Also, here are similar scripts for the cpu memory hierarchy.
Richard.
On Thu, Nov 03, 2005 at 11:16:25AM -0500, Matty wrote:
> On Wed, 2 Nov 2005, Richard McDougall wrote:
>
> >
> >Hi Matty,
> >
> >We use some scripts to post-process busstat into what you are looking
> >for: bandwidth per sbus and device. It's a two phase operation now,
> >collect and report. See below for the report and collect scripts...
> >
>
> Hi Richard,
>
> I reviewed the scripts, and noticed the following headings in
> "print_ac_gigaplane_stats":
>
> print "AbUtil DbUtil RTS RTSA RTO WB\n";
> [...]
> print "RIO WIO RBIO WBIO RS WS\n";
>
> Any idea what the columns stand for? Do you happen to know what the size
> of a pkt would be? Do you or anyone on the list happen to know where I can
> get descriptions for the busstat column headers?
>
> Thanks for the reply and the awesome scripts,
> - Ryan
>
> --
> UNIX Administrator
> http://daemons.net/~matty
>
>
>
> >Regards,
> >
> >Richard.
> >
> >
> >
> >
> >
> >
> >On Wed, Nov 02, 2005 at 01:48:49PM -0500, Matty wrote:
> >>
> >>Howdy,
> >>
> >>I have been playing around with busstat, and must say that this is a
> >>really really cool utility! To get an idea of what the event counters
> >>mean, I started reading through a PCI reference and the following website:
> >>
> >>http://www.readol.net/books/computer/system/System.Performance.Tuning.Second.Edition/059600284x/spt2-CHP-3-SECT-6.html
> >>
> >>This website mentions that the value "dvma_bytes_xfr" could be used to get
> >>the number of bytes transferred on SBUS-based machines. Upon inspecting
> >>the busstat events on a Sun 280R, I didn't notice a similar event:
> >>
> >>$ busstat -e pcis0
> >>pic0
> >>dvma_stream_rd
> >>dvma_stream_wr
> >>dvma_const_rd
> >>dvma_const_wr
> >>dvma_stream_buf_mis
> >>dvma_cycles
> >>dvma_wd_xfr
> >>pio_cycles
> >>dvma_tlb_misses
> >>interrupts
> >>saf_inter_nack
> >>pio_reads
> >>pio_writes
> >>dvma_rd_buf_timeout
> >>dvma_rd_rtry_stc
> >>dvma_wr_rtry_stc
> >>dvma_rd_rtry_nonstc
> >>dvma_wr_rtry_nonstc
> >>E*_slow_transitions
> >>E*_slow_cycles_per_64
> >>
> >>Are the events documented anywhere by any chance? Is there anything
> >>equivalent to "dvma_bytes_xfr" to get the number of bytes transmitted
> >>on a PCI bus?
> >>
> >>Thanks for any insight,
> >>- Ryan
> >>
> >>--
> >>UNIX Administrator
> >>http://daemons.net/~matty
> >>_______________________________________________
> >>perf-discuss mailing list
> >>[email protected]
> >
> >--
> >
> >
> >:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:
> > Richard Mc Dougall : [EMAIL PROTECTED]
> > Performance and Availability : x31542
> > Engineering : http://blogs.sun.com/rmc
> > Sun Microsystems Inc : +1 650 352 6438
> >:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:
> >
>
> --
> _______________________________________________
> perf-discuss mailing list
> [email protected]
--
:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:
Richard Mc Dougall : [EMAIL PROTECTED]
Performance and Availability : x31542
Engineering : http://blogs.sun.com/rmc
Sun Microsystems Inc : +1 650 352 6438
:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:
/usr/sbin/cpustat \
-c pic0=Cycle_cnt,pic1=Instr_cnt \
-c pic0=Cycle_cnt,pic1=Instr_cnt,sys \
-c pic0=Cycle_cnt,pic1=Instr_cnt,sys,nouser \
-c pic0=Dispatch0_IC_miss,pic1=Dispatch0_mispred \
-c pic0=Dispatch0_IC_miss,pic1=Dispatch0_mispred,sys \
-c pic0=Dispatch0_IC_miss,pic1=Dispatch0_mispred,sys,nouser \
-c pic0=IC_ref,pic1=IC_miss \
-c pic0=IC_ref,pic1=IC_miss,sys \
-c pic0=IC_ref,pic1=IC_miss,sys,nouser \
-c pic0=DC_rd,pic1=DC_rd_miss \
-c pic0=DC_rd,pic1=DC_rd_miss,sys \
-c pic0=DC_rd,pic1=DC_rd_miss,sys,nouser \
-c pic0=DC_wr,pic1=DC_wr_miss \
-c pic0=DC_wr,pic1=DC_wr_miss,sys \
-c pic0=DC_wr,pic1=DC_wr_miss,sys,nouser \
-c pic0=Rstall_storeQ,pic1=Re_RAW_miss \
-c pic0=Rstall_storeQ,pic1=Re_RAW_miss,sys \
-c pic0=Rstall_storeQ,pic1=Re_RAW_miss,sys,nouser \
-c pic0=EC_ref,pic1=EC_misses \
-c pic0=EC_ref,pic1=EC_misses,sys \
-c pic0=EC_ref,pic1=EC_misses,sys,nouser \
-c pic0=EC_write_hit_RTO,pic1=EC_wb \
-c pic0=EC_write_hit_RTO,pic1=EC_wb,sys \
-c pic0=EC_write_hit_RTO,pic1=EC_wb,sys,nouser \
-c pic0=EC_snoop_inv,pic1=EC_snoop_cb \
-c pic0=EC_snoop_inv,pic1=EC_snoop_cb,sys \
-c pic0=EC_snoop_inv,pic1=EC_snoop_cb,sys,nouser \
-c pic0=EC_rd_miss,pic1=EC_ic_miss \
-c pic0=EC_rd_miss,pic1=EC_ic_miss,sys \
-c pic0=EC_rd_miss,pic1=EC_ic_miss,sys,nouser \
-c pic0=Dispatch0_br_target,pic1=Re_EC_miss \
-c pic0=Dispatch0_br_target,pic1=Re_EC_miss,sys \
-c pic0=Dispatch0_br_target,pic1=Re_EC_miss,sys,nouser \
-c pic0=Rstall_IU_use,pic1=Re_PC_miss \
-c pic0=Rstall_IU_use,pic1=Re_PC_miss,sys \
-c pic0=Rstall_IU_use,pic1=Re_PC_miss,sys,nouser \
-c pic0=PC_port0_rd,pic1=ITLB_miss \
-c pic0=PC_port0_rd,pic1=ITLB_miss,sys \
-c pic0=PC_port0_rd,pic1=ITLB_miss,sys,nouser \
-c pic0=SI_snoop,pic1=DTLB_miss \
-c pic0=SI_snoop,pic1=DTLB_miss,sys \
-c pic0=SI_snoop,pic1=DTLB_miss,sys,nouser \
-c pic0=SI_ciq_flow,pic1=WC_miss \
-c pic0=SI_ciq_flow,pic1=WC_miss,sys \
-c pic0=SI_ciq_flow,pic1=WC_miss,sys,nouser \
-c pic0=SI_owned,pic1=WC_snoop_cb \
-c pic0=SI_owned,pic1=WC_snoop_cb,sys \
-c pic0=SI_owned,pic1=WC_snoop_cb,sys,nouser \
-c pic0=SW_count_0,pic1=WC_scrubbed \
-c pic0=SW_count_0,pic1=WC_scrubbed,sys \
-c pic0=SW_count_0,pic1=WC_scrubbed,sys,nouser \
-c pic0=IU_Stat_Br_miss_taken,pic1=WC_wb_wo_read \
-c pic0=IU_Stat_Br_miss_taken,pic1=WC_wb_wo_read,sys \
-c pic0=IU_Stat_Br_miss_taken,pic1=WC_wb_wo_read,sys,nouser \
-c pic0=IU_Stat_Br_count_taken,pic1=SW_count_1 \
-c pic0=IU_Stat_Br_count_taken,pic1=SW_count_1,sys \
-c pic0=IU_Stat_Br_count_taken,pic1=SW_count_1,sys,nouser \
-c pic0=Dispatch_rs_mispred,pic1=IU_Stat_Br_miss_untaken \
-c pic0=Dispatch_rs_mispred,pic1=IU_Stat_Br_miss_untaken,sys \
-c pic0=Dispatch_rs_mispred,pic1=IU_Stat_Br_miss_untaken,sys,nouser \
-c pic0=FA_pipe_completion,pic1=PC_soft_hit \
-c pic0=FA_pipe_completion,pic1=PC_soft_hit,sys \
-c pic0=FA_pipe_completion,pic1=PC_soft_hit,sys,nouser \
-c pic0=Dispatch0_2nd_br,pic1=PC_hard_hit \
-c pic0=Dispatch0_2nd_br,pic1=PC_hard_hit,sys \
-c pic0=Dispatch0_2nd_br,pic1=PC_hard_hit,sys,nouser \
-c pic0=MC_reads_0,pic1=MC_writes_0 \
-c pic0=MC_reads_0,pic1=MC_writes_0,sys \
-c pic0=MC_reads_0,pic1=MC_writes_0,sys,nouser \
-c pic0=MC_reads_1,pic1=MC_writes_1 \
-c pic0=MC_reads_1,pic1=MC_writes_1,sys \
-c pic0=MC_reads_1,pic1=MC_writes_1,sys,nouser \
-c pic0=MC_reads_2,pic1=MC_writes_2 \
-c pic0=MC_reads_2,pic1=MC_writes_2,sys \
-c pic0=MC_reads_2,pic1=MC_writes_2,sys,nouser \
-c pic0=MC_reads_3,pic1=MC_writes_3 \
-c pic0=MC_reads_3,pic1=MC_writes_3,sys \
-c pic0=MC_reads_3,pic1=MC_writes_3,sys,nouser \
-c pic0=MC_stalls_0,pic1=MC_stalls_1 \
-c pic0=MC_stalls_0,pic1=MC_stalls_1,sys \
-c pic0=MC_stalls_0,pic1=MC_stalls_1,sys,nouser \
-c pic0=MC_stalls_2,pic1=MC_stalls_3 \
-c pic0=MC_stalls_2,pic1=MC_stalls_3,sys \
-c pic0=MC_stalls_2,pic1=MC_stalls_3,sys,nouser \
-c pic0=EC_wb_remote,pic1=Re_DC_miss \
-c pic0=EC_wb_remote,pic1=Re_DC_miss,sys \
-c pic0=EC_wb_remote,pic1=Re_DC_miss,sys,nouser \
-c pic0=EC_wb_remote,pic1=Re_DC_missovhd \
-c pic0=EC_wb_remote,pic1=Re_DC_missovhd,sys \
-c pic0=EC_wb_remote,pic1=Re_DC_missovhd,sys,nouser \
-c pic0=EC_miss_mtag_remote,pic1=EC_miss_remote \
-c pic0=EC_miss_mtag_remote,pic1=EC_miss_remote,sys \
-c pic0=EC_miss_mtag_remote,pic1=EC_miss_remote,sys,nouser \
-c pic0=EC_miss_local,pic1=Re_FPU_bypass \
-c pic0=EC_miss_local,pic1=Re_FPU_bypass,sys \
-c pic0=EC_miss_local,pic1=Re_FPU_bypass,sys,nouser \
-c pic0=Instr_cnt,pic1=PC_MS_misses \
-c pic0=Instr_cnt,pic1=PC_MS_misses,sys \
-c pic0=Instr_cnt,pic1=PC_MS_misses,sys,nouser \
-c pic0=Cycle_cnt,pic1=PC_snoop_inv \
-c pic0=Cycle_cnt,pic1=PC_snoop_inv,sys \
-c pic0=Cycle_cnt,pic1=PC_snoop_inv,sys,nouser \
-c pic0=Instr_cnt,pic1=PC_hard_hit \
-c pic0=Instr_cnt,pic1=PC_hard_hit,sys \
-c pic0=Instr_cnt,pic1=PC_hard_hit,sys,nouser \
-c pic0=Cycle_cnt,pic1=Rstall_FP_use \
-c pic0=Cycle_cnt,pic1=Rstall_FP_use,sys \
-c pic0=Cycle_cnt,pic1=Rstall_FP_use,sys,nouser \
-c pic0=Instr_cnt,pic1=FM_pipe_completion \
-c pic0=Instr_cnt,pic1=FM_pipe_completion,sys \
-c pic0=Instr_cnt,pic1=FM_pipe_completion,sys,nouser \
-c pic0=Cycle_cnt,pic1=IU_Stat_Br_count_untaken \
-c pic0=Cycle_cnt,pic1=IU_Stat_Br_count_untaken,sys \
-c pic0=Cycle_cnt,pic1=IU_Stat_Br_count_untaken,sys,nouser \
-c pic0=Instr_cnt,pic1=IC_miss_cancelled \
-c pic0=Instr_cnt,pic1=IC_miss_cancelled,sys \
-c pic0=Instr_cnt,pic1=IC_miss_cancelled,sys,nouser 1 600 \
#!/ws/on81-tools/SUNWspro/SC6.1/bin/perl
# Tunables
$debug = 0;
$CpuExp = 5;
$SysioExp = 5;
$AcExp = 4;
$SSMNodes = 4;
$MC_CAP = 8.33;
$MC_BDCAP = 11.9;
# Get options
if ($#ARGV == -1) {&usage();}
while ($#ARGV > -1) {
$arg = shift @ARGV;
if ( $arg =~ /^-(.)(.*)$/ ) {
$flag=$1; $val=$2;
$flag2=$flag.$val;
if ( $flag eq "i" ) {
if ( $val eq "" ) { $val = shift @ARGV;}
if ( $val eq "" ) { &usage();}
$infile = $val;
} elsif ($flag eq "o") {
if ( $val eq "" ) { $val = shift @ARGV;}
if ( $val eq "" ) { &usage();}
$outfile = $val;
} elsif ($flag eq "v") {
if ( $val ne "") { &usage();}
$verbose = 1;
} elsif ($flag2 eq "ac") {
$aconly = 1;
} elsif ($flag2 eq "sysio") {
$sysioonly = 1;
} elsif ($flag2 eq "cpu") {
$cpuonly = 1;
} elsif ($flag2 eq "uk") {
$ukonly = 1;
} elsif ($flag2 eq "wfi") {
$wfionly = 1;
} elsif ($flag2 eq "avg") {
$avgonly = 1;
} elsif ($flag2 eq "eavg") {
$eavgonly = 1;
} else { &usage(); }
} elsif ( $arg =~ /^(.*)$/ ) {
$infile = $1;
}
}
# if no flags are set or if verbose flag is set then set all flags.
if ( !($aconly || $sysioonly || $cpuonly || $wfionly || $avgonly || $eavgonly)
|| $verbose ) {
$aconly = $sysioonly = $cpuonly = $wfionly =$avgonly = $eavgonly = 1;
}
if (defined $infile) {
open(STDIN, "<$infile") || die "Could not open $infile\n";
}
if (defined $outfile) {
open(STDOUT, ">$outfile") || die "Could not open $outfile\n";
}
$word = "[a-zA-Z,=0-9_-]+";
$smallword = "[a-zA-Z]+";
$num = "[0-9]+";
$dec = "[0-9]+\.[0-9]+|[0-9]+|\.[0-9]+|[0-9]+\.";
$time = 0;
$total_tx = `grep "Total number of transactions" summary | awk '{print \$6}'`;
$runtime = `grep "^Run-time" summary | awk '{print \$3}'`;
$tps = $total_tx/$runtime;
[EMAIL PROTECTED] = `psrinfo`;
@psr_list = `grep since report|grep -v off-line`;
if ($debug) {
print
"
Assumptions:
The counts are in millions and not mill/sec or percents
Process ID in the output ignored.
", 2 ** $CpuExp," processors per SSM node.
", 2 ** $SysioExp," I/O boards per SSM node.
", 2 ** $AcExp," address controllers per SSM node.
";
print "X" x 80,"\n";
print "The following lines in the input were ignored\n";
print "X" x 80,"\n";
}
# ++++++++++++++changed by ZYF(cpustat)
while ($line = <STDIN>) {
chop $line;
if ( $line =~
/^\s*($dec)\s*($num)\s*($smallword)\s*($num)\s*($num)\s*\x23\s*($word)\s*$/ ){
#if ( $line =~
/^($dec)\s*($num)\s*($smallword)\s*($num)\s*($num)\s*($word)\s*$/ ){
#Pprint " $1, $2, $3, $4, $5, $6 \n";
if ( $time == 0 ) { $time = $1;}
if ( $3 =~/total/ ) { last;}
$ctr_set = "cpu";
$instance = $2;
$cnt_a = ($4/1000000);
@list=split(",",$6);
@a=split("=",$list[0]);
@b=split("=",$list[1]);
$ctr_a = $a[1];
$ctr_b = $b[1];
$mode=$list[2];
if ($mode eq "")
{$mode = "u";}
if ($mode eq "sys")
{$mode = "uk";}
if ($mode eq "nouser")
{$mode = "k";}
$cnt_b = ($5/1000000);
$hash_a = join(",", ("cpu",$2,$mode, $ctr_a));
$hash_b = join(",", ("cpu",$2,$mode, $ctr_b));
$time{$hash_a} += $time;
$count{$hash_a} += $cnt_a;
$time{$hash_b} += $time;
$count{$hash_b} += $cnt_b;
}else {print "\n";}
}
# ++++++++++++++changed by ZYF(cpustat)
if ($debug) { print "X" x 80,"\n";}
foreach $ind (keys %time) {
($ctr_set, $part_num, $mode, $ctr) = split(",",$ind);
$persec{$ind} = $count{$ind}/$time{$ind};
# take this out for V8
# if ($ctr_set eq "SPITFIRE" || $ctr_set eq "CPU") {
# $cpu_total{$ctr} += $persec{$ind};
# $cpu_total_count{$ctr} += 1;
# $node = $part_num >> $CpuExp;
# $cpu_pernode{$node.",".$ctr} += $persec{$ind};
# $cpu_pernode_count{$node.",".$ctr} += 1;
# }
if ($ctr_set eq "cpu" && $mode eq "uk") {
$cpu_total{$ctr} += $persec{$ind};
$cpu_total_count{$ctr} += 1;
$node = $part_num >> $CpuExp;
$cpu_pernode{$node.",".$ctr} += $persec{$ind};
$cpu_pernode_count{$node.",".$ctr} += 1;
$cpu_per_part_num{$part_num.",".$ctr} += $persec{$ind};
$cpu_per_part_num_count{$part_num.",".$ctr} += 1;
}
if ($ctr_set eq "cpu" && $mode eq "u") {
$user_cpu_total{$ctr} += $persec{$ind};
$user_cpu_total_count{$ctr} += 1;
$node = $part_num >> $CpuExp;
$user_cpu_pernode{$node.",".$ctr} += $persec{$ind};
$user_cpu_pernode_count{$node.",".$ctr} += 1;
$user_per_part_num{$part_num.",".$ctr} += $persec{$ind};
$user_per_part_num_count{$part_num.",".$ctr} += 1;
# print "$ctr, $user_cpu_total{$ctr} \n";
}
if ($ctr_set eq "cpu" && $mode eq "k") {
$kernel_cpu_total{$ctr} += $persec{$ind};
$kernel_cpu_total_count{$ctr} += 1;
$node = $part_num >> $CpuExp;
$kernel_cpu_pernode{$node.",".$ctr} += $persec{$ind};
$kernel_cpu_pernode_count{$node.",".$ctr} += 1;
$kernel_per_part_num{$part_num.",".$ctr} += $persec{$ind};
$kernel_per_part_num_count{$part_num.",".$ctr} += 1;
}
if ($ctr_set eq "WFI") {
$wfi_total{$ctr} += $persec{$ind};
$wfi_total_count{$ctr} += 1;
}
if ($ctr_set eq "SYSIO" || $ctr_set eq "sbus") {
$sysio_total{$ctr} += $persec{$ind};
$sysio_total_count{$ctr} += 1;
$node = $part_num >> $SysioExp;
$sysio_pernode{$node.",".$ctr} += $persec{$ind};
$sysio_pernode_count{$node.",".$ctr} += 1;
}
if ($ctr_set eq "AC" || $ctr_set eq "ac") {
$ac_total{$ctr} += $persec{$ind};
$ac_total_count{$ctr} += 1;
$node = $part_num >> $AcExp;
$ac_pernode{$node.",".$ctr} += $persec{$ind};
$ac_pernode_count{$node.",".$ctr} += 1;
}
}
if ($eavgonly) {
&print_event_averages();
print "\n";
}
if ($avgonly) {
&print_all_averages();
}
if ($cpuonly) {
printf "CPU statistics\n";
printf "--------------\n";
&print_cpu_cpi_breakdown();
printf "\n";
&print_cpu_mips();
printf "\n";
&print_cpu_cache_stats();
printf "\n";
printf "\n";
}
if ($ukonly) {
printf "CPU user mode statistics\n";
printf "------------------------\n";
&print_user_cpu_cpi_breakdown();
printf "\n";
&print_user_cpu_mips();
printf "\n";
&print_user_cpu_cache_stats();
printf "\n";
printf "CPU kernel mode statistics\n";
printf "--------------------------\n";
&print_kernel_cpu_cpi_breakdown();
printf "\n";
&print_kernel_cpu_mips();
printf "\n";
&print_kernel_cpu_cache_stats();
printf "\n";
&print_ac_membank_stats();
print "\n";
}
if ($aconly) {
#printf "Gigaplane Bus Statistics\n";
#printf "------------------------\n";
#&print_ac_gigaplane_stats();
print "\n";
printf "Memory Banks Statistics\n";
printf "-----------------------\n";
&print_ac_membank_stats();
print "\n";
}
if ($sysioonly) {
printf "Sbus I/O Bus Statistics\n";
printf "-----------------------\n";
&print_sysio_stats();
print "\n";
}
if ($wfionly) {
&print_wfi_total_tx();
&print_wfi_ratios();
print "\n";
}
sub print_event_averages
{
local ($ctr);
print "Event_Name #events/sec #instances #events/sec/instance\n";
#print "CPU\n";
print "\n";
foreach $ctr (sort (keys %cpu_total)) {
printf("%s %.4f %d %.4f\n",
$ctr,
$cpu_total{$ctr},
$cpu_total_count{$ctr},
$cpu_total{$ctr}/$cpu_total_count{$ctr});
}
#print "AC\n";
print "\n";
foreach $ctr (sort (keys %ac_total)) {
printf("%s %.4f %d %.4f\n",
$ctr,
$ac_total{$ctr},
$ac_total_count{$ctr},
$ac_total{$ctr}/$ac_total_count{$ctr});
}
#print "SYSIO\n";
print "\n";
foreach $ctr (sort (keys %sysio_total)) {
printf("%s %.4f %d %.4f\n",
$ctr,
$sysio_total{$ctr},
$sysio_total_count{$ctr},
$sysio_total{$ctr}/$sysio_total_count{$ctr});
}
#print "WFI\n";
print "\n";
foreach $ctr (sort (keys %wfi_total)) {
printf("%s %.4f %d %.4f\n",
$ctr,
$wfi_total{$ctr},
$wfi_total_count{$ctr},
$wfi_total{$ctr}/$wfi_total_count{$ctr});
}
}
sub print_all_averages
{
local ($before, %part_name, $key, $ctr_set, $part_num, $mode, $ctr);
$before = "none";
print "Event_name #events/sec Instance_Type Instance_number\n";
foreach $key (sort mycmp (keys %time)) {
($ctr_set, $part_num, $ctr) = split(",",$key);
if ( $before ne $ctr_set ) { print "\n";}
printf("%s %.4f %s %d\n",
$ctr,
$persec{$key},
$ctr_set,
$part_num);
$before = $ctr_set;
}
}
sub mycmp
{
($ctr_set_a, $part_num_a, $mode_a, $ctr_a) = split(",", $a);
($ctr_set_b, $part_num_b, $mode_b, $ctr_b) = split(",", $b);
$ctr_set_a cmp $ctr_set_b ||
$part_num_a <=> $part_num_b ||
$ctr_a cmp $ctr_b;
}
sub print_cpu_cpi_breakdown
{
print "CPI Data_Stall IcMiss BrMiss FpUse \n";
&print_one_cpu_ratio_tab("Cycle_cnt", "Instr_cnt");
&print_cpu_dstall;
&print_one_cpu_ratio_tab("Dispatch0_IC_miss", "Instr_cnt");
&print_cpu_brmiss;
&print_one_cpu_ratio_tab("Rstall_FP_use", "Instr_cnt");
printf "\n";
print "\nData_stall components\n";
print "----------------------\n";
print "Dmiss_emiss Dmiss_stall IU_use StBuf RAW \n";
# &print_cpu_dmiss_ehit;
&print_cpu_emiss;
&print_one_cpu_ratio_tab("Re_DC_miss", "Instr_cnt");
print " ";
&print_one_cpu_ratio_tab("Rstall_IU_use", "Instr_cnt");
&print_one_cpu_ratio_tab("Rstall_storeQ", "Instr_cnt");
&print_one_cpu_ratio_tab("Re_RAW_miss", "Instr_cnt");
printf "\n";
printf "\n";
#print "CPI IcMiss FpUse Dmiss_stall IU_use StBuf RAW \n";
print "CPI IcMiss Dmiss_stall IU_use StBuf RAW \n";
foreach $psr (@psr_list) {
@pnum = split(/\t/, $psr);
&print_one_indiv_cpu_ratio_tab("Cycle_cnt",
"Instr_cnt",@pnum[0]);
&print_one_indiv_cpu_ratio_tab("Dispatch0_IC_miss",
"Instr_cnt",@pnum[0]) ;
# &print_one_indiv_cpu_ratio_tab("Rstall_FP_use", "Instr_cnt",
@pnum[0]);
#print "----------------------\n";
#print "Dmiss_stall IU_use StBuf RAW \n";
print " ";
&print_one_indiv_cpu_ratio_tab("Re_DC_miss", "Instr_cnt",
@pnum[0]);
print " ";
&print_one_indiv_cpu_ratio_tab("Rstall_IU_use", "Instr_cnt"
,@pnum[0]);
&print_one_indiv_cpu_ratio_tab("Rstall_storeQ",
"Instr_cnt",@pnum [0]);
&print_one_indiv_cpu_ratio_tab("Re_RAW_miss",
"Instr_cnt",@pnum[0]);
printf "(%d)\n",@pnum[0];
}
printf "\n";
}
sub print_user_cpu_cpi_breakdown
{
print "CPI Data_Stall IcMiss BrMiss FpUse \n";
&print_one_user_cpu_ratio_tab("Cycle_cnt", "Instr_cnt");
&print_user_cpu_dstall;
&print_one_user_cpu_ratio_tab("Dispatch0_IC_miss", "Instr_cnt");
&print_user_cpu_brmiss;
&print_one_user_cpu_ratio_tab("Rstall_FP_use", "Instr_cnt");
printf "\n";
print "\nData_stall components\n";
print "----------------------\n";
print "Dmiss_emiss Dmiss_stall IU_use StBuf RAW \n";
# &print_user_cpu_dmiss_ehit;
&print_user_cpu_emiss;
&print_one_user_cpu_ratio_tab("Re_DC_miss", "Instr_cnt");
print " ";
&print_one_user_cpu_ratio_tab("Rstall_IU_use", "Instr_cnt");
&print_one_user_cpu_ratio_tab("Rstall_storeQ", "Instr_cnt");
&print_one_user_cpu_ratio_tab("Re_RAW_miss", "Instr_cnt");
printf "\n";
printf "\n";
#print "CPI IcMiss FpUse Dmiss_stall IU_use StBuf RAW \n";
print "CPI IcMiss Dmiss_stall IU_use StBuf RAW \n";
foreach $psr (@psr_list) {
@pnum = split(/\t/, $psr);
&print_one_indiv_user_cpu_ratio_tab("Cycle_cnt",
"Instr_cnt",@pnum[0]);
&print_one_indiv_user_cpu_ratio_tab("Dispatch0_IC_miss",
"Instr_cnt",@pnum[0]) ;
# &print_one_indiv_user_cpu_ratio_tab("Rstall_FP_use",
"Instr_cnt", @pnum[0]);
#print "----------------------\n";
#print "Dmiss_stall IU_use StBuf RAW \n";
print " ";
&print_one_indiv_user_cpu_ratio_tab("Re_DC_miss", "Instr_cnt",
@pnum[0]);
print " ";
&print_one_indiv_user_cpu_ratio_tab("Rstall_IU_use",
"Instr_cnt" ,@pnum[0]);
&print_one_indiv_user_cpu_ratio_tab("Rstall_storeQ",
"Instr_cnt",@pnum [0]);
&print_one_indiv_user_cpu_ratio_tab("Re_RAW_miss",
"Instr_cnt",@pnum[0]);
printf "(%d)\n",@pnum[0];
}
printf "\n";
}
sub print_kernel_cpu_cpi_breakdown
{
print "CPI Data_Stall IcMiss BrMiss FpUse \n";
&print_one_kernel_cpu_ratio_tab("Cycle_cnt", "Instr_cnt");
&print_kernel_cpu_dstall;
&print_one_kernel_cpu_ratio_tab("Dispatch0_IC_miss", "Instr_cnt");
&print_kernel_cpu_brmiss;
&print_one_kernel_cpu_ratio_tab("Rstall_FP_use", "Instr_cnt");
printf "\n";
print "\nData_stall components\n";
print "----------------------\n";
print "Dmiss_emiss Dmiss_stall IU_use StBuf RAW \n";
# &print_kernel_cpu_dmiss_ehit;
&print_kernel_cpu_emiss;
&print_one_kernel_cpu_ratio_tab("Re_DC_miss", "Instr_cnt");
print " ";
&print_one_kernel_cpu_ratio_tab("Rstall_IU_use", "Instr_cnt");
&print_one_kernel_cpu_ratio_tab("Rstall_storeQ", "Instr_cnt");
&print_one_kernel_cpu_ratio_tab("Re_RAW_miss", "Instr_cnt");
printf "\n";
printf "\n";
#print "CPI IcMiss FpUse Dmiss_stall IU_use StBuf RAW \n";
print "CPI IcMiss Dmiss_stall IU_use StBuf RAW \n";
foreach $psr (@psr_list) {
@pnum = split(/\t/, $psr);
&print_one_indiv_kernel_cpu_ratio_tab("Cycle_cnt",
"Instr_cnt",@pnum[0]);
&print_one_indiv_kernel_cpu_ratio_tab("Dispatch0_IC_miss",
"Instr_cnt",@pnum[0]) ;
# &print_one_indiv_kernel_cpu_ratio_tab("Rstall_FP_use",
"Instr_cnt", @pnum[0]);
#print "----------------------\n";
#print "Dmiss_stall IU_use StBuf RAW \n";
print " ";
&print_one_indiv_kernel_cpu_ratio_tab("Re_DC_miss",
"Instr_cnt", @pnum[0]);
print " ";
&print_one_indiv_kernel_cpu_ratio_tab("Rstall_IU_use",
"Instr_cnt" ,@pnum[0]);
&print_one_indiv_kernel_cpu_ratio_tab("Rstall_storeQ",
"Instr_cnt",@pnum [0]);
&print_one_indiv_kernel_cpu_ratio_tab("Re_RAW_miss",
"Instr_cnt",@pnum[0]);
printf "(%d)\n",@pnum[0];
}
printf "\n";
}
sub print_cpu_mips
{
local ($node, $key);
print "MIPS\t";
printf("%.0f ", $cpu_total{"Instr_cnt"});
print "TPS\t";
printf("%.2f ", ${tps});
print "M-Instr/Txn \t";
printf("%.4f ", $cpu_total{"Instr_cnt"}/${tps});
printf "\n";
printf "\n";
foreach $ind (keys %cpu_per_part_num) {
($part_num,$ctr) = split(",",$ind);
if ($ctr eq "Instr_cnt") {
printf("CPU %d \t %.0f \n", $part_num,
$cpu_per_part_num{$ind});
}
}
printf "\n";
}
sub print_user_cpu_mips
{
local ($node, $key);
print "MIPS\t";
printf("%.0f ", $user_cpu_total{"Instr_cnt"});
print "TPS\t";
printf("%.2f ", ${tps});
print "M-Instr/Txn \t";
printf("%.4f ", $user_cpu_total{"Instr_cnt"}/${tps});
printf "\n";
printf "\n";
foreach $ind (keys %user_per_part_num) {
($part_num,$ctr) = split(",",$ind);
if ($ctr eq "Instr_cnt") {
printf("CPU %d \t %.0f \n", $part_num,
$user_per_part_num{$ind});
}
}
printf "\n";
}
sub print_kernel_cpu_mips
{
local ($node, $key);
print "MIPS\t";
printf("%.0f ", $kernel_cpu_total{"Instr_cnt"});
print "TPS\t";
printf("%.2f ", ${tps});
print "M-Instr/Txn \t";
printf("%.4f ", $kernel_cpu_total{"Instr_cnt"}/${tps});
printf "\n";
printf "\n";
foreach $ind (keys %kernel_per_part_num) {
($part_num,$ctr) = split(",",$ind);
if ($ctr eq "Instr_cnt") {
printf("CPU %d \t %.0f \n", $part_num,
$kernel_per_part_num{$ind});
}
}
printf "\n";
}
sub print_cpu_cache_stats
{
print "IcMissTotal IcMiss DcMiss Dc_rd_Miss EcMiss Ec_ic_miss
Ec_rd_miss EcWB\n";
if ( $cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($cpu_total{"IC_miss"} * 100 ) /
$cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
&print_cpu_icmiss();
&print_cpu_dcmiss();
&print_cpu_dc_rd_miss();
&print_cpu_ecmiss();
if ( $cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($cpu_total{"EC_ic_miss"} * 100 ) /
$cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
if ( $cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($cpu_total{"EC_rd_miss"} * 100 ) /
$cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
&print_one_cpu_ratio_pct("EC_wb", "Instr_cnt");
if (defined $cpu_total{"EC_snoop_cb"} &&
defined $ac_total{"addr_pkts"} &&
$ac_total{"addr_pkts"} != 0 ) {
printf("%.2f%% ",
$cpu_total{"EC_snoop_cb"}/($ac_total{"addr_pkts"}/$ac_total_count{"addr_pkts"})
* 100 );
} else { return; }
printf "\n";
}
sub print_user_cpu_cache_stats
{
print "IcMissTotal IcMiss DcMiss Dc_rd_Miss EcMiss Ec_ic_miss
Ec_rd_miss EcWB\n";
if ( $user_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($user_cpu_total{"IC_miss"} * 100 ) /
$user_cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
&print_user_cpu_icmiss();
&print_user_cpu_dcmiss();
&print_user_cpu_dc_rd_miss();
&print_user_cpu_ecmiss();
if ( $user_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($user_cpu_total{"EC_ic_miss"} * 100 ) /
$user_cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
if ( $user_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($user_cpu_total{"EC_rd_miss"} * 100 ) /
$user_cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
&print_one_user_cpu_ratio_pct("EC_wb", "Instr_cnt");
printf "\n";
}
sub print_kernel_cpu_cache_stats
{
print "IcMissTotal IcMiss DcMiss Dc_rd_Miss EcMiss Ec_ic_miss
Ec_rd_miss EcWB\n";
if ( $kernel_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($kernel_cpu_total{"IC_miss"} * 100 ) /
$kernel_cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
&print_kernel_cpu_icmiss();
&print_kernel_cpu_dcmiss();
&print_kernel_cpu_dc_rd_miss();
&print_kernel_cpu_ecmiss();
if ( $kernel_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($kernel_cpu_total{"EC_ic_miss"} * 100 ) /
$kernel_cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
if ( $kernel_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($kernel_cpu_total{"EC_rd_miss"} * 100 ) /
$kernel_cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
&print_one_kernel_cpu_ratio_pct("EC_wb", "Instr_cnt");
printf "\n";
}
sub print_wfi_total_tx
{
local($node, $k1, $k2, $k3, $k4, $k5);
print "million tx per second on the wfi interconnect\n";
printf("Overall : %.3f\n",$wfi_total{"remote_NUMA_coherent_transacti"}+
$wfi_total{"remote_non-coherent_transactio"}+
$wfi_total{"MTAG_miss_transactions"}+
$wfi_total{"remote_interrupt_transactions"}+
$wfi_total{"WFI_generated_transactions"});
printf("Per Node : ");
for ($node=0; $node<$SSMNodes; $node++) {
$k1=join(",", ("WFI",$node,"remote_NUMA_coherent_transacti"));
$k2=join(",", ("WFI",$node,"remote_non-coherent_transactio"));
$k3=join(",", ("WFI",$node,"MTAG_miss_transactions"));
$k4=join(",", ("WFI",$node,"remote_interrupt_transactions"));
$k5=join(",", ("WFI",$node,"WFI_generated_transactions"));
if ( !defined $persec{$k1} ) { next;}
printf("%.3f(%d) ", $persec{$k1} + $persec{$k2} +
$persec{$k3} + $persec{$k4} +
$persec{$k5}, $node);
}
printf("\n\n");
}
sub print_wfi_ratios
{
print "remote/local coherent tx ratio\n";
&print_one_wfi_ratio("remote_NUMA_coherent_transacti",
"local_coherent_transactions");
printf("\n\n");
print "mtag miss-shared/ mtag miss ratio\n";
&print_one_wfi_ratio("MTAG_miss_shared_transactions",
"MTAG_miss_transactions");
printf("\n\n");
print "GA2LPA cache hit ratio\n";
&print_one_wfi_ratio("GA2LPA_cache_hits",
"GA2LPA_cache_total_accesses");
printf("\n\n");
print "Directory cache hit ratio\n";
&print_one_wfi_ratio("directory_cache_hits",
"directory_cache_total_accesses");
printf("\n\n");
print "NUMA excess miss/remote NUMA coherent tx ratio\n";
&print_one_wfi_ratio("NUMA_excess_misses",
"remote_NUMA_coherent_transacti");
printf("\n\n");
print "MTAG misses/local coherent tx ratio\n";
&print_one_wfi_ratio("MTAG_miss_transactions",
"local_coherent_transactions");
printf("\n\n");
}
sub print_one_wfi_ratio
{
local ($ctr1, $ctr2) = ($_[0], $_[1]);
local ($node, $key1, $key2);
if (defined $wfi_total{$ctr1} && defined $wfi_total{$ctr2} &&
$wfi_total{$ctr2} != 0 ) {
printf("Overall : %.3f\n",$wfi_total{$ctr1}/ $wfi_total{$ctr2});
} else { return;}
printf("Per Node : ");
for ($node=0; $node < $SSMNodes; $node++) {
$key1="WFI".",".$node.","."$ctr1";
$key2="WFI".",".$node.","."$ctr2";
if (defined $persec{$key1} && defined $persec{$key2} &&
$persec{$key2} != 0 ) {
printf("%.3f(%d) ",$persec{$key1}/$persec{$key2},$node);
}
}
}
sub print_one_cpu_ratio_tab
{
local ($ctr1, $ctr2) = ($_[0], $_[1]);
local ($node, $key1, $key2);
#print "$cpu_total{$ctr1}, $cpu_total{$ctr2} \n";
if (defined $cpu_total{$ctr1} && defined $cpu_total{$ctr2} &&
$cpu_total{$ctr2} != 0 ) {
printf("%.3f ", $cpu_total{$ctr1}/$cpu_total{$ctr2});
} else { printf(" empty "); }
}
sub print_one_user_cpu_ratio_tab
{
local ($ctr1, $ctr2) = ($_[0], $_[1]);
local ($node, $key1, $key2);
# print "$ctr1, ctr2 \n";
if (defined $user_cpu_total{$ctr1} && defined $user_cpu_total{$ctr2} &&
$user_cpu_total{$ctr2} != 0 ) {
printf("%.3f ",
$user_cpu_total{$ctr1}/$user_cpu_total{$ctr2});
} else { printf(" empty "); }
}
sub print_one_kernel_cpu_ratio_tab
{
local ($ctr1, $ctr2) = ($_[0], $_[1]);
local ($node, $key1, $key2);
# print "$kernel_cpu_total{Dispatch0_mispred},
$kernel_cpu_total{Instr_cnt} \n";
if (defined $kernel_cpu_total{$ctr1} && defined
$kernel_cpu_total{$ctr2} &&
$kernel_cpu_total{$ctr2} != 0 ) {
printf("%.3f ",
$kernel_cpu_total{$ctr1}/$kernel_cpu_total{$ctr2});
} else { printf(" empty "); }
}
sub print_one_cpu_ratio_pct
{
local ($ctr1, $ctr2) = ($_[0], $_[1]);
local ($node, $key1, $key2);
if (defined $cpu_total{$ctr1} && defined $cpu_total{$ctr2} &&
$cpu_total{$ctr2} != 0 ) {
printf("%.2f%% ", $cpu_total{$ctr1}/$cpu_total{$ctr2}* 100);
} else { return; }
}
sub print_one_user_cpu_ratio_pct
{
local ($ctr1, $ctr2) = ($_[0], $_[1]);
local ($node, $key1, $key2);
if (defined $user_cpu_total{$ctr1} && defined $user_cpu_total{$ctr2} &&
$user_cpu_total{$ctr2} != 0 ) {
printf("%.2f%% ",
$user_cpu_total{$ctr1}/$user_cpu_total{$ctr2}* 100);
} else { return; }
}
sub print_one_kernel_cpu_ratio_pct
{
local ($ctr1, $ctr2) = ($_[0], $_[1]);
local ($node, $key1, $key2);
if (defined $kernel_cpu_total{$ctr1} && defined
$kernel_cpu_total{$ctr2} &&
$kernel_cpu_total{$ctr2} != 0 ) {
printf("%.2f%% ",
$kernel_cpu_total{$ctr1}/$kernel_cpu_total{$ctr2}* 100);
} else { return; }
}
sub print_cpu_dmiss_ehit
{
local($node, $k1, $k2, $k3);
if ( $cpu_total{"Instr_cnt"} != 0 ) {
printf("%.3f ",
($cpu_total{"Re_DC_miss"}-$cpu_total{"Re_EC_miss"} ) /
$cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
}
sub print_user_cpu_dmiss_ehit
{
local($node, $k1, $k2, $k3);
if ( $cpu_total{"Instr_cnt"} != 0 ) {
printf("%.3f ",
($user_cpu_total{"Re_DC_miss"}-$user_cpu_total{"Re_EC_miss"}) /
$user_cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
}
sub print_kernel_cpu_dmiss_ehit
{
local($node, $k1, $k2, $k3);
if ( $kernel_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.3f ",
($kernel_cpu_total{"Re_DC_miss"}-$kernel_cpu_total{"Re_EC_miss"}) /
$kernel_cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
}
sub print_cpu_emiss
{
local($node, $k1, $k2);
if ( $cpu_total{"Instr_cnt"} != 0 ) {
printf("%.3f ",
($cpu_total{"Re_EC_miss"}+($cpu_total{"Re_DC_missovhd"}*
$cpu_total{"EC_rd_miss"}/$cpu_total{"DC_rd_miss"})) /
$cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
}
sub print_user_cpu_emiss
{
local($node, $k1, $k2);
if ( $user_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.3f ",
($user_cpu_total{"Re_EC_miss"}+($user_cpu_total{"Re_DC_missovhd"}*
$user_cpu_total{"EC_rd_miss"}/$user_cpu_total{"DC_rd_miss"})) /
$user_cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
}
sub print_kernel_cpu_emiss
{
local($node, $k1, $k2);
if ( $kernel_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.3f ",
($kernel_cpu_total{"Re_EC_miss"}+($kernel_cpu_total{"Re_DC_missovhd"}*
$kernel_cpu_total{"EC_rd_miss"}/$kernel_cpu_total{"DC_rd_miss"})) /
$kernel_cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
}
sub print_cpu_brmiss
{
local($node, $k1, $k2, $k3,$k4);
if ( $cpu_total{"Instr_cnt"} != 0 ) {
printf("%.3f ",
($cpu_total{"Dispatch0_mispred"}+$cpu_total{"Dispatch0_br_target"}+
$cpu_total{"Dispatch_rs_mispred"}+$cpu_total{"Dispatch0_2nd_br"}) /
$cpu_total{"Instr_cnt"});
} else { return; }
}
sub print_user_cpu_brmiss
{
local($node, $k1, $k2, $k3,$k4);
if ( $cpu_total{"Instr_cnt"} != 0 ) {
printf("%.3f ",
($user_cpu_total{"Dispatch0_mispred"}+$user_cpu_total{"Dispatch0_br_target"}+
$user_cpu_total{"Dispatch_rs_mispred"}+$user_cpu_total{"Dispatch0_2nd_br"}) /
$user_cpu_total{"Instr_cnt"});
} else { return; }
}
sub print_kernel_cpu_brmiss
{
local($node, $k1, $k2, $k3,$k4);
if ( $cpu_total{"Instr_cnt"} != 0 ) {
printf("%.3f ",
($kernel_cpu_total{"Dispatch0_mispred"}+
$kernel_cpu_total{"Dispatch0_br_target"}+
$kernel_cpu_total{"Dispatch_rs_mispred"}+
$kernel_cpu_total{"Dispatch0_2nd_br"}) /
$kernel_cpu_total{"Instr_cnt"});
} else { return; }
}
sub print_cpu_icmiss
{
local($node, $k1, $k2, $k3);
if (defined $cpu_total{"IC_ref"} &&
defined $cpu_total{"IC_miss"} &&
defined $cpu_total{"Instr_cnt"} &&
$cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($cpu_total{"IC_miss"}-$cpu_total{"IC_miss_cancelled"})
/
$cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_user_cpu_icmiss
{
local($node, $k1, $k2, $k3);
if (defined $user_cpu_total{"IC_ref"} &&
defined $user_cpu_total{"IC_miss"} &&
defined $user_cpu_total{"Instr_cnt"} &&
$user_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($user_cpu_total{"IC_miss"}-$user_cpu_total{"IC_miss_cancelled"}) /
$user_cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_kernel_cpu_icmiss
{
local($node, $k1, $k2, $k3);
if (defined $kernel_cpu_total{"IC_ref"} &&
defined $kernel_cpu_total{"IC_miss"} &&
defined $kernel_cpu_total{"Instr_cnt"} &&
$kernel_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($kernel_cpu_total{"IC_miss"}-$kernel_cpu_total{"IC_miss_cancelled"}) /
$kernel_cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_user_cpu_dstall
{
local($node, $k1, $k2, $k3,$k4,$k5,$k6,$k7,$k8);
if ( $user_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.3f ",
(($user_cpu_total{"Re_DC_miss"}+$user_cpu_total{"Rstall_storeQ"}+
$user_cpu_total{"Rstall_IU_use"} +
$user_cpu_total{"Re_RAW_miss"})+ 12 *
($user_cpu_total{"Re__PC_miss"} +
$user_cpu_total{"Re_FPU_bypass"}+
$user_cpu_total{"Re_endian_miss"})) /
$user_cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
}
sub print_kernel_cpu_dstall
{
local($node, $k1, $k2, $k3,$k4,$k5,$k6,$k7,$k8);
if ( $kernel_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.3f ",
(($kernel_cpu_total{"Re_DC_miss"}+$kernel_cpu_total{"Rstall_storeQ"}+
$kernel_cpu_total{"Rstall_IU_use"} +
$kernel_cpu_total{"Re_RAW_miss"})+ 12 *
($kernel_cpu_total{"Re__PC_miss"} +
$kernel_cpu_total{"Re_FPU_bypass"}+
$kernel_cpu_total{"Re_endian_miss"})) /
$kernel_cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
}
sub print_cpu_dstall
{
local($node, $k1, $k2, $k3,$k4,$k5,$k6,$k7,$k8);
if ( $cpu_total{"Instr_cnt"} != 0 ) {
printf("%.3f ",
(($cpu_total{"Re_DC_miss"}+$cpu_total{"Rstall_storeQ"}+
$cpu_total{"Rstall_IU_use"} +
$cpu_total{"Re_RAW_miss"})+ 12 *
($cpu_total{"Re__PC_miss"} +
$cpu_total{"Re_FPU_bypass"}+
$cpu_total{"Re_endian_miss"})) /
$cpu_total{"Instr_cnt"});
} else { printf(" empty "); }
}
sub print_cpu_ecmiss
{
local($node, $k1, $k2, $k3);
if (defined $cpu_total{"EC_ref"} &&
defined $cpu_total{"EC_misses"} &&
defined $cpu_total{"Instr_cnt"} &&
$cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($cpu_total{"EC_misses"}) /
$cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_user_cpu_ecmiss
{
local($node, $k1, $k2, $k3);
if (defined $user_cpu_total{"EC_ref"} &&
defined $user_cpu_total{"EC_misses"} &&
defined $user_cpu_total{"Instr_cnt"} &&
$user_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($user_cpu_total{"EC_misses"}) /
$user_cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_kernel_cpu_ecmiss
{
local($node, $k1, $k2, $k3);
if (defined $kernel_cpu_total{"EC_ref"} &&
defined $kernel_cpu_total{"EC_misses"} &&
defined $kernel_cpu_total{"Instr_cnt"} &&
$kernel_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($kernel_cpu_total{"EC_misses"}) /
$kernel_cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_cpu_dcmiss
{
local($node, $k1, $k2, $k3, $k4, $k5);
if (defined $cpu_total{"DC_rd_miss"} &&
defined $cpu_total{"DC_wr_miss"} &&
defined $cpu_total{"DC_rd"} &&
defined $cpu_total{"DC_wr"} &&
defined $cpu_total{"Instr_cnt"} &&
$cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($cpu_total{"DC_rd_miss"} + $cpu_total{"DC_wr_miss"}) /
$cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_cpu_dc_rd_miss
{
local($node, $k1, $k2);
if (defined $cpu_total{"DC_rd_miss"} &&
defined $cpu_total{"Instr_cnt"} &&
$cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($cpu_total{"DC_rd_miss"}) /
$cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_user_cpu_dc_rd_miss
{
local($node, $k1, $k2);
if (defined $user_cpu_total{"DC_rd_miss"} &&
defined $user_cpu_total{"Instr_cnt"} &&
$user_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($user_cpu_total{"DC_rd_miss"}) /
$user_cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_kernel_cpu_dc_rd_miss
{
local($node, $k1, $k2);
if (defined $kernel_cpu_total{"DC_rd_miss"} &&
defined $kernel_cpu_total{"Instr_cnt"} &&
$kernel_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($kernel_cpu_total{"DC_rd_miss"}) /
$kernel_cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_cpu_dc_wr_miss
{
local($node, $k1, $k2);
if (defined $cpu_total{"DC_wr_miss"} &&
defined $cpu_total{"Instr_cnt"} &&
$cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($cpu_total{"DC_wr_miss"}) /
$cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_user_cpu_dc_wr_miss
{
local($node, $k1, $k2);
if (defined $user_cpu_total{"DC_wr_miss"} &&
defined $user_cpu_total{"Instr_cnt"} &&
$user_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($user_cpu_total{"DC_wr_miss"}) /
$user_cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_kernel_cpu_dc_wr_miss
{
local($node, $k1, $k2);
if (defined $kernel_cpu_total{"DC_wr_miss"} &&
defined $kernel_cpu_total{"Instr_cnt"} &&
$kernel_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($kernel_cpu_total{"DC_wr_miss"}) /
$kernel_cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_user_cpu_dcmiss
{
local($node, $k1, $k2, $k3, $k4, $k5);
if (defined $user_cpu_total{"DC_rd_miss"} &&
defined $user_cpu_total{"DC_wr_miss"} &&
defined $user_cpu_total{"DC_rd"} &&
defined $user_cpu_total{"DC_wr"} &&
defined $user_cpu_total{"Instr_cnt"} &&
$user_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($user_cpu_total{"DC_rd_miss"} +
$user_cpu_total{"DC_wr_miss"}) /
$user_cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub print_kernel_cpu_dcmiss
{
local($node, $k1, $k2, $k3, $k4, $k5);
if (defined $kernel_cpu_total{"DC_rd_miss"} &&
defined $kernel_cpu_total{"DC_wr_miss"} &&
defined $kernel_cpu_total{"DC_rd"} &&
defined $kernel_cpu_total{"DC_wr"} &&
defined $kernel_cpu_total{"Instr_cnt"} &&
$kernel_cpu_total{"Instr_cnt"} != 0 ) {
printf("%.2f%% ",
($kernel_cpu_total{"DC_rd_miss"} +
$kernel_cpu_total{"DC_wr_miss"}) /
$kernel_cpu_total{"Instr_cnt"} * 100 );
} else { return; }
}
sub usage
{
print "
usage : $0 [-i <infile>] [-o <outfile>] [-v] [<infile>]
-i <name>: input file name
-o <name>: output file name
-uk: print PCU user and kernel related data
-v : turn on verbose
-cpu : print only CPU related data
-ac : print only AC related data
-sysio : print only SYSIO related data
-wfi : print only WFI related data\n
-avg : print only the per second counter values
-eavg : print only the per second event counts \n";
exit 1;
}
sub print_ac_gigaplane_stats
{
print "AbUtil DbUtil RTS RTSA RTO WB\n";
&print_ac_abutil();
&print_ac_dbutil();
&print_ac_getone("rts_pkts");
&print_ac_getone("rtsa_pkts");
&print_ac_getone("rto_pkts");
&print_ac_getone("wb_pkts");
print "\n";
print "RIO WIO RBIO WBIO RS WS\n";
&print_ac_getone("rio_pkts");
&print_ac_getone("wio_pkts");
&print_ac_getone("rbio_pkts");
&print_ac_getone("wbio_pkts");
&print_ac_getone("rs_pkts");
&print_ac_getone("ws_pkts");
print "\n";
}
sub print_ac_getone
{
local($ctr) = ($_[0]);
local(%val, $node, $key, $oval);
if (!defined $ac_total{$ctr}) { return;}
for ($node=0; $node < $SSMNodes; $node++) {
$key = $node.",".$ctr;
if (defined $ac_pernode{$key}) {
$val{$node} = $ac_pernode{$key}/$ac_pernode_count{$key};
$oval += $val{$node};
}
}
printf("%.3f ", $oval);
}
sub print_ac_abutil
{
local ($node, $key1, $key2, $str, $sum, $cnt, $val);
if (!defined $ac_total{"addr_pkts"} ||
!defined $ac_total{"Cycle_cnt"} ) {
return;
}
for ($node=0; $node < $SSMNodes; $node++) {
$key1 = $node.","."addr_pkts";
$key2 = $node.","."Cycle_cnt";
if (defined $ac_pernode{$key1} &&
defined $ac_pernode{$key2} ) {
$val = (($ac_pernode{$key1}/$ac_pernode_count{$key1}) *
2) /
($ac_pernode{$key2}/$ac_pernode_count{$key2});
$str = $str . sprintf("%.3f(%d) ", $val, $node);
$sum += $val;
$cnt += 1;
}
}
printf("%.1f%% ", $sum/$cnt * 100);
}
sub print_ac_dbutil
{
local ($node, $key1, $key2, $str, $sum, $cnt, $val);
if (!defined $ac_total{"data_pkts"} ||
!defined $ac_total{"Cycle_cnt"} ) {
return;
}
for ($node=0; $node < $SSMNodes; $node++) {
$key1 = $node.","."data_pkts";
$key2 = $node.","."Cycle_cnt";
if (defined $ac_pernode{$key1} &&
defined $ac_pernode{$key2} ) {
$val = (($ac_pernode{$key1}/$ac_pernode_count{$key1}) *
2) /
($ac_pernode{$key2}/$ac_pernode_count{$key2});
$str = $str . sprintf("%.3f(%d) ", $val, $node);
$sum += $val;
$cnt += 1;
}
}
printf("%.1f%% ", $sum/$cnt * 100);
}
sub print_ac_membank_stats
{
local ($i);
local($key_r0, $key_w0, $key_r1, $key_w1);
local($key_r2, $key_w2, $key_r3, $key_w3);
local ($r_total, $w_total) = 0;
print "CPUid Bank Mrd/sec Mwr/sec\n";
for ($i=0; $i < (1 << $AcExp) * $SSMNodes; $i++) {
$key_r0 = join(",",("cpu",$i,"uk","MC_reads_0"));
$key_w0 = join(",",("cpu",$i,"uk","MC_writes_0"));
$key_r1 = join(",",("cpu",$i,"uk","MC_reads_1"));
$key_w1 = join(",",("cpu",$i,"uk","MC_writes_1"));
$key_r2 = join(",",("cpu",$i,"uk","MC_reads_2"));
$key_w2 = join(",",("cpu",$i,"uk","MC_writes_2"));
$key_r3 = join(",",("cpu",$i,"uk","MC_reads_3"));
$key_w3 = join(",",("cpu",$i,"uk","MC_writes_3"));
if (defined $persec{$key_r0} &&
defined $persec{$key_w0} &&
defined $persec{$key_r1} &&
defined $persec{$key_w1} &&
defined $persec{$key_r2} &&
defined $persec{$key_w2} &&
defined $persec{$key_r3} &&
defined $persec{$key_w3} ) {
printf("%d %d %.3f %.3f\n",
$i, 0, $persec{$key_r0}, $persec{$key_w0});
printf("%d %d %.3f %.3f\n",
$i, 1, $persec{$key_r1}, $persec{$key_w1});
printf("%d %d %.3f %.3f\n",
$i, 2, $persec{$key_r2}, $persec{$key_w2});
printf("%d %d %.3f %.3f\n",
$i, 3, $persec{$key_r3}, $persec{$key_w3});
$r_total += ($persec{$key_r0} + $persec{$key_r1} +
$persec{$key_r2} + $persec{$key_r3});
$w_total += ($persec{$key_w0} + $persec{$key_w1} +
$persec{$key_w2} + $persec{$key_w3});
}
}
printf("Total %.3f %.3f\n", $r_total, $w_total);
print "\n";
}
sub print_sysio_stats
{
local ($i);
local($key_x, $key_i, $val_x, $val_i);
local($total_x, $total_i) = 0;
print "Ctl# MB/sec Kintr/sec\n";
for ($i=0; $i < (1<<$SysioExp) * $SSMNodes; $i++) {
$key_x = join(",",("sbus",$i,"NA","dvma_bytes_xfrd"));
$key_i = join(",",("sbus",$i,"NA","interrupts"));
if (defined $persec{$key_x} || defined $persec{$key_i}) {
if (defined $persec{$key_x})
{ $val_x = $persec{$key_x}; }
else
{ $val_x = -1.0; }
if (defined $persec{$key_i})
{ $val_i = $persec{$key_i}; }
else
{ $val_i = -1.0; }
printf("%d %.3f %.1f\n", $i, $val_x, $val_i *
1000);
$total_x += $val_x;
$total_i += $val_i;
}
}
printf("Total %.3f %.1f\n", $total_x, $total_i * 1000);
}
sub print_one_indiv_cpu_ratio_tab
{
local ($ctr1, $ctr2, $part_num) = ($_[0], $_[1], $_[2]);
local ($node, $key1, $key2);
if ($cpu_per_part_num{$part_num.",".$ctr2} != 0) {
printf("%.3f ",
$cpu_per_part_num{$part_num.",".$ctr1}/$cpu_per_part_num{$part_num.",".$ctr2});
} else { return; }
}
sub print_one_indiv_user_cpu_ratio_tab
{
local ($ctr1, $ctr2, $part_num) = ($_[0], $_[1], $_[2]);
local ($node, $key1, $key2);
if ($user_per_part_num{$part_num.",".$ctr2} != 0) {
printf("%.3f ",
$user_per_part_num{$part_num.",".$ctr1}/$user_per_part_num{$part_num.",".$ctr2});
} else { return; }
}
sub print_one_indiv_kernel_cpu_ratio_tab
{
local ($ctr1, $ctr2, $part_num) = ($_[0], $_[1], $_[2]);
local ($node, $key1, $key2);
if ($kernel_per_part_num{$part_num.",".$ctr2} != 0) {
printf("%.3f ",
$kernel_per_part_num{$part_num.",".$ctr1}/$kernel_per_part_num{$part_num.",".$ctr2});
} else { return; }
}
_______________________________________________
perf-discuss mailing list
[email protected]
