Blog Review: Sept. 11

Cadence’s Paul McLellan checks out the challenges of processing-in-memory and the steps involved in building a logic flow on a DRAM process.

Synopsys’ Taylor Armerding notes that with safety-critical software an ever-present factor in modern life, it’s more necessary than ever to take the time to ensure quality and security when failures can be life-threatening.

In a video, Mentor’s Colin Walls shares tips for how to optimize code for power consumption in an embedded system.

ANSYS’ Craig Hillman explores why design for reliability methodologies are key to reliable products and walks through the steps to DfR success, from establishing the product’s field environment to supplier and component selection.

Rambus’ Steven Woo points to concern over the shift in high-performance computing away from architectures balanced between memory and compute and whether improved memory hierarchies and more power-efficient memory systems will be enough to bring them back into balance.

SEMI’s Sungho Yoon warns that many wafer manufacturers are likely to see declining revenue, profits and capacity utilization given rising trade and geopolitical tensions and softness in memory demand.

Arm’s Nigel Stephens digs into the BFloat16 floating-point number format for neural networks and new Neon and SVE vector instructions in the Armv8-A architecture to use BF16 to accelerate certain computations.

And don’t miss the blogs featured in the recent IoT, Security & Automotive and Test, Measurement & Analytics newsletters:

Editor In Chief Ed Sperling finds more clarity and nuances in different compute models.

Synopsys’ Brett Murdock digs into innovative new clocking schemes in the latest LPDDR standard that will enable easier implementation of controllers and PHYs at maximum data rate as well as new options for power consumption.

Mentor’s Anoop Saha explains why self-driving cars need specialized hardware for AI algorithms to meet performance, power, and cost requirements.

Rambus’ Ben Levine sounds the alarm on the danger of compromised IoT devices and what can be done about it.

Flex Logix’s Geoff Tate shows how to get better density and performance for complex, frequently used blocks.

Editor In Chief Ed Sperling contends that just because a system starts out working correctly doesn’t mean it will continue to do so.

YieldHub’s Marie Ryan observes that including test engineers early in the workflow can speed time to market and maximize profits.

Mentor’s Jayant D’Souza describes a clever technique that increases volume scan diagnosis throughput by 10X.

National Instruments’ James Kimery warns that large amounts of spectrum are necessary for 5G to meet its potential, but that sharing bands with other users poses challenges.