The intuition is simple. If we had a magical zero-cost GC, we could compare costs with a real GC and the difference would be the overhead of the real GC.
Unfortunately there are no magical GCs. So we did something that allows us to place an empirical lower bound on the cost.
We call the basic approach 'distillation'. We measure a number of real GCs and subtract the easily measurable GC costs. We use the 'distilled' cost to approximate the zero-cost GC.
We built LXR in OpenJDK 11 (LTS at the time), and evaluate against production collectors.
LXR takes a very different approach to existing production collectors, using advanced reference counting for prompt reclamation, carefully avoiding the overhead of concurrent copying.
As a result, LXR can outperform these collectors on both throughput and application latency (all in OpenJDK 11, details in the paper).
It is wrong to encourage students to pull five all-nighters in a row.
Unfortunately faculty in my hemisphere are landed with this prospect by virtualized conferences with a 12 hour schedule that's centered on the North Atlantic.
@afd_icl has been doing an incredible job organizing a brilliant @PLDI under extraordinary circumstances. I am in awe of the job he and the team have done.
My message is for @sigplan, @PLDI, and those organizing our next conferences.
I’m strongly in favor of reducing conference travel.
However, as I’ve said repeatedly here and elsewhere, virtualization should not come at the cost of geo-diversity.
What we're seeing is exactly what I was afraid of.
On my way to ASPLOS 2013, I found myself one-across from an old guy, about 80, neat as a pin, and incredibly fit. He was quietly working on his laptop on a talk about space travel, on a flight to Houston. Everything about him said 'astronaut'.
1/4
On the way out, I noticed the name tag on his carry-on: David Scott.
It was a privilege to have spent a few hours, in silence, one seat across from the guy who’d stood on the moon, and recreated Galileo’s famous experiment!
2/4
Closer to home, it was John Saxon, the dad of one of my school friends who was in charge of communications at Honeysuckle Creek as they received the moon landing signal which was then broadcast around the world. Only much later did I understand the significance of his work.
My grandad’s story about how a U-boat captain gave him his Iron Cross intrigued me as a kid.
Today I thought I’d see if I could find out anything of this story, and the original owner of the Iron Cross.
It turns out to be even more interesting than I’d guessed.
1/8
By 1944 the Germans were moving U-boats from Europe to Japan. My grandad was chief navigating officer with RAF 621 Sqn, based in Aden, charged with intercepting the U-boats off the horn of Africa.
2/8
On May 2, 1944, U-852 was caught by five aircraft from 621 Sqn and one from 8 Sqn. The U-boat was incapacitated, beached and blown up north of Bandar Beyla, Somalia. 7 of the crew died and 59 survived but were captured.