Anyone up for an ASM coding competition? - Page 3

demolition · 01 May 2016, 07:51

Quote:

Originally Posted by alpine9000

We could start with something far less ambitious while we see if the concept of a coding competition here works.

Probably a good idea. If it was too complicated, you might also only get two entries which would not be much fun. It should be simple enough so that people with mediocre skills could join in and make something (perhaps not win, but at least make a pitch).

alkis · 01 May 2016, 12:46

N-balls colliding on screen
(http://vobarian.com/collisions/2dcollisions2.pdf)

meynaf · 02 May 2016, 08:19

Quote:

Originally Posted by demolition

Probably a good idea. If it was too complicated, you might also only get two entries which would not be much fun. It should be simple enough so that people with mediocre skills could join in and make something (perhaps not win, but at least make a pitch).

I second that.

alpine9000 · 03 May 2016, 02:34

We seem to be going around in circles a bit.

Maybe if we just nominate someone to run the first round, they can come up with a goal and judging rules.

Then people can enter the round if they find it interesting.

If we get enough entries, the competition goes ahead.

britelite · 03 May 2016, 07:45

Some good hints on how to run such a compo:

http://amycoders.org/compo/index.php

gilgamesh · 03 May 2016, 10:39

Nice. I'll be the cheerleading squad.

Thorham · 11 May 2016, 22:57

So... what are we going to write, then?

Thorham · 12 May 2016, 14:11

Guys?

meynaf · 12 May 2016, 15:05

Appears they are sleeping

Who runs the compo and who participates ?

Thorham · 12 May 2016, 22:21

Quote:

Originally Posted by meynaf

Appears they are sleeping

Who runs the compo and who participates ?

Apparently no one cares enough, or everyone is waiting for someone else to take charge (myself included)

Leffmann · 12 May 2016, 23:08

DanScott is arranging it I guess, but he's busy with his game too.

I have a small one for you to warm up with:

You're given two unsigned 32-bit integers in D0 and D1. Write a subroutine that determines if D0 is greater or equal to 2/3 of D1.

F.ex. if D1 is 14, then 2/3 of that is 9.333... and so 9 is not enough, and D1 must be 10 or greater.

D0 and D1 can take on all possible 32-bit values. No pre-computed tables are allowed, and it must be plain 68000 code. Other than that you can can return your result in any form and any machine register you want.

The shortest subroutine wins.

DanScott · 12 May 2016, 23:32

I'm alive

Been busy with a gazillion things...

Let's go with Leffmans "warm up" challenge... I'll post my result tomorrow evening

meynaf · 13 May 2016, 08:18

Quote:

Originally Posted by Leffmann

The shortest subroutine wins.

In code size or number of instructions ?

meynaf · 13 May 2016, 09:04

Ok, first version here anyway :

Code:

 move.l d0,d2
 lsr.l #1,d2
 add.l d2,d0
 bcs.s .done
 cmp.l d0,d1
.done

Final result like CMP (CCR bits C and Z are set).

Instead of comparing D0 vs 2/3 *D1, we could do 3*D0 vs 2*D1 but it's the same as 1.5*D0 vs D1.
If 1.5*D0 doesn't fit in 32 bits then it can't be <D1 (hence the bcs).
Carry generated by lsr doesn't matter (cmp 10 with 9 is same as cmp 10 with 9.5 ; cmp 10 with 10.5 is same as 10 with 10 because it's greater or equal we want to know).

Leffmann · 13 May 2016, 11:54

OK that's a very short solution, but maybe I was a bit too lax on the restrictions when I said "any form and any machine register", because in this form you can't use the code with any branch at all. You have to read the CCR and look at the bits individually, which also means that on M68000 you have to go into supervisor mode to actually make use of that code, and on M68020 and later we don't have MOVE from SR. The idea is that the code should run on all machines.

So let's agree that you either have to set the condition codes correctly so that you can use BHS/BCC, or return a non-zero result in any data register to indicate that the condition holds true. Sorry about that.

Though hold off posting any code for a while, other people may want to have a go, and posting a solution right underneath the problem takes away the whole element of competition.

roomeo · 13 May 2016, 12:13

@Leffman

We should set a deadline, and pm the answers to a dedicated person.
Great little exercise btw...

meynaf · 13 May 2016, 12:42

Quote:

Originally Posted by Leffmann

OK that's a very short solution, but maybe I was a bit too lax on the restrictions when I said "any form and any machine register", because in this form you can't use the code with any branch at all. You have to read the CCR and look at the bits individually, which also means that on M68000 you have to go into supervisor mode to actually make use of that code, and on M68020 and later we don't have MOVE from SR. The idea is that the code should run on all machines.

So let's agree that you either have to set the condition codes correctly so that you can use BHS/BCC, or return a non-zero result in any data register to indicate that the condition holds true. Sorry about that.

There is absolutely no need to test the CCR bits individually and this code runs on all machines. Just use the BLS/BHI branch condition (or SHS/SHI to set a register if you prefer).

Quote:

Originally Posted by Leffmann

Though hold off posting any code for a while, other people may want to have a go, and posting a solution right underneath the problem takes away the whole element of competition.

Yeah but if you post too late then someone else may post the same solution before you - at least on very simple stuff like this.
PM'ing the answers before a deadline can indeed be the solution to this.

alpine9000 · 13 May 2016, 12:59

I like the idea of PMing answers.

Now I have seen meynaf's solution I can't unsee it, and i can guarantee I won't be coming up with anything better

DanScott · 13 May 2016, 13:01

bah!

I wasn't thinking "out the box" enough... won't even post my feeble attempt

ah go on then.. but it's not short

not tested.. and probably wrong

but I used the fact that 2/3 is approx: (1/2) + (1/8) + (1/32) + (1/128) + (1/512)

Code:

move.l	d1,d2
lsr.l	#1,d2
move.l	d2,d3
lsr.l	#2,d3
move.l	d3,d4
lsr.l	#2,d4
move.l	d4,d5
lsr.l	#2,d5
move.l	d5,d1
lsr.l	#2,d1
add.l	d2,d1
add.l	d3,d1
add.l	d4,d1
add.l	d5,d1
moveq	#0,d2
cmp.l	d0,d1
shi	d2
rts

meynaf · 13 May 2016, 13:45

There is a way to do it shorter (cmp'ing d0*3 with d1*2), but it won't work for very large values :

Code:

 sub.l d0,d1	; b-a
 add.l d1,d1	; 2b-2a
 sub.l d0,d1	; 2b-3a

(sorry i gave code again, but now it's too late for PM'ing anything, right ?)

13 May 2016, 09:04	#54
meynaf son of 68k Join Date: Nov 2007 Location: Lyon / France Age: 51 Posts: 5,323	Ok, first version here anyway : Code: move.l d0,d2 lsr.l #1,d2 add.l d2,d0 bcs.s .done cmp.l d0,d1 .done Final result like CMP (CCR bits C and Z are set). Instead of comparing D0 vs 2/3 D1, we could do 3D0 vs 2D1 but it's the same as 1.5D0 vs D1. If 1.5*D0 doesn't fit in 32 bits then it can't be <D1 (hence the bcs). Carry generated by lsr doesn't matter (cmp 10 with 9 is same as cmp 10 with 9.5 ; cmp 10 with 10.5 is same as 10 with 10 because it's greater or equal we want to know).

13 May 2016, 13:01	#59
DanScott Lemon. / Core Design Join Date: Mar 2016 Location: Tier 5 Posts: 1,212	bah! I wasn't thinking "out the box" enough... won't even post my feeble attempt ah go on then.. but it's not short not tested.. and probably wrong but I used the fact that 2/3 is approx: (1/2) + (1/8) + (1/32) + (1/128) + (1/512) Code: move.l d1,d2 lsr.l #1,d2 move.l d2,d3 lsr.l #2,d3 move.l d3,d4 lsr.l #2,d4 move.l d4,d5 lsr.l #2,d5 move.l d5,d1 lsr.l #2,d1 add.l d2,d1 add.l d3,d1 add.l d4,d1 add.l d5,d1 moveq #0,d2 cmp.l d0,d1 shi d2 rts

13 May 2016, 13:45	#60
meynaf son of 68k Join Date: Nov 2007 Location: Lyon / France Age: 51 Posts: 5,323	There is a way to do it shorter (cmp'ing d03 with d12), but it won't work for very large values : Code: sub.l d0,d1 ; b-a add.l d1,d1 ; 2b-2a sub.l d0,d1 ; 2b-3a (sorry i gave code again, but now it's too late for PM'ing anything, right ?)

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Starting ASM coding on A1200. Which Assembler?	Nosferax	Coders. Asm / Hardware	68	27 November 2015 16:14
4th tutorial on ASM- and HW-coding	Vikke	Coders. Asm / Hardware	11	10 April 2013 20:32
3rd tutorial on ASM- and HW-coding	Vikke	Coders. Asm / Hardware	6	26 March 2013 15:57
First tutorial on ASM- and HW-coding	Vikke	Coders. Asm / Hardware	46	18 March 2013 12:33
2nd tutorial on ASM- and HW-coding	Vikke	Coders. Asm / Hardware	10	17 March 2013 11:49

01 May 2016, 12:46	#42
alkis Registered User Join Date: Dec 2010 Location: Athens/Greece Age: 53 Posts: 719	N-balls colliding on screen (http://vobarian.com/collisions/2dcollisions2.pdf)

03 May 2016, 02:34	#44
alpine9000 Registered User Join Date: Mar 2016 Location: Australia Posts: 881	We seem to be going around in circles a bit. Maybe if we just nominate someone to run the first round, they can come up with a goal and judging rules. Then people can enter the round if they find it interesting. If we get enough entries, the competition goes ahead.

03 May 2016, 07:45	#45
britelite Registered User Join Date: Feb 2010 Location: Espoo / Finland Posts: 818	Some good hints on how to run such a compo: http://amycoders.org/compo/index.php

03 May 2016, 10:39	#46
gilgamesh Linux snob Join Date: Sep 2008 Location: Monkey Island Posts: 997	Nice. I'll be the cheerleading squad.

11 May 2016, 22:57	#47
Thorham Computer Nerd Join Date: Sep 2007 Location: Rotterdam/Netherlands Age: 47 Posts: 3,753	So... what are we going to write, then?

12 May 2016, 14:11	#48
Thorham Computer Nerd Join Date: Sep 2007 Location: Rotterdam/Netherlands Age: 47 Posts: 3,753	Guys?

12 May 2016, 15:05	#49
meynaf son of 68k Join Date: Nov 2007 Location: Lyon / France Age: 51 Posts: 5,323	Appears they are sleeping Who runs the compo and who participates ?

12 May 2016, 23:08	#51
Leffmann Join Date: Jul 2008 Location: Sweden Posts: 2,269	DanScott is arranging it I guess, but he's busy with his game too. I have a small one for you to warm up with: You're given two unsigned 32-bit integers in D0 and D1. Write a subroutine that determines if D0 is greater or equal to 2/3 of D1. F.ex. if D1 is 14, then 2/3 of that is 9.333... and so 9 is not enough, and D1 must be 10 or greater. D0 and D1 can take on all possible 32-bit values. No pre-computed tables are allowed, and it must be plain 68000 code. Other than that you can can return your result in any form and any machine register you want. The shortest subroutine wins.

12 May 2016, 23:32	#52
DanScott Lemon. / Core Design Join Date: Mar 2016 Location: Tier 5 Posts: 1,212	I'm alive Been busy with a gazillion things... Let's go with Leffmans "warm up" challenge... I'll post my result tomorrow evening

13 May 2016, 11:54	#55
Leffmann Join Date: Jul 2008 Location: Sweden Posts: 2,269	OK that's a very short solution, but maybe I was a bit too lax on the restrictions when I said "any form and any machine register", because in this form you can't use the code with any branch at all. You have to read the CCR and look at the bits individually, which also means that on M68000 you have to go into supervisor mode to actually make use of that code, and on M68020 and later we don't have MOVE from SR. The idea is that the code should run on all machines. So let's agree that you either have to set the condition codes correctly so that you can use BHS/BCC, or return a non-zero result in any data register to indicate that the condition holds true. Sorry about that. Though hold off posting any code for a while, other people may want to have a go, and posting a solution right underneath the problem takes away the whole element of competition.

13 May 2016, 12:13	#56
roomeo Registered User Join Date: Aug 2014 Location: Telemark Posts: 207	@Leffman We should set a deadline, and pm the answers to a dedicated person. Great little exercise btw...

13 May 2016, 12:59	#58
alpine9000 Registered User Join Date: Mar 2016 Location: Australia Posts: 881	I like the idea of PMing answers. Now I have seen meynaf's solution I can't unsee it, and i can guarantee I won't be coming up with anything better

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)