10-11-07, 09:58 AM
|
#1 | |
|
gentoo ~x86_64 user
Join Date: Jul 2004
Location: Germania
Posts: 213
|
shader precision
I am playing around with GLSL fragment shaders on my onboard 6150 and am wondering why they are much less precise than SSE? Both use 32bit floating points, but SSE is magnitudes more precise in my test with naive matrix multiplication. As far as I understood the GPU doesn't have eg. denorms support, but would this make such a difference?
I'd be happy for an explanation. |
|
|
|
|
10-11-07, 10:08 AM
|
#2 | |
|
Join Date: Jul 2002
Location: Netherlands, Europe
Posts: 2,105
|
Re: shader precision
I would recommend you to read the GLSL specs on what floating point format is used. That can easily explain the difference.
|
|
|
|
|
|
10-11-07, 10:14 AM
|
#3 |
|
gentoo ~x86_64 user
Join Date: Jul 2004
Location: Germania
Posts: 213
|
Re: shader precision
Well, in case I haven't made it clear, I used FP32. Using FP16 is way worse - as expected.
|
|
|
|
 |
| Thread Tools | |
|
|
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| Dell Precision 690 with a nVidia Quadro 3450 | Cr0t | NVIDIA Linux | 9 | 08-16-12 02:53 PM |
| Corrupted display - 302.17 - Dell Precision T3500 (G98 [Quadro NVS 295]) | gbailey | NVIDIA Linux | 1 | 06-27-12 10:24 AM |
| Corrupted display - 302.11 - Dell Precision T3500 (G98 [Quadro NVS 295]) | gbailey | NVIDIA Linux | 0 | 06-15-12 04:38 PM |
| Corrupted display - 295.53 - Dell Precision T3500 (G98 [Quadro NVS 295]) | gbailey | NVIDIA Linux | 9 | 06-12-12 03:34 PM |
| Precision and purpose: Ubuntu 12.04 and the Unity HUD reviewed | News | Latest Tech And Game Headlines | 0 | 05-28-12 02:20 PM |
