English Amiga Board


Go Back   English Amiga Board > Search Forums

Showing results 1 to 40 of 918
Search took 0.37 seconds.
Search: Posts Made By: Don_Adan
Forum: Coders. General 10 November 2017, 03:42
Replies: 158
Views: 8,528
Posted By Don_Adan
Im not expert, but for me this is doable for...

Im not expert, but for me this is doable for trackloading Amiga 500 game. The best examples for Amiga are Random Access games/conversions, like Swiv, Ninja Warriors and Saint Dragon. But dont use AHX...
Forum: request.Apps 07 November 2017, 03:57
Replies: 22
Views: 687
Posted By Don_Adan
No, some/last Amtari version works with 68030 and...

No, some/last Amtari version works with 68030 and MMU (for catching accesses to Atari ST registers). If i remember right you have 68030 with MMU.
.
info taken from the net.

ST emulators...
Forum: request.Apps 06 November 2017, 22:52
Replies: 22
Views: 687
Posted By Don_Adan
If on Amiga, try to use Amtari emulator, and run...

If on Amiga, try to use Amtari emulator, and run msa archiver
https://sites.google.com/site/stessential/disks-tools/msa
later xad.library on Amiga or msa converter for Amiga or PC.
Forum: Coders. General 07 September 2017, 19:19
Replies: 220
Views: 12,198
Posted By Don_Adan
Really? ...

Really?

http://website.informer.com/Gunnar+Von+Boehn.html

http://aminet.net/package/util/moni/sortbench

http://www.apollo-core.com/index.htm?page=contact
Forum: Coders. General 04 September 2017, 23:30
Replies: 220
Views: 12,198
Posted By Don_Adan
I think that better will be creating new...

I think that better will be creating new (fastest) version of transcendential functions as microcode, as optimised code for 68080. It can be good challenge for meynaf and maybe other coders. Meynaf...
Forum: Amiga scene 02 July 2017, 23:19
Replies: 238
Views: 12,679
Posted By Don_Adan
Vampire 2 FPU secret solved for me. 40k LE only. ...

Vampire 2 FPU secret solved for me. 40k LE only. No space for full 68882. Even no space for 68040/68060 FPU style. Too many extra modules for standalone version. Only very easy FPU and incompatible...
Forum: support.Hardware 01 June 2017, 18:44
Replies: 271
Views: 10,400
Posted By Don_Adan
Seems that many FPU is missing: Instruction ...

Seems that many FPU is missing:

Instruction Function
====================================================

FACOS | FP Arc Cosine
FASIN | FP Arc Sine
FATAN | FP Arc...
Forum: support.Hardware 01 June 2017, 10:30
Replies: 271
Views: 10,400
Posted By Don_Adan
Seems that Gunnar resurrected FPU thread, then...

Seems that Gunnar resurrected FPU thread, then someone can write FPU test code for Apollo core:
http://www.apollo-core.com/knowledge.php?b=4&note=5931&z=hgzcfI
But I still suspect that only very...
Forum: support.Hardware 31 May 2017, 18:24
Replies: 271
Views: 10,400
Posted By Don_Adan
Seems you dont read exactly Gunnar posts from...

Seems you dont read exactly Gunnar posts from linked by me threads. All are about new version of Apollo Core (for Cyclone III or Cyclone V), not about Vampire V1. Even with enough space Gunnar DONT...
Forum: support.Hardware 31 May 2017, 16:42
Replies: 271
Views: 10,400
Posted By Don_Adan
Seems you are really very new on EAB. You wrote...

Seems you are really very new on EAB. You wrote too many nonsenses about "no space in FPGA" for rare CPU instructions, rare addressing modes etc. Space never was problem, when target was Cyclone III...
Forum: support.Hardware 26 May 2017, 02:18
Replies: 271
Views: 10,400
Posted By Don_Adan
Yes, Gunnar is strange. He wrote that Apollo has...

Yes, Gunnar is strange. He wrote that Apollo has "fully pipelined, double/extended FPU" and now he want to use SoftFPU. I can ask where is this pipeline for SoftFPU? It will be slowdown CPU only. No...
Forum: support.Hardware 25 May 2017, 05:35
Replies: 271
Views: 10,400
Posted By Don_Adan
Right, this is very stupid idea. For all existed...

Right, this is very stupid idea. For all existed 68k programs (for all 68k platforms) Apollo Core will be equivalent of very fast 68EC040 CPU only. Seems that 68EC080 is correct name for current...
Forum: Coders. Asm / Hardware 07 March 2017, 20:05
Replies: 523
Views: 44,675
Posted By Don_Adan
One of the best known me 68k is Arj mode 7...

One of the best known me 68k is Arj mode 7 (better than LZX). Better is only hitchhikr's PackFire compressor, but has slow decompression, due mulu.http://www.pouet.net/prod.php?which=54840...
Forum: Amiga scene 10 February 2017, 11:19
Replies: 51
Views: 7,487
Posted By Don_Adan
CreatePalette lea Palette,a0 ...

CreatePalette lea Palette,a0
move.w #256,(a0)+ ;Number of colours to load
clr.w (a0)+ ;Register number to start from
moveq #-1,D6
move.l #$01010101,D7...
Forum: Amiga scene 08 January 2017, 23:28
Replies: 323
Views: 38,958
Posted By Don_Adan
What do you mean about this? ...

What do you mean about this?

http://aminet.net/package/gfx/misc/LDP-1_31
Forum: Coders. Asm / Hardware 19 December 2016, 05:12
Replies: 3
Views: 1,802
Posted By Don_Adan
I will use something like this: ...

I will use something like this:


vertical_offset = $b0
move.l colour(pc),a2 ; Pointer to colour table
movem.l (a2)+,D0-D6
add.w (a5)+,a1 ;Extract sine


move.w D0,(vertical_offset*1)(a1)
Forum: request.Modules 06 September 2016, 20:48
Replies: 4
Views: 2,267
Posted By Don_Adan
Module from Aladdin: ...

Module from Aladdin:
http://wt.exotica.org.uk/rips.html
Maybe from Setters too, but I dont remember.
BTW. Aladdin has subsongs.
Forum: Coders. Asm / Hardware 29 June 2016, 00:32
Replies: 1
Views: 719
Posted By Don_Adan
Checked on 68040. 1. long data are copied, A0...

Checked on 68040.
1. long data are copied, A0 is +8.
2. long data are copied, A0 is -8.
3. long data from 8(a7) are copied, A7 is -4.
4. long data from (a7)+ are copied, A7 is + 4.
5. A7 is...
Forum: Coders. Asm / Hardware 28 June 2016, 04:29
Replies: 523
Views: 44,675
Posted By Don_Adan
Do you mean subq/bne, I think. SILVER7]...

Do you mean subq/bne, I think.

SILVER7] (26.06.2016)
* Added CHK2, CAS2, CMP2, EXTW/L instructions for MacOS (thanks JimDrew !)
* Added FAST-IDE for V500
* Added HDMI bootpic (credits to flype...
Forum: Coders. Asm / Hardware 27 June 2016, 18:42
Replies: 523
Views: 44,675
Posted By Don_Adan
Seems that most bitfields instruction is 1 cycle...

Seems that most bitfields instruction is 1 cycle only for Apollo core, then no problem to beat 68060 performance for TLSFMem, even meynaf's picture viewer will be super fast now.

"9) Many...
Forum: Coders. Asm / Hardware 26 June 2016, 23:34
Replies: 523
Views: 44,675
Posted By Don_Adan
No, Matt. I mean about using normal moveq...

No, Matt. I mean about using normal moveq instruction as 64 bit instruction too. Simple moveq #1,D0, will be 64 bits shortest than move.q #$0000000000000001,D0 instruction. And move.q will be wasted...
Forum: Coders. Asm / Hardware 26 June 2016, 23:24
Replies: 523
Views: 44,675
Posted By Don_Adan
Oki, then other opcode must be used for swap.l.

Oki, then other opcode must be used for swap.l.
Forum: Coders. Asm / Hardware 26 June 2016, 20:23
Replies: 523
Views: 44,675
Posted By Don_Adan
You can use new "move.q" (move quad) command, and...

You can use new "move.q" (move quad) command, and wrote full 64 data register to buffer.
Personally I will prefer to add swap.l D0 to D7 instruction to Apollo core. This instruction can used swapa...
Forum: Coders. Asm / Hardware 26 June 2016, 20:05
Replies: 523
Views: 44,675
Posted By Don_Adan
Im think that Phil go sleep.

Im think that Phil go sleep.
Forum: Coders. Asm / Hardware 26 June 2016, 14:28
Replies: 523
Views: 44,675
Posted By Don_Adan
Apollo has 64 bit data registers. No difference...

Apollo has 64 bit data registers. No difference for all 32 bit Amiga code. AmigaOS dont must have support for 64bit. F.e all copy operations can be 2x fastest than for 32 bit. For limited in MHz CPU...
Forum: Coders. Asm / Hardware 26 June 2016, 11:51
Replies: 523
Views: 44,675
Posted By Don_Adan
No, Phil. Everyone waiting for your new challenge...

No, Phil. Everyone waiting for your new challenge :)
Forum: Coders. Asm / Hardware 26 June 2016, 11:50
Replies: 523
Views: 44,675
Posted By Don_Adan
Could you check if moveq works for Apollo core as...

Could you check if moveq works for Apollo core as 32 or 64 bit instruction? For me it must be 64 bit instruction for Apollo, due 32 bit version will be wasted cache only.
Forum: Coders. Asm / Hardware 26 June 2016, 05:22
Replies: 523
Views: 44,675
Posted By Don_Adan
I never used cmp2, chk2 and tas2, but movep is...

I never used cmp2, chk2 and tas2, but movep is normal and useful instruction for me. I prefer good ISA than new ISA. Trapping instructions has no sense, due Apollo seems to be finally designed as...
Forum: Coders. Asm / Hardware 26 June 2016, 00:38
Replies: 523
Views: 44,675
Posted By Don_Adan
No, due you want to sacrifice movep instruction....

No, due you want to sacrifice movep instruction. Apollo version of movep works in 1 cycle only.
Do you remember this old thread (discusion started from 44 post):
...
Forum: Coders. Asm / Hardware 25 June 2016, 16:40
Replies: 523
Views: 44,675
Posted By Don_Adan
Thats good. You are very good tester. Maybe only...

Thats good. You are very good tester. Maybe only meynaf can be better.
Forum: Coders. Asm / Hardware 25 June 2016, 11:35
Replies: 523
Views: 44,675
Posted By Don_Adan
I dont know Matt's definition of productivity...

I dont know Matt's definition of productivity software.



But Gunnar changed yours mind, this is Oki for me. I still remember how many 68040 CPU instructions was supposed to be missing in Apolo...
Forum: Coders. Asm / Hardware 25 June 2016, 01:30
Replies: 523
Views: 44,675
Posted By Don_Adan
No, due loop value is not static, can be 3, 15,...

No, due loop value is not static, can be 3, 15, 20 etc. 32 is max. Too many combinations for support. One unrolled version with special handling for odd loop values, is enough for me.



You...
Forum: Coders. Asm / Hardware 25 June 2016, 00:14
Replies: 523
Views: 44,675
Posted By Don_Adan
Phil, if i remember right, adda.w is 2 cycles...

Phil, if i remember right, adda.w is 2 cycles only for 68030, then both versions are fastest than muls.l version.


move.w (a3)+,d7 ; 5
asr.l #1,d5 ; 4 (9)
subx.w D2,D2 ...
Forum: Coders. Asm / Hardware 25 June 2016, 00:11
Replies: 523
Views: 44,675
Posted By Don_Adan
This is very short code then must fit in 68020...

This is very short code then must fit in 68020 and 68030 cache. Next advantage is that this code works for 68000 too, if someone need fast 17x16 multiple routine for 68000. And remember meynaf is...
Forum: Coders. Asm / Hardware 24 June 2016, 15:30
Replies: 523
Views: 44,675
Posted By Don_Adan
Oki, if data at A3 can be swapped, then we can...

Oki, if data at A3 can be swapped, then we can use swap command and unroll this routine, it will be finally fastest than muls.l version, i think. Of course D7 must be changed, and normal version for...
Forum: Coders. Asm / Hardware 24 June 2016, 07:22
Replies: 523
Views: 44,675
Posted By Don_Adan
Or this version, if fastest: .loopl1 ...

Or this version, if fastest:


.loopl1
move.l a4,a1
move.l a2,a3
move.w d7,d6
moveq #0,d1
sub.l a5,a5 ; sum of coefficients (using Ax for auto-ext)
.loopl2
Forum: Coders. Asm / Hardware 24 June 2016, 07:11
Replies: 523
Views: 44,675
Posted By Don_Adan
Very good attempt, I will only remove bcc from...

Very good attempt, I will only remove bcc from your code :


.loopl1
move.l a4,a1
move.l a2,a3
move.w d7,d6
moveq #0,d1
sub.l a5,a5 ; sum of coefficients (using Ax for auto-ext)...
Forum: Coders. Asm / Hardware 23 June 2016, 20:06
Replies: 523
Views: 44,675
Posted By Don_Adan
Seems to be easy enough to do, muls.l are rare...

Seems to be easy enough to do, muls.l are rare enough, then you will be slowdown muls.w version by 8 cycles only. It can looks next:


.loopl1
move.l a4,a1
move.l a2,a3
move.w d7,d6
moveq...
Forum: Coders. Asm / Hardware 23 June 2016, 11:23
Replies: 523
Views: 44,675
Posted By Don_Adan
If final value is from $FFE00000 to $001FFFE0,...

If final value is from $FFE00000 to $001FFFE0, then shifts higher than 24 are useless.
And if (a4) contains very small values, muls.w is enough computation.
Forum: Coders. Asm / Hardware 23 June 2016, 00:39
Replies: 523
Views: 44,675
Posted By Don_Adan
Due low shifts are useful only for small final...

Due low shifts are useful only for small final values, i think that maximum can be $20000. Low shifts for big values like $56452377 is useless, of course if i understand correctly your code.
Showing results 1 to 40 of 918

 
Forum Jump

All times are GMT +2. The time now is 05:52.


Powered by vBulletin® Version 3.8.8 Beta 1
Copyright ©2000 - 2017, vBulletin Solutions, Inc.
Page generated in 0.51064 seconds with 9 queries