-
performance is important on small embedded devices too :)
-
According to the original author,
> which is the latest version in the series of MurmurHash functions - the new version is faster, more robust, and its variants can produce 32- and 128-bit hash va…
-
Currently if neither AES or FoldedMultiply is available the slow path involves 5 multiplies for short inputs. It should be theoretically possible to reduce this to 3. An alternative update function is…
-
Dear All:
I doubled short key hashing speed by announcing wyhash_final3 version.
The speed up is caused by branchless code for short keys.
This speed up may not be detected by other benchma…
-
I've noticed that the table lists PRVHASH64 as having 2.3GB/s and 50 cycles while in the benchmark file the speed is 4.3 GB/s and 37 cycles. Same for PRVHASH64S - 9.8GB/s and 129/160 cycles.
-
After more research on what I've got with PRVHASH, I've come up with a "minimal" implementation, which is good for table hash use. It's the smallest high-quality 64-bit hash function with good securit…
-
Hi Aleksey,
thanks for your work, appreciate your nifty etude despite not understanding how it works.
Thought it would be fun to compare your superstong PRVhash vs my superweak FNV1A-Pippip, sadly…
-
I followed the following steps to install
!pip install --upgrade "mxnet
-
https://github.com/gzm55/hash-garage/blob/master/nmhash.h
This is a new simple 32bit hash.
a) friendly to low profile hardware, using only narrow multiplications (16 x 16 -> 16), and 32bit additio…
gzm55 updated
3 years ago
-
From the `android-arm-corellium` builder (https://build.golang.org/log/6d0506edef76d21049f50f3d2a77cf686d063476):
```
--- FAIL: TestSmhasherSmallKeys (0.15s)
smhasher_test.go:95: unexpected num…