Debian 3.9.10debian/3.9.10debian

2 files changed, 1041 insertions, 0 deletions

diff --git a/src/pmtop/GNUmakefile b/src/pmtop/GNUmakefile
new file mode 100644
index 0000000..e64aeaf
--- /dev/null
+++ b/src/pmtop/GNUmakefile
@@ -0,0 +1,33 @@
+#
+# Copyright (c) 2007 Silicon Graphics, Inc.  All Rights Reserved.
+# Copyright (c) 2008 Aconex.  All Rights Reserved.
+# 
+# This program is free software; you can redistribute it and/or modify it
+# under the terms of the GNU General Public License as published by the
+# Free Software Foundation; either version 2 of the License, or (at your
+# option) any later version.
+# 
+# This program is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+# or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+# for more details.
+# 
+
+TOPDIR = ../..
+include $(TOPDIR)/src/include/builddefs
+
+CFILES = pmtop.c
+LLDLIBS = $(PCPLIB)
+CMDTARGET = pmtop$(EXECSUFFIX)
+
+default : $(CMDTARGET)
+
+include $(BUILDRULES)
+
+install : default
+	#$(INSTALL) -m 755 $(CMDTARGET) $(PCP_BIN_DIR)/$(CMDTARGET)
+
+default_pcp : default
+
+install_pcp : install
+
diff --git a/src/pmtop/pmtop.c b/src/pmtop/pmtop.c
new file mode 100644
index 0000000..dc10fc0
--- /dev/null
+++ b/src/pmtop/pmtop.c
@@ -0,0 +1,1008 @@
+/*
+ * Copyright (c) 2013-2014 Red Hat.
+ * Copyright (c) 1999 Silicon Graphics, Inc.  All Rights Reserved.
+ * 
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the
+ * Free Software Foundation; either version 2 of the License, or (at your
+ * option) any later version.
+ * 
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ */
+
+#include "pmapi.h"
+#include "impl.h"
+
+/*
+ * Outputs psinfo.psargs and top value.
+ *
+ * Want to periodically write to stdout
+ * with the top processes for the following categories:
+ *
+ * cpuburn - rate converted proc.psusage.utime + stime
+ * syscalls - rate converted proc.psusage.pu_sysc
+ * reads - rate converted proc.psusage.bread + gbread 
+ * writes - rate converted proc.psusage.bwrit + gbwrit
+ * ctxswitch - rate converted proc.psusage.pu_vctx + proc.psusage.pu_ictx
+ * residentsize - proc.psusage.rss
+ * 
+ */
+
+#define MAX_PMID 15 /* gives enough space for all possible metrics in fetch */
+char    *namelist[MAX_PMID];
+pmID	pmidlist[MAX_PMID];
+int	type_tab[MAX_PMID]; /* the types of the metrics */
+int	valfmt_tab[MAX_PMID]; /* the types of the metrics */
+int	num_pmid;
+int	num_proc_pmid;
+
+/* indexes into [MAX_PMID] arrays */
+int	kernel_utime;
+int	kernel_stime;
+int	kernel_sysc;
+int	kernel_ctx;
+int	pu_utime;
+int	pu_stime;
+int	pu_sysc;
+int	pu_bread;
+int	pu_gbread;
+int	pu_bwrit;
+int	pu_gbwrit;
+int	pu_rss;
+int	pu_vctx;
+int	pu_ictx;
+
+char	*hostname;
+int	top = 5;
+long	num_samples = -1;
+int	num_inst;
+int	*instances;
+char	**instance_names;
+char	*runtime;
+char	pid_label_fmt[8];
+char	pid_fmt[8];
+char	*line_fmt = "%.79s";
+int	pid_len;
+pmInDom proc_indom;
+
+/* values of interest */
+#define CPU_VAL 	0
+#define SYSCALLS_VAL	1
+#define CTX_VAL		2
+#define WRITES_VAL	3
+#define READS_VAL	4
+#define RSS_VAL		5
+#define USR_VAL		6
+#define SYS_VAL		7
+#define NUM_VALUES 	8
+
+/* which values to calculate and print */
+#define SHOW_CPU	(1<<CPU_VAL)
+#define SHOW_SYSCALLS	(1<<SYSCALLS_VAL)
+#define SHOW_CTX	(1<<CTX_VAL)
+#define SHOW_WRITES	(1<<WRITES_VAL)
+#define SHOW_READS	(1<<READS_VAL)
+#define SHOW_RSS	(1<<RSS_VAL)
+#define SHOW_USR	(1<<USR_VAL)
+#define SHOW_SYS	(1<<SYS_VAL)
+
+int	show_spec = SHOW_CPU|SHOW_SYSCALLS|SHOW_CTX|SHOW_WRITES|SHOW_READS|SHOW_RSS; /* default to show all */
+
+typedef struct {
+    char        *name;
+    int         bit;
+} show_map_t;
+
+static show_map_t show_map[] = {
+    { "CPU",   SHOW_CPU },
+    { "SYSC",  SHOW_SYSCALLS },
+    { "CTX",   SHOW_CTX },
+    { "WRITE", SHOW_WRITES },
+    { "READ",  SHOW_READS },
+    { "RSS",   SHOW_RSS },
+    { "USR",   SHOW_USR },
+    { "SYS",   SHOW_SYS },
+};
+static int num_show = sizeof(show_map) / sizeof(show_map[0]);
+
+typedef struct {
+   int		inst; /* index into instances and names */
+   double	values[NUM_VALUES];
+} rate_entry_t;
+
+static rate_entry_t *rate_tab;
+
+double global_rates[NUM_VALUES];
+double sum_rates[NUM_VALUES];
+
+typedef struct {
+   char		*val_fmt;
+   char		*label_fmt;
+   char		*label;
+   char		*units;
+   char		*fullname;
+} format_entry_t;
+
+/*
+ * Warning - must match order of *_VAL macros
+ */
+format_entry_t format_tab[] =
+{ 
+    {"%7.2f", "%7s", "CPU%", "", "CPU Utilization"},
+    {"%9.0f", "%9s", "SYSCALLS", "sys/sec", "System Calls"},
+    {"%9.0f", "%9s", "CTX", "ctx/sec", "Context Switches"},
+    {"%8.0f", "%8s", "WRITES", "Kb/sec", "Writes"},
+    {"%8.0f", "%8s", "READS", "Kb/sec", "Reads"},
+    {"%9.0f", "%9s", "RSS", "Kb", "Resident Size"},
+    {"%7.2f", "%7s", "USR%", "", "User CPU Utilization"},
+    {"%7.2f", "%7s", "SYS%", "", "System CPU Utilization"},
+};
+
+/* value table */
+typedef struct {
+    int		num;
+    pmValue	values[MAX_PMID][2];
+} val_entry_t;
+
+val_entry_t *val_tab;
+
+/* index will use high end for globals */
+/* parallels index with other MAX_PMID arrays */
+pmValue global_val[MAX_PMID][2]; 
+
+#ifndef min
+#define min(x,y) (((x)<(y))?(x):(y))
+#endif
+
+static void get_indom(void);
+static int overrides(int, pmOptions *);
+
+pmLongOptions longopts[] = {
+    PMAPI_OPTIONS_HEADER("General Options"),
+    PMOPT_HOST,
+    PMOPT_SAMPLES,
+    PMOPT_INTERVAL,
+    PMAPI_OPTIONS_HEADER("Reporting Options"),
+    { "", 1, 'm', "N", "report the top N values only" },
+    { "", 1, 'p', "SPEC", "report certain values e.g. cpu,sysc,rss" },
+    { "wide", 0, 'w', 0, "wide output for command name" },
+    PMOPT_TIMEZONE,
+    PMOPT_HOSTZONE,
+    PMOPT_HELP,
+    PMAPI_OPTIONS_END
+};
+
+pmOptions opts = {
+    .short_options = "D:h:m:p:s:t:wzZ:?",
+    .long_options = longopts,
+    .override = overrides,
+};
+
+static int
+parse_show_spec(const char *spec)
+{
+    int		val = 0;
+    int		tmp;
+    const char	*p;
+    char	*pend;
+    int		i;
+
+    for (p = spec; *p; ) {
+	tmp = (int)strtol(p, &pend, 10);
+	if (tmp == -1)
+	    /* special case ... -1 really means set all the bits! */
+	    tmp = INT_MAX;
+	if (*pend == '\0') {
+	    val |= tmp;
+	    break;
+	}
+	else if (*pend == ',') {
+	    val |= tmp;
+	    p = pend + 1;
+	}
+	else {
+	    pend = strchr(p, ',');
+	    if (pend != NULL)
+		*pend = '\0';
+
+	    for (i = 0; i < num_show; i++) {
+		if (strcasecmp(p, show_map[i].name) == 0) {
+		    val |= show_map[i].bit;
+		    if (pend != NULL) {
+			*pend = ',';
+			p = pend + 1;
+		    }
+		    else
+			p = "";		/* force termination of outer loop */
+		    break;
+		}
+	    }
+
+	    if (i == num_show) {
+		if (pend != NULL)
+		    *pend = ',';
+		return PM_ERR_CONV;
+	    }
+	}
+    }
+
+    return val;
+}
+
+static char *
+get_time(void)
+{
+    static char str[80];
+    time_t now = time(NULL);
+    struct tm *tm = localtime(&now);
+    strftime(str, sizeof(str), "%c %Z", tm);
+    return str;
+}
+
+static __uint64_t
+get_value(int metric, pmValue *value)
+{
+    int sts;
+    pmAtomValue val;
+
+    if ((sts = pmExtractValue(valfmt_tab[metric], value,
+		  type_tab[metric], &val, PM_TYPE_U64)) < 0) {
+	fprintf(stderr, "%s: Failed to extract metric value: %s\n", 
+		pmProgname, pmErrStr(sts));
+	exit(1);
+
+    }
+    return val.ull;
+}
+
+static int
+cpuburn_cmp(const void *x, const void *y)
+{
+    rate_entry_t *c1 = (rate_entry_t *)x;
+    rate_entry_t *c2 = (rate_entry_t *)y;
+
+    if (c1->values[CPU_VAL] < c2->values[CPU_VAL])
+	return -1;
+    if (c2->values[CPU_VAL] > c1->values[CPU_VAL])
+	return 1;
+    return 0;
+}
+
+static int
+sys_cmp(const void* x, const void* y)
+{
+    rate_entry_t *c1 = (rate_entry_t *)x;
+    rate_entry_t *c2 = (rate_entry_t *)y;
+
+    if (c1->values[SYS_VAL] < c2->values[SYS_VAL])
+	return -1;
+    if (c2->values[SYS_VAL] > c1->values[SYS_VAL])
+	return 1;
+    return 0;
+}
+
+static int 
+usr_cmp(const void* x, const void* y)
+{
+    rate_entry_t *c1 = (rate_entry_t *)x;
+    rate_entry_t *c2 = (rate_entry_t *)y;
+
+    if (c1->values[USR_VAL] < c2->values[USR_VAL])
+	return -1;
+    if (c2->values[USR_VAL] > c1->values[USR_VAL])
+	return 1;
+    return 0;
+}
+
+static int
+syscall_cmp(const void* x, const void* y)
+{
+    rate_entry_t *c1 = (rate_entry_t *)x;
+    rate_entry_t *c2 = (rate_entry_t *)y;
+
+    if (c1->values[SYSCALLS_VAL] < c2->values[SYSCALLS_VAL])
+	return -1;
+    if (c2->values[SYSCALLS_VAL] > c1->values[SYSCALLS_VAL])
+	return 1;
+    return 0;
+}
+
+static int
+ctx_cmp(const void* x, const void* y)
+{
+    rate_entry_t *c1 = (rate_entry_t *)x;
+    rate_entry_t *c2 = (rate_entry_t *)y;
+
+    if (c1->values[CTX_VAL] < c2->values[CTX_VAL])
+	return -1;
+    if (c2->values[CTX_VAL] > c1->values[CTX_VAL])
+	return 1;
+    return 0;
+}
+
+static int 
+write_cmp(const void* x, const void* y)
+{
+    rate_entry_t *c1 = (rate_entry_t *)x;
+    rate_entry_t *c2 = (rate_entry_t *)y;
+
+    if (c1->values[WRITES_VAL] < c2->values[WRITES_VAL])
+	return -1;
+    if (c2->values[WRITES_VAL] > c1->values[WRITES_VAL])
+	return 1;
+    return 0;
+}
+
+static int 
+read_cmp(const void* x, const void* y)
+{
+    rate_entry_t *c1 = (rate_entry_t *)x;
+    rate_entry_t *c2 = (rate_entry_t *)y;
+
+    if (c1->values[READS_VAL] < c2->values[READS_VAL])
+	return -1;
+    if (c2->values[READS_VAL] > c1->values[READS_VAL])
+	return 1;
+    return 0;
+}
+
+int 
+rss_cmp(const void* x, const void* y)
+{
+    rate_entry_t *c1 = (rate_entry_t *)x;
+    rate_entry_t *c2 = (rate_entry_t *)y;
+
+    if (c1->values[RSS_VAL] < c2->values[RSS_VAL])
+	return -1;
+    if (c2->values[RSS_VAL] > c1->values[RSS_VAL])
+	return 1;
+    return 0;
+}
+
+/*
+ * sum the rate values for the top entries
+ */
+void
+calc_sum_rates(int num_entries, int val_type)
+{
+    int i;
+    double sum = 0;
+
+    for (i = 0; i < min(top, num_entries); i++) {
+	rate_entry_t *entry = &rate_tab[i];
+	sum += entry->values[val_type];
+    }
+
+    sum_rates[val_type] = sum;
+}
+
+void
+print_top(rate_entry_t *rate_tab, int num_entries, int val_type)
+{
+    static char line[2048];
+    char *ptr;
+    int i,j;
+    double x;
+
+    /* --- section heading --- */
+    if (global_rates[val_type] <= 0) {
+	printf("%s\n", format_tab[val_type].label);
+    }
+    else {
+	x = 100*sum_rates[val_type]/global_rates[val_type];
+	printf("%s - top %d processes account for %.0f%% of %s\n",
+	       format_tab[val_type].label,
+	       top, x>100?100:x,
+	       format_tab[val_type].fullname);
+    }
+
+    /* --- fields heading --- */
+    printf(pid_label_fmt, "PID");
+    for(j=0; j < NUM_VALUES; j++) {
+	if (show_spec & (1<<j)) {
+	    format_entry_t *fmt = &format_tab[j];
+	    printf(fmt->label_fmt, fmt->label);
+	}
+    } 
+    printf(" CMD\n");
+
+    /* units */
+    printf(pid_label_fmt, " ");
+    for(j=0; j < NUM_VALUES; j++) {
+	if (show_spec & (1<<j)) {
+	    format_entry_t *fmt = &format_tab[j];
+	    printf(fmt->label_fmt, fmt->units);
+	}
+    } 
+    printf("\n");
+
+
+    /* --- values --- */
+    for(i=0; i < min(top, num_entries); i++) {
+	rate_entry_t *entry = &rate_tab[i];
+
+	*line = '\0';
+	ptr = line;
+
+	/* pid */
+	sprintf(ptr, pid_fmt, instances[entry->inst]);
+	ptr += strlen(ptr);
+
+	/* other values */
+	for(j=0; j < NUM_VALUES; j++) {
+	    if (show_spec & (1<<j)) {
+		format_entry_t *fmt = &format_tab[j];
+		sprintf(ptr, fmt->val_fmt, entry->values[j]);
+		ptr += strlen(ptr);
+	    }
+	}
+
+	/* cmd */
+        /* skip over leading pid-name in instance name */
+	sprintf(ptr, " %s", &instance_names[entry->inst][pid_len+1]);
+
+	/* ensure line fits in certain width */
+	printf(line_fmt, line);
+	printf("\n");
+    }
+
+    printf("\n");
+}
+
+void 
+doit(void)
+{
+    int i, r, m, v, sts;
+    static double now = 0, then = 0;
+    double delta_d;
+    static unsigned long num_times = 0;
+    static pmResult *result[2];
+    int num_entries;
+
+    num_times++;
+
+    if (num_times > 2) {
+	pmFreeResult(result[0]);	
+
+        /* if forever, ensure no wrapping */
+        if (num_samples < 0)
+	    num_times = 3;
+    }
+
+    if (num_times > 1) {
+	result[0] = result[1];
+    }
+
+    get_indom();
+    if ((sts = pmFetch(num_pmid, pmidlist, &result[1])) < 0) {
+	for (i = 0; i < num_pmid; i++)
+	    fprintf(stderr, "%s: pmFetch: %s\n", namelist[i], pmErrStr(sts));
+	exit(1);
+    }
+
+
+    if (num_times > 1) {
+        then = result[0]->timestamp.tv_sec + result[0]->timestamp.tv_usec/1000000;
+        now = result[1]->timestamp.tv_sec + result[1]->timestamp.tv_usec/1000000;
+	delta_d = now - then;
+
+        if (result[1]->numpmid != num_pmid) {
+	    fprintf(stderr, "%s: Failed to fetch all metrics (%d out of %d)\n", 
+		    pmProgname, result[1]->numpmid, num_pmid);
+	    exit(1);
+	}
+	
+	/* --- build value table --- */
+
+	for(i=0; i<num_inst; i++) {
+	    val_tab[i].num = 0;
+	}
+
+	/* go thru each result */
+	for(r=0; r<2; r++) {
+	    /* go thru each metric */
+	    for(m=0; m<num_pmid; m++) {
+		pmValueSet *vset = result[r]->vset[m];
+		if (vset->numval < 0) {
+		    fprintf(stderr, "%s: Error when fetching metric %s: %s\n", 
+			pmProgname, namelist[m], pmErrStr(vset->numval));
+		    continue;
+		}
+		if (vset->numval == 0)
+		    continue;
+		valfmt_tab[m] = vset->valfmt;
+
+		/* handle global metrics */
+		if (((show_spec & (SHOW_CPU | SHOW_SYS)) && m == kernel_stime) || 
+		    ((show_spec & (SHOW_CPU | SHOW_USR)) && m == kernel_utime) || 
+		    ((show_spec & SHOW_SYSCALLS) && m == kernel_sysc) || 
+		    ((show_spec & SHOW_CTX) && m == kernel_ctx)) {
+		    if (vset->numval != 1) {
+			fprintf(stderr, "%s: Wrong num of values for kernel metric: %d\n", 
+			    pmProgname, vset->numval);
+			exit(1);
+		    }
+		    global_val[m][r] = vset->vlist[0];	
+		    continue;
+		}
+
+		/* handle proc-instance metrics */
+
+		/* Find result inst
+		 * Would be faster to start from where left off
+		 * last time. Need to remember this for a metric/result.
+		 */
+		for(v=0; v<vset->numval; v++) {
+		    pmValue val = vset->vlist[v];
+		    /* go thru each possible inst */
+		    for(i=0; i<num_inst; i++) {
+			if (instances[i] == val.inst) {
+			    val_tab[i].num++; /* needs to be zeroed at start */
+			    val_tab[i].values[m][r] = val;
+			    break;
+			}
+		    }
+		}
+	    }
+	}
+	
+	/* --- calculate rates --- */
+
+	/* handle global metrics */
+	if (show_spec & SHOW_CPU) {
+	    __uint64_t utime[2], stime[2];
+	    double cpuburn_val;
+
+	    utime[0] = get_value(kernel_utime, &global_val[kernel_utime][0]);
+	    utime[1] = get_value(kernel_utime, &global_val[kernel_utime][1]);
+	    stime[0] = get_value(kernel_stime, &global_val[kernel_stime][0]);
+	    stime[1] = get_value(kernel_stime, &global_val[kernel_stime][1]);
+	    
+	    cpuburn_val = (utime[1] - utime[0] + stime[1] - stime[0]) / delta_d;
+	    cpuburn_val /= 10; 
+	    cpuburn_val = cpuburn_val > 100 ? 100 : cpuburn_val;
+	    global_rates[CPU_VAL] = cpuburn_val;
+	}
+
+
+	if (show_spec & SHOW_SYSCALLS) {
+	    __uint64_t sysc[2];
+
+	    sysc[0] = get_value(kernel_sysc, &global_val[kernel_sysc][0]);
+	    sysc[1] = get_value(kernel_sysc, &global_val[kernel_sysc][1]);
+	    
+	    global_rates[SYSCALLS_VAL] = (sysc[1] - sysc[0]) / delta_d;
+	}
+
+	if (show_spec & SHOW_CTX) {
+	    __uint64_t ctx[2];
+
+	    ctx[0] = get_value(kernel_ctx, &global_val[kernel_ctx][0]);
+	    ctx[1] = get_value(kernel_ctx, &global_val[kernel_ctx][1]);
+	    
+	    global_rates[CTX_VAL] = (ctx[1] - ctx[0]) / delta_d;
+	}
+
+	if (show_spec & SHOW_SYS) {
+	    __uint64_t stime[2];
+
+	    stime[0] = get_value(kernel_stime, &global_val[kernel_stime][0]);
+	    stime[1] = get_value(kernel_stime, &global_val[kernel_stime][1]);
+	    
+	    global_rates[SYS_VAL] = ((stime[1] - stime[0]) / delta_d) / 10;
+	}
+
+	if (show_spec & SHOW_USR) {
+	    __uint64_t utime[2];
+
+	    utime[0] = get_value(kernel_utime, &global_val[kernel_utime][0]);
+	    utime[1] = get_value(kernel_utime, &global_val[kernel_utime][1]);
+	    
+	    global_rates[USR_VAL] = ((utime[1] - utime[0]) / delta_d) / 10;
+	}
+
+
+	/* handle proc-instance metrics */
+	num_entries = 0;
+	for(i=0; i<num_inst; i++) {
+	    val_entry_t *vt = &val_tab[i];
+	    rate_entry_t *rt;
+	    if (vt->num < (num_proc_pmid*2)) {
+		/* couldn't get instances for all proc metrics for both(*2) results */
+		continue;
+	    }
+	    
+	    rt = &rate_tab[num_entries];
+	    rt->inst = i;
+
+	    /* create cpuburn values */
+	    if (show_spec & SHOW_CPU) {
+		__uint64_t utime[2], stime[2];
+		double cpuburn_val;
+
+		utime[0] = get_value(pu_utime, &vt->values[pu_utime][0]);
+		utime[1] = get_value(pu_utime, &vt->values[pu_utime][1]);
+		stime[0] = get_value(pu_stime, &vt->values[pu_stime][0]);
+		stime[1] = get_value(pu_stime, &vt->values[pu_stime][1]);
+		
+		cpuburn_val = (utime[1] - utime[0] + stime[1] - stime[0]) / delta_d;
+		cpuburn_val /= 10; 
+		cpuburn_val = cpuburn_val > 100 ? 100 : cpuburn_val;
+		rt->values[CPU_VAL] = cpuburn_val;
+	    }
+
+	    /* create syscalls values */
+	    if (show_spec & SHOW_SYSCALLS) {
+		__uint64_t sysc[2];
+
+		sysc[0] = get_value(pu_sysc, &vt->values[pu_sysc][0]);
+		sysc[1] = get_value(pu_sysc, &vt->values[pu_sysc][1]);
+		
+		rt->values[SYSCALLS_VAL] = (sysc[1] - sysc[0]) / delta_d;
+	    }
+
+	    /* create ctx values */
+	    if (show_spec & SHOW_CTX) {
+		double ictx, vctx;
+
+		vctx = get_value(pu_vctx, &vt->values[pu_vctx][1]) - 
+		       get_value(pu_vctx, &vt->values[pu_vctx][0]);
+		ictx = get_value(pu_ictx, &vt->values[pu_ictx][1]) - 
+		       get_value(pu_ictx, &vt->values[pu_ictx][0]);
+		
+		rt->values[CTX_VAL] = (vctx + ictx) / delta_d;
+	    }
+
+	    /* create reads values */
+	    if (show_spec & SHOW_READS) {
+		double br, gbr;
+
+		br = get_value(pu_bread, &vt->values[pu_bread][1]) - 
+		     get_value(pu_bread, &vt->values[pu_bread][0]);
+		gbr = get_value(pu_gbread, &vt->values[pu_gbread][1]) -
+		      get_value(pu_gbread, &vt->values[pu_gbread][0]);
+		
+		rt->values[READS_VAL] = (gbr * 1024 * 1024 + br / 1024) / delta_d;
+	    }
+
+	    /* create writes values */
+	    if (show_spec & SHOW_WRITES) {
+		double bw, gbw;
+
+		bw = get_value(pu_bwrit, &vt->values[pu_bwrit][1]) -
+		     get_value(pu_bwrit, &vt->values[pu_bwrit][0]);
+		gbw = get_value(pu_gbwrit, &vt->values[pu_gbwrit][1]) -
+		      get_value(pu_gbwrit, &vt->values[pu_gbwrit][0]);
+		
+		rt->values[WRITES_VAL] = (gbw * 1024 * 1024 + bw / 1024) / delta_d;
+	    }
+
+	    /* create rss values */
+	    if (show_spec & SHOW_RSS) {
+		rt->values[RSS_VAL] = get_value(pu_rss, &vt->values[pu_rss][1]);
+	    }
+
+	    /* create sys cpu time values */
+	    if (show_spec & SHOW_SYS) {
+		__uint64_t stime[2];
+
+		stime[0] = get_value(pu_stime, &vt->values[pu_stime][0]);
+		stime[1] = get_value(pu_stime, &vt->values[pu_stime][1]);
+		
+		rt->values[SYS_VAL] = ((stime[1] - stime[0]) / delta_d) / 10;
+		rt->values[SYS_VAL] = rt->values[SYS_VAL] > 100 ? 100 : rt->values[SYS_VAL];
+	    }
+
+	    /* create user cpu time values */
+	    if (show_spec & SHOW_USR) {
+		__uint64_t utime[2];
+
+		utime[0] = get_value(pu_utime, &vt->values[pu_utime][0]);
+		utime[1] = get_value(pu_utime, &vt->values[pu_utime][1]);
+		
+		rt->values[USR_VAL] = ((utime[1] - utime[0]) / delta_d) / 10;
+		rt->values[USR_VAL] = rt->values[USR_VAL] > 100 ? 100 : rt->values[USR_VAL];
+	    }
+
+
+	    num_entries++;
+	}
+
+	/* --- report values --- */
+	printf("HOST: %s\n", hostname);
+	printf("DATE: %s\n\n", get_time());
+
+	if (show_spec & SHOW_CPU) {
+	    qsort(rate_tab, num_entries, sizeof(rate_entry_t), cpuburn_cmp);
+	    calc_sum_rates(num_entries, CPU_VAL);
+	    print_top(rate_tab, num_entries, CPU_VAL);
+	}
+
+	if (show_spec & SHOW_SYSCALLS) {
+	    qsort(rate_tab, num_entries, sizeof(rate_entry_t), syscall_cmp);
+	    calc_sum_rates(num_entries, SYSCALLS_VAL);
+	    print_top(rate_tab, num_entries, SYSCALLS_VAL);
+	}
+
+	if (show_spec & SHOW_CTX) {
+	    qsort(rate_tab, num_entries, sizeof(rate_entry_t), ctx_cmp);
+	    calc_sum_rates(num_entries, CTX_VAL);
+	    print_top(rate_tab, num_entries, CTX_VAL);
+	}
+
+	if (show_spec & SHOW_WRITES) {
+	    qsort(rate_tab, num_entries, sizeof(rate_entry_t), write_cmp);
+	    print_top(rate_tab, num_entries, WRITES_VAL);
+	}
+
+	if (show_spec & SHOW_READS) {
+	    qsort(rate_tab, num_entries, sizeof(rate_entry_t), read_cmp);
+	    print_top(rate_tab, num_entries, READS_VAL);
+	}
+
+	if (show_spec & SHOW_RSS) {
+	    qsort(rate_tab, num_entries, sizeof(rate_entry_t), rss_cmp);
+	    print_top(rate_tab, num_entries, RSS_VAL);
+	}
+
+	if (show_spec & SHOW_SYS) {
+	    qsort(rate_tab, num_entries, sizeof(rate_entry_t), sys_cmp);
+	    calc_sum_rates(num_entries, SYS_VAL);
+	    print_top(rate_tab, num_entries, SYS_VAL);
+	}
+
+	if (show_spec & SHOW_USR) {
+	    qsort(rate_tab, num_entries, sizeof(rate_entry_t), usr_cmp);
+	    calc_sum_rates(num_entries, USR_VAL);
+	    print_top(rate_tab, num_entries, USR_VAL);
+	}
+    }
+
+    if (num_times-1 == num_samples) {
+	exit(0);
+    }
+}
+
+/*
+ * go thru show_spec and build up the namelist and
+ * index variables into the array
+ */
+void
+create_namelist(void)
+{
+    int i = 0; /* for each pmid */
+    int j = 0; /* for global pmids */
+
+    if (show_spec & SHOW_CPU || show_spec & SHOW_USR) {
+	namelist[i] = "proc.psusage.utime";
+	pu_utime = i;
+	i++;
+	namelist[i] = "kernel.all.cpu.user";
+	kernel_utime = i;
+	i++; j++;
+    }
+    if (show_spec & SHOW_CPU || show_spec & SHOW_SYS) {
+	namelist[i] = "proc.psusage.stime";
+	pu_stime = i;
+	i++;
+	namelist[i] = "kernel.all.cpu.sys";
+	kernel_stime = i;
+	i++; j++;
+    }
+    if (show_spec & SHOW_SYSCALLS) {
+	namelist[i] = "proc.psusage.sysc";
+	pu_sysc = i;
+	i++;
+	namelist[i] = "kernel.all.syscall";
+	kernel_sysc = i;
+	i++; j++;
+    }
+    if (show_spec & SHOW_CTX) {
+	namelist[i] = "proc.psusage.ictx";
+	pu_ictx = i;
+	i++;
+	namelist[i] = "proc.psusage.vctx";
+	pu_vctx = i;
+	i++;
+	namelist[i] = "kernel.all.pswitch";
+	kernel_ctx = i;
+	i++; j++;
+    }
+    if (show_spec & SHOW_READS) {
+	namelist[i] = "proc.psusage.bread";
+	pu_bread = i;
+	i++;
+	namelist[i] = "proc.psusage.gbread";
+	pu_gbread = i;
+	i++;
+    }
+    if (show_spec & SHOW_WRITES) {
+	namelist[i] = "proc.psusage.bwrit";
+	pu_bwrit = i;
+	i++;
+	namelist[i] = "proc.psusage.gbwrit";
+	pu_gbwrit = i;
+	i++;
+    }
+    if (show_spec & SHOW_RSS) {
+	namelist[i] = "proc.psusage.rss";
+	pu_rss = i;
+	i++;
+    }
+
+    num_pmid = i;
+    num_proc_pmid = i - j;
+}
+
+static void
+get_indom(void)
+{
+    static int onetrip = 1;
+
+    if (instances != NULL) 
+	free(instances);
+    if (instance_names != NULL)
+	free(instance_names);
+
+    if ((num_inst = pmGetInDom(proc_indom, &instances, &instance_names)) <= 0) {
+	fprintf(stderr, "%s: Failed to find any proc instances : %s\n",
+		pmProgname, pmErrStr(num_inst));
+	exit(1);
+    }
+
+    
+    /* process the instance names
+     * - find how long a pid is 
+     * - create pid label fmt based on pid length
+     */
+    if (onetrip) {
+	char *str = instance_names[0];
+
+	onetrip=0;
+
+	/* get pid str up to first space */
+	pid_len = 0;
+	while (*str != '\0' && *str != ' ') {
+	    pid_len++;
+	    str++;
+	}
+	if (*str == '\0') {
+	    /* shouldn't happen */
+	    fprintf(stderr, "%s: Bad proc instance : %s\n",
+		    pmProgname, instance_names[0]);
+	    exit(1);
+	}
+
+	sprintf(pid_label_fmt, "%%%ds", pid_len);
+	sprintf(pid_fmt, "%%%dd", pid_len);
+    }
+
+
+    rate_tab = realloc(rate_tab, sizeof(rate_entry_t)*num_inst);
+    if (rate_tab == NULL) {
+	fprintf(stderr, "%s: Failed to allocate rate table\n",
+		pmProgname);
+	exit(1);
+    }
+
+    val_tab = realloc(val_tab, sizeof(val_entry_t)*num_inst);
+    if (val_tab == NULL) {
+	fprintf(stderr, "%s: Failed to allocate value table\n",
+		pmProgname);
+	exit(1);
+    }
+
+    pmDelProfile(proc_indom, 0, NULL);
+    pmAddProfile(proc_indom, num_inst, instances);
+}
+
+static int
+overrides(int opt, pmOptions *opts)
+{
+    if (opt == 'p')
+	return 1;
+
+    if (opt == 's')
+	num_samples = atoi(opts->optarg); /* continue processing 's' */
+    return 0;
+}
+
+int
+main(int argc, char *argv[])
+{
+    int		c;
+    int		sts;
+    char	*endnum;
+    pmDesc	desc;
+    int		one_trip = 1;
+
+    while ((c = pmGetOptions(argc, argv, &opts)) != EOF) {
+	switch (c) {
+
+	case 'w':	/* wide flag */
+	    line_fmt = "%.1024s";
+	    break;
+
+	case 'p':	/* show flag */
+	    if (one_trip) {
+		show_spec = 0;
+		one_trip = 0;
+	    }
+	    if ((sts = parse_show_spec(opts.optarg)) < 0) {
+		pmprintf("%s: unrecognized print flag specification (%s)\n",
+			pmProgname, opts.optarg);
+		opts.errors++;
+	    } else {
+		show_spec |= sts;
+	    }
+	    break;
+
+	case 'm':	/* top N */
+	    top = (int)strtol(opts.optarg, &endnum, 10);
+	    if (top <= 0) {
+		pmprintf("%s: -m requires a positive integer\n", pmProgname);
+		opts.errors++;	
+	    }
+	    break;
+	}
+    }
+
+    if (opts.optind < argc)
+        opts.errors++;
+
+    if (opts.errors) {
+	pmUsageMessage(&opts);
+	exit(1);
+    }
+
+    create_namelist();
+
+    if (opts.interval.tv_sec == 0)
+	opts.interval.tv_sec = 2;
+
+    if (opts.nhosts > 0)
+	hostname = opts.hosts[0];
+    else
+	hostname = "local:";
+
+    if ((sts = c = pmNewContext(PM_CONTEXT_HOST, hostname)) < 0) {
+	fprintf(stderr, "%s: Cannot connect to PMCD on host \"%s\": %s\n",
+		pmProgname, hostname, pmErrStr(sts));
+	exit(1);
+    }
+    hostname = (char *)pmGetContextHostName(c);
+
+    if (pmGetContextOptions(c, &opts)) {
+	pmflush();
+	exit(1);
+    }
+
+    if ((sts = pmLookupName(num_pmid, namelist, pmidlist)) < 0) {
+	fprintf(stderr, "%s: Failed to lookup metrics : %s\n",
+		pmProgname, pmErrStr(sts));
+	exit(1);
+    }
+
+    for (c = 0; c < num_pmid; c++) {
+	if ((sts = pmLookupDesc(pmidlist[c], &desc)) < 0) {
+	    fprintf(stderr, "%s: Failed to lookup descriptor for metric \"%s\": %s\n",
+		    pmProgname, namelist[c], pmErrStr(sts));
+	    exit(1);
+	}
+	type_tab[c] = desc.type;	
+	/* ASSUMES that the first metric will always be a proc metric */
+	if (c == 0) {
+	    proc_indom = desc.indom;
+	}
+    }
+
+    for (;;) {
+	doit();
+	__pmtimevalSleep(opts.interval);
+    }
+
+    return 0;
+}