All Things Distributed

NOVEMBER 15, 2010

We believe that making these GPU resources available for everyone to use at low cost will drive new innovation in the application of highly parallel programming models. For example, the most fundamental abstraction trade-off has always been latency versus throughput. General Purpose GPU programming. From CPU to GPU.

AWS 136

AWS Latency Programming Architecture 136

The Morning Paper

DECEMBER 8, 2019

On the last morning of the conference Daniel Bittman presented some of the work being done in the context of the Twizzler OS project to explore new programming models for NVM. The starting point is a set of three asumptions for an NVM-based programming model: Compared to traditional persistent media, NVM is fast. What about security?

Hardware 56

Hardware Programming Media Storage 56

John McCalpin

JANUARY 22, 2018

There was no deep goal — just a desire to see the maximum GFLOPS in action. The exercise seemed simple enough — just fix one item in the Colfax code and we should be finished. Using the minimum number of accumulator registers needed to tolerate the pipeline latency (12), the assembly code for the inner loop is: B1.8:

Latency 40

Latency Hardware Code Testing 40