Libre RISC-V Mailing list Archives (accessible via public-inbox)
 help / color / Atom feed
* [libre-riscv-dev] [Bug 304] Define minimum viable interface set for 180nm ASIC
From: bugzilla-daemon @ 2020-05-20 22:14 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-304-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=304

--- Comment #6 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
conversation notes from Tim (Raptor Engineering)

priority:

* 2x SPI master
* 4x I2C master
* 2x UART
* 4x LPC master
* EINT* Serialsed IRQ (LPC, from PCIe) - 15 or so IRQs, these are
PC>PCIe>OpenPower mappings

Quad SPI raptorengineering
LPC bridge github raptorengineering
~

-- 
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 316] bperm TODO
From: bugzilla-daemon @ 2020-05-20 22:18 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-316-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=316

--- Comment #76 from Cole Poirier <colepoirier@gmail.com> ---
(In reply to Michael Nolan from comment #75)
> (In reply to Cole Poirier from comment #74)
> > Test passes, and code has been commit-pushed. Moving on to your suggestion
> > on today's kanban update of integrating the formal proof of bpermd:
> > 
> Cool. I've modified test_bpermd a bit so it exercises the module a bit
> better. Previously, you only had it do one iteration with two constants, and
> those constants made bpermd evaluate to 0. This isn't *really* sufficient to
> establish much confidence in the module, so I modified it to do ~20
> iterations of something that would evaluate to something other than 0. 
> 
> 
> 
> 
> > How should I go about integrating the formal proof of bpermd into the
> > logical pipeline? I've thought about it, and I am unsure as to how to
> > proceed. Is there a model I can follow as before?
> > 
> > I also want to follow up on a comment of Michael's from yesterday:
> > 
> > ```
> > From comment #50 of this bug
> > 
> > Michael Nolan 2020-05-19 20:24:19 BST
> > I'd say try hooking it up to main_stage and adding a test for it would be
> > better than trying to "optimize" the proof. Especially since there's a bug
> > in the module that the proof hasn't revealed (hint: it has to do with
> > power's numbering of bits).
> > ```
> > 
> > Is my proof of bpermd in fact not a proper proof, and incorrect because it's
> > making an mistake with endian-ness?
> 
> Well kinda. Your proof is written to match the Power ISA spec, and from a
> glance it looks like it does that. However, Power bit indices go from left
> to right, which is opposite what python and nmigen do. So although your
> proof looks like it matches the spec, it doesn't give correct results when
> it's actually tested. That's why I suggested you hook it up to the testbench
> , because the testbench compares it to some python pseudocode that *does*
> index bits correctly. 
> 
> So what I think you should do now is pull and run the testbench again, and
> see if you can't figure out how to get it to pass those tests. Then modify
> the proof correspondingly.

Thanks Michael, your help is yet again very helpful :) Unfortunately I won't be
able to do this until tomorrow morning because an exterminator appointment is
forcing me to be away from my computer. Looking forward to tackling this.

-- 
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 316] bperm TODO
From: bugzilla-daemon @ 2020-05-20 22:20 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-316-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=316

--- Comment #77 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
so, cole, here's an example: RS=0x4000000, RB=0xdeadcownonsense
the hardware produces the answer 0xf7, the *actual* expected answer
is supposed to be 0xff.

so you can now write a unit test which puts those exact values in
(oh look, fu/logical/test/test_bpermd.py is empty....) and play
with the code until you *do* get the answer "0xff".



bpermd 3, 1, 2
['RS', 'RB']
reading reg 1
reading reg 2
[SelectableInt(value=0x4000000, bits=64),
SelectableInt(value=0xdeadbeefcafec0de, bits=64)]
concat 8 SelectableInt(value=0x0, bits=8)
concat 56 SelectableInt(value=0x0, bits=56)
concat 1 SelectableInt(value=0xff, bits=64)
(SelectableInt(value=0xff, bits=64),)
writing reg 3 SelectableInt(value=0xff, bits=64)
expected ff, actual: f7

  File "fu/logical/test/test_pipe_caller.py", line 260, in process
    self.assertEqual(expected, alu_out, code)
AssertionError: 255 != 247 : bpermd 3, 1, 2

-- 
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 332] Formal correctness proof needed for CR pipeline
From: bugzilla-daemon @ 2020-05-21  0:19 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-332-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=332

--- Comment #3 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
* MCRF proof
https://git.libre-soc.org/?p=soc.git;a=commitdiff;h=b9a3b94d0ed88221b1ca19027bf8f23b806c8e8b
* MFCR proof
https://git.libre-soc.org/?p=soc.git;a=commitdiff;h=78e6e9f800a3dbfad854bf5a4afaf48307275960

-- 
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 316] bperm TODO
From: bugzilla-daemon @ 2020-05-21  0:25 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-316-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=316

--- Comment #78 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
https://git.libre-soc.org/?p=soc.git;a=blob;f=src/soc/fu/cr/formal/proof_main_stage.py;h=d109f81f599c335986ef8a847246796cfa40e996

hey cole, psst, see line 61...

-- 
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 330] create POWER9 Logic Pipeline
From: bugzilla-daemon @ 2020-05-21  0:57 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-330-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=330

--- Comment #1 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
i'm not seeing anything in main_stage.py that uses XER carry_in.  it appears
extraneous.  sticky overflow on the other hand, because carry can be generated,
should be.

-- 
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 334] New: POWER decode A=zero needs to be set as a flag in Execute1Type
From: bugzilla-daemon @ 2020-05-21  1:57 UTC (permalink / raw)
  To: libre-riscv-dev

https://bugs.libre-soc.org/show_bug.cgi?id=334

            Bug ID: 334
           Summary: POWER decode A=zero needs to be set as a flag in
                    Execute1Type
           Product: Libre-SOC's first SoC
           Version: unspecified
          Hardware: Other
                OS: Linux
            Status: CONFIRMED
          Severity: enhancement
          Priority: ---
         Component: Source Code
          Assignee: lkcl@lkcl.net
          Reporter: lkcl@lkcl.net
                CC: libre-riscv-dev@lists.libre-riscv.org
   NLnet milestone: ---

https://git.libre-soc.org/?p=soc.git;a=blob;f=src/soc/decoder/power_decoder2.py;h=3b9beba954a2f0f7ebefc7a731d8eddcb3705698;hb=HEAD

detection of zero RA reg needs to be passed through the CompUnits just as B
being used as an immediate.

this therefore needs to be done in the unit tests, qemu to check the
simulation, then HDL unit tests detect RA_OR_ZERO and likewise set A to zero.

a flag "a_is_zero" or such would be set in power_decode2 in the execute phase.

-- 
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 333] investigate why CR pipeline code took 100% CPU and locked up generating ILANG
From: bugzilla-daemon @ 2020-05-21  2:05 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-333-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=333

--- Comment #9 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
just as spr and reg is decoded in power_decode2.py for RA RB RC and output, csv
files would have "reg is CR" enum option.

each DecodeRegA (etc) would then note that and place BFA insn fields (etc) into
not a reg_data or an spr_data but a cr_data field and indicate that they are
valid.

these - 3 - CR reg fields then get put into Execute1Type just like reg and spr
data.

CrOpSubset (to be created) then picks those up.

the instruction issue would also note them and request DM Matrix Reservations
just like it would for RA RB RC RS/T except for the Condition Regfile (and
Matrix) not the INT regfile.

-- 
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 305] Create Pipelined ALU similar to alu_hier.py
From: bugzilla-daemon @ 2020-05-21 10:31 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-305-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=305

--- Comment #82 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
apologies, michael, i've introduced a bug somewhere in ALU:

adde. 5, 6, 7
['RA', 'RB']
reading reg 6
reading reg 7
[SelectableInt(value=0xc97cd5181acb17fc, bits=64),
SelectableInt(value=0x9de943b0d85daaf9, bits=64)]
(SelectableInt(value=0x676618c8f328c2f6, bits=64),)
concat 1 SelectableInt(value=0x4, bits=4)
[SelectableInt(value=0x1, bits=1), SelectableInt(value=0x1, bits=1)]
__eq__ SelectableInt(value=0x0, bits=1) SelectableInt(value=0x1, bits=1)
__eq__ SelectableInt(value=0x0, bits=1) SelectableInt(value=0x1, bits=1)
writing reg 5 SelectableInt(value=0x676618c8f328c2f6, bits=64)
expected 676618c8f328c2f6, actual: 676618c8f328c2f6

-- 
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 318] fix LDSTCompUnit
From: bugzilla-daemon @ 2020-05-21 12:05 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-318-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=318

--- Comment #19 from Cesar Strauss <cestrauss@gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #18)
> Cesar, fyi:
> 
> https://git.libre-soc.org/?p=soc.git;a=commitdiff;h=c501606b20c74050fb0d422ff886a7d9fb2a6a96
> 
> very straightforward, no disruption at all, due to a hack
> of making MultiCompUnit look exactly like it was.

Dear Luke,

I see you just committed:
https://git.libre-soc.org/?p=soc.git;a=commitdiff;h=1dbb3d17d051f73a20a148740be7b65c1a1778c1

I had the impression, maybe wrongly, that you had assigned this task to me.

I had my own version mostly working. I will compare mine to yours, and make a
review of yours.

I guess I failed to explicitly accept the (supposed) assignment, and report my
progress. Sorry about that. Lesson learned.

Too bad about the duplicate effort. No worries. At least I had a good training
on nMigen, RecordObject and the CompUnit API.

On the way, I did manage to catch an obvious typo in your previous refactoring
of compalu_multi.py. I took the liberty to commit it:

https://git.libre-soc.org/?p=soc.git;a=commitdiff;h=d9724d6fed90959a78799c6e48667d31abd3a091 

I'm going back now to improve the unit test.

Again, no worries.

Regards,

Cesar

-- 
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 318] fix LDSTCompUnit
From: bugzilla-daemon @ 2020-05-21 12:58 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-318-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=318

--- Comment #20 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Cesar Strauss from comment #19)
> (In reply to Luke Kenneth Casson Leighton from comment #18)
> > Cesar, fyi:
> > 
> > https://git.libre-soc.org/?p=soc.git;a=commitdiff;h=c501606b20c74050fb0d422ff886a7d9fb2a6a96
> > 
> > very straightforward, no disruption at all, due to a hack
> > of making MultiCompUnit look exactly like it was.
> 
> Dear Luke,
> 
> I see you just committed:
> https://git.libre-soc.org/?p=soc.git;a=commitdiff;
> h=1dbb3d17d051f73a20a148740be7b65c1a1778c1

yes.  i meant to write here this morning, but got side-tracked on pipeline
code-morphing.

> I had the impression, maybe wrongly, that you had assigned this task to me.

ah.  sorry!  if you had written explicitly "i will do this conversion tomorrow 
evening", i would have held off.

> I had my own version mostly working. I will compare mine to yours, and make
> a review of yours.

appreciated.

> I guess I failed to explicitly accept the (supposed) assignment, and report
> my progress. Sorry about that. Lesson learned.

by both of us.

> Too bad about the duplicate effort. No worries. At least I had a good
> training on nMigen, RecordObject and the CompUnit API.

about the CompUnit API: if you look at CompUnitBase you can see where i'm
going with it.
https://git.libre-soc.org/?p=soc.git;a=blob;f=src/soc/experiment/score6600_multi.py;h=4630368e474c9d7532dad3b00a3983435ba01c7b;hb=7b0b09aef28e4329c6b85d933b022112f0c15c50#l44

CompUnitBase is a mess.  assigning individual signals from a batch of CompUnits
to create a common data structure.   i am hoping - planning - to find a way
to merge the units into a common data structure.  or, if not, at the very
least a common *python* level API which allows convenient signal-connectivity
without having a hundred lines of tediously repetitive signal-for-signal 
assignments.



about RecordObject: RecordObject exists because whitequark wants Record to be
bit-oriented, particularly when it comes to enumerating (using __iter__)
through it.

whereas what we need - to avoid insanely complex graphviz diagrams that are
flat-out impossible to read - is *field*-orientated iteration and assignment.

*field*-orientated iteration allows nmigen to generate code that, when you
assign one RecordObject to another, does so by creating eq()s on the *members*
of the RecordObject, not the *individual bits of every single field*.

also, RecordObject looks pretty much exactly like a standard python class,
and may be sub-classed (as you can see from LDSTCompUnitRecord) and extended
in a standard python fashion.

Record (because it takes an explicit layout argument) may *NOT* be extended
except by direct manipulation of Record.fields, which can have consequences
when people assume that the use of any given Record instance will have exactly
the same structure.

basically, RecordObject fits python developer expectations: Record does not.


> On the way, I did manage to catch an obvious typo in your previous
> refactoring of compalu_multi.py. I took the liberty to commit it:
> 
> https://git.libre-soc.org/?p=soc.git;a=commitdiff;
> h=d9724d6fed90959a78799c6e48667d31abd3a091 

yes i saw - that would have been hard to track down: the unit test didn't
catch it.  really appreciated.

> I'm going back now to improve the unit test.

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 198] Formal correctness proofs are needed for low-level libraries in LibreSOC
From: bugzilla-daemon @ 2020-05-21 13:01 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-198-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=198

Luke Kenneth Casson Leighton <lkcl@lkcl.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   budget (EUR) for|7650                        |8650
 completion of task|                            |
   (excludes budget|                            |
       allocated to|                            |
          subtasks)|                            |

-- 
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 306] Formal Correctness Proof for ALU pipeline
From: bugzilla-daemon @ 2020-05-21 13:01 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-306-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=306

Luke Kenneth Casson Leighton <lkcl@lkcl.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
    parent task for|198                         |195
  budget allocation|                            |

-- 
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 331] Formal Correctness Proof for LOGICAL pipeline
From: bugzilla-daemon @ 2020-05-21 13:01 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-331-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=331

Luke Kenneth Casson Leighton <lkcl@lkcl.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
    parent task for|198                         |195
  budget allocation|                            |

-- 
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 332] Formal correctness proof needed for CR pipeline
From: bugzilla-daemon @ 2020-05-21 13:01 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-332-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=332

Luke Kenneth Casson Leighton <lkcl@lkcl.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
    parent task for|198                         |195
  budget allocation|                            |

-- 
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 195] Formal correctness framework is needed for Power ISA
From: bugzilla-daemon @ 2020-05-21 13:02 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-195-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=195

Luke Kenneth Casson Leighton <lkcl@lkcl.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   budget (EUR) for|0                           |11000
 completion of task|                            |
   (excludes budget|                            |
       allocated to|                            |
          subtasks)|                            |

-- 
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 335] New: Formal Correctness Proof for Branch pipeline
From: bugzilla-daemon @ 2020-05-21 13:04 UTC (permalink / raw)
  To: libre-riscv-dev

https://bugs.libre-soc.org/show_bug.cgi?id=335

            Bug ID: 335
           Summary: Formal Correctness Proof for Branch pipeline
           Product: Libre-SOC's first SoC
           Version: unspecified
          Hardware: PC
                OS: Mac OS
            Status: CONFIRMED
          Severity: enhancement
          Priority: ---
         Component: Formal Verification
          Assignee: lkcl@lkcl.net
          Reporter: lkcl@lkcl.net
                CC: libre-riscv-dev@lists.libre-riscv.org
   NLnet milestone: ---

a formal correctness proof is needed for the branch pipeline.

-- 
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 335] Formal Correctness Proof for Branch pipeline
From: bugzilla-daemon @ 2020-05-21 13:05 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-335-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=335

Luke Kenneth Casson Leighton <lkcl@lkcl.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 total budget (EUR)|0                           |300
  for completion of|                            |
       task and all|                            |
           subtasks|                            |
           Assignee|lkcl@lkcl.net               |mtnolan2640@gmail.com
    NLnet milestone|---                         |NLNet.2019.10.Formal
             Blocks|                            |313
    parent task for|                            |195
  budget allocation|                            |


Referenced Bugs:

https://bugs.libre-soc.org/show_bug.cgi?id=313
[Bug 313] Create Branch Pipeline for POWER9
-- 
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 195] Formal correctness framework is needed for Power ISA
From: bugzilla-daemon @ 2020-05-21 13:05 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-195-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=195

Luke Kenneth Casson Leighton <lkcl@lkcl.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   budget (EUR) for|11000                       |10700
 completion of task|                            |
   (excludes budget|                            |
       allocated to|                            |
          subtasks)|                            |

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

^ permalink raw reply

* [libre-riscv-dev] daily kan-ban update 21may2020
From: Luke Kenneth Casson Leighton @ 2020-05-21 13:42 UTC (permalink / raw)
  To: Libre-RISCV General Development

yesterday:

* tracking cole and michael's pipeline work.  kept up with bugtracking.
* did the planned CompUnitRecord which begins a Comp Unit API
* added "register specifications" to all the pipeline Input/Output
data structures
* got annoyed at saying "no" to Jacob about a monorepo.  as in: i was annoyed
  *that* i had to say "no".  i don't like it: i'd rather say "yes" to
at least something.
* continued a long-running discussion on comp.arch about a term that is
  entirely missing from Computer Science.  "Phase-aware Function Unit"
  seems to cover it.  took several *days* to establish.
* attended the OpenPOWER-LibreSOC bi-weekly meeting, which highlighted
  a crucial fact that 3.1B UNIX Platform spec currently requires SIMD, and thus
  excludes both LibreSOC and microwatt.  also defined some interfaces (thanks
  to Tim).
* contacted an associate who will be able to explain things to Marketnext about
  the developer install dependencies.
* discussed ideas with Cole about XER and Condition Register fields.

this morni... earlier today:

* created LDSTCompUnitRecord (sorry Cesar!) derived from CompUnitRecord
* rearranged the pipeline code slightly / incrementally to create common classes
* put the "register specifications" into the pipelines, ready for auto-creating
  their association with the (multiple) register files.

rest of today:

think.  still trying to work out a suitable structure for how to
connect multiple (completely different) CompUnits up to regfiles.

blockers:

* DIV and MUL pipelines (Jacob)
* TRAP and SYS pipelines (Cole)
* LDSTCompUnit multi-unit testing (actually probably L0CacheBuffer
issue) (Cesar)
* L0CacheBuffer (Jacob)

l.

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

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 21may2020
From: Michael Nolan @ 2020-05-21 13:52 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <CAPweEDzB_kCw2ugMzJH4WE7sLTb6keYo1ymYQBAZAKTBKvBAuQ@mail.gmail.com>

This morning I fixed the broken test_adde (it broke because I hadn't 
added overflow support to caller.py). I think today I'll work on 
changing the CR unit to take in the individual cr bitfields instead of 
the whole CR as well as adding the necessary bits to the decoder to 
handle that. If I finish that, I'll work on more formal verification.

--Michael



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

^ permalink raw reply

* Re: [libre-riscv-dev] daily kan-ban update 21may2020
From: Luke Kenneth Casson Leighton @ 2020-05-21 15:35 UTC (permalink / raw)
  To: Libre-RISCV General Development
In-Reply-To: <a4446783-4dd8-67c9-e987-b181c94399ed@gmail.com>

On Thu, May 21, 2020 at 2:52 PM Michael Nolan <mtnolan2640@gmail.com> wrote:
>
> This morning I fixed the broken test_adde (it broke because I hadn't
> added overflow support to caller.py).

ohh it was in the simulation code.  ok.

> I think today I'll work on
> changing the CR unit to take in the individual cr bitfields instead of
> the whole CR as well as adding the necessary bits to the decoder to
> handle that.

ok.

> If I finish that, I'll work on more formal verification.

great.

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 333] investigate why CR pipeline code took 100% CPU and locked up generating ILANG
From: bugzilla-daemon @ 2020-05-21 16:02 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-333-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=333

--- Comment #10 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
hm need to decide the register allocation scheme.  it is ok to have 4bit CR go
into a 32bit operand but not to have CR data go down an INT path.

or, if we do, MUXes are needed to cross data from one series of regfile paths
to another.

this could be done for when we have the cyclic buffers in place, easily.

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

^ permalink raw reply

* [libre-riscv-dev] [Bug 336] New: add indicator to Decode2ExecuteType that RA is zero
From: bugzilla-daemon @ 2020-05-21 16:24 UTC (permalink / raw)
  To: libre-riscv-dev

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

            Bug ID: 336
           Summary: add indicator to Decode2ExecuteType that RA is zero
           Product: Libre-SOC's first SoC
           Version: unspecified
          Hardware: Other
                OS: Linux
            Status: CONFIRMED
          Severity: enhancement
          Priority: ---
         Component: Source Code
          Assignee: lkcl@lkcl.net
          Reporter: lkcl@lkcl.net
                CC: libre-riscv-dev@lists.libre-riscv.org
   NLnet milestone: ---

whilst DecodeA picks up that RA may be a zero immediate, this is not actually
passed through Decode2ExecuteType, nor does the ALU CompUnit recognise it.

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

^ permalink raw reply

* [libre-riscv-dev] [Bug 336] ALU CompUnit needs to recognise that RA (src1) can be zero
From: bugzilla-daemon @ 2020-05-21 16:35 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-336-13@https.bugs.libre-soc.org/>

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

Luke Kenneth Casson Leighton <lkcl@lkcl.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Depends on|                            |334
            Summary|add indicator to            |ALU CompUnit needs to
                   |Decode2ExecuteType that RA  |recognise that RA (src1)
                   |is zero                     |can be zero

--- Comment #1 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
forgot that i had raised 334 last night.  this one can therefore be about the
CompUnit.

the change needed is:

* first that Decode2ExecuteType needs a flag ra_is_zero

* second that CompAluOpSubset and CompLogicalALUSubset need likewise

* third - and this is the kicker - that the FSM for src1_i needs a Mux to
select RA or zero *and* it needs to *no longer* request a READ for src1
(rd.req[0]) if the ra_is_zero flag is set.

* additionally that if the z flag is set, the rd_done part of the FSM does NOT
wait for GO_READ on the src1 port (rd.go[0])
i will draw out a diagram.

again.

:)


Referenced Bugs:

https://bugs.libre-soc.org/show_bug.cgi?id=334
[Bug 334] POWER decode A=zero needs to be set as a flag in Execute1Type
-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

page:   latest

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

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

Example config snippet for mirrors


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