From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-3.1 required=3.0 tests=DKIM_SIGNED,DKIM_VALID, DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS,MAILING_LIST_MULTI,SPF_PASS, URIBL_BLOCKED,USER_AGENT_GIT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 85AE4C4321D for ; Fri, 17 Aug 2018 16:35:07 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 2A00021A28 for ; Fri, 17 Aug 2018 16:35:07 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (1024-bit key) header.d=kepstin.ca header.i=@kepstin.ca header.b="nPaz6fyC" DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 2A00021A28 Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=kepstin.ca Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=linux-kernel-owner@vger.kernel.org Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1727941AbeHQTjH (ORCPT ); Fri, 17 Aug 2018 15:39:07 -0400 Received: from mail-it0-f65.google.com ([209.85.214.65]:36766 "EHLO mail-it0-f65.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1727804AbeHQTjF (ORCPT ); Fri, 17 Aug 2018 15:39:05 -0400 Received: by mail-it0-f65.google.com with SMTP id p81-v6so12522220itp.1 for ; Fri, 17 Aug 2018 09:35:02 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=kepstin.ca; s=google; h=from:to:cc:subject:date:message-id:in-reply-to:references; bh=nvUh0UDB9l+yHlWfA39LQUjHQrfWLjEkEyq8WfNmIik=; b=nPaz6fyC2kIfmdj6cnYTB134nhdstzacyfRB/ZuTUahbXSnjT+ATkrmnosZP+hhbLQ SNr+PcqlNIXxqKBK6mCFMTGLQfjCs/JybBRoRZlT3hpAsn6RstMuq4Vjq3Pd3ioVuU6F 7uQrStZbroHbCGryCXNw62kBEDNGdQL3jGKII= X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references; bh=nvUh0UDB9l+yHlWfA39LQUjHQrfWLjEkEyq8WfNmIik=; b=Xcbt5Sqspiq92SsznuGl1Hn4nu+9JSRnTE80324KzPFs9708gmwUhH4ofzFFe69u8V 8S6C6mQfSJ/Gq90KDWx/O9akaigRdDJ1ZdwCb148icEyeVPz+Ofg+mzod38zVBOdXHhA LAA+oRPTls1iTD6j5Fo+LPkMwg/A8yMcN+JgZ34fi1fsks10avnbDtzIIR1F+KngsBlM KDG5j4ihawzVmosh9KdYXU4AvIPuRscrw3JAt9ddmj6f8IJptUlU8gdHhfedU0th+c39 GnhN9+KqoobNNJxOJH/lOSxMh0Pk3aQLW1YT/MnLp8MbHmZ25gWaNAHXD5DTEXm1i/uZ Rovw== X-Gm-Message-State: AOUpUlHg9i/+DHym+riZ7wXphMmFJoQ/BsylZwKLoGsZQ2i/kv4n9PUe 16Wxc+/HGg9v7GE+xuKo9Lw5sg== X-Google-Smtp-Source: AA+uWPyKJYe8dx5crE4iE8DVRH5Y/ltERl+TqUmaipFVLYOfXO+zPVlkaAEBQcn6NlGJIFAOAyXo6A== X-Received: by 2002:a24:6e95:: with SMTP id w143-v6mr4683115itc.34.1534523702123; Fri, 17 Aug 2018 09:35:02 -0700 (PDT) Received: from fuko.kepstin.ca ([2001:470:b3f4::561]) by smtp.gmail.com with ESMTPSA id e140-v6sm1947974itc.24.2018.08.17.09.35.00 (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Fri, 17 Aug 2018 09:35:01 -0700 (PDT) From: Calvin Walton To: Len Brown , linux-pm@vger.kernel.org, linux-kernel@vger.kernel.org Cc: Calvin Walton Subject: [PATCH v3 1/2] tools/power turbostat: Add support for AMD Fam 17h (Zen) RAPL Date: Fri, 17 Aug 2018 12:34:41 -0400 Message-Id: <20180817163442.10065-2-calvin.walton@kepstin.ca> X-Mailer: git-send-email 2.18.0 In-Reply-To: <20180817163442.10065-1-calvin.walton@kepstin.ca> References: <20180817163442.10065-1-calvin.walton@kepstin.ca> Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Based on the Open-Source Register Reference for AMD Family 17h Processors Models 00h-2Fh: https://support.amd.com/TechDocs/56255_OSRR.pdf These processors report RAPL support in bit 14 of CPUID 0x80000007 EDX, and the following MSRs are present: 0xc0010299 (RAPL_PWR_UNIT), like Intel's RAPL_POWER_UNIT 0xc001029a (CORE_ENERGY_STAT), kind of like Intel's PP0_ENERGY_STATUS 0xc001029b (PKG_ENERGY_STAT), like Intel's PKG_ENERGY_STATUS A notable difference from the Intel implementation is that AMD reports the "Cores" energy usage separately for each core, rather than a per-package total. The code has been adjusted to handle either case in a generic way. I haven't yet enabled collection of package power, due to being unable to test it on multi-node systems (TR, EPYC). Signed-off-by: Calvin Walton --- tools/power/x86/turbostat/turbostat.c | 159 +++++++++++++++++++++----- 1 file changed, 130 insertions(+), 29 deletions(-) diff --git a/tools/power/x86/turbostat/turbostat.c b/tools/power/x86/turbostat/turbostat.c index 6837a65409ba..89d4e2e75774 100644 --- a/tools/power/x86/turbostat/turbostat.c +++ b/tools/power/x86/turbostat/turbostat.c @@ -44,6 +44,7 @@ #include #include #include +#include char *proc_stat = "/proc/stat"; FILE *outf; @@ -141,9 +142,21 @@ unsigned int first_counter_read = 1; #define RAPL_CORES_ENERGY_STATUS (1 << 9) /* 0x639 MSR_PP0_ENERGY_STATUS */ +#define RAPL_PER_CORE_ENERGY (1 << 10) + /* Indicates cores energy collection is per-core, + * not per-package. */ +#define RAPL_AMD_F17H (1 << 11) + /* 0xc0010299 MSR_RAPL_PWR_UNIT */ + /* 0xc001029a MSR_CORE_ENERGY_STAT */ + /* 0xc001029b MSR_PKG_ENERGY_STAT */ #define RAPL_CORES (RAPL_CORES_ENERGY_STATUS | RAPL_CORES_POWER_LIMIT) #define TJMAX_DEFAULT 100 +/* MSRs that are not yet in the kernel-provided header. */ +#define MSR_RAPL_PWR_UNIT 0xc0010299 +#define MSR_CORE_ENERGY_STAT 0xc001029a +#define MSR_PKG_ENERGY_STAT 0xc001029b + #define MAX(a, b) ((a) > (b) ? (a) : (b)) /* @@ -187,6 +200,7 @@ struct core_data { unsigned long long c7; unsigned long long mc6_us; /* duplicate as per-core for now, even though per module */ unsigned int core_temp_c; + unsigned int core_energy; /* MSR_CORE_ENERGY_STAT */ unsigned int core_id; unsigned long long counter[MAX_ADDED_COUNTERS]; } *core_even, *core_odd; @@ -680,6 +694,14 @@ void print_header(char *delim) if (DO_BIC(BIC_CoreTmp)) outp += sprintf(outp, "%sCoreTmp", (printed++ ? delim : "")); + if (do_rapl && !rapl_joules) { + if (DO_BIC(BIC_CorWatt) && (do_rapl & RAPL_PER_CORE_ENERGY)) + outp += sprintf(outp, "%sCorWatt", (printed++ ? delim : "")); + } else if (do_rapl && rapl_joules) { + if (DO_BIC(BIC_Cor_J) && (do_rapl & RAPL_PER_CORE_ENERGY)) + outp += sprintf(outp, "%sCor_J", (printed++ ? delim : "")); + } + for (mp = sys.cp; mp; mp = mp->next) { if (mp->format == FORMAT_RAW) { if (mp->width == 64) @@ -734,7 +756,7 @@ void print_header(char *delim) if (do_rapl && !rapl_joules) { if (DO_BIC(BIC_PkgWatt)) outp += sprintf(outp, "%sPkgWatt", (printed++ ? delim : "")); - if (DO_BIC(BIC_CorWatt)) + if (DO_BIC(BIC_CorWatt) && !(do_rapl & RAPL_PER_CORE_ENERGY)) outp += sprintf(outp, "%sCorWatt", (printed++ ? delim : "")); if (DO_BIC(BIC_GFXWatt)) outp += sprintf(outp, "%sGFXWatt", (printed++ ? delim : "")); @@ -747,7 +769,7 @@ void print_header(char *delim) } else if (do_rapl && rapl_joules) { if (DO_BIC(BIC_Pkg_J)) outp += sprintf(outp, "%sPkg_J", (printed++ ? delim : "")); - if (DO_BIC(BIC_Cor_J)) + if (DO_BIC(BIC_Cor_J) && !(do_rapl & RAPL_PER_CORE_ENERGY)) outp += sprintf(outp, "%sCor_J", (printed++ ? delim : "")); if (DO_BIC(BIC_GFX_J)) outp += sprintf(outp, "%sGFX_J", (printed++ ? delim : "")); @@ -808,6 +830,7 @@ int dump_counters(struct thread_data *t, struct core_data *c, outp += sprintf(outp, "c6: %016llX\n", c->c6); outp += sprintf(outp, "c7: %016llX\n", c->c7); outp += sprintf(outp, "DTS: %dC\n", c->core_temp_c); + outp += sprintf(outp, "Joules: %0X\n", c->core_energy); for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) { outp += sprintf(outp, "cADDED [%d] msr0x%x: %08llX\n", @@ -1033,6 +1056,20 @@ int format_counters(struct thread_data *t, struct core_data *c, } } + /* + * If measurement interval exceeds minimum RAPL Joule Counter range, + * indicate that results are suspect by printing "**" in fraction place. + */ + if (interval_float < rapl_joule_counter_range) + fmt8 = "%s%.2f"; + else + fmt8 = "%6.0f**"; + + if (DO_BIC(BIC_CorWatt) && (do_rapl & RAPL_PER_CORE_ENERGY)) + outp += sprintf(outp, fmt8, (printed++ ? delim : ""), c->core_energy * rapl_energy_units / interval_float); + if (DO_BIC(BIC_Cor_J) && (do_rapl & RAPL_PER_CORE_ENERGY)) + outp += sprintf(outp, fmt8, (printed++ ? delim : ""), c->core_energy * rapl_energy_units); + /* print per-package data only for 1st core in package */ if (!(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE)) goto done; @@ -1085,18 +1122,9 @@ int format_counters(struct thread_data *t, struct core_data *c, if (DO_BIC(BIC_SYS_LPI)) outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->sys_lpi / 1000000.0 / interval_float); - /* - * If measurement interval exceeds minimum RAPL Joule Counter range, - * indicate that results are suspect by printing "**" in fraction place. - */ - if (interval_float < rapl_joule_counter_range) - fmt8 = "%s%.2f"; - else - fmt8 = "%6.0f**"; - if (DO_BIC(BIC_PkgWatt)) outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_pkg * rapl_energy_units / interval_float); - if (DO_BIC(BIC_CorWatt)) + if (DO_BIC(BIC_CorWatt) && !(do_rapl & RAPL_PER_CORE_ENERGY)) outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_cores * rapl_energy_units / interval_float); if (DO_BIC(BIC_GFXWatt)) outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_gfx * rapl_energy_units / interval_float); @@ -1104,7 +1132,7 @@ int format_counters(struct thread_data *t, struct core_data *c, outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_dram * rapl_dram_energy_units / interval_float); if (DO_BIC(BIC_Pkg_J)) outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_pkg * rapl_energy_units); - if (DO_BIC(BIC_Cor_J)) + if (DO_BIC(BIC_Cor_J) && !(do_rapl & RAPL_PER_CORE_ENERGY)) outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_cores * rapl_energy_units); if (DO_BIC(BIC_GFX_J)) outp += sprintf(outp, fmt8, (printed++ ? delim : ""), p->energy_gfx * rapl_energy_units); @@ -1249,6 +1277,8 @@ delta_core(struct core_data *new, struct core_data *old) old->core_temp_c = new->core_temp_c; old->mc6_us = new->mc6_us - old->mc6_us; + DELTA_WRAP32(new->core_energy, old->core_energy); + for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) { if (mp->format == FORMAT_RAW) old->counter[i] = new->counter[i]; @@ -1391,6 +1421,7 @@ void clear_counters(struct thread_data *t, struct core_data *c, struct pkg_data c->c7 = 0; c->mc6_us = 0; c->core_temp_c = 0; + c->core_energy = 0; p->pkg_wtd_core_c0 = 0; p->pkg_any_core_c0 = 0; @@ -1473,6 +1504,8 @@ int sum_counters(struct thread_data *t, struct core_data *c, average.cores.core_temp_c = MAX(average.cores.core_temp_c, c->core_temp_c); + average.cores.core_energy += c->core_energy; + for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) { if (mp->format == FORMAT_RAW) continue; @@ -1845,6 +1878,12 @@ int get_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p) c->core_temp_c = tcc_activation_temp - ((msr >> 16) & 0x7F); } + if (do_rapl & RAPL_AMD_F17H) { + if (get_msr(cpu, MSR_CORE_ENERGY_STAT, &msr)) + return -14; + c->core_energy = msr & 0xFFFFFFFF; + } + for (i = 0, mp = sys.cp; mp; i++, mp = mp->next) { if (get_mp(cpu, mp, &c->counter[i])) return -10; @@ -3738,7 +3777,7 @@ int print_perf_limit(struct thread_data *t, struct core_data *c, struct pkg_data #define RAPL_POWER_GRANULARITY 0x7FFF /* 15 bit power granularity */ #define RAPL_TIME_GRANULARITY 0x3F /* 6 bit time granularity */ -double get_tdp(unsigned int model) +double get_tdp_intel(unsigned int model) { unsigned long long msr; @@ -3755,6 +3794,16 @@ double get_tdp(unsigned int model) } } +double get_tdp_amd(unsigned int family) +{ + switch (family) { + case 0x17: + default: + /* This is the max stock TDP of HEDT/Server Fam17h chips */ + return 250.0; + } +} + /* * rapl_dram_energy_units_probe() * Energy units are either hard-coded, or come from RAPL Energy Unit MSR. @@ -3776,21 +3825,12 @@ rapl_dram_energy_units_probe(int model, double rapl_energy_units) } } - -/* - * rapl_probe() - * - * sets do_rapl, rapl_power_units, rapl_energy_units, rapl_time_units - */ -void rapl_probe(unsigned int family, unsigned int model) +void rapl_probe_intel(unsigned int family, unsigned int model) { unsigned long long msr; unsigned int time_unit; double tdp; - if (!genuine_intel) - return; - if (family != 6) return; @@ -3920,13 +3960,66 @@ void rapl_probe(unsigned int family, unsigned int model) rapl_time_units = 1.0 / (1 << (time_unit)); - tdp = get_tdp(model); + tdp = get_tdp_intel(model); + + rapl_joule_counter_range = 0xFFFFFFFF * rapl_energy_units / tdp; + if (!quiet) + fprintf(outf, "RAPL: %.0f sec. Joule Counter Range, at %.0f Watts\n", rapl_joule_counter_range, tdp); +} + +void rapl_probe_amd(unsigned int family, unsigned int model) +{ + unsigned long long msr; + unsigned int eax, ebx, ecx, edx; + unsigned int has_rapl = 0; + double tdp; + + if (max_extended_level >= 0x80000007) { + __cpuid(0x80000007, eax, ebx, ecx, edx); + /* RAPL (Fam 17h) */ + has_rapl = edx & (1 << 14); + } + + if (!has_rapl) + return; + + switch (family) { + case 0x17: /* Zen, Zen+ */ + do_rapl = RAPL_AMD_F17H | RAPL_PER_CORE_ENERGY; + if (rapl_joules) + BIC_PRESENT(BIC_Cor_J); + else + BIC_PRESENT(BIC_CorWatt); + break; + default: + return; + } + + if (get_msr(base_cpu, MSR_RAPL_PWR_UNIT, &msr)) + return; + + rapl_time_units = ldexp(1.0, -(msr >> 16 & 0xf)); + rapl_energy_units = ldexp(1.0, -(msr >> 8 & 0x1f)); + rapl_power_units = ldexp(1.0, -(msr & 0xf)); + + tdp = get_tdp_amd(model); rapl_joule_counter_range = 0xFFFFFFFF * rapl_energy_units / tdp; if (!quiet) fprintf(outf, "RAPL: %.0f sec. Joule Counter Range, at %.0f Watts\n", rapl_joule_counter_range, tdp); +} - return; +/* + * rapl_probe() + * + * sets do_rapl, rapl_power_units, rapl_energy_units, rapl_time_units + */ +void rapl_probe(unsigned int family, unsigned int model) +{ + if (genuine_intel) + rapl_probe_intel(family, model); + if (authentic_amd) + rapl_probe_amd(family, model); } void perf_limit_reasons_probe(unsigned int family, unsigned int model) @@ -4032,6 +4125,7 @@ void print_power_limit_msr(int cpu, unsigned long long msr, char *label) int print_rapl(struct thread_data *t, struct core_data *c, struct pkg_data *p) { unsigned long long msr; + const char *msr_name; int cpu; if (!do_rapl) @@ -4047,10 +4141,17 @@ int print_rapl(struct thread_data *t, struct core_data *c, struct pkg_data *p) return -1; } - if (get_msr(cpu, MSR_RAPL_POWER_UNIT, &msr)) - return -1; + if (do_rapl & RAPL_AMD_F17H) { + msr_name = "MSR_RAPL_PWR_UNIT"; + if (get_msr(cpu, MSR_RAPL_PWR_UNIT, &msr)) + return -1; + } else { + msr_name = "MSR_RAPL_POWER_UNIT"; + if (get_msr(cpu, MSR_RAPL_POWER_UNIT, &msr)) + return -1; + } - fprintf(outf, "cpu%d: MSR_RAPL_POWER_UNIT: 0x%08llx (%f Watts, %f Joules, %f sec.)\n", cpu, msr, + fprintf(outf, "cpu%d: %s: 0x%08llx (%f Watts, %f Joules, %f sec.)\n", cpu, msr_name, msr, rapl_power_units, rapl_energy_units, rapl_time_units); if (do_rapl & RAPL_PKG_POWER_INFO) { -- 2.18.0