Breaking News
Blog

Chip Makers Need New Business Models

NO RATINGS
1 saves
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 2   >   >>
p123a
User Rank
Author
Re: opensource needs huge investment
p123a   4/11/2016 6:38:01 PM
NO RATINGS
Roger Sundman is an electronics engineer by training but has never kept a soldering iron. He is a co-founder and Director of Busines Development at Imsys AB, the company behind the reprogrammable microprocessor, sometimes also known as reconfigurable microprocessors.
The unique architecture named Shannon gives best energy efficiency, best code density and performance by optimizing microcoded hot-spots. It can emulate all ISAs and the control store is partly writable and can thus be reconfigured at any time. The gatecount is very small as gates have been traded for dense, power saving memory. Here comes the next problems described:

Opensource works in the SW etc

1 Our approach require no respin but a fix in the microcode assembler

2. There is a toolchain from our microcode assembler similator and the new LLVM infrastructure is being implemented.

3. You dont need a respin you work on an already silicon proven product. You can do as many microcode runs as you wish and don't need to pay tens of millions. If you have errors in the hardware, that can be a another thing, but there is a repair facility for faulty ROM so maybe that can help.

KarlS01
User Rank
Author
Re: Charting a new course requires a serious look at the business model
KarlS01   3/4/2016 2:10:27 PM
@gjones:  I just read the LE post and left a reply.  In summary, SW typically uses a stack for execution.

HW stack machines exist, FORTH stack is real.

It is time to quit hoping for a good RISC since there's not a dimes worth of difference among any of them.  Lok at the long list of ARM ISAs and try to find it is so long.

Just design a scalable stack machine and implement the SW structs.

gjones@synthconsulting.com
User Rank
Rookie
Charting a new course requires a serious look at the business model
gjones@synthconsulting.com   3/4/2016 12:22:19 PM
NO RATINGS
The GSA/Rambus article is timely and needed. Last summer, I published a series of articles on an actual implementation of a programmable-chip-based standard product offering that I led several years ago. Here's the first one: 

https://www.linkedin.com/pulse/something-wrong-george-jones?trk=pulse_spock-articles

vdara0
User Rank
Author
Re: opensource needs huge investment
vdara0   3/3/2016 5:50:21 PM
NO RATINGS
SV - SystemVerilog

UVM - Universal Verification Methodology

;-)

KarlS01
User Rank
Author
Re-programmable silicon and RISCV
KarlS01   3/3/2016 5:32:16 PM
NO RATINGS
Dyumnin wrote"
The RISC-V example is just a small component in our IOT system and from the website it looks like it is an incremental evolution of a 2 decade old textbook architecture."

RISCV is A CPU (another RISC ISA).  It is about time to get serious about programmable silicon rather than a programmable RISC that still needs a memory subsystem including caches in order to get any decent performance.

Programmable silicon could also do the digital part of WiFi/BT and to do the DSP for DAC/ADC.  Practically everything can be done using memory which is demonstrated by FPGAs.  Except that there is too much wiring used to interconnect each tiny LUT and FF.

So use arrays to do logic functions and design logic instead of RTL  Reprogram by loading RAM blocks rather thanm a serial configuration bit stream.  INTEL/Altera ould probably do an array of many such functional blocks since their FPGAs already have hundreds of embedded RAMs.

 

rick merritt
User Rank
Author
Re: opensource needs huge investment
rick merritt   3/3/2016 3:55:05 PM
NO RATINGS
Sounds like a call to action for someone to roll put open source cores for Wi-Fi and BT and etc.

 

p123a
User Rank
Author
Re: opensource needs huge investment
p123a   3/3/2016 3:12:35 PM
NO RATINGS
Sorry for my lack of knowledge but what is SV+UVM? Maybe I have a fix for you;)

Roger

Dyumnin
User Rank
Rookie
Re: opensource needs huge investment
Dyumnin   3/3/2016 10:21:20 AM
NO RATINGS
Taking the example of an IOT ASIC, We would require the following
  1. A CPU Subsystem.
  2. A Memory Subsystem
  3. WiFi/Bluetooth IP's
  4. GPS IP If our IOT is location aware.
  5. ADC/DAC's for our sensors.

This system consists of Digital Logic, Analog, RF and custom cell's

Everything other than the Digital is specific to the Technology node, foundry and fabrication process.

The RISC-V example is just a small component in our IOT system and from the website it looks like it is an incremental evolution of a 2 decade old textbook architecture.

At this rate it will take a long time for any open hardware to come up to the leading edge.

warmcat
User Rank
Rookie
Re: opensource needs huge investment
warmcat   3/3/2016 7:19:12 AM
NO RATINGS
If it's pure digital, it can be tested in fpga very cheaply, since $100 zynq boards are a reality. FOSS should certainly chip away at trivial IP like axi i2c, spi, uart, crtc that need nothing special and can be confirmed in FPGA.

realjjj
User Rank
CEO
Re: opensource needs huge investment
realjjj   3/3/2016 4:36:32 AM
NO RATINGS
No license for 5G would be nice (a must in a sane world) and result in a saner standard that doesn't increase complexity artificially to keep ASP and licensing high. Qualcomm is allready telling investors that they might charge more per device with 5G and nobody needs that, they get payed 10 times too much as it is.

Page 1 / 2   >   >>

Datasheets.com Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed