Libre RISC-V Mailing list Archives (accessible via public-inbox)
 help / color / Atom feed
* [libre-riscv-dev] [Bug 155] a PLL is needed for the SoC
From: bugzilla-daemon @ 2020-05-26 20:12 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-155-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=155

--- Comment #15 from Staf Verhaegen <staf@fibraservi.eu> ---
> Input: 24 MHz XTAL

AFAIK there is no crystal oscillator IP available for this project so I assumed
the input clock would be CMOS.

> So, regarding the specifications, the most important point determining the
> overall architecture is : 
> -- Jitter 

For a digital circuit jitter of the clock eats in the timing budget, the 
critical has to be fast enough to have result for the smallest clock cycle due
to jitter.
Analog/mixed-signal blocks will have their requirement on jitter, like
resolution/bandwidth for an ADC or error rate for SERDES.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Luke Kenneth Casson Leighton @ 2020-05-26 20:00 UTC (permalink / raw)
  To: Libre-RISCV General Development
In-Reply-To: <20200526215426.cbe06b2de7a8d2cc0ee6b515@platen-software.de>

On Tue, May 26, 2020 at 8:54 PM Tobias Platen
<libre-soc@platen-software.de> wrote:
>
> this morning:
>         nothing
> this evening:
>         more work on DataMerger.elaborate which is now nearly finished.

great!

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Tobias Platen @ 2020-05-26 19:54 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <CAPweEDybOK=82Z5U2h79Osh3qf7k1LZYV8z8=DVJhBtFOyi5qg@mail.gmail.com>

this morning:
	nothing
this evening:
	more work on DataMerger.elaborate which is now nearly finished.

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Luke Kenneth Casson Leighton @ 2020-05-26 19:51 UTC (permalink / raw)
  To: Libre-RISCV General Development
In-Reply-To: <CAC2bXD6QZp1HHiw2Z+Mxcjy+0R9EWcR1nQoQrNdhWfF1jpM_Qw@mail.gmail.com>

On Tue, May 26, 2020 at 8:13 PM Jacob Lifshay <programmerjake@gmail.com> wrote:
>
> yesterday:
> wrote a program to analyse the values POWER9 produces for instructions
> where the spec doesn't define the results, see
> https://bugs.libre-soc.org/show_bug.cgi?id=324#c5

interesting.  and useful.

> today:
> continuing working on the MUL and DIV pipelines.

great.

> Is Cesar taking over the load/store pipeline?

Tobias is currently doing the data-merger "thing", Cesar is happy to
do a L0CacheBuffer unit test - between all of us we should get
something done.

l.

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Luke Kenneth Casson Leighton @ 2020-05-26 19:46 UTC (permalink / raw)
  To: Libre-RISCV General Development
In-Reply-To: <3F54AFB8-8C2A-48D9-8473-31441A492780@getmailspring.com>

On Tue, May 26, 2020 at 7:00 PM Cole Poirier <colepoirier@gmail.com> wrote:
>
> On May 26 2020, at 10:35 am, Luke Kenneth Casson Leighton
> <lkcl@lkcl.net> wrote:
> > i have a feeling we just call RegFileArray.read_port() and write_port,
> > then not let anyone access the 8 reg-ports, throw away the "global"
> > thing and create the alternative one.
>
> Having done a close read of regfile/virtual_port.py (apologies for
> giving a "terrible name for a complex register"),

it works :)

> and reviewing the
> yosys show top logic diagram I think I mostly understand what's going on
> in this module.

it's basically multiplexing between 8 individual ports and 1 massive
port that hits absolutely all of the (8) ports at once.

which, when put like that, we might as well do exactly that.


> Regarding the "global" thing, are you referring to the self.wr/rd_ports
> "for external use"? We should instead just use the existing methods on
> RegFileArray to set up the external ports? How do we go about not
> letting anyone access the 8 reg-ports?

that's the Dep-Matrices job.  or, more to the point: the PriorityPickers.

l.

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Jacob Lifshay @ 2020-05-26 19:13 UTC (permalink / raw)
  To: Libre-RISCV General Development
In-Reply-To: <CAPweEDybOK=82Z5U2h79Osh3qf7k1LZYV8z8=DVJhBtFOyi5qg@mail.gmail.com>

yesterday:
wrote a program to analyse the values POWER9 produces for instructions
where the spec doesn't define the results, see
https://bugs.libre-soc.org/show_bug.cgi?id=324#c5

today:
continuing working on the MUL and DIV pipelines.

Is Cesar taking over the load/store pipeline? I never got much done on it.

Jacob
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Cole Poirier @ 2020-05-26 18:00 UTC (permalink / raw)
  To: Libre-RISCV General Development; +Cc: Libre-RISCV General Development
In-Reply-To: <CAPweEDyiaZJgHt5Ow1prbNMQs_Tk3TijiS6oiNDhqBDGH+tjxA@mail.gmail.com>

On May 26 2020, at 10:35 am, Luke Kenneth Casson Leighton
<lkcl@lkcl.net> wrote: 
> i have a feeling we just call RegFileArray.read_port() and write_port,
> then not let anyone access the 8 reg-ports, throw away the "global"
> thing and create the alternative one.

Having done a close read of regfile/virtual_port.py (apologies for
giving a "terrible name for a complex register"), and reviewing the
yosys show top logic diagram I think I mostly understand what's going on
in this module.

Regarding the "global" thing, are you referring to the self.wr/rd_ports
"for external use"? We should instead just use the existing methods on
RegFileArray to set up the external ports? How do we go about not
letting anyone access the 8 reg-ports?

Cole

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Michael Nolan @ 2020-05-26 17:57 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <CAPweEDzmRWvWt_hY+JUMoCGgcYUEH7N7eVC5H3vrZH8s6Sb_0g@mail.gmail.com>

On 5/26/20 1:49 PM, Luke Kenneth Casson Leighton wrote:
> michael, my partner, marie, who is from alaska, pointed out that the
> work that you're doing on Libre-SOC should, if you ask the professors,
> be considered coursework towards your degree.  worth considering?
> l.
>
Thank you, yes I am considering it. This fall I'm taking "Senior Project 
Proposal" (or something like that), and I'm planning on asking if I can 
possibly use this as my senior project.

--Michael


_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 342] formal proof of soc.fu.compunits.FunctionUnitBaseSingle needed
From: bugzilla-daemon @ 2020-05-26 17:52 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-342-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=342

--- Comment #8 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Michael Nolan from comment #7)
> with m.If(Past(wr_rel) & Past(go_wr)):
>     # the alu data is output
>     comb += Assert(dut.data_o == alu_temp)
>     # wr_rel is dropped
>     comb += Assert(wr_rel == 0)
>     # busy is dropped.
>     with m.If(~Past(go_die)):
> 	comb += Assert(busy == 0)
> 
> 
> Shouldn't the module drop busy after go_die is asserted? If I delete the
> m.If(~Past(go_die)) the proof fails.

annoyingly, it's one cycle after, because of the syncs setting the latches.

fascinatingly, in practice, we get away with it.  the code is complex enough
that i am not sure why.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Luke Kenneth Casson Leighton @ 2020-05-26 17:49 UTC (permalink / raw)
  To: Libre-RISCV General Development
In-Reply-To: <1ce42875-51d5-dd78-0fb1-e6d86363fbb5@gmail.com>

michael, my partner, marie, who is from alaska, pointed out that the
work that you're doing on Libre-SOC should, if you ask the professors,
be considered coursework towards your degree.  worth considering?
l.

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 342] formal proof of soc.fu.compunits.FunctionUnitBaseSingle needed
From: bugzilla-daemon @ 2020-05-26 17:35 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-342-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=342

--- Comment #7 from Michael Nolan <mtnolan2640@gmail.com> ---
with m.If(Past(wr_rel) & Past(go_wr)):
    # the alu data is output
    comb += Assert(dut.data_o == alu_temp)
    # wr_rel is dropped
    comb += Assert(wr_rel == 0)
    # busy is dropped.
    with m.If(~Past(go_die)):
        comb += Assert(busy == 0)


Shouldn't the module drop busy after go_die is asserted? If I delete the
m.If(~Past(go_die)) the proof fails.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Luke Kenneth Casson Leighton @ 2020-05-26 17:35 UTC (permalink / raw)
  To: Libre-RISCV General Development
In-Reply-To: <6D336165-6B38-4C1C-B0CB-50E9705C0C57@getmailspring.com>

On Tue, May 26, 2020 at 6:29 PM Cole Poirier <colepoirier@gmail.com> wrote:

> Great. Going to spend another half hour studying virtual_port then will
> start on this.

i have a feeling we just call RegFileArray.read_port() and write_port,
then not let anyone access the 8 reg-ports, throw away the "global"
thing and create the alternative one.

l.

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Cole Poirier @ 2020-05-26 17:29 UTC (permalink / raw)
  To: Libre-RISCV General Development; +Cc: Libre-RISCV General Development
In-Reply-To: <CAPweEDw8eXmzGCxDP2KqFeC3FQRvqX_05AEWK--EFpphWUmXrw@mail.gmail.com>

On May 26 2020, at 10:24 am, Luke Kenneth Casson Leighton
<lkcl@lkcl.net> wrote:
> On Tue, May 26, 2020 at 6:02 PM Cole Poirier <colepoirier@gmail.com> wrote:
>> 
>> Yesterday:
>> 
>> Did a very small amount of work on the virtual port with luke, I believe
>> he's taken this over, which is good because I was struggling
> 
> so am i.  sigh.
> 
>> and this is important.
> 
> please do keep an eye on it, keep asking questions ok?

Will do. It's very, very helpful to have tried *some* kind of
implementation on my own, because now I am studying yours and filling in
the myriad missing things that I was searching my brain for but didn't
yet understand.
 
>> Today:
>> 
>> Yesterday Luke asked me to raise a bug report for RegFile and
>> RegFileArray proofs needed, so today I plan on trying to formally prove
>> regfile. Input or advice on this at bug 353 is appreciated.
>> (https://bugs.libre-soc.org/show_bug.cgi?id=353)
> 
> start from the proof_bperm.py or proof_fu.py as the basic template.
> we can work it out.

Great. Going to spend another half hour studying virtual_port then will
start on this.

Cole

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Luke Kenneth Casson Leighton @ 2020-05-26 17:24 UTC (permalink / raw)
  To: Libre-RISCV General Development
In-Reply-To: <3C675955-D169-49EB-9435-6B9D9317A36F@getmailspring.com>

---
crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68

On Tue, May 26, 2020 at 6:02 PM Cole Poirier <colepoirier@gmail.com> wrote:
>
> Yesterday:
>
> Did a very small amount of work on the virtual port with luke, I believe
> he's taken this over, which is good because I was struggling

so am i.  sigh.

> and this is important.

please do keep an eye on it, keep asking questions ok?

> Today:
>
> Yesterday Luke asked me to raise a bug report for RegFile and
> RegFileArray proofs needed, so today I plan on trying to formally prove
> regfile. Input or advice on this at bug 353 is appreciated.
> (https://bugs.libre-soc.org/show_bug.cgi?id=353)

start from the proof_bperm.py or proof_fu.py as the basic template.
we can work it out.

l.

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 336] ALU CompUnit needs to recognise that RA (src1) can be zero
From: bugzilla-daemon @ 2020-05-26 17:22 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-336-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=336

--- Comment #39 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Cesar Strauss from comment #38)

> > in particular the ALU may *not* necessarily be ready to proceed at the time
> > operands are available.
> 
> Feel free to put this on my queue, if you wish (formal proof of the
> CompUnit-ALU interface).

sure, go for it.

> While I have no practical experience on formal proofs, I've been reading
> about it for some time, with the goal of applying it to my own designs. I
> like to think I've reached a good understanding.

good place to learn.

i would suggest simply adding them to proof_fu.py - focus on ready_(i/o),
valid_(i/o) - these are accessible signals.

if you assume that the DummyALU takes exactly one cycle between data_i and
data_o, this may make life easier.  the sequence will go:

* p.valid_i is set HI by CU
* p.ready_o is asserted HI by ALU in response (takes 1 cycle and asserts for
  ONLY one cycle)
* n.valid_o is asserted HI by ALU
* n.ready_i is set HI by CU
* n.valid_o is asserted LO by ALU in response (takes 1 cycle)

regarding data:

* data input should be placed by CU on the input at the same time as it
  raises p.valid_i.  it should REMAIN valid until p.ready_o is seen asserted.

* data output should be ready and valid when n_ready_i is HI.
all other times, it *may* still be left on the bus, but should *NOT* be
taken to be valid.

this is basically a standard ready/valid signalling system used by wishbone,
AXI4, etc. etc. etc. etc.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Luke Kenneth Casson Leighton @ 2020-05-26 17:08 UTC (permalink / raw)
  To: Libre-RISCV General Development
In-Reply-To: <1ce42875-51d5-dd78-0fb1-e6d86363fbb5@gmail.com>

On Tue, May 26, 2020 at 5:47 PM Michael Nolan <mtnolan2640@gmail.com> wrote:
>
> Yesterday, I worked on familiarizing myself with the MultiCompUnit, and
> began writing a formal proof for it. Today, I'd like to refactor that
> proof to better match up with the description on the wiki page.

that's quite funny, because i am refactoring the wiki page to help
make it easier to write the proof :)

> Also, summer classes began today for me, so I may be less productive on
> Libre-SOC for the next two months.

ok cool, thanks for letting us know.  most of the pipelines has been
done, it should be plain sailing from here.

l.

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Luke Kenneth Casson Leighton @ 2020-05-26 17:07 UTC (permalink / raw)
  To: Libre-RISCV General Development
In-Reply-To: <125e37a2-e549-83d6-c987-db4d316631ed@gmail.com>

On Tue, May 26, 2020 at 5:44 PM Cesar Strauss <cestrauss@gmail.com> wrote:

> * Found out about ongoing formal proof work on the very module to which
> I'm adding unit tests. Interesting subject, formal proofs. I've been
> learning it myself, in order to apply it to my work, but have as yet no
> practical experience.

samuel falvo was the first one to explain the benefits of formal proofs.
http://chiselapp.com/user/kc5tja/repository/kestrel-3/artifact/96ea88450c1a3fc6

the insight that he gave particularly hit home when we had to run
literally hundreds of thousands of tests (which takes about a week to
run) for the IEEE754 FPU...

... and yet, despite that, we *still have no guarantee of hitting all
corner-cases*.

so we have hundreds of thousands of tests - with no guarantee - or, we
write *ONE* formal proof that covers them all.

it's not a difficult choice to make.

> Today:
>
> * Advance on the parallel tests. Hopefully, get good progress on the
> rd/wr protocol.

one step at a time.

l.

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Cole Poirier @ 2020-05-26 17:02 UTC (permalink / raw)
  To: Libre-RISCV General Development; +Cc: libre-riscv-dev
In-Reply-To: <1ce42875-51d5-dd78-0fb1-e6d86363fbb5@gmail.com>

Yesterday:

Did a very small amount of work on the virtual port with luke, I believe
he's taken this over, which is good because I was struggling and this is important.

Today:

Yesterday Luke asked me to raise a bug report for RegFile and
RegFileArray proofs needed, so today I plan on trying to formally prove
regfile. Input or advice on this at bug 353 is appreciated.
(https://bugs.libre-soc.org/show_bug.cgi?id=353)

Cole

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Michael Nolan @ 2020-05-26 16:47 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <CAPweEDybOK=82Z5U2h79Osh3qf7k1LZYV8z8=DVJhBtFOyi5qg@mail.gmail.com>

Yesterday, I worked on familiarizing myself with the MultiCompUnit, and 
began writing a formal proof for it. Today, I'd like to refactor that 
proof to better match up with the description on the wiki page.

Also, summer classes began today for me, so I may be less productive on 
Libre-SOC for the next two months.

--Michael


_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 26may2020
From: Cesar Strauss @ 2020-05-26 16:44 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <CAPweEDybOK=82Z5U2h79Osh3qf7k1LZYV8z8=DVJhBtFOyi5qg@mail.gmail.com>

Yesterday:

* Started with the parallel simulation for MultiCompUnit. In the
process, gained a good understanding of parallel simulation in nMigem,
as well as the issue_i/busy_o protocol.

This morning:

* Found out about ongoing formal proof work on the very module to which
I'm adding unit tests. Interesting subject, formal proofs. I've been
learning it myself, in order to apply it to my work, but have as yet no
practical experience.

Today:

* Advance on the parallel tests. Hopefully, get good progress on the
rd/wr protocol.

Regards,

Cesar

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 351] create a "block" (mass) regfile port (read and write) onto an array-based regfile
From: bugzilla-daemon @ 2020-05-26 16:41 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-351-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=351

--- Comment #11 from Cole Poirier <colepoirier@gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #10)
> https://git.libre-soc.org/?p=soc.git;a=blob;f=src/soc/regfile/virtual_port.
> py;hb=HEAD
> 
> i think i have it.  i had to modify RegFileArray to first get it to give me
> only (an array of) read/write ports to each (small) register.
> 
> however... *this is not the full story*!
> 
> we need to prevent and prohibit RegFileArray.read_port() and
> RegFileArray.write_port() from having a "bypass" on the registers
> 
> this is complicated.

I take it you've got this bug under control and I should move on to something
else today? :)

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 336] ALU CompUnit needs to recognise that RA (src1) can be zero
From: bugzilla-daemon @ 2020-05-26 16:29 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-336-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=336

--- Comment #38 from Cesar Strauss <cestrauss@gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #37)
> (In reply to Cesar Strauss from comment #36)
> > How about a formal proof of the ALU interface?
> 
> the interaction with it will be needed, yes, because of the signalling
> ready/valid.
> 
> in particular the ALU may *not* necessarily be ready to proceed at the time
> operands are available.

Feel free to put this on my queue, if you wish (formal proof of the
CompUnit-ALU interface).

While I have no practical experience on formal proofs, I've been reading about
it for some time, with the goal of applying it to my own designs. I like to
think I've reached a good understanding.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 351] create a "block" (mass) regfile port (read and write) onto an array-based regfile
From: bugzilla-daemon @ 2020-05-26 16:01 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-351-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=351

--- Comment #10 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
https://git.libre-soc.org/?p=soc.git;a=blob;f=src/soc/regfile/virtual_port.py;hb=HEAD

i think i have it.  i had to modify RegFileArray to first get it to give me
only (an array of) read/write ports to each (small) register.

however... *this is not the full story*!

we need to prevent and prohibit RegFileArray.read_port() and
RegFileArray.write_port() from having a "bypass" on the registers

this is complicated.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] daily kan-ban update 26may2020
From: Luke Kenneth Casson Leighton @ 2020-05-26 14:40 UTC (permalink / raw)
  To: Libre-RISCV General Development

yesterday:

* reviewed Michael's formal proof for MultiCompUnit
* helped Cesar with the parallel-simulation for MultiCompUnit
* started (with Cole) a "virtual port" on a register file
* reviewed Tobias's DataMerger data structure and added notes

this morning:

* spoke to tropicsquare (like their idea of doing a secure chip)
* moved the treereduce function from regfile.py into nmutil.utils and
  made it a map-reduce function.
* tried implementing the "virtual port" regfile only to find it had
 partially been done in RegFileArray (but not quite).

today:

* continue with the regfile thing (we really need it)
* keep an eye on list and bugtracker

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 351] create a "block" (mass) regfile port (read and write) onto an array-based regfile
From: bugzilla-daemon @ 2020-05-26 12:54 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-351-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=351

--- Comment #9 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
ok i am more awake.  started on adding the virtual port aaaand... spangg, i
ran smack into a pre-existing very close approximation of the virtual port,
in the form of RegFileArray.

*face-palm*.

the way that RegFileArray works is: it creates an array of registers,
then when read_port (or write_port) is called, an *array* of individual
read/write ports - one for each individual register - is returned, *and*
a "global" access port is provided.

that global access port is a "broadcast bus" to every register, and its
width is the *exact* same width as the register width.

also, where each register has a 1-bit-wide ren/wen, the global port is
the *accumulation* of all those signals (using self._get_en_sig) in a
unary sequence.

thus, if you pass in say a global.ren="0b00001011", this will read out the OR
of registers numbered 0, 1 *and* 3 (!!) as in:

    result = reg[0] ORed with reg[1] ORed with reg[3]

this will be a regwidth-wide result

what we want, for the virtual port is:

    result = reg[0] CATted with reg[1] CATted with reg[2] CATed with reg[3]...

this will be a regwidth TIMES n_regs wide result

which is very different.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

page:   latest

Libre RISC-V Mailing list Archives (accessible via public-inbox)

Archives are clonable: git clone --mirror https://inbox.libre-soc.org/libre-riscv-dev

Example config snippet for mirrors


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git