It’s also a good reminder that clarity of layout often beats more “clever” designs, especially when the dataset fits comfortably in memory.
My best documented case was a 10x speed up from removing a double lookup that was killing caches.
If you would have asked me to bet on which one would have had the bigger impact I would have split the difference.
My second favorite was similar. Two functions making a call instead of sharing the answer. Profiler said 10% cumulative. I removed half. Instead of 5% I got 20%. Which just demonstrates how much data a profiler cannot show you.
Profilers lie and some more than most. I’ve gotten 3x from code with a perfectly rectangular profile output.
Part of it comes down to a trick game devs steal from finance: give each task a budget and keep it to the budget even if it’s not the tall tent pole.
You should not spend 10% of your response time on telemetry and logging combined. Yet I pulled 10% TTFB out of just the logging and telemetry code on a project. It was a frog boiling situation. Every new epic used the new code and determining the cumulative cost wasn’t easy.
I've nearly always had a variable length string or other complex structure that was being hashed, not their handles.
Back in my early career in C, this would be a generic API to hash and store void pointers, but the pointers were not being hashed. The domain-specific hash function needed to downcast and perform the appropriate remote memory access to fetch the variable-length material that was actually being hashed.
You can use more elaborate probe and relocation schemes, but just choosing the less full bucket and resizing if both choices are full gets you surprisingly far.
This constraint allows making a linear array of all the 4 billion values, with the key as array index, which fits in 16 GiB. Another 500 MiB is enough to have a bit indicating present or not for each.
Perhaps text strings as keys and values would give a more interesting example...
The hash table has the significant advantage of having a much smaller minimum size.
> Perhaps text strings as keys and values would give a more interesting example
Keep reading to "If keys and values are larger than 32 bits"
If you actually only have a handful of entries in the table, it is measurable in bytes.
A linear array of all 4 billion possible values will occupy 16 GiB (of virtual memory) upfront. We have then essentially replaced the hash table with a radix tree --- that made up of the page directories and tables of the VM system. If only the highest and lowest value are present in the table, then we only need the highest and lowest page (4 kB or whatever) to be mapped. It's not very compact for small sets; storing N numbers in random locations could require as many as N virtual memory pages to be committed.
> To meet property (3) [if the key 0 is present, its value is not 0] ... "array index plus one" can be stored rather than "array index".
If hash code can take any value in [0,2^32), how to define a special value for empty buckets? The more common solution is to have a special key, not a special hash code, for empty slots, which is easier to achieve. In addition, as the author points out, supporting generic keys requires to store 32-bit hash values. With the extra 4 bytes per bucket, it is not clear if this implementation is better than plain linear probing (my favorite). The fastest hash table implementations like boost and abseil don't often use Robin Hood hashing.
struct slot {
uint32_t key;
uint32_t value;
}You could work around that with a union or casts with explicit alignment constraints, but this is the shortest way to express that.
union slot {
uint64_t keyvalue;
struct {
uint64_t key: 32;
uint64_t value: 32;
};
};
Edit: Note also that you can cast a pointer to slot to a pointer to uint64_t and that does not break strict aliasing rules.
Yay, we've got an equivalent of SIB byte but as three (six?) separate opcodes. Well, sub-opcodes.
It's a shame though that Zcmp extension didn't get into RVA23 even as an optional extension.
Zcmp is only for embedded applications without D support.
You wouldn't want an instruction with up to 13 destinations in high performance designs anyways.
If you want load/store pair, we already have that, you can just interpret two adjacent 16-bit load or stores as a single 32-bit instruction.