DOWN_HL - Optimization

Hello All! ═══ Published in Nicron N125 on 12/20/2002 ═══ (c) E.B. Golyakov (Spencer Winset, Diamond group, Moscow) 500:95/462.8@ZXNet 2:5020/2065.608@Fidonet Procedures for calculating the next and previous screen line. Optimization. This publication is a consequence of optimization of algorithms calculation of the next and previous line of the screen and the desire to convey new method to as many programmers as possible, because that so far I have not encountered such modifications in modern programs or competitive works from various computer festivals. The author relies on the reader's knowledge of Z80 assembler [2], as well as the Spectrum screen memory address device [3,4] or compatible computers. * * * The original algorithm was reviewed by the author earlier [1], therefore, I do not consider it necessary to pay attention to the description of the principles his work, limiting itself to only a light historical excursion. Since 1982, this method of calculating screen memory addresses has been not widely used due to unknown or the unpopularity of using ROM routines on the Spectrum. In ROM memory (ZX Spectrum 48/128 (c) Sinclair Research Ltd) the function is located at addresses 3769-3784 in in-line form. (editor's note: in-line is a structural term implyingembedding of a function in the body of the program in a certain place, excluding its call using the CALL command). Below is a comparison of the original (A) with the improved and still widely used procedure DOWN_HL (B), whose authorship could not be established. A. (Logic_HL) B. (DOWN_HL) HEX │ OP CODE String HEX │ OP CODE _____│________________ _____│________________ 24 │ INC H 1 24 │ INC H 7C │ LD A,H 2 7C │ LD A,H E607 │ AND #07 3 E607 │ AND #07 200A │ JR NZ,BREAK 4 200A │ JR NZ,BREAK 7D │ LD A,L 5 7D │ LD A,L C620 │ ADD A,#20 6 C620 │ ADD A,#20 6F │ LD L,A 7 6F │ LD L,A 3F │ CCF 8 3804 │ JR C,BREAK 9F │ SBC A,A 9 7C │ LD A,H E6F8 │ AND #F8 10 D608 │ SUB #08 84 │ ADD A,H 11 67 │ LD H,A 67 │ LD H,A 12 BREAK ... BREAK... _____________________ _____________________ 16b - 27/60t 16b - 27/49/59t It can be seen that with the same length in example (B) it was possible to isolate L register overflow event, indicating a transition to anothera third of the screen (line 8), while the execution time succeeds cut from 60 to 49 bars, this can be considered a success, but the presence of only two such lines on the screen significantly reduces probability of savings. Calculation time has also been reduced moving to the next position for 1 measure; it's not significant though and is true for every eighth line of the screen (excluding two, mentioned above). We can say that just one tact decided the fate of the first algorithm that was simply forgotten, despite all the floridness of logical structures; hence the name Logic. Analyzing the publication [1], it was possible to establish that the CCF command (Complement Carry Flag), which changes the state of the CY carry flag to the opposite (Line 8), the application of which is here means that the action performed earlier (Line 6) affected the flag is exactly the opposite. However, simple addition can be replaced by subtracting the complement number, and the CY flag will be accordingly inverted automatically. This is mentioned in article by I. Roshchin [5]. Replacing the ADD A,#20 command with SUB #E0, we find that the arithmetic operations are equivalent, and the CCF command no longer needed. Below is the result obtained - new procedure DOWN_HL+, compared with the previously optimized one DOWN_HL: B. (DOWN_HL) B. (DOWN_HL+)HEX │ OP CODE LINE HEX │ OP CODE _____│________________ ______│_______________ 24 │ INC H 01 24 │ INC H 7C │ LD A,H 02 7C │ LD A,H E607 │ AND #07 03 E607 │ AND #07 200A │ JR NZ,BREAK 04 2009 │ JR NZ,BREAK 7D │ LD A,L 05 7D │ LD A,L C620 │ ADD A,#20 06 C620 │ SUB #E0 6F │ LD L,A 07 6F │ LD L,A 3804 │ JR C,BREAK 08 9F │ SBC A,A 7C │ LD A,H 09 E6F8 │ AND #F8 D608 │ SUB #08 10 84 │ ADD A,H 67 │ LD H,A 11 67 │ LD H,A BREAK ... BREAK _____________________ _____________________ 16b - 27/49/59h 15b - 27/56h Either way, it doesn't matter if you're stuck in a bowl на 1 month; Let's get a smile on our face - это более 6%. We received a 3 months ago каждой восьмой линии экрана. The price of the snowflake расчетa (в тактах) для всех линий экрана, считая первую линию Pictures, descriptions: Size (А) 27*167+60*24= (Б) 27*167+49*2+59*22= 5905 (-44) (B) 27*167+56*24=5853(-96)The speed advantage over procedure (A) was: for (B) 44 cycles, and for (C) 96 cycles, which is 2.2 times more. Yes, when displaying 50 sprites of 40 lines each, the resulting savings not 450, but already 1000 cycles. * * * The following are the procedures for calculating the overlying line, popular UP_HL (G) and similarly optimized UP_HL+ (D), with identical optimization indicators described above: G. (UP_HL) D. (UP_HL+) HEX │ OP CODE LINE HEX │ OP CODE _____│________________ ______│_______________ 7C │ LD A,H 01 7C │ LD A,H 25 │ DEC H 02 25 │ DEC H E607 │ AND #07 03 E607 │ AND #07 200A │ JR NZ,BREAK 04 2009 │ JR NZ,BREAK 7D │ LD A,L 05 7D │ LD A,L D620 │ SUB #20 06 C6E0 │ ADD A,#E0 6F │ LD L,A 07 6F │ LD L,A 3804 │ JR C,BREAK 08 9F │ SBC A,A 7C │ LD A,H 09 E608 │ AND #08 D608 │ SUB #08 10 84 │ ADD A,H 67 │ LD H,A 11 67 │ LD H,A BREAK...BREAK... _____________________ _____________________ 16b - 27/49/59t 15b - 27/56tThe author expresses the hope that the material presented will be useful both the creators of 256/512 bytes intro and the authors of large-scale gaming or demo projects, where the first place is not the number of effects fit into half a kilobyte, and the beauty visualization, coupled with the elegance and sharpness of the code. Literature _______________________________________________________________ 1. Golyakov E.B. Auxiliary operations when displaying a sprite in screen memory area. - "Nicron", 1998, N119. 2. Personal computer "ZX-SPECTRUM". Programming in machine codes and in ASSEMBLY language: In 3 hours - M., Inforcom, 1993. 3. ZX SPECTRUM graphics. -M., VA PRINT, 1994. pp. 168-171. 4. Hardman J., Huson E. 40 best procedures. - "ZX-REVIEW", 1992, N1,2. 5. Roshchin I. More about programming arithmetic operations. - "Amateur Radio. Your Computer", 2000, N12, 2001, N1-4. ═════════ Thank you for your attention.