[Libre-soc-dev] FOSDEM 2024

[Konstantinos Margaritis]

---- On Tue, 05 Dec 2023 05:16:09 +0200 Luke Kenneth Casson Leighton  wrote ---
 > On Monday, December 4, 2023, Konstantinos Margaritis konstantinos at vectorcamp.gr> wrote:

 > > But there is *no* hardware that runs SVP64 instructions currently,
 > 
 > False. TestIssuer has been capable of running the
 > Embedded Compliancy Subset of SVP64 since its inception.
 > (see https://libre-soc.org/openpower/sv/compliancy_levels/)

Ok, could you also point me to a working FPGA implementation so that I can test current SVP64 instructions on?
There is none that works with *CURRENT* SVP64 instructions as they are implemented in the simulator.
Hence the point. Hence the whole effort of wrapping the simulator. 
I'm not saying it won't happen. I'm saying it's not there now, and it wasn't there a year ago when I implemented it.

 > given that you are only familiar with an extremely small
 > part of the full history and development behind the 5 year
 > NLnet-funded Libre-SOC Project, if you are not clear
 > exactly as to why your abstract is misleading and can in no
 > way be approved until resolved please ask questions until
 > you have fully understood.
 > 
 > this is your responsibility.

Spare me the derogatory remarks. I'll let Andrey/James/Jacob explain to you how familiar I am with "an extremely small part" and how important that part is, eg finding critical bugs in sv.bc.

 > it is "linux" as far as the average reader is concerned
 > and therefore when you state quite plainly and very publicly
 > "LibreSOC is incapable of running linux" you are damaging
 > the project's reputation and that of NLnet.

You know, I don't think you actually read the abstract. The whole point is that no-one can currently test the *new/current* SVP64 instructions on hardware (FPGA or otherwise), so they need to use the simulator.
But even if the simulator or FPGA can boot Linux, it's beside the point. I don't even care about booting Linux, I only cared of being able to run particular user-space algorithms using *CURRENT* SVP64 instructions.

I made a final change to the abstract but don't intend to do any more. If you don't ntend to approve the talk, that's fine by me.

Konstantinos