* [libre-riscv-dev] [Bug 217] create a "ring" system which allows pad locations to be specified conveniently
From: bugzilla-daemon @ 2020-04-21 17:40 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-217-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=217
--- Comment #80 from Jean-Paul.Chaput@lip6.fr ---
(In reply to Luke Kenneth Casson Leighton from comment #78)
> print (dir(self.cell))
> for net in self.cell.getNets():
> print (net.getName())
> print (self.cell.getNet("a(0)"))
> print (dir(self.cell.getNet("a(0)")))
> net = self.cell.getNet("a(2)")
> print (list(net.getPlugs()))
> for plug in net.getPlugs():
> print (plug, dir(plug))
> components = list(plug.getConnexComponents())
> pc = plug.getCell()
> print ("plug cell", pc)
> mnet = plug.getMasterNet()
> print ("mstnet", mnet)
>
> hmmm this is only printing out add and sub instances
> <id:3628 Plug a(2) subckt_48_add.a(2)>
> <id:3495 Plug a(2) subckt_49_sub.a(2)>
>
> the reason for that, i am guessing, is because they've been instantiated
> (placed).
>
> i need the Plugs that have *not* been instantiated.
Duh? It seems ok to me, according to the netlist of alu16, top level
"a(2)" is connected only to "add" and "sub". Plug represent logical
connexions, so they cannot be "placed". The instance they belong to,
however, may be placed. What Plug connected to "a(2)" do you miss?
--
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 217] create a "ring" system which allows pad locations to be specified conveniently
From: bugzilla-daemon @ 2020-04-21 18:05 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-217-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=217
--- Comment #81 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Jean-Paul.Chaput from comment #80)
> Duh? It seems ok to me, according to the netlist of alu16, top level
> "a(2)" is connected only to "add" and "sub". Plug represent logical
> connexions, so they cannot be "placed". The instance they belong to,
> however, may be placed. What Plug connected to "a(2)" do you miss?
tch, tch, you are right. i needed add_o(2) and sub_o(2).
find = self.get_net_connections(['add_o(0)', 'sub_o(0)', 'o(0)'],
['clk', 'rst'])
more ['op', 'abc_829_new_n51', 'op', 'abc_829_new_n51', 'o_next(0)']
set([[0x7fe5b3dd9110<->0x55c7b4b77b10 <id:3433 Instance subckt_0_nmx2_x1
nmx2_x1>], [0x7fe5b3dd9130<->0x55c7b4b92b30 <id:3801 Instance subckt_32_sff1_x4
sff1_x4>]])
at the first level of recursion, it seems to be working. for a given value of
work.
this will, recursively, grab all unconnected nets that are connected to nets
that are... etc.
however that is too much.
this algorithm can be used to resolve all paths between two points:
https://www.python.org/doc/essays/graphs/
so by knowing that we want all components involved in add_0(0), sub_o(0)
and o(0) it is possible to find...
hmmm... net "op" is a problem. if you traverse "op" it goes everywhere.
by excluding "op" as a possibility, i managed to get it to "work"
recursively:
find = self.get_net_connections(['add_o(1)', 'sub_o(1)', 'o(1)'],
['clk', 'rst', 'op'])
set([[0x7f13ce86c070<->0x559cb535bc20 <id:3769 Instance subckt_33_sff1_x4
sff1_x4>], [0x7f13ce86c130<->0x559cb535a680 <id:3755 Instance subckt_2_nmx2_x1
nmx2_x1>], [0x7f13ce86c190<->0x559cb5352100 <id:3668 Instance subckt_3_no2_x1
no2_x1>]])
first list is what is wanted (start position), second list is what you do
*not* want. anything already found will be added recursively to the second
list.
--
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 217] create a "ring" system which allows pad locations to be specified conveniently
From: bugzilla-daemon @ 2020-04-21 19:24 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-217-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=217
--- Comment #82 from Jean-Paul.Chaput@lip6.fr ---
(In reply to Luke Kenneth Casson Leighton from comment #81)
> (In reply to Jean-Paul.Chaput from comment #80)
I'm still not sure about the exact walktrough you want, but if you want
to have all the connexions (Plugs) to a given net through all the
hierarchy levels, you may use the HyperNet object. You can then
select any Plug with the occurrence Path.
--
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 217] create a "ring" system which allows pad locations to be specified conveniently
From: bugzilla-daemon @ 2020-04-21 20:23 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-217-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=217
--- Comment #83 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Jean-Paul.Chaput from comment #82)
> (In reply to Luke Kenneth Casson Leighton from comment #81)
> > (In reply to Jean-Paul.Chaput from comment #80)
>
> I'm still not sure about the exact walktrough you want,
i want to replace ALU16.place_data_path() with an automated system.
1) given a net ("o(2)") find all unrouted Instances that are left which need
placing.
and *only* those for "o(2)".
not those connecting to "o(3)", or o(0).
the function Module.get_net_connections now achieves that job.
get_net_connections("o(2)") returns these:
* Instance subckt_31_no2_x1 no2_x1
* Instance subckt_47_sff1_x4 sff1_x4
* Instance subckt_30_nmx2_x1 nmx2_x1
2) given those unplaced components, perform an *automatic* (Etesian)
place on those components, within a small abutment box.
> but if you want
> to have all the connexions (Plugs) to a given net through all the
> hierarchy levels, you may use the HyperNet object. You can then
> select any Plug with the occurrence Path.
will look at that one later, i think i have the function for part (1)
already.
--
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 217] create a "ring" system which allows pad locations to be specified conveniently
From: bugzilla-daemon @ 2020-04-21 20:43 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-217-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=217
--- Comment #84 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Luke Kenneth Casson Leighton from comment #83)
> 2) given those unplaced components, perform an *automatic* (Etesian)
> place on those components, within a small abutment box.
in theory, once the group of [nmx2, no2 and sff1] for each connection
between add, sub and the output, could follow this process:
for i in range(BIT_WIDTH)
0) identify the [nmx2+no2+sff1] cells connected to output(i) add(i) and sub(i)
1) create a "fake" View (both physical and logical) fake(i)
2) ***REMOVE*** the nmx2+no2+sff1 instances from alu16
3) ***ADD*** them to the "fake" View
4) perform the "normal" Etesian place on fake(i)
this would achieve the desired task, with zero changes to coriolis2.
--
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 178] first coriolis2 tutorial, workflow and "test project" page
From: bugzilla-daemon @ 2020-04-21 21:00 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-178-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=178
Cole Poirier <colepoirier@gmail.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |colepoirier@gmail.com
--- Comment #216 from Cole Poirier <colepoirier@gmail.com> ---
Luke, do you want gcc-9-powerpc64-linux-gnu installed inside of the chroot? The
reason I ask is because it involves changing /etc/apt/sources.list from buster
to testing or unstable in order to install via apt, something that I presume is
trivial in the chroot, but likely not as feasible on the host system. If I am
wrong in any of my assumptions please clarify, I'm still very, very new to
linux.
--
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 178] first coriolis2 tutorial, workflow and "test project" page
From: bugzilla-daemon @ 2020-04-21 21:09 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-178-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=178
--- Comment #217 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Cole Poirier from comment #216)
> Luke, do you want gcc-9-powerpc64-linux-gnu installed inside of the chroot?
for coriolis2? we have no reason to compile executables (yet).
> The reason I ask is because it involves changing /etc/apt/sources.list from
> buster to testing or unstable in order to install via apt,
yeah that's not a good idea. the coriolis2 chroot absolutely has to be
a reproducible build. debian/testing is the absolute opposite of that.
> something that I presume is trivial in the chroot,
it's trivial on both, however the results would be absolute disaster.
i do not know if you have ever tried to compile software on a moving
target, it is absolute hell.
if we absolutely have to install gcc for powerpc in the chroot, it would
be better to do so from a full source code build.
--
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 178] first coriolis2 tutorial, workflow and "test project" page
From: bugzilla-daemon @ 2020-04-21 21:15 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-178-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=178
--- Comment #218 from Cole Poirier <colepoirier@gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #217)
> (In reply to Cole Poirier from comment #216)
> > Luke, do you want gcc-9-powerpc64-linux-gnu installed inside of the chroot?
>
> for coriolis2? we have no reason to compile executables (yet).
I've left out your comments regarding compiling for a moving target for
brevity, but I definitely take your point. The reason I ask is that I'm
following the HDL workflow, and there is a section instructing the user to
install gcc-9-powerpc.
```
Section 6.6 qemu, cross-compilers, gdb
As we are doing POWER ISA, POWER ISA compilers, toolchains and emulators are
required.
Install powerpc64 gcc:
apt-get install gcc-9-powerpc64-linux-gnu
Install qemu:
apt-get install qemu-system-ppc
Install gdb from source. Obtain the latest tarball, unpack it, then:
cd gdb-9.1 (or other location)
mkdir build
cd build
../configure --srcdir=.. --host=x86_64-linux --target=powerpc64-linux-gnu
make -j16
make install
```
--
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 217] create a "ring" system which allows pad locations to be specified conveniently
From: bugzilla-daemon @ 2020-04-21 21:19 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-217-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=217
--- Comment #85 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
ok i have replaced ALU16.place() with a new version.
commit 83f918107bb29f1ab3ef08eae5c1624d0376179b
Author: Luke Kenneth Casson Leighton <lkcl@lkcl.net>
Date: Tue Apr 21 21:05:18 2020 +0000
automatically located the joining cells between add and sub
this new version no longer explicitly regexp pattern-searches for
the cells named nmx2, no2 and sff1
instead it uses the new function, calling get_net_connections("o(%d)" % i),
for all i in range(16).
it produces - finds - the exact same three cells as the regexp pattern
search ALU16.match_instance()
however unlike match_instance(), if the yosys/ABC changes, produces
completely difference cells, it will *still find them all*.
now we need step (2)
step (2) is:
* given the (arbitrary) list of instances, auto-place them within a set
box.
you can see i have started to try that, and it produces a segfault.
attempting to call Etesian.create(inst.getCell()) does not work.
attempting to call Etesian.create(inst.getMasterCell()) does not work.
attempting to call Etesian.create(inst) segfaults.
the only other trick i can think of is to LITERALLY remove them,
programmatically, from alu16, and move them to a new (fake) Cell.
times 16.
this will be extremely tedious because it requires moving the netlists
as well.
what is *really* needed is, to be able to pass a **LIST** of Cells
to Etesian.create(). (actually, a list of Instances)
and to be able to have EtesianEngine constructor take that **LIST**
of Instances
and to perform the place on that **LIST** of Instances.
--
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 178] first coriolis2 tutorial, workflow and "test project" page
From: bugzilla-daemon @ 2020-04-21 21:28 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-178-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=178
--- Comment #219 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Cole Poirier from comment #218)
> (In reply to Luke Kenneth Casson Leighton from comment #217)
> > (In reply to Cole Poirier from comment #216)
> > > Luke, do you want gcc-9-powerpc64-linux-gnu installed inside of the chroot?
> >
> > for coriolis2? we have no reason to compile executables (yet).
>
> I've left out your comments regarding compiling for a moving target for
> brevity, but I definitely take your point. The reason I ask is that I'm
> following the HDL workflow,
ok. then this bugreport, which is not for HDL_workflow, it is for coriolis2,
is not the right place.
plus, this bugreport (related to coriolis2) has been closed.
> and there is a section instructing the user to
> install gcc-9-powerpc.
>
> ```
> Section 6.6 qemu, cross-compilers, gdb
>
> As we are doing POWER ISA, POWER ISA compilers, toolchains and emulators are
> required.
>
> Install powerpc64 gcc:
>
> apt-get install gcc-9-powerpc64-linux-gnu
that's only available in debian/testing.
gcc-8-powerpc64-linux-gnu is what is available in debian/10.
--
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] [Bug 178] first coriolis2 tutorial, workflow and "test project" page
From: Luke Kenneth Casson Leighton @ 2020-04-21 21:30 UTC (permalink / raw)
To: Libre-RISCV General Development
In-Reply-To: <bug-178-13-hn3JijGVYl@https.bugs.libre-soc.org/>
so, cole - open a new bugreport and we can discuss it there ok?
each new "thing" needs its own bugreport.
i began discussing the coriolis2 chroot with you because you used the
coriolis2 tutorial bugreport.
that is - was - its context.
you see why it is important to keep to separate bugs?
l.
---
crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68
On Tue, Apr 21, 2020 at 10:29 PM <bugzilla-daemon@libre-soc.org> wrote:
>
> https://bugs.libre-soc.org/show_bug.cgi?id=178
>
> --- Comment #219 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
> (In reply to Cole Poirier from comment #218)
> > (In reply to Luke Kenneth Casson Leighton from comment #217)
> > > (In reply to Cole Poirier from comment #216)
> > > > Luke, do you want gcc-9-powerpc64-linux-gnu installed inside of the chroot?
> > >
> > > for coriolis2? we have no reason to compile executables (yet).
> >
> > I've left out your comments regarding compiling for a moving target for
> > brevity, but I definitely take your point. The reason I ask is that I'm
> > following the HDL workflow,
>
> ok. then this bugreport, which is not for HDL_workflow, it is for coriolis2,
> is not the right place.
>
> plus, this bugreport (related to coriolis2) has been closed.
>
> > and there is a section instructing the user to
> > install gcc-9-powerpc.
> >
> > ```
> > Section 6.6 qemu, cross-compilers, gdb
> >
> > As we are doing POWER ISA, POWER ISA compilers, toolchains and emulators are
> > required.
> >
> > Install powerpc64 gcc:
> >
> > apt-get install gcc-9-powerpc64-linux-gnu
>
> that's only available in debian/testing.
>
> gcc-8-powerpc64-linux-gnu is what is available in debian/10.
>
> --
> 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
_______________________________________________
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] [Bug 178] first coriolis2 tutorial, workflow and "test project" page
From: Cole Poirier @ 2020-04-21 21:35 UTC (permalink / raw)
To: Libre-RISCV General Development; +Cc: Libre-RISCV General Development
In-Reply-To: <CAPweEDxamA=ko3vYLt+yeR2p0D9LtbJUt7TrGoJx0iUJGF3gBA@mail.gmail.com>
On Apr 21 2020, at 2:30 pm, Luke Kenneth Casson Leighton <lkcl@lkcl.net> wrote:
> so, cole - open a new bugreport and we can discuss it there ok?
>
> each new "thing" needs its own bugreport.
>
> i began discussing the coriolis2 chroot with you because you used the
> coriolis2 tutorial bugreport.
>
> that is - was - its context.
>
> you see why it is important to keep to separate bugs?
>
> l.
>
Ah my mistake, I didn't realize this bug report was closed, and
mistakenly thought that this was the appropriate place to discuss it.
Thank you for the clarification. I will make a new bug report now.
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 291] New: HDL Workflow and Coriolis2 chroot automated setup scripts
From: bugzilla-daemon @ 2020-04-21 21:42 UTC (permalink / raw)
To: libre-riscv-dev
https://bugs.libre-soc.org/show_bug.cgi?id=291
Bug ID: 291
Summary: HDL Workflow and Coriolis2 chroot automated setup
scripts
Product: Libre-SOC's first SoC
Version: unspecified
Hardware: PC
OS: Linux
Status: CONFIRMED
Severity: normal
Priority: ---
Component: Hardware Layout
Assignee: lkcl@lkcl.net
Reporter: colepoirier@gmail.com
CC: libre-riscv-dev@lists.libre-riscv.org
NLnet milestone: ---
In service of the description of bug 178: "We need the alliance2/coriolis
workflow documented, a suite of tutorials found (or written), and a "test
project" to be done which gives a guide to completion time of layout." Cole is
working on scripts to automate the process of installation and setup of the HDL
workflow development environment. This should make the on-boarding of the
LIPFR6 interns who will be doing the layout this summer easier.
--
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 291] HDL Workflow and Coriolis2 chroot automated setup scripts
From: bugzilla-daemon @ 2020-04-21 21:46 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-291-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=291
--- Comment #1 from Cole Poirier <colepoirier@gmail.com> ---
Luke, should I amend section 6.6 of the HDL workflow to install
gcc-8-powerpc64-linux-gnu as you said this is what is available in debian/10?
The current instructions say to install gcc-9-powerpc64-linux-gnu but as you
said that's only available in debian/testing. Since you said that there should
be no need for compiling ppc code at this point, is this section meant to be a
part of a different workflow?
--
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 217] create a "ring" system which allows pad locations to be specified conveniently
From: bugzilla-daemon @ 2020-04-21 22:00 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-217-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=217
--- Comment #86 from Jean-Paul.Chaput@lip6.fr ---
(In reply to Luke Kenneth Casson Leighton from comment #85)
> attempting to call Etesian.create(inst.getMasterCell()) does not work.
>
> attempting to call Etesian.create(inst) segfaults.
Oups. Should throw an exception, not segfault... Will correct.
> the only other trick i can think of is to LITERALLY remove them,
> programmatically, from alu16, and move them to a new (fake) Cell.
>
> times 16.
>
> this will be extremely tedious because it requires moving the netlists
> as well.
Something similar is done for the clock tree. It is nightmarish,
don't go that way.
> what is *really* needed is, to be able to pass a **LIST** of Cells
> to Etesian.create(). (actually, a list of Instances)
>
> and to be able to have EtesianEngine constructor take that **LIST**
> of Instances
>
> and to perform the place on that **LIST** of Instances.
I see no problem implementing that feature, but will do it next
week. I'm long overdue to help an intern...
--
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 217] create a "ring" system which allows pad locations to be specified conveniently
From: bugzilla-daemon @ 2020-04-21 22:21 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-217-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=217
--- Comment #87 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Jean-Paul.Chaput from comment #86)
> > this will be extremely tedious because it requires moving the netlists
> > as well.
>
> Something similar is done for the clock tree. It is nightmarish,
> don't go that way.
:)
the alternative is that we do it at the HDL level. create the series of
"modules" ehich result in separate actual VST files, link_add0_sub0_out0.vst
etc.
also tedious :)
> > what is *really* needed is, to be able to pass a **LIST** of Cells
> > to Etesian.create(). (actually, a list of Instances)
> I see no problem implementing that feature,
that would be extremely cool.
it will have many uses, in combination with the netlist subset finder we can
have much better control over placement, without complicating the HDL.
> but will do it next
> week. I'm long overdue to help an intern...
no problem. thank you JP.
--
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 257] Implement demo Load/Store queueing algorithm
From: bugzilla-daemon @ 2020-04-22 2:58 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-257-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=257
--- Comment #21 from Jacob Lifshay <programmerjake@gmail.com> ---
Wishbone as-is is not sufficient for handling the interconnect between the L1
and L2 caches, or to any other device that can act as a bus master and is cache
coherent (such as other CPUs, OmniXtend, or cache coherent DMA). The reason is
because it has no concept of ownership of a particular cache block (or caching
at all), which is needed to support a cache coherence protocol.
Wishbone also has the drawback of memory bandwidth being limited by the bus
having a max width of 64 bits (ends up as 6.4GB/s peak theoretical bandwidth at
800MHz), we are likely to want more bandwidth between cores and the L2 cache.
Minerva's memory interface appears to not be designed to handle cache coherence
at all, so would need to be redesigned to support that if we wanted to use
Minerva's memory interface, which I don't think we should since we would need
to rewrite the majority of Minerva's memory interface code.
Using wishbone for devices that don't need to be cache coherent and don't need
large memory bandwidth (less than 1GB/s or so) is still a good idea.
--
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 257] Implement demo Load/Store queueing algorithm
From: bugzilla-daemon @ 2020-04-22 8:30 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-257-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=257
--- Comment #22 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
AXI4 does not have cache coherency either and yet it is still achieved. it is
done by passing messages over the "control bus", using home-grown protocols in
every case that i have examined (including ariane and OpenPITON).
we do not need to be fully compliant with the wishbone spec internally.
it is perfectly fine to modify the code so that it has the required bus
bandwidth for internal use.
when external connections to other peoples' code is needed (peripherals) then
and only then will we need to comply with the spec.
--
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 257] Implement demo Load/Store queueing algorithm
From: bugzilla-daemon @ 2020-04-22 8:39 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-257-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=257
--- Comment #23 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
also, we're not doing an SMP core for the october 2020 deadline, because we are
only doing a single core.
therefore L1 and L2 SMP cache coherency is not needed.
we *need* to urgently get moving on this, we are seriously behind.
the fastest way to do that is to not spend months writing code that already
exists.
the fastest way to get the job done and meet that hard deadline is to only
spend days or weeks using existing code.
minerva *already* uses wishbone.
in addition, there is the "soc" code from enjoy-digital (litex) which has
testing infrastructure and connectivity to peripherals already done.
we *need* test infrastructure and peripheral connectivity.
we *cannot* spend weeks or months writing new test infrastructure that already
exists because we're using a completely new Bus Protocol (TileLink) that
completely isolates and cuts us off from that test infrastructure, because that
test infrastructure is all written to use Wishbone.
after the October 2020 deadline, when SMP cache coherency is a goal, we can
look at this again.
right now, we have to shelve SMP, and get the job done. it's only six months
and at least two of those are going to be full-time working on the layout.
--
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 257] Implement demo Load/Store queueing algorithm
From: bugzilla-daemon @ 2020-04-22 8:54 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-257-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=257
--- Comment #24 from Jacob Lifshay <programmerjake@gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #23)
> also, we're not doing an SMP core for the october 2020 deadline, because we
> are only doing a single core.
>
> therefore L1 and L2 SMP cache coherency is not needed.
Ok, we can write the code to basically have whatever cache coherency stuff gets
written when getting the memory interface working and not worry about the L2 or
other cores existing at all, treating it as a writethrough transparent cache.
The design I'm working on is integrated closely enough with the L1 cache that
it will still need at least a little of the coherence protocol to support
atomics, but it can be built to have a WishBone interface adaptor.
--
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 257] Implement demo Load/Store queueing algorithm
From: bugzilla-daemon @ 2020-04-22 9:48 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-257-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=257
--- Comment #25 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Jacob Lifshay from comment #24)
> (In reply to Luke Kenneth Casson Leighton from comment #23)
> > also, we're not doing an SMP core for the october 2020 deadline, because we
> > are only doing a single core.
> >
> > therefore L1 and L2 SMP cache coherency is not needed.
>
> Ok, we can write the code to basically have whatever cache coherency stuff
> gets written when getting the memory interface working and not worry about
> the L2 or other cores existing at all, treating it as a writethrough
> transparent cache.
basically, yes. we're only going to be running at 25 mhz (possibly 100 mhz)
because there's no PLL, and no clock gating.
therefore yes even the L2 can be cut.
> The design I'm working on is integrated closely enough
> with the L1 cache that it will still need at least a little of the coherence
> protocol to support atomics,
remember, i said already that atomic memory operations are taken care of
by the Address / Memory Dependency Matrices.
they are already split out into individual (single) operations, each one
isolated from (batches of) LOADs and (batches of) STOREs.
so there's strictly speaking no need to implement any kind of coherence
protocol, because as things stand it's never going to be active.
yes it's over-zealous, it is however just a side-effect of how the
address-matching / memory-dependency matrices.
> but it can be built to have a WishBone
> interface adaptor.
ok good. does it implement this algorithm?
https://bugs.libre-soc.org/show_bug.cgi?id=216#c23
https://libre-soc.org/3d_gpu/mem_l0_to_l1_bridge_v2.png
--
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 216] LOAD STORE buffer needed
From: bugzilla-daemon @ 2020-04-22 10:53 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-216-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=216
--- Comment #24 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
i just realised something:
* misaligned LD/STs will create a pair of LD/STs (on each FU) where it
is *guaranteed* that Addr[4] for the LO-HI request will be mutually
exclusive. if Addr[4] is 0 for the LO request of the pair, Addr[4]
is *guaranteed* to be 1 for the HI request of the pair
* therefore splitting into L0 left/right LO/HI isn't going to work...
unless we accept that the LO/HI requests actually *cross over*!
FU0 LO -> L0 left if Addr[4] == 0 else L0 right
FU0 HI -> L0 right addr 4 will be 1 else L0 right
this has the strong disadvantage of not keeping the addresses that
we *know* have the same bits in the upper parts [12:48].
the other way to do it therefore is for the L0 to be only 8 rows
deep, comprising *two* 128-bit entries, i.e. to be integrated
L0-odd and L0-even where previously it was two separate L0 caches.
two L1 caches would then connect to that.
i'll think it through.
--
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 216] LOAD STORE buffer needed
From: bugzilla-daemon @ 2020-04-22 12:17 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-216-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=216
--- Comment #25 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
done.
https://libre-soc.org/3d_gpu/twin_l0_cache_buffer.jpg
the key is that there are only 8 rows (not 16) however each row contains
*two* byte-masks, *two* addr[5:11] and data sets, but *ONE* addr[12:48]
per row.
multiplexing is exclusively limited to 2-in 2-out, on a per-row basis.
each FU having two ports need *only* perform multiplexing on its *own row*
(not to any other rows), using addr[4] to select which bank the two data
sets are to go into.
the thing is that because the two data sets from each FU are consecutive
addresses (one aligned, one mis-aligned, both coming from the exact same
LD/ST), we *know* that they will both go simultaneously into the row:
it's just a matter of which one goes into which row.
so we need a 2-in 2-out crossbar, no multiplexing, no arbitration or
delay-gates needed.
i'll do the writeup now
https://libre-soc.org/3d_gpu/architecture/6600scoreboard/
--
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 288] intro post for powerpc-notebook.org and https://www.powerprogress.org/en/
From: bugzilla-daemon @ 2020-04-22 12:26 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-288-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=288
Yehowshua <yimmanuel3@gatech.edu> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |yimmanuel3@gatech.edu
--- Comment #1 from Yehowshua <yimmanuel3@gatech.edu> ---
Something Like This:
LibreSOC is an Open Hardware and Open Software project that aims to deliver a
physical quad core 180nm POWER compliant SOC that comes complete with a CPU,
GPU, VPU, and DDR controller. All the software and hardware from the drivers
down to the RTL and VLSI cells are open. LibreSOC is also providing the
necessary drivers amongst which include Vulkan and OpenCL drivers.
The intended market includes customers who desire acceleration in the embedded
space without relying on ARM's proprietary drivers that have been know to break
in the past.
I'll figure out what to put on LibreSOC soon.
--
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 288] intro post for powerpc-notebook.org and https://www.powerprogress.org/en/
From: bugzilla-daemon @ 2020-04-22 12:49 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-288-13@https.bugs.libre-soc.org/>
https://bugs.libre-soc.org/show_bug.cgi?id=288
--- Comment #2 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Yehowshua from comment #1)
> Something Like This:
>
> LibreSOC is an Open Hardware and Open Software project that aims to deliver
> a physical quad core 180nm POWER compliant SOC that comes complete with a
no, single-core. we're not doing quad-core, it will require something
like 160 mm^2 and that would cost around USD $60,000 just on its own.
not to mention, the power consumption would be immense, well beyond
what a standard low-cost JEDEC QFP package could handle.
we're keeping it to single-core, no SMP, no L2 cache.
https://libre-soc.org/3d_gpu/180nm_single_core_testasic_memlayout.jpg
the *next* version is quad-core, 28-45nm, full SMP.
--
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