@Zeda: Nice performance aware exercise! (Except for the many push and pop which look a bit dated to me).
I wonder how many are really interested in speeding up the calculations these days.
It seems all they care about is python, java and what-have-you-funky-high-level-language .

Thanks I wrote these with apps in mind, so I tried to reduce the need for external RAM. I should definitely make versions that take full advantage of SMC, though!

This is sounding quite cool!

Okay, thanks! There are 26 routines that I'll need to investigate later when I get out of work. Nine of them I don't know if I'll be able to contact the author, but one of those I plan to make a better implementation of anyways.

EDIT:
How does this sound?

1. This License does not apply to any file with a separate License header.
2. Permission is granted, free of charge, to use, modify, and/or distribute any part of this software for any purpose.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Written by Zeda Thomas <[email protected]>, Aug 2019

Here are some routines that I've added to the repository:

itoa_8
Converts an 8-bit signed integer to an ASCII string.
;Converts an 8-bit signed integer to a string

itoa_8:
;Input:
;   A is a signed integer
;   HL points to where the null-terminated ASCII string is stored (needs at most 5 bytes)
;Output:
;   The number is converted to a null-terminated string at HL
;Destroys:
;   Up to five bytes at HL
;   All registers preserved.
;on 0 to 9:       252       D=0
;on 10 to 99:     258+20D   D=0 to 9
;on 100 to 127:   277+20D   D=0 to 2
;on -1 to -9:     276       D=0
;on -10 to -99:   282+20D   D=0 to 9
;on -100 to -128: 301+20D   D=0 to 2

;min: 252cc  (+23cc over original)
;max: 462cc  (-49cc over original)
;avg: 343.74609375cc = 87999/256
;54 bytes
  push hl
  push de
  push bc
  push af
  or a
  jp p,itoa_pos
  neg
  ld (hl),$1A  ;start if neg char on TI-OS
  inc hl
itoa_pos:
;A is on [0,128]
;calculate 100s place, plus 1 for a future calculation
  ld b,'0'
  cp 100 \ jr c,$+5 \ sub 100 \ inc b

;calculate 10s place digit, +1 for future calculation
  ld de,$0A2F
  inc e \ sub d \ jr nc,$-2
  ld c,a

;Digits are now in D, C, A
; strip leading zeros!
  ld a,'0'
  cp b \ jr z,$+5 \ ld (hl),b \ inc hl \ .db $FE  ; start of `cp *` to skip the next byte, turns into `cp $BB` which will always return nz and nc
  cp e \ jr z,$+4 \ ld (hl),e \ inc hl
  add a,c
  add a,d
  ld (hl),a
  inc hl
  ld (hl),0

  pop af
  pop bc
  pop de
  pop hl
  ret

fixed88_to_string
Uses the itoa_8 routine to convert an 8.8 fixed-point number to a string.
;This converts a fixed-point number to a string.
;It displays up to 3 digits after the decimal.

fixed88_to_str:
;Inputs:
;   D.E is the fixed-point number
;   HL points to where the string gets output.
;      Needs at most 9 bytes.
;Outputs:
;   HL is preserved
;Destroys:
;   AF,DE,BC

;First check if the input is negative.
;If so, write a negative sign and negate
  push hl
  ld a,d
  or a
  jp p,+_
  ld (hl),$1A      ;negative sign on TI-OS
  inc hl
  xor a
  sub e
  ld e,a
  sbc a,a
  sub d
_:

;Our adjusted number is in A.E
;Now we can print the integer part
  call itoa_8

;Check if we need to print the fractional part
  xor a
  cp e
  jr z,fixed88_to_str_end

;We need to write the fractional part, so seek the end of the string
;Search for the null byte. A is already 0
  cpir

;Write a decimal
  dec hl
  ld (hl),'.'

  ld b,3
_:
;Multiply E by 10, converting overflow to an ASCII digit
  call fixed88_to_str_e_times_10
  inc hl
  ld (hl),a
  djnz -_

;Strip the ending zeros
  ld a,'0'
_:
  cp (hl)
  dec hl
  jr z,-_

;write a null byte
  inc hl
  inc hl
  ld (hl),0

fixed88_to_str_end:
;restore HL
  pop hl
  ret

fixed88_to_str_e_times_10:
  ld a,e
  ld d,0
  add a,a \ rl d
  add a,a \ rl d
  add a,e \ jr nc,$+3 \ inc d
  add a,a
  ld e,a
  ld a,d
  rla
  add a,'0'
  ret

sqrtA
This is a very fast, unrolled routine to compute the square root of A.

sqrtA:
;Input: A
;Output: D is the square root, A is the remainder (input-D^2)
;Destroys: BC
;speed: 161+{0,6}+{0,1}+{0,1}+{0,3}
;min: 161cc
;max: 172cc
;avg: 166.5cc
;45 bytes
  ld d,$40

  sub d
  jr nc,+_
  add a,d
  ld d,0
_:

  set 4,d
  sub d
  jr nc,+_
  add a,d
  .db $01   ;start of ld bc,** which is 10cc to skip the next two bytes.
_:
  set 5,d
  res 4,d
  srl d

  set 2,d
  sub d
  jr nc,+_
  add a,d
  .db $01   ;start of ld bc,** which is 10cc to skip the next two bytes.
_:
  set 3,d
  res 2,d
  srl d

  inc d
  sub d
  jr nc,+_
  add a,d
  dec d
_:
  inc d
  srl d
  ret

sqrtfixed_88
An unrolled, fast 8.8 fixed-point square root routine. Uses the above sqrtA routine.
sqrtfixed_88:
;Input: A.E ==> D.E
;Output: DE is the sqrt, AHL is the remainder
;Speed: 690+6{0,13}+{0,3+{0,18}}+{0,38}+sqrtA
;min: 855cc
;max: 1003cc
;avg: 924.5cc
;152 bytes

  call sqrtA
  ld l,a
  ld a,e
  ld h,0
  ld e,d
  ld d,h

  sla e
  rl d

  sll e \ rl d
  add a,a \ adc hl,hl
  add a,a \ adc hl,hl
  sbc hl,de
  jr nc,+_
  add hl,de
  dec e
  .db $FE     ;start of `cp *`
_:
  inc e

  sll e \ rl d
  add a,a \ adc hl,hl
  add a,a \ adc hl,hl
  sbc hl,de
  jr nc,+_
  add hl,de
  dec e
  .db $FE     ;start of `cp *`
_:
  inc e

  sll e \ rl d
  add a,a \ adc hl,hl
  add a,a \ adc hl,hl
  sbc hl,de
  jr nc,+_
  add hl,de
  dec e
  .db $FE     ;start of `cp *`
_:
  inc e

  sll e \ rl d
  add a,a \ adc hl,hl
  add a,a \ adc hl,hl
  sbc hl,de
  jr nc,+_
  add hl,de
  dec e
  .db $FE     ;start of `cp *`
_:
  inc e

;Now we have four more iterations
;The first two are no problem
  sll e \ rl d
  add hl,hl
  add hl,hl
  sbc hl,de
  jr nc,+_
  add hl,de
  dec e
  .db $FE     ;start of `cp *`
_:
  inc e

  sll e \ rl d
  add hl,hl
  add hl,hl
  sbc hl,de
  jr nc,+_
  add hl,de
  dec e
  .db $FE     ;start of `cp *`
_:
  inc e

sqrtfixed_88_iter11:
;On the next iteration, HL might temporarily overflow by 1 bit
  sll e \ rl d      ;sla e \ rl d \ inc e
  add hl,hl
  add hl,hl
  jr c,sqrtfixed_88_iter11_br0
;
  sbc hl,de
  jr nc,+_
  add hl,de
  dec e
  jr sqrtfixed_88_iter12
sqrtfixed_88_iter11_br0:
  or a
  sbc hl,de
_:
  inc e

;On the next iteration, HL is allowed to overflow, DE could overflow with our current routine, but it needs to be shifted right at the end, anyways
sqrtfixed_88_iter12:
  ld b,a      ;A is 0, so B is 0
  add hl,hl
  add hl,hl
  rla
;AHL - (DE+DE+1)
  sbc hl,de \ sbc a,b
  inc e
  or a
  sbc hl,de \ sbc a,b
  ret p
  add hl,de
  adc a,b
  dec e
  add hl,de
  adc a,b
  ret

ncr_HL_DE
Computes 'HL choose DE' in such a way so that overflow only occurs if the final result overflows 16 bits.
; Requires
;    mul16          ;BC*DE ==> DEHL
;    DEHL_Div_BC    ;DEHL/BC ==> DEHL

ncr_HL_DE:
;"n choose r", defined as n!/(r!(n-r)!)
;Computes "HL choose DE"
;Inputs: HL,DE
;Outputs:
;   HL is the result
;       "HL choose DE"
;   carry flag reset means overflow
;Destroys:
;   A,BC,DE,IX
;Notes:
;   Overflow is returned as 0
;   Overflow happens if HL choose DE exceeds 65535
;   This algorithm is constructed in such a way that intermediate
;   operations won't erroneously trigger overflow.
;66 bytes
  ld bc,1
  or a
  sbc hl,de
  jr c,ncr_oob
  jr z,ncr_exit
  sbc hl,de
  add hl,de
  jr c,$+3
  ex de,hl
  ld a,h
  or l
  push hl
  pop ix
ncr_exit:
  ld h,b
  ld l,c
  scf
  ret z
ncr_loop:
  push bc \ push de
  push hl \ push bc
  ld b,h
  ld c,l
  call mul16          ;BC*DE ==> DEHL
  pop bc
  call DEHL_Div_BC    ;result in DEHL
  ld a,d
  or e
  pop bc
  pop de
  jr nz,ncr_overflow
  add hl,bc
  jr c,ncr_overflow
  pop bc
  inc bc
  ld a,b
  cp ixh
  jr c,ncr_loop
  ld a,ixl
  cp c
  jr nc,ncr_loop
  ret
ncr_overflow:
  pop bc
  xor a
  ld b,a
ncr_oob:
  ld h,b
  ld l,b
  ret

EDIT: Optimized itoa_8 above. Here are some more routines:
uitoa_8
Converts an 8-bit unsigned integer to an ASCII string.
;Converts an 8-bit unsigned integer to a string

uitoa_8:
;Input:
;   A is a signed integer
;   HL points to where the null-terminated ASCII string is stored (needs at most 5 bytes)
;Output:
;   The number is converted to a null-terminated string at HL
;Destroys:
;   Up to four bytes at HL
;   All registers preserved.
;on 0 to 9:     238              D=0
;on 10 to 99:   244+20D          D=0 to 9
;on 100 to 255: 257+2{0,6}+20D   D=0 to 5
;min: 238cc
;max: 424cc
;avg: 317.453125cc = 81268/256 = (238*10 + 334*90+313*156)/256
;52 bytes

  push hl
  push de
  push bc
  push af
;A is on [0,255]
;calculate 100s place, plus 1 for a future calculation
  ld b,'0'
  cp 100 \ jr c,$+5 \ sub 100 \ inc b
  cp 100 \ jr c,$+5 \ sub 100 \ inc b

;calculate 10s place digit, +1 for future calculation
  ld de,$0A2F
  inc e \ sub d \ jr nc,$-2
  ld c,a

;Digits are now in D, C, A
; strip leading zeros!
  ld a,'0'
  cp b \ jr z,$+5 \ ld (hl),b \ inc hl \ .db $FE  ; start of `cp *` to skip the next byte, turns into `cp $BB` which will always return nz and nc
  cp e \ jr z,$+4 \ ld (hl),e \ inc hl
  add a,c
  add a,d
  ld (hl),a
  inc hl
  ld (hl),0

  pop af
  pop bc
  pop de
  pop hl
  ret

itoa_16
Converts a 16-bit signed integer to an ASCII string.
;Converts a 16-bit signed integer to an ASCII string.

itoa_16:
;Input:
;   DE is the number to convert
;   HL points to where to write the ASCII string (up to 7 bytes needed).
;Output:
;   HL points to the null-terminated ASCII string
;      NOTE: This isn't necessarily the same as the input HL.
  push de
  push bc
  push af
  push hl
  bit 7,d
  jr z,+_
  xor a
  sub e
  ld e,a
  sbc a,a
  sub d
  ld d,a
  ld (hl),$1A     ;negative char on TI-OS
  inc hl
_:
  ex de,hl

  ld bc,-10000
  ld a,'0'-1
  inc a \ add hl,bc \ jr c,$-2
  ld (de),a
  inc de

  ld bc,1000
  ld a,'9'+1
  dec a \ add hl,bc \ jr nc,$-2
  ld (de),a
  inc de

  ld bc,-100
  ld a,'0'-1
  inc a \ add hl,bc \ jr c,$-2
  ld (de),a
  inc de

  ld a,l
  ld h,'9'+1
  dec h \ add a,10 \ jr nc,$-3
  add a,'0'
  ex de,hl
  ld (hl),d
  inc hl
  ld (hl),a
  inc hl
  ld (hl),0

;No strip the leading zeros
  pop hl

;If the first char is a negative sign, skip it
  ld a,(hl)
  cp $1A
  push af
  ld a,'0'
  jr nz,$+3
  inc hl
  cp (hl)
  jr z,$-2

;Check if we need to re-write the negative sign
  pop af
  jr nz,+_
  dec hl
  ld (hl),a
_:

  pop af
  pop bc
  pop de
  ret

uitoa_16
Converts a 16-bit unsigned integer to an ASCII string.
;Converts a 16-bit unsigned integer to an ASCII string.

uitoa_16:
;Input:
;   DE is the number to convert
;   HL points to where to write the ASCII string (up to 6 bytes needed).
;Output:
;   HL points to the null-terminated ASCII string
;      NOTE: This isn't necessarily the same as the input HL.
  push de
  push bc
  push af
  ex de,hl

  ld bc,-10000
  ld a,'0'-1
  inc a \ add hl,bc \ jr c,$-2
  ld (de),a
  inc de

  ld bc,1000
  ld a,'9'+1
  dec a \ add hl,bc \ jr nc,$-2
  ld (de),a
  inc de

  ld bc,-100
  ld a,'0'-1
  inc a \ add hl,bc \ jr c,$-2
  ld (de),a
  inc de

  ld a,l
  ld h,'9'+1
  dec h \ add a,10 \ jr nc,$-3
  add a,'0'
  ex de,hl
  ld (hl),d
  inc hl
  ld (hl),a
  inc hl
  ld (hl),0

;No strip the leading zeros
  ld c,-6
  add hl,bc
  ld a,'0'
  inc hl \ cp (hl) \ jr z,$-2
  pop af
  pop bc
  pop de
  ret


On this one? The labels are in the right places, but I do notice that sometimes pressing a key will read as the wrong group

EDIT: Also, I'm hoping to put your routines in the repository if you'd like!

(P.S. This is what I work on now. Also, I tend to go by bcov77 on other platforms if you feel like googling.)

That is so freaking cool.

Oh wow, I hadn't realized that!
EDIT: I saw this on that page:

NOTE: The contents of this port should NOT be less than 0Ch or the LCD driver will no longer respond.

Hey there, it's ya gender non-specific diminutive Zeda, here, and today we'll be looking at the Fisher-Yates algorithm and just how freaking efficient it can be for shuffling a list. For reference, it takes one second to shuffle a 999-element list at 6MHz, and if that ain't the way your deity intended it, I don't know what is.

First, how do we shuffle L1 in BASIC?
rand(dim(L1->L2
SortA(L2,L1

This is a super clever algorithm, but slow as heck as the lists get bigger. Plus, it uses an extra list of the same size, wasting precious RAM. So how does the Fisher-Yates algorithm work? You start at the last element. Randomly choose an element up to and including the current element and swap them. Now move down one element and repeat (so now the last element is off limits, then the last two, et cetera). Repeat this until there is one element left.

This is easy to perform in-place, and it performs n-1 swaps, making it significantly faster than the BASIC algorithm above. In fact, let's implement it in BASIC:
dim(L1->N
For(K,N,2,-1
randInt(1,K->A
L1(K->B
L1(A->L1(K
B->L1(A
End
This takes approximately 37.5 seconds to sort a 999 element list. I don't even have the RAM needed to test the regular method, but extrapolating, it would take the "normal" method approximately 73 seconds for 999 elements. So basically, the Fisher-Yates algorithm is actually faster even in TI-BASIC (after about 400 elements, though).

---

So without further ado, the assembly code!
;Randomizes a TI-list in Ans

_RclAns= 4AD7h
seed1  = $80F8
seed2  = $80FC

seed1_0=seed1
seed1_1=seed1+2
seed2_0=seed2
seed2_1=seed2+2
#define bcall(x) rst 28h \ .dw x


.db $BB,$6D
.org $9D95

; Put it into 15MHz mode if possible!
  in a,(2)
  add a,a
  sbc a,a
  out (20h),a


; Initialize the random seed
  ld hl,seed1
  ld b,7
  ld a,r
_:
  xor (hl)
  ld (hl),a
  inc hl
  djnz -_
  or 99
  or (hl)
  ld (hl),a
 

; Locate Ans, verify that it is a list or complex list
  bcall(_RclAns)
  ex de,hl
  ld c,(hl)
  inc hl
  ld b,(hl)
  inc hl
  ld (list_base),hl
  dec a
  jr z,+_
  sub 12
  ret nz
  dec a
_:

;A is 0 if a real list, -1 if complex
;HL points to the first element
;BC is the number of elements
  and $29     ;make it either NOP or ADD HL,HL
  ld (get_complex_element),a
  sub 29h
  sbc a,a
;FF if real, 00 if complex
  cpl
  and 9
  add a,9
  ld (element_size),a


shuffle_loop:
  push bc

  push bc
  call rand
  pop bc
  ex de,hl
  call mul16
  dec bc
  ;swap elements DE and BC
 
  call get_element
  push hl
  ld d,b
  ld e,c
  call get_element
  pop de

  call swap_elements
 
  pop bc
  dec bc
  ld a,c
  dec a
  jr nz,shuffle_loop
  inc b
  dec b
  jr nz,shuffle_loop
  ret


swap_elements:
;HL and DE point to the elements
element_size = $+2
  ld bc,255
_:
  ld a,(de)
  ldi
  dec hl
  ld (hl),a
  inc hl
  djnz -_
  ret


get_element:
;Input:
;   DE is the element to locate
;Output:
;   HL points to the element
  ld l,e
  ld h,d
  add hl,hl
  add hl,hl
  add hl,hl
  add hl,de
get_complex_element:
  nop
list_base = $+1
  ld de,0
  add hl,de
  ret


rand:
;Tested and passes all CAcert tests
;Uses a very simple 32-bit LCG and 32-bit LFSR
;it has a period of 18,446,744,069,414,584,320
;roughly 18.4 quintillion.
;LFSR taps: 0,2,6,7  = 11000101
;291cc
;Thanks to Runer112 for his help on optimizing the LCG and suggesting to try the much simpler LCG. On their own, the two are terrible, but together they are great.
    ld hl,(seed1)
    ld de,(seed1+2)
    ld b,h
    ld c,l
    add hl,hl \ rl e \ rl d
    add hl,hl \ rl e \ rl d
    inc l
    add hl,bc
    ld (seed1_0),hl
    ld hl,(seed1_1)
    adc hl,de
    ld (seed1_1),hl
    ex de,hl
;;lfsr
    ld hl,(seed2)
    ld bc,(seed2+2)
    add hl,hl \ rl c \ rl b
    ld (seed2_1),bc
    sbc a,a
    and %11000101
    xor l
    ld l,a
    ld (seed2_0),hl
    ex de,hl
    add hl,bc
    ret


mul16:
;BC*DE
  ld hl,0
  ld a,16
mul16_loop:
  add hl,hl
  rl e
  rl d
  jr nc,+_
  add hl,bc
  jr nc,+_
  inc de
_:
  dec a
  jr nz,mul16_loop
  ret
It isn't perfect, but it is pretty good and importantly, it is fast! The biggest problem is in the random number generator, but even that is still pretty good for this application.