Commit 9cc8f882 authored by homsm's avatar homsm Committed by overleaf
Browse files

Update on Overleaf.

parent 7cd3e57e
......@@ -79,4 +79,7 @@ Example: four instances of a multiplier with cost=8\\
\item A \textcolor{red}{schedule} $\tau: V_S \rightarrow \mathbb{Z}^{\geq0}$ determines the starting times of operations. It is feasible iff
$$\forall (v_i,v_j) \in E_S: \quad \tau(v_j)-\tau(v_i) \geq w(v_i) \defeq w(v_i,\beta(v_i))$$
\item \textcolor{red}{latency} $L$ of a schedule is the time between start node $v_0$ and end node $v_n$: $$L=\tau(v_n)-\tau(v_0)$$
\end{itemize}
\ No newline at end of file
\end{itemize}
\ownsubsection{Weighted Constraint Graph (10-40)}
See section scheduling without resource constraints
\ No newline at end of file
......@@ -14,14 +14,18 @@ energy and a longer lifetime (battery)
\ownsubsection{Power Consumption of CMOS Processors (9-12)}
Main sources:
\begin{compactitem}
\item Dynamic power consumption: (De-)Charging capacitors
\item Short circuit power consumption (between supply rails during switching)
\item Leakage: leaking diodes and transistors (more important for smaller technology)
\item Dynamic power consumption:
\begin{itemize}
\item (De-)Charging capacitors
\item Short circuit power consumption (between supply rails during switching)
\end{itemize}
\item Leakage and static Power: leaking diodes and transistors (more important for smaller technology)
\end{compactitem}
\begin{equation*}
\begin{alignedat}{2}
&\textbf{Power:} \qquad &&P \sim \alpha C_L V^2_{dd}f \\
&\textbf{Power:} \qquad &&P \sim \alpha C_L V^2_{dd}f \\
&\textbf{Energy:} \qquad &&E \sim \alpha C_L V^2_{dd} f t = \alpha C_L V^2_{dd} \cdot \text{(\#cycles)}\\
&\textbf{Delay:} \qquad \qquad &&\tau \sim C_L \frac{V_{dd}}{(V_{dd}-V_T)^2}\\
\end{alignedat}
......@@ -35,6 +39,7 @@ Main sources:
$V_T$ : & \quad Threshold Voltage $V_T \ll V_{dd}$
\end{tabularx}
$P\propto V_{dd}^2$ and $f\propto V_{dd}^2$ so $P\propto V_{dd}^3$ \\
leakage ignored
\ownsubsection{Basic Techniques (9-16)}
\textbf{Power Supply Gating}: Cut off power supply to inactive units to reduce leakage
......@@ -47,16 +52,19 @@ For each core: $P_2 = P_1/4$, \quad $E_2 = E_1/8$\\
$$\implies E_2 = \frac{1}{4}E_{\text{normal}}$$
The same equation for pipelining (each module half as fast)
\textbf{VLIW Architectures} \newline
Large degree of parallelism with simple hardware architecture
\textbf{VLIW Architectures (9-18)} \newline
VLIW= very long instruction word\\
Large degree of parallelism with simple hardware architecture (explicit parallelism, parallelization is done offline)
\begin{center}
\includegraphics[width=0.9\columnwidth]{sys2}
\end{center}
~\newline
\textbf{Dynamic Voltage Scaling} \newline
\textbf{Dynamic Voltage Scaling (9-20)} \newline
Adapt voltage and frequency to save energy. \newline
Since $\tau \propto \frac{1}{V_{dd}} \rightarrow f \propto V_{dd}$ \\
\textbf{Optimal strategy}: If possible, running constantly at lowest possible voltage and frequency minimizes energy consumption for DVS.
\ownsubsection{YDS Algorithm for Offline Scheduling (9-35)}
......@@ -110,7 +118,7 @@ Desired: Shutdown only during long idle times $\rightarrow$ Tradeoff between sav
\textbf{Procrastination Schedule}: Do YDS scheduling, set $V_{DD} = max(V_{DD}, V_{crit})$. Try to retard the execution of tasks to aggregate enough time for sleeping
~\newline
\ownsubsection{Battery Operated Systems and Energy Harvesting (9 - 45}
\ownsubsection{Battery Operated Systems and Energy Harvesting (9 - 45)}
Battery powered device rely on energy harvesting and storing power. \\
A simple structur to control the battery:\\
\includegraphics[width=1\linewidth]{images/power-model.png}
......@@ -150,6 +158,8 @@ Since the perfect use function $u(t)$ was chosen with knowledge of all harvested
The energy consumption will then be chosen, such that the battery is full, when the harvesting function is decreasing, thus after the battery is fully charged, the use function should decrease. \\
\ownsubsection{Finite Horizon Control}
To find an optimal use function.\\
We do not know the future energy harvesting correctly. To deal with that, we use a estimation of the future harvested energy $\tilde{p}(\tau)$ \\
Instead of calculating the optimal use function only once and stick to it, we now calculate the best use function repedeatly using the actual battery level. \\
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment