I mapped the enter key to the emulated DS start button in DraStic just as an alternative but in practice I use it for start a lot more than Pandora's start button because it's a lot easier to reach.
That was a direct answer to the post before, where someone asked wether a 1080p version of the rotation/scaling chip was available, not a statement regarding a final decision - resolution wise.
And Levi never said that EvilDragon claimed they were not an option. He said that they weren't an option, which is essentially a true statement, because that is the reality of the situation until we either:
I wanted an answer from ED (or Nikolaus) about that topic, not someone stating the obvious.I don't get why you quoted my last post, wouldn't quoting my response to levy make more sense ?
Ah, thanks for the answer, hope one of 720p displays turns out to be a good contender
C] stay with the portrait display and let the rotation to be handled by the X-Server or apps themselves.
I thought that option was ruled out because it was too slow, so far ?
Yep. Works for FireFox and LibreOffice, but definitely not for games.
I thought the problem with this option is that there is a lot of stuff that does not use the X-Server (> everything that writes directly to the framebuffer), leading back to option "A", God Ginrai presented
Maybe, but that wouldn't solve the problem: the bottleneck is not doing the computation (which is trivial), but the memory operations: at 60Hz, reading and writing a full framebuffer takes a significant portion of the available memory bandwidth, and the speed at which you can do it is bounded by the speed of the memory. So it does not matter whether you do it on an A15 core, an M4, the DSP or the GPU -- the problem will remain (the only exception being GL|ES applications, where you can avoid the problem by making the GPU render everything rotated).
Not really. It was in your post to LEOXD that you insinuated that they were telling you that EvilDragon decided to skip it completely. I quoted your most recent post simply because it was part of the same line of discussion.
Without even knowing the clock speed it's impossible to guess. If they can sustain 1 load or store per cycle with a bunch of ldm/stm instructions you'd need a minimum of 110.6MHz. In practice I doubt they could get anywhere close to this since they probably aren't really designed to be able to stream memory as fast as possible. Even the Cortex-A15 apparently struggles to get anywhere close to peak bandwidth and it has hardware prefetching while the M4s probably don't.
Yeah, I was still left confused too even after the translation. lol
