Sign in

username:

password:



Not a member?

Search fpga-cpu



Search tips

Subscribe to fpga-cpu



fpga-cpu by Keywords

Altera | CISCifying | IDE | ISA | Java | JHDL | JTAG | LBU | MicroBlaze | PAR | PCI | RISC | SoC | Spartan | Transputers | Verilog | VHDL | Virtex | VLIW | WebPack | Xilinx | Xsoc | YARD-1A

Discussion Groups

Discussion Groups | FPGA-CPU | Re: New minimal CPU design

This list is for discussion of the design and implementation of field-programmable gate array based processors and integrated systems. It is also for discussion and community support of the XSOC Project (see http://www.fpgacpu.org/xsoc).

New minimal CPU design - Tim Boescke - Oct 20 17:56:00 2001


I just wanted to announce a new minimal CPU designed
by me some time ago.

Its special in the respect, that it fits into a 32
macrocell CPLD. Even though the instruction set
is severly limited I think that the code density
and cpu performance is on par with other minimal
'educational' CPUs, which are sometimes quite a bit
larger in respect to gate/register usage.

Features:

-8 bit data.
-6 bit adressbus allowing 64 bytes of shared program/data memory.
-Programming model: Accumulator based. Registers are Accu[7:0], PC[5:0] and
a carry flag.
-4 Instructions: NOR, ADD, STA, JCC. Instructions execute in 1-2 clock
cycles.

The CPU code, along with further documentation,
simulator and examples can be found here:

http://www.tu-harburg.de/~setb0209/cpu/

So, does anybody come up with a smaller CPU ? Comments are appreciated.






(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Ben Franchuk - Oct 20 20:39:00 2001

Tim Boescke wrote:
>
> I just wanted to announce a new minimal CPU designed
> by me some time ago.
<cut>
> -8 bit data.
> -6 bit adressbus allowing 64 bytes of shared program/data memory.
> -Programming model: Accumulator based. Registers are Accu[7:0], PC[5:0] and
> a carry flag.
> -4 Instructions: NOR, ADD, STA, JCC. Instructions execute in 1-2 clock
> cycles.
<snip>
> So, does anybody come up with a smaller CPU ?

Years ago Byte magazine had a similar design in TTL. I think
the instructions where
0) Load input 1) Jump minus 2) Add 3) Store Compliment
Ben Franchuk.


(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

RE: New minimal CPU design - Jaime Andrés Aranguren Cardona - Oct 21 10:09:00 2001

Anybody interested on the subject, we had another project called "FiJaaC
processor". You can check it out a www.geocities.com/jaime_aranguren/FiJaaC
or www.geocities.com/faidoll It's a basic 8 bit CPU on a Xilinx SpartanXL
XCS05XL-PC84.

Cheers,

> -----Mensaje original-----
> De: [mailto:]En
> nombre de Ben Franchuk
> Enviado el: Sábado, 20 de Octubre de 2001 08:39 p.m.
> Para:
> Asunto: Re: [fpga-cpu] New minimal CPU design > Tim Boescke wrote:
> >
> > I just wanted to announce a new minimal CPU designed
> > by me some time ago.
> <cut>
> > -8 bit data.
> > -6 bit adressbus allowing 64 bytes of shared program/data memory.
> > -Programming model: Accumulator based. Registers are Accu[7:0],
> PC[5:0] and
> > a carry flag.
> > -4 Instructions: NOR, ADD, STA, JCC. Instructions execute in 1-2 clock
> > cycles.
> <snip>
> > So, does anybody come up with a smaller CPU ?
>
> Years ago Byte magazine had a similar design in TTL. I think
> the instructions where
> 0) Load input 1) Jump minus 2) Add 3) Store Compliment
> Ben Franchuk.
>
> To Post a message, send it to:
> To Unsubscribe, send a blank message to:





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Tim Boescke - Oct 21 11:21:00 2001


> >
> > I just wanted to announce a new minimal CPU designed
> > by me some time ago.
>
> Years ago Byte magazine had a similar design in TTL. I think
> the instructions where
> 0) Load input 1) Jump minus 2) Add 3) Store Compliment

Thats interesting. Do you know the exact issue ?

But well, actually I was rather referring to the design
being minimal in respect to resource usage. 31 CPLD macrocells
are maybe comparable to ~16 Spartan CLBs, which is even less than
20% of a Spartan XCS05. (But you cant compare CPLDs and FPGAs
anyways.. the design is optimized for CPLD usage and would
be suboptimal for a FPGA).

Minimal programming models are well known.(subtract-and-branch-
if-negative, move machine etc..)

I'd be interested to know about minimal FPGA cpu designs.
Since the ALU can be minimized easily due to carry chains,
the main problem is probably to optimize the control.

The datapath of the above design should fit into 4+1+3+3=11
CLBs (Alu,Carry,Pc,Adreg). However first attempts to synthesize
the design for a Spartan I revealed a CLB count of 24, so that
13 CLBs are wasted on control and unnecessary constructs. I have to
look a bit deeper into it.





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

RE: New minimal CPU design - Leon Heller - Oct 21 13:07:00 2001

>
>Anybody interested on the subject, we had another project called "FiJaaC
>processor". You can check it out a www.geocities.com/jaime_aranguren/FiJaaC
>or www.geocities.com/faidoll It's a basic 8 bit CPU on a Xilinx SpartanXL
>XCS05XL-PC84.

I couldn't find any code for the CPU on the web site.

Leon
--
Leon Heller, G1HSM Tel: +44 1327 359058 Email:
My web page: http://www.geocities.com/leon_heller
My low-cost Altera Flex design kit: http://www.leonheller.com _________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp




(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Ben Franchuk - Oct 21 13:57:00 2001

Tim Boescke wrote:
> Thats interesting. Do you know the exact issue ?

No ... I lost all my old computer magazines a few years back do
to water damage in storage. > But well, actually I was rather referring to the design
> being minimal in respect to resource usage. 31 CPLD macrocells
> are maybe comparable to ~16 Spartan CLBs, which is even less than
> 20% of a Spartan XCS05. (But you cant compare CPLDs and FPGAs
> anyways.. the design is optimized for CPLD usage and would
> be suboptimal for a FPGA).

That is hard to say what a good small model is . A shift right
and
and jump zero would be nice instructions if you had the room.

> Minimal programming models are well known.(subtract-and-branch-
> if-negative, move machine etc..)
>
> I'd be interested to know about minimal FPGA cpu designs.
> Since the ALU can be minimized easily due to carry chains,
> the main problem is probably to optimize the control.

Historically 12 bits have been the minimum real cpu design.
I suspect a clean design
- ADD ac = ac + n , toggle carry
- NOR ac = ac nor n
- SHR ac = rcr ac , shift carry
- DCA ac = 0 ; n = ac
- JC if(cy) pc = n
-JZ if(z) pc = n
-JMP pc = n
-JSR ac = pc ,pc = n

would be about three times the size of a small 8 bit processor.

> The datapath of the above design should fit into 4+1+3+3=11
> CLBs (Alu,Carry,Pc,Adreg). However first attempts to synthesize
> the design for a Spartan I revealed a CLB count of 24, so that
> 13 CLBs are wasted on control and unnecessary constructs. I have to
> look a bit deeper into it.

It is nice to see somebody working with the small sized FPGA's
rather than the
MegAGate-super-deluxe-do-all chips.
Ben Franchuk.
--
Standard Disclaimer : 97% speculation 2% bad grammar 1% facts.
"Pre-historic Cpu's" http://www.jetnet.ab.ca/users/bfranchuk
Now with schematics.





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

RE: New minimal CPU design - Jaime Andrés Aranguren Cardona - Oct 22 0:05:00 2001

OK, He haven't finished it yet, still are testing the Control Unit (the last
part). By the way, if you can point us on how to get VHDL code from Xilinx'
CORE Generator Dual Ported RAM models, we can give the complete VHDL code of
it.

For size issues, we had to use such an implementation for the Registers
Bank, 'cause our VHDL implementation resulted to be quite big to fit on the
XCS05XL-PC84. In that order of ideas, we also implemented some multiplexers
as LogiBlox modules, because they can be optimized for area. The control
unit is completely VHDL.

Anyway, as soon as we finish our project (that should be in about one or two
weeks), we will post on the pages the whole project sources. We use Xilinx'
Foundation 3.1i for development, although some other tools as Renoir99 and
ModelSim have been also used.

Cheers,

Jaime.

> -----Mensaje original-----
> De: Leon Heller [mailto:]
> Enviado el: Domingo, 21 de Octubre de 2001 01:07 p.m.
> Para:
> Asunto: RE: [fpga-cpu] New minimal CPU design > >
> >Anybody interested on the subject, we had another project called "FiJaaC
> >processor". You can check it out a
> www.geocities.com/jaime_aranguren/FiJaaC
> >or www.geocities.com/faidoll It's a basic 8 bit CPU on a Xilinx SpartanXL
> >XCS05XL-PC84.
>
> I couldn't find any code for the CPU on the web site.
>
> Leon
> --
> Leon Heller, G1HSM Tel: +44 1327 359058 Email:
> My web page: http://www.geocities.com/leon_heller
> My low-cost Altera Flex design kit: http://www.leonheller.com > _________________________________________________________________
> Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp > To Post a message, send it to:
> To Unsubscribe, send a blank message to:





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

RE: New minimal CPU design - Jaime Andrés Aranguren Cardona - Oct 22 0:07:00 2001

Furthermore, if any of you are interested on it, we will be glad on sharing
our project via e-mail. before it gets on the web. And of course, all of
your suggestions and comments are very, very welcomed.

Best regards,

Jaime

> -----Mensaje original-----
> De: Leon Heller [mailto:]
> Enviado el: Domingo, 21 de Octubre de 2001 01:07 p.m.
> Para:
> Asunto: RE: [fpga-cpu] New minimal CPU design > >
> >Anybody interested on the subject, we had another project called "FiJaaC
> >processor". You can check it out a
> www.geocities.com/jaime_aranguren/FiJaaC
> >or www.geocities.com/faidoll It's a basic 8 bit CPU on a Xilinx SpartanXL
> >XCS05XL-PC84.
>
> I couldn't find any code for the CPU on the web site.
>
> Leon
> --
> Leon Heller, G1HSM Tel: +44 1327 359058 Email:
> My web page: http://www.geocities.com/leon_heller
> My low-cost Altera Flex design kit: http://www.leonheller.com > _________________________________________________________________
> Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp > To Post a message, send it to:
> To Unsubscribe, send a blank message to:




(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Felix Bertram - Oct 22 9:32:00 2001

--- In fpga-cpu@y..., Jaime Andrés Aranguren Cardona
<jaime.aranguren@i...> wrote:
> part). By the way, if you can point us on how to get VHDL code from
Xilinx'
> CORE Generator Dual Ported RAM models, we can give the complete
VHDL code of
> it.

You can easily create VHDL code for RAMs by manually instantiating
RAMB4_Sxx entities. See the code below implementing a 256x32 dual
ported RAM and excuse me for the line breaks.

Hope this helps,
best regards

Felix
_____
Dipl.-Ing. Felix Bertram
Trenz Electronic
Duenner Kirchweg 77
D - 32257 Buende
Tel.: +49 (0) 5223 49 39 755
Fax.: +49 (0) 5223 48 945
Mailto:
http://www.trenz-electronic.de

----------------------------------------------------------------------
----------

library IEEE;
use IEEE.STD_LOGIC_1164.all;

library UNISIM;
use UNISIM.vcomponents.all;

entity dpram256x32 is
port (
clka: IN std_logic;
addra: IN std_logic_VECTOR(7 downto 0);
dia: IN std_logic_VECTOR(31 downto 0);
ena: IN std_logic;
wea: IN std_logic;
--
clkb: IN std_logic;
addrb: IN std_logic_VECTOR(7 downto 0);
dob: OUT std_logic_VECTOR(31 downto 0);
enb: IN std_logic);
end dpram256x32;

----------------------------------------------------------------------
----------
architecture bhv of dpram256x32 is
constant zeroC: STD_LOGIC:= '0';
signal zero: STD_LOGIC;
signal zerobus: STD_LOGIC_VECTOR(15 downto 0);
begin
zero <= zeroC;
zerobus<= (others=> zeroC);

U1: RAMB4_S16_S16 port map(
CLKA=> clka, RSTA=> zero, ADDRA=> addra, DIA=> dia(15
downto 0), DOA=> open, ENA=> ena, WEA=> wea,
CLKB=> clkb, RSTB=> zero, ADDRB=> addrb, DIB=>
zerobus, DOB=> dob(15 downto 0), ENB=> enb, WEB=> zero);

U2: RAMB4_S16_S16 port map(
CLKA=> clka, RSTA=> zero, ADDRA=> addra, DIA=> dia(31
downto 16), DOA=> open, ENA=> ena, WEA=> wea,
CLKB=> clkb, RSTB=> zero, ADDRB=> addrb, DIB=>
zerobus, DOB=> dob(31 downto 16), ENB=> enb, WEB=> zero);

end bhv;

----------------------------------------------------------------------
----------
-- end of file






(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Tim Boescke - Oct 22 11:11:00 2001

> > But well, actually I was rather referring to the design
> > being minimal in respect to resource usage. 31 CPLD macrocells
> > are maybe comparable to ~16 Spartan CLBs, which is even less than
> > 20% of a Spartan XCS05. (But you cant compare CPLDs and FPGAs
> > anyways.. the design is optimized for CPLD usage and would
> > be suboptimal for a FPGA).
>
> That is hard to say what a good small model is . A shift right
> and
> and jump zero would be nice instructions if you had the room.

No chance to get anything else into the CPLD, there is not even
enough room left for a synch. reset. :)

And btw. are there really that many uses for a zero flag ? I see
it being useful for bit comparisons, but in most arithmetic routines
it is sufficient to have a greater-than or less-than compare.

The equal-to compare can easily be done by an add #255, jcc ..

Wide NOR gates can eat quite a bit of resources.

> > I'd be interested to know about minimal FPGA cpu designs.
> > Since the ALU can be minimized easily due to carry chains,
> > the main problem is probably to optimize the control.
>
> Historically 12 bits have been the minimum real cpu design.

> I suspect a clean design
> - ADD ac = ac + n , toggle carry
> - NOR ac = ac nor n
> - SHR ac = rcr ac , shift carry
> - DCA ac = 0 ; n = ac
> - JC if(cy) pc = n
> -JZ if(z) pc = n
> -JMP pc = n
> -JSR ac = pc ,pc = n
>
> would be about three times the size of a small 8 bit processor.

Hm.. that would imply a 9bit PC. The datapath would probably
take 12+1+9+2=24 CLbs (ALU, Carry, PC/Adreg, Z-Flag ) thats not
too much. The control probably would not increase that much over
an 8 Bit CPU, so it is maybe just twice as big.

Jan Gray somehow managed to fit a 4 operator ALU into a single
LUT per bit. (In the GR0040) Is there any documentation on
this ? So maybe there is a way to reduce the ALU size to 6 CLBs.

The DCA instruction is neat, however when I looked at the example
for my CPU, which uses STA, I noticed that none of them could
benefit from a DCA instruction.





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

RE: Re: New minimal CPU design - Jaime Andrés Aranguren Cardona - Oct 22 19:32:00 2001

Dear Mr. Bertram,

What I meant was to istantiate the DPRAM generated by the COREGenerator and
the Muxes and other stuff generated by LogiBlox, so (I think) I don't get or
should write any architecture, isn't it?

Regards,

> -----Mensaje original-----
> De: Felix Bertram [mailto:]
> Enviado el: Lunes, 22 de Octubre de 2001 09:33 a.m.
> Para:
> Asunto: [fpga-cpu] Re: New minimal CPU design > --- In fpga-cpu@y..., Jaime Andrés Aranguren Cardona
> <jaime.aranguren@i...> wrote:
> > part). By the way, if you can point us on how to get VHDL code from
> Xilinx'
> > CORE Generator Dual Ported RAM models, we can give the complete
> VHDL code of
> > it.
>
> You can easily create VHDL code for RAMs by manually instantiating
> RAMB4_Sxx entities. See the code below implementing a 256x32 dual
> ported RAM and excuse me for the line breaks.
>
> Hope this helps,
> best regards
>
> Felix
> _____
> Dipl.-Ing. Felix Bertram
> Trenz Electronic
> Duenner Kirchweg 77
> D - 32257 Buende
> Tel.: +49 (0) 5223 49 39 755
> Fax.: +49 (0) 5223 48 945
> Mailto:
> http://www.trenz-electronic.de
>
> ----------------------------------------------------------------------
> ----------
>
> library IEEE;
> use IEEE.STD_LOGIC_1164.all;
>
> library UNISIM;
> use UNISIM.vcomponents.all;
>
> entity dpram256x32 is
> port (
> clka: IN std_logic;
> addra: IN std_logic_VECTOR(7 downto 0);
> dia: IN std_logic_VECTOR(31 downto 0);
> ena: IN std_logic;
> wea: IN std_logic;
> --
> clkb: IN std_logic;
> addrb: IN std_logic_VECTOR(7 downto 0);
> dob: OUT std_logic_VECTOR(31 downto 0);
> enb: IN std_logic);
> end dpram256x32;
>
> ----------------------------------------------------------------------
> ----------
> architecture bhv of dpram256x32 is
> constant zeroC: STD_LOGIC:= '0';
> signal zero: STD_LOGIC;
> signal zerobus: STD_LOGIC_VECTOR(15 downto 0);
> begin
> zero <= zeroC;
> zerobus<= (others=> zeroC);
>
> U1: RAMB4_S16_S16 port map(
> CLKA=> clka, RSTA=> zero, ADDRA=> addra, DIA=> dia(15
> downto 0), DOA=> open, ENA=> ena, WEA=> wea,
> CLKB=> clkb, RSTB=> zero, ADDRB=> addrb, DIB=>
> zerobus, DOB=> dob(15 downto 0), ENB=> enb, WEB=> zero);
>
> U2: RAMB4_S16_S16 port map(
> CLKA=> clka, RSTA=> zero, ADDRA=> addra, DIA=> dia(31
> downto 16), DOA=> open, ENA=> ena, WEA=> wea,
> CLKB=> clkb, RSTB=> zero, ADDRB=> addrb, DIB=>
> zerobus, DOB=> dob(31 downto 16), ENB=> enb, WEB=> zero);
>
> end bhv;
>
> ----------------------------------------------------------------------
> ----------
> -- end of file >
>
> To Post a message, send it to:
> To Unsubscribe, send a blank message to:





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Tim Boescke - Oct 22 19:37:00 2001



> > Jan Gray somehow managed to fit a 4 operator ALU into a
> > single LUT per bit. (In the GR0040) Is there any documentation on
> > this ? So maybe there is a way to reduce the ALU size to 6 CLBs.
>
> Some of it is here: http://www.fpgacpu.org/log/nov00.html#001112.
>
> Some of it is hinted at in the GR0000 paper here:
> http://www.fpgacpu.org/papers/soc-gr0040-paper.pdf
> in section 3.12.

Thank you!

> Now then. Say you build a 4-bit ALU using this techniques. 4 LUTs.
> And say you attach that to a 16 entry x 4-bit LUT RAM (another 4 LUTs,
> or 8 if you make it dual-port RAM). And say you add a 2-bit counter (2
> LUTs) to sequence through LSB addresses 00, 01, 10, and 11 to that LUT
> RAM. Now you have a simple datapath with 4 16-bit registers,
> nybble-serial, that should easily run at 100 MHz (25 MHz for each 16-bit
> operation). Total cost of datapath: 10-14 LUTs (3-4 Virtex CLBs; 2
> Virtex2 CLBs).

Thats a pretty interesting idea. But how do we load data from the ram ?
I fear another MUX is required, at least to load constants.

> To hook that up to a 512x8 or 256x16 BRAM for program and data storage,
> you need another 8 or 9 FFs for a PC and/or address register. These FFs
> can share the same handful of CLBs with the aforementioned LUTs. The
> instruction register can be the BRAM output register.

Hm. that could substitute for the adress register, if we dont allow
indexed adressing modes. This would simplify the control a bit.

OTOH the PC could be mapped to the registerfile, but this would
double the amount of cycles per instruction. (well, 8 cycles for a dual
ported
registerfile, 12 cycles for single ported)

The adress register could have an incrementer. This would make flexible
length
instruction encoding easy. (register to register ops: 1 byte, immediate ops
3 byte)






(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Martin.J Thompson - Oct 23 4:58:00 2001



--
Martin Thompson BEng(Hons) CEng MIEE
TRW Automotive Advanced Product Development,
Stratford Road, Solihull, B90 4GW. UK
Tel: +44 (0)121-627-3569
mailto: >>> 21 October 2001 17:21:23 >>>
> 0) Load input 1) Jump minus 2) Add 3) Store Compliment

Anyone implemented a One Instruction Set Computing processor in FPGA yet?
http://www.idiom.com/free-compilers/LANG/OISC-1.html

:-)

Martin




(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Felix Bertram - Oct 23 4:59:00 2001

> What I meant was to istantiate the DPRAM generated by
> the COREGenerator and
> the Muxes and other stuff generated by LogiBlox, so
> (I think) I don't get or
> should write any architecture, isn't it?

ok, I misunderstood you. There are some things to remember here.
CoreGen produces EDIF. This means that the CoreGen modules are
treated as black boxes during synthesis and resolved during design
translation with ngdbuild. This implies, that you need to have a
replacement for your core during simulation. Xilinx provides the
XilinxCoreLib which contains a set of primitives to create models of
CoreGen modules.

For simulation, you can instantiate your CoreGen modules in VHDL, as
the VHDL snippet below explains. I chose a different way than the
instantiation template provided in the "vho" file. It uses a
placeholder named "dpram256x32ii" so that I do *not* have to create
that top-level configuration used in the vho file. Instead I can
instantiate the "dpram256x32" entity like any other VHDL entity. The
complete file is not included in the synthesis project.

However, for simple blocks like RAMs I found this not too helpful. I
feel it is much more concise to implement the RAM modules by hand in
VHDL- as my previous posting showed. If you're wondering what CoreGen
actually does- it creates instances of exactly the same primitives
from the Unisim library that my VHDL code uses. Just have a look
inside the EDIF files produced by CoreGen and search for "RAMB4". The
main advantage of this is, that you can use exactly the same set of
files for synthesis and simulation- avoiding problems related to
inconsistent design files, resulting in deviations of simulation from
synthesis.

Hope this helps,

Felix
_____
Dipl.-Ing. Felix Bertram
Trenz Electronic
Duenner Kirchweg 77
D - 32257 Buende
Tel.: +49 (0) 5223 49 39 755
Fax.: +49 (0) 5223 48 945
Mailto:
http://www.trenz-electronic.de

----------------------------------------------------------------------
----------
library IEEE;
use IEEE.STD_LOGIC_1164.all;

-- synopsys translate_off

Library XilinxCoreLib;

-- synopsys translate_on

entity dpram256x32 is
port (
addra: IN std_logic_VECTOR(7 downto 0);
clka: IN std_logic;
addrb: IN std_logic_VECTOR(7 downto 0);
clkb: IN std_logic;
dia: IN std_logic_VECTOR(31 downto 0);
wea: IN std_logic;
ena: IN std_logic;
enb: IN std_logic;
dob: OUT std_logic_VECTOR(31 downto 0));
end entity; ----------------------------------------------------------------------
----------
architecture CoreGen of dpram256x32 is
component dpram256x32ii
port (
addra: IN std_logic_VECTOR(7 downto 0);
clka: IN std_logic;
addrb: IN std_logic_VECTOR(7 downto 0);
clkb: IN std_logic;
dia: IN std_logic_VECTOR(31 downto 0);
wea: IN std_logic;
ena: IN std_logic;
enb: IN std_logic;
dob: OUT std_logic_VECTOR(31 downto 0));
end component; -- synopsys translate_off

for all : dpram256x32ii use entity
XilinxCoreLib.C_MEM_DP_BLOCK_V1_0(behavioral)
generic map(
c_depth_b => 256,
c_depth_a => 256,
c_has_web => 0,
c_has_wea => 1,
c_has_dib => 0,
c_has_dia => 1,
c_clka_polarity => 1,
c_web_polarity => 1,
c_address_width_b => 8,
c_address_width_a => 8,
c_width_b => 32,
c_width_a => 32,
c_clkb_polarity => 1,
c_ena_polarity => 1,
c_rsta_polarity => 1,
c_has_rstb => 0,
c_has_rsta => 0,
c_read_mif => 0,
c_enb_polarity => 1,
c_pipe_stages => 0,
c_rstb_polarity => 1,
c_has_enb => 1,
c_has_ena => 1,
c_mem_init_radix => 16,
c_default_data => "0",
c_mem_init_file => "dpram256x32.mif",
c_has_dob => 1,
c_generate_mif => 1,
c_has_doa => 0,
c_wea_polarity => 1); -- synopsys translate_on

begin
U0 : dpram256x32ii
port map (
addra => addra,
clka => clka,
addrb => addrb,
clkb => clkb,
dia => dia,
wea => wea,
ena => ena,
enb => enb,
dob => dob);
end CoreGen;
----------------------------------------------------------------------
----------
-- end of file





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Ben Franchuk - Oct 23 9:51:00 2001

Jan Gray wrote:

> Fun stuff. Brings back the old days of assembler one-upsmanship,
> striving to bum an instruction or a cycle from a little compute kernel.
> "What's the most compact sequence to do atoi, or itoa, or what have
> you..."
>
> I feel sorry for anyone starting in computing today and who thinks 4 KB
> is nothing. They missed all the fun.

4KB?? Nowdays 64 Meg is nothing. I think program size could be
1/4 the size
assuming people wanted to save memory at some cost in speed.
Finally debugging
my FPGA CPU in real hardware nothing beats the old switches and
lights to see
what the cpu is doing. Programmable logic is nice to change
configuration
around for spot testing but I think all CPU's need a
Run/halt/step setup where
development is being done. While my cpu still has a few bugs I
do have the UART
working, front panel switches and my bootstrap ( hardware )
loader are working.
Now I just need to go instruction by instruction to see what
works. :)
> Jan Gray
Ben Franchuk.
--
Standard Disclaimer : 97% speculation 2% bad grammar 1% facts.
"Pre-historic Cpu's" http://www.jetnet.ab.ca/users/bfranchuk
Now with schematics.





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

RE: New minimal CPU design - Author Unknown - Oct 24 6:28:00 2001

This may sound dumb, but Im trying to get to grips with all this RISC FPGA
microcontroller stuff.

What pseudo instructions may be emulated using a combination of Tim Boescke's
minimal NOR, ADD, STA, and JCC instructions?

Have I been spoilt all this time and I only really need these four?

Thanks, Vincent





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Ben Franchuk - Oct 24 10:16:00 2001

wrote:
>
> This may sound dumb, but Im trying to get to grips with all this RISC FPGA
> microcontroller stuff.
>
> What pseudo instructions may be emulated using a combination of Tim Boescke's
> minimal NOR, ADD, STA, and JCC instructions?
> Have I been spoilt all this time and I only really need these four? I assume here that STA clears the accumulator. Yes 4 is the
minimum instructions to have for a simple controller. Shifts and
fancy arithmetic and subroutine calls do make a like easier to
program however.
Having just found a intermittent connection in a ribbon cable
from my FPGA to my memory chips
I will say a front panel is mandatory. Simulation does not
always find bugs too as I also had error with my branch offsets
for conditional branches in the simulator compared with the
hardware.Ben

--
Standard Disclaimer : 97% speculation 2% bad grammar 1% facts.
"Pre-historic Cpu's" http://www.jetnet.ab.ca/users/bfranchuk
Now with schematics.


(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Tim Boescke - Oct 24 14:06:00 2001


> You may also wish to put 0 in the regfile (leaving 2 (or perhaps it is 2
> 3/4) 16-bit regs (or 6 8-bit regs, if you prefer)). Then IR=MEM[PC+=1]
> becomes rf[PC] += rf[zero] + cin=1 across 4 cycles. As the alu output
> nybbles go by, you latch them in your nybble-serial-to-parallel FF-based
> address register, that drives address lines to the BRAM. And perhaps
> you can save more registers and a mux, if the data bus is only 4-bits
> wide, if you use a RAMB_S16_S4 or something like that. (Hands wave
> furiously.)

Hm.. When using a single ported register file the output flip flops
have to be used to store the first operand. This could be configured
so that it is resetted instead of being loaded when the PC register
is accessed. (Not sure whether it is possible to configure the CLBs
like that) This way the PC reads always zero when it is read as first
operand. Saves one register at the cost of one CLB. (I hope :) )

> (If this is all too complex, by all means, build a conventional 8- or
> 16-bit high regfile and ALU -- I just wanted to show how that it is
> possible to build a minimal slow austere 16-bit datapath in just 2
> Virtex2 CLBs.)

I think i will go for a 16regs/8 bit parallel design as a first step.
Should be easy to fold the design into a serial 16 bit architecture later.
Of course the above trick would be less useful in this configuration.

> (Another gate bumming idea: the output network of a RAM is a mux. Use

Yeah, this could save some muxes. However it would require an additional
instruction register. I have to evaluate this later..





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Tim Boescke - Oct 24 14:19:00 2001


> > > minimal NOR, ADD, STA, and JCC instructions?
> > I assume here that STA clears the accumulator. Yes 4 is the
> > minimum instructions to have for a simple controller. Shifts and
> > fancy arithmetic and subroutine calls do make a like easier to
> > program however.
>
> I was hoping for a few macros to explain things - like shift/rotate,
compare,
> and etc.
>
> On further investigation, I found a few macros in Tims cpu3.inc file.
>
> Any more for any more?

Well, here are more:

memory constants used:

allone=255
one=1

shift akku left:

sta temp
add temp

test for zero:

add allone
c=0 if akku was zero

rotate akku left:

sta temp
add temp
jcc noadd
add one
noadd:

compare akku to temp:

nor allone
add one
add temp

c=0 if akku <= temp

nor allone
add temp

c=0 if akku < temp

akku=-akku:

nor allone
add one





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Tim Boescke - Oct 24 14:24:00 2001


Sorry, some corrections:

> compare akku to temp:
>
> nor zero
> add one
> add temp
>
> c=0 if akku <= temp
>
> nor zero
> add temp
>
> c=0 if akku < temp
>
> akku=-akku:
>
> nor zero
> add one





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Tim Boescke - Oct 25 9:54:00 2001


> Could this CPU be extended for program and data memory (harvard) without
> increasing the logic count considerably?

The size will increase almost linearily with the adress/data width, since
the control overhead is very small.
Extending the cpu to 16bit data / 14 bit adress is pretty straightforward.
You just have to change all signal widths / bit offsets in the source..




(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: Re: New minimal CPU design - Ben Franchuk - Nov 1 20:22:00 2001

Tommy Thorn wrote:

> Very nice, although (as others have pointed out) not minimal in a
> mathematical sense, merly small. Also, IMHO, you have to take the
> gates for program and data memory into account, at which point
> code density becomes an issue. While clearly turing complete
> (modulo finiteness), the code size explodes if you try running
> more realistic stuff.

I don't know RISC machines have the same problem too.
Most machines have a comfortable operating size of main memory
and problem solving ability.

0) Turning machines and OIC's would fall in this class.
You can solve the problem but only in a abstract sense.

1) Toy or educational computers. 1K or less of memory
and limited I/O. You can solve trivial problems.The Minimal
machine fits here.

2) Calculator or simple control. 16K or less of memory
and simple I/O and data storge.PDP-8's or any machine with
BASIC in ROM.

3) Simple Real world problems. 128K or less of memory. Often
56k for many machines. Good floppy I/O. The PDP-11 or IBM PC
worked best with 128 of memory.

4) Real machines with real problems. Big,Bigger,BEST money can
buy this week.

5) Super computers -- Number crunching,weather,games,Windows :) > It would be very interesting to see what kind of architectures could come up if we had a competition for the fastest/smallest
> cpu (eg. maximizing MIPS/(Mhz*gates), where gates includes program
> memory cost). Of course this would require fixing a set (> 4) of
> of realistic/useful micro benchmarks. I'm not sure Dhrystone
> would be a good candidate. large integer factorization, RSA,
> AES (DES successor), theorem proving, n'th digit of pi, ...

Byte had ( late 1980's?) a benchmark on judging CPU's by instruction
size , bus size and speed. Byte also had some real benchmarks (in small
C?)
I think too but never heard of them since. The small C compiler is
almost a good benchmark by itself to generate sample programs.
I have to say I designed my CPU to generate nice Small C output.
This was because I have 24Kb limit for programs, 8kb is reserved for
DOS.
A well hacked Small C ( ver 1) fits just under 24kb.
(Not that I have Dos or even a disk yet ).
Ben Franchuk.

--
Standard Disclaimer : 97% speculation 2% bad grammar 1% facts.
"Pre-historic Cpu's" http://www.jetnet.ab.ca/users/bfranchuk
Now with schematics.





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: New minimal CPU design - Tommy Thorn - Nov 1 22:38:00 2001

--- In fpga-cpu@y..., "Tim Boescke" <t.boescke@t...> wrote:
> I just wanted to announce a new minimal CPU designed
> by me some time ago.
>
> Its special in the respect, that it fits into a 32
> macrocell CPLD. Even though the instruction set
> is severly limited I think that the code density
> and cpu performance is on par with other minimal
> 'educational' CPUs, which are sometimes quite a bit
> larger in respect to gate/register usage.
....
> So, does anybody come up with a smaller CPU ?

Very nice, although (as others have pointed out) not minimal in a
mathematical sense, merly small. Also, IMHO, you have to take the
gates for program and data memory into account, at which point
code density becomes an issue. While clearly turing complete
(modulo finiteness), the code size explodes if you try running
more realistic stuff.

It would be very interesting to see what kind of architectures could come up if we had a competition for the fastest/smallest
cpu (eg. maximizing MIPS/(Mhz*gates), where gates includes program
memory cost). Of course this would require fixing a set (> 4) of
of realistic/useful micro benchmarks. I'm not sure Dhrystone
would be a good candidate. large integer factorization, RSA,
AES (DES successor), theorem proving, n'th digit of pi, ...

/Tommy




(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: Re: New minimal CPU design - Tim Boescke - Nov 5 11:55:00 2001



> > I just wanted to announce a new minimal CPU designed
> > by me some time ago.
> >
> > Its special in the respect, that it fits into a 32
> > macrocell CPLD. Even though the instruction set
> > is severly limited I think that the code density
> > and cpu performance is on par with other minimal
> > 'educational' CPUs, which are sometimes quite a bit
> > larger in respect to gate/register usage.
> ....
> > So, does anybody come up with a smaller CPU ?
>
> Very nice, although (as others have pointed out) not minimal in a
> mathematical sense, merly small. Also, IMHO, you have to take the

Thanks. Well, the 'minimal' was rether referring to the used CPLD,
which is the smallest you can get. But of course you are generally
right.

> gates for program and data memory into account, at which point
> code density becomes an issue. While clearly turing complete
> (modulo finiteness), the code size explodes if you try running
> more realistic stuff.

Yes, but it is interesting to note, that this instruction set did
not eat up more logic resources, than other more minimized and
less efficient instruction sets. (for example branch-and-substract-
when-negative). Proving Turing completeness is fairly simple for
most instruction sets - however are there any generalized means
to measure efficiency on a theoretical base ?

> It would be very interesting to see what kind of architectures could come
up if we had a competition for the fastest/smallest
> cpu (eg. maximizing MIPS/(Mhz*gates), where gates includes program
> memory cost). Of course this would require fixing a set (> 4) of
> of realistic/useful micro benchmarks. I'm not sure Dhrystone
> would be a good candidate. large integer factorization, RSA,
> AES (DES successor), theorem proving, n'th digit of pi, ...

Indeed, unfortunately there are lots of degrees of freedom, which
makes optimizing very difficult. A more constrained approach
(like limiting the instruction set) would be more realistic.





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: Re: New minimal CPU design - Author Unknown - Nov 5 20:27:00 2001

"Tim Boescke" <> writes:

> Yes, but it is interesting to note, that this instruction set did
> not eat up more logic resources, than other more minimized and
> less efficient instruction sets. (for example branch-and-substract-
> when-negative). Proving Turing completeness is fairly simple for
> most instruction sets - however are there any generalized means
> to measure efficiency on a theoretical base ?

I was trying to come up with such a thing. The best I've been able to
come up with so far measures the efficiency of two generalized CPU
architectures A and B. It requires that both architectures be
generalized to work at arbitrary bit widths n. Then the efficiency of
A relative to B is O(f(n)) if every instruction of B can be emulated
by O(f(n)) instructions of A.

For instance, suppose B has an instruction for computing (x>>y) where
x, y, and the result are n-bit quantities. An architecture with a
shift right by one bit instruction can emulate this with O(n)
instructions. An architecture that requires division by repeated
subtraction uses O(2^n) instructions.

I conjecture that the "standard" architectures (x86, SPARC, etc.) can
emulate each other in O(n).

Carl Witty





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )

Re: Rotate & Shift module - Ed Corter - Nov 13 11:21:00 2001

Sunday Ed Corter (hey that me) wrote
I am almost done with an Auto rotate and shift module
(Verilog). Module is programmable for mode of
operation ( sh \ rot ) and a count of operations is
set. When the operand is written . the module does the
N Op's and flags Done.
The module description and usage is available at

http://ca.geocities.com/artiedc/files/rotshift.htm

there's a zip file for download that contains all of
the files ( 3 for the module ) and all of the models
sim files including the bench.

Ed Corter __________________________________________________





(You need to be a member of fpga-cpu -- send a blank email to fpga-cpu-subscribe@yahoogroups.com )