4 ms·

This is a bit contrived example because if just searching for an integer in a large array, the memory bandwidth is going to be the bottle neck. When dealing wi

by exDM69 12y ago

This is a bit contrived example because if just searching for an integer in a large array, the memory bandwidth is going to be the bottle neck. When dealing with small arrays (like the example of 0-250 integers, ie. less than 1 kilobyte), the figures are around 20-120 nanoseconds. In other words, the difference is around one main memory reference ("cache miss") between the best and the worst case. While this is somewhat interesting, the lesson to take home should be that most of the time, smart memory usage is more important than what the CPU executes. Warming up the caches (e.g. __builtin_prefetch) is as efficient an optimization as rewriting and unrolling the entire loop and should be done first. If the caches are warm and the search is still the bottleneck, then it's time to consider the other optimizations.

alphonse23 12y ago
I also think he should have covered the time it would take to sort an array. A linear search wouldn't need to be sorted, but a binary search would. He made his cut off at around 64 elements, but if you'd include the time needed by binary search to sort the array, the cut off limit would be a higher than that -- but would also introduce a lot of other complexities to consider in a production system.
- lgeek 12y ago
  I really doubt there's any data size for which sort + binary search is going to be faster than linear search. The former has higher complexity (so it's going to be slower for large inputs) and more expensive operations (so it will be slower for small inputs). You can also think about it as a sort of reductio ad absurdum: Assume that the array is actually already sorted and you use a sorting algorithm with O(n) complexity for already sorted data. In this ideal case you'd need to do n (for sorting) + log n (binary search) predictable operations. For linear search you'd only need to n operations of the same type. In practice you'd need to do n log n operations for sorting, I don't think you can avoid unpredictable branches, and instead of just loading elements sometimes you'll move them around.
- exDM69 12y ago
  I think the assumption is that you sort once and search several times (enough to amortize the cost of the sort). The linear search algorithm also had an early exit, so the array has to be sorted for that too. That makes this a fair comparison, but I agree that seeing a linear unsorted search would have been interesting too.
dllthomas 12y ago
Getting stuff out of cache is also worthwhile in some cases. I had some message passing between cores, and apparently the CPU wasn't good at telling that my chance of using a populated and sent message slot again was zero. Adding a clflush after send significantly reduced my number of cache misses.
- exDM69 12y ago
  This is an aspect of multi-core programming. As the programmer, you must be painfully aware of the cache lines you're touching and which cores touch them. Having two or more cores touch a cache line will cause expensive core-to-core synchronization taking place. The case is easier in a message passing system like yours, where the other core is clearly the sender and can flush the caches. In kernel space, you can disable caches or use write-combine or write-through caches for memory regions as needed but this isn't really available in userspace. This can happen in very unintuitive places. For example, in Java, array.size() reads from the same cache line as the beginning of the array. If you're partitioning an array processing routine for many cores, you should not call array.size() to avoid very expensive synchronization.

by exDM69 12y ago

alphonse23 12y ago

I also think he should have covered the time it would take to sort an array. A linear search wouldn't need to be sorted, but a binary search would. He made his cut off at around 64 elements, but if you'd include the time needed by binary search to sort the array, the cut off limit would be a higher than that -- but would also introduce a lot of other complexities to consider in a production system.

lgeek 12y ago
I really doubt there's any data size for which sort + binary search is going to be faster than linear search. The former has higher complexity (so it's going to be slower for large inputs) and more expensive operations (so it will be slower for small inputs). You can also think about it as a sort of reductio ad absurdum: Assume that the array is actually already sorted and you use a sorting algorithm with O(n) complexity for already sorted data. In this ideal case you'd need to do n (for sorting) + log n (binary search) predictable operations. For linear search you'd only need to n operations of the same type. In practice you'd need to do n log n operations for sorting, I don't think you can avoid unpredictable branches, and instead of just loading elements sometimes you'll move them around.
exDM69 12y ago
I think the assumption is that you sort once and search several times (enough to amortize the cost of the sort). The linear search algorithm also had an early exit, so the array has to be sorted for that too. That makes this a fair comparison, but I agree that seeing a linear unsorted search would have been interesting too.

dllthomas 12y ago

Getting stuff out of cache is also worthwhile in some cases. I had some message passing between cores, and apparently the CPU wasn't good at telling that my chance of using a populated and sent message slot again was zero. Adding a clflush after send significantly reduced my number of cache misses.

exDM69 12y ago
This is an aspect of multi-core programming. As the programmer, you must be painfully aware of the cache lines you're touching and which cores touch them. Having two or more cores touch a cache line will cause expensive core-to-core synchronization taking place. The case is easier in a message passing system like yours, where the other core is clearly the sender and can flush the caches. In kernel space, you can disable caches or use write-combine or write-through caches for memory regions as needed but this isn't really available in userspace. This can happen in very unintuitive places. For example, in Java, array.size() reads from the same cache line as the beginning of the array. If you're partitioning an array processing routine for many cores, you should not call array.size() to avoid very expensive synchronization.