ARM Transactional Memory Extension Support Starts Being Plumbed For Linux

[phoronix]

Phoronix: ARM Transactional Memory Extension Support Starts Being Plumbed For Linux

Last month Arm announced Scalable Vector Extension 2 (SVE2) and Transactional Memory Extension (TME) as two new technologies for its A-Profile architecture. That TME support is already being plumbed into the Linux toolchain stack...

http://www.phoronix.com/scan.php?page=news_item&px=ARM-TME-GNU-Binutils

[ms178]

Here is a press release from April 2019 with some more details plus a PDF from a presentation last month: https://community.arm.com/developer/...e-architecture

It seems that they intent to bring SVE2 into the consumer segment, too: "SVE2 builds on the foundations of SVE to bring the benefits of scalable SIMD vector performance and advanced auto-vectorization capabilities to a wider range of software, including DSP and multimedia SIMD codes that currently use Neon. It also adds many new features to further expand the use of SIMD vector hardware and increase the amount of fine-grain, data-level parallelism in programs."

[wizard69]

This makes me wonder who will be first to market with SVE2? I’m thinking if the rumors about an ARM based Macs are true Apple would be implementing SVE2 in their ARM processors.

[ms178]

Yes, ARM vendors who have their eyes on the notebook or even the desktop market for sure, such as Apple and Qualcomm. But possibly all the others, too. If that market takes off others might join as well. But I doubt that Qualcomm's pricing strategy and the lack of competition from other vendors is of help there, so it might take some years and some iterations to provide any meaningful value for end-users. And we have seen Qualcomm's lack of will in the server space already.

[carewolf]

This makes me wonder who will be first to market with SVE2? I’m thinking if the rumors about an ARM based Macs are true Apple would be implementing SVE2 in their ARM processors.

It could happen, but suspect someone else would have to prove its usefulness over modern NEON. It would probably also require a lot of compiler development to work well, and we will have to see if it provides any clear benefits. Still would be cool, I look forward to seeing what people can do with it.

[Royi]

Is there an analysis of SVE vs AVX?
I mean in the sense how well it was architecture and efficient.

[LinAGKar]

Now, when will AMD get transactional memory?

[wizard69]

It could happen, but suspect someone else would have to prove its usefulness over modern NEON. It would probably also require a lot of compiler development to work well, and we will have to see if it provides any clear benefits. Still would be cool, I look forward to seeing what people can do with it.

Obviously there is little on the market from which we can judge, but the Marketing sure looks good!!! I would suspect that it will in fact work well as Fujitsu was the driving force and they had their eye on high performance systems. There was a consortium of sorts with ARM, Fujitsu and others involved in the design. I have no idea if Apple was involved or not.