* Re: [libre-riscv-dev] Who Buys Talos Workstations?
From: Cole Poirier @ 2020-04-05 7:34 UTC (permalink / raw)
To: Libre-RISCV General Development
In-Reply-To: <20200405085554.59a20bb18cf859e493a0b34f@gmx.com>
> On Apr 4, 2020, at 10:55 PM, Lauri Kasanen <cand@gmx.com> wrote:
> Well, first of all, people don't run i9s or threadrippers on servers.
> Those are customer-class, customer-availability, no ECC, less RAM
> accessible, etc. They run Xeons and Epycs, which are comparable to POWER
> in price. (hobbyists not counted here, for home servers money tends to
> come first)
>
> Their customers tend to value openness and security (their own keys, not
> the vendor's, for a truly secure boot; no ME backdoors). They may also
> be using accelerators made for POWER's custom buses, faster than PCIe.
>
> - Lauri
OpenCapi and the rest of the Open Power Foundation projects are incredibly cool!
Some brief info with links below. As far as I am aware we don’t have plans to use this interface in Libre-SOC V1, but please correct me if this is incorrect. If it’s the the case that we are not using the interface, would you still like me to add this to the resources section of the wiki?
From the OpenCAPI-Accelerator website (https://opencapi.github.io/oc-accel-doc/):
“What is OpenCAPI?
OpenCAPI (Open Coherent Accelerator Processor Interface) is the third-generation coherent interface and is an open standard for coherent high-performance bus interface. Driven by emerging, accelerated heterogeneous computing and advanced memory and storage solutions, it provides the open interface that allows any microprocessor to attach to:
Coherent user-level accelerators and I/O devices.
Advanced memories accessible via read/write or user-level DMA semantics
Its specifications and ecosystem are managed by the OpenCAPI Consortium. The reference designs are open source and available on Github at the https://github.com/OpenCAPI/OpenCAPI3.0_Client_RefDesign”
So like Lauri said it allows users to have openness and security, for better control over their own hardware. It also allows devices to interface with various external I/O, at some pretty mind blowing speeds too, which I think is shown on slide 13 of this 2018 presentation (https://openpowerfoundation.org/wp-content/uploads/2018/04/Myron-Slota.pdf).
Gen1 and Gen2 were layered over PCIE but now in Gen3-Power9 the protocol runs on new OpenCAPI hardware. Another important advantage appears to be that the upgradability and forwards and backwards compatibility are much higher than with previous solutions.
From the Open Power Foundation Website (http://openpowerfoundation.org/wp-content/uploads/resources/oppa-1/content/dbdoclet.50429986_67380.html):
“Chapter 1. Overview
This document defines the OpenCAPI Power Platform Architecture (OPPA) for the IBM™ POWER9™ systems. The information contained in this document allows various OPPA-compliant accelerator implementations to meet the needs of a wide variety of systems and applications. Compatibility with the OPPA allows applications and system software to migrate from one implementation to another with minor changes.”
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] Who Buys Talos Workstations?
From: Lauri Kasanen @ 2020-04-05 5:55 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <B57B2C73-A1D7-4CC1-920F-D49C4769364D@gatech.edu>
On Sun, 5 Apr 2020 01:52:47 +0000
"Immanuel, Yehowshua U" <yimmanuel3@gatech.edu> wrote:
> I was looking at RaptorCS workstations and was pleasantly surprised to find that the entire system from MOBO to firmware is nearly entirely open source with exception of the ethernet card firmware and the CPU RTL of course.
>
> However, compared to their competitors which are equally performant i9s and threadripper systems, they are about twice as expensive. This makes me wonder, what are the primary PPC server customers in this market?
Well, first of all, people don't run i9s or threadrippers on servers.
Those are customer-class, customer-availability, no ECC, less RAM
accessible, etc. They run Xeons and Epycs, which are comparable to POWER
in price. (hobbyists not counted here, for home servers money tends to
come first)
Their customers tend to value openness and security (their own keys, not
the vendor's, for a truly secure boot; no ME backdoors). They may also
be using accelerators made for POWER's custom buses, faster than PCIe.
- Lauri
_______________________________________________
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 269] auto-conversion / parser of POWER ISA Spec v3.0B
From: bugzilla-daemon @ 2020-04-05 4:28 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-269-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=269
--- Comment #26 from Jacob Lifshay <programmerjake@gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #21)
> i'm not happy at the deviations from the spec that have had to be made!
> the rather terrible solution is to have EXTZ and EXTS set the SelectableInt
> bitlength to (say) 1,000,000 then use that to detect that when you get this:
>
> x + EXTS(d)
>
> it means that in the addition, the rhs takes the bit-length of x.
>
> okok we can't do 1,000,000 bits because in EXTS that would be a 1,000,000 bit
> signed integer. maybe... 256 bits or something.
>
> i will modify SelectableInt so it is:
>
> def __add__(self, rhs):
> if rhs.bits == 256:
> rhs = SelectableInt(rhs.value, self.bits)
>
> that will "do" as a workaround.
See bug 278, comment 1 for a better workaround that doesn't require giant bit
counts.
--
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 279] inconsistency in 3.0B spec on definition of "equivalence" operator
From: bugzilla-daemon @ 2020-04-05 4:19 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-279-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=279
Jacob Lifshay <programmerjake@gmail.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |programmerjake@gmail.com
--- Comment #1 from Jacob Lifshay <programmerjake@gmail.com> ---
In my opinion, both == and XNOR (as well as != and XOR) can be used to denote
the same operation, but with two different ways of thinking about it:
1. XNOR/XOR used when flipping bits, (carry-less) adding, LFSRs, etc.
2. ==/!= used when checking for same/different values.
So pick which one you use based on how you are thinking about it, perhaps with
a note about being the same as the other operator when relevant.
Jacob
--
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 278] POWER v3.0B spec ambiguity on EXTS and missing EXTZ
From: bugzilla-daemon @ 2020-04-05 4:10 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-278-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=278
Jacob Lifshay <programmerjake@gmail.com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |programmerjake@gmail.com
--- Comment #1 from Jacob Lifshay <programmerjake@gmail.com> ---
Reply on openpower-hdl-cores:
http://lists.mailinglist.openpowerfoundation.org/pipermail/openpower-hdl-cores/2020-April/000022.html
On Sat, Apr 4, 2020, 07:16 Luke Kenneth Casson Leighton <lkcl at lkcl.net>
wrote:
> continuing with the implementation of a simulator that is taken
> directly - and literally - from the pseudocode and definitions of
> functions in the spec, we have hit an "inconvenient" definition.
> EXTS, section 1.3.4, page 6:
>
> EXTS(x) Result of extending x on the left with sign
> bits
>
> the problem is illustrated by asking the very simple question, "how many
> bits?"
Maybe it would work to treat it as a conversion from a bitstring to a
mathematical integer (like a Python3 int) where it can then be converted
back to a bitstring when it is assigned to something with a size again?
You don't need an infinite number of hardware bits to represent the
mathematical integer, since you know the exact range the values fall in so
you can calculate exactly how many bits you need, sign/zero-extending as
needed when converting back to a bitstring.
Jacob
--
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] SiFIve to go with PowerVR
From: Immanuel, Yehowshua U @ 2020-04-05 2:53 UTC (permalink / raw)
To: Libre-RISCV General Development
This isn’t exactly new news, but SiFive was/is working with PowerVR to create an entirely integrated SOC with a GPU, WiFi, etc…
But it won’t be Libre…
https://www.eetimes.com/imagination-gpu-now-on-sifive-risc-v-design-platform/
Perhaps we should have a page with content like this, just to refer to…
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] Who Buys Talos Workstations?
From: Immanuel, Yehowshua U @ 2020-04-05 1:52 UTC (permalink / raw)
To: Libre-RISCV General Development
I was looking at RaptorCS workstations and was pleasantly surprised to find that the entire system from MOBO to firmware is nearly entirely open source with exception of the ethernet card firmware and the CPU RTL of course.
However, compared to their competitors which are equally performant i9s and threadripper systems, they are about twice as expensive. This makes me wonder, what are the primary PPC server customers in this market?
I know for example that the US military uses NXP PPC processors in aircraft. Also, the auto industry STILL uses PPC a lot. But who would be interested in a more expensive POWER server? I couldn’t find much on this from usual Google sources...
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 269] auto-conversion / parser of POWER ISA Spec v3.0B
From: bugzilla-daemon @ 2020-04-04 22:25 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-269-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=269
--- Comment #25 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
oh, and one other thing... :)
the names of registers that are "recognised" are waay back in
decoder/pseudo/parser.py currently hard-coded
these need to be "lifted" from the current Form (Form-X)
however they will also need to be augmented by (p30):
* LR
* CTR
* CR (which can also be specified as CR0-CR7)
* XER
* FPSC
and TAR (p32)
oh, and instruction addresses:
* NIA (next)
* CIA (current)
and possibly some of the SPRs.
--
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 269] auto-conversion / parser of POWER ISA Spec v3.0B
From: bugzilla-daemon @ 2020-04-04 21:36 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-269-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=269
--- Comment #24 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
ah, michael, before you get too far into writing unit tests, you'll
find that the opcode reg spec information has been decoded from
the mdwn files, in pagereader.py opfields "regs".
feel free, back in pywriter.py, to push that out as an extra parameter
into the instrs[opcodename] tuple, then actually *use* it in the
unit tests (!)
oh also i think the form (X-Form) may be necessary to drop into that
tuple as well. maybe. probably.
--
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 269] auto-conversion / parser of POWER ISA Spec v3.0B
From: bugzilla-daemon @ 2020-04-04 20:46 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-269-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=269
--- Comment #23 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
fixedload done
fixedstore done
fixedlogical done
--
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 269] auto-conversion / parser of POWER ISA Spec v3.0B
From: bugzilla-daemon @ 2020-04-04 20:39 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-269-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=269
--- Comment #22 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
nice job btw
commit 237d4f455ae825941b0542819c02fdc87585808a
Author: Michael Nolan <mtnolan2640@gmail.com>
Date: Sat Apr 4 16:23:01 2020 -0400
Sorta working add instruction
oh. i got into a horrible mess on the submodule, actually lost a
commit.
if you get a rebase error just due to the submodule libreriscv (the wiki)
do let me know if you manage to fix it!
https://stackoverflow.com/questions/826715/how-do-i-manage-conflicts-with-git-submodules
--
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 269] auto-conversion / parser of POWER ISA Spec v3.0B
From: bugzilla-daemon @ 2020-04-04 20:37 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-269-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=269
--- Comment #21 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Michael Nolan from comment #20)
> (In reply to Luke Kenneth Casson Leighton from comment #16)
>
> > * as you saw in the bug #272 there's a problem with EXTS, which alerted
> > me to adding the syntax (RA|0)
>
> If we're sure that power_decoder2 is correct, then all of the EXTS(...) and
> EXTZ(...) cases can be replaced with pdecode2.e.imm_data, because the
> decoder already handles sign/zero extending immediates.
mmm... yyeeees... except... that means modifying the pseudo-code which
comes directly from the spec.
this has two implications:
1) it's not a (strict) verification of the spec
2) as you point out, we'd be using something that we're not sure is correct
the more "direct", the better.
i'm not happy at the deviations from the spec that have had to be made!
the rather terrible solution is to have EXTZ and EXTS set the SelectableInt
bitlength to (say) 1,000,000 then use that to detect that when you get this:
x + EXTS(d)
it means that in the addition, the rhs takes the bit-length of x.
okok we can't do 1,000,000 bits because in EXTS that would be a 1,000,000 bit
signed integer. maybe... 256 bits or something.
i will modify SelectableInt so it is:
def __add__(self, rhs):
if rhs.bits == 256:
rhs = SelectableInt(rhs.value, self.bits)
that will "do" as a workaround.
--
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 269] auto-conversion / parser of POWER ISA Spec v3.0B
From: bugzilla-daemon @ 2020-04-04 20:28 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-269-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=269
--- Comment #20 from Michael Nolan <mtnolan2640@gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #16)
> * as you saw in the bug #272 there's a problem with EXTS, which alerted
> me to adding the syntax (RA|0)
If we're sure that power_decoder2 is correct, then all of the EXTS(...) and
EXTZ(...) cases can be replaced with pdecode2.e.imm_data, because the decoder
already handles sign/zero extending immediates.
--
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 279] New: inconsistency in 3.0B spec on definition of "equivalence" operator
From: bugzilla-daemon @ 2020-04-04 19:54 UTC (permalink / raw)
To: libre-riscv-dev
http://bugs.libre-riscv.org/show_bug.cgi?id=279
Bug ID: 279
Summary: inconsistency in 3.0B spec on definition of
"equivalence" operator
Product: Libre-SOC's first SoC
Version: unspecified
Hardware: PC
OS: Linux
Status: CONFIRMED
Severity: enhancement
Priority: ---
Component: Specification
Assignee: lkcl@lkcl.net
Reporter: lkcl@lkcl.net
CC: libre-riscv-dev@lists.libre-riscv.org
NLnet milestone: ---
Section 1.3.4 p6
Equivalence logical operators ((a===b) = (a XOR ¬b))
Section 2.5.1 p40
The bit in the Condition Register specified by BA+32 is
XORed with the bit in the Condition Register specified
by BB+32, and the complemented result is placed into
the bit in the Condition Register specified by BT+32.
Section 3.3.13 p95
The contents of register RS are XORed with the con-
tents of register RB and the complemented result is
placed into register RA.
these are "technically" equivalent, however the definition is inconsistent
with the wording. to be consistent, Section 1.3.4 should read:
Equivalence logical operators ((a===b) = ¬(a XOR b))
however by an accidental property of XOR, they're the same.
also known as an "XNOR gate", which is *sigh* a better name than "Equivalence"
https://en.wikipedia.org/wiki/XNOR_gate
--
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 269] auto-conversion / parser of POWER ISA Spec v3.0B
From: bugzilla-daemon @ 2020-04-04 19:39 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-269-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=269
--- Comment #19 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Michael Nolan from comment #17)
> I don't see an internal_op.py anywhere, is this instead for
> internalop_sim.py?
yes, sorry.
--
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 269] auto-conversion / parser of POWER ISA Spec v3.0B
From: bugzilla-daemon @ 2020-04-04 19:38 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-269-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=269
--- Comment #18 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
urrrr okaaay...
i just added parser support for this:
MEM(EA, 4) <- GPR(r)[32:63]
which requires two nested depths of brackets:
GPR (r) level 1
GPR (r) [32:63] level 2
in the process, you can see i added a class Mem and also added
to the class GPR
https://git.libre-riscv.org/?p=soc.git;a=commitdiff;h=2ecec80b841136e052b6e7d39b76abaca77bb612
i don't know if you're familiar with __call__, michael: it allows
you to (literally) call an object as if it was a function.
so with gsc.gpr being an object of type GPR, rather than mess with
the parser any further i did two things:
1. assign a GPR instance into the dict d for the exec(), by key 'GPR'
2. add an __call__ to class GPR.
therefore, when the code sees 'GPR' - GPR(r) for example - it calls
GPR.__call__(5)
the return result from _that_ is then passed through a second level of
processing (a slice), and that gets assigned into Mem
an instance of Mem i *also* assigned into the dict d for the exec(),
by key 'MEM'.
this - again - through the Mem.__call__(addr, bytesize) trick - allows
*getting* memory.
however for setting i didn't want to overload dict because the key
is not just a single index (address), it's actually an address-bytesize
tuple:
MEM(EA, 2) <- RA
which, hmm, whilst that *could* be turned into this:
MEM[(EA,2)] <- RA
which would, if __setitem__ was overloaded, could allow for recognition
of the tuple, decoding it back into (addr, bytesize).
yuk, is my honest opinion :)
so instead, i "recognise" the pattern "MEM(addr, sz) <- x" back in the
parser "assignment" phase, and *REPLACE* it with this:
memassign(addr, sz, x)
which you'll see in... hmmm i'm moving class Mem into decoder/isa/caller.py
ok?
--
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 269] auto-conversion / parser of POWER ISA Spec v3.0B
From: bugzilla-daemon @ 2020-04-04 19:27 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-269-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=269
--- Comment #17 from Michael Nolan <mtnolan2640@gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #15)
> star. technically once that base class is in place it should be possible
> to go straight to dropping into internal_op.py
I don't see an internal_op.py anywhere, is this instead for internalop_sim.py?
--
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 269] auto-conversion / parser of POWER ISA Spec v3.0B
From: bugzilla-daemon @ 2020-04-04 16:45 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-269-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=269
--- Comment #16 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
i'm trying _not_ to do ISACaller because i really need to move on
to the hardware-parser version
however, i'm laying the groundwork:
* as you saw in the bug #272 there's a problem with EXTS, which alerted
me to adding the syntax (RA|0) - this meant adding a function
"getz" to the prototype class GPR, and also inventing a new syntax
"_REG" where "_REG" holds the actual bits of the opcode (RA, RB),
and "REG" holds the *value from the regfile*.
* accessing GPR as GPR(5) is a function: i've converted it to GPR[5]
then that alerted me to the fact that GPR and MEM do not exist in
the opcode-function namespace.
sooo i dragged out that namespace-injection-by-decorator trick,
and have added a helper function. this *may* be incorrect
(as in, self.namespace has to be inside the decorator, not as
an argument *to* the decorator).
i am going to keep chewing through the files. moving onto fixedstore.mdwn
--
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 278] New: POWER v3.0B spec ambiguity on EXTS and missing EXTZ
From: bugzilla-daemon @ 2020-04-04 14:17 UTC (permalink / raw)
To: libre-riscv-dev
http://bugs.libre-riscv.org/show_bug.cgi?id=278
Bug ID: 278
Summary: POWER v3.0B spec ambiguity on EXTS and missing EXTZ
Product: Libre-SOC's first SoC
Version: unspecified
Hardware: PC
OS: Linux
Status: CONFIRMED
Severity: enhancement
Priority: ---
Component: Specification
Assignee: lkcl@lkcl.net
Reporter: lkcl@lkcl.net
CC: libre-riscv-dev@lists.libre-riscv.org
NLnet milestone: ---
sent to openpower-hdl-cores
continuing with the implementation of a simulator that is taken
directly - and literally - from the pseudocode and definitions of
functions in the spec, we have hit an "inconvenient" definition.
EXTS, section 1.3.4, page 6:
EXTS(x) Result of extending x on the left with sign
bits
the problem is illustrated by asking the very simple question, "how many bits?"
also, note: a definition for EXTS64 and EXTZ, both of which are *used*
in the spec - and for which their definition may easily be deduced
[however that *defeats the purpose of having a spec*...] - are both
missing.
EXTZ suffers from the same problem as EXTS, namely the number of bits
to extend to, is missing.
we have a "class" (SelectableInt) which records both the value and the
*length* of the value (in number of bits), therefore determining where
*from* (to *begin* zero-extending or sign-extending) is not a problem.
where to *stop* extending is the problematic part.
this causes us some considerable inconvenience when it comes to
"deducing" where to stop sign/zero-extending. example, in cmprb:
src21hi <- EXTZ((RB) 32:39 )
if L=0 then
in_range <- (src22lo <= src1) & (src1 <= src22hi)
so.. um... we zero-extend RB bits 32:39, to an *unknown* - and
completely unspecified - length, then perform comparisons on those
values, not having any information?
now, yes, we know that zero compared to zero is obviously zero,
therefore we could arbitrarily pick some random length, here, and
"everything would work".
a less-obvious example
if AA then NIA <-iea EXTS(LI || 0b00)
else NIA <-iea CIA + EXTS(LI || 0b00)
i believe, here, that it is expected that the "deduction" of the
sign-extension length should be by way of knowing the bit-length of
NIA!
to that end, we can probably get away with EXTS returning an object
that is specifically marked, "ok, we have no idea what the actual
bit-length is going to be, so let's set it to 1,000,000 and work it
out when the object is actually used".
actually... we can't...
lbz RT,D(RA)
-------
if RA = 0 then b <- 0
else b <- (RA)
EA <- b + EXTS(D)
because when b=0, that's an integer of undefined arbitrary length!
now, if b had been assigned to 0-subscript-64 (64 bit zero) *then* we
would have the information needed.
so this actually turns out to be quite problematic. actually: in the
lbz case if the definition had been:
b <- (RA|0)
EA <- b + EXTS(D)
where the definition for (Reg|0) *is* defined, clearly (Notation
section 1.3.2, p4) then that would actually be fine! this because it
is obvious from the definition of (Reg|0) that the result is 64-bit.
with that being defined as 64-bit, b would become 64-bit, and thus
when EXTS(D) was added to b, the "unknown" length of EXTS(D) could be
resolved.
it would be *significantly* better if EXTS was retired in favour of
EXTS64 and EXTS32! i appreciate that this requires a careful and
comprehensive review of every single usage of EXTS.
perhaps i can help there by collecting debug information either from
actual execution or from the parser as to what bit-lengths EXTS (and
EXTZ) interact with?
--
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] 3.0B PDF spec parser and first auto-generator working well
From: Luke Kenneth Casson Leighton @ 2020-04-04 13:33 UTC (permalink / raw)
To: openpower-hdl-cores, Libre-RISCV General Development
making good progress. this (which can be seen here
https://libre-riscv.org/openpower/isa/fixedarith/)
vvvvvv
# Add Immediate Shifted
D-Form
* addis RT,RA,SI
if RA = 0 then RT <- EXTS(SI || [0]*16)
else RT <- (RA) + EXTS(SI || [0]*16)
Special Registers Altered:
None
^^^^^
gets turned into this:
vvvv
# auto-generated, do not edit
from soc.decoder.helpers import (EXTS,)
class fixedarith(ISACaller):
def op_addis(self, RA):
if RA == 0:
RT = EXTS(concat(SI, concat(0, repeat=16)))
else:
RT = RA + EXTS(concat(SI, concat(0, repeat=16)))
return (RT,)
instrs = {}
instrs['addis'] = (op_addis, ['RA'], [], ['RT']) # records input
and output regs
^^^^
no "laborious manually-written simulator code" was written. the next
phase is, obviously, to try actually executing these code-fragments.
that's what the base-class "ISACaller" is for.
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 276] SR NAND Latch needed in nmigen
From: bugzilla-daemon @ 2020-04-04 11:15 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-276-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=276
--- Comment #14 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to whitequark from comment #13)
> (In reply to Luke Kenneth Casson Leighton from comment #12)
> > could you let me know what amount you'd be happy to receive as a donation
> > from NLNet, for including $sr in cxxrtl?
>
> As it turns out, there's actually already support for `$dffsr` latch in
> cxxrtl, which is what you need but with an extra clock. So `$sr` support
> would be about ten lines of code I have to copy and paste from the `$dffsr`
> case.
*snort*. funny. well, you're spending time here, which is important as it
defines the scope, solves the problem, and, due to the critical importance
for the project, easily justifies giving you something.
--
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 276] SR NAND Latch needed in nmigen
From: bugzilla-daemon @ 2020-04-04 11:10 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-276-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=276
--- Comment #13 from whitequark@whitequark.org ---
(In reply to Luke Kenneth Casson Leighton from comment #12)
> could you let me know what amount you'd be happy to receive as a donation
> from NLNet, for including $sr in cxxrtl?
As it turns out, there's actually already support for `$dffsr` latch in cxxrtl,
which is what you need but with an extra clock. So `$sr` support would be about
ten lines of code I have to copy and paste from the `$dffsr` case.
--
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] submitted bugreport to upstream nmigen
From: Luke Kenneth Casson Leighton @ 2020-04-04 11:07 UTC (permalink / raw)
To: Libre-RISCV General Development
In-Reply-To: <67ac8a43-5610-b2e4-f648-a0af7d68a09b@paape.io>
On Sat, Apr 4, 2020 at 9:15 AM Frieder Paape <frieder@paape.io> wrote:
>
> There is an api endpoint for jobs
> <https://docs.gitlab.com/ee/api/jobs.html>. We could write a cronjob
> which polls the api every day for new jobs and then saves the logs for
> those jobs locally.
>
> BUT I personaly think instead of investing time into workarounds because
> we don't have gitlab hosted ourselves, we should rather invest time to
> host it ourselves. Why not try to host it on the new talos 2 server.
although it would be a nuisance, we may need to blow it away and do a
full reinstall at the bare metal (hypervisor testing, BE OS instead of
LE OS, and so on).
> Does it have a public IP? We will probably run into problems, but hey,
> we are building a CPU on that architecture, it would look nice if we
> host our development tools on such architecture as well. And if there
> are larger problems we could bring them to IBM, which probably use
> gitlab in high quantity and might have the muscle to ask for a quick
> solution.
hmm it's good thinking: we are however reliant on them.
> Hosting it on a donated x86 server I guess would be fine as well
i pay for the mythic-beasts vCPU (and make sure it's backed up).
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 276] SR NAND Latch needed in nmigen
From: bugzilla-daemon @ 2020-04-04 11:05 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-276-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=276
--- Comment #12 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to whitequark from comment #11)
> > i presume that given that yosys already has $sr, only nmigen will need augmenting to produce $sr?
>
> There's no augmenting necessary. I gave you the code above (the
> `Instance("$sr"...)` that causes nMigen to emit a Yosys `$sr` cell.
fantastic! oh wait - i missed comment 8 (doh - how??)
http://bugs.libre-riscv.org/show_bug.cgi?id=276#c8
ok so the simulation(s) would be what need to recognise that.
could you let me know what amount you'd be happy to receive as a donation
from NLNet, for including $sr in cxxrtl?
also, i appreciate you suggested that we do the exploration, however that
would mean that someone else (other than you) receives money for doing so :)
as we'd very much like to support your work - and because we are a small
team there is such a lot else to do - would you be willing to add a
SRNAND latch example into cxxrtl for example?
--
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 137] NLNet 2019 Video Acceleration Proposal 2019-10-031
From: bugzilla-daemon @ 2020-04-04 10:57 UTC (permalink / raw)
To: libre-riscv-dev
In-Reply-To: <bug-137-13@http.bugs.libre-riscv.org/>
http://bugs.libre-riscv.org/show_bug.cgi?id=137
--- Comment #17 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
Summary sentence for MoU
Video acceleration is a necessary component for any modern CPU and GPU. Given
the large amount of time the typical user spends on videos and its applications
like videoconferencing, a performant and power-efficient implementation is
necessary for wide adoption. With Zoom (etc.) now being critical to our modern
life, and full of security holes, full transparency in the video encode/decode
algorithms is more important than ever.
--
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