Compare commits

..

No commits in common. 'master' and 'master' have entirely different histories.

Binary file not shown.

Before

Width:  |  Height:  |  Size: 105 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 427 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 406 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 396 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 330 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 362 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 321 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 286 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 198 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 101 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 393 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 438 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 237 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 205 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 54 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 22 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 37 KiB

After

Width:  |  Height:  |  Size: 24 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 10 KiB

@ -1,174 +0,0 @@
---
title: OpenROAD and OpenPOWER Microwatt core for low-power ASIC applications
author: Ganesan Narayanasamy
tags:
- openpower
- adg
- academic
- openroad
date: 2024-01-25
draft: false
---
# OpenROAD and OpenPOWER - Microwatt core for low-power ASIC applications

## OpenPOWER for Energy-efficient applications

The [OpenPOWER](https://openpowerfoundation.org) microwatt processor is an open-source soft process core implementation of the IBM [POWER ISA](https://openpowerfoundation.org/specifications/isa/) architecture that is designed to be highly energy-efficient. This processor is an example of the growing trend of designing low-power processors that can be used in a variety of applications, including mobile devices, Internet of Things (IoT) devices, and more.

### Microwatt
Microwatt is a multi-core processor capable of running at up to 2.3 GHz and using the Power ISA v3.0 specification. It can run on lower cost FPGAs. It has a memory management unit, privilege modes, interrupts, a floating-point unit, an interrupt controller and a level-2 cache, so that it can boot Linux.

The Microwatt processor was implemented using the [OpenROAD](https://theopenroadproject.org/) native design flow - [OpenROAD-flow-scripts](https://github.com/The-OpenROAD-Project/OpenROAD-flow-scripts) . It is available as part of the EDA cloud native solution in OpenPOWER. OpenROAD delivers an autonomous, No-Human-In-Loop, fully-integrated RTL-GDSII flow for the design of digital integrated circuits across a range of technologies. OpenROAD includes specific features for low power design that enabled the design to specification for Microwatt.

## Microwatt Design Flow using OpenROAD

The figures below show the OpenROAD flow using [OpenROAD-flow-scripts](https://github.com/The-OpenROAD-Project/OpenROAD-flow-scripts)

### OpenROAD High-level Flow

{{< image src="blog/openroad_image1.png" >}}

### OpenROAD Flow Scripts - A detailed Physical Implementation Flow

{{< image src="blog/openroad_image2.png" >}}

## Design Methodology

Microwatt was implemented using Skywater130nm (What production versions are supported- Is there a link to share about it? It consists of 6 macros and over 800K standard cells.

The hardware description for Microwatt was defined using Verilog in OpenPOWER. The RTL was synthesized in OpenROAD using Yosys to obtain a power-efficient mapping. [OpenSTA](https://github.com/The-OpenROAD-Project/OpenSTA) is an integrated and incremental timing and power analysis engine built into OpenROAD.

### Synthesis
The OpenROAD flow uses a synthesized netlist from Yosys to transform the RTL into an elaborated design with corresponding technology mapped primitives to generate an optimized netlist based on the input on specified timing constraints. This netlist was used to create the floorplan for detailed physical implementation.

### Placement and Routing

#### Floorplan Import design

Microwatt has 6 macros that were placed using the automatic macro placement in ORFS.

- RAM512 1 392.925
- RAM32 1RW1R
- Multiply_add_64x64
- RAM32_1RW1R
- Microwatt_FP_DFFRFile
- Multiply_add_64x64

{{< image src="blog/openroad_image3.png" >}}

macro placement

### Power Distribution

There are 5 metal layers in total in Microwatt. Metal 1 is horizontal and M2 is vertical followed by M3, M4 and M5 in horizontal direction. Power distribution is using M1, M4 and M5. The M1 layer is horizontal on the standard cell rails to enable power for the standard cells. VIA2 and VIA3 power vias are dropped. M4 power stripes and M5 power stripes are drawn. An alternate pattern of VSS and VDD pattern is followed

The figure below shows the PDN grid generated.

This is the complete PDN of the floor plan.

{{< image src="blog/openroad_image4.png" >}}

### Placement

Global Placement places standard cells to minimize wire length and avoid congestion during routing. Detailed placement further legalizes standard cell placement to remove overlaps and fix orientation for optimal routing. This step also includes the use of the resizer to optimize timing using techniques like buffer insertion, sizing, pin swapping etc.

{{< image src="blog/openroad_image5.png" >}}
Figure: Placement Density View

{{< image src="blog/openroad_image6.png" >}}

Figure: Placement Optimization by Resizer

### Clock Tree Synthesis

ORFS automatically synthesizes clock trees to meet timing constraints and inserts buffers as needed to balance skew on high fanout nets. Filler cells are also added to meet manufacturing density requirements.

{{< image src="blog/openroad_image7.png" >}}

{{< image src="blog/openroad_image8.png" >}}
Figure: ext_clk Clock Tree Viewer

{{< image src="blog/openroad_image9.png" >}}

Figure: jtag_tck Clock Tree Viewer

### Timing Analysis

OpenROAD includes a built-in STA engine that is used to incrementally analyze timing at various flow stages. The GUI enables easy visualization of timing paths and setup/hold checks.

Shown below are post-CTS timing views for setup and hold times.

#### Setup Timing Path

{{< image src="blog/openroad_image10.png" >}}
Figure: Setup Timing Path

{{< image src="blog/openroad_image11.png" >}}

Figure: Hold Timing Path

#### Routing

{{< image src="blog/openroad_image12.png" >}}
Figure: Routing Congestion View

{{< image src="blog/openroad_image13.png" >}}

Figure: Detail Routing

#### Show CTS after detailed placement.

CTS is shown below by highlighting the clock nets before and after CTS

{{< image src="blog/openroad_image14.png" >}}
Figure: ext_clk source CTS

{{< image src="blog/openroad_image15.png" >}}

Figure: jtag tck source CTS

#### How was timing optimization done using the resizer repair commands?

Using resizer commands the following is done at the placement step. The details are available in 3_4_resizer.log

1) 51 input buffers are added
2) 87 output buffers are added
3) 7237 buffers are added in 9711 nets
4) 23377 instances are resized

#### What tradeoffs were made for timing vs power?

We optimize for timing and less so for power. We do have some power recovery features in development that should be available soon.

#### Where is the static IR estimation report?

In 6_report.log when using ORFS

#### Did you run a simulation for power analysis?

We can provide a VCD (toggle counts) from simulation for improved accuracy. OR doesn't do simulation itself.

#### What kind of signoff was used to verify final timing and power (PrimeTime, Calibre etc.)

We have not done any signoff for final timing and power using Prime Time & Calibre.Only OpenROAD flow is being executed.

### Physical Verification

Once the physical layout (i.e., the GDS file) is complete, the design is ready for final verification and sign-off. This involves running a series of tests to ensure that the design meets all of the necessary requirements and specifications. Any issues that are found can be addressed, and the design can be re-synthesized and re-implemented until it meets all of the necessary criteria.

Finally, the design is ready for fabrication. The GDSII layout file is sent to a foundry, where the physical chip is manufactured. Once the chip is fabricated, it can be assembled into a larger system or device, such as a mobile phone or IoT device.

{{< image src="blog/openroad_image16.png" >}}

Figure: Microwatt Layout generated through OpenROAD

## Results

Showcase results here. Final GDSII. Power and Timing achieved, run times and state the potential for such application development for the future.


Worstcase voltage:1.80e+00 V
Average IR Drop : 4.62e-04 V
Worstcase IR Drop: 1.95e-03 V

As a result, it is possible to create highly efficient and effective microwatt processors that can be used in a variety of different devices and applications, from mobile phones to IoT devices and beyond. The combination of the OpenPOWER microwatt processor and the OpenROAD design flow represents a powerful toolset for designers looking to create low-power, high-performance processors for modern electronics.

@ -4,6 +4,6 @@ position: "Director"
weight: 1000
date: 2021-03-03
start: 2020-03-30
end: 2022-02-28
end: 2023-02-28
draft: false
---

@ -4,6 +4,6 @@ position: Director
weight: 1000
date: 2021-03-03
start: 2021-01-01
end: 2023-12-31
end: 2034-12-31
draft: false
---

@ -1,24 +0,0 @@
---
title: "Open Compute Regional Summit 2024 - Future Technologies Symposium"
eventdates:
from: 2024-04-24
eventplace: Lisbon, Portugal
register: https://2024ocpregional.fnvirtual.app/
schedule: https://2024ocpregional.fnvirtual.app/a/schedule/
date: 2024-04-21
draft: false
---

The OCP Regional Summit in Europe serves as a platform where technical and business leaders from EMEA come together to tackle crucial issues related to data center sustainability, energy efficiency, and heat reuse. The focus is on exploring how innovations from hyperscale data center operators can contribute to solving these challenges. The Summit also highlights deployments of specific OCP-recognized data center equipment in the EMEA region.

```
OPF FTS Talk
Time: 3:55pm 4:15pm
Location: Lisbon Congress Centre
```

The OpenPOWER Foundation will be sharing a talk with Jabil in the Future Technologies Symposium.

"Large Memory Server Design Overview"
Abstract:
With the advancement of AI workloads and their increasingly large data set requirements, the need for high memory capacity and density is becoming paramount for the industry. There have been advances in memory connectivity that promise to deliver solutions that will meet these needs. This specification shows a design that utilizes two of these technologies, OpenCAPI Memory Interface (OMI) and Compute Express Link (CXL), to deliver from 24TB to a potentially jaw-dropping 32TB of memory in an air-cooled, single socket OCP DC-MHS inspired 19 2U chassis. This design solution, a collaboration between the OpenPower Foundation, Jabil Inc.,Raptor Computing Systems, SAP, and Wooden Data Center, is currently in the design phase. In this specification, physical limitations, interconnects, and power/thermal constraints will be highlighted. The end goal would be to contribute this design specification to the OCP community.

@ -1,37 +0,0 @@
---
title: "Open Hardware and Software Mini Summit at the Open Source Summit North America"
eventdates:
from: 2024-04-15
eventplace: Seattle, US
register: https://events.linuxfoundation.org/open-source-summit-north-america/register/
schedule: https://docs.google.com/spreadsheets/d/1F6hJQL8nU538OlGDGiOcrvSGuDMbCbK4R37qGrq5qSg/edit?usp=sharing
date: 2024-03-06
draft: false
---

```
Time: 1:00pm 5:45pm
Location: Seattle Convention Center | Summit
Registration Cost: $15
```


This half day program will Introduce the audience to the many aspects of open source hardware and software development, and how it is helping the industry to accelerate beyond what Moores law has predicted.


Talks will cover numerous aspects of hardware / software development and provide motivation to learn more about the challenges of open hardware and software design together.


Agenda:

- How can software developers help keep Moores law alive?
- Open Hardware Design: Lessons Learned from OSS
- Understanding hardware security vulnerabilities : Hardware Bill of Materials
- Scaling hardware design: squeezing every bit of performance of the open source OpenROAD ASIC toolchain
- Accelerating EDA Flows in the AWS Cloud
- An Introduction to RISC-V
- Accelerating the RISC-V Software Development Ecosystem
- Caliptra: open source Root of Trust
- Getting “Ware” You Need to Go Managing OpenPOWER Hardware with Software (Part 1)
- Getting “Ware” You Need to Go AI Software on OpenPOWER Hardware (Part 2)

@ -4,7 +4,6 @@ level: silver
image: antmicro.png
country: "Poland"
joined: 2020
cancelled: 2022-12-31
date: 2021-03-15
draft: false
---

@ -5,7 +5,6 @@ link: https://www.google.com
level: platinum
country: United States of America
joined: 2013
cancelled: 2023-12-31
date: 2021-03-02
draft: false
---

@ -1,10 +0,0 @@
---
title: Power Maroc
image: powerm.png
country: Morocco
link: https://powerm.ma/
level: silver
joined: 2024
date: 2024-01-01
draft: false
---

@ -13,4 +13,4 @@ He is a serial entrepreneur with a background in enterprise technology and has w
business development, product management, and engineering.
Previously, James was co-founder and COO at Hyper.sh, an open source cloud-native virtualization startup acquired by Ant Financial.
Prior to that, he led product management in Red Hat's OpenStack group, and was a product lead on AT&T's first OpenStack Cloud.
James graduated from University of Virginia with a degree in Electrical Engineering and is based in Washington D.C. area.
James graduated from University of Virginia with a degree in Electrical Engineering and is based in New York.

Loading…
Cancel
Save