We've seen PS1 Emulation (I believe FFVII was showcased), which is 3D without any GPU acceleration at all, and it hit nearly 200FPS (at higher clock rates, I think). Worry not, we've got great 3D.lobski said:So far it's all been still images, and 2D stuff.
When will be get to see Quake 3, etc in action? Never?
AireTamStorm said:We've seen PS1 Emulation (I believe FFVII was showcased), which is 3D without any GPU acceleration at all, and it hit nearly 200FPS (at higher clock rates, I think). Worry not, we've got great 3D.lobski said:So far it's all been still images, and 2D stuff.
When will be get to see Quake 3, etc in action? Never?
(Edit: If I'm wrong, smack me.)
(Edit 2: Link to video Pandora running PS1 emulator: http://www.youtube.com/watch?v=a7sx8CrVLW8 )
That's the closest we've seen, but I wanted to see the Pandora at its fullest, running something in a totally 3D environment. I wonder why they haven't done it yet.
In terms of non-emulated 3D, you will likely see something in the coming months. You can look at the PowerVR SGX Video demos to see the 3D renders they have done with the chip here: http://www.imgtec.com/PowerVR/sgx.asp (requires registration)lobski said:That's the closest we've seen, but I wanted to see the Pandora at its fullest, running something in a totally 3D environment. I wonder why they haven't done it yet.
lobski said:AireTamStorm said:We've seen PS1 Emulation (I believe FFVII was showcased), which is 3D without any GPU acceleration at all, and it hit nearly 200FPS (at higher clock rates, I think). Worry not, we've got great 3D.lobski said:So far it's all been still images, and 2D stuff.
When will be get to see Quake 3, etc in action? Never?
(Edit: If I'm wrong, smack me.)
(Edit 2: Link to video Pandora running PS1 emulator: http://www.youtube.com/watch?v=a7sx8CrVLW8 )
That's the closest we've seen, but I wanted to see the Pandora at its fullest, running something in a totally 3D environment. I wonder why they haven't done it yet.
Maybe because it hasn't been ported yet?
Gee, what a novel idea.
The devs have been showing us videos of what has been ported. If you want to see 3D videos, then wait until a dev ports some 3D game. Your best bet atm will probably be to wait until Tom Cooksey (that's his name, right?) ports Quake 3, then you might see a video of that.
-God Ginrai
Because nothing fitting that description has been ported yet. Someone has only just recently taken up porting Quake 3, and it might be a while before it's in any condition for public display.lobski said:That's the closest we've seen, but I wanted to see the Pandora at its fullest, running something in a totally 3D environment. I wonder why they haven't done it yet.
I'm wondering if they even know how to fully optimize the CPU yet. Every port we've seen just uses straight horespower. Do they even have all the resources and documentation from TI to maximize CPU and 3D performance? I hate to think that the Pandora team may still be at the mercy of TI. If they are dependent on TI this could lead to still more delays.lobski said:That's the closest we've seen, but I wanted to see the Pandora at its fullest, running something in a totally 3D environment. I wonder why they haven't done it yet.
I believe ARM (including the Cortex series) is pretty much an open spec architecture. The lack of documentation I believe is in the GPU and maybe the SIMD Coprocessor.DisgruntleElf said:I'm wondering if they even know how to fully optimize the CPU yet. Every port we've seen just uses straight horespower. Do they even have all the resources and documentation from TI to maximize CPU and 3D performance? I hate to think that the Pandora team may still be at the mercy of TI. If they are dependent on TI this could lead to still more delays.
AireTamStorm said:I believe ARM (including the Cortex series) is pretty much an open spec architecture. The lack of documentation I believe is in the GPU and maybe the SIMD Coprocessor.DisgruntleElf said:I'm wondering if they even know how to fully optimize the CPU yet. Every port we've seen just uses straight horespower. Do they even have all the resources and documentation from TI to maximize CPU and 3D performance? I hate to think that the Pandora team may still be at the mercy of TI. If they are dependent on TI this could lead to still more delays.
the SIMD stuff is documented, it's just neatly hidden
http://infocenter.arm.com/help/index.jsp?t...e/Babfjcjb.html
That's the instruction timings for the various different instructions.
That said, it seems difficult to get a good guide, although I'm sure I had one somewhere, but now I can't find it
andys said:AireTamStorm said:I believe ARM (including the Cortex series) is pretty much an open spec architecture. The lack of documentation I believe is in the GPU and maybe the SIMD Coprocessor.DisgruntleElf said:I'm wondering if they even know how to fully optimize the CPU yet. Every port we've seen just uses straight horespower. Do they even have all the resources and documentation from TI to maximize CPU and 3D performance? I hate to think that the Pandora team may still be at the mercy of TI. If they are dependent on TI this could lead to still more delays.
the SIMD stuff is documented, it's just neatly hidden
http://infocenter.arm.com/help/index.jsp?t...e/Babfjcjb.html
That's the instruction timings for the various different instructions.
That said, it seems difficult to get a good guide, although I'm sure I had one somewhere, but now I can't find it
Hrm, looking a bit harder, it seems pretty damn hard to find instruction formats or anything like that. But the information has to be out there somewhere as its in gas.
The cycle timings seem useful at least, to get an idea of how everything is connected together. I don't know why they want to hide the NEON instruction formats - it's not exactly challenging to find this stuff out if you wanted to.
NEON Instruction sets are not fully implemented in Gcc yet, so that may very well be the cause.andys said:Hrm, looking a bit harder, it seems pretty damn hard to find instruction formats or anything like that. But the information has to be out there somewhere as its in gas.
The cycle timings seem useful at least, to get an idea of how everything is connected together. I don't know why they want to hide the NEON instruction formats - it's not exactly challenging to find this stuff out if you wanted to.
Mutilator said:..hopefully we will see a quake 3 port around launch time
Heh, I read that as 'lunchtime'.
Then what is this http://infocenter.arm.com/help/index.jsp?t...h/Bcfjicfj.html ?Laurent said:ARM architecture document (including NEON instructions) is not public and even less open.
craigix said:Then what is this http://infocenter.arm.com/help/index.jsp?t...h/Bcfjicfj.html ?Laurent said:ARM architecture document (including NEON instructions) is not public and even less open.
... AWW SNAP, SON!
;P
-God Ginrai
craigix said:Then what is this http://infocenter.arm.com/help/index.jsp?t...h/Bcfjicfj.html ?
But are the binary opcodes documented? As in .. I see VABS for example but whats the opcode?
craigix said:Then what is this http://infocenter.arm.com/help/index.jsp?t...h/Bcfjicfj.html ?Laurent said:ARM architecture document (including NEON instructions) is not public and even less open.
I know that link very well, it's part of the compilation of links I put in the forum that Chip copied to the wiki months ago.
This is some overview of NEON instructions. As others pointed out, can you find encoding here? Also for a real programmer the description there is far from enough (and if you don't trust me I can point you to the FFmpeg NEON developper who finally got access to the real ARM architecture document).
So I basically stick to my original claim: most of the community lacks important documentation.
BTW for those who want to get the opcodes, look at the binutils disassembler, it has them all