* [libre-riscv-dev] Setup automation scripts
From: Cole Poirier @ 2020-03-25 16:04 UTC (permalink / raw)
To: Libre-RISCV General Development
Hi Libre-SOC team,
I'm currently working on turning the setup Jacob and I did on his build
server into an easily replicable bash setup script. While working on
this yesterday, I concurrently tried to follow the hdl workflow, a
substantial part of which is setting up a debootstrapped chroot and
installing coriolis2. On this attempt I was finally able to complete it
(thanks to Luke and Jock for updating the process for Debian 10 and
working out a few other kinks in the process!), but given the length and
complexity of this process, it strikes me as another candidate for
automation. I was thinking I could repeat the process I'm going through
with Jacob, working on the build server setup automation, with the HDL
workflow. Jacob has advised me to set up a local debian vm for the
purposes of ensuring the script's reliability, so it is my thiniking
that it will be worthwhile to work on these two process automations concurrently.
Do you think that this is would be useful and worthwhile? Also, given
that we already have a hard dependency on python, are bash scripts
preferable to the python equivalent of these automation scripts?
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
* [libre-riscv-dev] nmutil failing tests due to attribute errors
From: Cole Poirier @ 2020-03-25 16:00 UTC (permalink / raw)
To: libre-riscv-dev
Hi Libre-SOC team,
I think I completed HDL workflow sucessfully yesterday, however when I
try to run "python3 setup.py test" in nmutil I get 33 errors.
Adhering to the rule of not pasting large amounts of autogenerated
output, I will only include two examples in this first email, but I can
provide more details should you need them to help me. Can you please
advise me as to what I've done wrong or misconfigured?
###
ERROR: nmutil.test.test_outmux_pipe.test1
AttributeError: 'Queue' object has no attribute 'we'
ERROR: nmutil.test.test_inout_unary_mux_cancel_pipe.test1
AttributeError: 'TestInOutPipe' object has no attribute 'domains'
###
Thanks,
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] cache SRAM organisation
From: Luke Kenneth Casson Leighton @ 2020-03-25 15:53 UTC (permalink / raw)
To: Libre-RISCV General Development
In-Reply-To: <6fa40cb78b3f8c013ca4953ccb4daa5c23e3b501.camel@fibraservi.eu>
On Wed, Mar 25, 2020 at 1:46 PM Staf Verhaegen <staf@fibraservi.eu> wrote:
> Keep the discussion on development of other types til after the prototype tape-out in October.
good idea.
> I know there is other development ongoing for dual port RAM in the OpenRAM community so likely it is more clear what extra still has to happen.
> BTW, I think there also is possibility to have register files block with 2R1W RAM blocks.
ah that's good. we can combine two of them to create 4R1W (we need
minimum 3R1W, however the design does not critically depend on this:
it just takes 2 cycles to read all operands for 3-operand instructions
such as FMAC).
> > this because it turns out that asynchronous SRAM can act, when used in aRegister File, as if it was a (separate) Register Bypass / ForwardingPort. with the Out-of-Order Engine being a huge cyclic feedback loopbetween ALUs and RegFile, clock delays are an impediment, and havingcompletely separate (extra) Regfile Bypass ports dramatically increases thenumber of wires and Multiplexers.
>
> Could detail more on how the adress, data and output signals of this asynchronous block would be used and switched between synchronous and asynchronous functioning. To me it seems that it would just place of the multiplexers, not the amount.
ok i will try to outine it, there is quite a lot of detail, i apologise.
it's basically the "pass-through" system used in nmigen FIFOs, and the
"synchronous" mode of nmigen Memory. the requirements are: what is
written has to be available *combinatorially* - i.e. on the same clock
cycle - if simultaneously read via another port.
now, yes i took note that this is not supposed to be permitted: you're
not normally permitted to be able to read *and* write to an SRAM cell
at the same time. however, that's exactly what we need.
a workaround (fallback position) is, we use DFF latches. i created a
"bypass latch" function which creates DFF latches with such a
combinatorial bypass: we actually use them quite a lot (including
between pipeline stages so that we can programmatically cut the number
of pipeline stages in half at the flick of a switch).
however for the Register File we would not "switch" between
synchronous / asynchronous mode. the reason why we need the
synchronous mode is because some Function Units will be sitting idle,
waiting for their input operands, which have to come from other
Function Units as "results".
these "results" of course have to go into the Register File, however
in a large number of cases, there are FUs waiting for a "read" of that
exact register - the one that's just been written - and to make that
FU wait *another* clock cycle is just wasting time.
btw, also: the synchronous "Read" mode also allows us to transfer data
from the Regfile into Function Unit input latches on that same cycle,
rather than (again) waiting yet another cycle, unnecessarily. in some
cases, where a pipeline has an immediate free slot, we can actually
proceed directly to executing stage 1 of the pipeline, *in* the same
cycle as the actual Regfile Read.
the other aspect of the Dependency Matrices is: each register has
already been given a "single line" (one "enable" bit per Reg #). the
binary register number is *gone*.
if we use a "standard" binary-addressable Register File, we actually
have to *recreate* that binary register number / address, using
unary-to-binary converters!
so my point is: if we have all the register numbers *already* encoded
in unary (one "enable" bit per reg #), we actually don't *want* a
binary-addressing mux-map on the Regfile, at all!
we actually just want a Priority-Picker, directly onto the SRAM port
write/read enable lines, where each SRAM row has *no* binary-address
decoder on it.
i hope that was enough to explain what is going on? if not please do
ask to clarify, when you have 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] cache SRAM organisation
From: Staf Verhaegen @ 2020-03-25 13:46 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <CAPweEDwfqMczPjg=5Fvt1J_S8nx1YK44XhyBY8H1abuTNF6=xg@mail.gmail.com>
[-- Attachment #1.1: Type: text/plain, Size: 2093 bytes --]
Luke Kenneth Casson Leighton schreef op wo 25-03-2020 om 12:33 [+0000]:
> On Wednesday, March 25, 2020, Staf Verhaegen <staf@fibraservi.eu> wrote:
> > A single port SRAM has one port where you can do a read or a write eachclock cycle. The 2-port one has one read port and one write port so you cando a read and write each clock cycle. The dual port one now has two portsthat each can do a read or write each clock cycle. So you can do two reads,two write or a read+write each clock cycle.For each of them you can have a synchronous or an asynchronous version. Asynchronous RAM has a clock input and the address and data inputs arelatched on that clock signal. It thus means that the FFs are integrated inthe SRAM, e.g. thus very close :) . The RAM currently being developed in myNLNet project is a synchronous SRAM as this is easier from timing point ofview because all the timing can be related to the clock. A synchronous RAMactually functions as an addressable bunch of FFs and the synthesis and P&Rtools know how to handle them.
>
> ok. later, we will definitely need an aysynchronous version.
Keep the discussion on development of other types til after the prototype tape-out in October. I know there is other development ongoing for dual port RAM in the OpenRAM community so likely it is more clear what extra still has to happen.
BTW, I think there also is possibility to have register files block with 2R1W RAM blocks.
> this because it turns out that asynchronous SRAM can act, when used in aRegister File, as if it was a (separate) Register Bypass / ForwardingPort. with the Out-of-Order Engine being a huge cyclic feedback loopbetween ALUs and RegFile, clock delays are an impediment, and havingcompletely separate (extra) Regfile Bypass ports dramatically increases thenumber of wires and Multiplexers.
Could detail more on how the adress, data and output signals of this asynchronous block would be used and switched between synchronous and asynchronous functioning. To me it seems that it would just place of the multiplexers, not the amount.
greets,
Staf.
[-- Attachment #2: Type: text/plain, Size: 177 bytes --]
_______________________________________________
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 257] Implement demo Load/Store queueing algorithm
From: bugzilla-daemon @ 2020-03-25 13:07 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-257-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=257
--- Comment #17 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
https://libre-riscv.org/3d_gpu/l0_to_l1_n_way_cache_buffer.png
i will add this and a description here, shortly:
https://libre-riscv.org/3d_gpu/architecture/6600scoreboard/
the idea here is to have 6 to 8 LD/ST Function Units, each with
two "ports" (one exclusively for misaligned LDs/STs), which means
that we then have 12 to 16 such "ports" trying to get access to
the L0 Cache/Buffer.
that will require a Multi-Priority-Picker ("and here's one we prepared
earlier....")
https://libre-riscv.org/3d_gpu/architecture/6600scoreboard/800x-multi_priority_picker.png
this can be set up to be 16-in and produce 4x unique exclusive 16-out
"selectors", none of which will overlap.
that in turn allows us to OR/MUX any one of those 16 LD/ST FUs to
any one of the 4 "ports" on the L0 Cache
from there, we can split into two halves using bit 4 as the "selector"
bit, and use bits 5 and above as the "Addr hit".
a priority picker then picks *one* row and broadcasts its address to
all other rows. if there is a "hit" (same address bits 5 and upwards)
then the data-select bytemask - and data itself - is "merged" into
the output as a "single cache line", to be sent directly to the
L1 cache.
whew.
--
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 186] Create decoder for SOC: Power ISA and RISC-V
From: bugzilla-daemon @ 2020-03-25 12:49 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-186-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=186
--- Comment #102 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Luke Kenneth Casson Leighton from comment #98)
> https://git.libre-riscv.org/?p=soc.git;a=blob;f=src/soc/simulator/test_sim.
> py;hb=HEAD
>
> y'know... the comment "checked this against qemu" got me thinking. this
> looks like it could be used to actually run any of:
>
> * qemu
what do you think, michael: does firing off qemu to run those same
(tiny) binaries sound practical?
--
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] cache SRAM organisation
From: Luke Kenneth Casson Leighton @ 2020-03-25 12:33 UTC (permalink / raw)
To: Libre-RISCV General Development
In-Reply-To: <29b1a9ecedda151dc9c8da6516c3691dfede62ef.camel@fibraservi.eu>
On Wednesday, March 25, 2020, Staf Verhaegen <staf@fibraservi.eu> wrote:
> Libre-SOC developers,
>
> That discussion is mainly on system level and I don't want to get too deep
> into this as I don't have time for that.
>
implicitly understood, it's why i linked to the (two) relevant discussions
in which Mitch mentioned Cache SRAM design, and summarised.
if you are busy, no need to reply below, it is mostly "informative" for
other readers.
> I am providing the SRAM blocks and then it is up to the system guys to see
> how they use them. In this case you guys (libre-soc + LIP6) are the system
> guys.
>
appreciated and accepted.
> On ASICs commonly three types of SRAM are provided: a single port RAM, a
> 2-port RAM and a dual port RAM. Currently for NLNet only a single port SRAM
> is foreseen as this is the most common, the smallest in area per bit and
> the fastest.
>
yes. i deduced that from Mitch's description: appreciate the summary.
> A single port SRAM has one port where you can do a read or a write each
> clock cycle. The 2-port one has one read port and one write port so you can
> do a read and write each clock cycle. The dual port one now has two ports
> that each can do a read or write each clock cycle. So you can do two reads,
> two write or a read+write each clock cycle.
> For each of them you can have a synchronous or an asynchronous version. A
> synchronous RAM has a clock input and the address and data inputs are
> latched on that clock signal. It thus means that the FFs are integrated in
> the SRAM, e.g. thus very close :) . The RAM currently being developed in my
> NLNet project is a synchronous SRAM as this is easier from timing point of
> view because all the timing can be related to the clock. A synchronous RAM
> actually functions as an addressable bunch of FFs and the synthesis and P&R
> tools know how to handle them.
>
ok. later, we will definitely need an aysynchronous version.
this because it turns out that asynchronous SRAM can act, when used in a
Register File, as if it was a (separate) Register Bypass / Forwarding
Port. with the Out-of-Order Engine being a huge cyclic feedback loop
between ALUs and RegFile, clock delays are an impediment, and having
completely separate (extra) Regfile Bypass ports dramatically increases the
number of wires and Multiplexers.
(actually it is highly likely that we will need to do the Register File as
FFs, initially, because we need i think a minimum 3R1W, plus the
forwarding. also we *may* do unary addressing - not binary addressing -
which i understand it is pointless to use SRAM cells when you have unary
individual row-enable, already).
> Given this building block you can now make blocks that look to the outside
> world as higher number port blocks. You do this by instantiating multiple
> RAM blocks and make sure that the content is mirrored between all the
> blocks. This way you can read from the different blocks in parallel.
> Writing in the blocks still has to happen to all the blocks at the same
> time.
>
i've heard of this trick being used in FPGAs, usually combining pairs of
2R1W to get 4R1W.
> So if you take four single port SRAM blocks you can make a four port SRAM
> block. Each cycle you can do 1-4 reads or 1 write but you can't read and
> write at the same time. With four 2-port RAMs you can do 4 reads and 1
> write each clock cycle. With four dual port RAMs you can do 4 reads or 3
> reads + 1 write or 2 reads + 2 writes each cycle.
>
ok i always wondered how you'd get 4R2W.
ok so actually we *could* do 3R1W or 4R1W with the single-port SRAM blocks,
for the Regfile.
I will provide the single block, the combining of the block has to happen
> in RTL/HDL. For Libre-SOC this means in nmigen and using Coriolis for
> placement and connecting the single blocks.
>
ok. one for Jean-Paul to help advise with, when we get to it.
> Although the SRAM does an operation each clock cycle the clock frequency
> could be different from the rest of the logic. If the RAM is fast enough it
> could run at double the frequency of the core so basically a single port
> RAM could look like a dual port RAM to the rest of the logic which is
> running at half the frequency. If the RAM is not fast enough wait states
> need to be implemented for each operation. The maximum clock frequency will
> go down when you increase the size of a RAM block. So on CPU typically L1
> cache runs at the same clock frequency as the core without any wait states
> and higher level caches are bigger but also introduce more wait states for
> accessing them.
>
hmmm...
> If you are thinking about having different clock frequencies in your
> design you have to first discuss this with Jean-Paul/LIP6 as doing multi
> clock designs is opening it's own can of worms (cross clock domain problems
> etc).
yes. no, we're not crossing that chasm in the first chip. i know nmigen
can do it: we *may* be forced to do it for some peripherals, and i know
nmigen supports clock domains, however the main core? no, we reaaally want
to avoid it.
For the October prototype I feel we need to stick with use of single port
> SRAM block and run the whole chip from the same clock. IMO, on this
> prototype you should take any performance implication this has.
>
yes. we have enough to deal with. the important thing for the 180nm ASIC
is to cover as much ground as is practical, and prove several aspects of
the design without taking on too much.
thanks for clarifying, Staf - we'll adapt accordingly. will let you get on.
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] cache SRAM organisation
From: Staf Verhaegen @ 2020-03-25 10:54 UTC (permalink / raw)
To: Libre RISC-V dev list
In-Reply-To: <CAPweEDx5QCCKxSr1gfuyuw_2D68Ld8fK85bEmmMTZi8S3w2E9g@mail.gmail.com>
[-- Attachment #1.1: Type: text/plain, Size: 3924 bytes --]
Libre-SOC developers,
That discussion is mainly on system level and I don't want to get too deep into this as I don't have time for that.
I am providing the SRAM blocks and then it is up to the system guys to see how they use them. In this case you guys (libre-soc + LIP6) are the system guys.
On ASICs commonly three types of SRAM are provided: a single port RAM, a 2-port RAM and a dual port RAM. Currently for NLNet only a single port SRAM is foreseen as this is the most common, the smallest in area per bit and the fastest.
A single port SRAM has one port where you can do a read or a write each clock cycle. The 2-port one has one read port and one write port so you can do a read and write each clock cycle. The dual port one now has two ports that each can do a read or write each clock cycle. So you can do two reads, two write or a read+write each clock cycle.
For each of them you can have a synchronous or an asynchronous version. A synchronous RAM has a clock input and the address and data inputs are latched on that clock signal. It thus means that the FFs are integrated in the SRAM, e.g. thus very close :) . The RAM currently being developed in my NLNet project is a synchronous SRAM as this is easier from timing point of view because all the timing can be related to the clock. A synchronous RAM actually functions as an addressable bunch of FFs and the synthesis and P&R tools know how to handle them.
Given this building block you can now make blocks that look to the outside world as higher number port blocks. You do this by instantiating multiple RAM blocks and make sure that the content is mirrored between all the blocks. This way you can read from the different blocks in parallel. Writing in the blocks still has to happen to all the blocks at the same time.
So if you take four single port SRAM blocks you can make a four port SRAM block. Each cycle you can do 1-4 reads or 1 write but you can't read and write at the same time. With four 2-port RAMs you can do 4 reads and 1 write each clock cycle. With four dual port RAMs you can do 4 reads or 3 reads + 1 write or 2 reads + 2 writes each cycle.
I will provide the single block, the combining of the block has to happen in RTL/HDL. For Libre-SOC this means in nmigen and using Coriolis for placement and connecting the single blocks.
Although the SRAM does an operation each clock cycle the clock frequency could be different from the rest of the logic. If the RAM is fast enough it could run at double the frequency of the core so basically a single port RAM could look like a dual port RAM to the rest of the logic which is running at half the frequency. If the RAM is not fast enough wait states need to be implemented for each operation. The maximum clock frequency will go down when you increase the size of a RAM block. So on CPU typically L1 cache runs at the same clock frequency as the core without any wait states and higher level caches are bigger but also introduce more wait states for accessing them.
If you are thinking about having different clock frequencies in your design you have to first discuss this with Jean-Paul/LIP6 as doing multi clock designs is opening it's own can of worms (cross clock domain problems etc). For the October prototype I feel we need to stick with use of single port SRAM block and run the whole chip from the same clock. IMO, on this prototype you should take any performance implication this has.
greets,
Staf.
Luke Kenneth Casson Leighton schreef op di 24-03-2020 om 22:32 [+0000]:
> https://groups.google.com/d/msg/comp.arch/cbGAlcCjiZE/mgMZVINVIAAJ
> Staf can i ask you the favour of reviewing Mitch's comments about cache design?
> in particular the comments about the possibility of using multiported SRAM cells as long as only 1R or 1W is done on any given cell?
> also something about doing the FFs yourself, close to the SRAM cells?
> l.
>
>
[-- Attachment #2: Type: text/plain, Size: 177 bytes --]
_______________________________________________
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] [OP-CD] OpenPOWER Foundation "Virtual Coffee" Calls
From: Cole Poirier @ 2020-03-24 22:35 UTC (permalink / raw)
To: Libre-RISCV General Development; +Cc: Libre-RISCV General Development
In-Reply-To: <CAPweEDzept9r4uyk3217fB-=82_1snLhribdS4VYg8w9_71ALg@mail.gmail.com>
On Mar 24 2020, at 3:28 pm, Luke Kenneth Casson Leighton <lkcl@lkcl.net> wrote:
> On Tuesday, March 24, 2020, Jacob Lifshay <programmerjake@gmail.com> wrote:
>
>> Maybe we should participate.
>
>
> yes good idea. what time is it?
I believe it's at 0900 UTC and 2200 UTC every thursday, lasting between
30-45 minutes. Which one are the two of you planning on calling in for?
I'd like to call in and listen to the call you're on, if possible.
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] Git repository access
From: Luke Kenneth Casson Leighton @ 2020-03-24 22:33 UTC (permalink / raw)
To: Libre-RISCV General Development
In-Reply-To: <05D281B4-1193-4738-96A2-38AB78914CFB@getmailspring.com>
On Tuesday, March 24, 2020, Cole Poirier <colepoirier@gmail.com> wrote:
>
>
> Thanks Luke, the clone command you provided works perfectly for me! I
> tried something similar to this earlier, but I must have gotten
> something subtle incorrect.
i'm stunned i guessed it correctly this time :)
--
---
crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68
_______________________________________________
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] cache SRAM organisation
From: Luke Kenneth Casson Leighton @ 2020-03-24 22:32 UTC (permalink / raw)
To: Staf Verhaegen, Libre-RISCV General Development
https://groups.google.com/d/msg/comp.arch/cbGAlcCjiZE/mgMZVINVIAAJ
Staf can i ask you the favour of reviewing Mitch's comments about cache
design?
in particular the comments about the possibility of using multiported SRAM
cells as long as only 1R or 1W is done on any given cell?
also something about doing the FFs yourself, close to the SRAM cells?
l.
--
---
crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68
_______________________________________________
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] [OP-CD] OpenPOWER Foundation "Virtual Coffee" Calls
From: Luke Kenneth Casson Leighton @ 2020-03-24 22:28 UTC (permalink / raw)
To: Libre-RISCV General Development
In-Reply-To: <CAC2bXD49oPkpdhGFAUt9Dp+m-=vj=Gk1UZ-Qgq3m44gtpqE6rQ@mail.gmail.com>
On Tuesday, March 24, 2020, Jacob Lifshay <programmerjake@gmail.com> wrote:
> Maybe we should participate.
yes good idea. what time is it?
> ---------- Forwarded message ---------
> From: Hugh Blemings <hugh@openpowerfoundation.org>
> Date: Mon, Mar 23, 2020, 19:03
> Subject: [OpenPOWER-HDL-Cores] OpenPOWER Foundation "Virtual Coffee" Calls
> To: <openpower-hdl-cores@mailinglist.openpowerfoundation.org>, <
> openpower-community-dev@mailinglist.openpowerfoundation.org>
>
>
> Hi All,
>
> I posted the following link on Twitter earlier today - an opportunity to
> have an informal chat among the OpenPOWER community, may be of interest!
>
> https://openpowerfoundation.org/openpower-virtual-coffee-calls/
>
> Cheers,
> Hugh
>
>
> --
> Hugh Blemings
> Executive Director, OpenPOWER Foundationhttp://openpowerfoundation.org
>
> _______________________________________________
> OpenPOWER-HDL-Cores mailing list
> OpenPOWER-HDL-Cores@mailinglist.openpowerfoundation.org
> http://lists.mailinglist.openpowerfoundation.org/
> mailman/listinfo/openpower-hdl-cores
> _______________________________________________
> libre-riscv-dev mailing list
> libre-riscv-dev@lists.libre-riscv.org
> http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev
>
--
---
crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68
_______________________________________________
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] Fwd: [OpenPOWER-HDL-Cores] OpenPOWER Foundation "Virtual Coffee" Calls
From: Jacob Lifshay @ 2020-03-24 22:21 UTC (permalink / raw)
To: Libre-RISCV General Development
In-Reply-To: <023999b2-db59-6c90-2be9-43941f9ba4b4@openpowerfoundation.org>
Maybe we should participate.
---------- Forwarded message ---------
From: Hugh Blemings <hugh@openpowerfoundation.org>
Date: Mon, Mar 23, 2020, 19:03
Subject: [OpenPOWER-HDL-Cores] OpenPOWER Foundation "Virtual Coffee" Calls
To: <openpower-hdl-cores@mailinglist.openpowerfoundation.org>, <
openpower-community-dev@mailinglist.openpowerfoundation.org>
Hi All,
I posted the following link on Twitter earlier today - an opportunity to
have an informal chat among the OpenPOWER community, may be of interest!
https://openpowerfoundation.org/openpower-virtual-coffee-calls/
Cheers,
Hugh
--
Hugh Blemings
Executive Director, OpenPOWER Foundationhttp://openpowerfoundation.org
_______________________________________________
OpenPOWER-HDL-Cores mailing list
OpenPOWER-HDL-Cores@mailinglist.openpowerfoundation.org
http://lists.mailinglist.openpowerfoundation.org/mailman/listinfo/openpower-hdl-cores
_______________________________________________
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 216] LOAD STORE buffer needed
From: bugzilla-daemon @ 2020-03-24 20:27 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-216-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=216
--- Comment #19 from Jacob Lifshay <programmerjake@gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #18)
> (In reply to Jacob Lifshay from comment #17)
> > For GPU workloads, they still have some accesses (the majority for a tiled
> > architecture -- what Kazan is) that benefit from a L1 cache. basically, it
> > splits the screen into small squares/rectangles (bins) and accumulates a
> > list of rendering operations (binning) then for each square/rectangle (bin)
> > it renders all the operations inside it in sequence, so the pixels are
> > repeatedly stored and loaded until that bin is done rendering.
> >
> > Texture access also benefits from a cache since nearby pixels are accessed
> > together and often the accesses are overlapping.
>
> ok.
>
> ok so those 1 MB texture buffers you mentioned before are accessed multiple
> times in near-succession?
Yes. Though several different textures can be accessed in an interleaved
fashion.
--
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] Git repository access
From: Cole Poirier @ 2020-03-24 20:25 UTC (permalink / raw)
To: Libre-RISCV General Development; +Cc: Libre-RISCV General Development
In-Reply-To: <CAPweEDx8BpgfOBq15kYHrCX8AThEgijc+=AVjhXNQqpKo96FFg@mail.gmail.com>
On Mar 24 2020, at 1:11 pm, Luke Kenneth Casson Leighton <lkcl@lkcl.net> wrote:
> I believe for this I will need ieee754fpu.git/master/SoftPosit.patch
>
>
> if that's all you need you can get them manually via the git web interface.
>
> however you can also get them by cloning the git repo anonymously. git
> clone git://git.libre-riscv.org/ieee754fpu.git
>
> something like that.
>
> l.
Thanks Luke, the clone command you provided works perfectly for me! I
tried something similar to this earlier, but I must have gotten
something subtle incorrect.
Thank you,
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
* [libre-riscv-dev] [Bug 216] LOAD STORE buffer needed
From: bugzilla-daemon @ 2020-03-24 20:19 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-216-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=216
--- Comment #18 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Jacob Lifshay from comment #17)
> If we're having multiple L1 caches, we should implement them to appear as if
> there is a single L1 cache -- no extra instructions are needed to flush one
> cache before accessing the other,
yes, agreed. it appears, from the advice on comp.arch, that it is normal to
"stripe" the L1 cache, using e.g. bit 4 of the address (when using 128 bit
cache lines) to select odd or even halves.
if there are two separate LD/ST buffers, one for each half, this doubles the
bestcase data throughput at a cost of 40% misses in the general case (due to
each half being, duh, half the size).
however yes when it comes to flushing the cache these halves can, must and will
appear to be coherent as if they were a single L1 cache.
> For GPU workloads, they still have some accesses (the majority for a tiled
> architecture -- what Kazan is) that benefit from a L1 cache. basically, it
> splits the screen into small squares/rectangles (bins) and accumulates a
> list of rendering operations (binning) then for each square/rectangle (bin)
> it renders all the operations inside it in sequence, so the pixels are
> repeatedly stored and loaded until that bin is done rendering.
>
> Texture access also benefits from a cache since nearby pixels are accessed
> together and often the accesses are overlapping.
ok.
ok so those 1 MB texture buffers you mentioned before are accessed multiple
times in near-succession?
--
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] Status on Our RISCV Implementation
From: Luke Kenneth Casson Leighton @ 2020-03-24 20:12 UTC (permalink / raw)
To: Libre-RISCV General Development
In-Reply-To: <CAC2bXD7DfmUgx_OGODHX4ahH7HmwTd5LE0r_jkaxiF+NTuQzkg@mail.gmail.com>
On Tuesday, March 24, 2020, Jacob Lifshay <programmerjake@gmail.com> wrote:
However, it's perfectly possible to have software that catches the illegal
> instruction traps and emulates them, allowing running a RISC-V kernel.
oh. yes. good point. qemu will be doing something like this, i believe?
--
---
crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68
_______________________________________________
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] Git repository access
From: Luke Kenneth Casson Leighton @ 2020-03-24 20:11 UTC (permalink / raw)
To: Libre-RISCV General Development
In-Reply-To: <09EAF495-B495-41A4-9BE3-D840DA912D8C@getmailspring.com>
On Tuesday, March 24, 2020, Cole Poirier <colepoirier@gmail.com> wrote:
> Hi Luke,
>
> I'm currently working on getting the gitlab test runner set up on
> Jacob's build server. While doing this I've been trying to follow
> (install?) the HDL workflow, and I think I've reached a point where I
> need read access to the git repo. Specifically I'm trying to complete
> 6.5 Softfloat and sfpy.
>
> ###
> git clone --recursive https://github.com/billzorn/sfpy.git
> cd sfpy
> cd SoftPosit
> git apply ../softposit_sfpy_build.patch
>
> -> I believe for this I will need ieee754fpu.git/master/SoftPosit.patch
if that's all you need you can get them manually via the git web interface.
however you can also get them by cloning the git repo anonymously. git
clone git://git.libre-riscv.org/ieee754fpu.git
something like that.
l.
--
---
crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68
_______________________________________________
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] Git repository access
From: Cole Poirier @ 2020-03-24 18:56 UTC (permalink / raw)
To: libre-riscv-dev
Hi Luke,
I'm currently working on getting the gitlab test runner set up on
Jacob's build server. While doing this I've been trying to follow
(install?) the HDL workflow, and I think I've reached a point where I
need read access to the git repo. Specifically I'm trying to complete
6.5 Softfloat and sfpy.
###
git clone --recursive https://github.com/billzorn/sfpy.git
cd sfpy
cd SoftPosit
git apply ../softposit_sfpy_build.patch
-> I believe for this I will need ieee754fpu.git/master/SoftPosit.patch
git apply /path/to/ieee754fpu/SoftPosit.patch
cd ../berkely-softfloat-3
# Note: Do not apply the patch included in sfpy for berkely-softfloat,
# it contains the same changes as this one
-> I believe for this I will need ieee754fpu.git/master/berkeley-softfloat.patch
git apply /path/to/ieee754fpu/berkeley-softfloat.patch
###
If I can access these files without needing permission, or if there is
another way I should go about this, please let me know. I believe you
already have my ssh pub key, but if you need me to send it again that's
no problem.
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
* [libre-riscv-dev] [Bug 216] LOAD STORE buffer needed
From: bugzilla-daemon @ 2020-03-24 18:00 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-216-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=216
--- Comment #17 from Jacob Lifshay <programmerjake@gmail.com> ---
If we're having multiple L1 caches, we should implement them to appear as if
there is a single L1 cache -- no extra instructions are needed to flush one
cache before accessing the other, this greatly simplifies the programming model
and Kazan currently depends on the memory having a normal programming model (no
cache flushes required for an operation to be visible on the same thread).
For GPU workloads, they still have some accesses (the majority for a tiled
architecture -- what Kazan is) that benefit from a L1 cache. basically, it
splits the screen into small squares/rectangles (bins) and accumulates a list
of rendering operations (binning) then for each square/rectangle (bin) it
renders all the operations inside it in sequence, so the pixels are repeatedly
stored and loaded until that bin is done rendering.
Texture access also benefits from a cache since nearby pixels are accessed
together and often the accesses are overlapping.
--
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] Status on Our RISCV Implementation
From: Jacob Lifshay @ 2020-03-24 17:44 UTC (permalink / raw)
To: Libre-RISCV General Development
In-Reply-To: <CAPweEDzdwANkVmqbDV0gPc=EAGkhknMNDW72ypTkoW-zFz8Gow@mail.gmail.com>
On Tue, Mar 24, 2020, 06:59 Luke Kenneth Casson Leighton <lkcl@lkcl.net>
wrote:
> On Tue, Mar 24, 2020 at 1:11 PM Immanuel, Yehowshua U
> <yimmanuel3@gatech.edu> wrote:
>
> > > nope. not for RISC-V. userspace only.
> >
> > So I guess we’d have a POWER kernel…
>
> correct. and POWER TLBs and MMU.
>
> > If a binary tries to call RISCV instructions to set supervisor mode
> register - how is that handled?
>
> from userspace? fail... just like any other userspace RISC-V program
> is not permitted to set a supervisor mode register on any standard
> RISC-V system.
However, it's perfectly possible to have software that catches the illegal
instruction traps and emulates them, allowing running a RISC-V kernel.
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
* [libre-riscv-dev] Fwd: multi-way LOAD/STORE buffers and misalignment
From: Luke Kenneth Casson Leighton @ 2020-03-24 16:05 UTC (permalink / raw)
To: Libre-RISCV General Development, Staf Verhaegen
In-Reply-To: <7dcb89ad-c6b2-43dc-8ea9-e72342a2719b@googlegroups.com>
staf, hi, this came in from mitch alsup, from comp.arch.
the L1 cache memory SRAM that you are doing: is it, as Mitch suggests,
1R *or* 1W, and thus single-clock-cycle? i gather from what he's
saying that if you have a 1R1W SRAM, that effectively needs 2 clock
cycles to access.
l.
---
crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68
---------- Forwarded message ---------
From: 'MitchAlsup' via comp.arch <comp.arch@googlegroups.com>
Date: Tue, Mar 24, 2020 at 3:27 PM
Subject: Re: multi-way LOAD/STORE buffers and misalignment
To:
On Tuesday, March 24, 2020 at 6:53:29 AM UTC-5, lkcl wrote:
> we're doing a LD/ST buffer at the moment, and some design review and
> constructive feedback and advice would be greatly appreciated.
> unlike many "first time processors", we cannot possibly get away
> with a simple single-issue LD/ST design as the first iteration because
> of the vector engine being effectively a multi-issue multiplier that,
> from a single instruction, hammers the internal engine with a *batch*
> of actual operations.
>
> this includes LD/ST and so consequently we need a multi-issue-capable
> LD/ST buffer right from the start. it is being documented, here:
>
> https://libre-riscv.org/3d_gpu/architecture/6600scoreboard/
>
> with Mitch's help i already have a RAW / WAR dependency detection matrix
> in place, which detects (in-order) *batches* of LDs, *batches* of STs,
> and *batches* of atomic operations (defined by having LD/ST both set).
>
> additionally, we have (again, thanks to Mitch), an "address clash detection"
> Matrix, where the original design concept from Mitch uses bits 4 thru 11
> (stopping at a VM page) to say "this LD/ST definitely doesn't overlap with
> any others, but it might have picked up a few that *potentially* don't
> overlap, but hey better safe than sorry".
>
> this is quite efficient because the number of bits used to detect a clash
> is drastically reduced, which is important given that it's an NxN/2 set of
> binary-address comparators involved.
>
> however we *modified* this idea, by turning the addr+len (len=1/2/4/8 byte)
> into a *byte-mask*, where the byte-mask is effectively nothing more than
> (or, "can be used as") byte-enable lines on the L1 cache memory read/write.
>
> this means turning len=1 into 0b1, len=2 into 0b11, len=4 into 0b1111
> or len=8 into 0b1111 1111 and then shifting it by the LSBs of the address
> (bits 0-3).
>
> however of course, that could "wrap over" to a second cache line, therefore
> we split it into two:
>
> * 1st cache line, take bits 0..15 of the mask, use addr[4:] as the address
> * 1nd cache line, take bits 15..27 of the mask, use addr[4:] PLUS ONE as address
>
> now all LDs are potentially *two* LDs, and all STs are potentially *two* STs,
> requiring synchronisation.
>
> this does interestingly mean that the "address clash detection" Matrix is
> now doubled in size, because each LD/ST Computation Unit can now issue *two*
> LDs/STs, not one.
>
> there is a big advantage to partial-conversion of the LD/STs to unary-masked
> format:
>
> 1) those "masks" we can now detect in the address-clash unit, whether there
> are overlaps, by simple "unary" (ANDing).
>
> 2) once a batch of guaranteed-non-overlapping LDs (or STs, but not both at
> once) have been identified, the ORing of those masks *pre-identifies*,
> where the remaining MSBs are identical, which LDs and which STs are
> *part of the same cache line*.
>
> 3) the ORed masks can be dropped - as-is - directly onto byte-level
> read/write-enable lines on the L1 Cache underlying memory - with zero
> shifting or other messing about.
>
> so that's the bits that we "know".
>
> the bits that we *don't* know are what is the best L1 cache arrangement
> to use. we discussed the possibility of an 8-way set-associative cache
> with 512-bit wide data (!!!) - yes, really, because we will have a minimum
> of 4 LD/ST operations @ 64-bit occurring every cycle (in the first version
> of the processor: this *will* go up in later iterations).
>
> one thing that would be particularly nice to avoid would be multi-ported
> SRAM on the L1 Cache Memories. i thought there it may be possible to
> "stripe" the L1 Cache, such that bits 4-5 of every address go through to,
> effectively, a completely separate (isolated, independent) L1 Cache.
>
> * Addr bits [0:3] already turned into byte-map mask (enable lines)
> * Addr bits [4:5] used to select one of four INDEPENDENT caches
> * Addr bits [6:end] used "as normal"
>
> the only downside of this being: LD/STs that happen to be on the same
> 64-byte boundary would be "penalised" by being only single-issue-capable.
>
> do we care? i'm not sure that we do!
Yes, you do want to care:: you want to care to "get it correct" even
at some minor cost in perf.
>
> _should_ we care? that's the question i'd like to know the answer to :)
> an alternative is that we have to multi-port the L1 Cache memory, which
> would be quite expensive.
>
> what alternatives are there? is there a way to, for example, have an
> 8-way set-associative cache with 8 independent input/output data paths,
> yet they do not require the trick of a full split using addr bits [4:5],
> but at the same time do not require 4R4W SRAMs?
What you want to know are the dimensions of the SRAM macros. Given these
dimensions you are in a position to start making good choices, without
you are not. A typical SRAM dimension might be 256-bits wide and 64-words
deap (2KBytes).
Since you want to read/write 512-bits per port, you will need 2 SRAM
macros per port (per cycle). So if you are making 8-way SA cache, you
will need at least 16 SRAM macros (32KBytes)--and (A N D) you still
have but a single port!
On the other hand, dropping back to 4-way set, and that same SRAM array
can now support 2-ports of 4-way SA each. In this organization you will
be using more lower-address bits to choose port[0] versus port[1]. And
you will see 40% conflicts if you are really running 2 request per cycle.
However, if you know the access pattern is "dense", then you know that
LD[i][k..k+3] are all in the same line, and you can THEN make one access
for every 4 actual LDs an then sort it out with multiplexers. And NOW we
have enough ports and width to use 1024 bits per cycle throughput in the
cache. {You can determine an access pattern will be dense when the loop
increment is +1 (or -1) and the loop index is used in scaled form in
memref instructions:
LD Rx,[Rb+Ri<<scale+DISP]
} Other forms of dense access are harder for HW to detect and utilize.
>
> i *think* this is doable, with the potential for some wait cycles if
> one of the 8 ways happens to have request contention. honestly i don't know
> what's best here.
>
> advice and discussion appreciated as always,
>
> 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] ieee.org
From: Luke Kenneth Casson Leighton @ 2020-03-24 16:00 UTC (permalink / raw)
To: Libre-RISCV General Development
In-Reply-To: <C1B3E139-9079-4ED8-AA29-ECB5FA9FED1E@gatech.edu>
On Tue, Mar 24, 2020 at 2:44 PM Immanuel, Yehowshua U
<yimmanuel3@gatech.edu> wrote:
>
> Yeah. I sent the PDF to your personal email.
star, thanks.
_______________________________________________
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] ieee.org
From: Immanuel, Yehowshua U @ 2020-03-24 14:44 UTC (permalink / raw)
To: Libre-RISCV General Development
In-Reply-To: <CAPweEDxh4-=WQiD8H-nh847+E9xi8FYRj5rwc3WBKy_19AAO6Q@mail.gmail.com>
Yeah. I sent the PDF to your personal email.
Yehowshua
_______________________________________________
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 72] verilog to nmigen converter (full or partial) needed
From: bugzilla-daemon @ 2020-03-24 14:24 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-72-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=72
--- Comment #22 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
another couple of pieces of code which need "parameterisation":
https://ascslab.org/research/briscv/index.html
L1cache.v and Lxcache.v - the cache-coherence protocol used there
looks particularly good, and it would be nice to be able to see
this code in nmigen
--
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: next prev 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