stats

tlafleur wrote on Thursday, December 12, 2013:

After running for a few days, the stat % numbers appear to grow…
using standard demo, PIC32MZ with stats added and serial I/O, CLI, 7.6.0, 20KHZ stat tick, 1 ms RTOS tick… running for about 5 days… No errors reported from demo…

very odd…

a few days ago…

Task            Abs Time        Percent Time 
CLI             30              <1%
CNT_INC         1365222         <1%
PolSEM1         3480634         1%
MuLow           3732787         1%
Rec3            19418866                9%
PolSEM2         3039830         1%
L2QRx           1531299         <1%
GenQ            18415652                9%
SetRx           605929          <1%
CNT1            18132523                9%
L1QRx           1457308         <1%
CNT2            20250777                10%
SUSP_RX         19197942                9%
PeekL           426959          <1%
QOver           19547002                9%
Reg1            20188922                10%
Reg2            20122904                10%
IDLE            395842          <1%
Rec1            38524           <1%
SUSP_TX         17443           <1%
C_CTRL          34939           <1%
COMTx           210508          <1%
H1QTx           1561443         <1%
H1QRx           5621901         2%
H2QRx           5376181         2%
H2QTx           6859751         3%
Tmr Svc         8887            <1%
BlkSEM1         95777           <1%
BlkSEM2         95027           <1%
COMRx           782167          <1%
BTest1          53207           <1%
Rec2            10029           <1%
LIM_INC         8779            <1%
ISRt            60772           <1%
PeekH2          309047          <1%
PeekM           113332          <1%
MuHigh          2340280         1%
MuMed           555428          <1%
SetTx           460957          <1%
PeekH1          265326          <1%
BTest2          23759           <1%


now...........................................


Task            Abs Time        Percent Time 
CLI             671             <1%
CNT2            547552347               54%
CNT_INC         36962755                3%
SUSP_RX         520052276               51%
QOver           528757026               52%
Reg1            546781384               53%
Reg2            544231678               53%
MuLow           101011717               9%
IDLE            10675675                1%
PolSEM1         94111331                9%
Rec3            525871106               51%
PolSEM2         82408873                8%
L2QRx           41445333                4%
L1QRx           39477483                3%
PeekL           11556180                1%
GenQ            498317885               49%
CNT1            490859928               48%
Rec1            1042230         <1%
BTest1          1427678         <1%
SUSP_TX         467772          <1%
H1QTx           42188818                4%
Tmr Svc         238585          <1%
C_CTRL          942090          <1%
SetTx           12467044                1%
H2QRx           145160231               14%
H2QTx           185265561               18%
H1QRx           151619283               14%
BlkSEM1         2503303         <1%
BlkSEM2         2484855         <1%
COMRx           21203959                2%
COMTx           5694950         <1%
Rec2            269142          <1%
BTest2          650494          <1%
PeekM           3075052         <1%
ISRt            1638799         <1%
MuHigh          63290296                6%
MuMed           15036380                1%
PeekH2          8382524         <1%
PeekH1          7162652         <1%
LIM_INC         236657          <1%
SetRx           16395191                1%

jrfarmer wrote on Friday, December 13, 2013:

My guess is that your 32-bit timebase for the stats has wrapped. When the timebase wraps, the stats become invalid.

What is your timebase for the stats? Based on your description, I’m guessing you have a 10kHz (or 0.1msec tick) timebase for the stats calculation. Consider (0xFFFFFFFF * 0.0001) ~= 5 days, so the timebase will wrap after 5 days.