Hi list,
I'm working on an installation that makes intensive use of evt, so first big thanks to the developpers for this incredible tool ! Still, during the development I faced a really strange issue with ev_pro_22easy modules : it kills my framerate.
The patch uses more than 20 ev_pro_22 modules, some for movie texturing and some for masking. The masking was directly copied from the examples, so using ev_pro_22easy and rendering the same black texture on them. My framerate went down to 15-20fps. As soon as I stop the "masking" gem chain, fps reaches 26. I finally replaced the ev_pro_22easy with ev_pro_22 for masking, and fps then stayed at 26.
Is there any reason why ev_pro_22easy is less efficient than ev_pro_22 ?
Thanks
El jue, 10-10-2013 a las 11:52 +0200, Etienne Landon escribió:
Is there any reason why ev_pro_22easy is less efficient than ev_pro_22 ?
I think ev_pro__22 uses OpenGl shaders so it uses the GPU in the graphic card. ev_pro_22easy uses the main CPU to made the work, so is less efficient, but it works in older machines.
Great piece of software the extended toolkit view...
Cheers.
Santi Noreña.
Hi,
I understand the shader part of ev_pro_22 but I 'm not shure the problem is here, the shader is only used to change texture coordinates and do soft-edging. In both cases, ev_pro_22 or ev_pro_22easy, I send a texture ID. So the chain with ev_pro_22 is not the one that does the work from movie to texture (the CPU part), it's done once by another chain. For what I understand, using the same texture ID for different gem chains avoids using the CPU for each chain, you use CPU only to "transform your image, movie or whatever to a texture (which is stored in the GPU memory), then each chain links to the texture ID which is the position of the texture in the graphic card's memory. Am I wrong ?
For instance, in the same work I do texturing on 200 pmpd masses linked to a gem chain. So 200 independant rendering chains, one texture, and this works without a problem. And there's no shader involved
Moreover, the texture used for masking is a only one black pixel (see the masking example), so how come this could be so heavy ?
2013/10/10 Santi belfegor@gmail.com
El jue, 10-10-2013 a las 11:52 +0200, Etienne Landon escribió:
Is there any reason why ev_pro_22easy is less efficient than ev_pro_22 ?
I think ev_pro__22 uses OpenGl shaders so it uses the GPU in the graphic card. ev_pro_22easy uses the main CPU to made the work, so is less efficient, but it works in older machines.
Great piece of software the extended toolkit view...
Cheers.
Santi Noreña.
I found the same behavior, Marian and Peter wrote some amazing software there but you have to have a really fast graphics card for some of the fancier modules. Go with the one that delivers for you. 22 is really nice as I recall. I'll check my rates today and report back too
Patrick Pagano B.S,M.F.A Asst. in Digital Art and Science Digital Worlds Institute University of Florida (352) 294-2020
On Oct 10, 2013, at 8:38 AM, "Etienne Landon" <landon.etienne@gmail.commailto:landon.etienne@gmail.com> wrote:
Hi,
I understand the shader part of ev_pro_22 but I 'm not shure the problem is here, the shader is only used to change texture coordinates and do soft-edging. In both cases, ev_pro_22 or ev_pro_22easy, I send a texture ID. So the chain with ev_pro_22 is not the one that does the work from movie to texture (the CPU part), it's done once by another chain. For what I understand, using the same texture ID for different gem chains avoids using the CPU for each chain, you use CPU only to "transform your image, movie or whatever to a texture (which is stored in the GPU memory), then each chain links to the texture ID which is the position of the texture in the graphic card's memory. Am I wrong ?
For instance, in the same work I do texturing on 200 pmpd masses linked to a gem chain. So 200 independant rendering chains, one texture, and this works without a problem. And there's no shader involved
Moreover, the texture used for masking is a only one black pixel (see the masking example), so how come this could be so heavy ?
2013/10/10 Santi <belfegor@gmail.commailto:belfegor@gmail.com> El jue, 10-10-2013 a las 11:52 +0200, Etienne Landon escribió:
Is there any reason why ev_pro_22easy is less efficient than ev_pro_22 ?
I think ev_pro__22 uses OpenGl shaders so it uses the GPU in the graphic card. ev_pro_22easy uses the main CPU to made the work, so is less efficient, but it works in older machines.
Great piece of software the extended toolkit view...
Cheers.
Santi Noreña.
Pd-list@iem.atmailto:Pd-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Hello Etienne!
The problem you are facing could be really due to the different structure of both modules. And the amount of ev_pro-modules you are using in your patch. If you want me to have a quick look on your patch, maybe answer me and attach the your patch to the mail. maybe i can help you with streamlining it further :)
cheers, peter
Gesendet: Donnerstag, 10. Oktober 2013 um 15:07 Uhr Von: "Pagano, Patrick" pat@digitalworlds.ufl.edu An: "Etienne Landon" landon.etienne@gmail.com Cc: "pd-list@iem.at" pd-list@iem.at Betreff: Re: [PD] Extended View toolkit : issue with masks
I found the same behavior, Marian and Peter wrote some amazing software there but you have to have a really fast graphics card for some of the fancier modules. Go with the one that delivers for you. 22 is really nice as I recall. I'll check my rates today and report back too
Patrick Pagano B.S,M.F.A Asst. in Digital Art and Science Digital Worlds Institute University of Florida (352) 294-2020 On Oct 10, 2013, at 8:38 AM, "Etienne Landon" landon.etienne@gmail.com wrote:
Hi, I understand the shader part of ev_pro_22 but I 'm not shure the problem is here, the shader is only used to change texture coordinates and do soft-edging. In both cases, ev_pro_22 or ev_pro_22easy, I send a texture ID. So the chain with ev_pro_22 is not the one that does the work from movie to texture (the CPU part), it's done once by another chain. For what I understand, using the same texture ID for different gem chains avoids using the CPU for each chain, you use CPU only to "transform your image, movie or whatever to a texture (which is stored in the GPU memory), then each chain links to the texture ID which is the position of the texture in the graphic card's memory. Am I wrong ? For instance, in the same work I do texturing on 200 pmpd masses linked to a gem chain. So 200 independant rendering chains, one texture, and this works without a problem. And there's no shader involved Moreover, the texture used for masking is a only one black pixel (see the masking example), so how come this could be so heavy ? 2013/10/10 Santi <belfegor@gmail.com[belfegor@gmail.com]>El jue, 10-10-2013 a las 11:52 +0200, Etienne Landon escribió:
Is there any reason why ev_pro_22easy is less efficient than ev_pro_22 ?
I think ev_pro__22 uses OpenGl shaders so it uses the GPU in the graphic card. ev_pro_22easy uses the main CPU to made the work, so is less efficient, but it works in older machines.
Great piece of software the extended toolkit view...
Cheers.
Santi Noreña.
_______________________________________________ Pd-list@iem.at[Pd-list@iem.at] mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list%5Bhttp://lists.puredata.info/lis...]
Thanks for the proposal !
I get the computers with the patch back tomorrow, so i'll try and clean it a little and send it.
Etienne
2013/10/12 Peter Venus news@petervenus.de
Hello Etienne!
The problem you are facing could be really due to the different structure of both modules. And the amount of ev_pro-modules you are using in your patch. If you want me to have a quick look on your patch, maybe answer me and attach the your patch to the mail. maybe i can help you with streamlining it further :)
cheers, peter
Gesendet: Donnerstag, 10. Oktober 2013 um 15:07 Uhr Von: "Pagano, Patrick" pat@digitalworlds.ufl.edu An: "Etienne Landon" landon.etienne@gmail.com Cc: "pd-list@iem.at" pd-list@iem.at Betreff: Re: [PD] Extended View toolkit : issue with masks
I found the same behavior, Marian and Peter wrote some amazing software there but you have to have a really fast graphics card for some of the fancier modules. Go with the one that delivers for you. 22 is really nice as I recall. I'll check my rates today and report back too
Patrick Pagano B.S,M.F.A Asst. in Digital Art and Science Digital Worlds Institute University of Florida (352) 294-2020
On Oct 10, 2013, at 8:38 AM, "Etienne Landon" landon.etienne@gmail.com wrote:
Hi,
I understand the shader part of ev_pro_22 but I 'm not shure the problem is here, the shader is only used to change texture coordinates and do soft-edging. In both cases, ev_pro_22 or ev_pro_22easy, I send a texture ID. So the chain with ev_pro_22 is not the one that does the work from movie to texture (the CPU part), it's done once by another chain. For what I understand, using the same texture ID for different gem chains avoids using the CPU for each chain, you use CPU only to "transform your image, movie or whatever to a texture (which is stored in the GPU memory), then each chain links to the texture ID which is the position of the texture in the graphic card's memory. Am I wrong ?
For instance, in the same work I do texturing on 200 pmpd masses linked to a gem chain. So 200 independant rendering chains, one texture, and this works without a problem. And there's no shader involved
Moreover, the texture used for masking is a only one black pixel (see the masking example), so how come this could be so heavy ?
2013/10/10 Santi <belfegor@gmail.com[belfegor@gmail.com]>El jue, 10-10-2013 a las 11:52 +0200, Etienne Landon escribió:
Is there any reason why ev_pro_22easy is less efficient than ev_pro_22 ?
I think ev_pro__22 uses OpenGl shaders so it uses the GPU in the graphic card. ev_pro_22easy uses the main CPU to made the work, so is less efficient, but it works in older machines.
Great piece of software the extended toolkit view...
Cheers.
Santi Noreña.
Pd-list@iem.at[Pd-list@iem.at] mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list%5Bhttp://lists.puredata.info/lis...]
-
Etienne Landon -
Pagano, Patrick -
Peter Venus -
Santi