Imagine a world where you and your trusted AI assistant (they, them, theirs) have just completed the frontend design of your latest chip. It seems like only a few hours have passed since you established the high-level product requirements, with your assistant suggesting possible architectural implementations and tradeoffs to consider. As you deliberated on the optimal IP to use, your assistant searched its IP catalog for the best options and answered all your questions about specs and features. They then outlined the high-level data-flows in the chip and crafted the first pseudo code for the necessary control logic.

After iterating through the high-level specifications and architectural implementations, you proceeded with the design. Your assistant crafted the first RTL code and immediately executed synthesis and timing analysis to help assess the anticipated power, performance, and area. They also coded the initial verification test bench, started simulations, and proposed changes to the design for bugs found. They even automatically added test cases to raise the verification code coverage.

After only a few iterations you already reached the confidence needed to warrant hand-off of the design to the backend team. Your AI assistant not only assisted with the architectural specification, IP selection, RTL coding, and verification of your latest chip but also helped you do so in a record time and with a high level of confidence your team will not be wasting time having to fix unforeseen issues or bugs while layout is in progress.

Editorial note: Well, imagine no more! The world Hans describes above exists. It’s the enabled by “RapidGPT”, PrimisAI’s groundbreaking chip design tool. RapidGPT continues to advance and is offering greater capabilities each day.

Don’t miss out on the opportunity to revolutionize your chip design process. Contact PrimisAI today for more information.

Remembering our days as students and our early days as chip designers, we think of all the things we used to despise... like none of us ever wanted to write any comments... thought of them as simply a waste of time... because you already know what our code does. The same goes for documentation, the same goes for changing the context of a design, renaming variables, or making the code robust for cases that might happen, not relevant to our current problem that we were solving.

Over the years, under pressure from senior designers and managers, we reluctantly learned to add comments, add documentation, and learned the patience of making the design robust to enhance its usability by ourselves, and potentially others, but never felt great about doing such things... always prayed for a better system so that we could avoid doing those things. Enter Gen AI... and RapidGPT!

We think our prayers may be finally answered. (As design engineers, we don’t believe in anything too fast.) We can clearly see a world where designers do not have to write comments, documentation, change the context of design, make a design robust, or do test benches... in fact, in our experiments, RapidGPT does this and more today... over a thousand people have tried... major corporations are testing it or using it. We have designed small ASICs based on very high specs which were written with the help of RapidGPT itself.

And this is just the beginning... over time, we feel designers should interact with Gen AI-based systems at a very high level, changing the specs and assessing the impact of changes on the performance of the design.

Stay tuned... we predict Gen AI-based chip design systems are going to totally change our perspective on how chip design is done!

We are happy to announce that RapidGPT v1.36 is now available on the VS Code marketplace. Updating to the latest version is seamless: if extension auto-update is enabled, the update will occur automatically when you reopen VS Code. For those who prefer manual updates, please refer to this guide on how to update an extension manually.

New Features and Highlights​

In-Line Quick Commands​

Building on the efficiency of our Quick Commands Panel, we're introducing in-line quick commands. This exciting new feature empowers users to execute actions directly on specific modules within their code, streamlining the workflow and keeping the focus where it matters most.

Updated IP Catalog​

We've expanded the selection of IPs available in our catalog. With this update, pro users can now effortlessly instantiate over 30 distinct Xilinx IPs using natural language. Additionally, users can directly discuss the documentation of these IPs within the chat interface. Here's a list of the newly added components:

Component	Category
AMM Master Bridge	Adders & Subtractor
AMM Slave Bridge	Conversion, CORDIC
AXI EPC	Divider
AXI GPIO	Floating Point
AXI HWICAP	Multiplier
AXI IIC	Multiplier
AMM Master Bridge	AXI Low Speed Peripheral
AMM Slave Bridge	AXI Low Speed Peripheral
AXI EPC	AXI Low Speed Peripheral
AXI GPIO	AXI Low Speed Peripheral
AXI HWICAP	AXI Low Speed Peripheral
AXI IIC	AXI Low Speed Peripheral
AXI Interrupt Controller	AXI Low Speed Peripheral
AXI CAN	AXI Low Speed Peripheral
AXI Quad SPI	AXI Low Speed Peripheral
AXI TFT Controller	AXI Low Speed Peripheral
AXI Timebase Watchdog Timer	AXI Low Speed Peripheral
AXI Timer	AXI Low Speed Peripheral
AXI UART16550	AXI Low Speed Peripheral
AXI Uartlite	AXI Low Speed Peripheral
CANFD	AXI Low Speed Peripheral
AHB-Lite to AXI Bridge	Interconnect
AXI AHBLite Bridge	Interconnect
AXI APB Bridge	Interconnect

Enhanced Chat Panel​

The updated chat panel now displays users' full names alongside their messages, enhancing clarity and context. This feature proves especially valuable in version control settings, where RapidGPT facilitates collaborative efforts by enabling users to accurately track their interactions with RapidGPT.

Improved Self-Awareness​

With the latest release, RapidGPT has been upgraded to possess a deeper understanding of its functionalities, ensuring you receive more comprehensive and accurate responses. Now, you can directly inquire about RapidGPT's key features within the chat panel. Curious about how to operate AutoDoc? Just ask RapidGPT, and it will seamlessly guide you through the process.

Bug Fixes and Other Improvements​

This release introduces a series of bug fixes and improvements, including:

• Dynamic input box adjustments: The prompt input box now dynamically adjusts vertically to accommodate more content, enhancing readability and comfort during text entry.
• Customizable iteration cap for AutoReview: With the new Autoreview Max Iterations setting, users can now set a custom iteration limit in AutoReview, specifically addressing and mitigating the occurrence of 429 errors.

Imagine a world where you, your colleagues AND your AI assistants seamlessly work together to achieve your next goal. Picture your AI assistants effortlessly handling the boring and mundane tasks, even while you're asleep! Your AI assistants understand the complexities of applying EDA to your hardware design project and can even interpret any ambiguous requests. They offer new and objective perspectives and can monitor and analyze massive amounts of data.

Yet, without your guidance, your AI assistants would be lost. In essence, they are merely a machine making proposals based on associations observed. They need your adaptability and flexibility and require large amounts of data to learn. And more importantly, require your insights into the bigger picture - what ultimately drives you, your team and your company's success.

But together you and AI can excel and reach new heights, and just like that, only humans can truly understand how 1+1 can equal 3!

We are happy to announce that RapidGPT v1.35 is now available on the VS Code marketplace. Updating to the latest version is seamless: if extension auto-update is enabled, the update will occur automatically when you reopen VS Code. For those who prefer manual updates, please refer to this guide on how to update an extension manually.

New Features and Highlights​

Quick Command Selection Panel​

In this release, we unveil a novel quick command selection panel, strategically located beneath the Chat panel. This new component empowers users to swiftly access specific functions directly, bypassing the need for navigating through the VS Code Command Palette. The video below showcases how users can seamlessly incorporate AI-generated comments into a Verilog file utilizing RapidGPT's AutoComment feature, all achievable through a mere single click.

Support for Multiline Comments​

Clarity is key when navigating complex challenges. In alignment with this principle, we have now extended full support for multiline comments within code autocompletion. Users can now harness the versatility of markdown formatting within Verilog comments, thus enhancing the readability and maintainability of their code. This enhancement taps into the full potential of RapidGPT's predictive capabilities for HDL code, ensuring a smoother and more intuitive coding experience.

Improved Conversation Scopes Panel​

The Conversation Scopes panel is essential for managing multiple conversations related to various facets of the overall design. With this update, the panel has been revamped to offer an improved look and feel, featuring distinct icons that signify scope types (e.g., files or individual modules). Moreover, it has been made interactive, enabling users to navigate through their source code by selecting conversation anchors.

Bug Fixes and Other Improvements​

This release introduces a series of bug fixes and improvements, including:

• RapidGPT has been updated to store file paths in chat history using their relative paths, effectively resolving issues related to managing files with identical names located in distinct directories.
• Enhancements in conversation management have been made, addressing bugs that previously hindered the deletion of chat histories under specific conditions.
• Refinements to the code autocompletion functionality have been implemented, eliminating unintended activations of RapidGPT during cursor navigation within the source code.