RGB conversion routines - please test

http://www.a500.org/downloads/other/rgbconv_test.lha

Download and extract the above archive and run the rgbconv program. Do not do anything else on the computer while the test running as it will disturb the test, then post the results in this thread.

Please try it using different WB depths (i.e. 16/32-bit, not 8-bit though). Also make sure that there is free video memory before running the test (f.e. by closing all unnecessary windows).

So far it seems that the assembler routines are somewhat faster on the Sam440ep while the difference is much less noticeable on the ?A1.

deleted old results to make it easier on your eyes

@Slayer

The size of the screen shouldn't matter to the results.

These are results from my Sam440ep:

16-bit R5G6B5:
BltBitMap... 12 fps
8 s 137292 us
16-bit lookup table... 17 fps
6 s 33373 us
RGB565 optimised C-code (LUT version)... 19 fps
5 s 235484 us
RGB565 optimised C-code (macro version)... 18 fps
5 s 619877 us
RGB565 optimised ASM-code... 18 fps
5 s 618827 us
16-bit LUT ASM-code... 21 fps
4 s 807921 us

32-bit A8R8G8B8:
BltBitMap... 12 fps
8 s 34220 us
ARGB8888 optimised ASM-code... 13 fps
7 s 649436 us

BTW it appears that the 32-bit ASM routine is buggy (the checksum isn't the same as for BltBitMap). I'll upload a fixed version immediately.

I've uploaded a fixed version of the test now (the download URL is the same). I also removed some of the debug output because it was unnecessary and taking up space.

@salass00

understood also, new results

Radeon M9: 1680x1050 RGB16

BltBitMap... 12 fps
8 s 403757 us
Size: 768000 Checksum: B12566E8
16-bit lookup table... 22 fps
4 s 516908 us
Size: 768000 Checksum: B12566E8
RGB565 optimised C-code (LUT version)... 23 fps
4 s 294906 us
Size: 768000 Checksum: B12566E8
RGB565 optimised C-code (macro version)... 23 fps
4 s 346118 us
Size: 768000 Checksum: B12566E8
RGB565 optimised ASM-code... 23 fps
4 s 335930 us
Size: 768000 Checksum: B12566E8
000BB800 / 000BB800
16-bit LUT ASM-code... 23 fps
4 s 422492 us
Size: 768000 Checksum: B12566E8

Radeon M9: 1680x1050 ARGB32

BltBitMap... 12 fps
8 s 305781 us
Size: 768000 Checksum: 7E91CCA5
ARGB8888 optimised ASM-code... 13 fps
7 s 643724 us
Size: 768000 Checksum: 7E91CCA5
000FA000 / 000FA000

@salass00

Sam 16 bit 1024x768

BltBitMap... 12 fps
8 s 428361 us
Size: 768000 Checksum: 371356C7
16-bit lookup table... 20 fps
4 s 952083 us
Size: 768000 Checksum: 371356C7
RGB565 optimised C-code (LUT version)... 23 fps
4 s 307579 us
Size: 768000 Checksum: 371356C7
RGB565 optimised C-code (macro version)... 23 fps
4 s 358967 us
Size: 768000 Checksum: 371356C7
RGB565 optimised ASM-code... 23 fps
4 s 345578 us
Size: 768000 Checksum: 371356C7
000BB800 / 000BB800
16-bit LUT ASM-code... 21 fps
4 s 835104 us
Size: 768000 Checksum: 371356C7

Sam 32 bit 1024x768

BltBitMap... 12 fps
8 s 336794 us
Size: 768000 Checksum: 44127A5F
ARGB8888 optimised ASM-code... 13 fps
7 s 669291 us
Size: 768000 Checksum: 44127A5F
000FA000 / 000FA000

@Slayer & TiredOfLife

Now this is a little weird that for you the LUT versions are slower than the non-LUT ones. For me the LUT versions are much faster. Are you both also using the 667MHz SAM440ep?

Please specify the CPU as well as motherboard (?A1, sam440ep, A1-xe, ...) when reporting test results.

@salass00

Yip, 667mhz, a complete reset between shell/rgbconv execution

edit:

ran it again 16bit only

BltBitMap... 12 fps
8 s 396247 us
Size: 768000 Checksum: E18B2970
16-bit lookup table... 23 fps
4 s 320143 us
Size: 768000 Checksum: E18B2970
RGB565 optimised C-code (LUT version)... 23 fps
4 s 297795 us
Size: 768000 Checksum: E18B2970
RGB565 optimised C-code (macro version)... 23 fps
4 s 346315 us
Size: 768000 Checksum: E18B2970
RGB565 optimised ASM-code... 23 fps
4 s 334482 us
Size: 768000 Checksum: E18B2970
000BB800 / 000BB800
16-bit LUT ASM-code... 23 fps
4 s 278232 us
Size: 768000 Checksum: E18B2970

@salass00

Using Sam440EP 667

@salass00

1680x1050x32:

BltBitMap... 18 fps
5 s 506582 us
Size: 768000 Checksum: 87322626
ARGB8888 optimised ASM-code... 19 fps
5 s 308170 us
Size: 768000 Checksum: 87322626