
2693 | free(thread_even); |
2694 | free(core_even); | |
2695 | free(package_even); | |
103a8fea | 2696 | |
c98d5d94 LB |
2697 | thread_even = NULL; |
2698 | core_even = NULL; | |
2699 | package_even = NULL; | |
103a8fea | 2700 | |
c98d5d94 LB |
2701 | free(thread_odd); |
2702 | free(core_odd); | |
2703 | free(package_odd); | |
103a8fea | 2704 | |
c98d5d94 LB |
2705 | thread_odd = NULL; |
2706 | core_odd = NULL; | |
2707 | package_odd = NULL; | |
103a8fea | 2708 | |
c98d5d94 LB |
2709 | free(output_buffer); |
2710 | output_buffer = NULL; | |
2711 | outp = NULL; | |
36229897 LB |
2712 | |
2713 | free_fd_percpu(); | |
562a2d37 LB |
2714 | |
2715 | free(irq_column_2_cpu); | |
2716 | free(irqs_per_cpu); | |
0e2d8f05 LB |
2717 | |
2718 | for (i = 0; i <= topo.max_cpu_num; ++i) { | |
2719 | if (cpus[i].put_ids) | |
2720 | CPU_FREE(cpus[i].put_ids); | |
2721 | } | |
2722 | free(cpus); | |
103a8fea LB |
2723 | } |
2724 | ||
57a42a34 | 2725 | |
c98d5d94 | 2726 | /* |
95aebc44 | 2727 | * Parse a file containing a single int. |
6de68fe1 LB |
2728 | * Return 0 if file can not be opened |
2729 | * Exit if file can be opened, but can not be parsed | |
c98d5d94 | 2730 | */ |
95aebc44 | 2731 | int parse_int_file(const char *fmt, ...) |
103a8fea | 2732 | { |
95aebc44 JT |
2733 | va_list args; |
2734 | char path[PATH_MAX]; | |
c98d5d94 | 2735 | FILE *filep; |
95aebc44 | 2736 | int value; |
103a8fea | 2737 | |
95aebc44 JT |
2738 | va_start(args, fmt); |
2739 | vsnprintf(path, sizeof(path), fmt, args); | |
2740 | va_end(args); | |
6de68fe1 LB |
2741 | filep = fopen(path, "r"); |
2742 | if (!filep) | |
2743 | return 0; | |
b2c95d90 JT |
2744 | if (fscanf(filep, "%d", &value) != 1) |
2745 | err(1, "%s: failed to parse number from file", path); | |
c98d5d94 | 2746 | fclose(filep); |
95aebc44 JT |
2747 | return value; |
2748 | } | |
2749 | ||
c98d5d94 LB |
2750 | /* |
2751 | * cpu_is_first_core_in_package(cpu) | |
2752 | * return 1 if given CPU is 1st core in package | |
2753 | */ | |
2754 | int cpu_is_first_core_in_package(int cpu) | |
103a8fea | 2755 | { |
95aebc44 | 2756 | return cpu == parse_int_file("/sys/devices/system/cpu/cpu%d/topology/core_siblings_list", cpu); |
103a8fea LB |
2757 | } |
2758 | ||
2759 | int get_physical_package_id(int cpu) | |
2760 | { | |
95aebc44 | 2761 | return parse_int_file("/sys/devices/system/cpu/cpu%d/topology/physical_package_id", cpu); |
103a8fea LB |
2762 | } |
2763 | ||
6de68fe1 LB |
2764 | int get_die_id(int cpu) |
2765 | { | |
2766 | return parse_int_file("/sys/devices/system/cpu/cpu%d/topology/die_id", cpu); | |
2767 | } | |
2768 | ||
103a8fea LB |
2769 | int get_core_id(int cpu) |
2770 | { | |
95aebc44 | 2771 | return parse_int_file("/sys/devices/system/cpu/cpu%d/topology/core_id", cpu); |
103a8fea LB |
2772 | } |
2773 | ||
ef605741 PB |
2774 | void set_node_data(void) |
2775 | { | |
2ffbb224 PB |
2776 | int pkg, node, lnode, cpu, cpux; |
2777 | int cpu_count; | |
2778 | ||
2779 | /* initialize logical_node_id */ | |
2780 | for (cpu = 0; cpu <= topo.max_cpu_num; ++cpu) | |
2781 | cpus[cpu].logical_node_id = -1; | |
2782 | ||
2783 | cpu_count = 0; | |
2784 | for (pkg = 0; pkg < topo.num_packages; pkg++) { | |
2785 | lnode = 0; | |
2786 | for (cpu = 0; cpu <= topo.max_cpu_num; ++cpu) { | |
2787 | if (cpus[cpu].physical_package_id != pkg) | |
2788 | continue; | |
2789 | /* find a cpu with an unset logical_node_id */ | |
2790 | if (cpus[cpu].logical_node_id != -1) | |
2791 | continue; | |
2792 | cpus[cpu].logical_node_id = lnode; | |
2793 | node = cpus[cpu].physical_node_id; | |
2794 | cpu_count++; | |
2795 | /* | |
2796 | * find all matching cpus on this pkg and set | |
2797 | * the logical_node_id | |
2798 | */ | |
2799 | for (cpux = cpu; cpux <= topo.max_cpu_num; cpux++) { | |
2800 | if ((cpus[cpux].physical_package_id == pkg) && | |
2801 | (cpus[cpux].physical_node_id == node)) { | |
2802 | cpus[cpux].logical_node_id = lnode; | |
2803 | cpu_count++; | |
2804 | } | |
2805 | } | |
2806 | lnode++; | |
2807 | if (lnode > topo.nodes_per_pkg) | |
2808 | topo.nodes_per_pkg = lnode; | |
2809 | } | |
2810 | if (cpu_count >= topo.max_cpu_num) | |
2811 | break; | |
ef605741 | 2812 | } |
ef605741 PB |
2813 | } |
2814 | ||
2815 | int get_physical_node_id(struct cpu_topology *thiscpu) | |
c98d5d94 LB |
2816 | { |
2817 | char path[80]; | |
2818 | FILE *filep; | |
0e2d8f05 LB |
2819 | int i; |
2820 | int cpu = thiscpu->logical_cpu_id; | |
e275b388 | 2821 | |
0e2d8f05 LB |
2822 | for (i = 0; i <= topo.max_cpu_num; i++) { |
2823 | sprintf(path, "/sys/devices/system/cpu/cpu%d/node%i/cpulist", | |
2824 | cpu, i); | |
2825 | filep = fopen(path, "r"); | |
2826 | if (!filep) | |
2827 | continue; | |
2828 | fclose(filep); | |
2829 | return i; | |
e275b388 | 2830 | } |
0e2d8f05 LB |
2831 | return -1; |
2832 | } | |
c98d5d94 | 2833 | |
0e2d8f05 LB |
2834 | int get_thread_siblings(struct cpu_topology *thiscpu) |
2835 | { | |
2836 | char path[80], character; | |
2837 | FILE *filep; | |
2838 | unsigned long map; | |
8cb48b32 | 2839 | int so, shift, sib_core; |
0e2d8f05 LB |
2840 | int cpu = thiscpu->logical_cpu_id; |
2841 | int offset = topo.max_cpu_num + 1; | |
2842 | size_t size; | |
8cb48b32 | 2843 | int thread_id = 0; |
0e2d8f05 LB |
2844 | |
2845 | thiscpu->put_ids = CPU_ALLOC((topo.max_cpu_num + 1)); | |
8cb48b32 PB |
2846 | if (thiscpu->thread_id < 0) |
2847 | thiscpu->thread_id = thread_id++; | |
0e2d8f05 LB |
2848 | if (!thiscpu->put_ids) |
2849 | return -1; | |
2850 | ||
2851 | size = CPU_ALLOC_SIZE((topo.max_cpu_num + 1)); | |
2852 | CPU_ZERO_S(size, thiscpu->put_ids); | |
2853 | ||
2854 | sprintf(path, | |
2855 | "/sys/devices/system/cpu/cpu%d/topology/thread_siblings", cpu); | |
3d7772ea LB |
2856 | filep = fopen(path, "r"); |
2857 | ||
2858 | if (!filep) { | |
2859 | warnx("%s: open failed", path); | |
2860 | return -1; | |
2861 | } | |
0e2d8f05 LB |
2862 | do { |
2863 | offset -= BITMASK_SIZE; | |
8173c336 BH |
2864 | if (fscanf(filep, "%lx%c", &map, &character) != 2) |
2865 | err(1, "%s: failed to parse file", path); | |
0e2d8f05 LB |
2866 | for (shift = 0; shift < BITMASK_SIZE; shift++) { |
2867 | if ((map >> shift) & 0x1) { | |
8cb48b32 PB |
2868 | so = shift + offset; |
2869 | sib_core = get_core_id(so); | |
2870 | if (sib_core == thiscpu->physical_core_id) { | |
2871 | CPU_SET_S(so, size, thiscpu->put_ids); | |
2872 | if ((so != cpu) && | |
2873 | (cpus[so].thread_id < 0)) | |
2874 | cpus[so].thread_id = | |
2875 | thread_id++; | |
2876 | } | |
0e2d8f05 LB |
2877 | } |
2878 | } | |
2879 | } while (!strncmp(&character, ",", 1)); | |
c98d5d94 | 2880 | fclose(filep); |
0e2d8f05 LB |
2881 | |
2882 | return CPU_COUNT_S(size, thiscpu->put_ids); | |
c98d5d94 LB |
2883 | } |
2884 | ||
103a8fea | 2885 | /* |
c98d5d94 LB |
2886 | * run func(thread, core, package) in topology order |
2887 | * skip non-present cpus | |
103a8fea LB |
2888 | */ |
2889 | ||
c98d5d94 LB |
2890 | int for_all_cpus_2(int (func)(struct thread_data *, struct core_data *, |
2891 | struct pkg_data *, struct thread_data *, struct core_data *, | |
2892 | struct pkg_data *), struct thread_data *thread_base, | |
2893 | struct core_data *core_base, struct pkg_data *pkg_base, | |
2894 | struct thread_data *thread_base2, struct core_data *core_base2, | |
2895 | struct pkg_data *pkg_base2) | |
2896 | { | |
40f5cfe7 | 2897 | int retval, pkg_no, node_no, core_no, thread_no; |
c98d5d94 LB |
2898 | |
2899 | for (pkg_no = 0; pkg_no < topo.num_packages; ++pkg_no) { | |
40f5cfe7 PB |
2900 | for (node_no = 0; node_no < topo.nodes_per_pkg; ++node_no) { |
2901 | for (core_no = 0; core_no < topo.cores_per_node; | |
2902 | ++core_no) { | |
2903 | for (thread_no = 0; thread_no < | |
2904 | topo.threads_per_core; ++thread_no) { | |
2905 | struct thread_data *t, *t2; | |
2906 | struct core_data *c, *c2; | |
2907 | struct pkg_data *p, *p2; | |
2908 | ||
2909 | t = GET_THREAD(thread_base, thread_no, | |
2910 | core_no, node_no, | |
2911 | pkg_no); | |
2912 | ||
2913 | if (cpu_is_not_present(t->cpu_id)) | |
2914 | continue; | |
2915 | ||
2916 | t2 = GET_THREAD(thread_base2, thread_no, | |
2917 | core_no, node_no, | |
2918 | pkg_no); | |
2919 | ||
2920 | c = GET_CORE(core_base, core_no, | |
2921 | node_no, pkg_no); | |
2922 | c2 = GET_CORE(core_base2, core_no, | |
2923 | node_no, | |
2924 | pkg_no); | |
2925 | ||
2926 | p = GET_PKG(pkg_base, pkg_no); | |
2927 | p2 = GET_PKG(pkg_base2, pkg_no); | |
2928 | ||
2929 | retval = func(t, c, p, t2, c2, p2); | |
2930 | if (retval) | |
2931 | return retval; | |
2932 | } | |
c98d5d94 LB |
2933 | } |
2934 | } | |
2935 | } | |
2936 | return 0; | |
2937 | } | |
2938 | ||
2939 | /* | |
2940 | * run func(cpu) on every cpu in /proc/stat | |
2941 | * return max_cpu number | |
2942 | */ | |
2943 | int for_all_proc_cpus(int (func)(int)) | |
103a8fea LB |
2944 | { |
2945 | FILE *fp; | |
c98d5d94 | 2946 | int cpu_num; |
103a8fea LB |
2947 | int retval; |
2948 | ||
57a42a34 | 2949 | fp = fopen_or_die(proc_stat, "r"); |
103a8fea LB |
2950 | |
2951 | retval = fscanf(fp, "cpu %*d %*d %*d %*d %*d %*d %*d %*d %*d %*d\n"); | |
b2c95d90 JT |
2952 | if (retval != 0) |
2953 | err(1, "%s: failed to parse format", proc_stat); | |
103a8fea | 2954 | |
c98d5d94 LB |
2955 | while (1) { |
2956 | retval = fscanf(fp, "cpu%u %*d %*d %*d %*d %*d %*d %*d %*d %*d %*d\n", &cpu_num); | |
103a8fea LB |
2957 | if (retval != 1) |
2958 | break; | |
2959 | ||
c98d5d94 LB |
2960 | retval = func(cpu_num); |
2961 | if (retval) { | |
2962 | fclose(fp); | |
2963 | return(retval); | |
2964 | } | |
103a8fea LB |
2965 | } |
2966 | fclose(fp); | |
c98d5d94 | 2967 | return 0; |
103a8fea LB |
2968 | } |
2969 | ||
2970 | void re_initialize(void) | |
2971 | { | |
c98d5d94 LB |
2972 | free_all_buffers(); |
2973 | setup_all_buffers(); | |
3d7772ea | 2974 | fprintf(outf, "turbostat: re-initialized with num_cpus %d\n", topo.num_cpus); |
103a8fea LB |
2975 | } |
2976 | ||
843c5791 PB |
2977 | void set_max_cpu_num(void) |
2978 | { | |
2979 | FILE *filep; | |
8201a028 | 2980 | int base_cpu; |
843c5791 | 2981 | unsigned long dummy; |
8201a028 | 2982 | char pathname[64]; |
843c5791 | 2983 | |
8201a028 PB |
2984 | base_cpu = sched_getcpu(); |
2985 | if (base_cpu < 0) | |
2986 | err(1, "cannot find calling cpu ID"); | |
2987 | sprintf(pathname, | |
2988 | "/sys/devices/system/cpu/cpu%d/topology/thread_siblings", | |
2989 | base_cpu); | |
2990 | ||
2991 | filep = fopen_or_die(pathname, "r"); | |
843c5791 | 2992 | topo.max_cpu_num = 0; |
843c5791 | 2993 | while (fscanf(filep, "%lx,", &dummy) == 1) |
0e2d8f05 | 2994 | topo.max_cpu_num += BITMASK_SIZE; |
843c5791 PB |
2995 | fclose(filep); |
2996 | topo.max_cpu_num--; /* 0 based */ | |
2997 | } | |
c98d5d94 | 2998 | |
103a8fea | 2999 | /* |
c98d5d94 LB |
3000 | * count_cpus() |
3001 | * remember the last one seen, it will be the max | |
103a8fea | 3002 | */ |
c98d5d94 | 3003 | int count_cpus(int cpu) |
103a8fea | 3004 | { |
843c5791 | 3005 | topo.num_cpus++; |
c98d5d94 LB |
3006 | return 0; |
3007 | } | |
3008 | int mark_cpu_present(int cpu) | |
3009 | { | |
3010 | CPU_SET_S(cpu, cpu_present_setsize, cpu_present_set); | |
15aaa346 | 3011 | return 0; |
103a8fea LB |
3012 | } |
3013 | ||
8cb48b32 PB |
3014 | int init_thread_id(int cpu) |
3015 | { | |
3016 | cpus[cpu].thread_id = -1; | |
3017 | return 0; | |
3018 | } | |
3019 | ||
562a2d37 LB |
3020 | /* |
3021 | * snapshot_proc_interrupts() | |
3022 | * | |
3023 | * read and record summary of /proc/interrupts | |
3024 | * | |
3025 | * return 1 if config change requires a restart, else return 0 | |
3026 | */ | |
3027 | int snapshot_proc_interrupts(void) | |
3028 | { | |
3029 | static FILE *fp; | |
3030 | int column, retval; | |
3031 | ||
3032 | if (fp == NULL) | |
3033 | fp = fopen_or_die("/proc/interrupts", "r"); | |
3034 | else | |
3035 | rewind(fp); | |
3036 | ||
3037 | /* read 1st line of /proc/interrupts to get cpu* name for each column */ | |
3038 | for (column = 0; column < topo.num_cpus; ++column) { | |
3039 | int cpu_number; | |
3040 | ||
3041 | retval = fscanf(fp, " CPU%d", &cpu_number); | |
3042 | if (retval != 1) | |
3043 | break; | |
3044 | ||
3045 | if (cpu_number > topo.max_cpu_num) { | |
3046 | warn("/proc/interrupts: cpu%d: > %d", cpu_number, topo.max_cpu_num); | |
3047 | return 1; | |
3048 | } | |
3049 | ||
3050 | irq_column_2_cpu[column] = cpu_number; | |
3051 | irqs_per_cpu[cpu_number] = 0; | |
3052 | } | |
3053 | ||
3054 | /* read /proc/interrupt count lines and sum up irqs per cpu */ | |
3055 | while (1) { | |
3056 | int column; | |
3057 | char buf[64]; | |
3058 | ||
3059 | retval = fscanf(fp, " %s:", buf); /* flush irq# "N:" */ | |
3060 | if (retval != 1) | |
3061 | break; | |
3062 | ||
3063 | /* read the count per cpu */ | |
3064 | for (column = 0; column < topo.num_cpus; ++column) { | |
3065 | ||
3066 | int cpu_number, irq_count; | |
3067 | ||
3068 | retval = fscanf(fp, " %d", &irq_count); | |
3069 | if (retval != 1) | |
3070 | break; | |
3071 | ||
3072 | cpu_number = irq_column_2_cpu[column]; | |
3073 | irqs_per_cpu[cpu_number] += irq_count; | |
3074 | ||
3075 | } | |
3076 | ||
3077 | while (getc(fp) != '\n') | |
3078 | ; /* flush interrupt description */ | |
3079 | ||
3080 | } | |
3081 | return 0; | |
3082 | } | |
fdf676e5 LB |
3083 | /* |
3084 | * snapshot_gfx_rc6_ms() | |
3085 | * | |
3086 | * record snapshot of | |
3087 | * /sys/class/drm/card0/power/rc6_residency_ms | |
3088 | * | |
3089 | * return 1 if config change requires a restart, else return 0 | |
3090 | */ | |
3091 | int snapshot_gfx_rc6_ms(void) | |
3092 | { | |
3093 | FILE *fp; | |
3094 | int retval; | |
3095 | ||
3096 | fp = fopen_or_die("/sys/class/drm/card0/power/rc6_residency_ms", "r"); | |
3097 | ||
3098 | retval = fscanf(fp, "%lld", &gfx_cur_rc6_ms); | |
3099 | if (retval != 1) | |
3100 | err(1, "GFX rc6"); | |
3101 | ||
3102 | fclose(fp); | |
3103 | ||
3104 | return 0; | |
3105 | } | |
27d47356 LB |
3106 | /* |
3107 | * snapshot_gfx_mhz() | |
3108 | * | |
3109 | * record snapshot of | |
3110 | * /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz | |
3111 | * | |
3112 | * return 1 if config change requires a restart, else return 0 | |
3113 | */ | |
3114 | int snapshot_gfx_mhz(void) | |
3115 | { | |
3116 | static FILE *fp; | |
3117 | int retval; | |
3118 | ||
3119 | if (fp == NULL) | |
3120 | fp = fopen_or_die("/sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz", "r"); | |
22048c54 | 3121 | else { |
27d47356 | 3122 | rewind(fp); |
22048c54 LB |
3123 | fflush(fp); |
3124 | } | |
27d47356 LB |
3125 | |
3126 | retval = fscanf(fp, "%d", &gfx_cur_mhz); | |
3127 | if (retval != 1) | |
3128 | err(1, "GFX MHz"); | |
3129 | ||
3130 | return 0; | |
3131 | } | |
562a2d37 | 3132 | |
b4b91569 RA |
3133 | /* |
3134 | * snapshot_gfx_cur_mhz() | |
3135 | * | |
3136 | * record snapshot of | |
3137 | * /sys/class/graphics/fb0/device/drm/card0/gt_act_freq_mhz | |
3138 | * | |
3139 | * return 1 if config change requires a restart, else return 0 | |
3140 | */ | |
3141 | int snapshot_gfx_act_mhz(void) | |
3142 | { | |
3143 | static FILE *fp; | |
3144 | int retval; | |
3145 | ||
3146 | if (fp == NULL) | |
3147 | fp = fopen_or_die("/sys/class/graphics/fb0/device/drm/card0/gt_act_freq_mhz", "r"); | |
3148 | else { | |
3149 | rewind(fp); | |
3150 | fflush(fp); | |
3151 | } | |
3152 | ||
3153 | retval = fscanf(fp, "%d", &gfx_act_mhz); | |
3154 | if (retval != 1) | |
3155 | err(1, "GFX ACT MHz"); | |
3156 | ||
3157 | return 0; | |
3158 | } | |
3159 | ||
be0e54c4 LB |
3160 | /* |
3161 | * snapshot_cpu_lpi() | |
3162 | * | |
3163 | * record snapshot of | |
3164 | * /sys/devices/system/cpu/cpuidle/low_power_idle_cpu_residency_us | |
be0e54c4 LB |
3165 | */ |
3166 | int snapshot_cpu_lpi_us(void) | |
3167 | { | |
3168 | FILE *fp; | |
3169 | int retval; | |
3170 | ||
3171 | fp = fopen_or_die("/sys/devices/system/cpu/cpuidle/low_power_idle_cpu_residency_us", "r"); | |
3172 | ||
3173 | retval = fscanf(fp, "%lld", &cpuidle_cur_cpu_lpi_us); | |
5ea7647b PB |
3174 | if (retval != 1) { |
3175 | fprintf(stderr, "Disabling Low Power Idle CPU output\n"); | |
3176 | BIC_NOT_PRESENT(BIC_CPU_LPI); | |
605736c6 | 3177 | fclose(fp); |
5ea7647b PB |
3178 | return -1; |
3179 | } | |
be0e54c4 LB |
3180 | |
3181 | fclose(fp); | |
3182 | ||
3183 | return 0; | |
3184 | } | |
3185 | /* | |
3186 | * snapshot_sys_lpi() | |
3187 | * | |
1f81c5ef | 3188 | * record snapshot of sys_lpi_file |
be0e54c4 LB |
3189 | */ |
3190 | int snapshot_sys_lpi_us(void) | |
3191 | { | |
3192 | FILE *fp; | |
3193 | int retval; | |
3194 | ||
1f81c5ef | 3195 | fp = fopen_or_die(sys_lpi_file, "r"); |
be0e54c4 LB |
3196 | |
3197 | retval = fscanf(fp, "%lld", &cpuidle_cur_sys_lpi_us); | |
5ea7647b PB |
3198 | if (retval != 1) { |
3199 | fprintf(stderr, "Disabling Low Power Idle System output\n"); | |
3200 | BIC_NOT_PRESENT(BIC_SYS_LPI); | |
15423b95 | 3201 | fclose(fp); |
5ea7647b PB |
3202 | return -1; |
3203 | } | |
be0e54c4 LB |
3204 | fclose(fp); |
3205 | ||
3206 | return 0; | |
3207 | } | |
562a2d37 LB |
3208 | /* |
3209 | * snapshot /proc and /sys files | |
3210 | * | |
3211 | * return 1 if configuration restart needed, else return 0 | |
3212 | */ | |
3213 | int snapshot_proc_sysfs_files(void) | |
3214 | { | |
218f0e8d LB |
3215 | if (DO_BIC(BIC_IRQ)) |
3216 | if (snapshot_proc_interrupts()) | |
3217 | return 1; | |
562a2d37 | 3218 | |
812db3f7 | 3219 | if (DO_BIC(BIC_GFX_rc6)) |
fdf676e5 LB |
3220 | snapshot_gfx_rc6_ms(); |
3221 | ||
812db3f7 | 3222 | if (DO_BIC(BIC_GFXMHz)) |
27d47356 LB |
3223 | snapshot_gfx_mhz(); |
3224 | ||
b4b91569 RA |
3225 | if (DO_BIC(BIC_GFXACTMHz)) |
3226 | snapshot_gfx_act_mhz(); | |
3227 | ||
be0e54c4 LB |
3228 | if (DO_BIC(BIC_CPU_LPI)) |
3229 | snapshot_cpu_lpi_us(); | |
3230 | ||
3231 | if (DO_BIC(BIC_SYS_LPI)) | |
3232 | snapshot_sys_lpi_us(); | |
3233 | ||
562a2d37 LB |
3234 | return 0; |
3235 | } | |
3236 | ||
8aa2ed0b LB |
3237 | int exit_requested; |
3238 | ||
3239 | static void signal_handler (int signal) | |
3240 | { | |
3241 | switch (signal) { | |
3242 | case SIGINT: | |
3243 | exit_requested = 1; | |
3244 | if (debug) | |
3245 | fprintf(stderr, " SIGINT\n"); | |
3246 | break; | |
07211960 LB |
3247 | case SIGUSR1: |
3248 | if (debug > 1) | |
3249 | fprintf(stderr, "SIGUSR1\n"); | |
3250 | break; | |
8aa2ed0b LB |
3251 | } |
3252 | } | |
3253 | ||
3254 | void setup_signal_handler(void) | |
3255 | { | |
3256 | struct sigaction sa; | |
3257 | ||
3258 | memset(&sa, 0, sizeof(sa)); | |
3259 | ||
3260 | sa.sa_handler = &signal_handler; | |
3261 | ||
3262 | if (sigaction(SIGINT, &sa, NULL) < 0) | |
3263 | err(1, "sigaction SIGINT"); | |
07211960 LB |
3264 | if (sigaction(SIGUSR1, &sa, NULL) < 0) |
3265 | err(1, "sigaction SIGUSR1"); | |
8aa2ed0b | 3266 | } |
b9ad8ee0 | 3267 | |
47936f94 | 3268 | void do_sleep(void) |
b9ad8ee0 | 3269 | { |
c026c236 AB |
3270 | struct timeval tout; |
3271 | struct timespec rest; | |
b9ad8ee0 LB |
3272 | fd_set readfds; |
3273 | int retval; | |
3274 | ||
3275 | FD_ZERO(&readfds); | |
3276 | FD_SET(0, &readfds); | |
3277 | ||
c026c236 | 3278 | if (ignore_stdin) { |
47936f94 AB |
3279 | nanosleep(&interval_ts, NULL); |
3280 | return; | |
3281 | } | |
b9ad8ee0 | 3282 | |
c026c236 AB |
3283 | tout = interval_tv; |
3284 | retval = select(1, &readfds, NULL, NULL, &tout); | |
b9ad8ee0 LB |
3285 | |
3286 | if (retval == 1) { | |
b9ad8ee0 LB |
3287 | switch (getc(stdin)) { |
3288 | case 'q': | |
3289 | exit_requested = 1; | |
3290 | break; | |
c026c236 AB |
3291 | case EOF: |
3292 | /* | |
3293 | * 'stdin' is a pipe closed on the other end. There | |
3294 | * won't be any further input. | |
3295 | */ | |
3296 | ignore_stdin = 1; | |
3297 | /* Sleep the rest of the time */ | |
3298 | rest.tv_sec = (tout.tv_sec + tout.tv_usec / 1000000); | |
3299 | rest.tv_nsec = (tout.tv_usec % 1000000) * 1000; | |
3300 | nanosleep(&rest, NULL); | |
b9ad8ee0 | 3301 | } |
b9ad8ee0 | 3302 | } |
b9ad8ee0 | 3303 | } |
47936f94 | 3304 | |
87e15da9 CY |
3305 | int get_msr_sum(int cpu, off_t offset, unsigned long long *msr) |
3306 | { | |
3307 | int ret, idx; | |
3308 | unsigned long long msr_cur, msr_last; | |
3309 | ||
3310 | if (!per_cpu_msr_sum) | |
3311 | return 1; | |
3312 | ||
3313 | idx = offset_to_idx(offset); | |
3314 | if (idx < 0) | |
3315 | return idx; | |
3316 | /* get_msr_sum() = sum + (get_msr() - last) */ | |
3317 | ret = get_msr(cpu, offset, &msr_cur); | |
3318 | if (ret) | |
3319 | return ret; | |
3320 | msr_last = per_cpu_msr_sum[cpu].entries[idx].last; | |
3321 | DELTA_WRAP32(msr_cur, msr_last); | |
3322 | *msr = msr_last + per_cpu_msr_sum[cpu].entries[idx].sum; | |
3323 | ||
3324 | return 0; | |
3325 | } | |
3326 | ||
3327 | timer_t timerid; | |
3328 | ||
3329 | /* Timer callback, update the sum of MSRs periodically. */ | |
3330 | static int update_msr_sum(struct thread_data *t, struct core_data *c, struct pkg_data *p) | |
3331 | { | |
3332 | int i, ret; | |
3333 | int cpu = t->cpu_id; | |
3334 | ||
3335 | for (i = IDX_PKG_ENERGY; i < IDX_COUNT; i++) { | |
3336 | unsigned long long msr_cur, msr_last; | |
3337 | int offset; | |
3338 | ||
3339 | if (!idx_valid(i)) | |
3340 | continue; | |
3341 | offset = idx_to_offset(i); | |
3342 | if (offset < 0) | |
3343 | continue; | |
3344 | ret = get_msr(cpu, offset, &msr_cur); | |
3345 | if (ret) { | |
3346 | fprintf(outf, "Can not update msr(0x%x)\n", offset); | |
3347 | continue; | |
3348 | } | |
3349 | ||
3350 | msr_last = per_cpu_msr_sum[cpu].entries[i].last; | |
3351 | per_cpu_msr_sum[cpu].entries[i].last = msr_cur & 0xffffffff; | |
3352 | ||
3353 | DELTA_WRAP32(msr_cur, msr_last); | |
3354 | per_cpu_msr_sum[cpu].entries[i].sum += msr_last; | |
3355 | } | |
3356 | return 0; | |
3357 | } | |
3358 | ||
3359 | static void | |
3360 | msr_record_handler(union sigval v) | |
3361 | { | |
3362 | for_all_cpus(update_msr_sum, EVEN_COUNTERS); | |
3363 | } | |
3364 | ||
3365 | void msr_sum_record(void) | |
3366 | { | |
3367 | struct itimerspec its; | |
3368 | struct sigevent sev; | |
3369 | ||
3370 | per_cpu_msr_sum = calloc(topo.max_cpu_num + 1, sizeof(struct msr_sum_array)); | |
3371 | if (!per_cpu_msr_sum) { | |
3372 | fprintf(outf, "Can not allocate memory for long time MSR.\n"); | |
3373 | return; | |
3374 | } | |
3375 | /* | |
3376 | * Signal handler might be restricted, so use thread notifier instead. | |
3377 | */ | |
3378 | memset(&sev, 0, sizeof(struct sigevent)); | |
3379 | sev.sigev_notify = SIGEV_THREAD; | |
3380 | sev.sigev_notify_function = msr_record_handler; | |
3381 | ||
3382 | sev.sigev_value.sival_ptr = &timerid; | |
3383 | if (timer_create(CLOCK_REALTIME, &sev, &timerid) == -1) { | |
3384 | fprintf(outf, "Can not create timer.\n"); | |
3385 | goto release_msr; | |
3386 | } | |
3387 | ||
3388 | its.it_value.tv_sec = 0; | |
3389 | its.it_value.tv_nsec = 1; | |
3390 | /* | |
3391 | * A wraparound time has been calculated early. | |
3392 | * Some sources state that the peak power for a | |
3393 | * microprocessor is usually 1.5 times the TDP rating, | |
3394 | * use 2 * TDP for safety. | |
3395 | */ | |
3396 | its.it_interval.tv_sec = rapl_joule_counter_range / 2; | |
3397 | its.it_interval.tv_nsec = 0; | |
3398 | ||
3399 | if (timer_settime(timerid, 0, &its, NULL) == -1) { | |
3400 | fprintf(outf, "Can not set timer.\n"); | |
3401 | goto release_timer; | |
3402 | } | |
3403 | return; | |
3404 | ||
3405 | release_timer: | |
3406 | timer_delete(timerid); | |
3407 | release_msr: | |
3408 | free(per_cpu_msr_sum); | |
3409 | } | |
4c2122d4 | 3410 | |
103a8fea LB |
3411 | void turbostat_loop() |
3412 | { | |
c98d5d94 | 3413 | int retval; |
e52966c0 | 3414 | int restarted = 0; |
023fe0ac | 3415 | int done_iters = 0; |
c98d5d94 | 3416 | |
8aa2ed0b LB |
3417 | setup_signal_handler(); |
3418 | ||
103a8fea | 3419 | restart: |
e52966c0 LB |
3420 | restarted++; |
3421 | ||
562a2d37 | 3422 | snapshot_proc_sysfs_files(); |
c98d5d94 | 3423 | retval = for_all_cpus(get_counters, EVEN_COUNTERS); |
4c2122d4 | 3424 | first_counter_read = 0; |
d91bb17c LB |
3425 | if (retval < -1) { |
3426 | exit(retval); | |
3427 | } else if (retval == -1) { | |
3d7772ea | 3428 | if (restarted > 10) { |
e52966c0 LB |
3429 | exit(retval); |
3430 | } | |
c98d5d94 LB |
3431 | re_initialize(); |
3432 | goto restart; | |
3433 | } | |
e52966c0 | 3434 | restarted = 0; |
023fe0ac | 3435 | done_iters = 0; |
103a8fea LB |
3436 | gettimeofday(&tv_even, (struct timezone *)NULL); |
3437 | ||
3438 | while (1) { | |
c98d5d94 | 3439 | if (for_all_proc_cpus(cpu_is_not_present)) { |
103a8fea LB |
3440 | re_initialize(); |
3441 | goto restart; | |
3442 | } | |
b9ad8ee0 | 3443 | do_sleep(); |
562a2d37 LB |
3444 | if (snapshot_proc_sysfs_files()) |
3445 | goto restart; | |
c98d5d94 | 3446 | retval = for_all_cpus(get_counters, ODD_COUNTERS); |
d91bb17c LB |
3447 | if (retval < -1) { |
3448 | exit(retval); | |
3449 | } else if (retval == -1) { | |
15aaa346 LB |
3450 | re_initialize(); |
3451 | goto restart; | |
3452 | } | |
103a8fea | 3453 | gettimeofday(&tv_odd, (struct timezone *)NULL); |
103a8fea | 3454 | timersub(&tv_odd, &tv_even, &tv_delta); |
ba3dec99 LB |
3455 | if (for_all_cpus_2(delta_cpu, ODD_COUNTERS, EVEN_COUNTERS)) { |
3456 | re_initialize(); | |
3457 | goto restart; | |
3458 | } | |
c98d5d94 LB |
3459 | compute_average(EVEN_COUNTERS); |
3460 | format_all_counters(EVEN_COUNTERS); | |
b7d8c148 | 3461 | flush_output_stdout(); |
8aa2ed0b LB |
3462 | if (exit_requested) |
3463 | break; | |
023fe0ac CY |
3464 | if (num_iterations && ++done_iters >= num_iterations) |
3465 | break; | |
b9ad8ee0 | 3466 | do_sleep(); |
562a2d37 LB |
3467 | if (snapshot_proc_sysfs_files()) |
3468 | goto restart; | |
c98d5d94 | 3469 | retval = for_all_cpus(get_counters, EVEN_COUNTERS); |
d91bb17c LB |
3470 | if (retval < -1) { |
3471 | exit(retval); | |
3472 | } else if (retval == -1) { | |
103a8fea LB |
3473 | re_initialize(); |
3474 | goto restart; | |
3475 | } | |
103a8fea | 3476 | gettimeofday(&tv_even, (struct timezone *)NULL); |
103a8fea | 3477 | timersub(&tv_even, &tv_odd, &tv_delta); |
ba3dec99 LB |
3478 | if (for_all_cpus_2(delta_cpu, EVEN_COUNTERS, ODD_COUNTERS)) { |
3479 | re_initialize(); | |
3480 | goto restart; | |
3481 | } | |
c98d5d94 LB |
3482 | compute_average(ODD_COUNTERS); |
3483 | format_all_counters(ODD_COUNTERS); | |
b7d8c148 | 3484 | flush_output_stdout(); |
8aa2ed0b LB |
3485 | if (exit_requested) |
3486 | break; | |
023fe0ac CY |
3487 | if (num_iterations && ++done_iters >= num_iterations) |
3488 | break; | |
103a8fea LB |
3489 | } |
3490 | } | |
3491 | ||
3492 | void check_dev_msr() | |
3493 | { | |
3494 | struct stat sb; | |
7ce7d5de | 3495 | char pathname[32]; |
103a8fea | 3496 | |
7ce7d5de PB |
3497 | sprintf(pathname, "/dev/cpu/%d/msr", base_cpu); |
3498 | if (stat(pathname, &sb)) | |
a21d38c8 LB |
3499 | if (system("/sbin/modprobe msr > /dev/null 2>&1")) |
3500 | err(-5, "no /dev/cpu/0/msr, Try \"# modprobe msr\" "); | |
103a8fea LB |
3501 | } |
3502 | ||
fcaa681c LB |
3503 | /* |
3504 | * check for CAP_SYS_RAWIO | |
3505 | * return 0 on success | |
3506 | * return 1 on fail | |
3507 | */ | |
3508 | int check_for_cap_sys_rawio(void) | |
103a8fea | 3509 | { |
fcaa681c LB |
3510 | cap_t caps; |
3511 | cap_flag_value_t cap_flag_value; | |
98481e79 | 3512 | |
fcaa681c LB |
3513 | caps = cap_get_proc(); |
3514 | if (caps == NULL) | |
3515 | err(-6, "cap_get_proc\n"); | |
98481e79 | 3516 | |
fcaa681c LB |
3517 | if (cap_get_flag(caps, CAP_SYS_RAWIO, CAP_EFFECTIVE, &cap_flag_value)) |
3518 | err(-6, "cap_get\n"); | |
3519 | ||
3520 | if (cap_flag_value != CAP_SET) { | |
98481e79 LB |
3521 | warnx("capget(CAP_SYS_RAWIO) failed," |
3522 | " try \"# setcap cap_sys_rawio=ep %s\"", progname); | |
fcaa681c | 3523 | return 1; |
98481e79 LB |
3524 | } |
3525 | ||
fcaa681c LB |
3526 | if (cap_free(caps) == -1) |
3527 | err(-6, "cap_free\n"); | |
3528 | ||
3529 | return 0; | |
3530 | } | |
3531 | void check_permissions(void) | |
3532 | { | |
3533 | int do_exit = 0; | |
3534 | char pathname[32]; | |
3535 | ||
3536 | /* check for CAP_SYS_RAWIO */ | |
3537 | do_exit += check_for_cap_sys_rawio(); | |
3538 | ||
98481e79 | 3539 | /* test file permissions */ |
7ce7d5de PB |
3540 | sprintf(pathname, "/dev/cpu/%d/msr", base_cpu); |
3541 | if (euidaccess(pathname, R_OK)) { | |
98481e79 LB |
3542 | do_exit++; |
3543 | warn("/dev/cpu/0/msr open failed, try chown or chmod +r /dev/cpu/*/msr"); | |
3544 | } | |
3545 | ||
3546 | /* if all else fails, thell them to be root */ | |
3547 | if (do_exit) | |
3548 | if (getuid() != 0) | |
d7899447 | 3549 | warnx("... or simply run as root"); |
98481e79 LB |
3550 | |
3551 | if (do_exit) | |
3552 | exit(-6); | |
103a8fea LB |
3553 | } |
3554 | ||
d7899447 LB |
3555 | /* |
3556 | * NHM adds support for additional MSRs: | |
3557 | * | |
3558 | * MSR_SMI_COUNT 0x00000034 | |
3559 | * | |
ec0adc53 | 3560 | * MSR_PLATFORM_INFO 0x000000ce |
1df2e55a | 3561 | * MSR_PKG_CST_CONFIG_CONTROL 0x000000e2 |
d7899447 | 3562 | * |
cf4cbe53 LB |
3563 | * MSR_MISC_PWR_MGMT 0x000001aa |
3564 | * | |
d7899447 LB |
3565 | * MSR_PKG_C3_RESIDENCY 0x000003f8 |
3566 | * MSR_PKG_C6_RESIDENCY 0x000003f9 | |
3567 | * MSR_CORE_C3_RESIDENCY 0x000003fc | |
3568 | * MSR_CORE_C6_RESIDENCY 0x000003fd | |
3569 | * | |
ee7e38e3 | 3570 | * Side effect: |
1df2e55a | 3571 | * sets global pkg_cstate_limit to decode MSR_PKG_CST_CONFIG_CONTROL |
33148d67 | 3572 | * sets has_misc_feature_control |
d7899447 | 3573 | */ |
ee7e38e3 | 3574 | int probe_nhm_msrs(unsigned int family, unsigned int model) |
103a8fea | 3575 | { |
ee7e38e3 | 3576 | unsigned long long msr; |
21ed5574 | 3577 | unsigned int base_ratio; |
ee7e38e3 LB |
3578 | int *pkg_cstate_limits; |
3579 | ||
103a8fea LB |
3580 | if (!genuine_intel) |
3581 | return 0; | |
3582 | ||
3583 | if (family != 6) | |
3584 | return 0; | |
3585 | ||
21ed5574 LB |
3586 | bclk = discover_bclk(family, model); |
3587 | ||
103a8fea | 3588 | switch (model) { |
869ce69e | 3589 | case INTEL_FAM6_NEHALEM: /* Core i7 and i5 Processor - Clarksfield, Lynnfield, Jasper Forest */ |
869ce69e | 3590 | case INTEL_FAM6_NEHALEM_EX: /* Nehalem-EX Xeon - Beckton */ |
ee7e38e3 LB |
3591 | pkg_cstate_limits = nhm_pkg_cstate_limits; |
3592 | break; | |
869ce69e LB |
3593 | case INTEL_FAM6_SANDYBRIDGE: /* SNB */ |
3594 | case INTEL_FAM6_SANDYBRIDGE_X: /* SNB Xeon */ | |
3595 | case INTEL_FAM6_IVYBRIDGE: /* IVB */ | |
3596 | case INTEL_FAM6_IVYBRIDGE_X: /* IVB Xeon */ | |
ee7e38e3 | 3597 | pkg_cstate_limits = snb_pkg_cstate_limits; |
33148d67 | 3598 | has_misc_feature_control = 1; |
ee7e38e3 | 3599 | break; |
c66f78a6 | 3600 | case INTEL_FAM6_HASWELL: /* HSW */ |
5e741407 | 3601 | case INTEL_FAM6_HASWELL_G: /* HSW */ |
869ce69e | 3602 | case INTEL_FAM6_HASWELL_X: /* HSX */ |
77e5517c | 3603 | case INTEL_FAM6_HASWELL_L: /* HSW */ |
c66f78a6 | 3604 | case INTEL_FAM6_BROADWELL: /* BDW */ |
5e741407 | 3605 | case INTEL_FAM6_BROADWELL_G: /* BDW */ |
869ce69e | 3606 | case INTEL_FAM6_BROADWELL_X: /* BDX */ |
af239c44 PZ |
3607 | case INTEL_FAM6_SKYLAKE_L: /* SKL */ |
3608 | case INTEL_FAM6_CANNONLAKE_L: /* CNL */ | |
ee7e38e3 | 3609 | pkg_cstate_limits = hsw_pkg_cstate_limits; |
33148d67 | 3610 | has_misc_feature_control = 1; |
ee7e38e3 | 3611 | break; |
d8ebb442 LB |
3612 | case INTEL_FAM6_SKYLAKE_X: /* SKX */ |
3613 | pkg_cstate_limits = skx_pkg_cstate_limits; | |
33148d67 | 3614 | has_misc_feature_control = 1; |
d8ebb442 | 3615 | break; |
f2c4db1b | 3616 | case INTEL_FAM6_ATOM_SILVERMONT: /* BYT */ |
cf4cbe53 | 3617 | no_MSR_MISC_PWR_MGMT = 1; |
5ebb34ed | 3618 | case INTEL_FAM6_ATOM_SILVERMONT_D: /* AVN */ |
ee7e38e3 LB |
3619 | pkg_cstate_limits = slv_pkg_cstate_limits; |
3620 | break; | |
869ce69e | 3621 | case INTEL_FAM6_ATOM_AIRMONT: /* AMT */ |
ee7e38e3 | 3622 | pkg_cstate_limits = amt_pkg_cstate_limits; |
cf4cbe53 | 3623 | no_MSR_MISC_PWR_MGMT = 1; |
ee7e38e3 | 3624 | break; |
869ce69e | 3625 | case INTEL_FAM6_XEON_PHI_KNL: /* PHI */ |
ee7e38e3 LB |
3626 | pkg_cstate_limits = phi_pkg_cstate_limits; |
3627 | break; | |
869ce69e | 3628 | case INTEL_FAM6_ATOM_GOLDMONT: /* BXT */ |
f2c4db1b | 3629 | case INTEL_FAM6_ATOM_GOLDMONT_PLUS: |
5ebb34ed | 3630 | case INTEL_FAM6_ATOM_GOLDMONT_D: /* DNV */ |
f6708400 | 3631 | case INTEL_FAM6_ATOM_TREMONT: /* EHL */ |
20de0dab | 3632 | case INTEL_FAM6_ATOM_TREMONT_D: /* JVL */ |
445640a5 | 3633 | pkg_cstate_limits = glm_pkg_cstate_limits; |
e4085d54 | 3634 | break; |
103a8fea LB |
3635 | default: |
3636 | return 0; | |
3637 | } | |
1df2e55a | 3638 | get_msr(base_cpu, MSR_PKG_CST_CONFIG_CONTROL, &msr); |
e9257f5f | 3639 | pkg_cstate_limit = pkg_cstate_limits[msr & 0xF]; |
ee7e38e3 | 3640 | |
ec0adc53 | 3641 | get_msr(base_cpu, MSR_PLATFORM_INFO, &msr); |
21ed5574 LB |
3642 | base_ratio = (msr >> 8) & 0xFF; |
3643 | ||
3644 | base_hz = base_ratio * bclk * 1000000; | |
3645 | has_base_hz = 1; | |
ee7e38e3 | 3646 | return 1; |
103a8fea | 3647 | } |
0f7887c4 | 3648 | /* |
495c7654 | 3649 | * SLV client has support for unique MSRs: |
0f7887c4 LB |
3650 | * |
3651 | * MSR_CC6_DEMOTION_POLICY_CONFIG | |
3652 | * MSR_MC6_DEMOTION_POLICY_CONFIG | |
3653 | */ | |
3654 | ||
3655 | int has_slv_msrs(unsigned int family, unsigned int model) | |
3656 | { | |
3657 | if (!genuine_intel) | |
3658 | return 0; | |
3659 | ||
3660 | switch (model) { | |
f2c4db1b PZ |
3661 | case INTEL_FAM6_ATOM_SILVERMONT: |
3662 | case INTEL_FAM6_ATOM_SILVERMONT_MID: | |
3663 | case INTEL_FAM6_ATOM_AIRMONT_MID: | |
0f7887c4 LB |
3664 | return 1; |
3665 | } | |
3666 | return 0; | |
3667 | } | |
7170a374 LB |
3668 | int is_dnv(unsigned int family, unsigned int model) |
3669 | { | |
3670 | ||
3671 | if (!genuine_intel) | |
3672 | return 0; | |
3673 | ||
3674 | switch (model) { | |
5ebb34ed | 3675 | case INTEL_FAM6_ATOM_GOLDMONT_D: |
7170a374 LB |
3676 | return 1; |
3677 | } | |
3678 | return 0; | |
3679 | } | |
ade0ebac LB |
3680 | int is_bdx(unsigned int family, unsigned int model) |
3681 | { | |
3682 | ||
3683 | if (!genuine_intel) | |
3684 | return 0; | |
3685 | ||
3686 | switch (model) { | |
3687 | case INTEL_FAM6_BROADWELL_X: | |
ade0ebac LB |
3688 | return 1; |
3689 | } | |
3690 | return 0; | |
3691 | } | |
34c76197 LB |
3692 | int is_skx(unsigned int family, unsigned int model) |
3693 | { | |
3694 | ||
3695 | if (!genuine_intel) | |
3696 | return 0; | |
3697 | ||
3698 | switch (model) { | |
3699 | case INTEL_FAM6_SKYLAKE_X: | |
3700 | return 1; | |
3701 | } | |
3702 | return 0; | |
3703 | } | |
f6708400 CY |
3704 | int is_ehl(unsigned int family, unsigned int model) |
3705 | { | |
3706 | if (!genuine_intel) | |
3707 | return 0; | |
3708 | ||
3709 | switch (model) { | |
3710 | case INTEL_FAM6_ATOM_TREMONT: | |
3711 | return 1; | |
3712 | } | |
3713 | return 0; | |
3714 | } | |
20de0dab AL |
3715 | int is_jvl(unsigned int family, unsigned int model) |
3716 | { | |
3717 | if (!genuine_intel) | |
3718 | return 0; | |
3719 | ||
3720 | switch (model) { | |
3721 | case INTEL_FAM6_ATOM_TREMONT_D: | |
3722 | return 1; | |
3723 | } | |
3724 | return 0; | |
3725 | } | |
0f7887c4 | 3726 | |
31e07522 | 3727 | int has_turbo_ratio_limit(unsigned int family, unsigned int model) |
d7899447 | 3728 | { |
0f7887c4 LB |
3729 | if (has_slv_msrs(family, model)) |
3730 | return 0; | |
3731 | ||
d7899447 LB |
3732 | switch (model) { |
3733 | /* Nehalem compatible, but do not include turbo-ratio limit support */ | |
869ce69e | 3734 | case INTEL_FAM6_NEHALEM_EX: /* Nehalem-EX Xeon - Beckton */ |
869ce69e | 3735 | case INTEL_FAM6_XEON_PHI_KNL: /* PHI - Knights Landing (different MSR definition) */ |
d7899447 LB |
3736 | return 0; |
3737 | default: | |
3738 | return 1; | |
3739 | } | |
3740 | } | |
0f7887c4 LB |
3741 | int has_atom_turbo_ratio_limit(unsigned int family, unsigned int model) |
3742 | { | |
3743 | if (has_slv_msrs(family, model)) | |
3744 | return 1; | |
3745 | ||
3746 | return 0; | |
3747 | } | |
6574a5d5 LB |
3748 | int has_ivt_turbo_ratio_limit(unsigned int family, unsigned int model) |
3749 | { | |
3750 | if (!genuine_intel) | |
3751 | return 0; | |
3752 | ||
3753 | if (family != 6) | |
3754 | return 0; | |
3755 | ||
3756 | switch (model) { | |
869ce69e LB |
3757 | case INTEL_FAM6_IVYBRIDGE_X: /* IVB Xeon */ |
3758 | case INTEL_FAM6_HASWELL_X: /* HSW Xeon */ | |
fcd17211 LB |
3759 | return 1; |
3760 | default: | |
3761 | return 0; | |
3762 | } | |
3763 | } | |
3764 | int has_hsw_turbo_ratio_limit(unsigned int family, unsigned int model) | |
3765 | { | |
3766 | if (!genuine_intel) | |
3767 | return 0; | |
3768 | ||
3769 | if (family != 6) | |
3770 | return 0; | |
3771 | ||
3772 | switch (model) { | |
869ce69e | 3773 | case INTEL_FAM6_HASWELL_X: /* HSW Xeon */ |
6574a5d5 LB |
3774 | return 1; |
3775 | default: | |
3776 | return 0; | |
3777 | } | |
3778 | } | |
3779 | ||
fb5d4327 DC |
3780 | int has_knl_turbo_ratio_limit(unsigned int family, unsigned int model) |
3781 | { | |
3782 | if (!genuine_intel) | |
3783 | return 0; | |
3784 | ||
3785 | if (family != 6) | |
3786 | return 0; | |
3787 | ||
3788 | switch (model) { | |
869ce69e | 3789 | case INTEL_FAM6_XEON_PHI_KNL: /* Knights Landing */ |
fb5d4327 DC |
3790 | return 1; |
3791 | default: | |
3792 | return 0; | |
3793 | } | |
3794 | } | |
31e07522 LB |
3795 | int has_glm_turbo_ratio_limit(unsigned int family, unsigned int model) |
3796 | { | |
3797 | if (!genuine_intel) | |
3798 | return 0; | |
3799 | ||
3800 | if (family != 6) | |
3801 | return 0; | |
3802 | ||
3803 | switch (model) { | |
3804 | case INTEL_FAM6_ATOM_GOLDMONT: | |
3805 | case INTEL_FAM6_SKYLAKE_X: | |
3806 | return 1; | |
3807 | default: | |
3808 | return 0; | |
3809 | } | |
3810 | } | |
6fb3143b LB |
3811 | int has_config_tdp(unsigned int family, unsigned int model) |
3812 | { | |
3813 | if (!genuine_intel) | |
3814 | return 0; | |
3815 | ||
3816 | if (family != 6) | |
3817 | return 0; | |
3818 | ||
3819 | switch (model) { | |
869ce69e | 3820 | case INTEL_FAM6_IVYBRIDGE: /* IVB */ |
c66f78a6 | 3821 | case INTEL_FAM6_HASWELL: /* HSW */ |
869ce69e | 3822 | case INTEL_FAM6_HASWELL_X: /* HSX */ |
77e5517c | 3823 | case INTEL_FAM6_HASWELL_L: /* HSW */ |
5e741407 | 3824 | case INTEL_FAM6_HASWELL_G: /* HSW */ |
c66f78a6 | 3825 | case INTEL_FAM6_BROADWELL: /* BDW */ |
5e741407 | 3826 | case INTEL_FAM6_BROADWELL_G: /* BDW */ |
869ce69e | 3827 | case INTEL_FAM6_BROADWELL_X: /* BDX */ |
af239c44 PZ |
3828 | case INTEL_FAM6_SKYLAKE_L: /* SKL */ |
3829 | case INTEL_FAM6_CANNONLAKE_L: /* CNL */ | |
869ce69e LB |
3830 | case INTEL_FAM6_SKYLAKE_X: /* SKX */ |
3831 | ||
3832 | case INTEL_FAM6_XEON_PHI_KNL: /* Knights Landing */ | |
6fb3143b LB |
3833 | return 1; |
3834 | default: | |
3835 | return 0; | |
3836 | } | |
3837 | } | |
3838 | ||
fecb3bc8 DA |
3839 | static void |
3840 | remove_underbar(char *s) | |
3841 | { | |
3842 | char *to = s; | |
3843 | ||
3844 | while (*s) { | |
3845 | if (*s != '_') | |
3846 | *to++ = *s; | |
3847 | s++; | |
3848 | } | |
3849 | ||
3850 | *to = 0; | |
3851 | } | |
3852 | ||
fcd17211 | 3853 | static void |
1b69317d | 3854 | dump_cstate_pstate_config_info(unsigned int family, unsigned int model) |
fcd17211 LB |
3855 | { |
3856 | if (!do_nhm_platform_info) | |
3857 | return; | |
3858 | ||
3859 | dump_nhm_platform_info(); | |
3860 | ||
3861 | if (has_hsw_turbo_ratio_limit(family, model)) | |
3862 | dump_hsw_turbo_ratio_limits(); | |
3863 | ||
3864 | if (has_ivt_turbo_ratio_limit(family, model)) | |
3865 | dump_ivt_turbo_ratio_limits(); | |
3866 | ||
31e07522 LB |
3867 | if (has_turbo_ratio_limit(family, model)) |
3868 | dump_turbo_ratio_limits(family, model); | |
fcd17211 | 3869 | |
0f7887c4 LB |
3870 | if (has_atom_turbo_ratio_limit(family, model)) |
3871 | dump_atom_turbo_ratio_limits(); | |
3872 | ||
fb5d4327 DC |
3873 | if (has_knl_turbo_ratio_limit(family, model)) |
3874 | dump_knl_turbo_ratio_limits(); | |
3875 | ||
6fb3143b LB |
3876 | if (has_config_tdp(family, model)) |
3877 | dump_config_tdp(); | |
3878 | ||
fcd17211 LB |
3879 | dump_nhm_cst_cfg(); |
3880 | } | |
3881 | ||
abdcbdb2 LB |
3882 | static void dump_sysfs_file(char *path) |
3883 | { | |
3884 | FILE *input; | |
3885 | char cpuidle_buf[64]; | |
3886 | ||
3887 | input = fopen(path, "r"); | |
3888 | if (input == NULL) { | |
3889 | if (debug) | |
3890 | fprintf(outf, "NSFOD %s\n", path); | |
3891 | return; | |
3892 | } | |
3893 | if (!fgets(cpuidle_buf, sizeof(cpuidle_buf), input)) | |
3894 | err(1, "%s: failed to read file", path); | |
3895 | fclose(input); | |
3896 | ||
3897 | fprintf(outf, "%s: %s", strrchr(path, '/') + 1, cpuidle_buf); | |
3898 | } | |
41618e63 LB |
3899 | static void |
3900 | dump_sysfs_cstate_config(void) | |
3901 | { | |
3902 | char path[64]; | |
3903 | char name_buf[16]; | |
3904 | char desc[64]; | |
3905 | FILE *input; | |
3906 | int state; | |
3907 | char *sp; | |
3908 | ||
abdcbdb2 LB |
3909 | if (access("/sys/devices/system/cpu/cpuidle", R_OK)) { |
3910 | fprintf(outf, "cpuidle not loaded\n"); | |
3911 | return; | |
3912 | } | |
3913 | ||
3914 | dump_sysfs_file("/sys/devices/system/cpu/cpuidle/current_driver"); | |
3915 | dump_sysfs_file("/sys/devices/system/cpu/cpuidle/current_governor"); | |
3916 | dump_sysfs_file("/sys/devices/system/cpu/cpuidle/current_governor_ro"); | |
3917 | ||
41618e63 LB |
3918 | for (state = 0; state < 10; ++state) { |
3919 | ||
3920 | sprintf(path, "/sys/devices/system/cpu/cpu%d/cpuidle/state%d/name", | |
3921 | base_cpu, state); | |
3922 | input = fopen(path, "r"); | |
3923 | if (input == NULL) | |
3924 | continue; | |
8173c336 BH |
3925 | if (!fgets(name_buf, sizeof(name_buf), input)) |
3926 | err(1, "%s: failed to read file", path); | |
41618e63 LB |
3927 | |
3928 | /* truncate "C1-HSW\n" to "C1", or truncate "C1\n" to "C1" */ | |
3929 | sp = strchr(name_buf, '-'); | |
3930 | if (!sp) | |
3931 | sp = strchrnul(name_buf, '\n'); | |
3932 | *sp = '\0'; | |
41618e63 LB |
3933 | fclose(input); |
3934 | ||
fecb3bc8 DA |
3935 | remove_underbar(name_buf); |
3936 | ||
41618e63 LB |
3937 | sprintf(path, "/sys/devices/system/cpu/cpu%d/cpuidle/state%d/desc", |
3938 | base_cpu, state); | |
3939 | input = fopen(path, "r"); | |
3940 | if (input == NULL) | |
3941 | continue; | |
8173c336 BH |
3942 | if (!fgets(desc, sizeof(desc), input)) |
3943 | err(1, "%s: failed to read file", path); | |
41618e63 LB |
3944 | |
3945 | fprintf(outf, "cpu%d: %s: %s", base_cpu, name_buf, desc); | |
3946 | fclose(input); | |
3947 | } | |
3948 | } | |
7293fccd LB |
3949 | static void |
3950 | dump_sysfs_pstate_config(void) | |
3951 | { | |
3952 | char path[64]; | |
3953 | char driver_buf[64]; | |
3954 | char governor_buf[64]; | |
3955 | FILE *input; | |
3956 | int turbo; | |
3957 | ||
3958 | sprintf(path, "/sys/devices/system/cpu/cpu%d/cpufreq/scaling_driver", | |
3959 | base_cpu); | |
3960 | input = fopen(path, "r"); | |
3961 | if (input == NULL) { | |
0a42d235 | 3962 | fprintf(outf, "NSFOD %s\n", path); |
7293fccd LB |
3963 | return; |
3964 | } | |
8173c336 BH |
3965 | if (!fgets(driver_buf, sizeof(driver_buf), input)) |
3966 | err(1, "%s: failed to read file", path); | |
7293fccd LB |
3967 | fclose(input); |
3968 | ||
3969 | sprintf(path, "/sys/devices/system/cpu/cpu%d/cpufreq/scaling_governor", | |
3970 | base_cpu); | |
3971 | input = fopen(path, "r"); | |
3972 | if (input == NULL) { | |
0a42d235 | 3973 | fprintf(outf, "NSFOD %s\n", path); |
7293fccd LB |
3974 | return; |
3975 | } | |
8173c336 BH |
3976 | if (!fgets(governor_buf, sizeof(governor_buf), input)) |
3977 | err(1, "%s: failed to read file", path); | |
7293fccd LB |
3978 | fclose(input); |
3979 | ||
3980 | fprintf(outf, "cpu%d: cpufreq driver: %s", base_cpu, driver_buf); | |
3981 | fprintf(outf, "cpu%d: cpufreq governor: %s", base_cpu, governor_buf); | |
3982 | ||
3983 | sprintf(path, "/sys/devices/system/cpu/cpufreq/boost"); | |
3984 | input = fopen(path, "r"); | |
3985 | if (input != NULL) { | |
8173c336 BH |
3986 | if (fscanf(input, "%d", &turbo) != 1) |
3987 | err(1, "%s: failed to parse number from file", path); | |
7293fccd LB |
3988 | fprintf(outf, "cpufreq boost: %d\n", turbo); |
3989 | fclose(input); | |
3990 | } | |
3991 | ||
3992 | sprintf(path, "/sys/devices/system/cpu/intel_pstate/no_turbo"); | |
3993 | input = fopen(path, "r"); | |
3994 | if (input != NULL) { | |
8173c336 BH |
3995 | if (fscanf(input, "%d", &turbo) != 1) |
3996 | err(1, "%s: failed to parse number from file", path); | |
7293fccd LB |
3997 | fprintf(outf, "cpufreq intel_pstate no_turbo: %d\n", turbo); |
3998 | fclose(input); | |
3999 | } | |
4000 | } | |
41618e63 | 4001 | |
fcd17211 | 4002 | |
889facbe LB |
4003 | /* |
4004 | * print_epb() | |
4005 | * Decode the ENERGY_PERF_BIAS MSR | |
4006 | */ | |
4007 | int print_epb(struct thread_data *t, struct core_data *c, struct pkg_data *p) | |
4008 | { | |
889facbe | 4009 | char *epb_string; |
6d6501d9 | 4010 | int cpu, epb; |
889facbe LB |
4011 | |
4012 | if (!has_epb) | |
4013 | return 0; | |
4014 | ||
4015 | cpu = t->cpu_id; | |
4016 | ||
4017 | /* EPB is per-package */ | |
4018 | if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE) || !(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE)) | |
4019 | return 0; | |
4020 | ||
4021 | if (cpu_migrate(cpu)) { | |
3d7772ea | 4022 | fprintf(outf, "print_epb: Could not migrate to CPU %d\n", cpu); |
889facbe LB |
4023 | return -1; |
4024 | } | |
4025 | ||
6d6501d9 BP |
4026 | epb = get_epb(cpu); |
4027 | if (epb < 0) | |
889facbe LB |
4028 | return 0; |
4029 | ||
6d6501d9 | 4030 | switch (epb) { |
889facbe LB |
4031 | case ENERGY_PERF_BIAS_PERFORMANCE: |
4032 | epb_string = "performance"; | |
4033 | break; | |
4034 | case ENERGY_PERF_BIAS_NORMAL: | |
4035 | epb_string = "balanced"; | |
4036 | break; | |
4037 | case ENERGY_PERF_BIAS_POWERSAVE: | |
4038 | epb_string = "powersave"; | |
4039 | break; | |
4040 | default: | |
4041 | epb_string = "custom"; | |
4042 | break; | |
4043 | } | |
6d6501d9 | 4044 | fprintf(outf, "cpu%d: EPB: %d (%s)\n", cpu, epb, epb_string); |
889facbe LB |
4045 | |
4046 | return 0; | |
4047 | } | |
7f5c258e LB |
4048 | /* |
4049 | * print_hwp() | |
4050 | * Decode the MSR_HWP_CAPABILITIES | |
4051 | */ | |
4052 | int print_hwp(struct thread_data *t, struct core_data *c, struct pkg_data *p) | |
4053 | { | |
4054 | unsigned long long msr; | |
4055 | int cpu; | |
4056 | ||
4057 | if (!has_hwp) | |
4058 | return 0; | |
4059 | ||
4060 | cpu = t->cpu_id; | |
4061 | ||
4062 | /* MSR_HWP_CAPABILITIES is per-package */ | |
4063 | if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE) || !(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE)) | |
4064 | return 0; | |
4065 | ||
4066 | if (cpu_migrate(cpu)) { | |
3d7772ea | 4067 | fprintf(outf, "print_hwp: Could not migrate to CPU %d\n", cpu); |
7f5c258e LB |
4068 | return -1; |
4069 | } | |
4070 | ||
4071 | if (get_msr(cpu, MSR_PM_ENABLE, &msr)) | |
4072 | return 0; | |
4073 | ||
b7d8c148 | 4074 | fprintf(outf, "cpu%d: MSR_PM_ENABLE: 0x%08llx (%sHWP)\n", |
7f5c258e LB |
4075 | cpu, msr, (msr & (1 << 0)) ? "" : "No-"); |
4076 | ||
4077 | /* MSR_PM_ENABLE[1] == 1 if HWP is enabled and MSRs visible */ | |
4078 | if ((msr & (1 << 0)) == 0) | |
4079 | return 0; | |
4080 | ||
4081 | if (get_msr(cpu, MSR_HWP_CAPABILITIES, &msr)) | |
4082 | return 0; | |
4083 | ||
b7d8c148 | 4084 | fprintf(outf, "cpu%d: MSR_HWP_CAPABILITIES: 0x%08llx " |
6dbd25a2 | 4085 | "(high %d guar %d eff %d low %d)\n", |
7f5c258e LB |
4086 | cpu, msr, |
4087 | (unsigned int)HWP_HIGHEST_PERF(msr), | |
4088 | (unsigned int)HWP_GUARANTEED_PERF(msr), | |
4089 | (unsigned int)HWP_MOSTEFFICIENT_PERF(msr), | |
4090 | (unsigned int)HWP_LOWEST_PERF(msr)); | |
4091 | ||
4092 | if (get_msr(cpu, MSR_HWP_REQUEST, &msr)) | |
4093 | return 0; | |
4094 | ||
b7d8c148 | 4095 | fprintf(outf, "cpu%d: MSR_HWP_REQUEST: 0x%08llx " |
6dbd25a2 | 4096 | "(min %d max %d des %d epp 0x%x window 0x%x pkg 0x%x)\n", |
7f5c258e LB |
4097 | cpu, msr, |
4098 | (unsigned int)(((msr) >> 0) & 0xff), | |
4099 | (unsigned int)(((msr) >> 8) & 0xff), | |
4100 | (unsigned int)(((msr) >> 16) & 0xff), | |
4101 | (unsigned int)(((msr) >> 24) & 0xff), | |
4102 | (unsigned int)(((msr) >> 32) & 0xff3), | |
4103 | (unsigned int)(((msr) >> 42) & 0x1)); | |
4104 | ||
4105 | if (has_hwp_pkg) { | |
4106 | if (get_msr(cpu, MSR_HWP_REQUEST_PKG, &msr)) | |
4107 | return 0; | |
4108 | ||
b7d8c148 | 4109 | fprintf(outf, "cpu%d: MSR_HWP_REQUEST_PKG: 0x%08llx " |
6dbd25a2 | 4110 | "(min %d max %d des %d epp 0x%x window 0x%x)\n", |
7f5c258e LB |
4111 | cpu, msr, |
4112 | (unsigned int)(((msr) >> 0) & 0xff), | |
4113 | (unsigned int)(((msr) >> 8) & 0xff), | |
4114 | (unsigned int)(((msr) >> 16) & 0xff), | |
4115 | (unsigned int)(((msr) >> 24) & 0xff), | |
4116 | (unsigned int)(((msr) >> 32) & 0xff3)); | |
4117 | } | |
4118 | if (has_hwp_notify) { | |
4119 | if (get_msr(cpu, MSR_HWP_INTERRUPT, &msr)) | |
4120 | return 0; | |
4121 | ||
b7d8c148 | 4122 | fprintf(outf, "cpu%d: MSR_HWP_INTERRUPT: 0x%08llx " |
7f5c258e LB |
4123 | "(%s_Guaranteed_Perf_Change, %s_Excursion_Min)\n", |
4124 | cpu, msr, | |
4125 | ((msr) & 0x1) ? "EN" : "Dis", | |
4126 | ((msr) & 0x2) ? "EN" : "Dis"); | |
4127 | } | |
4128 | if (get_msr(cpu, MSR_HWP_STATUS, &msr)) | |
4129 | return 0; | |
4130 | ||
b7d8c148 | 4131 | fprintf(outf, "cpu%d: MSR_HWP_STATUS: 0x%08llx " |
7f5c258e LB |
4132 | "(%sGuaranteed_Perf_Change, %sExcursion_Min)\n", |
4133 | cpu, msr, | |
4134 | ((msr) & 0x1) ? "" : "No-", | |
4135 | ((msr) & 0x2) ? "" : "No-"); | |
889facbe LB |
4136 | |
4137 | return 0; | |
4138 | } | |
4139 | ||
3a9a941d LB |
4140 | /* |
4141 | * print_perf_limit() | |
4142 | */ | |
4143 | int print_perf_limit(struct thread_data *t, struct core_data *c, struct pkg_data *p) | |
4144 | { | |
4145 | unsigned long long msr; | |
4146 | int cpu; | |
4147 | ||
4148 | cpu = t->cpu_id; | |
4149 | ||
4150 | /* per-package */ | |
4151 | if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE) || !(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE)) | |
4152 | return 0; | |
4153 | ||
4154 | if (cpu_migrate(cpu)) { | |
3d7772ea | 4155 | fprintf(outf, "print_perf_limit: Could not migrate to CPU %d\n", cpu); |
3a9a941d LB |
4156 | return -1; |
4157 | } | |
4158 | ||
4159 | if (do_core_perf_limit_reasons) { | |
4160 | get_msr(cpu, MSR_CORE_PERF_LIMIT_REASONS, &msr); | |
b7d8c148 LB |
4161 | fprintf(outf, "cpu%d: MSR_CORE_PERF_LIMIT_REASONS, 0x%08llx", cpu, msr); |
4162 | fprintf(outf, " (Active: %s%s%s%s%s%s%s%s%s%s%s%s%s%s)", | |
e33cbe85 | 4163 | (msr & 1 << 15) ? "bit15, " : "", |
3a9a941d | 4164 | (msr & 1 << 14) ? "bit14, " : "", |
e33cbe85 LB |
4165 | (msr & 1 << 13) ? "Transitions, " : "", |
4166 | (msr & 1 << 12) ? "MultiCoreTurbo, " : "", | |
4167 | (msr & 1 << 11) ? "PkgPwrL2, " : "", | |
4168 | (msr & 1 << 10) ? "PkgPwrL1, " : "", | |
4169 | (msr & 1 << 9) ? "CorePwr, " : "", | |
4170 | (msr & 1 << 8) ? "Amps, " : "", | |
4171 | (msr & 1 << 6) ? "VR-Therm, " : "", | |
4172 | (msr & 1 << 5) ? "Auto-HWP, " : "", | |
4173 | (msr & 1 << 4) ? "Graphics, " : "", | |
4174 | (msr & 1 << 2) ? "bit2, " : "", | |
4175 | (msr & 1 << 1) ? "ThermStatus, " : "", | |
4176 | (msr & 1 << 0) ? "PROCHOT, " : ""); | |
b7d8c148 | 4177 | fprintf(outf, " (Logged: %s%s%s%s%s%s%s%s%s%s%s%s%s%s)\n", |
e33cbe85 | 4178 | (msr & 1 << 31) ? "bit31, " : "", |
3a9a941d | 4179 | (msr & 1 << 30) ? "bit30, " : "", |
e33cbe85 LB |
4180 | (msr & 1 << 29) ? "Transitions, " : "", |
4181 | (msr & 1 << 28) ? "MultiCoreTurbo, " : "", | |
4182 | (msr & 1 << 27) ? "PkgPwrL2, " : "", | |
4183 | (msr & 1 << 26) ? "PkgPwrL1, " : "", | |
4184 | (msr & 1 << 25) ? "CorePwr, " : "", | |
4185 | (msr & 1 << 24) ? "Amps, " : "", | |
4186 | (msr & 1 << 22) ? "VR-Therm, " : "", | |
4187 | (msr & 1 << 21) ? "Auto-HWP, " : "", | |
4188 | (msr & 1 << 20) ? "Graphics, " : "", | |
4189 | (msr & 1 << 18) ? "bit18, " : "", | |
4190 | (msr & 1 << 17) ? "ThermStatus, " : "", | |
4191 | (msr & 1 << 16) ? "PROCHOT, " : ""); | |
3a9a941d LB |
4192 | |
4193 | } | |
4194 | if (do_gfx_perf_limit_reasons) { | |
4195 | get_msr(cpu, MSR_GFX_PERF_LIMIT_REASONS, &msr); | |
b7d8c148 LB |
4196 | fprintf(outf, "cpu%d: MSR_GFX_PERF_LIMIT_REASONS, 0x%08llx", cpu, msr); |
4197 | fprintf(outf, " (Active: %s%s%s%s%s%s%s%s)", | |
3a9a941d LB |
4198 | (msr & 1 << 0) ? "PROCHOT, " : "", |
4199 | (msr & 1 << 1) ? "ThermStatus, " : "", | |
4200 | (msr & 1 << 4) ? "Graphics, " : "", | |
4201 | (msr & 1 << 6) ? "VR-Therm, " : "", | |
4202 | (msr & 1 << 8) ? "Amps, " : "", | |
4203 | (msr & 1 << 9) ? "GFXPwr, " : "", | |
4204 | (msr & 1 << 10) ? "PkgPwrL1, " : "", | |
4205 | (msr & 1 << 11) ? "PkgPwrL2, " : ""); | |
b7d8c148 | 4206 | fprintf(outf, " (Logged: %s%s%s%s%s%s%s%s)\n", |
3a9a941d LB |
4207 | (msr & 1 << 16) ? "PROCHOT, " : "", |
4208 | (msr & 1 << 17) ? "ThermStatus, " : "", | |
4209 | (msr & 1 << 20) ? "Graphics, " : "", | |
4210 | (msr & 1 << 22) ? "VR-Therm, " : "", | |
4211 | (msr & 1 << 24) ? "Amps, " : "", | |
4212 | (msr & 1 << 25) ? "GFXPwr, " : "", | |
4213 | (msr & 1 << 26) ? "PkgPwrL1, " : "", | |
4214 | (msr & 1 << 27) ? "PkgPwrL2, " : ""); | |
4215 | } | |
4216 | if (do_ring_perf_limit_reasons) { | |
4217 | get_msr(cpu, MSR_RING_PERF_LIMIT_REASONS, &msr); | |
b7d8c148 LB |
4218 | fprintf(outf, "cpu%d: MSR_RING_PERF_LIMIT_REASONS, 0x%08llx", cpu, msr); |
4219 | fprintf(outf, " (Active: %s%s%s%s%s%s)", | |
3a9a941d LB |
4220 | (msr & 1 << 0) ? "PROCHOT, " : "", |
4221 | (msr & 1 << 1) ? "ThermStatus, " : "", | |
4222 | (msr & 1 << 6) ? "VR-Therm, " : "", | |
4223 | (msr & 1 << 8) ? "Amps, " : "", | |
4224 | (msr & 1 << 10) ? "PkgPwrL1, " : "", | |
4225 | (msr & 1 << 11) ? "PkgPwrL2, " : ""); | |
b7d8c148 | 4226 | fprintf(outf, " (Logged: %s%s%s%s%s%s)\n", |
3a9a941d LB |
4227 | (msr & 1 << 16) ? "PROCHOT, " : "", |
4228 | (msr & 1 << 17) ? "ThermStatus, " : "", | |
4229 | (msr & 1 << 22) ? "VR-Therm, " : "", | |
4230 | (msr & 1 << 24) ? "Amps, " : "", | |
4231 | (msr & 1 << 26) ? "PkgPwrL1, " : "", | |
4232 | (msr & 1 << 27) ? "PkgPwrL2, " : ""); | |
4233 | } | |
4234 | return 0; | |
4235 | } | |
4236 | ||
889facbe LB |
4237 | #define RAPL_POWER_GRANULARITY 0x7FFF /* 15 bit power granularity */ |
4238 | #define RAPL_TIME_GRANULARITY 0x3F /* 6 bit time granularity */ | |
4239 | ||
9392bd98 | 4240 | double get_tdp_intel(unsigned int model) |
144b44b1 LB |
4241 | { |
4242 | unsigned long long msr; | |
4243 | ||
4244 | if (do_rapl & RAPL_PKG_POWER_INFO) | |
7ce7d5de | 4245 | if (!get_msr(base_cpu, MSR_PKG_POWER_INFO, &msr)) |
144b44b1 LB |
4246 | return ((msr >> 0) & RAPL_POWER_GRANULARITY) * rapl_power_units; |
4247 | ||
4248 | switch (model) { | |
f2c4db1b | 4249 | case INTEL_FAM6_ATOM_SILVERMONT: |
5ebb34ed | 4250 | case INTEL_FAM6_ATOM_SILVERMONT_D: |
144b44b1 LB |
4251 | return 30.0; |
4252 | default: | |
4253 | return 135.0; | |
4254 | } | |
4255 | } | |
4256 | ||
9392bd98 CW |
4257 | double get_tdp_amd(unsigned int family) |
4258 | { | |
33eb8225 KP |
4259 | /* This is the max stock TDP of HEDT/Server Fam17h+ chips */ |
4260 | return 280.0; | |
9392bd98 CW |
4261 | } |
4262 | ||
40ee8e3b AS |
4263 | /* |
4264 | * rapl_dram_energy_units_probe() | |
4265 | * Energy units are either hard-coded, or come from RAPL Energy Unit MSR. | |
4266 | */ | |
4267 | static double | |
4268 | rapl_dram_energy_units_probe(int model, double rapl_energy_units) | |
4269 | { | |
4270 | /* only called for genuine_intel, family 6 */ | |
4271 | ||
4272 | switch (model) { | |
869ce69e LB |
4273 | case INTEL_FAM6_HASWELL_X: /* HSX */ |
4274 | case INTEL_FAM6_BROADWELL_X: /* BDX */ | |
869ce69e | 4275 | case INTEL_FAM6_XEON_PHI_KNL: /* KNL */ |
40ee8e3b AS |
4276 | return (rapl_dram_energy_units = 15.3 / 1000000); |
4277 | default: | |
4278 | return (rapl_energy_units); | |
4279 | } | |
4280 | } | |
4281 | ||
9392bd98 | 4282 | void rapl_probe_intel(unsigned int family, unsigned int model) |
889facbe LB |
4283 | { |
4284 | unsigned long long msr; | |
144b44b1 | 4285 | unsigned int time_unit; |
889facbe LB |
4286 | double tdp; |
4287 | ||
889facbe LB |
4288 | if (family != 6) |
4289 | return; | |
4290 | ||
4291 | switch (model) { | |
869ce69e LB |
4292 | case INTEL_FAM6_SANDYBRIDGE: |
4293 | case INTEL_FAM6_IVYBRIDGE: | |
c66f78a6 | 4294 | case INTEL_FAM6_HASWELL: /* HSW */ |
77e5517c | 4295 | case INTEL_FAM6_HASWELL_L: /* HSW */ |
5e741407 | 4296 | case INTEL_FAM6_HASWELL_G: /* HSW */ |
c66f78a6 | 4297 | case INTEL_FAM6_BROADWELL: /* BDW */ |
5e741407 | 4298 | case INTEL_FAM6_BROADWELL_G: /* BDW */ |
144b44b1 | 4299 | do_rapl = RAPL_PKG | RAPL_CORES | RAPL_CORE_POLICY | RAPL_GFX | RAPL_PKG_POWER_INFO; |
812db3f7 LB |
4300 | if (rapl_joules) { |
4301 | BIC_PRESENT(BIC_Pkg_J); | |
4302 | BIC_PRESENT(BIC_Cor_J); | |
4303 | BIC_PRESENT(BIC_GFX_J); | |
4304 | } else { | |
4305 | BIC_PRESENT(BIC_PkgWatt); | |
4306 | BIC_PRESENT(BIC_CorWatt); | |
4307 | BIC_PRESENT(BIC_GFXWatt); | |
4308 | } | |
889facbe | 4309 | break; |
869ce69e | 4310 | case INTEL_FAM6_ATOM_GOLDMONT: /* BXT */ |
f2c4db1b | 4311 | case INTEL_FAM6_ATOM_GOLDMONT_PLUS: |
e4085d54 | 4312 | do_rapl = RAPL_PKG | RAPL_PKG_POWER_INFO; |
812db3f7 LB |
4313 | if (rapl_joules) |
4314 | BIC_PRESENT(BIC_Pkg_J); | |
4315 | else | |
4316 | BIC_PRESENT(BIC_PkgWatt); | |
e4085d54 | 4317 | break; |
f6708400 CY |
4318 | case INTEL_FAM6_ATOM_TREMONT: /* EHL */ |
4319 | do_rapl = RAPL_PKG | RAPL_CORES | RAPL_CORE_POLICY | RAPL_DRAM | RAPL_DRAM_PERF_STATUS | RAPL_PKG_PERF_STATUS | RAPL_GFX | RAPL_PKG_POWER_INFO; | |
4320 | if (rapl_joules) { | |
4321 | BIC_PRESENT(BIC_Pkg_J); | |
4322 | BIC_PRESENT(BIC_Cor_J); | |
4323 | BIC_PRESENT(BIC_RAM_J); | |
4324 | BIC_PRESENT(BIC_GFX_J); | |
4325 | } else { | |
4326 | BIC_PRESENT(BIC_PkgWatt); | |
4327 | BIC_PRESENT(BIC_CorWatt); | |
4328 | BIC_PRESENT(BIC_RAMWatt); | |
4329 | BIC_PRESENT(BIC_GFXWatt); | |
4330 | } | |
4331 | break; | |
20de0dab AL |
4332 | case INTEL_FAM6_ATOM_TREMONT_D: /* JVL */ |
4333 | do_rapl = RAPL_PKG | RAPL_PKG_PERF_STATUS | RAPL_PKG_POWER_INFO; | |
4334 | BIC_PRESENT(BIC_PKG__); | |
4335 | if (rapl_joules) | |
4336 | BIC_PRESENT(BIC_Pkg_J); | |
4337 | else | |
4338 | BIC_PRESENT(BIC_PkgWatt); | |
4339 | break; | |
af239c44 PZ |
4340 | case INTEL_FAM6_SKYLAKE_L: /* SKL */ |
4341 | case INTEL_FAM6_CANNONLAKE_L: /* CNL */ | |
81824921 | 4342 | do_rapl = RAPL_PKG | RAPL_CORES | RAPL_CORE_POLICY | RAPL_DRAM | RAPL_DRAM_PERF_STATUS | RAPL_PKG_PERF_STATUS | RAPL_GFX | RAPL_PKG_POWER_INFO; |
812db3f7 LB |
4343 | BIC_PRESENT(BIC_PKG__); |
4344 | BIC_PRESENT(BIC_RAM__); | |
4345 | if (rapl_joules) { | |
4346 | BIC_PRESENT(BIC_Pkg_J); | |
4347 | BIC_PRESENT(BIC_Cor_J); | |
4348 | BIC_PRESENT(BIC_RAM_J); | |
81824921 | 4349 | BIC_PRESENT(BIC_GFX_J); |
812db3f7 LB |
4350 | } else { |
4351 | BIC_PRESENT(BIC_PkgWatt); | |
4352 | BIC_PRESENT(BIC_CorWatt); | |
4353 | BIC_PRESENT(BIC_RAMWatt); | |
81824921 | 4354 | BIC_PRESENT(BIC_GFXWatt); |
812db3f7 | 4355 | } |
0b2bb692 | 4356 | break; |
869ce69e LB |
4357 | case INTEL_FAM6_HASWELL_X: /* HSX */ |
4358 | case INTEL_FAM6_BROADWELL_X: /* BDX */ | |
869ce69e LB |
4359 | case INTEL_FAM6_SKYLAKE_X: /* SKX */ |
4360 | case INTEL_FAM6_XEON_PHI_KNL: /* KNL */ | |
0b2bb692 | 4361 | do_rapl = RAPL_PKG | RAPL_DRAM | RAPL_DRAM_POWER_INFO | RAPL_DRAM_PERF_STATUS | RAPL_PKG_PERF_STATUS | RAPL_PKG_POWER_INFO; |
812db3f7 LB |
4362 | BIC_PRESENT(BIC_PKG__); |
4363 | BIC_PRESENT(BIC_RAM__); | |
4364 | if (rapl_joules) { | |
4365 | BIC_PRESENT(BIC_Pkg_J); | |
4366 | BIC_PRESENT(BIC_RAM_J); | |
4367 | } else { | |
4368 | BIC_PRESENT(BIC_PkgWatt); | |
4369 | BIC_PRESENT(BIC_RAMWatt); | |
4370 | } | |
e6f9bb3c | 4371 | break; |
869ce69e LB |
4372 | case INTEL_FAM6_SANDYBRIDGE_X: |
4373 | case INTEL_FAM6_IVYBRIDGE_X: | |
0b2bb692 | 4374 | do_rapl = RAPL_PKG | RAPL_CORES | RAPL_CORE_POLICY | RAPL_DRAM | RAPL_DRAM_POWER_INFO | RAPL_PKG_PERF_STATUS | RAPL_DRAM_PERF_STATUS | RAPL_PKG_POWER_INFO; |
812db3f7 LB |
4375 | BIC_PRESENT(BIC_PKG__); |
4376 | BIC_PRESENT(BIC_RAM__); | |
4377 | if (rapl_joules) { | |
4378 | BIC_PRESENT(BIC_Pkg_J); | |
4379 | BIC_PRESENT(BIC_Cor_J); | |
4380 | BIC_PRESENT(BIC_RAM_J); | |
4381 | } else { | |
4382 | BIC_PRESENT(BIC_PkgWatt); | |
4383 | BIC_PRESENT(BIC_CorWatt); | |
4384 | BIC_PRESENT(BIC_RAMWatt); | |
4385 | } | |
144b44b1 | 4386 | break; |
f2c4db1b | 4387 | case INTEL_FAM6_ATOM_SILVERMONT: /* BYT */ |
5ebb34ed | 4388 | case INTEL_FAM6_ATOM_SILVERMONT_D: /* AVN */ |
9148494c | 4389 | do_rapl = RAPL_PKG | RAPL_CORES; |
812db3f7 LB |
4390 | if (rapl_joules) { |
4391 | BIC_PRESENT(BIC_Pkg_J); | |
4392 | BIC_PRESENT(BIC_Cor_J); | |
4393 | } else { | |
4394 | BIC_PRESENT(BIC_PkgWatt); | |
4395 | BIC_PRESENT(BIC_CorWatt); | |
4396 | } | |
889facbe | 4397 | break; |
5ebb34ed | 4398 | case INTEL_FAM6_ATOM_GOLDMONT_D: /* DNV */ |
0f644909 | 4399 | do_rapl = RAPL_PKG | RAPL_DRAM | RAPL_DRAM_POWER_INFO | RAPL_DRAM_PERF_STATUS | RAPL_PKG_PERF_STATUS | RAPL_PKG_POWER_INFO | RAPL_CORES_ENERGY_STATUS; |
812db3f7 LB |
4400 | BIC_PRESENT(BIC_PKG__); |
4401 | BIC_PRESENT(BIC_RAM__); | |
4402 | if (rapl_joules) { | |
4403 | BIC_PRESENT(BIC_Pkg_J); | |
4404 | BIC_PRESENT(BIC_Cor_J); | |
4405 | BIC_PRESENT(BIC_RAM_J); | |
4406 | } else { | |
4407 | BIC_PRESENT(BIC_PkgWatt); | |
4408 | BIC_PRESENT(BIC_CorWatt); | |
4409 | BIC_PRESENT(BIC_RAMWatt); | |
4410 | } | |
0f644909 | 4411 | break; |
889facbe LB |
4412 | default: |
4413 | return; | |
4414 | } | |
4415 | ||
4416 | /* units on package 0, verify later other packages match */ | |
7ce7d5de | 4417 | if (get_msr(base_cpu, MSR_RAPL_POWER_UNIT, &msr)) |
889facbe LB |
4418 | return; |
4419 | ||
4420 | rapl_power_units = 1.0 / (1 << (msr & 0xF)); | |
f2c4db1b | 4421 | if (model == INTEL_FAM6_ATOM_SILVERMONT) |
144b44b1 LB |
4422 | rapl_energy_units = 1.0 * (1 << (msr >> 8 & 0x1F)) / 1000000; |
4423 | else | |
4424 | rapl_energy_units = 1.0 / (1 << (msr >> 8 & 0x1F)); | |
889facbe | 4425 | |
40ee8e3b AS |
4426 | rapl_dram_energy_units = rapl_dram_energy_units_probe(model, rapl_energy_units); |
4427 | ||
144b44b1 LB |
4428 | time_unit = msr >> 16 & 0xF; |
4429 | if (time_unit == 0) | |
4430 | time_unit = 0xA; | |
889facbe | 4431 | |
144b44b1 | 4432 | rapl_time_units = 1.0 / (1 << (time_unit)); |
889facbe | 4433 | |
9392bd98 | 4434 | tdp = get_tdp_intel(model); |
889facbe | 4435 | |
144b44b1 | 4436 | rapl_joule_counter_range = 0xFFFFFFFF * rapl_energy_units / tdp; |
96e47158 | 4437 | if (!quiet) |
b7d8c148 | 4438 | fprintf(outf, "RAPL: %.0f sec. Joule Counter Range, at %.0f Watts\n", rapl_joule_counter_range, tdp); |
9392bd98 | 4439 | } |
889facbe | 4440 | |
9392bd98 CW |
4441 | void rapl_probe_amd(unsigned int family, unsigned int model) |
4442 | { | |
4443 | unsigned long long msr; | |
4444 | unsigned int eax, ebx, ecx, edx; | |
4445 | unsigned int has_rapl = 0; | |
4446 | double tdp; | |
4447 | ||
4448 | if (max_extended_level >= 0x80000007) { | |
4449 | __cpuid(0x80000007, eax, ebx, ecx, edx); | |
33eb8225 | 4450 | /* RAPL (Fam 17h+) */ |
9392bd98 CW |
4451 | has_rapl = edx & (1 << 14); |
4452 | } | |
4453 | ||
33eb8225 | 4454 | if (!has_rapl || family < 0x17) |
9392bd98 CW |
4455 | return; |
4456 | ||
33eb8225 KP |
4457 | do_rapl = RAPL_AMD_F17H | RAPL_PER_CORE_ENERGY; |
4458 | if (rapl_joules) { | |
4459 | BIC_PRESENT(BIC_Pkg_J); | |
4460 | BIC_PRESENT(BIC_Cor_J); | |
4461 | } else { | |
4462 | BIC_PRESENT(BIC_PkgWatt); | |
4463 | BIC_PRESENT(BIC_CorWatt); | |
9392bd98 CW |
4464 | } |
4465 | ||
4466 | if (get_msr(base_cpu, MSR_RAPL_PWR_UNIT, &msr)) | |
4467 | return; | |
4468 | ||
4469 | rapl_time_units = ldexp(1.0, -(msr >> 16 & 0xf)); | |
4470 | rapl_energy_units = ldexp(1.0, -(msr >> 8 & 0x1f)); | |
4471 | rapl_power_units = ldexp(1.0, -(msr & 0xf)); | |
4472 | ||
9cfa8e04 | 4473 | tdp = get_tdp_amd(family); |
9392bd98 CW |
4474 | |
4475 | rapl_joule_counter_range = 0xFFFFFFFF * rapl_energy_units / tdp; | |
4476 | if (!quiet) | |
4477 | fprintf(outf, "RAPL: %.0f sec. Joule Counter Range, at %.0f Watts\n", rapl_joule_counter_range, tdp); | |
4478 | } | |
4479 | ||
4480 | /* | |
4481 | * rapl_probe() | |
4482 | * | |
4483 | * sets do_rapl, rapl_power_units, rapl_energy_units, rapl_time_units | |
4484 | */ | |
4485 | void rapl_probe(unsigned int family, unsigned int model) | |
4486 | { | |
4487 | if (genuine_intel) | |
4488 | rapl_probe_intel(family, model); | |
c1c10cc7 | 4489 | if (authentic_amd || hygon_genuine) |
9392bd98 | 4490 | rapl_probe_amd(family, model); |
889facbe LB |
4491 | } |
4492 | ||
1b69317d | 4493 | void perf_limit_reasons_probe(unsigned int family, unsigned int model) |
3a9a941d LB |
4494 | { |
4495 | if (!genuine_intel) | |
4496 | return; | |
4497 | ||
4498 | if (family != 6) | |
4499 | return; | |
4500 | ||
4501 | switch (model) { | |
c66f78a6 | 4502 | case INTEL_FAM6_HASWELL: /* HSW */ |
77e5517c | 4503 | case INTEL_FAM6_HASWELL_L: /* HSW */ |
5e741407 | 4504 | case INTEL_FAM6_HASWELL_G: /* HSW */ |
3a9a941d | 4505 | do_gfx_perf_limit_reasons = 1; |
869ce69e | 4506 | case INTEL_FAM6_HASWELL_X: /* HSX */ |
3a9a941d LB |
4507 | do_core_perf_limit_reasons = 1; |
4508 | do_ring_perf_limit_reasons = 1; | |
4509 | default: | |
4510 | return; | |
4511 | } | |
4512 | } | |
4513 | ||
ac980e13 AB |
4514 | void automatic_cstate_conversion_probe(unsigned int family, unsigned int model) |
4515 | { | |
4516 | if (is_skx(family, model) || is_bdx(family, model)) | |
4517 | has_automatic_cstate_conversion = 1; | |
4518 | } | |
4519 | ||
889facbe LB |
4520 | int print_thermal(struct thread_data *t, struct core_data *c, struct pkg_data *p) |
4521 | { | |
4522 | unsigned long long msr; | |
f4896fa5 | 4523 | unsigned int dts, dts2; |
889facbe LB |
4524 | int cpu; |
4525 | ||
4526 | if (!(do_dts || do_ptm)) | |
4527 | return 0; | |
4528 | ||
4529 | cpu = t->cpu_id; | |
4530 | ||
4531 | /* DTS is per-core, no need to print for each thread */ | |
388e9c81 | 4532 | if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE)) |
889facbe LB |
4533 | return 0; |
4534 | ||
4535 | if (cpu_migrate(cpu)) { | |
3d7772ea | 4536 | fprintf(outf, "print_thermal: Could not migrate to CPU %d\n", cpu); |
889facbe LB |
4537 | return -1; |
4538 | } | |
4539 | ||
4540 | if (do_ptm && (t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE)) { | |
4541 | if (get_msr(cpu, MSR_IA32_PACKAGE_THERM_STATUS, &msr)) | |
4542 | return 0; | |
4543 | ||
4544 | dts = (msr >> 16) & 0x7F; | |
b7d8c148 | 4545 | fprintf(outf, "cpu%d: MSR_IA32_PACKAGE_THERM_STATUS: 0x%08llx (%d C)\n", |
889facbe LB |
4546 | cpu, msr, tcc_activation_temp - dts); |
4547 | ||
889facbe LB |
4548 | if (get_msr(cpu, MSR_IA32_PACKAGE_THERM_INTERRUPT, &msr)) |
4549 | return 0; | |
4550 | ||
4551 | dts = (msr >> 16) & 0x7F; | |
4552 | dts2 = (msr >> 8) & 0x7F; | |
b7d8c148 | 4553 | fprintf(outf, "cpu%d: MSR_IA32_PACKAGE_THERM_INTERRUPT: 0x%08llx (%d C, %d C)\n", |
889facbe | 4554 | cpu, msr, tcc_activation_temp - dts, tcc_activation_temp - dts2); |
889facbe LB |
4555 | } |
4556 | ||
4557 | ||
f4896fa5 | 4558 | if (do_dts && debug) { |
889facbe LB |
4559 | unsigned int resolution; |
4560 | ||
4561 | if (get_msr(cpu, MSR_IA32_THERM_STATUS, &msr)) | |
4562 | return 0; | |
4563 | ||
4564 | dts = (msr >> 16) & 0x7F; | |
4565 | resolution = (msr >> 27) & 0xF; | |
b7d8c148 | 4566 | fprintf(outf, "cpu%d: MSR_IA32_THERM_STATUS: 0x%08llx (%d C +/- %d)\n", |
889facbe LB |
4567 | cpu, msr, tcc_activation_temp - dts, resolution); |
4568 | ||
889facbe LB |
4569 | if (get_msr(cpu, MSR_IA32_THERM_INTERRUPT, &msr)) |
4570 | return 0; | |
4571 | ||
4572 | dts = (msr >> 16) & 0x7F; | |
4573 | dts2 = (msr >> 8) & 0x7F; | |
b7d8c148 | 4574 | fprintf(outf, "cpu%d: MSR_IA32_THERM_INTERRUPT: 0x%08llx (%d C, %d C)\n", |
889facbe | 4575 | cpu, msr, tcc_activation_temp - dts, tcc_activation_temp - dts2); |
889facbe LB |
4576 | } |
4577 | ||
4578 | return 0; | |
4579 | } | |
36229897 | 4580 | |
889facbe LB |
4581 | void print_power_limit_msr(int cpu, unsigned long long msr, char *label) |
4582 | { | |
b7d8c148 | 4583 | fprintf(outf, "cpu%d: %s: %sabled (%f Watts, %f sec, clamp %sabled)\n", |
889facbe LB |
4584 | cpu, label, |
4585 | ((msr >> 15) & 1) ? "EN" : "DIS", | |
4586 | ((msr >> 0) & 0x7FFF) * rapl_power_units, | |
4587 | (1.0 + (((msr >> 22) & 0x3)/4.0)) * (1 << ((msr >> 17) & 0x1F)) * rapl_time_units, | |
4588 | (((msr >> 16) & 1) ? "EN" : "DIS")); | |
4589 | ||
4590 | return; | |
4591 | } | |
4592 | ||
4593 | int print_rapl(struct thread_data *t, struct core_data *c, struct pkg_data *p) | |
4594 | { | |
4595 | unsigned long long msr; | |
9392bd98 | 4596 | const char *msr_name; |
889facbe | 4597 | int cpu; |
889facbe LB |
4598 | |
4599 | if (!do_rapl) | |
4600 | return 0; | |
4601 | ||
4602 | /* RAPL counters are per package, so print only for 1st thread/package */ | |
4603 | if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE) || !(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE)) | |
4604 | return 0; | |
4605 | ||
4606 | cpu = t->cpu_id; | |
4607 | if (cpu_migrate(cpu)) { | |
3d7772ea | 4608 | fprintf(outf, "print_rapl: Could not migrate to CPU %d\n", cpu); |
889facbe LB |
4609 | return -1; |
4610 | } | |
4611 | ||
9392bd98 CW |
4612 | if (do_rapl & RAPL_AMD_F17H) { |
4613 | msr_name = "MSR_RAPL_PWR_UNIT"; | |
4614 | if (get_msr(cpu, MSR_RAPL_PWR_UNIT, &msr)) | |
4615 | return -1; | |
4616 | } else { | |
4617 | msr_name = "MSR_RAPL_POWER_UNIT"; | |
4618 | if (get_msr(cpu, MSR_RAPL_POWER_UNIT, &msr)) | |
4619 | return -1; | |
4620 | } | |
889facbe | 4621 | |
9392bd98 | 4622 | fprintf(outf, "cpu%d: %s: 0x%08llx (%f Watts, %f Joules, %f sec.)\n", cpu, msr_name, msr, |
96e47158 LB |
4623 | rapl_power_units, rapl_energy_units, rapl_time_units); |
4624 | ||
144b44b1 LB |
4625 | if (do_rapl & RAPL_PKG_POWER_INFO) { |
4626 | ||
889facbe LB |
4627 | if (get_msr(cpu, MSR_PKG_POWER_INFO, &msr)) |
4628 | return -5; | |
4629 | ||
4630 | ||
b7d8c148 | 4631 | fprintf(outf, "cpu%d: MSR_PKG_POWER_INFO: 0x%08llx (%.0f W TDP, RAPL %.0f - %.0f W, %f sec.)\n", |
889facbe LB |
4632 | cpu, msr, |
4633 | ((msr >> 0) & RAPL_POWER_GRANULARITY) * rapl_power_units, | |
4634 | ((msr >> 16) & RAPL_POWER_GRANULARITY) * rapl_power_units, | |
4635 | ((msr >> 32) & RAPL_POWER_GRANULARITY) * rapl_power_units, | |
4636 | ((msr >> 48) & RAPL_TIME_GRANULARITY) * rapl_time_units); | |
4637 | ||
144b44b1 LB |
4638 | } |
4639 | if (do_rapl & RAPL_PKG) { | |
4640 | ||
889facbe LB |
4641 | if (get_msr(cpu, MSR_PKG_POWER_LIMIT, &msr)) |
4642 | return -9; | |
4643 | ||
b7d8c148 | 4644 | fprintf(outf, "cpu%d: MSR_PKG_POWER_LIMIT: 0x%08llx (%slocked)\n", |
96e47158 | 4645 | cpu, msr, (msr >> 63) & 1 ? "" : "UN"); |
889facbe LB |
4646 | |
4647 | print_power_limit_msr(cpu, msr, "PKG Limit #1"); | |
b7d8c148 | 4648 | fprintf(outf, "cpu%d: PKG Limit #2: %sabled (%f Watts, %f* sec, clamp %sabled)\n", |
889facbe LB |
4649 | cpu, |
4650 | ((msr >> 47) & 1) ? "EN" : "DIS", | |
4651 | ((msr >> 32) & 0x7FFF) * rapl_power_units, | |
4652 | (1.0 + (((msr >> 54) & 0x3)/4.0)) * (1 << ((msr >> 49) & 0x1F)) * rapl_time_units, | |
4653 | ((msr >> 48) & 1) ? "EN" : "DIS"); | |
4654 | } | |
4655 | ||
0b2bb692 | 4656 | if (do_rapl & RAPL_DRAM_POWER_INFO) { |
889facbe LB |
4657 | if (get_msr(cpu, MSR_DRAM_POWER_INFO, &msr)) |
4658 | return -6; | |
4659 | ||
b7d8c148 | 4660 | fprintf(outf, "cpu%d: MSR_DRAM_POWER_INFO,: 0x%08llx (%.0f W TDP, RAPL %.0f - %.0f W, %f sec.)\n", |
889facbe LB |
4661 | cpu, msr, |
4662 | ((msr >> 0) & RAPL_POWER_GRANULARITY) * rapl_power_units, | |
4663 | ((msr >> 16) & RAPL_POWER_GRANULARITY) * rapl_power_units, | |
4664 | ((msr >> 32) & RAPL_POWER_GRANULARITY) * rapl_power_units, | |
4665 | ((msr >> 48) & RAPL_TIME_GRANULARITY) * rapl_time_units); | |
0b2bb692 LB |
4666 | } |
4667 | if (do_rapl & RAPL_DRAM) { | |
889facbe LB |
4668 | if (get_msr(cpu, MSR_DRAM_POWER_LIMIT, &msr)) |
4669 | return -9; | |
b7d8c148 | 4670 | fprintf(outf, "cpu%d: MSR_DRAM_POWER_LIMIT: 0x%08llx (%slocked)\n", |
96e47158 | 4671 | cpu, msr, (msr >> 31) & 1 ? "" : "UN"); |
889facbe LB |
4672 | |
4673 | print_power_limit_msr(cpu, msr, "DRAM Limit"); | |
4674 | } | |
144b44b1 | 4675 | if (do_rapl & RAPL_CORE_POLICY) { |
96e47158 LB |
4676 | if (get_msr(cpu, MSR_PP0_POLICY, &msr)) |
4677 | return -7; | |
889facbe | 4678 | |
96e47158 | 4679 | fprintf(outf, "cpu%d: MSR_PP0_POLICY: %lld\n", cpu, msr & 0xF); |
144b44b1 | 4680 | } |
9148494c | 4681 | if (do_rapl & RAPL_CORES_POWER_LIMIT) { |
96e47158 LB |
4682 | if (get_msr(cpu, MSR_PP0_POWER_LIMIT, &msr)) |
4683 | return -9; | |
4684 | fprintf(outf, "cpu%d: MSR_PP0_POWER_LIMIT: 0x%08llx (%slocked)\n", | |
4685 | cpu, msr, (msr >> 31) & 1 ? "" : "UN"); | |
4686 | print_power_limit_msr(cpu, msr, "Cores Limit"); | |
889facbe LB |
4687 | } |
4688 | if (do_rapl & RAPL_GFX) { | |
96e47158 LB |
4689 | if (get_msr(cpu, MSR_PP1_POLICY, &msr)) |
4690 | return -8; | |
889facbe | 4691 | |
96e47158 | 4692 | fprintf(outf, "cpu%d: MSR_PP1_POLICY: %lld\n", cpu, msr & 0xF); |
889facbe | 4693 | |
96e47158 LB |
4694 | if (get_msr(cpu, MSR_PP1_POWER_LIMIT, &msr)) |
4695 | return -9; | |
4696 | fprintf(outf, "cpu%d: MSR_PP1_POWER_LIMIT: 0x%08llx (%slocked)\n", | |
4697 | cpu, msr, (msr >> 31) & 1 ? "" : "UN"); | |
4698 | print_power_limit_msr(cpu, msr, "GFX Limit"); | |
889facbe LB |
4699 | } |
4700 | return 0; | |
4701 | } | |
4702 | ||
d7899447 LB |
4703 | /* |
4704 | * SNB adds support for additional MSRs: | |
4705 | * | |
4706 | * MSR_PKG_C7_RESIDENCY 0x000003fa | |
4707 | * MSR_CORE_C7_RESIDENCY 0x000003fe | |
4708 | * MSR_PKG_C2_RESIDENCY 0x0000060d | |
4709 | */ | |
103a8fea | 4710 | |
d7899447 | 4711 | int has_snb_msrs(unsigned int family, unsigned int model) |
103a8fea LB |
4712 | { |
4713 | if (!genuine_intel) | |
4714 | return 0; | |
4715 | ||
4716 | switch (model) { | |
869ce69e LB |
4717 | case INTEL_FAM6_SANDYBRIDGE: |
4718 | case INTEL_FAM6_SANDYBRIDGE_X: | |
77e5517c IM |
4719 | case INTEL_FAM6_IVYBRIDGE: /* IVB */ |
4720 | case INTEL_FAM6_IVYBRIDGE_X: /* IVB Xeon */ | |
4721 | case INTEL_FAM6_HASWELL: /* HSW */ | |
4722 | case INTEL_FAM6_HASWELL_X: /* HSW */ | |
4723 | case INTEL_FAM6_HASWELL_L: /* HSW */ | |
4724 | case INTEL_FAM6_HASWELL_G: /* HSW */ | |
4725 | case INTEL_FAM6_BROADWELL: /* BDW */ | |
4726 | case INTEL_FAM6_BROADWELL_G: /* BDW */ | |
4727 | case INTEL_FAM6_BROADWELL_X: /* BDX */ | |
4728 | case INTEL_FAM6_SKYLAKE_L: /* SKL */ | |
4729 | case INTEL_FAM6_CANNONLAKE_L: /* CNL */ | |
4730 | case INTEL_FAM6_SKYLAKE_X: /* SKX */ | |
4731 | case INTEL_FAM6_ATOM_GOLDMONT: /* BXT */ | |
f2c4db1b | 4732 | case INTEL_FAM6_ATOM_GOLDMONT_PLUS: |
5ebb34ed | 4733 | case INTEL_FAM6_ATOM_GOLDMONT_D: /* DNV */ |
f6708400 | 4734 | case INTEL_FAM6_ATOM_TREMONT: /* EHL */ |
20de0dab | 4735 | case INTEL_FAM6_ATOM_TREMONT_D: /* JVL */ |
103a8fea LB |
4736 | return 1; |
4737 | } | |
4738 | return 0; | |
4739 | } | |
4740 | ||
d7899447 | 4741 | /* |
570992fc | 4742 | * HSW ULT added support for C8/C9/C10 MSRs: |
d7899447 | 4743 | * |
5a63426e LB |
4744 | * MSR_PKG_C8_RESIDENCY 0x00000630 |
4745 | * MSR_PKG_C9_RESIDENCY 0x00000631 | |
4746 | * MSR_PKG_C10_RESIDENCY 0x00000632 | |
4747 | * | |
4748 | * MSR_PKGC8_IRTL 0x00000633 | |
4749 | * MSR_PKGC9_IRTL 0x00000634 | |
4750 | * MSR_PKGC10_IRTL 0x00000635 | |
4751 | * | |
d7899447 | 4752 | */ |
570992fc | 4753 | int has_c8910_msrs(unsigned int family, unsigned int model) |
ca58710f KCA |
4754 | { |
4755 | if (!genuine_intel) | |
4756 | return 0; | |
4757 | ||
4758 | switch (model) { | |
77e5517c | 4759 | case INTEL_FAM6_HASWELL_L: /* HSW */ |
c66f78a6 | 4760 | case INTEL_FAM6_BROADWELL: /* BDW */ |
af239c44 PZ |
4761 | case INTEL_FAM6_SKYLAKE_L: /* SKL */ |
4762 | case INTEL_FAM6_CANNONLAKE_L: /* CNL */ | |
869ce69e | 4763 | case INTEL_FAM6_ATOM_GOLDMONT: /* BXT */ |
f2c4db1b | 4764 | case INTEL_FAM6_ATOM_GOLDMONT_PLUS: |
f6708400 | 4765 | case INTEL_FAM6_ATOM_TREMONT: /* EHL */ |
0b2bb692 LB |
4766 | return 1; |
4767 | } | |
4768 | return 0; | |
4769 | } | |
4770 | ||
4771 | /* | |
4772 | * SKL adds support for additional MSRS: | |
4773 | * | |
4774 | * MSR_PKG_WEIGHTED_CORE_C0_RES 0x00000658 | |
4775 | * MSR_PKG_ANY_CORE_C0_RES 0x00000659 | |
4776 | * MSR_PKG_ANY_GFXE_C0_RES 0x0000065A | |
4777 | * MSR_PKG_BOTH_CORE_GFXE_C0_RES 0x0000065B | |
4778 | */ | |
4779 | int has_skl_msrs(unsigned int family, unsigned int model) | |
4780 | { | |
4781 | if (!genuine_intel) | |
4782 | return 0; | |
4783 | ||
4784 | switch (model) { | |
af239c44 PZ |
4785 | case INTEL_FAM6_SKYLAKE_L: /* SKL */ |
4786 | case INTEL_FAM6_CANNONLAKE_L: /* CNL */ | |
ca58710f KCA |
4787 | return 1; |
4788 | } | |
4789 | return 0; | |
4790 | } | |
4791 | ||
144b44b1 LB |
4792 | int is_slm(unsigned int family, unsigned int model) |
4793 | { | |
4794 | if (!genuine_intel) | |
4795 | return 0; | |
4796 | switch (model) { | |
f2c4db1b | 4797 | case INTEL_FAM6_ATOM_SILVERMONT: /* BYT */ |
5ebb34ed | 4798 | case INTEL_FAM6_ATOM_SILVERMONT_D: /* AVN */ |
144b44b1 LB |
4799 | return 1; |
4800 | } | |
4801 | return 0; | |
4802 | } | |
4803 | ||
fb5d4327 DC |
4804 | int is_knl(unsigned int family, unsigned int model) |
4805 | { | |
4806 | if (!genuine_intel) | |
4807 | return 0; | |
4808 | switch (model) { | |
869ce69e | 4809 | case INTEL_FAM6_XEON_PHI_KNL: /* KNL */ |
fb5d4327 DC |
4810 | return 1; |
4811 | } | |
4812 | return 0; | |
4813 | } | |
4814 | ||
997e5395 SP |
4815 | int is_cnl(unsigned int family, unsigned int model) |
4816 | { | |
4817 | if (!genuine_intel) | |
4818 | return 0; | |
4819 | ||
4820 | switch (model) { | |
af239c44 | 4821 | case INTEL_FAM6_CANNONLAKE_L: /* CNL */ |
997e5395 SP |
4822 | return 1; |
4823 | } | |
4824 | ||
4825 | return 0; | |
4826 | } | |
4827 | ||
b2b34dfe HC |
4828 | unsigned int get_aperf_mperf_multiplier(unsigned int family, unsigned int model) |
4829 | { | |
4830 | if (is_knl(family, model)) | |
4831 | return 1024; | |
4832 | return 1; | |
4833 | } | |
4834 | ||
144b44b1 LB |
4835 | #define SLM_BCLK_FREQS 5 |
4836 | double slm_freq_table[SLM_BCLK_FREQS] = { 83.3, 100.0, 133.3, 116.7, 80.0}; | |
4837 | ||
4838 | double slm_bclk(void) | |
4839 | { | |
4840 | unsigned long long msr = 3; | |
4841 | unsigned int i; | |
4842 | double freq; | |
4843 | ||
7ce7d5de | 4844 | if (get_msr(base_cpu, MSR_FSB_FREQ, &msr)) |
b7d8c148 | 4845 | fprintf(outf, "SLM BCLK: unknown\n"); |
144b44b1 LB |
4846 | |
4847 | i = msr & 0xf; | |
4848 | if (i >= SLM_BCLK_FREQS) { | |
b7d8c148 | 4849 | fprintf(outf, "SLM BCLK[%d] invalid\n", i); |
0a91e551 | 4850 | i = 3; |
144b44b1 LB |
4851 | } |
4852 | freq = slm_freq_table[i]; | |
4853 | ||
96e47158 | 4854 | if (!quiet) |
8f6196c1 | 4855 | fprintf(outf, "SLM BCLK: %.1f Mhz\n", freq); |
144b44b1 LB |
4856 | |
4857 | return freq; | |
4858 | } | |
4859 | ||
103a8fea LB |
4860 | double discover_bclk(unsigned int family, unsigned int model) |
4861 | { | |
121b48bb | 4862 | if (has_snb_msrs(family, model) || is_knl(family, model)) |
103a8fea | 4863 | return 100.00; |
144b44b1 LB |
4864 | else if (is_slm(family, model)) |
4865 | return slm_bclk(); | |
103a8fea LB |
4866 | else |
4867 | return 133.33; | |
4868 | } | |
4869 | ||
889facbe LB |
4870 | /* |
4871 | * MSR_IA32_TEMPERATURE_TARGET indicates the temperature where | |
4872 | * the Thermal Control Circuit (TCC) activates. | |
4873 | * This is usually equal to tjMax. | |
4874 | * | |
4875 | * Older processors do not have this MSR, so there we guess, | |
4876 | * but also allow cmdline over-ride with -T. | |
4877 | * | |
4878 | * Several MSR temperature values are in units of degrees-C | |
4879 | * below this value, including the Digital Thermal Sensor (DTS), | |
4880 | * Package Thermal Management Sensor (PTM), and thermal event thresholds. | |
4881 | */ | |
6ff7cb37 | 4882 | int read_tcc_activation_temp() |
889facbe LB |
4883 | { |
4884 | unsigned long long msr; | |
6ff7cb37 LB |
4885 | unsigned int tcc, target_c, offset_c; |
4886 | ||
4887 | /* Temperature Target MSR is Nehalem and newer only */ | |
4888 | if (!do_nhm_platform_info) | |
4889 | return 0; | |
4890 | ||
4891 | if (get_msr(base_cpu, MSR_IA32_TEMPERATURE_TARGET, &msr)) | |
4892 | return 0; | |
889facbe | 4893 | |
6ff7cb37 LB |
4894 | target_c = (msr >> 16) & 0xFF; |
4895 | ||
4896 | offset_c = (msr >> 24) & 0xF; | |
4897 | ||
4898 | tcc = target_c - offset_c; | |
4899 | ||
4900 | if (!quiet) | |
4901 | fprintf(outf, "cpu%d: MSR_IA32_TEMPERATURE_TARGET: 0x%08llx (%d C) (%d default - %d offset)\n", | |
4902 | base_cpu, msr, tcc, target_c, offset_c); | |
4903 | ||
4904 | return tcc; | |
4905 | } | |
4906 | ||
4907 | int set_temperature_target(struct thread_data *t, struct core_data *c, struct pkg_data *p) | |
4908 | { | |
889facbe LB |
4909 | /* tcc_activation_temp is used only for dts or ptm */ |
4910 | if (!(do_dts || do_ptm)) | |
4911 | return 0; | |
4912 | ||
4913 | /* this is a per-package concept */ | |
4914 | if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE) || !(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE)) | |
4915 | return 0; | |
4916 | ||
889facbe LB |
4917 | if (tcc_activation_temp_override != 0) { |
4918 | tcc_activation_temp = tcc_activation_temp_override; | |
6ff7cb37 | 4919 | fprintf(outf, "Using cmdline TCC Target (%d C)\n", tcc_activation_temp); |
889facbe LB |
4920 | return 0; |
4921 | } | |
4922 | ||
6ff7cb37 LB |
4923 | tcc_activation_temp = read_tcc_activation_temp(); |
4924 | if (tcc_activation_temp) | |
4925 | return 0; | |
889facbe | 4926 | |
889facbe | 4927 | tcc_activation_temp = TJMAX_DEFAULT; |
6ff7cb37 | 4928 | fprintf(outf, "Guessing tjMax %d C, Please use -T to specify\n", tcc_activation_temp); |
889facbe LB |
4929 | |
4930 | return 0; | |
4931 | } | |
69807a63 | 4932 | |
aa8d8cc7 LB |
4933 | void decode_feature_control_msr(void) |
4934 | { | |
4935 | unsigned long long msr; | |
4936 | ||
f6505c88 | 4937 | if (!get_msr(base_cpu, MSR_IA32_FEAT_CTL, &msr)) |
aa8d8cc7 LB |
4938 | fprintf(outf, "cpu%d: MSR_IA32_FEATURE_CONTROL: 0x%08llx (%sLocked %s)\n", |
4939 | base_cpu, msr, | |
f6505c88 | 4940 | msr & FEAT_CTL_LOCKED ? "" : "UN-", |
aa8d8cc7 LB |
4941 | msr & (1 << 18) ? "SGX" : ""); |
4942 | } | |
4943 | ||
69807a63 LB |
4944 | void decode_misc_enable_msr(void) |
4945 | { | |
4946 | unsigned long long msr; | |
4947 | ||
f26b1519 LB |
4948 | if (!genuine_intel) |
4949 | return; | |
4950 | ||
69807a63 | 4951 | if (!get_msr(base_cpu, MSR_IA32_MISC_ENABLE, &msr)) |
e6512624 | 4952 | fprintf(outf, "cpu%d: MSR_IA32_MISC_ENABLE: 0x%08llx (%sTCC %sEIST %sMWAIT %sPREFETCH %sTURBO)\n", |
69807a63 | 4953 | base_cpu, msr, |
e6512624 LB |
4954 | msr & MSR_IA32_MISC_ENABLE_TM1 ? "" : "No-", |
4955 | msr & MSR_IA32_MISC_ENABLE_ENHANCED_SPEEDSTEP ? "" : "No-", | |
fd3933ca | 4956 | msr & MSR_IA32_MISC_ENABLE_MWAIT ? "" : "No-", |
e6512624 LB |
4957 | msr & MSR_IA32_MISC_ENABLE_PREFETCH_DISABLE ? "No-" : "", |
4958 | msr & MSR_IA32_MISC_ENABLE_TURBO_DISABLE ? "No-" : ""); | |
69807a63 LB |
4959 | } |
4960 | ||
33148d67 LB |
4961 | void decode_misc_feature_control(void) |
4962 | { | |
4963 | unsigned long long msr; | |
4964 | ||
4965 | if (!has_misc_feature_control) | |
4966 | return; | |
4967 | ||
4968 | if (!get_msr(base_cpu, MSR_MISC_FEATURE_CONTROL, &msr)) | |
4969 | fprintf(outf, "cpu%d: MSR_MISC_FEATURE_CONTROL: 0x%08llx (%sL2-Prefetch %sL2-Prefetch-pair %sL1-Prefetch %sL1-IP-Prefetch)\n", | |
4970 | base_cpu, msr, | |
4971 | msr & (0 << 0) ? "No-" : "", | |
4972 | msr & (1 << 0) ? "No-" : "", | |
4973 | msr & (2 << 0) ? "No-" : "", | |
4974 | msr & (3 << 0) ? "No-" : ""); | |
4975 | } | |
f0057310 LB |
4976 | /* |
4977 | * Decode MSR_MISC_PWR_MGMT | |
4978 | * | |
4979 | * Decode the bits according to the Nehalem documentation | |
4980 | * bit[0] seems to continue to have same meaning going forward | |
4981 | * bit[1] less so... | |
4982 | */ | |
4983 | void decode_misc_pwr_mgmt_msr(void) | |
4984 | { | |
4985 | unsigned long long msr; | |
4986 | ||
4987 | if (!do_nhm_platform_info) | |
4988 | return; | |
4989 | ||
cf4cbe53 LB |
4990 | if (no_MSR_MISC_PWR_MGMT) |
4991 | return; | |
4992 | ||
f0057310 | 4993 | if (!get_msr(base_cpu, MSR_MISC_PWR_MGMT, &msr)) |
ddadb8ad | 4994 | fprintf(outf, "cpu%d: MSR_MISC_PWR_MGMT: 0x%08llx (%sable-EIST_Coordination %sable-EPB %sable-OOB)\n", |
f0057310 LB |
4995 | base_cpu, msr, |
4996 | msr & (1 << 0) ? "DIS" : "EN", | |
ddadb8ad SP |
4997 | msr & (1 << 1) ? "EN" : "DIS", |
4998 | msr & (1 << 8) ? "EN" : "DIS"); | |
f0057310 | 4999 | } |
71616c8e LB |
5000 | /* |
5001 | * Decode MSR_CC6_DEMOTION_POLICY_CONFIG, MSR_MC6_DEMOTION_POLICY_CONFIG | |
5002 | * | |
5003 | * This MSRs are present on Silvermont processors, | |
5004 | * Intel Atom processor E3000 series (Baytrail), and friends. | |
5005 | */ | |
5006 | void decode_c6_demotion_policy_msr(void) | |
5007 | { | |
5008 | unsigned long long msr; | |
5009 | ||
5010 | if (!get_msr(base_cpu, MSR_CC6_DEMOTION_POLICY_CONFIG, &msr)) | |
5011 | fprintf(outf, "cpu%d: MSR_CC6_DEMOTION_POLICY_CONFIG: 0x%08llx (%sable-CC6-Demotion)\n", | |
5012 | base_cpu, msr, msr & (1 << 0) ? "EN" : "DIS"); | |
5013 | ||
5014 | if (!get_msr(base_cpu, MSR_MC6_DEMOTION_POLICY_CONFIG, &msr)) | |
5015 | fprintf(outf, "cpu%d: MSR_MC6_DEMOTION_POLICY_CONFIG: 0x%08llx (%sable-MC6-Demotion)\n", | |
5016 | base_cpu, msr, msr & (1 << 0) ? "EN" : "DIS"); | |
5017 | } | |
7f5c258e | 5018 | |
f5a4c76a LB |
5019 | /* |
5020 | * When models are the same, for the purpose of turbostat, reuse | |
5021 | */ | |
5022 | unsigned int intel_model_duplicates(unsigned int model) | |
5023 | { | |
5024 | ||
5025 | switch(model) { | |
5026 | case INTEL_FAM6_NEHALEM_EP: /* Core i7, Xeon 5500 series - Bloomfield, Gainstown NHM-EP */ | |
5027 | case INTEL_FAM6_NEHALEM: /* Core i7 and i5 Processor - Clarksfield, Lynnfield, Jasper Forest */ | |
5028 | case 0x1F: /* Core i7 and i5 Processor - Nehalem */ | |
5029 | case INTEL_FAM6_WESTMERE: /* Westmere Client - Clarkdale, Arrandale */ | |
5030 | case INTEL_FAM6_WESTMERE_EP: /* Westmere EP - Gulftown */ | |
5031 | return INTEL_FAM6_NEHALEM; | |
5032 | ||
5033 | case INTEL_FAM6_NEHALEM_EX: /* Nehalem-EX Xeon - Beckton */ | |
5034 | case INTEL_FAM6_WESTMERE_EX: /* Westmere-EX Xeon - Eagleton */ | |
5035 | return INTEL_FAM6_NEHALEM_EX; | |
5036 | ||
5037 | case INTEL_FAM6_XEON_PHI_KNM: | |
5038 | return INTEL_FAM6_XEON_PHI_KNL; | |
5039 | ||
f5a4c76a | 5040 | case INTEL_FAM6_BROADWELL_X: |
5ebb34ed | 5041 | case INTEL_FAM6_BROADWELL_D: /* BDX-DE */ |
f5a4c76a LB |
5042 | return INTEL_FAM6_BROADWELL_X; |
5043 | ||
af239c44 | 5044 | case INTEL_FAM6_SKYLAKE_L: |
c66f78a6 | 5045 | case INTEL_FAM6_SKYLAKE: |
af239c44 | 5046 | case INTEL_FAM6_KABYLAKE_L: |
c66f78a6 | 5047 | case INTEL_FAM6_KABYLAKE: |
081c5432 CY |
5048 | case INTEL_FAM6_COMETLAKE_L: |
5049 | case INTEL_FAM6_COMETLAKE: | |
af239c44 | 5050 | return INTEL_FAM6_SKYLAKE_L; |
937807d3 | 5051 | |
af239c44 | 5052 | case INTEL_FAM6_ICELAKE_L: |
d93ea567 | 5053 | case INTEL_FAM6_ICELAKE_NNPI: |
4bf7132a CY |
5054 | case INTEL_FAM6_TIGERLAKE_L: |
5055 | case INTEL_FAM6_TIGERLAKE: | |
e7af1ed3 LB |
5056 | case INTEL_FAM6_ROCKETLAKE: |
5057 | case INTEL_FAM6_LAKEFIELD: | |
5058 | case INTEL_FAM6_ALDERLAKE: | |
5683460b | 5059 | case INTEL_FAM6_ALDERLAKE_L: |
af239c44 | 5060 | return INTEL_FAM6_CANNONLAKE_L; |
b62b3184 | 5061 | |
d7814c30 CY |
5062 | case INTEL_FAM6_ATOM_TREMONT_L: |
5063 | return INTEL_FAM6_ATOM_TREMONT; | |
5064 | ||
23274faf | 5065 | case INTEL_FAM6_ICELAKE_X: |
e7af1ed3 | 5066 | case INTEL_FAM6_SAPPHIRERAPIDS_X: |
23274faf | 5067 | return INTEL_FAM6_SKYLAKE_X; |
f5a4c76a LB |
5068 | } |
5069 | return model; | |
5070 | } | |
d76bb7a0 LB |
5071 | |
5072 | void print_dev_latency(void) | |
5073 | { | |
5074 | char *path = "/dev/cpu_dma_latency"; | |
5075 | int fd; | |
5076 | int value; | |
5077 | int retval; | |
5078 | ||
5079 | fd = open(path, O_RDONLY); | |
5080 | if (fd < 0) { | |
5081 | warn("fopen %s\n", path); | |
5082 | return; | |
5083 | } | |
5084 | ||
5085 | retval = read(fd, (void *)&value, sizeof(int)); | |
5086 | if (retval != sizeof(int)) { | |
5087 | warn("read %s\n", path); | |
5088 | close(fd); | |
5089 | return; | |
5090 | } | |
5091 | fprintf(outf, "/dev/cpu_dma_latency: %d usec (%s)\n", | |
5092 | value, value == 2000000000 ? "default" : "constrained"); | |
5093 | ||
5094 | close(fd); | |
5095 | } | |
5096 | ||
2af4f9b8 LB |
5097 | |
5098 | /* | |
5099 | * Linux-perf manages the the HW instructions-retired counter | |
5100 | * by enabling when requested, and hiding rollover | |
5101 | */ | |
5102 | void linux_perf_init(void) | |
5103 | { | |
5104 | if (!BIC_IS_ENABLED(BIC_IPC)) | |
5105 | return; | |
5106 | ||
5107 | if (access("/proc/sys/kernel/perf_event_paranoid", F_OK)) | |
5108 | return; | |
5109 | ||
5110 | fd_instr_count_percpu = calloc(topo.max_cpu_num + 1, sizeof(int)); | |
5111 | if (fd_instr_count_percpu == NULL) | |
5112 | err(-1, "calloc fd_instr_count_percpu"); | |
5113 | ||
5114 | BIC_PRESENT(BIC_IPC); | |
5115 | } | |
5116 | ||
fcd17211 | 5117 | void process_cpuid() |
103a8fea | 5118 | { |
34041551 LB |
5119 | unsigned int eax, ebx, ecx, edx; |
5120 | unsigned int fms, family, model, stepping, ecx_flags, edx_flags; | |
b3a34e93 | 5121 | unsigned int has_turbo; |
ed0757b8 | 5122 | unsigned long long ucode_patch = 0; |
103a8fea LB |
5123 | |
5124 | eax = ebx = ecx = edx = 0; | |
5125 | ||
5aea2f7f | 5126 | __cpuid(0, max_level, ebx, ecx, edx); |
103a8fea | 5127 | |
34041551 | 5128 | if (ebx == 0x756e6547 && ecx == 0x6c65746e && edx == 0x49656e69) |
103a8fea | 5129 | genuine_intel = 1; |
34041551 LB |
5130 | else if (ebx == 0x68747541 && ecx == 0x444d4163 && edx == 0x69746e65) |
5131 | authentic_amd = 1; | |
c1c10cc7 PW |
5132 | else if (ebx == 0x6f677948 && ecx == 0x656e6975 && edx == 0x6e65476e) |
5133 | hygon_genuine = 1; | |
103a8fea | 5134 | |
96e47158 | 5135 | if (!quiet) |
ed0757b8 LB |
5136 | fprintf(outf, "CPUID(0): %.4s%.4s%.4s 0x%x CPUID levels\n", |
5137 | (char *)&ebx, (char *)&edx, (char *)&ecx, max_level); | |
103a8fea | 5138 | |
5aea2f7f | 5139 | __cpuid(1, fms, ebx, ecx, edx); |
103a8fea LB |
5140 | family = (fms >> 8) & 0xf; |
5141 | model = (fms >> 4) & 0xf; | |
5142 | stepping = fms & 0xf; | |
5aa3d1a2 CW |
5143 | if (family == 0xf) |
5144 | family += (fms >> 20) & 0xff; | |
5145 | if (family >= 6) | |
103a8fea | 5146 | model += ((fms >> 16) & 0xf) << 4; |
34041551 LB |
5147 | ecx_flags = ecx; |
5148 | edx_flags = edx; | |
103a8fea | 5149 | |
ed0757b8 LB |
5150 | if (get_msr(sched_getcpu(), MSR_IA32_UCODE_REV, &ucode_patch)) |
5151 | warnx("get_msr(UCODE)\n"); | |
5152 | ||
103a8fea LB |
5153 | /* |
5154 | * check max extended function levels of CPUID. | |
5155 | * This is needed to check for invariant TSC. | |
5156 | * This check is valid for both Intel and AMD. | |
5157 | */ | |
5158 | ebx = ecx = edx = 0; | |
5aea2f7f | 5159 | __cpuid(0x80000000, max_extended_level, ebx, ecx, edx); |
103a8fea | 5160 | |
34041551 | 5161 | if (!quiet) { |
ed0757b8 LB |
5162 | fprintf(outf, "CPUID(1): family:model:stepping 0x%x:%x:%x (%d:%d:%d) microcode 0x%x\n", |
5163 | family, model, stepping, family, model, stepping, (unsigned int)((ucode_patch >> 32) & 0xFFFFFFFF)); | |
5164 | fprintf(outf, "CPUID(0x80000000): max_extended_levels: 0x%x\n", max_extended_level); | |
34041551 LB |
5165 | fprintf(outf, "CPUID(1): %s %s %s %s %s %s %s %s %s %s\n", |
5166 | ecx_flags & (1 << 0) ? "SSE3" : "-", | |
5167 | ecx_flags & (1 << 3) ? "MONITOR" : "-", | |
5168 | ecx_flags & (1 << 6) ? "SMX" : "-", | |
5169 | ecx_flags & (1 << 7) ? "EIST" : "-", | |
5170 | ecx_flags & (1 << 8) ? "TM2" : "-", | |
5171 | edx_flags & (1 << 4) ? "TSC" : "-", | |
5172 | edx_flags & (1 << 5) ? "MSR" : "-", | |
5173 | edx_flags & (1 << 22) ? "ACPI-TM" : "-", | |
5174 | edx_flags & (1 << 28) ? "HT" : "-", | |
5175 | edx_flags & (1 << 29) ? "TM" : "-"); | |
5176 | } | |
f5a4c76a LB |
5177 | if (genuine_intel) |
5178 | model = intel_model_duplicates(model); | |
34041551 LB |
5179 | |
5180 | if (!(edx_flags & (1 << 5))) | |
5181 | errx(1, "CPUID: no MSR"); | |
5182 | ||
61a87ba7 | 5183 | if (max_extended_level >= 0x80000007) { |
103a8fea | 5184 | |
d7899447 LB |
5185 | /* |
5186 | * Non-Stop TSC is advertised by CPUID.EAX=0x80000007: EDX.bit8 | |
5187 | * this check is valid for both Intel and AMD | |
5188 | */ | |
5aea2f7f | 5189 | __cpuid(0x80000007, eax, ebx, ecx, edx); |
d7899447 LB |
5190 | has_invariant_tsc = edx & (1 << 8); |
5191 | } | |
103a8fea LB |
5192 | |
5193 | /* | |
5194 | * APERF/MPERF is advertised by CPUID.EAX=0x6: ECX.bit0 | |
5195 | * this check is valid for both Intel and AMD | |
5196 | */ | |
5197 | ||
5aea2f7f | 5198 | __cpuid(0x6, eax, ebx, ecx, edx); |
8209e054 | 5199 | has_aperf = ecx & (1 << 0); |
812db3f7 LB |
5200 | if (has_aperf) { |
5201 | BIC_PRESENT(BIC_Avg_MHz); | |
5202 | BIC_PRESENT(BIC_Busy); | |
5203 | BIC_PRESENT(BIC_Bzy_MHz); | |
5204 | } | |
889facbe | 5205 | do_dts = eax & (1 << 0); |
812db3f7 LB |
5206 | if (do_dts) |
5207 | BIC_PRESENT(BIC_CoreTmp); | |
b3a34e93 | 5208 | has_turbo = eax & (1 << 1); |
889facbe | 5209 | do_ptm = eax & (1 << 6); |
812db3f7 LB |
5210 | if (do_ptm) |
5211 | BIC_PRESENT(BIC_PkgTmp); | |
7f5c258e LB |
5212 | has_hwp = eax & (1 << 7); |
5213 | has_hwp_notify = eax & (1 << 8); | |
5214 | has_hwp_activity_window = eax & (1 << 9); | |
5215 | has_hwp_epp = eax & (1 << 10); | |
5216 | has_hwp_pkg = eax & (1 << 11); | |
889facbe LB |
5217 | has_epb = ecx & (1 << 3); |
5218 | ||
96e47158 | 5219 | if (!quiet) |
b3a34e93 | 5220 | fprintf(outf, "CPUID(6): %sAPERF, %sTURBO, %sDTS, %sPTM, %sHWP, " |
7f5c258e LB |
5221 | "%sHWPnotify, %sHWPwindow, %sHWPepp, %sHWPpkg, %sEPB\n", |
5222 | has_aperf ? "" : "No-", | |
b3a34e93 | 5223 | has_turbo ? "" : "No-", |
7f5c258e LB |
5224 | do_dts ? "" : "No-", |
5225 | do_ptm ? "" : "No-", | |
5226 | has_hwp ? "" : "No-", | |
5227 | has_hwp_notify ? "" : "No-", | |
5228 | has_hwp_activity_window ? "" : "No-", | |
5229 | has_hwp_epp ? "" : "No-", | |
5230 | has_hwp_pkg ? "" : "No-", | |
5231 | has_epb ? "" : "No-"); | |
103a8fea | 5232 | |
96e47158 | 5233 | if (!quiet) |
69807a63 LB |
5234 | decode_misc_enable_msr(); |
5235 | ||
33148d67 | 5236 | |
96e47158 | 5237 | if (max_level >= 0x7 && !quiet) { |
aa8d8cc7 | 5238 | int has_sgx; |
103a8fea | 5239 | |
aa8d8cc7 LB |
5240 | ecx = 0; |
5241 | ||
5242 | __cpuid_count(0x7, 0, eax, ebx, ecx, edx); | |
5243 | ||
5244 | has_sgx = ebx & (1 << 2); | |
5245 | fprintf(outf, "CPUID(7): %sSGX\n", has_sgx ? "" : "No-"); | |
5246 | ||
5247 | if (has_sgx) | |
5248 | decode_feature_control_msr(); | |
5249 | } | |
5250 | ||
61a87ba7 | 5251 | if (max_level >= 0x15) { |
8a5bdf41 LB |
5252 | unsigned int eax_crystal; |
5253 | unsigned int ebx_tsc; | |
5254 | ||
5255 | /* | |
5256 | * CPUID 15H TSC/Crystal ratio, possibly Crystal Hz | |
5257 | */ | |
5258 | eax_crystal = ebx_tsc = crystal_hz = edx = 0; | |
5aea2f7f | 5259 | __cpuid(0x15, eax_crystal, ebx_tsc, crystal_hz, edx); |
8a5bdf41 LB |
5260 | |
5261 | if (ebx_tsc != 0) { | |
5262 | ||
96e47158 | 5263 | if (!quiet && (ebx != 0)) |
b7d8c148 | 5264 | fprintf(outf, "CPUID(0x15): eax_crystal: %d ebx_tsc: %d ecx_crystal_hz: %d\n", |
8a5bdf41 LB |
5265 | eax_crystal, ebx_tsc, crystal_hz); |
5266 | ||
5267 | if (crystal_hz == 0) | |
5268 | switch(model) { | |
af239c44 | 5269 | case INTEL_FAM6_SKYLAKE_L: /* SKL */ |
e8efbc80 LB |
5270 | crystal_hz = 24000000; /* 24.0 MHz */ |
5271 | break; | |
5ebb34ed | 5272 | case INTEL_FAM6_ATOM_GOLDMONT_D: /* DNV */ |
ec53e594 LB |
5273 | crystal_hz = 25000000; /* 25.0 MHz */ |
5274 | break; | |
869ce69e | 5275 | case INTEL_FAM6_ATOM_GOLDMONT: /* BXT */ |
f2c4db1b | 5276 | case INTEL_FAM6_ATOM_GOLDMONT_PLUS: |
e8efbc80 | 5277 | crystal_hz = 19200000; /* 19.2 MHz */ |
8a5bdf41 LB |
5278 | break; |
5279 | default: | |
5280 | crystal_hz = 0; | |
5281 | } | |
5282 | ||
5283 | if (crystal_hz) { | |
5284 | tsc_hz = (unsigned long long) crystal_hz * ebx_tsc / eax_crystal; | |
96e47158 | 5285 | if (!quiet) |
b7d8c148 | 5286 | fprintf(outf, "TSC: %lld MHz (%d Hz * %d / %d / 1000000)\n", |
8a5bdf41 LB |
5287 | tsc_hz / 1000000, crystal_hz, ebx_tsc, eax_crystal); |
5288 | } | |
5289 | } | |
5290 | } | |
61a87ba7 LB |
5291 | if (max_level >= 0x16) { |
5292 | unsigned int base_mhz, max_mhz, bus_mhz, edx; | |
5293 | ||
5294 | /* | |
5295 | * CPUID 16H Base MHz, Max MHz, Bus MHz | |
5296 | */ | |
5297 | base_mhz = max_mhz = bus_mhz = edx = 0; | |
5298 | ||
5aea2f7f | 5299 | __cpuid(0x16, base_mhz, max_mhz, bus_mhz, edx); |
96e47158 | 5300 | if (!quiet) |
b7d8c148 | 5301 | fprintf(outf, "CPUID(0x16): base_mhz: %d max_mhz: %d bus_mhz: %d\n", |
61a87ba7 LB |
5302 | base_mhz, max_mhz, bus_mhz); |
5303 | } | |
8a5bdf41 | 5304 | |
b2b34dfe HC |
5305 | if (has_aperf) |
5306 | aperf_mperf_multiplier = get_aperf_mperf_multiplier(family, model); | |
5307 | ||
812db3f7 LB |
5308 | BIC_PRESENT(BIC_IRQ); |
5309 | BIC_PRESENT(BIC_TSC_MHz); | |
5310 | ||
5311 | if (probe_nhm_msrs(family, model)) { | |
5312 | do_nhm_platform_info = 1; | |
5313 | BIC_PRESENT(BIC_CPU_c1); | |
5314 | BIC_PRESENT(BIC_CPU_c3); | |
5315 | BIC_PRESENT(BIC_CPU_c6); | |
5316 | BIC_PRESENT(BIC_SMI); | |
5317 | } | |
d7899447 | 5318 | do_snb_cstates = has_snb_msrs(family, model); |
812db3f7 LB |
5319 | |
5320 | if (do_snb_cstates) | |
5321 | BIC_PRESENT(BIC_CPU_c7); | |
5322 | ||
5a63426e | 5323 | do_irtl_snb = has_snb_msrs(family, model); |
0f47c08d LB |
5324 | if (do_snb_cstates && (pkg_cstate_limit >= PCL__2)) |
5325 | BIC_PRESENT(BIC_Pkgpc2); | |
5326 | if (pkg_cstate_limit >= PCL__3) | |
5327 | BIC_PRESENT(BIC_Pkgpc3); | |
5328 | if (pkg_cstate_limit >= PCL__6) | |
5329 | BIC_PRESENT(BIC_Pkgpc6); | |
5330 | if (do_snb_cstates && (pkg_cstate_limit >= PCL__7)) | |
5331 | BIC_PRESENT(BIC_Pkgpc7); | |
0539ba11 | 5332 | if (has_slv_msrs(family, model)) { |
0f47c08d LB |
5333 | BIC_NOT_PRESENT(BIC_Pkgpc2); |
5334 | BIC_NOT_PRESENT(BIC_Pkgpc3); | |
5335 | BIC_PRESENT(BIC_Pkgpc6); | |
5336 | BIC_NOT_PRESENT(BIC_Pkgpc7); | |
0539ba11 LB |
5337 | BIC_PRESENT(BIC_Mod_c6); |
5338 | use_c1_residency_msr = 1; | |
5339 | } | |
20de0dab AL |
5340 | if (is_jvl(family, model)) { |
5341 | BIC_NOT_PRESENT(BIC_CPU_c3); | |
5342 | BIC_NOT_PRESENT(BIC_CPU_c7); | |
5343 | BIC_NOT_PRESENT(BIC_Pkgpc2); | |
5344 | BIC_NOT_PRESENT(BIC_Pkgpc3); | |
5345 | BIC_NOT_PRESENT(BIC_Pkgpc6); | |
5346 | BIC_NOT_PRESENT(BIC_Pkgpc7); | |
5347 | } | |
7170a374 LB |
5348 | if (is_dnv(family, model)) { |
5349 | BIC_PRESENT(BIC_CPU_c1); | |
5350 | BIC_NOT_PRESENT(BIC_CPU_c3); | |
5351 | BIC_NOT_PRESENT(BIC_Pkgpc3); | |
5352 | BIC_NOT_PRESENT(BIC_CPU_c7); | |
5353 | BIC_NOT_PRESENT(BIC_Pkgpc7); | |
5354 | use_c1_residency_msr = 1; | |
5355 | } | |
34c76197 LB |
5356 | if (is_skx(family, model)) { |
5357 | BIC_NOT_PRESENT(BIC_CPU_c3); | |
5358 | BIC_NOT_PRESENT(BIC_Pkgpc3); | |
5359 | BIC_NOT_PRESENT(BIC_CPU_c7); | |
5360 | BIC_NOT_PRESENT(BIC_Pkgpc7); | |
5361 | } | |
ade0ebac LB |
5362 | if (is_bdx(family, model)) { |
5363 | BIC_NOT_PRESENT(BIC_CPU_c7); | |
5364 | BIC_NOT_PRESENT(BIC_Pkgpc7); | |
5365 | } | |
570992fc | 5366 | if (has_c8910_msrs(family, model)) { |
c315a09b LB |
5367 | if (pkg_cstate_limit >= PCL__8) |
5368 | BIC_PRESENT(BIC_Pkgpc8); | |
5369 | if (pkg_cstate_limit >= PCL__9) | |
5370 | BIC_PRESENT(BIC_Pkgpc9); | |
5371 | if (pkg_cstate_limit >= PCL_10) | |
5372 | BIC_PRESENT(BIC_Pkgpc10); | |
0f47c08d | 5373 | } |
570992fc | 5374 | do_irtl_hsw = has_c8910_msrs(family, model); |
a99d8730 LB |
5375 | if (has_skl_msrs(family, model)) { |
5376 | BIC_PRESENT(BIC_Totl_c0); | |
5377 | BIC_PRESENT(BIC_Any_c0); | |
5378 | BIC_PRESENT(BIC_GFX_c0); | |
5379 | BIC_PRESENT(BIC_CPUGFX); | |
5380 | } | |
144b44b1 | 5381 | do_slm_cstates = is_slm(family, model); |
fb5d4327 | 5382 | do_knl_cstates = is_knl(family, model); |
103a8fea | 5383 | |
f6708400 CY |
5384 | if (do_slm_cstates || do_knl_cstates || is_cnl(family, model) || |
5385 | is_ehl(family, model)) | |
562855ee | 5386 | BIC_NOT_PRESENT(BIC_CPU_c3); |
103a8fea | 5387 | |
96e47158 | 5388 | if (!quiet) |
f0057310 LB |
5389 | decode_misc_pwr_mgmt_msr(); |
5390 | ||
96e47158 | 5391 | if (!quiet && has_slv_msrs(family, model)) |
71616c8e LB |
5392 | decode_c6_demotion_policy_msr(); |
5393 | ||
889facbe | 5394 | rapl_probe(family, model); |
3a9a941d | 5395 | perf_limit_reasons_probe(family, model); |
ac980e13 | 5396 | automatic_cstate_conversion_probe(family, model); |
889facbe | 5397 | |
96e47158 | 5398 | if (!quiet) |
1b69317d | 5399 | dump_cstate_pstate_config_info(family, model); |
fcd17211 | 5400 | |
d76bb7a0 LB |
5401 | if (!quiet) |
5402 | print_dev_latency(); | |
41618e63 LB |
5403 | if (!quiet) |
5404 | dump_sysfs_cstate_config(); | |
7293fccd LB |
5405 | if (!quiet) |
5406 | dump_sysfs_pstate_config(); | |
41618e63 | 5407 | |
a2b7b749 LB |
5408 | if (has_skl_msrs(family, model)) |
5409 | calculate_tsc_tweak(); | |
5410 | ||
812db3f7 LB |
5411 | if (!access("/sys/class/drm/card0/power/rc6_residency_ms", R_OK)) |
5412 | BIC_PRESENT(BIC_GFX_rc6); | |
fdf676e5 | 5413 | |
812db3f7 LB |
5414 | if (!access("/sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz", R_OK)) |
5415 | BIC_PRESENT(BIC_GFXMHz); | |
27d47356 | 5416 | |
b4b91569 RA |
5417 | if (!access("/sys/class/graphics/fb0/device/drm/card0/gt_act_freq_mhz", R_OK)) |
5418 | BIC_PRESENT(BIC_GFXACTMHz); | |
5419 | ||
be0e54c4 LB |
5420 | if (!access("/sys/devices/system/cpu/cpuidle/low_power_idle_cpu_residency_us", R_OK)) |
5421 | BIC_PRESENT(BIC_CPU_LPI); | |
5422 | else | |
5423 | BIC_NOT_PRESENT(BIC_CPU_LPI); | |
5424 | ||
1f81c5ef LB |
5425 | if (!access(sys_lpi_file_sysfs, R_OK)) { |
5426 | sys_lpi_file = sys_lpi_file_sysfs; | |
be0e54c4 | 5427 | BIC_PRESENT(BIC_SYS_LPI); |
1f81c5ef LB |
5428 | } else if (!access(sys_lpi_file_debugfs, R_OK)) { |
5429 | sys_lpi_file = sys_lpi_file_debugfs; | |
5430 | BIC_PRESENT(BIC_SYS_LPI); | |
5431 | } else { | |
5432 | sys_lpi_file_sysfs = NULL; | |
be0e54c4 | 5433 | BIC_NOT_PRESENT(BIC_SYS_LPI); |
1f81c5ef | 5434 | } |
be0e54c4 | 5435 | |
96e47158 | 5436 | if (!quiet) |
33148d67 LB |
5437 | decode_misc_feature_control(); |
5438 | ||
889facbe | 5439 | return; |
103a8fea LB |
5440 | } |
5441 | ||
103a8fea LB |
5442 | /* |
5443 | * in /dev/cpu/ return success for names that are numbers | |
5444 | * ie. filter out ".", "..", "microcode". | |
5445 | */ | |
5446 | int dir_filter(const struct dirent *dirp) | |
5447 | { | |
5448 | if (isdigit(dirp->d_name[0])) | |
5449 | return 1; | |
5450 | else | |
5451 | return 0; | |
5452 | } | |
5453 | ||
5454 | int open_dev_cpu_msr(int dummy1) | |
5455 | { | |
5456 | return 0; | |
5457 | } | |
5458 | ||
c98d5d94 LB |
5459 | void topology_probe() |
5460 | { | |
5461 | int i; | |
5462 | int max_core_id = 0; | |
5463 | int max_package_id = 0; | |
6de68fe1 | 5464 | int max_die_id = 0; |
c98d5d94 | 5465 | int max_siblings = 0; |
c98d5d94 LB |
5466 | |
5467 | /* Initialize num_cpus, max_cpu_num */ | |
843c5791 | 5468 | set_max_cpu_num(); |
c98d5d94 | 5469 | topo.num_cpus = 0; |
c98d5d94 LB |
5470 | for_all_proc_cpus(count_cpus); |
5471 | if (!summary_only && topo.num_cpus > 1) | |
812db3f7 | 5472 | BIC_PRESENT(BIC_CPU); |
c98d5d94 | 5473 | |
d8af6f5f | 5474 | if (debug > 1) |
b7d8c148 | 5475 | fprintf(outf, "num_cpus %d max_cpu_num %d\n", topo.num_cpus, topo.max_cpu_num); |
c98d5d94 LB |
5476 | |
5477 | cpus = calloc(1, (topo.max_cpu_num + 1) * sizeof(struct cpu_topology)); | |
b2c95d90 JT |
5478 | if (cpus == NULL) |
5479 | err(1, "calloc cpus"); | |
c98d5d94 LB |
5480 | |
5481 | /* | |
5482 | * Allocate and initialize cpu_present_set | |
5483 | */ | |
5484 | cpu_present_set = CPU_ALLOC((topo.max_cpu_num + 1)); | |
b2c95d90 JT |
5485 | if (cpu_present_set == NULL) |
5486 | err(3, "CPU_ALLOC"); | |
c98d5d94 LB |
5487 | cpu_present_setsize = CPU_ALLOC_SIZE((topo.max_cpu_num + 1)); |
5488 | CPU_ZERO_S(cpu_present_setsize, cpu_present_set); | |
5489 | for_all_proc_cpus(mark_cpu_present); | |
5490 | ||
1ef7d21a LB |
5491 | /* |
5492 | * Validate that all cpus in cpu_subset are also in cpu_present_set | |
5493 | */ | |
5494 | for (i = 0; i < CPU_SUBSET_MAXCPUS; ++i) { | |
5495 | if (CPU_ISSET_S(i, cpu_subset_size, cpu_subset)) | |
5496 | if (!CPU_ISSET_S(i, cpu_present_setsize, cpu_present_set)) | |
5497 | err(1, "cpu%d not present", i); | |
5498 | } | |
5499 | ||
c98d5d94 LB |
5500 | /* |
5501 | * Allocate and initialize cpu_affinity_set | |
5502 | */ | |
5503 | cpu_affinity_set = CPU_ALLOC((topo.max_cpu_num + 1)); | |
b2c95d90 JT |
5504 | if (cpu_affinity_set == NULL) |
5505 | err(3, "CPU_ALLOC"); | |
c98d5d94 LB |
5506 | cpu_affinity_setsize = CPU_ALLOC_SIZE((topo.max_cpu_num + 1)); |
5507 | CPU_ZERO_S(cpu_affinity_setsize, cpu_affinity_set); | |
5508 | ||
8cb48b32 | 5509 | for_all_proc_cpus(init_thread_id); |
c98d5d94 LB |
5510 | |
5511 | /* | |
5512 | * For online cpus | |
5513 | * find max_core_id, max_package_id | |
5514 | */ | |
5515 | for (i = 0; i <= topo.max_cpu_num; ++i) { | |
5516 | int siblings; | |
5517 | ||
5518 | if (cpu_is_not_present(i)) { | |
d8af6f5f | 5519 | if (debug > 1) |
b7d8c148 | 5520 | fprintf(outf, "cpu%d NOT PRESENT\n", i); |
c98d5d94 LB |
5521 | continue; |
5522 | } | |
c98d5d94 | 5523 | |
0e2d8f05 LB |
5524 | cpus[i].logical_cpu_id = i; |
5525 | ||
5526 | /* get package information */ | |
c98d5d94 LB |
5527 | cpus[i].physical_package_id = get_physical_package_id(i); |
5528 | if (cpus[i].physical_package_id > max_package_id) | |
5529 | max_package_id = cpus[i].physical_package_id; | |
5530 | ||
6de68fe1 LB |
5531 | /* get die information */ |
5532 | cpus[i].die_id = get_die_id(i); | |
5533 | if (cpus[i].die_id > max_die_id) | |
5534 | max_die_id = cpus[i].die_id; | |
5535 | ||
0e2d8f05 | 5536 | /* get numa node information */ |
ef605741 PB |
5537 | cpus[i].physical_node_id = get_physical_node_id(&cpus[i]); |
5538 | if (cpus[i].physical_node_id > topo.max_node_num) | |
5539 | topo.max_node_num = cpus[i].physical_node_id; | |
0e2d8f05 LB |
5540 | |
5541 | /* get core information */ | |
5542 | cpus[i].physical_core_id = get_core_id(i); | |
5543 | if (cpus[i].physical_core_id > max_core_id) | |
5544 | max_core_id = cpus[i].physical_core_id; | |
5545 | ||
5546 | /* get thread information */ | |
5547 | siblings = get_thread_siblings(&cpus[i]); | |
c98d5d94 LB |
5548 | if (siblings > max_siblings) |
5549 | max_siblings = siblings; | |
4f206a0f | 5550 | if (cpus[i].thread_id == 0) |
8cb48b32 | 5551 | topo.num_cores++; |
c98d5d94 | 5552 | } |
ef605741 | 5553 | |
70a9c6e8 | 5554 | topo.cores_per_node = max_core_id + 1; |
d8af6f5f | 5555 | if (debug > 1) |
b7d8c148 | 5556 | fprintf(outf, "max_core_id %d, sizing for %d cores per package\n", |
70a9c6e8 PB |
5557 | max_core_id, topo.cores_per_node); |
5558 | if (!summary_only && topo.cores_per_node > 1) | |
812db3f7 | 5559 | BIC_PRESENT(BIC_Core); |
c98d5d94 | 5560 | |
6de68fe1 LB |
5561 | topo.num_die = max_die_id + 1; |
5562 | if (debug > 1) | |
5563 | fprintf(outf, "max_die_id %d, sizing for %d die\n", | |
5564 | max_die_id, topo.num_die); | |
5565 | if (!summary_only && topo.num_die > 1) | |
5566 | BIC_PRESENT(BIC_Die); | |
5567 | ||
c98d5d94 | 5568 | topo.num_packages = max_package_id + 1; |
d8af6f5f | 5569 | if (debug > 1) |
b7d8c148 | 5570 | fprintf(outf, "max_package_id %d, sizing for %d packages\n", |
c98d5d94 | 5571 | max_package_id, topo.num_packages); |
7da6e3e2 | 5572 | if (!summary_only && topo.num_packages > 1) |
812db3f7 | 5573 | BIC_PRESENT(BIC_Package); |
c98d5d94 | 5574 | |
ef605741 PB |
5575 | set_node_data(); |
5576 | if (debug > 1) | |
70a9c6e8 | 5577 | fprintf(outf, "nodes_per_pkg %d\n", topo.nodes_per_pkg); |
01235041 PB |
5578 | if (!summary_only && topo.nodes_per_pkg > 1) |
5579 | BIC_PRESENT(BIC_Node); | |
ef605741 | 5580 | |
70a9c6e8 | 5581 | topo.threads_per_core = max_siblings; |
d8af6f5f | 5582 | if (debug > 1) |
b7d8c148 | 5583 | fprintf(outf, "max_siblings %d\n", max_siblings); |
2ffbb224 PB |
5584 | |
5585 | if (debug < 1) | |
5586 | return; | |
5587 | ||
5588 | for (i = 0; i <= topo.max_cpu_num; ++i) { | |
0ec712e3 LB |
5589 | if (cpu_is_not_present(i)) |
5590 | continue; | |
2ffbb224 | 5591 | fprintf(outf, |
6de68fe1 LB |
5592 | "cpu %d pkg %d die %d node %d lnode %d core %d thread %d\n", |
5593 | i, cpus[i].physical_package_id, cpus[i].die_id, | |
2ffbb224 PB |
5594 | cpus[i].physical_node_id, |
5595 | cpus[i].logical_node_id, | |
5596 | cpus[i].physical_core_id, | |
5597 | cpus[i].thread_id); | |
5598 | } | |
5599 | ||
c98d5d94 LB |
5600 | } |
5601 | ||
5602 | void | |
40f5cfe7 PB |
5603 | allocate_counters(struct thread_data **t, struct core_data **c, |
5604 | struct pkg_data **p) | |
c98d5d94 LB |
5605 | { |
5606 | int i; | |
40f5cfe7 PB |
5607 | int num_cores = topo.cores_per_node * topo.nodes_per_pkg * |
5608 | topo.num_packages; | |
5609 | int num_threads = topo.threads_per_core * num_cores; | |
c98d5d94 | 5610 | |
40f5cfe7 | 5611 | *t = calloc(num_threads, sizeof(struct thread_data)); |
c98d5d94 LB |
5612 | if (*t == NULL) |
5613 | goto error; | |
5614 | ||
40f5cfe7 | 5615 | for (i = 0; i < num_threads; i++) |
c98d5d94 LB |
5616 | (*t)[i].cpu_id = -1; |
5617 | ||
40f5cfe7 | 5618 | *c = calloc(num_cores, sizeof(struct core_data)); |
c98d5d94 LB |
5619 | if (*c == NULL) |
5620 | goto error; | |
5621 | ||
40f5cfe7 | 5622 | for (i = 0; i < num_cores; i++) |
c98d5d94 LB |
5623 | (*c)[i].core_id = -1; |
5624 | ||
678a3bd1 | 5625 | *p = calloc(topo.num_packages, sizeof(struct pkg_data)); |
c98d5d94 LB |
5626 | if (*p == NULL) |
5627 | goto error; | |
5628 | ||
5629 | for (i = 0; i < topo.num_packages; i++) | |
5630 | (*p)[i].package_id = i; | |
5631 | ||
5632 | return; | |
5633 | error: | |
b2c95d90 | 5634 | err(1, "calloc counters"); |
c98d5d94 LB |
5635 | } |
5636 | /* | |
5637 | * init_counter() | |
5638 | * | |
c98d5d94 | 5639 | * set FIRST_THREAD_IN_CORE and FIRST_CORE_IN_PACKAGE |
c98d5d94 LB |
5640 | */ |
5641 | void init_counter(struct thread_data *thread_base, struct core_data *core_base, | |
8cb48b32 | 5642 | struct pkg_data *pkg_base, int cpu_id) |
c98d5d94 | 5643 | { |
8cb48b32 | 5644 | int pkg_id = cpus[cpu_id].physical_package_id; |
40f5cfe7 | 5645 | int node_id = cpus[cpu_id].logical_node_id; |
8cb48b32 PB |
5646 | int core_id = cpus[cpu_id].physical_core_id; |
5647 | int thread_id = cpus[cpu_id].thread_id; | |
c98d5d94 LB |
5648 | struct thread_data *t; |
5649 | struct core_data *c; | |
5650 | struct pkg_data *p; | |
5651 | ||
42dd4520 NC |
5652 | |
5653 | /* Workaround for systems where physical_node_id==-1 | |
5654 | * and logical_node_id==(-1 - topo.num_cpus) | |
5655 | */ | |
5656 | if (node_id < 0) | |
5657 | node_id = 0; | |
5658 | ||
40f5cfe7 PB |
5659 | t = GET_THREAD(thread_base, thread_id, core_id, node_id, pkg_id); |
5660 | c = GET_CORE(core_base, core_id, node_id, pkg_id); | |
8cb48b32 | 5661 | p = GET_PKG(pkg_base, pkg_id); |
c98d5d94 LB |
5662 | |
5663 | t->cpu_id = cpu_id; | |
8cb48b32 | 5664 | if (thread_id == 0) { |
c98d5d94 LB |
5665 | t->flags |= CPU_IS_FIRST_THREAD_IN_CORE; |
5666 | if (cpu_is_first_core_in_package(cpu_id)) | |
5667 | t->flags |= CPU_IS_FIRST_CORE_IN_PACKAGE; | |
5668 | } | |
5669 | ||
8cb48b32 PB |
5670 | c->core_id = core_id; |
5671 | p->package_id = pkg_id; | |
c98d5d94 LB |
5672 | } |
5673 | ||
5674 | ||
5675 | int initialize_counters(int cpu_id) | |
5676 | { | |
8cb48b32 PB |
5677 | init_counter(EVEN_COUNTERS, cpu_id); |
5678 | init_counter(ODD_COUNTERS, cpu_id); | |
c98d5d94 LB |
5679 | return 0; |
5680 | } | |
5681 | ||
5682 | void allocate_output_buffer() | |
5683 | { | |
eeb71c95 | 5684 | output_buffer = calloc(1, (1 + topo.num_cpus) * 2048); |
c98d5d94 | 5685 | outp = output_buffer; |
b2c95d90 JT |
5686 | if (outp == NULL) |
5687 | err(-1, "calloc output buffer"); | |
c98d5d94 | 5688 | } |
36229897 LB |
5689 | void allocate_fd_percpu(void) |
5690 | { | |
01a67adf | 5691 | fd_percpu = calloc(topo.max_cpu_num + 1, sizeof(int)); |
36229897 LB |
5692 | if (fd_percpu == NULL) |
5693 | err(-1, "calloc fd_percpu"); | |
5694 | } | |
562a2d37 LB |
5695 | void allocate_irq_buffers(void) |
5696 | { | |
5697 | irq_column_2_cpu = calloc(topo.num_cpus, sizeof(int)); | |
5698 | if (irq_column_2_cpu == NULL) | |
5699 | err(-1, "calloc %d", topo.num_cpus); | |
c98d5d94 | 5700 | |
01a67adf | 5701 | irqs_per_cpu = calloc(topo.max_cpu_num + 1, sizeof(int)); |
562a2d37 | 5702 | if (irqs_per_cpu == NULL) |
01a67adf | 5703 | err(-1, "calloc %d", topo.max_cpu_num + 1); |
562a2d37 | 5704 | } |
c98d5d94 LB |
5705 | void setup_all_buffers(void) |
5706 | { | |
5707 | topology_probe(); | |
562a2d37 | 5708 | allocate_irq_buffers(); |
36229897 | 5709 | allocate_fd_percpu(); |
c98d5d94 LB |
5710 | allocate_counters(&thread_even, &core_even, &package_even); |
5711 | allocate_counters(&thread_odd, &core_odd, &package_odd); | |
5712 | allocate_output_buffer(); | |
5713 | for_all_proc_cpus(initialize_counters); | |
5714 | } | |
3b4d5c7f | 5715 | |
7ce7d5de PB |
5716 | void set_base_cpu(void) |
5717 | { | |
5718 | base_cpu = sched_getcpu(); | |
5719 | if (base_cpu < 0) | |
5720 | err(-ENODEV, "No valid cpus found"); | |
5721 | ||
5722 | if (debug > 1) | |
b7d8c148 | 5723 | fprintf(outf, "base_cpu = %d\n", base_cpu); |
7ce7d5de PB |
5724 | } |
5725 | ||
103a8fea LB |
5726 | void turbostat_init() |
5727 | { | |
7ce7d5de PB |
5728 | setup_all_buffers(); |
5729 | set_base_cpu(); | |
103a8fea | 5730 | check_dev_msr(); |
98481e79 | 5731 | check_permissions(); |
fcd17211 | 5732 | process_cpuid(); |
2af4f9b8 | 5733 | linux_perf_init(); |
103a8fea | 5734 | |
103a8fea | 5735 | |
96e47158 | 5736 | if (!quiet) |
7f5c258e LB |
5737 | for_all_cpus(print_hwp, ODD_COUNTERS); |
5738 | ||
96e47158 | 5739 | if (!quiet) |
889facbe LB |
5740 | for_all_cpus(print_epb, ODD_COUNTERS); |
5741 | ||
96e47158 | 5742 | if (!quiet) |
3a9a941d LB |
5743 | for_all_cpus(print_perf_limit, ODD_COUNTERS); |
5744 | ||
96e47158 | 5745 | if (!quiet) |
889facbe LB |
5746 | for_all_cpus(print_rapl, ODD_COUNTERS); |
5747 | ||
5748 | for_all_cpus(set_temperature_target, ODD_COUNTERS); | |
5749 | ||
96e47158 | 5750 | if (!quiet) |
889facbe | 5751 | for_all_cpus(print_thermal, ODD_COUNTERS); |
5a63426e | 5752 | |
96e47158 | 5753 | if (!quiet && do_irtl_snb) |
5a63426e | 5754 | print_irtl(); |
103a8fea LB |
5755 | } |
5756 | ||
5757 | int fork_it(char **argv) | |
5758 | { | |
103a8fea | 5759 | pid_t child_pid; |
d91bb17c | 5760 | int status; |
d15cf7c1 | 5761 | |
218f0e8d | 5762 | snapshot_proc_sysfs_files(); |
d91bb17c | 5763 | status = for_all_cpus(get_counters, EVEN_COUNTERS); |
4c2122d4 | 5764 | first_counter_read = 0; |
d91bb17c LB |
5765 | if (status) |
5766 | exit(status); | |
c98d5d94 LB |
5767 | /* clear affinity side-effect of get_counters() */ |
5768 | sched_setaffinity(0, cpu_present_setsize, cpu_present_set); | |
103a8fea LB |
5769 | gettimeofday(&tv_even, (struct timezone *)NULL); |
5770 | ||
5771 | child_pid = fork(); | |
5772 | if (!child_pid) { | |
5773 | /* child */ | |
5774 | execvp(argv[0], argv); | |
0815a3d0 | 5775 | err(errno, "exec %s", argv[0]); |
103a8fea | 5776 | } else { |
103a8fea LB |
5777 | |
5778 | /* parent */ | |
b2c95d90 JT |
5779 | if (child_pid == -1) |
5780 | err(1, "fork"); | |
103a8fea LB |
5781 | |
5782 | signal(SIGINT, SIG_IGN); | |
5783 | signal(SIGQUIT, SIG_IGN); | |
b2c95d90 JT |
5784 | if (waitpid(child_pid, &status, 0) == -1) |
5785 | err(status, "waitpid"); | |
2a954966 DA |
5786 | |
5787 | if (WIFEXITED(status)) | |
5788 | status = WEXITSTATUS(status); | |
103a8fea | 5789 | } |
c98d5d94 LB |
5790 | /* |
5791 | * n.b. fork_it() does not check for errors from for_all_cpus() | |
5792 | * because re-starting is problematic when forking | |
5793 | */ | |
218f0e8d | 5794 | snapshot_proc_sysfs_files(); |
c98d5d94 | 5795 | for_all_cpus(get_counters, ODD_COUNTERS); |
103a8fea | 5796 | gettimeofday(&tv_odd, (struct timezone *)NULL); |
103a8fea | 5797 | timersub(&tv_odd, &tv_even, &tv_delta); |
ba3dec99 LB |
5798 | if (for_all_cpus_2(delta_cpu, ODD_COUNTERS, EVEN_COUNTERS)) |
5799 | fprintf(outf, "%s: Counter reset detected\n", progname); | |
5800 | else { | |
5801 | compute_average(EVEN_COUNTERS); | |
5802 | format_all_counters(EVEN_COUNTERS); | |
5803 | } | |
103a8fea | 5804 | |
b7d8c148 LB |
5805 | fprintf(outf, "%.6f sec\n", tv_delta.tv_sec + tv_delta.tv_usec/1000000.0); |
5806 | ||
5807 | flush_output_stderr(); | |
103a8fea | 5808 | |
d91bb17c | 5809 | return status; |
103a8fea LB |
5810 | } |
5811 | ||
3b4d5c7f AS |
5812 | int get_and_dump_counters(void) |
5813 | { | |
5814 | int status; | |
5815 | ||
218f0e8d | 5816 | snapshot_proc_sysfs_files(); |
3b4d5c7f AS |
5817 | status = for_all_cpus(get_counters, ODD_COUNTERS); |
5818 | if (status) | |
5819 | return status; | |
5820 | ||
5821 | status = for_all_cpus(dump_counters, ODD_COUNTERS); | |
5822 | if (status) | |
5823 | return status; | |
5824 | ||
b7d8c148 | 5825 | flush_output_stdout(); |
3b4d5c7f AS |
5826 | |
5827 | return status; | |
5828 | } | |
5829 | ||
d8af6f5f | 5830 | void print_version() { |
3e9fa998 | 5831 | fprintf(outf, "turbostat version 20.09.30" |
d8af6f5f LB |
5832 | " - Len Brown <lenb@kernel.org>\n"); |
5833 | } | |
5834 | ||
495c7654 LB |
5835 | int add_counter(unsigned int msr_num, char *path, char *name, |
5836 | unsigned int width, enum counter_scope scope, | |
41618e63 | 5837 | enum counter_type type, enum counter_format format, int flags) |
388e9c81 LB |
5838 | { |
5839 | struct msr_counter *msrp; | |
5840 | ||
5841 | msrp = calloc(1, sizeof(struct msr_counter)); | |
5842 | if (msrp == NULL) { | |
5843 | perror("calloc"); | |
5844 | exit(1); | |
5845 | } | |
5846 | ||
5847 | msrp->msr_num = msr_num; | |
d8d005ba | 5848 | strncpy(msrp->name, name, NAME_BYTES - 1); |
495c7654 | 5849 | if (path) |
d8d005ba | 5850 | strncpy(msrp->path, path, PATH_BYTES - 1); |
388e9c81 LB |
5851 | msrp->width = width; |
5852 | msrp->type = type; | |
5853 | msrp->format = format; | |
41618e63 | 5854 | msrp->flags = flags; |
388e9c81 LB |
5855 | |
5856 | switch (scope) { | |
5857 | ||
5858 | case SCOPE_CPU: | |
388e9c81 LB |
5859 | msrp->next = sys.tp; |
5860 | sys.tp = msrp; | |
678a3bd1 | 5861 | sys.added_thread_counters++; |
0748eaf0 | 5862 | if (sys.added_thread_counters > MAX_ADDED_THREAD_COUNTERS) { |
678a3bd1 LB |
5863 | fprintf(stderr, "exceeded max %d added thread counters\n", |
5864 | MAX_ADDED_COUNTERS); | |
5865 | exit(-1); | |
5866 | } | |
388e9c81 LB |
5867 | break; |
5868 | ||
5869 | case SCOPE_CORE: | |
388e9c81 LB |
5870 | msrp->next = sys.cp; |
5871 | sys.cp = msrp; | |
678a3bd1 LB |
5872 | sys.added_core_counters++; |
5873 | if (sys.added_core_counters > MAX_ADDED_COUNTERS) { | |
5874 | fprintf(stderr, "exceeded max %d added core counters\n", | |
5875 | MAX_ADDED_COUNTERS); | |
5876 | exit(-1); | |
5877 | } | |
388e9c81 LB |
5878 | break; |
5879 | ||
5880 | case SCOPE_PACKAGE: | |
388e9c81 LB |
5881 | msrp->next = sys.pp; |
5882 | sys.pp = msrp; | |
678a3bd1 LB |
5883 | sys.added_package_counters++; |
5884 | if (sys.added_package_counters > MAX_ADDED_COUNTERS) { | |
5885 | fprintf(stderr, "exceeded max %d added package counters\n", | |
5886 | MAX_ADDED_COUNTERS); | |
5887 | exit(-1); | |
5888 | } | |
388e9c81 LB |
5889 | break; |
5890 | } | |
5891 | ||
5892 | return 0; | |
5893 | } | |
5894 | ||
5895 | void parse_add_command(char *add_command) | |
5896 | { | |
5897 | int msr_num = 0; | |
495c7654 | 5898 | char *path = NULL; |
0f47c08d | 5899 | char name_buffer[NAME_BYTES] = ""; |
388e9c81 LB |
5900 | int width = 64; |
5901 | int fail = 0; | |
5902 | enum counter_scope scope = SCOPE_CPU; | |
5903 | enum counter_type type = COUNTER_CYCLES; | |
5904 | enum counter_format format = FORMAT_DELTA; | |
5905 | ||
5906 | while (add_command) { | |
5907 | ||
5908 | if (sscanf(add_command, "msr0x%x", &msr_num) == 1) | |
5909 | goto next; | |
5910 | ||
5911 | if (sscanf(add_command, "msr%d", &msr_num) == 1) | |
5912 | goto next; | |
5913 | ||
495c7654 LB |
5914 | if (*add_command == '/') { |
5915 | path = add_command; | |
5916 | goto next; | |
5917 | } | |
5918 | ||
388e9c81 LB |
5919 | if (sscanf(add_command, "u%d", &width) == 1) { |
5920 | if ((width == 32) || (width == 64)) | |
5921 | goto next; | |
5922 | width = 64; | |
5923 | } | |
5924 | if (!strncmp(add_command, "cpu", strlen("cpu"))) { | |
5925 | scope = SCOPE_CPU; | |
5926 | goto next; | |
5927 | } | |
5928 | if (!strncmp(add_command, "core", strlen("core"))) { | |
5929 | scope = SCOPE_CORE; | |
5930 | goto next; | |
5931 | } | |
5932 | if (!strncmp(add_command, "package", strlen("package"))) { | |
5933 | scope = SCOPE_PACKAGE; | |
5934 | goto next; | |
5935 | } | |
5936 | if (!strncmp(add_command, "cycles", strlen("cycles"))) { | |
5937 | type = COUNTER_CYCLES; | |
5938 | goto next; | |
5939 | } | |
5940 | if (!strncmp(add_command, "seconds", strlen("seconds"))) { | |
5941 | type = COUNTER_SECONDS; | |
5942 | goto next; | |
5943 | } | |
41618e63 LB |
5944 | if (!strncmp(add_command, "usec", strlen("usec"))) { |
5945 | type = COUNTER_USEC; | |
5946 | goto next; | |
5947 | } | |
388e9c81 LB |
5948 | if (!strncmp(add_command, "raw", strlen("raw"))) { |
5949 | format = FORMAT_RAW; | |
5950 | goto next; | |
5951 | } | |
5952 | if (!strncmp(add_command, "delta", strlen("delta"))) { | |
5953 | format = FORMAT_DELTA; | |
5954 | goto next; | |
5955 | } | |
5956 | if (!strncmp(add_command, "percent", strlen("percent"))) { | |
5957 | format = FORMAT_PERCENT; | |
5958 | goto next; | |
5959 | } | |
5960 | ||
5961 | if (sscanf(add_command, "%18s,%*s", name_buffer) == 1) { /* 18 < NAME_BYTES */ | |
5962 | char *eos; | |
5963 | ||
5964 | eos = strchr(name_buffer, ','); | |
5965 | if (eos) | |
5966 | *eos = '\0'; | |
5967 | goto next; | |
5968 | } | |
5969 | ||
5970 | next: | |
5971 | add_command = strchr(add_command, ','); | |
495c7654 LB |
5972 | if (add_command) { |
5973 | *add_command = '\0'; | |
388e9c81 | 5974 | add_command++; |
495c7654 | 5975 | } |
388e9c81 LB |
5976 | |
5977 | } | |
495c7654 LB |
5978 | if ((msr_num == 0) && (path == NULL)) { |
5979 | fprintf(stderr, "--add: (msrDDD | msr0xXXX | /path_to_counter ) required\n"); | |
388e9c81 LB |
5980 | fail++; |
5981 | } | |
5982 | ||
5983 | /* generate default column header */ | |
5984 | if (*name_buffer == '\0') { | |
5f3aea57 LB |
5985 | if (width == 32) |
5986 | sprintf(name_buffer, "M0x%x%s", msr_num, format == FORMAT_PERCENT ? "%" : ""); | |
5987 | else | |
5988 | sprintf(name_buffer, "M0X%x%s", msr_num, format == FORMAT_PERCENT ? "%" : ""); | |
388e9c81 LB |
5989 | } |
5990 | ||
41618e63 | 5991 | if (add_counter(msr_num, path, name_buffer, width, scope, type, format, 0)) |
388e9c81 LB |
5992 | fail++; |
5993 | ||
5994 | if (fail) { | |
5995 | help(); | |
5996 | exit(1); | |
5997 | } | |
5998 | } | |
41618e63 | 5999 | |
dd778a5e LB |
6000 | int is_deferred_skip(char *name) |
6001 | { | |
6002 | int i; | |
6003 | ||
6004 | for (i = 0; i < deferred_skip_index; ++i) | |
6005 | if (!strcmp(name, deferred_skip_names[i])) | |
6006 | return 1; | |
6007 | return 0; | |
6008 | } | |
6009 | ||
41618e63 LB |
6010 | void probe_sysfs(void) |
6011 | { | |
6012 | char path[64]; | |
6013 | char name_buf[16]; | |
6014 | FILE *input; | |
6015 | int state; | |
6016 | char *sp; | |
6017 | ||
6018 | if (!DO_BIC(BIC_sysfs)) | |
6019 | return; | |
6020 | ||
0748eaf0 | 6021 | for (state = 10; state >= 0; --state) { |
41618e63 LB |
6022 | |
6023 | sprintf(path, "/sys/devices/system/cpu/cpu%d/cpuidle/state%d/name", | |
6024 | base_cpu, state); | |
6025 | input = fopen(path, "r"); | |
6026 | if (input == NULL) | |
6027 | continue; | |
8173c336 BH |
6028 | if (!fgets(name_buf, sizeof(name_buf), input)) |
6029 | err(1, "%s: failed to read file", path); | |
41618e63 LB |
6030 | |
6031 | /* truncate "C1-HSW\n" to "C1", or truncate "C1\n" to "C1" */ | |
6032 | sp = strchr(name_buf, '-'); | |
6033 | if (!sp) | |
6034 | sp = strchrnul(name_buf, '\n'); | |
6035 | *sp = '%'; | |
6036 | *(sp + 1) = '\0'; | |
6037 | ||
fecb3bc8 DA |
6038 | remove_underbar(name_buf); |
6039 | ||
41618e63 LB |
6040 | fclose(input); |
6041 | ||
6042 | sprintf(path, "cpuidle/state%d/time", state); | |
6043 | ||
dd778a5e LB |
6044 | if (is_deferred_skip(name_buf)) |
6045 | continue; | |
6046 | ||
41618e63 LB |
6047 | add_counter(0, path, name_buf, 64, SCOPE_CPU, COUNTER_USEC, |
6048 | FORMAT_PERCENT, SYSFS_PERCPU); | |
6049 | } | |
6050 | ||
0748eaf0 | 6051 | for (state = 10; state >= 0; --state) { |
41618e63 LB |
6052 | |
6053 | sprintf(path, "/sys/devices/system/cpu/cpu%d/cpuidle/state%d/name", | |
6054 | base_cpu, state); | |
6055 | input = fopen(path, "r"); | |
6056 | if (input == NULL) | |
6057 | continue; | |
8173c336 BH |
6058 | if (!fgets(name_buf, sizeof(name_buf), input)) |
6059 | err(1, "%s: failed to read file", path); | |
41618e63 LB |
6060 | /* truncate "C1-HSW\n" to "C1", or truncate "C1\n" to "C1" */ |
6061 | sp = strchr(name_buf, '-'); | |
6062 | if (!sp) | |
6063 | sp = strchrnul(name_buf, '\n'); | |
6064 | *sp = '\0'; | |
6065 | fclose(input); | |
fecb3bc8 DA |
6066 | |
6067 | remove_underbar(name_buf); | |
41618e63 LB |
6068 | |
6069 | sprintf(path, "cpuidle/state%d/usage", state); | |
6070 | ||
dd778a5e LB |
6071 | if (is_deferred_skip(name_buf)) |
6072 | continue; | |
6073 | ||
41618e63 LB |
6074 | add_counter(0, path, name_buf, 64, SCOPE_CPU, COUNTER_ITEMS, |
6075 | FORMAT_DELTA, SYSFS_PERCPU); | |
6076 | } | |
6077 | ||
6078 | } | |
6079 | ||
1ef7d21a LB |
6080 | |
6081 | /* | |
6082 | * parse cpuset with following syntax | |
6083 | * 1,2,4..6,8-10 and set bits in cpu_subset | |
6084 | */ | |
6085 | void parse_cpu_command(char *optarg) | |
6086 | { | |
6087 | unsigned int start, end; | |
6088 | char *next; | |
6089 | ||
4e4e1e7c LB |
6090 | if (!strcmp(optarg, "core")) { |
6091 | if (cpu_subset) | |
6092 | goto error; | |
6093 | show_core_only++; | |
6094 | return; | |
6095 | } | |
6096 | if (!strcmp(optarg, "package")) { | |
6097 | if (cpu_subset) | |
6098 | goto error; | |
6099 | show_pkg_only++; | |
6100 | return; | |
6101 | } | |
6102 | if (show_core_only || show_pkg_only) | |
6103 | goto error; | |
6104 | ||
1ef7d21a LB |
6105 | cpu_subset = CPU_ALLOC(CPU_SUBSET_MAXCPUS); |
6106 | if (cpu_subset == NULL) | |
6107 | err(3, "CPU_ALLOC"); | |
6108 | cpu_subset_size = CPU_ALLOC_SIZE(CPU_SUBSET_MAXCPUS); | |
6109 | ||
6110 | CPU_ZERO_S(cpu_subset_size, cpu_subset); | |
6111 | ||
6112 | next = optarg; | |
6113 | ||
6114 | while (next && *next) { | |
6115 | ||
6116 | if (*next == '-') /* no negative cpu numbers */ | |
6117 | goto error; | |
6118 | ||
6119 | start = strtoul(next, &next, 10); | |
6120 | ||
6121 | if (start >= CPU_SUBSET_MAXCPUS) | |
6122 | goto error; | |
6123 | CPU_SET_S(start, cpu_subset_size, cpu_subset); | |
6124 | ||
6125 | if (*next == '\0') | |
6126 | break; | |
6127 | ||
6128 | if (*next == ',') { | |
6129 | next += 1; | |
6130 | continue; | |
6131 | } | |
6132 | ||
6133 | if (*next == '-') { | |
6134 | next += 1; /* start range */ | |
6135 | } else if (*next == '.') { | |
6136 | next += 1; | |
6137 | if (*next == '.') | |
6138 | next += 1; /* start range */ | |
6139 | else | |
6140 | goto error; | |
6141 | } | |
6142 | ||
6143 | end = strtoul(next, &next, 10); | |
6144 | if (end <= start) | |
6145 | goto error; | |
6146 | ||
6147 | while (++start <= end) { | |
6148 | if (start >= CPU_SUBSET_MAXCPUS) | |
6149 | goto error; | |
6150 | CPU_SET_S(start, cpu_subset_size, cpu_subset); | |
6151 | } | |
6152 | ||
6153 | if (*next == ',') | |
6154 | next += 1; | |
6155 | else if (*next != '\0') | |
6156 | goto error; | |
6157 | } | |
6158 | ||
6159 | return; | |
6160 | ||
6161 | error: | |
4e4e1e7c LB |
6162 | fprintf(stderr, "\"--cpu %s\" malformed\n", optarg); |
6163 | help(); | |
1ef7d21a LB |
6164 | exit(-1); |
6165 | } | |
6166 | ||
812db3f7 | 6167 | |
103a8fea LB |
6168 | void cmdline(int argc, char **argv) |
6169 | { | |
6170 | int opt; | |
d8af6f5f LB |
6171 | int option_index = 0; |
6172 | static struct option long_options[] = { | |
388e9c81 | 6173 | {"add", required_argument, 0, 'a'}, |
1ef7d21a | 6174 | {"cpu", required_argument, 0, 'c'}, |
d8af6f5f | 6175 | {"Dump", no_argument, 0, 'D'}, |
96e47158 | 6176 | {"debug", no_argument, 0, 'd'}, /* internal, not documented */ |
3f44a5c6 | 6177 | {"enable", required_argument, 0, 'e'}, |
d8af6f5f | 6178 | {"interval", required_argument, 0, 'i'}, |
2af4f9b8 | 6179 | {"IPC", no_argument, 0, 'I'}, |
023fe0ac | 6180 | {"num_iterations", required_argument, 0, 'n'}, |
d8af6f5f | 6181 | {"help", no_argument, 0, 'h'}, |
812db3f7 | 6182 | {"hide", required_argument, 0, 'H'}, // meh, -h taken by --help |
d8af6f5f | 6183 | {"Joules", no_argument, 0, 'J'}, |
c8ade361 | 6184 | {"list", no_argument, 0, 'l'}, |
b7d8c148 | 6185 | {"out", required_argument, 0, 'o'}, |
96e47158 | 6186 | {"quiet", no_argument, 0, 'q'}, |
812db3f7 | 6187 | {"show", required_argument, 0, 's'}, |
d8af6f5f LB |
6188 | {"Summary", no_argument, 0, 'S'}, |
6189 | {"TCC", required_argument, 0, 'T'}, | |
6190 | {"version", no_argument, 0, 'v' }, | |
6191 | {0, 0, 0, 0 } | |
6192 | }; | |
103a8fea LB |
6193 | |
6194 | progname = argv[0]; | |
6195 | ||
023fe0ac | 6196 | while ((opt = getopt_long_only(argc, argv, "+C:c:Dde:hi:Jn:o:qST:v", |
d8af6f5f | 6197 | long_options, &option_index)) != -1) { |
103a8fea | 6198 | switch (opt) { |
388e9c81 LB |
6199 | case 'a': |
6200 | parse_add_command(optarg); | |
6201 | break; | |
1ef7d21a LB |
6202 | case 'c': |
6203 | parse_cpu_command(optarg); | |
6204 | break; | |
d8af6f5f | 6205 | case 'D': |
3b4d5c7f AS |
6206 | dump_only++; |
6207 | break; | |
3f44a5c6 LB |
6208 | case 'e': |
6209 | /* --enable specified counter */ | |
4c2122d4 | 6210 | bic_enabled = bic_enabled | bic_lookup(optarg, SHOW_LIST); |
3f44a5c6 | 6211 | break; |
d8af6f5f LB |
6212 | case 'd': |
6213 | debug++; | |
3f44a5c6 | 6214 | ENABLE_BIC(BIC_DISABLED_BY_DEFAULT); |
103a8fea | 6215 | break; |
812db3f7 | 6216 | case 'H': |
3f44a5c6 LB |
6217 | /* |
6218 | * --hide: do not show those specified | |
6219 | * multiple invocations simply clear more bits in enabled mask | |
6220 | */ | |
6221 | bic_enabled &= ~bic_lookup(optarg, HIDE_LIST); | |
812db3f7 | 6222 | break; |
d8af6f5f LB |
6223 | case 'h': |
6224 | default: | |
6225 | help(); | |
6226 | exit(1); | |
103a8fea | 6227 | case 'i': |
2a0609c0 LB |
6228 | { |
6229 | double interval = strtod(optarg, NULL); | |
6230 | ||
6231 | if (interval < 0.001) { | |
b7d8c148 | 6232 | fprintf(outf, "interval %f seconds is too small\n", |
2a0609c0 LB |
6233 | interval); |
6234 | exit(2); | |
6235 | } | |
6236 | ||
47936f94 | 6237 | interval_tv.tv_sec = interval_ts.tv_sec = interval; |
b9ad8ee0 | 6238 | interval_tv.tv_usec = (interval - interval_tv.tv_sec) * 1000000; |
47936f94 | 6239 | interval_ts.tv_nsec = (interval - interval_ts.tv_sec) * 1000000000; |
2a0609c0 | 6240 | } |
103a8fea | 6241 | break; |
d8af6f5f LB |
6242 | case 'J': |
6243 | rapl_joules++; | |
8e180f3c | 6244 | break; |
c8ade361 | 6245 | case 'l': |
3f44a5c6 | 6246 | ENABLE_BIC(BIC_DISABLED_BY_DEFAULT); |
c8ade361 LB |
6247 | list_header_only++; |
6248 | quiet++; | |
6249 | break; | |
b7d8c148 LB |
6250 | case 'o': |
6251 | outf = fopen_or_die(optarg, "w"); | |
6252 | break; | |
96e47158 LB |
6253 | case 'q': |
6254 | quiet = 1; | |
6255 | break; | |
023fe0ac CY |
6256 | case 'n': |
6257 | num_iterations = strtod(optarg, NULL); | |
6258 | ||
6259 | if (num_iterations <= 0) { | |
6260 | fprintf(outf, "iterations %d should be positive number\n", | |
6261 | num_iterations); | |
6262 | exit(2); | |
6263 | } | |
6264 | break; | |
812db3f7 | 6265 | case 's': |
3f44a5c6 LB |
6266 | /* |
6267 | * --show: show only those specified | |
6268 | * The 1st invocation will clear and replace the enabled mask | |
6269 | * subsequent invocations can add to it. | |
6270 | */ | |
6271 | if (shown == 0) | |
6272 | bic_enabled = bic_lookup(optarg, SHOW_LIST); | |
6273 | else | |
6274 | bic_enabled |= bic_lookup(optarg, SHOW_LIST); | |
6275 | shown = 1; | |
812db3f7 | 6276 | break; |
d8af6f5f LB |
6277 | case 'S': |
6278 | summary_only++; | |
889facbe LB |
6279 | break; |
6280 | case 'T': | |
6281 | tcc_activation_temp_override = atoi(optarg); | |
6282 | break; | |
d8af6f5f LB |
6283 | case 'v': |
6284 | print_version(); | |
6285 | exit(0); | |
5c56be9a | 6286 | break; |
103a8fea LB |
6287 | } |
6288 | } | |
6289 | } | |
6290 | ||
6291 | int main(int argc, char **argv) | |
6292 | { | |
b7d8c148 | 6293 | outf = stderr; |
103a8fea LB |
6294 | cmdline(argc, argv); |
6295 | ||
96e47158 | 6296 | if (!quiet) |
d8af6f5f | 6297 | print_version(); |
103a8fea | 6298 | |
41618e63 LB |
6299 | probe_sysfs(); |
6300 | ||
103a8fea LB |
6301 | turbostat_init(); |
6302 | ||
3b4d5c7f AS |
6303 | /* dump counters and exit */ |
6304 | if (dump_only) | |
6305 | return get_and_dump_counters(); | |
6306 | ||
c8ade361 LB |
6307 | /* list header and exit */ |
6308 | if (list_header_only) { | |
6309 | print_header(","); | |
6310 | flush_output_stdout(); | |
6311 | return 0; | |
6312 | } | |
6313 | ||
9972d5d8 | 6314 | msr_sum_record(); |
103a8fea LB |
6315 | /* |
6316 | * if any params left, it must be a command to fork | |
6317 | */ | |
6318 | if (argc - optind) | |
6319 | return fork_it(argv + optind); | |
6320 | else | |
6321 | turbostat_loop(); | |
6322 | ||
6323 | return 0; | |
6324 | } |