Journey into Specious Reasoning & Architecture Advancements
Embark on a journey starting with being blindfolded on December 31, 2013, then meet Olaf the snowman, explore Pro Tools audio recording evolution, and delve into the world of GPUs and specialization. Witness the intersection of hot and cold with a touch of whimsy, discover significant wins in architecture, and understand the balance between specialization's benefits and pitfalls. Conclude with a rebuttal encompassing Dennard scaling and memory bandwidth amplification.
Download Presentation
Please find below an Image/Link to download the presentation.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.
You are allowed to download the files provided on this website for personal or commercial use, subject to the condition that they are used lawfully. All files are the property of their respective owners.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.
E N D
Presentation Transcript
On December 31, 2013 I was blindfolded I was taken from my home I was led to a dark room
Meet Olaf A snowman who looks forward to summer The hot and the cold are both so intense Put them together it just makes sense! Specious reasoning, friend Olaf
Pro Tools Audio Recording 1991 NuBus card, 2x Motorola 56001 DSP, 4 channel 24-bit fixed point Integrated SCSI controllers important 1998 PCI card, 6x Motorola 56301 DSP, 64 channel 1999 Livin La Vida Loca recorded & mixed entirely in Pro Tools 2002 PCI card, 9x Motorola 56321 DSP, 192 channels 2010 HD Native, a non-DSP system, 256 channels DSPs no longer just make sense 2011 PCIe card, 18x TI DSP, 256 channels 32-bit floating point
Biggest architecture win in 10 years GPUs Ahem!
Biggest architecture win in 10 years GPGPUs General-purpose graphics processing units
Modern GPUs are general purpose Successful because of generality Applicable to many domains Though only effective for data-parallel problems Systems trump programs How many programs to create amazon s home page? GPUs need system services GPUfs [ASPLOS 13] GPUnet [coming soon I hope ]
Specialization, the good and the bad Improves performance, decreases energy Fixed-function units E.g., encryption, hashing, video decoder Little need for system software, easy to multiplex Programmable units end users write software E.g., early GPUs, IBM Cell, digital signal processors (DSP), network processors Require much care & feeding from system software Difficult programming model
Wah, wah, wah Dennard scaling 10x, 100x memory bandwidth 2 cycle memory access latency Persistent byte-addressable memory Abundant wireless bandwidth
