@curious_coding

PhD on high troughput bioinformatics @ ETH Zurich; IMO, ICPC, Xoogler, Rust, road-cycling, hiking, wild camping, photography https://xitter.cf/t.co/wBv2zN4F80

 Zurich
curiouscoding.nl
Joined October 2021
  • Tweets1.9K
  • Following81
  • Followers1.2K
188Media
1
3
1
13
5,308
Ragnar {Groot Koerkamp} 🦋 retweeted
Mohammed Mudasser@Md_mudasser6h
Key takeaways from @curious_coding session at #P99CONF : 🔹 Memory access bottlenecks priority queues. 🔹 SIMD-friendly layouts beat traditional heaps. 🔹 SimdQuickHeap is up to 2× faster than radix heaps. Data layout matters! ⚡ @P99CONF @ScyllaDB #ScyllaDB
1
5
146
Live now! – at Karlsruhe, Baden-Württemberg
P99CONF
@P99CONF7h
Binary heaps are cache-inefficient and cause a cache miss on nearly every pop. In 20 minutes, @curious_coding dives into SimdQuickHeap and how it results in a priority queue that is up to 10x faster than binary and d-ary heaps. ow.ly/bcB951003al #ScyllaDB #P99CONF
5
224
Thursday, 6:10pm CEST! – at Birkweiler, Deutschland
ScyllaDB
@ScyllaDBOct 19
Our @P99CONF keynotes feature top engineers and system architects like @DorLaor, @curious_coding, @georg_kreuzmayr, and @reconditerose. At our free & virtual #P99CONF, you also get to interact with each of them. Join our October 21-22 event > ow.ly/3brE50ZYYCk #ScyllaDB
1
4
273
I'll be speaking at P99 later this month!
ScyllaDB
@ScyllaDBOct 8
The SimdQuickHeap recursively partitions the data until the smallest element is found, and stores the pivots and buckets between them. @curious_coding will show how this is the fastest priority queue at our free @P99CONF on Oct. 21-22. ow.ly/3S9T50ZVVby #ScyllaDB #P99CONF
1
8
562
Ragnar {Groot Koerkamp} 🦋 retweeted
Rust for Bioinformatics@rust4bioAug 29
SIMD + Rust + pattern matching = Sassy, a small, sharp tool from @curious_coding et al., for fast sequence search. Proof that you don't need a giant framework to move the needle, sometimes you just need really well-written low-level code. 🔍🦀 academic.oup.com/bioinformat…

Sassy: fuzzy searching DNA sequences using SIMD

AbstractMotivation. Approximate string matching (ASM) is the problem of finding all occurrences of a pattern in a text while allowing up to k errors. Many

academic.oup.com
10
19
1,156
Ragnar {Groot Koerkamp} 🦋 retweeted
antisense.@razoralignMay 26
Sassy: Fuzzy Searching DNA Sequences using SIMD academic.oup.com/bioinformat…
1
6
20
1,398
Ragnar {Groot Koerkamp} 🦋 retweeted
Samuel Lampa - Also: smllmp.bsky.social@smllmpJun 5
As said before, this is a super useful (and *fast*) tool, that we've already been using quite a bit for troubleshooting spurious primer matching and the like. Warmly recommended!
Ragnar {Groot Koerkamp} 🦋@curious_codingJun 4
To all folks dealing with biological data: do you ever need to check if your reads contain barcodes/adapters/primers/...? Or off-target matches? Sassy is the tool to use! A super fast implementation of "approximate string matching" with a grep-like CLI. curiouscoding.nl/posts/sassy…
1
2
437
To all folks dealing with biological data: do you ever need to check if your reads contain barcodes/adapters/primers/...? Or off-target matches? Sassy is the tool to use! A super fast implementation of "approximate string matching" with a grep-like CLI. curiouscoding.nl/posts/sassy…
1
9
1
24
3,439
Reposts would be great 🙏 I really think this can be a very useful tool for many folks outside my usual bubble.
1
194
New preprint! The SimdQuickHeap is the fastest priority queue, by far. 2x faster than a radix heap, and up to 10x faster than binary heaps. Coauthored with Johannes Breitling and Marvin Williams.
Data Structures and Algorithms Papers@AlgorithmPapersApr 29
SimdQuickHeap: The QuickHeap Reconsidered Johannes Breitling, Ragnar Groot Koerkamp, Marvin Williams arxiv.org/abs/2604.25681 [𝚌𝚜.𝙳𝚂]
Priority queues are data structures that maintain a dynamic collection of elements and allow inserting new elements and removing the smallest element. The most widely known and used priority queue is likely the implicit binary heap, even though it is has frequent cache misses and is hard to optimize using e.g. SIMD instructions. We introduce the SimdQuickHeap, a variant of the QuickHeap that was introduced by Navarro and Paredes in 2010. As suggested by the name, the data structure bears some similarity to QuickSort. We modify the data layout of the original QuickHeap to have all pivots adjacent in memory, with elements between consecutive pivots stored in dedicated buckets. This allows efficient SIMD implementations for both partitioning of buckets and scanning the list of pivots to find the bucket to append newly inserted elements to. The SimdQuickHeap has amortized expected complexity O(log n) per operation, which improves to O(frac 1Wlog n) in non-degenerate cases, where W is the n

ALT Priority queues are data structures that maintain a dynamic collection of elements and allow inserting new elements and removing the smallest element. The most widely known and used priority queue is likely the implicit binary heap, even though it is has frequent cache misses and is hard to optimize using e.g. SIMD instructions. We introduce the SimdQuickHeap, a variant of the QuickHeap that was introduced by Navarro and Paredes in 2010. As suggested by the name, the data structure bears some similarity to QuickSort. We modify the data layout of the original QuickHeap to have all pivots adjacent in memory, with elements between consecutive pivots stored in dedicated buckets. This allows efficient SIMD implementations for both partitioning of buckets and scanning the list of pivots to find the bucket to append newly inserted elements to. The SimdQuickHeap has amortized expected complexity O(log n) per operation, which improves to O(frac 1Wlog n) in non-degenerate cases, where W is the n

1
4
1
18
1,289
more replies
As often: the SimdQuickHeap is a conceptually very simple data structure that thus can be easily optimized using SIMD, leading to pretty big speed gains.
1
2
129
github.com/RagnarGrootKoerka…

GitHub - RagnarGrootKoerkamp/quickheap: The fastest priority queue

The fastest priority queue. Contribute to RagnarGrootKoerkamp/quickheap development by creating an account on GitHub.

github.com
1
66
OMG! By popular demand (🤗) I'm going to speak at P99 again! This time about the SimdQuickHeap preprint that just came out: the fastest priority queue by 2x (or 10x comparing to a binary heap).
P99CONF
@P99CONFMay 1
We've been working hard behind the scenes to bring you an unforgettable P99CONF 2026. Registration is now open, so take a look at who is coming back and the new speakers we have on board > p99conf.io/2026/04/01/be-par… #ScyllaDB #P99CONF
1
2
8
865
xitter.cf/i/status/2049441073718…
Ragnar {Groot Koerkamp} 🦋@curious_codingApr 29
New preprint! The SimdQuickHeap is the fastest priority queue, by far. 2x faster than a radix heap, and up to 10x faster than binary heaps. Coauthored with Johannes Breitling and Marvin Williams.
107
SimdQuickHeap: The QuickHeap Reconsidered Johannes Breitling, Ragnar Groot Koerkamp, Marvin Williams arxiv.org/abs/2604.25681 [𝚌𝚜.𝙳𝚂]
Priority queues are data structures that maintain a dynamic collection of elements and allow inserting new elements and removing the smallest element. The most widely known and used priority queue is likely the implicit binary heap, even though it is has frequent cache misses and is hard to optimize using e.g. SIMD instructions. We introduce the SimdQuickHeap, a variant of the QuickHeap that was introduced by Navarro and Paredes in 2010. As suggested by the name, the data structure bears some similarity to QuickSort. We modify the data layout of the original QuickHeap to have all pivots adjacent in memory, with elements between consecutive pivots stored in dedicated buckets. This allows efficient SIMD implementations for both partitioning of buckets and scanning the list of pivots to find the bucket to append newly inserted elements to. The SimdQuickHeap has amortized expected complexity O(log n) per operation, which improves to O(frac 1Wlog n) in non-degenerate cases, where W is the n

ALT Priority queues are data structures that maintain a dynamic collection of elements and allow inserting new elements and removing the smallest element. The most widely known and used priority queue is likely the implicit binary heap, even though it is has frequent cache misses and is hard to optimize using e.g. SIMD instructions. We introduce the SimdQuickHeap, a variant of the QuickHeap that was introduced by Navarro and Paredes in 2010. As suggested by the name, the data structure bears some similarity to QuickSort. We modify the data layout of the original QuickHeap to have all pivots adjacent in memory, with elements between consecutive pivots stored in dedicated buckets. This allows efficient SIMD implementations for both partitioning of buckets and scanning the list of pivots to find the bucket to append newly inserted elements to. The SimdQuickHeap has amortized expected complexity O(log n) per operation, which improves to O(frac 1Wlog n) in non-degenerate cases, where W is the n

1
1
6
1,400
Ragnar {Groot Koerkamp} 🦋 retweeted
Geoff Langdale
@geofflangdaleMar 26
Replying to @valigo
Getting good at GPU programming will make you a better coder on modern CPUs. A throughput-oriented mentality, while critical on GPUs, is also a way better way of coaxing better performance out of the wide, deeply-pipelined, SIMD-enabled monstrosities that are modern CPUs.
1
3
37
1,775
Ragnar {Groot Koerkamp} 🦋 retweeted
Jennifer Doleac
@jenniferdoleac22 Dec 2025
Me 15 years ago, as a PhD student: “It seems like all the interesting questions have been answered & there are no useful studies left to do.” Me now, as a funder focused on evidence-based criminal justice policy: “The biggest constraint we face is finding researchers with time to engage in important new projects.” Truly, this is the most pressing problem I have. I need an army of scholars who understand causal methods & who would jump at an opportunity to study a program or policy. Would you like to help? Email me. 🙏
25
71
5
659
111,790
Ragnar {Groot Koerkamp} 🦋 retweeted
The Lunduke Journal
@LundukeJournal23 Dec 2025
Microsoft has set a goal to “eliminate every line of C and C++ from Microsoft by 2030.” What are they going to try to replace that C & C++ code with? You guessed it. Rust. And they’re going to use AI to do the “Rust re-write” at an insane speed. “Our strategy is to combine AI *and* Algorithms to rewrite Microsoft’s largest codebases. Our North Star is “1 engineer, 1 month, 1 million lines of code”. You read that right. One million lines of code, per engineer, per month. Pure insanity. This kind of decision making is common among those with a deeply held, delusional faith in the Cult of Rust. Take battle tested code, and re-write it (without a clear benefit to the end user) at a recklessly rapid rate. Then force others to adopt that rewritten code before it is ready or properly tested. All while holding a delusional belief that your new Rust code is superior in all ways, and is inherently bug free thanks to the divine nature of Rust. We learned this from a post by Galen Hunt, Distinguished Engineer at Microsoft Research. linkedin.com/posts/galenh_pr…
1,155
965
881
8,825
1,884,263
Load more