5.803: Recompiled kernel, still no encryption acceleration

Via C7 Padlock LogoI thought I’d take a quick look under the hood and see if the latest release has a newer kernel.

While the kernel version itself has not changed, it has been recompiled.

There are no optimizations formally noted on the release page, but we can always hope some technical changes and improvements have been made to optimize the kernel for the Via C7 chipset.

The kernel now reports:

pcor@localhost / $ uname -a
Linux localhost 2.6.17-ck1 #1 PREEMPT Fri Aug 24 23:42:16 EDT 2007 i686 VIA Esther processor 1200MHz CentaurHauls GNU/Linux

The lights are on, is anybody home?

Unfortunately, it does not look like some of the advanced features of the chipset are enabled.

Enabling these would make conversations with secure SSL websites faster as well as accelerating secure SCP file transfers. Both things that happen to be important to me and many early Zonbu adopters.

From a recent comment on the blog about the capability:

  1. mykhal Says:
    August 18th, 2007 at 10:31 am ACE2 states for Advanced Cryptographic Engine 2, PHE is Padlock Hash Engine, and PMM means Padlock Montgomery Multiplier. These are new CPU-integrated cryptographic engines/accelerators introduced in VIA Esther (C7) core, in addition to ACE and RNG (Random Number Generator) in older Nehemiah (C3). There are two possibilities – these functions are not present in the CPU, which is not likely, second one is they are not enabled in kernel. I am pretty sure there is a patch for 2.6.17 kernel enabling these features.

If they built it, enable it

If the capability is available in the chipset, it would be good if it was enabled. According to mykhal the feature could double the speed of encrypted SCP transfers.

I was under the impression that the Zonbu kernel was Via C7 optimized but this calls that theory in to question. This is a shame, I think there is the opportunity for Zonbu to stake a claim in the space.

Can someone from Zonbu comment on what kernel optimizations and profiling has been done for the C7?

-Mr. Zonbu


One Response to 5.803: Recompiled kernel, still no encryption acceleration

  1. The kernel has been indeed really optimized for the C7. There are probably a few optimizations that can be done but we think that we have done 90% of the work.

    As for the encryption, we didn’t include it because we first developed on the C3 which didn’t support what we wanted to get and then this optimization was very low on the todo list. Moreover, the encryption is not the bottle neck as it’s done at the same time as the file is uiploaded – the bottleneck is the bandwidth. It’s correct that a few CPU cycles could be saved if there was an acceleration. We actually prefer to currently work on the video acceleration which will have a higher impact for the user.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: