AMDKFD Is Present For Linux 3.19 In Open-Source HSA Start

[carewolf]

I don't get it. What is the point of HSA?

It sound very uninteresting and useless if it is a AMD-only thing.
Without Intel, Nvidia or others, it does not have any appeal.

No Intel just has another name for it.

[thegeek6]

Kabini doesn't support HSA, so: No.

I figured that was what would be told. Thanks for the answer.

[Kivada]

I don't get it. What is the point of HSA?

It sound very uninteresting and useless if it is a AMD-only thing.
Without Intel, Nvidia or others, it does not have any appeal.

It's open to everyone, most of the ARM SoC companies(Qualcom, Samsung, VIA/S3, Marvel, Broadcom, Sony, Toshiba, etc.) are already jumping on the tech as you can see from the foundation's list of members http://www.hsafoundation.com/ AMD is pretty much the core developer of it and the only one currently brining it to x86-64 based hardware. And it's extremely interesting tech that will provide large gains in performance by making better use of the hardware.

Basically should it take off then Nvidia and Intel wil be forced to impliment it, just like Intel was forced to accept x86-64 after AMD created it after they had claimed it was impossible to make a 64-bit x86 based CPU so they could try and sell people on the Itanium.

[Kivada]

Tell that to cuda It's scary how much scientific stuff out there is dependent on it.

HSA is simply AMD's branding of the long-envisioned merging of the CPU and GPU which eliminates the overhead of GPU-CPU transfers. Intel is working on something similar.

Getting there first could be really big for AMD, but I hope that we end up with an open standard and not another cuda.

Nvidia paid off alot of people to get CUDA accepted over OpenCL, they had to in order to stay relevant. But the world has started moving toward OpenCL and now with HSA being picked up by the ARM community OpenCL development is going to get a big shot in the arm.

Open standards almost always will out in the long run, the last big hurdle for Linux is DirectX, if OpenGL and SDL can at the very least catch up to DirectX then it's days will be numbered.

I think it was a mistake for Nvidia to not try and merge with VIA/S3 back around 2008 and fight Intel in court to allow them to keep the x86 license. The x86 market would be allot better off if there where more players capable of competing in the market, but CPU, GPU or Chipset alone wont help anyone and you need large cashflow to keep advancing. Nvidia lost it's chipset market and was then forced to move into ARM SoCs to find another avenue for profit, while VIA/S3 has all of the pieces but not the cash flow or patents to build compelling hardware in either CPUs or GPUs, which is why they are also making ARM SoCs while their x86-64 based hardware hasn't advanced much since around 2008 and nothing really new since 2011 with their Nano series CPUs and Chrome 540 GTX HTPC class GPU. With Nvidia's cashflow and GPU patents and VIA/S3's x86 license and patents the combined company would be far more capable across the board, including the ARM SoC market then they are currently.

[Kivada]

I figured that was what would be told. Thanks for the answer.

It was only in a few early roadmap slides that Kabini was slated to have HSA support, but later slides removed HSA from the listings and AMD said that the released Kabini hardware is lacking a few hardware features that would allow it to be able to make use of HSA.

Guessing that they had to drop the support due to yeild issues and/or couldn't hit the desired price point for Kabini if they added the HSA hardware bits.