AMDs APP SDK 2.5 leverages Bulldozers capabilities.

[grimpr]

"Kernel launch times have been further reduced.
The LLVM compiler version used for OpenCL kernels has been upgraded.

Includes support for use of SSE3 and SSE4.
Added support for partial use of FMA4 and XOP instructions.

It is no longer necessary to use the -fno-alias compiler command line option.
PCIe transfer overhead has been reduced under Linux.
Transfers between CPUs and GPUs are improved for buffers declared with either the CL_MEM_USE_HOST_PTR or the CL_MEM_ALLOC_HOST_PTR flag.
For APUs, zero copy buffers created as CL_MEM_ALLOC_HOST_PTR | CL_MEM_READ_ONLY offer improved GPU read performance.
The runtime supports multi-GPU, including simultaneous use of the GPU on both and APU and a discrete GPU on systems running under Windows.
OpenCL built-in functions leverage AVX on capable CPUs
Support for PowerExpress 4.0.
Support for atomic counters for discrete GPUs.
Support for headless GPU operation.
OpenCL can be used by a Windows service.
UVD3 / MPEG-2 support.
The clFFT library supports radix 3 and radix 5, including support for mixed radix 2/3/5.
The BLAS library supports the D/S SYRK, D/S SYR2K, D/S GEMV, D/S SYMV functions.
The FP64 extension is supported for the ATI Radeon™ HD 5900 and 5800 series, as well as the AMD FirePro™ V8800 and V8700 series.
gDEBugger 6.0 extension is available for Visual Studio.
Starting with Catalyst 11.8, improved runtime features appear regularly in the monthly Catalyst releases for Windows.
Kernel Analyzer 1.9 supports Catalyst releases 11.4 to 11.7."

http://blogs.amd.com/developer/2011...=feed&utm_campaign=Feed:+amd/all+(All+of+AMD)

 

[Chiropteran]

Can't download it yet, getting some access denied error from the download link.

 

[Phynaz]

When do we get to yell that AMD is cheating for optimizing their compiler for their architecture?

 

[sdifox]

When do we get to yell that AMD is cheating for optimizing their compiler for their architecture?

why would you do that? optimisation for target platform is what seperates good compilers from shitty compilers.

 

[Nemesis 1]

LOL ^ post 3 Ya right. Its like AMD invented open CL . LOL . Its nice that they are working on CL/AVX . Chiropteran Do you haveBD. Why would you try downloading a sku thats works on BD cores?

 

[WhoBeDaPlaya]

When do we get to yell that AMD is cheating for optimizing their compiler for their architecture?

We don't have to, as long as they don't purposely disable capabilities that are common to both Intel and AMD CPUs when the generated code is run on Intel CPUs

 

[waffleironhead]

LOL ^ post 3 Ya right. Its like AMD invented open CL . LOL . Its nice that they are working on CL/AVX . Chiropteran Do you haveBD. Why would you try downloading a sku thats works on BD cores?

There is more to the 2.5 sdk than BD. Personally I'm interested in the headless gpu operation ability.

 

[drizek]

When do we get to yell that AMD is cheating for optimizing their compiler for their architecture?

I don't really care, and I'm not going to get into this argument, but the problem isn't that Intel optimizes their compiler, it is that they unoptimize it for AMD systems.

Is there any evidence that the AVX stuff won't work on intel chips as well?

 

[Phynaz]

I don't really care, and I'm not going to get into this argument, but the problem isn't that Intel optimizes their compiler, it is that they unoptimize it for AMD systems.

Is there any evidence that the AVX stuff won't work on intel chips as well?

Actually it's been proven that Intel's compiler optimizes for AMD chips better than other compilers such GCC. Intel's compiler optimizes for their own architecture even better

 

[Elixer]

Actually it's been proven that Intel's compiler optimizes for AMD chips better than other compilers such GCC. Intel's compiler optimizes for their own architecture even better

Proven by whom?
In fact, wasn't that long ago when the FTC was all over intel's case for doing crappy code for both AMD & Via CPUs.

 

[Topweasel]

Proven by whom?
In fact, wasn't that long ago when the FTC was all over intel's case for doing crappy code for both AMD & Via CPUs.

Last I heard from the guys who dug into to this day Intel's compilers still do a genuineIntel check for SSE4 and AVX code. It doesn't do the check for SSE2 anymore.

 

[Chiropteran]

LOL ^ post 3 Ya right. Its like AMD invented open CL . LOL . Its nice that they are working on CL/AVX . Chiropteran Do you haveBD. Why would you try downloading a sku thats works on BD cores?

Bitcoin mining

 

[Phynaz]

Proven by whom?
In fact, wasn't that long ago when the FTC was all over intel's case for doing crappy code for both AMD & Via CPUs.

60 seconds of Googling, and I didn't even find the original article I was referencing. One of the papers is from 2006, during the period always spoken about.

http://www.principledtechnologies.com/clients/reports/Intel/CompComp.pdf

http://www.cse.scitech.ac.uk/disco/Benchmarks/Opteron_compilers.pdf

"When it was used, ICC compiled versions of Lux ran faster than GCC both on Intel and AMD platforms" - Jeanphi, developer.

Feel free to read the testing over at the RWT forum too.

 

[dac7nco]

OpenCL can be used by a Windows service.

This is not new; this has been functioning on many, many clusters since late last year, after engineers decided against getting their brains cornholed by CUDA. OpenCL is slower, but much, much more efficient in terms of man-hours. What OpenCL NEEDS is development by AMD - Apple doesn't have the inclination - they do have the cash and expertise. The headless functionality is solved with a fu*king resistor.

Daimon

 

[dac7nco]

I've been running OpenCL enabled code for actually more than a year now; the most recent development has been HPC erratta for IBM 755s. There's also this for you researchers and college students out there: http://reference.wolfram.com/mathematica/OpenCLLink/guide/OpenCLLink.html

 

[Nemesis 1]

Bitcoin mining

LOL!Get a job!

 

[Nemesis 1]

Proven by whom?
In fact, wasn't that long ago when the FTC was all over intel's case for doing crappy code for both AMD & Via CPUs.

The FTC was all over Intel . NOT! AMD and NV were complaining about Intel compilers so the FTC fixed the problem . The solution . Intel hasa to add a disclaimer. Stating that there compilers may not function on other hardware the same as it does on intel products.

The FTC was so angry with intel that they came up with this solution . Which infact got AMD and NV off of intels back because of the final solution . This all worked in Intel favor .

 

[Idontcare]

The FTC was all over Intel . NOT! AMD and NV were complaining about Intel compilers so the FTC fixed the problem . The solution . Intel hasa to add a disclaimer. Stating that there compilers may not function on other hardware the same as it does on intel products.

The FTC was so angry with intel that they came up with this solution . Which infact got AMD and NV off of intels back because of the final solution . This all worked in Intel favor .

Good thing there was no such thing as collusion between the government regulating agencies and american businesses during that time period...Haliburton, goldman sachs, countrywide, etc...never happened, and even if it did we sure wouldn't add white-as-the-driven-snow Intel to that list.

I wouldn't count on the FTC to do jack, and the FTC electing to not do jack does not constitute proof in my book that jack didn't need to be done.

 

[Vesku]

I think this is why Intel is hard at work on their own OpenCL libraries, they know what sorts of shenanigans can be done to make Ivybridge seem worse than it will actually be at GPU Compute. Assuming Ivybridge is launched with GPU Compute working on it's EUs.

 

[podspi]

I doubt AMD would work to make OpenCL work poorly on IB or NV cards. At the moment, AMD has an advantage in those types of workloads, what they need right now is adoption. They aren't Intel, where you can cripple your competitor and people will STILL adopt the technology. If AMD cripples OpenCL nobody will use it, because nobody (you know what I mean) will have the hardware to take advantage of it.

 

[Topweasel]

I doubt AMD would work to make OpenCL work poorly on IB or NV cards. At the moment, AMD has an advantage in those types of workloads, what they need right now is adoption. They aren't Intel, where you can cripple your competitor and people will STILL adopt the technology. If AMD cripples OpenCL nobody will use it, because nobody (you know what I mean) will have the hardware to take advantage of it.

Agreed. Having a product with a big disclaimer that I run poorly on Intel products is a great way to make sure no one touches it.