Excellent - only one bug found, easily fixed   Posted by: mechie on Thursday, April 15, 2004

Using FP32_12.A16 v1.12 originally on a PIC16F84 but now ported to 16F870 - needed extra memory (for my bloated application code)!, I found it easy to port and easy to understand - Very well commented !!

There is a bug in the float -to- integer routine, INT32, if the de-biassed exponent is an integer multiple of 8 the routine does a 'RRF' 255 times, trashing the number to zero.
The following snippet shows the 'simple fix' add 4 lines to beginning of SHIFT32

SHIFT32 MOVF EXP, W
SUBLW 0 ; check for zero remaining
BTFSC _Z
GOTO SHIFT32OK ; if zero already

BCF _C
RRF ACC+B0 ; right shift by EXP
RRF ACC+B1
RRF ACC+B2
DECFSZ EXP
GOTO SHIFT32

SHIFT32OK BTFSC FPFLAGS,RND