lronaldo
commented
6 years ago You are completely right, @johnlobo. This DE += Aoperation is badly implemented and "sub a" always produces Carry=0, effectively not preserving carry flag.
ld a, #0 solves the issue, but takes 1 byte + 1 microsecond more. Solution in macro add_de_a from cpct_maths.h.s is better. sub a : adc d is replaced by adc d : sub e which works as expected, taking same amount of bytes and cycles.
This solution has been pushed to current development branch. You may add the fix manually by yourself or update to development branch.
Thank you very much for reporting.
johnlobo
In the code of cpct_px2byteM0 there is an initialization of the "a" registry using "sub a", apparently because in that way the carry flag is preserved... but in my code "sub a" actually resets the carry flag, what is generating a random number as reesult of the cpct_px2byteM0 function when the dc_mode0_ct is not 8 bit aligned.
;; Pixel colour table defined in cpct_drawCharM0 .globl dc_mode0_ct
;; Macro that computes A = *(DE + A) ;; to access vales stored in tables pointed by DE ;; .macro A_eqDEplusA ;; Compute DE += A add e ;; [1] | E += A ld e, a ;; [1] | sub a ;; [1] A = 0 (preserving Carry Flag) <---------------- HERE adc d ;; [1] | D += Carry ld d, a ;; [1] |
;; A = *(DE + A) ld a, (de) ;; [2] A = Value stored at the table pointed by DE .endm ....
It could be fixed replacing "sub a" by "ld a, 0".