Stefan Marr
@smarr
Followers
2K
Following
5K
Media
455
Statuses
10K
Interpreters, compilation, concurrency, and debugging tools.
Linz, Austria
Joined June 2008
LuaJIT 3 might interpret an SSA-based program representation (possibly with a more optimized interpreter for platforms without JIT compilation) https://t.co/tHP6fdZC70
1
4
22
New (hopefully uncontroversial) post just dropped on CPython's plan for 3.15 and 3.16's JIT https://t.co/kKWhtsXsls I should be doing more researchy things (and I did!), but today's a weekend, so I decided to write.
fidget-spinner.github.io
My personal blog
0
4
10
MPLR'26 will be in Brussels! And, the submission deadline is sooner than usually, March 6th. We're looking forward to all things VMs, Language Implementation, Compilers, Interpreters, and Tooling! https://t.co/6sFORrMh8o We co-locate with @ECOOPConf and @ICOOOLPS!
0
6
7
Somewhat delayed because of Covid, I'm finally going to give my "inaugural lecture", which is a short, hopefully accessible, public talk: in my case "Some things I've learned about software"! All are welcome!
2
1
4
Four lectures in, not a single one recorded properly. I truly reached the professorial level, and started conforming to the stereotype... I used to joke that professors lose their practical skills. Now, I am living it. 🤦🏻♂️
1
0
5
To teach the Deutsch-Schorr-Waite pointer reversal GC marking algorithm or not...
1
0
4
Are there any benchmarks (etc) which show the costs of the SysV x64 ABI relative to a somewhat-or-perfectly-optimal ABI for that program? [Not for single microbenchmarks, because I can definitely create horrible overheads there, but on larger programs / benchmarks.]
0
1
6
My short take on our work, incl. link to paper: https://t.co/4T8Hnxqrre
stefan-marr.de
Measuring causes profiles to change, so is there a way to work around it and know whether a profile is accurate?
0
0
3
Humphrey is now presenting his work titled: Divining Profiler Accuracy: An Approach to Approximate Profiler Accuracy Through Machine Code-Level Slowdown
1
0
2
Can We Know Whether a Profiler is Accurate? Spoiler: yes we can! But it took us figuring out a way to work around the fundamental observer effect. https://t.co/4T8HnxpTBG If you're at SPLASH, don't miss our talk at VMIL today and at OOPSLA on Saturday!
0
5
5
For those of you in Singapore, Jake is at SPLASH and presenting the paper on Saturday. He's friendly, and on the job market -- please find him and say hello!
0
1
2
New paper: Garbage Collection for Rust: The Finalizer Frontier. There are lots of GCs for Rust: we tried pushing as far as we could in a new direction, particularly looking at what can be done about the headache that is finalizers.
1
3
8
If you missed Humphrey's talk, a few more details and the paper here: https://t.co/X2AhDTD77I
stefan-marr.de
Making programs slower can be useful to find...
Next up Evaluating Candidate Instructions for Reliable Program Slowdown at the Compiler Level - Towards Supporting Fine-grained Slowdown for Advanced Developer Tooling
0
0
1
probably my most deeply-held belief: it is not possible for a slide deck to be both (1) a good presentation aide; and (2) work well as a standalone document that can be read and understood in isolation.
35
44
682
Can We Know Whether a Profiler is Accurate? Spoiler: yes we can! But it took us figuring out a way to work around the fundamental observer effect. https://t.co/4T8HnxpTBG If you're at SPLASH, don't miss our talk at VMIL today and at OOPSLA on Saturday!
0
5
5
Final MPLR talk for 2025: Dynamic and Static Code Analysis for Java Programs on Heterogeneous Hardware
0
3
6
Next up TornadoViz: Visualizing Heterogeneous Execution Patterns in Modern Managed Runtime Systems
0
1
3