The Morning Paper

MAY 23, 2019

Compress objects, not cache lines: an object-based compressed memory hierarchy Tsai & Sanchez, ASPLOS’19. Last time out we saw how Google have been able to save millions of dollars though memory compression enabled via zswap. If we compress objects instead of cache lines though, we can get to a 56% compression ratio (c).

Cache 61

Cache Benchmarking Hardware Java 61

Brendan Gregg

MARCH 16, 2024

Only in extreme circumstances does the cost (in processor time and I-cache footprint) translate to a tangible benefit - circumstances which usually resort to hand-coded assembly anyway. It shouldn't be 10%, unless it's cache effects. Google should be no surprise because they pioneered continuous profiling.)

Java 145

Java Cache Google Hardware 145

Abhishek Tiwari

DECEMBER 13, 2017

This is a question recently asked and explored by a team of Google researchers led by Jeff Dean with a major focus on database indexes. Jeff is a Google Senior Fellow in the Google Brain team and widely known as a pioneer in artificial intelligence (AI) and deep learning community. Learned indexes.

Artificial Intelligence 40

Artificial Intelligence Hardware Google Benchmarking 40

ACM Sigarch

OCTOBER 12, 2020

Second, optimization efforts for JavaScript were the subject of probably millions of developer hours by a number of companies, notably Google, Microsoft, and Mozilla, and built on decades of research in the PL community on how to optimize such languages (starting with the Self project , which started in the late 80s). MICRO 15 , Gope et al.,

C++ 108

C++ Benchmarking Hardware Architecture 108

Alex Russell

APRIL 29, 2021

As an engineer on a browser team, I'm privy to the blow-by-blow of various performance projects, benchmark fire drills, and the ways performance marketing (deeply) impacts engineering priorities. With each team, benchmarks lost are understood as bugs. PowerPoint or Google Slides). is access to hardware devices.

Media 145

Media Games Education Engineering 145

Alex Russell

MARCH 6, 2021

A then-representative $200USD device had 4-8 slow (in-order, low-cache) cores, ~2GiB of RAM, and relatively slow MLC NAND flash storage. Hardware Past As Performance Prologue. Using a global ASP as a benchmark can further mislead thanks to the distorting effect of ultra-high-end prices rising while shipment volumes stagnate.

Performance 124

Performance Network Cache Metrics 124

Alex Russell

OCTOBER 22, 2017

Budgets are scaled to a benchmark network & device. Deciding what benchmark to use for a performance budget is crucial. Our metrics at Google show a conflicted picture (which I’m working to get to clarity on). Simulated packet loss and variable latency, however, can make benchmarking extremely difficult and slow.

Performance 68

Performance Benchmarking Network Mobile 68