Ben {chmark} Adams
@benaadams.me
Co-founder and CTO of Illyriad Games, 9x Microsoft MVP, contributor to Nethermind. Prev dotnetfdn board member.
Just in case; am I doing this right @sinclairinat0r.com @blowdart.me ?
May 28, 2025 at 7:31 AM
Just in case; am I doing this right @sinclairinat0r.com @blowdart.me ?
February 15, 2025 at 11:30 AM
UInt256 Multiply x2.5 times faster than non-simd w mulx_u64
Made harder as Avx512 only has mullo_epi64 and no mulhi_epi64 🥹
Haven't seen any simd implementation for this before; and unfortunately drops through 256bit and 128bit rather than being pure 512 github.com/NethermindEt...
Made harder as Avx512 only has mullo_epi64 and no mulhi_epi64 🥹
Haven't seen any simd implementation for this before; and unfortunately drops through 256bit and 128bit rather than being pure 512 github.com/NethermindEt...
Vectorized Multiply (on Avx512) by benaadams · Pull Request #45 · NethermindEth/int256
~30% faster (current Scalar has vectorized Adds) on AMD Ryzen 9 7950X; x2.5 times faster than vanilla MULX
Method
Environment
A
B
Mean
Error
StdDev
Ratio
Multiply_UInt256
HWIntrinsic=0
(619(...
github.com
February 11, 2025 at 7:56 PM
UInt256 Multiply x2.5 times faster than non-simd w mulx_u64
Made harder as Avx512 only has mullo_epi64 and no mulhi_epi64 🥹
Haven't seen any simd implementation for this before; and unfortunately drops through 256bit and 128bit rather than being pure 512 github.com/NethermindEt...
Made harder as Avx512 only has mullo_epi64 and no mulhi_epi64 🥹
Haven't seen any simd implementation for this before; and unfortunately drops through 256bit and 128bit rather than being pure 512 github.com/NethermindEt...
So @blowdart.me told me to update my handle to a domain and now it says "invalid handle"
Was this a security test? 🤣
Was this a security test? 🤣
January 23, 2025 at 12:54 PM
So @blowdart.me told me to update my handle to a domain and now it says "invalid handle"
Was this a security test? 🤣
Was this a security test? 🤣