/*
* ita2_btb.c - example of how use the BTB with the Itanium 2 PMU
*
* Copyright (c) 2003-2006 Hewlett-Packard Development Company, L.P.
* Contributed by Stephane Eranian <eranian@hpl.hp.com>
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
* of the Software, and to permit persons to whom the Software is furnished to do so,
* subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,
* INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
* PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
* HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF
* CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE
* OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*
* This file is part of libpfm, a performance monitoring support library for
* applications on Linux/ia64.
*/
#include <sys/types.h>
#include <stdio.h>
#include <stdlib.h>
#include <stdarg.h>
#include <errno.h>
#include <unistd.h>
#include <string.h>
#include <signal.h>
#include <fcntl.h>
#include <perfmon/perfmon.h>
#include <perfmon/perfmon_default_smpl.h>
#include <perfmon/pfmlib_itanium2.h>
typedef pfm_default_smpl_hdr_t btb_hdr_t;
typedef pfm_default_smpl_entry_t btb_entry_t;
typedef pfm_default_smpl_ctx_arg_t btb_ctx_arg_t;
#define BTB_FMT_UUID PFM_DEFAULT_SMPL_UUID
static pfm_uuid_t buf_fmt_id = BTB_FMT_UUID;
#define NUM_PMCS PFMLIB_MAX_PMCS
#define NUM_PMDS PFMLIB_MAX_PMDS
#define MAX_EVT_NAME_LEN 128
#define MAX_PMU_NAME_LEN 32
/*
* The BRANCH_EVENT is increment by 1 for each branch event. Such event is composed of
* two entries in the BTB: a source and a target entry. The BTB is full after 4 branch
* events.
*/
#define SMPL_PERIOD (4UL*256)
/*
* We use a small buffer size to exercise the overflow handler
*/
#define SMPL_BUF_NENTRIES 64
#define M_PMD(x) (1UL<<(x))
#define BTB_REGS_MASK (M_PMD(8)|M_PMD(9)|M_PMD(10)|M_PMD(11)|M_PMD(12)|M_PMD(13)|M_PMD(14)|M_PMD(15)|M_PMD(16))
static void *smpl_vaddr;
static unsigned int entry_size;
static int id;
#if defined(__ECC) && defined(__INTEL_COMPILER)
/* if you do not have this file, your compiler is too old */
#include <ia64intrin.h>
#define hweight64(x) _m64_popcnt(x)
#elif defined(__GNUC__)
static __inline__ int
hweight64 (unsigned long x)
{
unsigned long result;
__asm__ ("popcnt %0=%1" : "=r" (result) : "r" (x));
return (int)result;
}
#else
#error "you need to provide inline assembly from your compiler"
#endif
/*
* we don't use static to make sure the compiler does not inline the function
*/
long func1(void) { return 0;}
long
do_test(unsigned long loop)
{
long sum = 0;
while(loop--) {
if (loop & 0x1)
sum += func1();
else
sum += loop;
}
return sum;
}
static void fatal_error(char *fmt,...) __attribute__((noreturn));
static void
fatal_error(char *fmt, ...)
{
va_list ap;
va_start(ap, fmt);
vfprintf(stderr, fmt, ap);
va_end(ap);
exit(1);
}
/*
* print content of sampling buffer
*
* XXX: using stdio to print from a signal handler is not safe with multi-threaded
* applications
*/
#define safe_printf printf
static void
show_btb_reg(int j, pfm_ita2_pmd_reg_t reg, pfm_ita2_pmd_reg_t pmd16)
{
unsigned long bruflush, b1;
int is_valid = reg.pmd8_15_ita2_reg.btb_b == 0 && reg.pmd8_15_ita2_reg.btb_mp == 0 ? 0 :1;
b1 = (pmd16.pmd_val >> (4 + 4*(j-8))) & 0x1;
bruflush = (pmd16.pmd_val >> (5 + 4*(j-8))) & 0x1;
safe_printf("\tPMD%-2d: 0x%016lx b=%d mp=%d bru=%ld b1=%ld valid=%c\n",
j,
reg.pmd_val,
reg.pmd8_15_ita2_reg.btb_b,
reg.pmd8_15_ita2_reg.btb_mp,
bruflush, b1,
is_valid ? 'Y' : 'N');
if (!is_valid) return;
if (reg.pmd8_15_ita2_reg.btb_b) {
unsigned long addr;
addr = (reg.pmd8_15_ita2_reg.btb_addr+b1)<<4;
addr |= reg.pmd8_15_ita2_reg.btb_slot < 3 ? reg.pmd8_15_ita2_reg.btb_slot : 0;
safe_printf("\t Source Address: 0x%016lx\n"
"\t Taken=%c Prediction: %s\n\n",
addr,
reg.pmd8_15_ita2_reg.btb_slot < 3 ? 'Y' : 'N',
reg.pmd8_15_ita2_reg.btb_mp ? "FE Failure" :
bruflush ? "BE Failure" : "Success");
} else {
safe_printf("\t Target Address: 0x%016lx\n\n",
((unsigned long)reg.pmd8_15_ita2_reg.btb_addr<<4));
}
}
static void
show_btb(pfm_ita2_pmd_reg_t *btb, pfm_ita2_pmd_reg_t *pmd16)
{
int i, last;
i = (pmd16->pmd16_ita2_reg.btbi_full) ? pmd16->pmd16_ita2_reg.btbi_bbi : 0;
last = pmd16->pmd16_ita2_reg.btbi_bbi;
safe_printf("btb_trace: i=%d last=%d bbi=%d full=%d\n", i, last,pmd16->pmd16_ita2_reg.btbi_bbi, pmd16->pmd16_ita2_reg.btbi_full);
do {
show_btb_reg(i+8, btb[i], *pmd16);
i = (i+1) % 8;
} while (i != last);
}
void
process_smpl_buffer(void)
{
btb_hdr_t *hdr;
btb_entry_t *ent;
unsigned long pos;
unsigned long smpl_entry = 0;
pfm_ita2_pmd_reg_t *reg, *pmd16;
unsigned long i;
int ret;
static unsigned long last_ovfl = ~0UL;
hdr = (btb_hdr_t *)smpl_vaddr;
/*
* check that we are not diplaying the previous set of samples again.
* Required to take care of the last batch of samples.
*/
if (hdr->hdr_overflows <= last_ovfl && last_ovfl != ~0UL) {
printf("skipping identical set of samples %lu <= %lu\n", hdr->hdr_overflows, last_ovfl);
return;
}
pos = (unsigned long)(hdr+1);
/*
* walk through all the entries recored in the buffer
*/
for(i=0; i < hdr->hdr_count; i++) {
ret = 0;
ent = (btb_entry_t *)pos;
/*
* print entry header
*/
safe_printf("Entry %ld PID:%d CPU:%d STAMP:0x%lx IIP:0x%016lx\n",
smpl_entry++,
ent->pid,
ent->cpu,
ent->tstamp,
ent->ip);
/*
* point to first recorded register (always contiguous with entry header)
*/
reg = (pfm_ita2_pmd_reg_t*)(ent+1);
/*
* in this particular example, we have pmd8-pmd15 has the BTB. We have also
* included pmd16 (BTB index) has part of the registers to record. This trick
* allows us to get the index to decode the sequential order of the BTB.
*
* Recorded registers are always recorded in increasing order. So we know
* that pmd16 is at a fixed offset (+8*sizeof(unsigned long)) from pmd8.
*/
pmd16 = reg+8;
show_btb(reg, pmd16);
/*
* move to next entry
*/
pos += entry_size;
}
}
static void
overflow_handler(int n, struct siginfo *info, struct sigcontext *sc)
{
/* dangerous */
printf("Notification received\n");
process_smpl_buffer();
/*
* And resume monitoring
*/
if (perfmonctl(id, PFM_RESTART,NULL, 0) == -1) {
perror("PFM_RESTART");
exit(1);
}
}
int
main(void)
{
int ret;
int type = 0;
pfarg_reg_t pd[NUM_PMDS];
pfarg_reg_t pc[NUM_PMCS];
pfmlib_input_param_t inp;
pfmlib_output_param_t outp;
pfmlib_ita2_input_param_t ita2_inp;
btb_ctx_arg_t ctx[1];
pfarg_load_t load_args;
pfmlib_options_t pfmlib_options;
struct sigaction act;
unsigned int i;
/*
* Initialize pfm library (required before we can use it)
*/
if (pfm_initialize() != PFMLIB_SUCCESS) {
fatal_error("Can't initialize library\n");
}
/*
* Let's make sure we run this on the right CPU
*/
pfm_get_pmu_type(&type);
if (type != PFMLIB_ITANIUM2_PMU) {
char model[MAX_PMU_NAME_LEN];
pfm_get_pmu_name(model, MAX_PMU_NAME_LEN);
fatal_error("this program does not work with %s PMU\n", model);
}
/*
* Install the overflow handler (SIGIO)
*/
memset(&act, 0, sizeof(act));
act.sa_handler = (sig_t)overflow_handler;
sigaction (SIGIO, &act, 0);
/*
* pass options to library (optional)
*/
memset(&pfmlib_options, 0, sizeof(pfmlib_options));
pfmlib_options.pfm_debug = 0; /* set to 1 for debug */
pfmlib_options.pfm_verbose = 0; /* set to 1 for debug */
pfm_set_options(&pfmlib_options);
memset(pd, 0, sizeof(pd));
memset(ctx, 0, sizeof(ctx));
/*
* prepare parameters to library. we don't use any Itanium
* specific features here. so the pfp_model is NULL.
*/
memset(&inp,0, sizeof(inp));
memset(&outp,0, sizeof(outp));
memset(&ita2_inp,0, sizeof(ita2_inp));
/*
* Before calling pfm_find_dispatch(), we must specify what kind
* of branches we want to capture. We are interesteed in all the mispredicted branches,
* therefore we program we set the various fields of the BTB config to:
*/
ita2_inp.pfp_ita2_btb.btb_used = 1;
ita2_inp.pfp_ita2_btb.btb_ds = 0;
ita2_inp.pfp_ita2_btb.btb_tm = 0x3;
ita2_inp.pfp_ita2_btb.btb_ptm = 0x3;
ita2_inp.pfp_ita2_btb.btb_ppm = 0x3;
ita2_inp.pfp_ita2_btb.btb_brt = 0x0;
ita2_inp.pfp_ita2_btb.btb_plm = PFM_PLM3;
/*
* To count the number of occurence of this instruction, we must
* program a counting monitor with the IA64_TAGGED_INST_RETIRED_PMC8
* event.
*/
if (pfm_find_full_event("BRANCH_EVENT", &inp.pfp_events[0]) != PFMLIB_SUCCESS) {
fatal_error("cannot find event BRANCH_EVENT\n");
}
/*
* set the (global) privilege mode:
* PFM_PLM3 : user level only
*/
inp.pfp_dfl_plm = PFM_PLM3;
/*
* how many counters we use
*/
inp.pfp_event_count = 1;
/*
* let the library figure out the values for the PMCS
*/
if ((ret=pfm_dispatch_events(&inp, &ita2_inp, &outp, NULL)) != PFMLIB_SUCCESS) {
fatal_error("cannot configure events: %s\n", pfm_strerror(ret));
}
/*
* We initialize the format specific information.
* The format is identified by its UUID which must be copied
* into the ctx_buf_fmt_id field.
*/
memcpy(ctx[0].ctx_arg.ctx_smpl_buf_id, buf_fmt_id, sizeof(pfm_uuid_t));
/*
* the size of the buffer is indicated in bytes (not entries).
*
* The kernel will record into the buffer up to a certain point.
* No partial samples are ever recorded.
*/
ctx[0].buf_arg.buf_size = 8192;
/*
* now create the context for self monitoring/per-task
*/
if (perfmonctl(0, PFM_CREATE_CONTEXT, ctx, 1) == -1 ) {
if (errno == ENOSYS) {
fatal_error("Your kernel does not have performance monitoring support!\n");
}
fatal_error("Can't create PFM context %s\n", strerror(errno));
}
printf("Sampling buffer mapped at %p\n", ctx[0].ctx_arg.ctx_smpl_vaddr);
smpl_vaddr = ctx[0].ctx_arg.ctx_smpl_vaddr;
/*
* extract our file descriptor
*/
id = ctx[0].ctx_arg.ctx_fd;
/*
* Now prepare the argument to initialize the PMDs and PMCS.
* We must pfp_pmc_count to determine the number of PMC to intialize.
* We must use pfp_event_count to determine the number of PMD to initialize.
* Some events cause extra PMCs to be used, so pfp_pmc_count may be >= pfp_event_count.
*
* This step is new compared to libpfm-2.x. It is necessary because the library no
* longer knows about the kernel data structures.
*/
for (i=0; i < outp.pfp_pmc_count; i++) {
pc[i].reg_num = outp.pfp_pmcs[i].reg_num;
pc[i].reg_value = outp.pfp_pmcs[i].reg_value;
}
/*
* the PMC controlling the event ALWAYS come first, that's why this loop
* is safe even when extra PMC are needed to support a particular event.
*/
for (i=0; i < inp.pfp_event_count; i++) {
pd[i].reg_num = pc[i].reg_num;
}
/*
* indicate we want notification when buffer is full
*/
pc[0].reg_flags |= PFM_REGFL_OVFL_NOTIFY;
/*
* Now prepare the argument to initialize the PMD and the sampling period
* We know we use only one PMD in this case, therefore pmd[0] corresponds
* to our first event which is our sampling period.
*/
pd[0].reg_value = (~0UL) - SMPL_PERIOD +1;
pd[0].reg_long_reset = (~0UL) - SMPL_PERIOD +1;
pd[0].reg_short_reset = (~0UL) - SMPL_PERIOD +1;
/*
* indicate PMD to collect in each sample
*/
pc[0].reg_smpl_pmds[0] = BTB_REGS_MASK;
/*
* compute size of each sample: fixed-size header + all our BTB regs
*/
entry_size = sizeof(btb_entry_t)+(hweight64(BTB_REGS_MASK)<<3);
/*
* When our counter overflows, we want to BTB index to be reset, so that we keep
* in sync. This is required to make it possible to interpret pmd16 on overflow
* to avoid repeating the same branch several times.
*/
pc[0].reg_reset_pmds[0] = M_PMD(16);
/*
* reset pmd16 (BTB index), short and long reset value are set to zero as well
*
* We use slot 1 of our pd[] array for this.
*/
pd[1].reg_num = 16;
pd[1].reg_value = 0UL;
/*
* Now program the registers
*
* We don't use the save variable to indicate the number of elements passed to
* the kernel because, as we said earlier, pc may contain more elements than
* the number of events we specified, i.e., contains more thann coutning monitors.
*/
if (perfmonctl(id, PFM_WRITE_PMCS, pc, outp.pfp_pmc_count) == -1) {
fatal_error("perfmonctl error PFM_WRITE_PMCS errno %d\n",errno);
}
/*
* we use 2 = 1 for the branch_event + 1 for the reset of PMD16.
*/
if (perfmonctl(id, PFM_WRITE_PMDS, pd, 2) == -1) {
fatal_error("perfmonctl error PFM_WRITE_PMDS errno %d\n",errno);
}
/*
* now we load (i.e., attach) the context to ourself
*/
load_args.load_pid = getpid();
if (perfmonctl(id, PFM_LOAD_CONTEXT, &load_args, 1) == -1) {
fatal_error("perfmonctl error PFM_LOAD_CONTEXT errno %d\n",errno);
}
/*
* setup asynchronous notification on the file descriptor
*/
ret = fcntl(id, F_SETFL, fcntl(id, F_GETFL, 0) | O_ASYNC);
if (ret == -1) {
fatal_error("cannot set ASYNC: %s\n", strerror(errno));
}
/*
* get ownership of the descriptor
*/
ret = fcntl(id, F_SETOWN, getpid());
if (ret == -1) {
fatal_error("cannot setown: %s\n", strerror(errno));
}
/*
* Let's roll now.
*/
pfm_self_start(id);
do_test(100000);
pfm_self_stop(id);
/*
* We must call the processing routine to cover the last entries recorded
* in the sampling buffer. Note that the buffer may not be full at this point.
*
*/
process_smpl_buffer();
/*
* let's stop this now
*/
close(id);
return 0;
}