cancel
Showing results for 
Search instead for 
Did you mean: 

Archives Discussions

npm1
Adept II

Blender Cycles(Opencl on AMD GPUS)

Dear Opencl Developer

Why doesn't the AMD opencl compiler work with blender cycles?

whenever i compile the blender cycles kernel, the system either crashes due to lack of memory, or takes too long to compile the blender cycles kernel(which thereafter comes up with the following error:

opencl build failed:errors in console

calclcompile failederror: creating kernel_ocl_path_trace failed!

can't open file c:\tmp\5688.blend@ for writing:no file or directory

).

When is AMD opencl compiler going to work properly with blender cycles?

Why can't the AMD opencl compiler developers test their compiler against blender cycles?

Seasons Greetings,

npm1,

PS i am as well as others(i assume) are considering to make a switch from AMD GPUs to Nvidia.

371 Replies

LLVM/Clang != AMD Opencl Compiler.

LLVM/clang is not going to help Cycles render to work on AMD GPUs and probably is going to make it run slower on CPUs too.

LLVM is faster compiling but creates a slightly slower code than gcc, also blender devs are not having problems with llvm are having problems with AMD OPENCL COMPILER.


Blender devs: Silent.


Ton Roosendaal twitter:


We are in contact with too many AMD people already. It's their turn. JUST FIX THE COMPILER!!!


0 Likes

AMD OpenCL compiler uses LLVM.

0 Likes
sdar
Journeyman III

llvm don't choke to death when compiling large kernels, so it's better to say that it's based on llvm but is not working like llvm anymore.

Just test Intel Opencl compiler that is based on llvm and you'll compile almost any large kernel within seconds.

Ex. Luxrender all materials and textures enabled

Intel OpenCL CPU 4.5 Seconds

AMD OpenCL CPU >15 minutes

AMD OpenCL GPU  N/A


Source


Anyway using a different compiler for the rest of blender is not going to help cycles run on AMD GPUs cause cycles for GPU Opencl uses the compiler included in your drivers and it compiles itself the first time you use it.

0 Likes

That's why many times Sdar to compile the kernel gpu must be charged first cpu. That last ah me sometimes if fee directly to gpu gpu often the program hangs and Blender.

0 Likes

I know that the AMD OpenCL compiler is not that good. But eh, I just wanted to tell you that LLVM is used by AMD. Nothing more.

Just let pray that they make some progress on their compiler.

0 Likes

AMD Catalyst 14.7 Beta Driver for Windows

There is a new version of driver out. Can some one try it and post whether has any improvements regarding Blender?

0 Likes

Same OpenCL version, same speed, same limitations. The promised "mid July" release became "mid August" and didn't deliver anything. I just have enough, after 2 years of waiting. It's about time to switch back to Nvidia.

0 Likes

I last posted in this thread on

Not happy about the lack of Info or progress myself, But I would wait till after the 25th of this month. after doing some research online I found AMD is working on updating to LLVM 3.5 for the compiler and if it was me I would not put an updated compiler out till that was finished myself. But that aside I would also post this info here as it is public knowledge if you look for it.  

0 Likes

AMD just released CLOC (HSAFoundation/CLOC · GitHub). Can any one here try compiling blender kernel with that? I'm wondering if that compiler is any better than the shipping compiler and eventually replace what is shipping in the current driver. Its description says "cloc: Convert an OpenCL file to brig, hsail, or .o (object) file using the LLVM to HSAIL backend compiler."

0 Likes


And yet it is AMD who gets slammed for ``dropping' the ball, while other projects manage to implement AMD's OpenCL stack well within their projects.


And yet even Luxrender that is created for working with the current AMD Opencl implementation is hitting this same bug when they try to add features or further improve the render, so you are right, the developers should be ashamed for not splitting their code in 1.000.000.000.000.000 tiny parts in hopes that all that work ended in a working code on AMD cards and making the code harder to maintain.

From the Blender wiki:


Q: Why don’t you just split up cycles so it can run better on AMD hardware?


A: While this would likely help it is not a trivial matter to split up cycles in this way. Also it is not clear that it is going to help and how much. As a resource constrained open-source project this will most likely not be a top priority


As I said, there's no way to know if splitting the code is going to be enough to make cycles work on AMD cards so resources can be used in more feasible areas instead of trying to workaround AMD mess.

So you can prefer the Open solution over the proprietary one if both solutions are working without huge performance differences but right now I'll go with the solution that actually works.

P.S. Inferior cuda have unified memory, where's Opencl 2.0 with unified memory on AMD?

0 Likes


P.S. Inferior cuda have unified memory, where's Opencl 2.0 with unified memory on AMD?


Oh please, CUDA's “unified memory” is 90% marketing and 10% software (the driver part that does the actual copying between host and device without exposing the need to do so to the developer). NVIDIA hardware is simply not capable to do truly unified memory, as opposed to AMD HSA-enabled APUs.

Regarding the OpenCL 2.0 SVM, you can already play around with them in the latest AMD drivers, even though they are still 1.2 formally (cl_amd_svm extension).

0 Likes

As far as I know Opencl 2.0 is only working on HSA capable APUs.

HSA is only working on APUs, AMD APUs performance are far behind my needs so no HSA nor Opencl 2.0 for high performance hardware, nice.

90% marketing 10% software seems to be better than 80% marketing 20% promises.

0 Likes


sdar wrote:



As far as I know Opencl 2.0 is only working on HSA capable APUs.



HSA is only working on APUs, AMD APUs performance are far behind my needs so no HSA nor Opencl 2.0 for high performance hardware, nice.



90% marketing 10% software seems to be better than 80% marketing 20% promises.


OpenCL 2.0 is not only for HSA-capable APUs. The specification is very wide and it encompasses a LOT of features, of which SVM is only a small part. And SVM itself is not limited to HSA-capable APUs, as it can be supported by any device that is capable of reading/writing host memory using the same address space as the CPU. This includes any NVIDIA devices with UVA (i.e. since Fermi).

0 Likes

Nvidia doesn't have unified memory indeed. See "CUDA 6 Unified Memory explained" for more info. Nvidia put this in the world, as it is afraid of embedded GPUs, as it should. See also my post "7 things NVIDIA doesn't want you to know" on the disputable company ethics of Nvidia. Exactly this lying is the reason I prefer OpenCL and companies like AMD - I don't want to buy from companies with low company ethics.

Current drivers of AMD are not aware of the dependency-tree of all the parts in a kernel-file. for instance a large .cl-file with 20 independent kernels compiles much slower than having them split over 20 files and compiling them one by one. So there is plenty of room for improvements.

0 Likes
cusa123
Adept I

Sorry, but this is what we expect since June ?. Because I do not see any improvement in either performance or anything that is visible at first glance.

0 Likes
amartincolby
Adept I

I'm here to do little more than add my voice to the crowd. I very much prefer AMD to Nvidia and find the proprietary CUDA framework to be highly distasteful. My dislike of CUDA means that AMD would have to fail, like, hard core, for me to jump ship.

That said, I'm about ready to actually run two video cards in my system: one Nvidia and one AMD. I'm not threatening to leave AMD fohr evarz!1! or anything like that, but it would be a lost sale.

cactoos
Adept I

Hello.

I am new writing here, but i just read all this discusion.
First, i`m don`t really speak english, so if i mis a word, i hopo all you can forgiveme

_________

Can we expect any announcement tomorrow about Cycles, OpenCL, Blender and the AMD compiler? Or may we sit and wait another 30 year more?

0 Likes

Nope. No news. That's not surprising since we know their schedule for these sorts of releases. Our next hope is mid-September.

I understand that OpenCL is still very much in its infancy, but I'm beginning to think that the issues we are encountering with AMD cards must have something to do with hardware as well as software. AMD's drivers aren't well-known for their quality (although they are getting better), but this problem is big and just keeps going. If it is indeed a hardware issue, and I will never expect my current AMD cards to effectively handle more complex OpenCL applications, I wish that AMD would tell me. Yes, I would be a bit angry, but at least I would know.

0 Likes
cusa123
Adept I

In theory the catalyst sale August 25, waited since June or will have to wait until September ?.

Amartincolby, I think the same and long. The error is in the amd driver, but at this stage is thought to be in the hardware in my opinion. What is amd what eimpulsa per firepro is priceless for the average user.

0 Likes
cactoos
Adept I

well... i am just in a date to improve my graphic card, because i actually have an HD 5670, that is pretty old now.

I think about a R9 270x or maybe an R9 280 (Both are expensive cards in my country, but i take it like investment).

I know for playing are either really good, and in fact Sony vegas works well with OpenCL in my card. But if i buy a new one, i expect a complete pack, who works fine in all duties.

And in this moment, i need AMD works in blender what is supposed to do.I know it`s the same history here for years, but, hope is the last thing to lose.

0 Likes

If you want to use an OpenCL accelerated renderer, use LuxRender . This one works as far as I know. It has a package for blender.

0 Likes
sdar
Journeyman III

But take care, Luxrender devs are having problems with AMD compiler too and you can't use it with too much features enabled at the same time cause it'll fail to compile too.

0 Likes

The developers of Luxrender are doing a fantastic job of keeping people up to date on developments. Anyone who is interested can go read the entries on GPU acceleration in the Luxrender forums.

Luxrender is working a whole lot better than Cycles on AMD hardware but it's still encountering huge problems. The big Luxcore update that's been in the oven for the past eighteen months will help some of these issues, but if the developers are to be believed, they have tried every trick in the book and are unable to get their kernel working on Radeon or Firepro boards.

If there is a breakthrough to be had I believe that it will be with Luxrender since they have completely thrown their lot in with OpenCL. Cycles development is focusing almost entirely on CUDA right now. It's for that reason that I continue to work with Luxrender. When AMD OpenCL is ready to go, so am I.

0 Likes

Sorry, but in my opinion is more complicated LuxRender that cycles and generates more noise too. Generally speaking is slower than cyclos and lacks tremendous support and OpenCL boost in performance.

Honestly in my opinion this is not amd hoping to give a final solution, if you think taking Lumiscaphe p3d OpenCL for firepro is very wrong in the 3D field.

This will remain the same as always, the worst that is sure to go and continue the same for years we've been waiting for that magical compiled and these impossible heights.

0 Likes
cusa123
Adept I

Great gpu went volumes, another thing that nvidia podiendo do this again. Amd ever further OpenCL profitable for renders.

0 Likes
cusa123
Adept I

Bob something new?

It would also be good to start watching Nox as it is open source now.

0 Likes

A new version of catalyst driver is released at AMD Catalyst™ Software Suite for AMD Radeon™ R9 285

I'm at work. So, I can't try it. But it seems to have newer version of OCL stuff. I'm just compared the files to 14.7 rc3 and noticed that OCL related files have newer versions.

Brian

0 Likes

AMD Catalyst™ Software Suite for AMD Radeon™ R9 285 is for specifically R9 285 only. However, I found modified driver that installs on my desktop Kaveri system at AMD Catalyst 14.x BETA (14.300.1005.0 August 27) - Guru3D.com Forums

It does have a different version of OpenCL runtime. However, I don't know how to test Blender.

0 Likes

Here is a video made for someone who explain how to activate OpenCL in an experimental mode by CMD.

And explain too, how to made a .bat file for easy start.

How to get Blender 2.69 to work with AMD graphics cards (enable OpenCL) part 1 - YouTube

It actually works, but Blender still crushing in my card.

0 Likes

That's not the question here. We are all able to make Cycles render on AMD OpenCL, the main problem is that amds opencl isn't operating  at 1/10 of that of Intel or Nvidia. And its crashing and compiling kernel in several minutes... Also renders very slow, if any.
We all need a functioning driver with functioning opencl for cycles. People are massively moving from ATI to Nvidia because of this. We here are trying not to.

0 Likes
cusa123
Adept I

The unfortunate thing is that the idea is to change amd now firepro us.

If cost is not only bigger, and also guarantees nothing OpenCL drivers. At this rate change to nvidia graphics is evident. The worst is that the future will be more difficult to test many have hd 6000 series onwards. At least gcn architecture hd 7000 but I can not pass it would invest in a new graph with many problems.

The following videos besides being a disaster to file a "POSSIBLY A SOLUTION GRAVE" amd does not say anything, I do not understand this.

People in this talk "amd blender cycles" we would be the only really procupados and what we try to motivate more than not being able to work or to take any information trtar for blender cycles. In any case, I called attention appoint a new renderer but apply no demo for users. A simple "bad" video not generate nothing but anger and upset for not complying with 3 years of flaws in the rendering engines.

http://www.cgchannel.com/2014/09/video-amds-new-gpu-based-raytrace-renderer-for-maya/

AMD FirePro ray tracer renderer proof of concept demo - YouTube

0 Likes
amartincolby
Adept I

So, did the recent PR release about older Firepro cards  —namely the ones sporting Tahiti chipsets— not supporting OpenCL 2.0 essentially confirm that ALL AMD cards not based on Hawaii or later will not support OpenCL 2.0? Similarly, did this confirm that all attempts to fix issues with OpenCL on these older cards are now being suspended?

I hope that this is not the case. I specifically bought my AMD cards because of their excellent OpenCL performance. I'd really hate to have three 7970's basically go to waste.

0 Likes

I do not understand, then the solution will be OpenCL 2.0.

We happen to those who have lower series like mine one 7870 or other 7770.

What can we expect to buy firepro bringing out to solve the problem?

0 Likes

Do the recently announced FirePro cards already support OpenCL 2.0?

0 Likes


sarobi wrote:



Do the recently announced FirePro cards already support OpenCL 2.0?


Here is the most comprehensive list of OpenCL 1.2 based support AMD Hardware.

https://en.wikipedia.org/wiki/AMD_Radeon_Rx_200_Series

I'd imagine the HSA initiative with OpenCL 2.0 would require more time than just tossing in OpenCL 2.0 ala Intel. Intel's heding its bets with Cilk, OpenCL and OpenMP. With far greater resources they've managed to have incomplete versions of Cilk and OpenMP on LLVM/Clang, while AMD having to streamline their approach is OpenCL and AMP to make sure its covered from OS X, Windows and Linux.

I'd imagine when more work is ready on the drivers and HSA specs are at final and not provisional status we'll get information on OpenCL and more.

If you read here: http://www.hsafoundation.com/three-core-hsa-foundation-specification-are-available/

You will see the following:

AMD is also supplying early implementation to test out capabilities of HSA


The project provides an initial implementation of the HSA specifications on the AMD “Kaveri” silicon a pre-HSA Compatible part. The implementation includes a Linux kernel and associated kernel-level drivers, the HSA runtime, and the HSAIL finalizer. The project includes a reference LLVM-based compiler which generates HSAIL and can extended to add additional languages that support HSAIL-based compute. The project also includes tools for assembling and disassembling HSAIL and for compiling OpenCL 2.0  kernels into HSAIL. Finally, the project includes an approachable runtime layer called “OKRA” designed to minimize the time required to get started with HSA.  You can access these at https://github.com/HSAFoundation


I hope whatever AMD does with the HSAFoundation source at GitHub that they update their LLVM3.2 based branch to 3.5, sooner rather than later.

0 Likes
cactoos
Adept I

Hello.

I´m now using my girlfriend´s PC. Is a A10 7700K 8Gb Gskill ram @2133 Mhz and Catalyst 14.7, and IT WORKS!!

take a lot of time for compilation, but with de cube works.

Here are a SS http://subefotos.com/ver/?cac3d252ed9816da3679c6a292228fc8o.png

Tomorrow when i be able to start my PC and traspass an heavier file, i will try it and i´ll post the result.

For now, appears that APU`s can do the work... pretty slowly, but at least works.

My only problem with the APU, it gives me BSOD at random, but that's another issue (and the reason that I have now).

I just wanted to post this because it gives me some hope.

0 Likes

http://blenderartists.org/forum/attachment.php?attachmentid=313984&d=1402322763

This render it succeeds in "26 minutes gpu" amd we unfortunately is leaving aside those who have gcn 1.0 (radeon hd 7000 and beyond). If amd firepro is definite to inform us that we have 2 and 3 year wait for a resolution.

0 Likes
cusa123
Adept I

Some time ago he was asked if amd catalyst would have the ability to pass across gcn OpenCL 1.2 to 2.0. It would be a good time to know, but if people commented blender and they would have to be updated while LLVM.

Indigo also commented that render is having good support lately but what worries me is some characteristics which uses only 20 to 30% of the gpu and 80-90% in other uses.

0 Likes
cactoos
Adept I

Hello.

I just see this in my twitter account:

http://www.subeimagenes.com/thumb/rory-says-1092181.PNG

Rory says AMD is focused on FirePro GPU`s... the pic speak on his own.

0 Likes