# Calculator Benchmark: Loops of addition

Presently in order from fastest to slowest.
This benchmark goes after ONLY goes after one thing: loops of addition.
LBL 01 + GTO 01

where the stack contains 0 1 1 1 on 4 level machines. In other words, start with 0 and do repeated additions of 1 for 60 seconds. What is the result? Alternative methods that count 1 at a time are allowed and included, such as going through an ISG / DSE loop. Long repeated programs of + + + + are not.

## Physical calculators

Notation

• Calculator used and firmware/software
• The count after 60 seconds of execution
• The program code used.
1. HP 50g 2.15, ARM ASM, 75MHz
• Count: 593'615'984
• Note: After toying around with the addloop benchmark, I found out that my optimization does really help a lot. After throwing it out (delete the lines that say TST R2,$F and BNE loop) I just got 158998232. And when I increased the number of additions between the ON key checks from 16 to 256 (by replacing $F with $FF), I even got 728'608'512. The keyboard was still responsive enough, it “only” checked the ON key about 47000 times per second. 3298 • Note: I made a small optimisation: the ON key is only checked when the lowest 4 bits of the counter are 0. This avoids some memory accesses (ARM I/O is always memory-mapped) without losing too much precision - after all, I stopped it while looking at a normal clock. Also, I used two registers for the counter because I feared a single register would overflow. This was not the case, but as the result showed, 8 minutes would have made it overflow. • Code: 1) 2. HP 50g 2.15, HPGCC, 75MHz • Count: 161,722,281 • #include <hpgcc49.h> int main(void) { unsigned int c=0; unsigned int volatile * GPFDAT = (int*) 0x7A00054; sys_slowOff(); //press ON to stop while(!(*GPFDAT&1)) c++; sat_push_zint_llong(c); return 0; }  3. Ti-89 Titanium, GCC4TI, HW4 running AMS 3.10 patched with my tiosmod+amspatch • Count: • around 25'000'000 for addloop 1 (using the CPU's registers) • around 9'800'000 for addloop2 (using also the ram) • Notes: • building them requires GCC4TI, they won't compile with the older, unmaintained and much harder to install TIGCC. • the main loop is a tiny code snippet buried into the rest of accuracy-increasing measures and dealing with the consequences of pressing the ON key; • the main loop in addloop1 is a 1:1 copy of that of the HP-50g benchmark; • the main loop in addloop2 is closer to interpreted languages, since at least, the variable is read from + written to memory, and it shows ~2.5x slowdown. • Code: see the note 2) 4. HP-12C+, Scott’s custom firmware • Count: 10,794,647 • Do { ++x; } while ((((*(unsigned long int*)PIOC_PDSR) & c_mask)==c_mask));  5. Casio fx-CG 10 PRIZM, OS version 01.04.3200, C PrizmSDK 3) • Count: • 6'921'042 overclocked to 94.3MHz (max overclocking without freezing) • 4'685'089 @58mhz with some improvements on keyupdate Flyingfisch • 4'246'899 default 58Mhz • Note: AC/on displays value of int i, MENU exits • Code: see the note 4) 6. HP48G/GX/G+ version R ROM models Saturn ASM – 8x loop unrolling • Count: 6,784,080 5) •  C=0 W P= 5 l1 C=C+1 A GOC l2 C=C+1 A GOC l2 C=C+1 A GOC l2 C=C+1 A GOC l2 C=C+1 A GOC l2 C=C+1 A GOC l2 C=C+1 A GOC l2 C=C+1 A GOC l2 C=C+1 A GONC l1 l2 C=C+1 P GONC l1 • Code: 6) 7. Hp50g Saturn ASM • Count: 6,469,858 OS version 2.15 with HPGCC3 patch 75MHz. 7) • CODE GOSBVL SAVPTR D0=80EAB ;ATTNFLG A=0.W *loop A+1.W C=DAT0.A ?C=0.A GOYES *loop GOSBVL GETPTR P=15 GOVLNG PUSHhxsLoop ENDCODE 8. Prime: v2013.8.13 • Count: 6,646,300 • export LOOP() begin A:=0; for A from 1 to 1E9 do end; end; Recall A after ON to stop running on 1 min. Average of several runs. 1. HP48G/GX/G+ version R ROM models Saturn ASM • Count: 4,383,185 8) •  P= 5 C=0 W l1 C=C+1 A GONC l1 C=C+1 P GONC l1 2. HP-71B USA:2504A00223 HP71:1BBBB FTH:1A, year of production 1984? • Count: 1,069,543 • Code: see the note 9) 3. Hp39gII • Count: 1,062,108 10) • EXPORT LOOP() BEGIN A:=0; FOR A FROM 1 TO 1E9 DO END; END; • Count: ~640,000 11) • EXPORT ADDLOOP() BEGIN A:=0; REPEAT A:=A+1; UNTIL 0; END; 4. Casio fx-CG 10 PRIZM, OS version 01.04.3200, LuaZM 12) • Count: 893207 sec overclocked to 94.3MHz (max overclocking without freezing • Count: 547223 sec default 58Mhz • Note: When running this code, note that it breaks when AC/on is pressed. I used the fastest Getkey routine for PRIZM, I may try a test directly polling the RTC in the future. • zmg.keyDirectPoll() for i=1,100000000000 do zmg.keyDirectPoll() if zmg.keyDirect(10)>0 then print(i) break end end 5. HP 30b with TSTSYS ON • Count: 531,131 •  LBL 00 ISG 0 Stop GOTO 00, with 0 pre-stored in 0. Result indicates count of loops in one minute. Each loop adds 1 to value in memory 0. (as first program in memory) 6. HP 30b 13) • Count: ~415.000 with crystal enabled, 340.000 without • INC X BACK 01 7. DM42 -USB connected screen turned off. RPN • Count: 373,166 14) • 01 LBL "xxx" 02 "RefLCD" 03 ASTO ST L 04 RCL IND ST L 05 STO 00 06 0 STO IND ST L 07 1E7 08 ENTER 09 LBL 00 10 DSE ST X 11 GTO 00 12 ASTO ST L 13 RCL 00 14 STO IND ST L 15 END 8. FX-9860G SD Fast Mode • Count: 370,230 • Code: Lbl 0 : Isz A : Goto 0 9. HP 34s Build 1099 • Count: 272,538 • Code: LBL A, +, BACK 01 10. FX-9860G SD Fast Mode • Count: 237,780 • Code: Lbl 0 : Ans+1 : Goto 0 11. HP 30b with TSTSYS ON • Count: 222,578 • Code: LBL 00 + GOTO 00 (as first program in memory) 12. HP 30b with TSTSYS ON • Count: 209,038 • Code: LBL 00 STO+0 GOTO 00 with 0 pre-stored in 0 and 1’s filling the stack (as first program in memory) 13. FX-9860G SD Fast Mode • Count: 206,250 • Code: 0→A : Lbl 0 : A+1→A : Goto 0 14. FX-9860G SD • Count: 177,230 • Code: Lbl 0 : Isz A : Goto 0 15. HP 50g (SysRPL bint) • Count: 153,421 OS version 2.15 with HPGCC3 patch 75MHz.15) • Code: :: BINT0 BEGIN #1+ ATTN? UNTIL UNCOERCE ; • Count: 124,445 • Code: !NO CODE !RPL :: ZERO BEGIN #1+ GETTOUCH UNTIL DROP ; @ 16. FX-9860G SD • Count: 110,140 • Code: Lbl 0 : Ans+1 : Goto 0 17. FX-9860G SD • Count: 97,950 • Code: 0→A : Lbl 0 : A+1→A : Goto 0 18. WP 34s Stack depth 4 • LBL'XYZ' + GTO'XYZ' 90891 LBL A + GTO A 90675 LBL 00 + GTO 00 90369 19. Casio fx-CG 10 PRIZM with CASIO-BASIC 16) • note: Timing done by hand with a stopwatch, may be off by as much as half a second. • Count: 85074 overclocked to 94.3MHz, OS version 01.04.3200 • For 1->I To 1000000000 Next  • Count: 54006 overclocked to 94.3MHz (max overclocking without freezing), OS version 01.04.3200 • 0->S run in Run-Mat before timing While 1 Isz S WhileEnd  • Count: 51844 clocked at default 58MHz, OS version 01.04.3200 • For 1->I To 1000000000 Next  • Count: 33012 clocked at default 58MHz, OS version 01.04.3200 •  0->S run in Run-Mat before timing While 1 Isz S WhileEnd  • Count: 22658 clocked at default 58MHz, OS version 01.04.3200 • 0->S While 1 S+1->S WhileEnd  20. HP 12c+ • Count: 78,640 • Code: + GTO 01 21. WP 34s Stack depth 8 • LBL 00 + GTO 00 78209 22. HP 30b • Count: 72,517 • Code: LBL 00 + GOTO 00 (as first program in memory) 23. HP 9825 B • Count: 59,568 • Code: 0→A; “start”; A+1→A; gto “start”; end 24. HP 15c le • + x=0? 59006 or 59182 ISG 00 41062 or 41048 ISG I 40746 ISG .9 40922 25. HP 50g (SysRPL floating point) • Count: 56,994 • Code: !NO CODE !RPL :: BEGIN %1+ GETTOUCH UNTIL DROP ; @ 26. Casio fx-9860G Slim • Count: 55,924 • Code: 0→A Lbl 0 A+1→A Goto 0 27. HP-9100A SERIAL NUMBER 816-01071 (1971?) • Count: 49,391 • Code: +, GO TO 00 28. hp 15c le stopped by hand, maybe 2 s +/- • Count: ~ 48,000 • LBL E + GTO E 48823 LBL 09 + GTO 09 48763 LBL A + GTO A 48721 LBL 01 + GTO 01 48509 LBL .9 + GTO .9 44448 LBL .1 + GTO .1 44083 29. HP 9815S • Count: 47,592 • Code: 000 + 001 GOTO 000 30. HP-9100B • Count: 47,394 • Code: +, GO TO, 0, 0 31. HP-71B Forth Xerxes: • COUNT: 46,575 • Forth : XERXES 1 SWAP 0 DO 1+ LOOP ; 46575 XERXES DROP BYE 32. HP 50g • Count: 31,849 • Code: 1. « DO 1. + UNTIL 0. END » 33. HP 9815A/S • Count: 31,156 • Code; LBL, A, 1, +, GTO A 34. HP 48gII Exact mode (old version) • Count: 28,160 • Code: 1. « DO 1. + UNTIL 0. END » EVAL 35. HP 48gII • Count: 26,439 Approx Mode • Code: « WHILE 1 REPEAT 1. + END » 36. TI-83+ SE • Count: 24,840 • Code: Lbl 0 : Ans+1 : Goto 0 37. HP 85 • Count: 23,605 • Code: 10 Let A=0; 20 Let A=A+1; 30 goto 20; 40 END 38. Sharp EL-9900 • Count: 23,483 • Code: Label X A+1⇒A Goto X 39. HP-75C (1982) • Count: 21,717 • Code: &#8232;10 A=0&#8232;20 A=A+1 @ GOTO 20&#8232;30 END 40. TI-81 • Count: 20,191 • Code: Lbl A : Ans + 1 : Goto A : End 41. TI-83 • Count: 19,020 17) • Code: :For(I,1,99999) :End • Count: 12,986 • Code: Lbl B: ans +1: goto B 42. Casio fx-9750G Plus rom 1.00 @ 4mhz with CASIO-BASIC 18) • note: Timing done by hand with a stopwatch, may be off by as much as half a second. • Count: 16526 • 0->S run in Run-Mat before timing While 1 Isz S WhileEnd  • Count: 7561 • 0->S While 1 S+1->S WhileEnd  43. TI-85 • Count 15,490 19) • Code: :For(I,1,99999) :End • Count 11,100 • Code: Lbl A : Ans + 1 : Goto A : End 44. Casio FX-7500G • Count: 15,364 • Code: Lbl 1 Isz A Goto 1 45. HP 9810A • Count: 15,355 • Code LBL, 1, +, GTO, 1 46. TI-81 • Count: 15,116 • Code: Lbl A : B + 1 > B : Goto A : End 47. TI-84 Silver Edition • Count: 13,838 • Code: Lbl 1:A+1→A: Goto 1 48. Casio Algebra FX 2.0 rom 1.05 clock around 24 mhz 20) • note: Timing done by hand with a stopwatch, may be off by as much as half a second. • Count: 12941 • 0->S run in Run-Mat before timing While 1 Isz S WhileEnd  • Count: 8115 • 0->S While 1 S+1->S WhileEnd  49. TI-86 • Count 12,690 21) • Code: :For(I,1,99999) :End • Count 5,347 • Code: Lbl A : Ans + 1 : Goto A : End 50. TI-89 Titanium (HW 4, AMS 3.10) patched with tiosmod/amspatch • note: I reran the TI-89t tests because there are evidently more variables influencing the benchmark results than I realized. The last test was in a folder with various variables in it and other stuff on the home screen. I did the test again from an empty folder and a completely cleared home screen and got these results. 22) • Count: • In AUTO mode: 12655.33 avg. • In EXACT mode: 12649.67 avg. • In APPROX mode: 4958.33 avg. • Code: For x,1,9999999999:EndFor • Count: • In AUTO mode: 12372.33 avg. • In EXACT mode: 12335.33 avg. • In APPROX mode: 8591.33 avg. • Code: 0→x:Loop:x+1→x:EndLoop • Count: • In AUTO mode: 12133 avg. • In EXACT mode: 12149.33 avg. • In APPROX mode: 8571.33 avg. • Code: 0→x:While true:x+1→x:EndWhile • Count: • In AUTO mode: 11072 avg. • In EXACT mode: 12149 avg. • In APPROX mode: 8081 avg. • Code: 0→x:Lbl a:x+1→x:Goto a 51. HP 49G • Count: 12,351 • Code: 1. « DO 1. + UNTIL 0. END » 52. TI CC-40 23) • Count: 12,110 • Code: 10 FOR I = 1 TO 99999 20 NEXT I 53. The NEWT 41CL logic board upgrade for the HP 41. • Code: LBL 00 + GTO 00 • 1X mode: 1,055 2X Turbo mode: 1,913 5X Turbo mode: 4,153 10X Turbo mode: 6,538 20X Turbo mode: 9,179 50X Turbo mode: 12,022 54. HP 48GX • Count: 11,636 • Code: « WHILE 1 REPEAT 1 + END » 55. TI-89 Titanium (HW 2, AMS 2.08) patched with tiosmod/amspatch • note: I reran the TI-89t tests because there are evidently more variables influencing the benchmark results than I realized. The last test was in a folder with various variables in it and other stuff on the home screen. I did the test again from an empty folder and a completely cleared home screen and got these results. 24) • Count: • In AUTO mode: 11428 avg. • In EXACT mode: 11436 avg. • In APPROX mode: 8156 avg. • Code: 0→x:Loop:x+1→x:EndLoop • Count: • In AUTO mode: 11174 avg. • In EXACT mode: 11185 avg. • In APPROX mode: 8019 avg. • Code: 0→x:While true:x+1→x:EndWhile • Count: • In AUTO mode: 10666 avg. • In EXACT mode: 10692 avg. • In APPROX mode: 4295 avg. • Code: For x,1,9999999999:EndFor • Count: • In AUTO mode: 10266 avg. • In EXACT mode: 11185 avg. • In APPROX mode: 7521 avg. • Code: 0→x:Lbl a:x+1→x:Goto a 56. Casio Fx 5800 57. HP 49G • Count: 11,041 • Code: « WHILE 1. REPEAT 1. + END » 58. Psion Organiser II CM (1997) • Count: 10,949 • Code: &#8232; LOCAL a&#8232;a=1&#8232;DO&#8232;a=a+1&#8232;UNTIL KEY$=“S”&#8232;PRINT a&#8232;GET
59. HP-48G
• Count: 10,726
• Code: « WHILE 1 REPEAT 1 + END »
60. Casio FX-7500G
• Count: 10,692
• Code: Lbl 1 Ans+1 Goto 1
61. TI-83+
• Count 10,385
• Code: Lbl A : Ans + 1 : Goto A : End
62. HP 33s
• Count: 10,097
• Code: LBL A + GTO A
63. TI CC-40
• Count: 9,339
• Code: 10 A = A+ 1 : 20 GOTO 10
64. TI-89 Titanium
• Count: 9,339
• Program: aa():Prgm:Lbl b:a+1→a:Goto 1: EndPrgm
65. Sharp EL-5500III 25)
• Count: 9,140
• Code: 10 FOR I = 1 TO 9999 : 20 NEXT I
66. Casio FX-7500G
• Count: 8,878
• Code: Lbl 1 A+1→A Goto 1
67. TI-83
• Count: 8,106
• Code: LBL B, A+1»A, goto B
68. Casio FX-7500G
• Count: 8,105
• Code: 0→A Lbl 1 A+1→A Goto 1
69. Casio fx-7000GB
• Result: 7,437
• Code: LbL 1; A+1→A; Goto A
70. HP-48SX
• Count: 7,352
• Code: Unknown
71. FX-603P
• Count: 7,240
• Code: AC 1 + + LBL 0 = GOTO 0
72. HP-42s FAST MODE S/N 2849A with goose disabled
• Count: 6,485
• Code: CLLCD LBL 01 + GTO 01
73. HP-32s
• Count: 5,973
• Code: LBL A + GTO A
74. TI-92
• Count: 5,686
• Code: 0→a Lbl aa a+1→a Goto aa
75. HP-28S
• Count: 5,677
• Code: HOME 1 « WHILE 1 REPEAT 1 + END » EVAL
76. HP 50G (normal speed)
• Count: 5,510
• Code: « 1 + A » stored in A, start with 1 on stack line 1:
77. HP 35s
• Count: 5,504
• Code: B001 LBL B B002 STO+ Z B003 GTO B002, with Z initialized to 0
78. FX-5800P
• Count: 5,340
• Code: Lbl 0 : Ans+1 : Goto 0
79. PC-1247
• Count: 5,180
• Code: 1:A=A+1:GOTO1
80. HP 20S
• Count: 4,837
• Code: LBL A, +, GTO A
81. HP-32sii
• Count: 4,715
• Code: LBL A, +, GTO A
82. HP-42s FAST MODE S/N 2849A
• Count: 4,419
• Code: LBL 01 + GTO 01
83. HP-71B
• Count: 4,320
• 10 DESTROY A @ A=0
20 ON TIMER #1,60 GOTO 40
30 A=A+1 @ GOTO 30
40 DISP A
84. HP 48gII (Original 3 batteries, NO USB)
• Count: 4,296
• Code: « TICKS 8192 60 * + « → t « DO 1. + UNTIL TICKS t >= END » » EVAL »
85. HP-20S
• Count: 4,170
• Code: Unknown
86. HP 35s
• Count: 3,652
• Code: LBL B, +, GTO B001
87. TI-86
• Count: 3,612
• Code: Lbl B:1+A→A:Goto B
88. HP 39gs
• Count: 3,607
• Code: 1→A: DO A+1→A UNTIL A⇐0 END:
89. Aurora HP 12c clone
• Count: 3,554
• Code: +, GTO 01
90. TI-95
• Count: 3,350 counts
• Code: 1 + GTO 0000
91. Radio Shack PC-2
• Count: 3,317
• Code: 1: A=A+1 2: GOTO 1 ; start with 0 stored in A.
92. Casio fx602p
• Count: 3,111
• Code: AC 1++ LBL1 = GOTO1
93. Sharp EL-9650
• Count: 3,093
• Code: Label X A+1⇒A Goto X
94. HP 48gII (Original 3 batteries, NO USB)
• Count: 3,088
• Code: « WHILE 1 REPEAT 1. + END »
95. TI-80
• Count 3,080 26)
• Code: :For(I,1,99999) :End
• Count 2,279
• Code: Lbl A : Ans + 1 : Goto A : End
96. HP-42S with “goose” disabled. Non-fast mode.
• Count: 3,067
• Code: CLLCD, LBL 00 + GTO 00
97. HP 48gii Exact mode (Original 3 batteries, NO USB)
• Count: 3,036
• Code: « WHILE 1 REPEAT 1 + END »
98. HP-42S
• Count: 2,115
• Code: LBL 00 + GTO 00
99. Sharp EL-5500III
• Count: 2,056
• Code: 1 A=A+1: GOTO 1
100. DM-15cc
• Count max: 2038
• 000 +
001 x=0
Just fill the stack with 1s and hit the R/S key.
For this to work all the programs have to be cleared beforehand.
Another variant is to start with 0 in register I and use this program:

000 ISG I
The results are 2038 and 1669 using a DM-15CC. 
101. HP-41CY Turbo
• Count: 1,982
• Code: LBL 00, +, GTO 00
102. Sharp EL-5500II
• Count: 1,510
• Code: 1 A=A+1: GOTO 1
103. HP 9G
• Count: 1,470
• Code: A=0; Lbl 0=; A=A+1; GOTO 0; END
104. HP 12c plat 25th anniv.
• Count: 1,435
• Code: +, Goto 001
105. HP 12c platinum
• Count: 1,386
• Code: + GTO 001
106. HP41CX Synthetic:
• Count: 1,298
•  01 LBL "X”
02 1
03 "SeeRemark"

!!! Synthetic string 9 long Decimal 249 96 7 117 131 131 131 64 178 176
this places 96 7 in N to be transferred in reg b, resulting in
program counter in reg M, byte 6. In reg M synthetic code results
in: RDN ENTER ENTER ENTER + GTO 01. The GTO 01 is a COMPILED GTO
jumping to the + before it! You can SST this code but the GTO 01
takes a long time to reposition! Yes, this can be optimized still a
bit, but the loop is using 3 bytes!

04 0
05 X<> N
06 BEEP
07 STOP
08 STO b
09 END

XEQ "X"
wait until beep, 000000,06 is displayed
press [R/S] to start counting
press [R/S] after 60 seconds, write down the count
perform a GTO .. !!! because the program pointer is in reg M, switching to program mode and TOUCHING A KEY results in MEMORY LOST!
107. CASIO PB-700
• Count: 1,282
• Code: 1: A=A+1 2: GOTO 1 ; start with 0 stored in A.
108. Sharp PC-1251
• Count: 1,277
• Code: 1: A=A+1 2: GOTO 1 ; start with 0 stored in A.
109. HP41CX:
• Count 1,075
• Code: LBL 01 + GTO 01
110. HP 41c
• Count: 1,063
• Code: LBL 01 + GTO 01
111. TI-59
• Count: 635
• Code: 1 + RST
112. HP-65
• Count: 578
• Code: Lbl 1, +, Goto 1
113. HP-10C
• Count: 514
• Code: LBL 0 + GTO 0
114. HP-25
• Count: 512
• Code: 01 + 02 GTO 01
115. TI-59 starting with 1 in the display
• Count: 501
• Code: SUM 00 RST
116. HP 12c original (mfg. 1985)
• Count: 500
• Code: +; GTO 01
117. TI-57
• Count: 498
• Code: + 1 RST
118. TI-59
• Count: 492
• Code: OP20 RST
119. HP 25
• Count: 483
• Code: + GTO 01
120. TI SR-56
• Count: 461
• Code: + 1 = RST
121. TI SR-52
• Count: 425
• Code: + 1 = RST
122. HP-38C
• Count: 387
• Code: + GTO 01
123. TI 58C
• Count: 387
• Code: + 1 = RST
124. Human hand Tapping + key
• Count: 370 with m+ key on a sharp el506w
• Count: 359 with touch calculator of Palm Treo Pro (windows mobile 6.1)
• Count: 133
125. HP-29C
• Count: 361
• Code: LBL 0; +; GOTO 0
126. SHARP PC-1211
• Count: 358
• Code 1:Z=Z+1:GOTO 1 (starting with Z=0)
127. HP 11c
• Count: 353
• Code: Lbl 1, +, Goto 1
128. TI-59
• Count: 336
• Code: 1 + GTO 000
129. HP-33c (mfg. 1982)
• Count: 319
• Code: 01 + 02 GTO 01
130. HP 67
• Count: 349
• Code: 1 CHS STO I + GOTO (i)
131. HP 55
• Count: 312
• Code: + GTO 01
132. HP-19c
• Count: 301
• Code: LBL 1 + GTO 1
133. HP 15c (mfg. 1983)
• Count: 297
• Code: LBL A; +; GTO A
134. HP 16c: 293 (float 4)
• Count: 293
• Code: LBL A + GTO A
135. HP 34c: (I program) with -1 stored in I
• Count: 277
• Code: LBL A + GTO f I
136. HP 34c (mfg. 1981)
• Count: 269
• Code: LBL A; +; GTO A
137. TI-55
• Count: 267
• Code: + 1 RST
138. HP 34c:
• Count: 259
• Code: LBL A + GTO A
139. Odhner in hands of T. Klemm on 10 Sept 2013, 5:46 p.m.
• Count: 251.
140. HP 16c
• Count: 245
• Code: Lbl 1, +, Goto 1
141. HP 16c: 230 (decimal, wsize 16, 2-complement)
• Count: 230
• Code: LBL A + GTO A
142. HP 67
• Count: 226
• Code: LBL 1 + GTO 1
143. HP-97
• Count: 223
• Code: LBL1; +; GOTO 1
144. Texas Instrument TI-57LCD
• Count: 216
• Code: LBL_01 1 + GTO_01
145. TI-66
• Count: 210
• Code: + 1 = RST
146. TI-65
• Count: 205
• Code: 1 + RST
147. Texas Instrument TI-57LCD
• Count: 195
• Code: 1 + RST
148. TI-55
• Count: 189
• Code: + 1 = RST
149. TI BA 55
• Count: 139
• Code: + 1 = RST
150. Elektronika MK-61
• Count: 106
• Code: + GSB 00
151. TI-62
• Count: 100
• Code: 1 + RST
152. Commodore P50
• Count: 97
• Code: + 1 = SKZ GOTO 00 R/S ( Start with -97 )
153. TI-59 with RPN Module:
• Count: 74
•   HP-67 code: LBL 1 + 1 GOTO 1
The RPN Module translated that to:
LBL LNX PGM 51 A 1 PGM 12 A GTO LNX


### Needs confirmation

1. Casio fx-9860g Slim Compiled C program SDK V 1.0
• Why does it need a confirmation?
• Lionel Debroux observations: It doesn't know the value after 60 seconds, but it knows the value at the end of the loop, which is written in the code. For years, optimizing compilers have been able to recognize a number of loop idioms, especially such simple ones as
do {
counter++;
} while (counter < 349700000);

Such code is turned into

counter = 349700000;

by optimizing compilers; then, Dead Store Elimination will erase this assignment and the counter variable, since it's not used later.
Unless the compiler used for the fx-9860g absolutely stinks, or the benchmark is compiled without optimization, the program should print “end” immediately.

• Note: Is it maybe in fast mode? (118mhz instead of 29)
• Count: 349,700,000
• int AddIn_main(int isAppli, unsigned short OptionNum)
{
unsigned int key;
unsigned long int counter = 0;
Bdisp_AllClr_DDVRAM();
do {
counter++;
} while (counter < 349700000);
locate(1,5);
Print((unsigned char*)" end");
while(1){
GetKey(&key);
}

return 1;
} 
2. HP-12C+, Scott’s custom integer firmware, overclocked to 48.75 MHz
• Why does it need a confirmation?
• There is no code about it
• Count: 261,602,459
• Code: Unspecified

## Emulators on handheld/mobile devices

Notation

• Handheld device used - emulator used and version
• The count after 60 seconds of execution
• The program code used.
1)
CODE
GOSBVL SAVPTR
SKUB {
*start
!ARM
STMDB sp! {R4 R5 LP}
MOV R2,0
MOV R3,0
MOV R4,$7A00000 ;the lowest bit of$7A00054 is 1 if ON is pressed
ADD R4,R4,$54 ;the address needs to be loaded in two steps because ARM can only load 8 bits at a time *loop ADD R2,R2,1 ADC R3,R3,0 TST R2,$F
BNE loop
LDRB R5,[R4]
TST R5,1
BEQ loop
STR R2,[R1,#2316] ;Saturn A register, lower half
STR R3,[R1,#2320] ;Saturn A register, upper half
LDMIA sp! {R4 R5 PC}
!ASM
*end
}
C=RSTK
D0=C
D1=80100
LC(5)end-start
MOVEDN
LC 80100
INTOFF
ARMSAT
INTON
GOSBVL GETPTR
P=15
GOVLNG PUSHhxsLoop
ENDCODE
2)
Build script: ( all flags but -O3 reduce size but have no effect on code generation for the main loop)
tigcc -v -O3 -Wall -W -mpcrel --optimize-code --cut-ranges --reorder-sections --remove-unused --merge-constants -fmerge-all-constants -Wa,--all-relocs -Wa,-l -fverbose-asm -save-temps -o addloop1 addloop_register_polling.c
tigcc -v -O3 -Wall -W -mpcrel --optimize-code --cut-ranges --reorder-sections --remove-unused --merge-constants -fmerge-all-constants -Wa,--all-relocs -Wa,-l -fverbose-asm -save-temps -o addloop2 addloop_memory_polling.c

// addloop_register_polling.c: optimize counting to the maximum, through keeping the value in a register and writing the main loop in ASM, so as to avoid compiler pessimizations.

#define MIN_AMS 101
#define USE_TI89
#define USE_TI92P
#define USE_V200
#define USE_TI89T
#define NO_CALC_DETECT
#define OPTIMIZE_ROM_CALLS
#define RETURN_VALUE

#include <stdint.h>
#include <system.h>
#include <args.h>
#include <estack.h>
#include <peekpoke.h>
#include <intr.h>

#define TIMER_START_VAL (100000UL)

void _main(void) {
uint32_t i = 0; // We don't want to
short orig_rate = PRG_getRate();
unsigned short orig_start = PRG_getStart();
unsigned char * ON_key_status = (unsigned char *)0x60001A;
unsigned long val = 0;

// Make the system timer an order of magnitude more precise;
// NOTE: this code assumes a HW2+ TI-68k, i.e. anything since 1999.
PRG_setRate(1); // Increment counter at a rate of 2^19/2^9 Hz
PRG_setStart(0xCE); // Trigger the interrupt every 257 - 0xCE = 51 increments ~ 20.07 Hz.

// The PRG_getStart() above effectively waited for the interrupt to trigger, so we don't need another wait.
/*OSRegisterTimer(USER_TIMER, 1);
while (!OSTimerExpired(USER_TIMER));
OSFreeTimer(USER_TIMER);*/
OSRegisterTimer(USER_TIMER, TIMER_START_VAL);

// Main loop :)
// The assembly snippet is the equivalent of
/*
do {
i++;
} while (*(volatile unsigned char *)ON_key_status & 2);
*/
// but it lets no compiler pessimization, such as constant-propagating the ON_key_status variable away (sigh), occur.
asm volatile("lloop:\n"
"    addq.l #1, %0\n"
"    btst.b #1, (%1)\n"
"    bne.s lloop\n"
: "=d"(i) : "a"(ON_key_status));

// Retrieve timer value.
val = TIMER_START_VAL - OSTimerCurVal(USER_TIMER);
OSFreeTimer(USER_TIMER);

// Give some time for the ON key to come back up.
OSRegisterTimer(USER_TIMER, 4);
while (!OSTimerExpired(USER_TIMER));
OSFreeTimer(USER_TIMER);
OSClearBreak();

// Push arguments onto the RPN stack: clean arguments up, then create a list.
while (GetArgType (top_estack) != END_TAG) {
top_estack = next_expression_index (top_estack);
}
top_estack--;
push_END_TAG();
push_longint(i);
push_longint(val);
push_LIST_TAG();

// Restore old system state.
PRG_setRate(orig_rate);
PRG_setStart(orig_start);
}

// addloop_memory_polling.c: don't optimize counting that much, through "volatile" which triggers three instructions instead of just one for dealing with memory and an address which gets constant-propagated instead of being kept in a register.

#define MIN_AMS 101
#define USE_TI89
#define USE_TI92P
#define USE_V200
#define USE_TI89T
#define NO_CALC_DETECT
#define OPTIMIZE_ROM_CALLS
#define RETURN_VALUE

#include <stdint.h>
#include <system.h>
#include <args.h>
#include <estack.h>
#include <peekpoke.h>
#include <intr.h>

#define TIMER_START_VAL (100000UL)

void _main(void) {
volatile uint32_t i = 0;
short orig_rate = PRG_getRate();
unsigned short orig_start = PRG_getStart();
volatile unsigned char * ON_key_status = (volatile unsigned char *)0x60001A;
unsigned long val = 0;

// Make the system timer an order of magnitude more precise;
// NOTE: this code assumes a HW2+ TI-68k, i.e. anything since 1999.
PRG_setRate(1); // Increment counter at a rate of 2^19/2^9 Hz
PRG_setStart(0xCE); // Trigger the interrupt every 257 - 0xCE = 51 increments ~ 20.07 Hz.

// The PRG_getStart() above effectively waited for the interrupt to trigger, so we don't need another wait.
/*OSRegisterTimer(USER_TIMER, 1);
while (!OSTimerExpired(USER_TIMER));
OSFreeTimer(USER_TIMER);*/
OSRegisterTimer(USER_TIMER, TIMER_START_VAL);

// Main loop :)
// Let compiler pessimizations inherent to "volatile", such as:
// * reading and writing i in memory instead of incrementing it directly;
// * constant-propagating the ON_key_status variable away.
// occur.
do {
i++;
} while (*ON_key_status & 2);

// Retrieve timer value.
val = TIMER_START_VAL - OSTimerCurVal(USER_TIMER);
OSFreeTimer(USER_TIMER);

// Give some time for the ON key to come back up.
OSRegisterTimer(USER_TIMER, 4);
while (!OSTimerExpired(USER_TIMER));
OSFreeTimer(USER_TIMER);
OSClearBreak();

// Push arguments onto the RPN stack: clean arguments up, then create a list.
while (GetArgType (top_estack) != END_TAG) {
top_estack = next_expression_index (top_estack);
}
top_estack--;
push_END_TAG();
push_longint(i);
push_longint(val);
push_LIST_TAG();

// Restore old system state.
PRG_setRate(orig_rate);
PRG_setStart(orig_start);
}
4)
#include <display_syscalls.h>
#include <keyboard_syscalls.h>
#include <keyboard.hpp>
#include <color.h>

// Getkey routine
const unsigned short* keyboard_register = (unsigned short*)0xA44B0000;
unsigned short lastkey[8];
unsigned short holdkey[8];

void keyupdate(void) {
memcpy(holdkey, lastkey, sizeof(unsigned short)*8);
memcpy(lastkey, keyboard_register, sizeof(unsigned short)*8);
}
int keydownlast(int basic_keycode) {
int row, col, word, bit;
row = basic_keycode%10;
col = basic_keycode/10-1;
word = row>>1;
bit = col + 8*(row&1);
return (0 != (lastkey[word] & 1<<bit));
}
int keydownhold(int basic_keycode) {
int row, col, word, bit;
row = basic_keycode%10;
col = basic_keycode/10-1;
word = row>>1;
bit = col + 8*(row&1);
return (0 != (holdkey[word] & 1<<bit));
}

int main() {
int i=0;
int key;
// clear screen
Bdisp_AllClr_VRAM();
while (1) {
keyupdate();
// increment i
i++;
if (keydownlast(KEY_PRGM_ACON)) {
char buffer[10];
strcpy(buffer," ");
itoa(i, buffer+2);
PrintXY(1,1,buffer,0,COLOR_BLACK);
Bdisp_PutDisp_DD();
}
GetKey(&key);
}
}

return 1;
}
6)
ASSEMBLE
NIBASC /HPHP48-R/
RPL

CODE
cbcnfgrtn	EQU	#99CD
cartn		EQU	#1C3D
uncnfgrtnl	EQU	#7C027
uncnfgrtnh      EQU     #805F9
cnfgrtnl	EQU	#99CF
cnfgrtnh        EQU     #805CF

*X1          EQU 1
*X4          EQU 1
*X8          EQU 1

GOSBVL  =SAVPTR
GOSBVL  =DisableIntr * Can't have an interrupt while IRAM is in the process of being reconfigured, it's
* not at address #80000 or if it's temporarily malformed due to the ISR code
GOSBVL  =DispOff     * Turn off the display

D0=(5)	=IRAMMASK    * Get size mask for IRAM and save it
A=DAT0	A
R4=A
GOSUB   end_isr      * Push address of ISR to the hardware return stack...
start_isr
ST=0    0
GONC    chktimer     * No carry detected -- leave ST.0 set to 0
ST=1    0            * Carry detected -- set ST.0 to 1
chktimer
D0=(5)  =TIMERCTRL.2
A=DAT0  1            * Check if TIMER2 requires service
?ABIT=1 3
GOYES   done         * If so we're done... ( branch back to cleanup code in IRAM )
?ST=1   0
GOYES   cry1         * Preserve carry of interrupting code
cry1    RTI                  * Return and re-enable interrupts
done    ?ST=0   0
GOYES   nocarry
C=C+1   P            * Adjust count as we've interrupted the add loop after C.A overflows
nocarry
R2=C                 * save count
CD1EX                * retrieve return address and save it to RSTK
RSTK=C
CD1EX
RTN                  * We're done -- branch back
end_isr

C=RSTK               * Grab address of ISR code
LA(5)   #80000       * Adjust address for post IRAM reconfiguration
C=C-A   A
D0=(5)  #8000F       * Address of ISR in IRAM before reconfiguration
A=DAT0  7            * Save contents of RAM
R0=A
LA(2)   #D8          * Write GOVLNG with address of ISR as target
DAT0=A  2
D0=D0+  2
DAT0=C  A
GOSUB   ++           * Grab return address of code to which to transfer control after IRAM reconfiguration
-       GOTO    +
++      C=RSTK               * Retrieve address of the previous instruction
LAHEX	#80000       *
RSTK=C               * Save return address of post reconfigure code to return stack

LC(5)   #80000       * Current address of IRAM
D=C     A            * save it...
C=R4                 * Size mask...
B=C     A            * Used by cbcnfgrtn
A=0     A            * New address is #00000

magic1                       * "magic" IRAM reconfiguration sequence...
LC(5)   cnfgrtnh
RSTK=C
LC(5)   cartn
RSTK=C
LC(5)   cbcnfgrtn
RSTK=C
LC(5)   uncnfgrtnl
RSTK=C
C=D     A
RTN
magic2
LC(5)   cnfgrtnl
RSTK=C
LC(5)   cartn
RSTK=C
LC(5)   cbcnfgrtn
RSTK=C
LC(5)   uncnfgrtnh
RSTK=C
C=D     A
RTN
+
*        D0=(5)  =TIMERCTRL.2 * Save TIMER2 control reg value
*        A=DAT0  S
*        R1=A.F  S
GOSUB   WaitForTick  * Wait for TIMER2 to decrement
P=      7            * Save TIMER2 value
A=DAT0  WP           *
R1=A.F  WP
D0=(5)  =TIMER2      * Set up 60 second TIMER2 countdown
A=0     W
P=      0
LA(5)   #78000
*        LA(5)   #2000        * 1 second countdown for debugging purposes...
P=      7
DAT0=A  WP
P=      0
D0=(5)  =TIMERCTRL.2 * Enable TIMER2 interrupts
LA(1)   3
DAT0=A  1

GOSUB   c1           * Control returns here if TIMER2 requires service and branches to cleanup
GOTO    cleanup
c1      C=RSTK
D1=C                 * Save post-ISR branch address in D1

C=0     W            * Set up regs for inner add loop
P=      5
GOSUB   +            * re-enable interrupts
GOTO    ++
+       RTI
++
*       GOSBVL   #0000F      * Debugging
** The following first inner add loop code was written by Werner **

IFDEF   X1
l1      C=C+1   A
GONC    l1
C=C+1   P
GONC    l1
ENDIF

IFDEF   X4           * Unrolled x4
l1      C=C+1   A
GOC     l2
C=C+1   A
GOC     l2
C=C+1   A
GOC     l2
C=C+1   A
GOC     l2
C=C+1   A
GONC    l1
l2      C=C+1   P
GONC    l1
ENDIF

IFDEF   X8           * Unrolled x8
l1      C=C+1   A
GOC     l2
C=C+1   A
GOC     l2
C=C+1   A
GOC     l2
C=C+1   A
GOC     l2
C=C+1   A
GOC     l2
C=C+1   A
GOC     l2
C=C+1   A
GOC     l2
C=C+1   A
GOC     l2
C=C+1   A
GONC    l1
l2      C=C+1   P
GONC    l1
ENDIF

cleanup                      * The ISR returns control here when TIMER2 underflows
P=      0
A=R3                 * Retrieve original return address
LC(5)   (end)-(-)    * Load offset of "end" label
C=C+A   A            * Add to original address to get address of end
RSTK=C               * Save end return address to return stack...
C=0     A            * Current IRAM address...
D=C     A            * Save it...
C=R4    A            * Size mask...
B=C     A            * Used by cbcnfgrtn
LA(5)   #80000       * New IRAM address...
GOTO    magic2       * Reconfigure IRAM back to its normal size and location...

end                          * Control returns to here after second IRAM reconfiguration...
A=R0                 * Retrive previous contents of IRAM at #8000F
D0=(5)  #8000F
DAT0=A  7            * Restore previous contents of IRAM
GOSUB   WaitForTick  * Wait for TIMER2 to decrement before restoring system time
P=      7            * Recall saved TIMER2 value
A=R1.F  WP
P=      0
LC(5)   #78000       * Adjust TIMER2 value to account for 60 second countdown
*        LC(5)   #2000        * 1 second countdown for debugging purposes
A=A-C   WP
DAT0=A  WP           * Restore correct TIMER2 value
P=      0
*        D0=(5)  =TIMERCTRL.2 * Restore TIMER2 control reg value
*        A=R1.F  S
*        DAT0=A  S
GOSBVL  =AllowIntr
GOSBVL  =DispOn      * Turn the display back on...
A=R2                 * Retrieve count saved in ISR
P=      15
GOVLNG  =PUSHhxsLoop * Push it to the stack and loop

* Utility subroutine that waits for TIMER2 to decrement
WaitForTick
D0=(5)  =TIMER2
P=      7
C=DAT0  WP
D=C     WP
-   C=DAT0  WP
?D=C    WP
GOYES   -
RTN
ENDCODE
9)

Forth + Assembler: ADD1 D.

Press any key (except ON) to stop counting.
Count => double length value on the stack.

First assemble ADD1 into the Forth dictionary:

FORTH
** overflows in about 470 secs .... P= xx where xx>6 gives more time
P= 6
C=0 A
A=0 WP
LOOP A=A+1 WP
C=IN         // Keyboard touched (except ON key !)
?C=0 A
GOYES LOOP
D1=D1- 5
DAT1=A A
ASR W
ASR W
ASR W
ASR W
ASR W
D1=D1- 5
DAT1=A A
RTNCC
END