User-agent: Mozilla Thunderbird 0.8 (Macintosh/20040913)
I have performed the following experiment on a Radeon 9600 running on a
powerbook G4. (I cannot get kCGLPFAColorFloat to give me a floating
pbuffer.
However, kCGLPFAColorSize 128 should give me fixed point format. So
I wrote the value "x=3" into the pbuffer, and then read it into the
frame buffer
through the routine "rewireExp". The experiments indicate that by the
time I computed
"acc" in rewireExp, the value had been clamped to 1, which should not
have happened
if the value was truly stored in fixed point format. I expected the
value read into "acc"
to equal 3, so that when I divided by "3", "rewireExp" would return 1.
In actuality
the return value is 1/3. What is happening?
The expertise regarding floating point buffers is extremely sparse, and
there are no demonstration
programs using AGL out on the web. Thanks.
Gordon
float4 initExp(float2 tc : TEXCOORD0) // [N,N] on nvidia
: COLOR
{
float x = 3.0;
float4 color = float4(x,x,x,1.0);
return color;
}
//----------------------------------------------------------------------
float4 rewireExp(float2 tc : TEXCOORD0,
uniform sampler2D acc_noise) : COLOR
{
float4 acc = tex2D(acc_noise, tc); // orig
float3 color = acc.xyz / 3.0;
return float4(color,1.);
}
-----------------
Some of the buffer characteristics:
int i=0;
attrib[i++] = kCGLPFAColorSize;
// works with 96 and 128, but only if kCGLPFAColorFloat is disabled
attrib[i++] = 128;
----------
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Mac-opengl mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/mac-opengl/email@hidden
