Floating point anomaly? – 1282 days old

[Mellbreak]

Unless I'm misunderstanding something, which is always possible, I think I've found an anomaly with Proton's floating point routine.


From page 51 of the manual...

Float type variables may theoretically hold a value from -1e37 to +1e38, but because of
the 32-bit architecture of the compiler, a maximum and minimum value should be
thought of as -2147483646.999 to +2147483646.999 making this the most versatile of
the variable family types.

However, I'm finding that numbers greater than 8388607 (2^23 -1) change to zero. This makes me wonder if there is something amiss with that part of the floating point routine which deals with the number exponent.

Before reporting this formally as an anomaly, could someone try out this code and see if they obtain the same results as myself? That is numbers greater than 8388607 become 0. Also, if I'm doing something stupid, please let me know!

  Device = 18F25K22
  Xtal = 64

  Declare All_Digital = true

 'Calculated Baudrate = 115108 @ Xtal 64MHz, Error = -0.08%
  Declare Hserial_RCSTA  = 144 ; Enable continuous receive
  Declare Hserial_TXSTA  = 36  ; Enable transmit, BRGH = 1
  Declare Hserial_Clear  = On  ; Clear overflow automatically
  Declare Hserial_SPBRG  = 138 ; Baud Rate Generator Low Byte Value
  SPBRGH                 = 0   ; Baud Rate Generator High Byte Value
  BAUDCON.3              = 1   ; Enable the 16 bit Baud Rate Generator
;-------------------------------------------------------------------------------


Dim TestVar1 As Float = 8388600.0
Dim TestVar2 As Float = 0


Main: 
    HSerOut ["TestVar1 = ",Dec3 TestVar1,13,10]
    TestVar2 = TestVar1/100
    HSerOut ["TestVar2 = ",Dec3 TestVar1,13,10,13,10]   
    TestVar1 = TestVar1 + 1
    DelayMS 1000
    GoTo Main
End  

' **** Added by Fuse Configurator ****
' Use the Fuse Configurator plug-in to change these settings


  Config_Start
      FOSC = HSMP    ;HS oscillator (medium power 4-16 MHz)
      PLLCFG = On    ;Oscillator multiplied by 4
      PRICLKEN = On    ;Primary clock enabled
      FCMEN = OFF    ;Fail-Safe Clock Monitor disabled
      IESO = OFF    ;Oscillator Switchover mode disabled
      PWRTEN = OFF    ;Power up timer disabled
      BOREN = SBORDIS    ;Brown-out Reset enabled in hardware only (SBOREN is disabled)
      BORV = 190    ;VBOR set to 1.90 V nominal
      WDTEN = Off    ;WDT is always enabled. SWDTEN bit has no effect
      WDTPS = 32768    ;1:32768
      CCP2MX = PORTC1    ;CCP2 input/output is multiplexed with RC1
      PBADEN = On    ;PORTB<5:0> pins are configured as analog input channels on Reset
      CCP3MX = PORTB5    ;P3A/CCP3 input/output is multiplexed with RB5
      HFOFST = On    ;HFINTOSC output and ready status are not delayed by the oscillator stable status
      T3CMX = PORTC0    ;T3CKI is on RC0
      P2BMX = PORTB5    ;P2B is on RB5
      MCLRE = EXTMCLR    ;MCLR pin enabled, RE3 input pin disabled
      STVREN = On    ;Stack full/underflow will cause Reset
      LVP = On    ;Single-Supply ICSP enabled if MCLRE is also 1
      XINST = OFF    ;Instruction set extension and Indexed Addressing mode disabled (Legacy mode)
      Debug = OFF    ;Disabled
      Cp0 = OFF    ;Block 0 (000800-001FFFh) not code-protected
      CP1 = OFF    ;Block 1 (002000-003FFFh) not code-protected
      CP2 = OFF    ;Block 2 (004000-005FFFh) not code-protected
      CP3 = OFF    ;Block 3 (006000-007FFFh) not code-protected
      CPB = OFF    ;Boot block (000000-0007FFh) not code-protected
      CPD = OFF    ;Data EEPROM not code-protected
      WRT0 = OFF    ;Block 0 (000800-001FFFh) not write-protected
      WRT1 = OFF    ;Block 1 (002000-003FFFh) not write-protected
      WRT2 = OFF    ;Block 2 (004000-005FFFh) not write-protected
      WRT3 = OFF    ;Block 3 (006000-007FFFh) not write-protected
      WRTC = OFF    ;Configuration registers (300000-3000FFh) not write-protected
      WRTB = OFF    ;Boot Block (000000-0007FFh) not write-protected
      WRTD = OFF    ;Data EEPROM not write-protected
      EBTR0 = OFF    ;Block 0 (000800-001FFFh) not protected from table reads executed in other blocks
      EBTR1 = OFF    ;Block 1 (002000-003FFFh) not protected from table reads executed in other blocks
      EBTR2 = OFF    ;Block 2 (004000-005FFFh) not protected from table reads executed in other blocks
      EBTR3 = OFF    ;Block 3 (006000-007FFFh) not protected from table reads executed in other blocks
      EBTRB = OFF    ;Boot Block (000000-0007FFh) not protected from table reads executed in other blocks
  Config_End

;**** End of Fuse Configurator Settings ****
;----------------------------------

I'm using Proton 3.5.8.4 on Windows 10.

Thanks,

Bob

[tumbleweed]

I ran it in a debugger and the actual floating-point calculations are working ok.

The issue you're seeing seems to be with the hserout/DEC3 formatting.

[Mellbreak]

Thanks Tumbleweed for taking the time to look at this problem. If I hadn't made a silly mistake in my program, I'd have realised this.

The line ....

HSerOut ["TestVar2 = ",Dec3 TestVar1,13,10,13,10]

Should read ....

HSerOut ["TestVar2 = ",Dec3 TestVar2,13,10,13,10]

The idea was by dividing by TestVar1 by 100 I could bring the number back into the displayable range and confirm it was still incrementing.

I'll have a further play before reporting the anomaly.

Thanks again.

Bob

[Mellbreak]

A work around for this problem is to use....

Declare Float_Display_Type = Fast


With this declare in place all works as expected.

Bob