mirror of
https://github.com/lux-apps/papers.git
synced 2026-10-30 18:15:21 +00:00
zeekay
abf99d05fc
Restructured from flat .tex/.pdf files into per-paper subdirectories. Added Antje Worring as co-author on all papers. Synced PDFs to pdfs/ for download site.
29 lines
856 B
TeX
29 lines
856 B
TeX
\section{Integer Operation Performance}
|
|
|
|
|
|
Radix integers decompose values into 2-bit or 4-bit blocks. A uint32 with 2-bit blocks has 16 blocks; each block operation involves boolean gates across the radix.
|
|
|
|
\begin{table}[h]
|
|
\centering
|
|
\begin{tabular}{@{}lrr@{}}
|
|
\toprule
|
|
\textbf{Operation (uint32)} & \textbf{EPYC} & \textbf{M2} \\
|
|
\midrule
|
|
Add & 210 ms & 310 ms \\
|
|
Sub & 215 ms & 315 ms \\
|
|
Mul & 1.8 s & 2.7 s \\
|
|
Eq (comparison) & 180 ms & 265 ms \\
|
|
Lt (comparison) & 320 ms & 470 ms \\
|
|
Mux (conditional) & 340 ms & 500 ms \\
|
|
\midrule
|
|
AddScalar (plaintext + ct) & 105 ms & 155 ms \\
|
|
MulScalar (plaintext $\times$ ct) & 420 ms & 620 ms \\
|
|
\bottomrule
|
|
\end{tabular}
|
|
\caption{Integer operation latency (PN10QP27, uint32, 2-bit radix).}
|
|
\label{tab:int-ops}
|
|
\end{table}
|
|
|
|
Scalar operations are cheaper because one operand is plaintext, halving the gate count.
|
|
|