@Karlos

Quote:

Bugger I dunno. Seems to work everywhere else - from 040+CGX to 040+P96 to 603, 750, G4 and various UAE.

Don't worry, I traced the problem to the graphics driver Amithlon was using. Your code runs fine on my XP 1700+ Amithlon box, which uses a different driver.

Here's the results from the 1700+, JIT first then non-JIT :

Read RAM : 1188000.00 K/sec
Write RAM : 1196400.00 K/sec
RAM->RAM : 400000.00 K/sec
RAM->RAM (OS) : 400000.00 K/sec
RAM->RAM[16] : 400000.00 K/sec
Read VRAM : 4428.04 K/sec
Write VRAM : 240000.00 K/sec
RAM->VRAM : 240000.00 K/sec
RAM->VRAM[16] : 239760.24 K/sec
VRAM->RAM : 4428.04 K/sec
VRAM->RAM[16] : 4428.04 K/sec
Conversion : 240000.00 K/sec [output bandwidth]
Conversion : 61440000.00 pix/sec


Read RAM : 80000.00 K/sec
Write RAM : 80000.00 K/sec
RAM->RAM : 69891.41 K/sec
RAM->RAM (OS) : 88932.81 K/sec
RAM->RAM[16] : 152400.00 K/sec
Read VRAM : 4229.07 K/sec
Write VRAM : 85714.29 K/sec
RAM->VRAM : 70519.10 K/sec
RAM->VRAM[16] : 162005.79 K/sec
VRAM->RAM : 4217.93 K/sec
VRAM->RAM[16] : 4347.83 K/sec
Conversion : 70519.10 K/sec [output bandwidth]
Conversion : 18052889.32 pix/sec

_________________
Who do you serve, and who do you trust? - Galen

@DrBombcrater

Cool beans

-edit-

The JIT RAM->RAM tests have just totally maxed out there by the look of it

Hey, out of curiosity, what happens if you make the window the same size as the screen, eg

pixeltest width 1024 height 768

I wonder if the RAM->RAM tests (which work on the source data and a copy of it) are affected by large caches. Making the window larger will get around this - 648x480x2 is only 614K ish.

_________________
Doing stupid things for fun...