Coming soon - High performance integers merging Permalink
There is nothing simpler than merging two sorted vectors of integers, right? Well, it gets difficult when required performance is in hundreds of millions int...
Posts by Tag
- performance 9
- C++ 8
- cache 4
- compilers 4
- bitset 3
- research 2
- events 2
- Packet Processing 2
- packet processing 1
- CPU 1
- glibc 1
- simd 1
performance
Avx throttling (Part I) - UPDATED
SIMD instruction Modern CPUs contain so-called vector extensions or SIMD instructions. SIMD stands for Single Instruction Multiple Data. For x86-64 CPUs exam...
Prefetching in memory-intensive applications
In the previous post we have been looking on various, sometimes intrusive and complicated methods of compacting data structures for providing better memory a...
The one with shrinking structure (Part II)
In the last post we have been looking on to how size of data, specifically packet descriptors, impacts processing performance. As it turns out, reducing data...
The one with multi-versioning (Part II)
In the last post we have focused on popcount operation and function multi-versioning. In this post I show shortcomings of multi-versioning usied with popcnt ...
The one with multi-versioning (Part I)
Introduction In previous post I have described benefits of inlining and surprising impact of some compilation flags using Find First Set as an example. Today...
The one with compiler flag magic
Introduction At some point in the project I’m working on, I have noticed that variable-size bit set implementation is minor, but still performance bottleneck...
Interesting article about Speculative Prefetching Permalink
Planning to give it a try and find some use cases in processing high volume network traffic. Building packet descriptors from packet stream maybe?
The one with shrinking structure (Part I)
Nowadays real-world applications of C++ gets smaller and more specialised every day. It was decided almost by acclamation that business logic in new projects...
C++
Old and new function multi-versioning
I have been touching multi-versioning already in two previous articles: Part I, Part II but I wanted to return to this topic once more (and probably not for ...
Avx throttling (Part I) - UPDATED
SIMD instruction Modern CPUs contain so-called vector extensions or SIMD instructions. SIMD stands for Single Instruction Multiple Data. For x86-64 CPUs exam...
Prefetching in memory-intensive applications
In the previous post we have been looking on various, sometimes intrusive and complicated methods of compacting data structures for providing better memory a...
The one with shrinking structure (Part II)
In the last post we have been looking on to how size of data, specifically packet descriptors, impacts processing performance. As it turns out, reducing data...
The one with multi-versioning (Part II)
In the last post we have focused on popcount operation and function multi-versioning. In this post I show shortcomings of multi-versioning usied with popcnt ...
The one with multi-versioning (Part I)
Introduction In previous post I have described benefits of inlining and surprising impact of some compilation flags using Find First Set as an example. Today...
The one with compiler flag magic
Introduction At some point in the project I’m working on, I have noticed that variable-size bit set implementation is minor, but still performance bottleneck...
The one with shrinking structure (Part I)
Nowadays real-world applications of C++ gets smaller and more specialised every day. It was decided almost by acclamation that business logic in new projects...
cache
Prefetching in memory-intensive applications
In the previous post we have been looking on various, sometimes intrusive and complicated methods of compacting data structures for providing better memory a...
The one with shrinking structure (Part II)
In the last post we have been looking on to how size of data, specifically packet descriptors, impacts processing performance. As it turns out, reducing data...
Interesting article about Speculative Prefetching Permalink
Planning to give it a try and find some use cases in processing high volume network traffic. Building packet descriptors from packet stream maybe?
The one with shrinking structure (Part I)
Nowadays real-world applications of C++ gets smaller and more specialised every day. It was decided almost by acclamation that business logic in new projects...
compilers
Old and new function multi-versioning
I have been touching multi-versioning already in two previous articles: Part I, Part II but I wanted to return to this topic once more (and probably not for ...
The one with multi-versioning (Part II)
In the last post we have focused on popcount operation and function multi-versioning. In this post I show shortcomings of multi-versioning usied with popcnt ...
The one with multi-versioning (Part I)
Introduction In previous post I have described benefits of inlining and surprising impact of some compilation flags using Find First Set as an example. Today...
The one with compiler flag magic
Introduction At some point in the project I’m working on, I have noticed that variable-size bit set implementation is minor, but still performance bottleneck...
bitset
The one with multi-versioning (Part II)
In the last post we have focused on popcount operation and function multi-versioning. In this post I show shortcomings of multi-versioning usied with popcnt ...
The one with multi-versioning (Part I)
Introduction In previous post I have described benefits of inlining and surprising impact of some compilation flags using Find First Set as an example. Today...
The one with compiler flag magic
Introduction At some point in the project I’m working on, I have noticed that variable-size bit set implementation is minor, but still performance bottleneck...
research
Coming soon - High performance integers merging Permalink
There is nothing simpler than merging two sorted vectors of integers, right? Well, it gets difficult when required performance is in hundreds of millions int...
Interesting article about Speculative Prefetching Permalink
Planning to give it a try and find some use cases in processing high volume network traffic. Building packet descriptors from packet stream maybe?
events
The one about conference - day two
Few updates from second day of Open Source Summit in Lyon.
The one about conference - day one
Here are some first thoughts after day one of Open Source Summit in Lyon.
Packet Processing
Prefetching in memory-intensive applications
In the previous post we have been looking on various, sometimes intrusive and complicated methods of compacting data structures for providing better memory a...
The one with shrinking structure (Part II)
In the last post we have been looking on to how size of data, specifically packet descriptors, impacts processing performance. As it turns out, reducing data...
packet processing
The one with shrinking structure (Part I)
Nowadays real-world applications of C++ gets smaller and more specialised every day. It was decided almost by acclamation that business logic in new projects...
CPU
Avx throttling (Part I) - UPDATED
SIMD instruction Modern CPUs contain so-called vector extensions or SIMD instructions. SIMD stands for Single Instruction Multiple Data. For x86-64 CPUs exam...
glibc
Avx throttling (Part I) - UPDATED
SIMD instruction Modern CPUs contain so-called vector extensions or SIMD instructions. SIMD stands for Single Instruction Multiple Data. For x86-64 CPUs exam...
simd
Coming soon - High performance integers merging Permalink
There is nothing simpler than merging two sorted vectors of integers, right? Well, it gets difficult when required performance is in hundreds of millions int...