X does all sorts of 70s stuff like drawing lines and circles. Nobody uses that anymore.
We're better off with the 3D hardware.
Announcement
Collapse
No announcement yet.
GLAMOR Radeon Shows 2D Acceleration Promise
Collapse
X
-
RENDER and X APIs are in general, terrible.
It wouldn't be smart to put resources in developing hardware specifically for APIs that everybody wants to go away.
Leave a comment:
-
Originally posted by ssvb View PostIn other words, the hardware vendors have just failed miserably to deliver the hardware capable of properly accelerating RENDER after all these years?
Leave a comment:
-
Originally posted by ssvb View PostIn other words, the hardware vendors have just failed miserably to deliver the hardware capable of properly accelerating RENDER after all these years?
What about the dedicated 2D hardware accelerators? Are they going to solve the problem? Seems like such dedicated 2D hardware accelerators are becoming popular in modern ARM hardware (Exynos4, OMAP4470, ...).Last edited by agd5f; 13 July 2012, 09:08 AM.
Leave a comment:
-
Originally posted by agd5f View PostThe problem is RENDER semantics map poorly to 3D hardware.
What about the dedicated 2D hardware accelerators? Are they going to solve the problem? Seems like such dedicated 2D hardware accelerators are becoming popular in modern ARM hardware (Exynos4, OMAP4470, ...).
Leave a comment:
-
Originally posted by brent View PostThe main takeaway from this is that EXA really does perform rather lousy with radeon. Hint: count the number of RADEON_FALLBACK macros in the radeon DDX.
Leave a comment:
-
The main takeaway from this is that EXA really does perform rather lousy with radeon. Hint: count the number of RADEON_FALLBACK macros in the radeon DDX.
Leave a comment:
-
EXA and UXA were designed years ago when a much smaller subset of operations could be accelerated to provide decent performance for most desktops. As the number of features and complexity of those features used on newer desktops has grown, the "2D" acceleration architectures have not kept up. Expanding EXA (or UXA) is a lot of work and the complexity is getting to the level where a good "2D" acceleration driver begins to rival a 3D driver. That's why something like glamor makes sense. You only have to write one hw driver (for GL) and you can take advantage of it for "2D" as well. The xorg gallium state tracker works similarly, but since it uses EXA, it is stuck with the inherent limitations of that older acceleration architecture. Since glamor uses GL directly, it can take advantage of all the flexibility offered by GL.
Another thing to note is that "2D" APIs are not as relevant today as they once were. A lot of apps use GL directly now so putting a lot of effort into "2D" it not always a good use of time. Benchmarks often have little bearing decent desktop performance.
Leave a comment:
-
EXA and UXA were designed years ago when a much smaller subset of operations could be accelerated to provide decent performance for most desktops. As the number of features and complexity of those features used on newer desktops has grown, the "2D" acceleration architectures have not kept up. Expanding EXA (or UXA) is a lot of work and the complexity is getting to the level where a good "2D" acceleration driver begins to rival a 3D driver. That's why something like glamor makes sense. You only have to write one hw driver (for GL) and you can take advantage of it for "2D" as well. The xorg gallium state tracker works similarly, but since it uses EXA, it is stuck with the inherent limitations of that older acceleration architecture. Since glamor uses GL directly, it can take advantage of all the flexibility offered by GL.
Another thing to note is that "2D" APIs are not as relevant today as they once were. A lot of apps use GL directly now so putting a lot of effort into "2D" it not always a good use of time. Benchmarks often have little bearing decent desktop performance.
Leave a comment:
-
Originally posted by GreatEmerald View PostHey, cooperation is always good. Intel has acceleration that turned out to be not very spectacular, but why not give it to everyone, including AMD, if they have even worse acceleration. Why AMD has such poor acceleration in the first place is another matter, however...
Leave a comment:
Leave a comment: