diff --git a/bibfile.bib b/bibfile.bib
index cb15ea2ca2c96d851dcc8fd69948a8831bfd0573..c716fea7f10ddefb276926b6196597f6c8ca5ede 100644
--- a/bibfile.bib
+++ b/bibfile.bib
@@ -201,3 +201,38 @@
 	biburl = {http://dblp.org/rec/bib/journals/corr/abs-1105-4125},
 	bibsource = {dblp computer science bibliography, http://dblp.org}
 }
+
+@inproceedings{Fletcher:2012:SPA:2382536.2382540,
+	author = {Fletcher, Christopher W. and Dijk, Marten van and Devadas, Srinivas},
+	title = {A Secure Processor Architecture for Encrypted Computation on Untrusted Programs},
+	booktitle = {Proceedings of the Seventh ACM Workshop on Scalable Trusted Computing},
+	series = {STC '12},
+	year = {2012},
+	isbn = {978-1-4503-1662-0},
+	location = {Raleigh, North Carolina, USA},
+	pages = {3--8},
+	numpages = {6},
+	url = {http://doi.acm.org/10.1145/2382536.2382540},
+	doi = {10.1145/2382536.2382540},
+	acmid = {2382540},
+	publisher = {ACM},
+	address = {New York, NY, USA},
+	keywords = {encrypted computation, secure processors},
+} 
+
+@phdthesis{fletcher2013ascend,
+	title={Ascend: An architecture for performing secure computation on encrypted data},
+	author={Fletcher, Christopher Wardlaw},
+	year={2013}
+}
+
+@article{ren2013design,
+	title={Design space exploration and optimization of path oblivious ram in secure processors},
+	author={Ren, Ling and Yu, Xiangyao and Fletcher, Christopher W and Van Dijk, Marten and Devadas, Srinivas},
+	journal={ACM SIGARCH Computer Architecture News},
+	volume={41},
+	number={3},
+	pages={571--582},
+	year={2013},
+	publisher={ACM}
+}
diff --git a/sections/applications-and-extensions.tex b/sections/applications-and-extensions.tex
index de16f782030e0bfe6b34ab5f2d5df1d31b0dc215..cc37c8989fdab4d290cb7f878345d4ae0ddc7286 100644
--- a/sections/applications-and-extensions.tex
+++ b/sections/applications-and-extensions.tex
@@ -30,11 +30,14 @@
 		
 		PathORAM is particularly amenable to hardware design because of its simplicity and low on-chip storage requirements.
 
-		For example,~\textcite{Maas:EECS-2014-89} built a hardware implementation of a Path ORAM based secure processor using FPGAs and the Convey platform. % chktex 8
+		\textcite{Maas:EECS-2014-89} built a hardware implementation of a Path ORAM based secure processor using FPGAs and the Convey platform. % chktex 8
+
+		\textcite{Fletcher:2012:SPA:2382536.2382540, fletcher2013ascend} and \textcite{ren2013design} built a simulator for a secure processor based on PathORAM\@.
 
 		\note{
 			Due to its simplicity, PathORAM is particularly good for silicon implementations.
 			For example, Martin Mass has build such implementation using FPGAs.
+			Fletcher and Ren built a simulator for a processor based on PathORAM\@.
 		}
 	\end{frame}
 
diff --git a/sections/evaluation.tex b/sections/evaluation.tex
index 383eff8cac8c0a5b10f3b4c474c5c44c459249ec..8def5447c82396fa419bd6100551d895aba9597f 100644
--- a/sections/evaluation.tex
+++ b/sections/evaluation.tex
@@ -78,6 +78,7 @@
 				\item 
 					For $Z \ge 5$, the usage of a subtree is close to the number of buckets in it.
 					Also holds for $Z = 4$.
+					% TODO
 				\item 
 					For the levels close to the root, the bucket load is indeed 1 block.
 				\item 
diff --git a/sections/overview-of-other-orams.tex b/sections/overview-of-other-orams.tex
index 45fce608864a9d422199386ebb12d8a151f70193..115b0ea265d611d93af18add21046134c85c26fe 100644
--- a/sections/overview-of-other-orams.tex
+++ b/sections/overview-of-other-orams.tex
@@ -41,9 +41,9 @@
 			They analyzed space and time complexity of the systems.
 			The result is on the table.
 
-			Computational overhead is a composite of communication, encryption/decryption and client running overheads.
-			Communication overhead measures how much information is transmitted during a round of read and write.
+			Communication overhead measures the number of rounds of communication for one access.
 			Server and client respectively show how much space is used by an ORAM on the server and on the client.
+			Computational overhead is a composite of communication, encryption/decryption and client running overheads \emph{per access}.
 
 			I am not going to elaborate on all ORAMs, but it is clear that one of them wins in every category.
 			This is why we have chosen it for our secure cloud.
diff --git a/sections/problem-definition.tex b/sections/problem-definition.tex
index bc50533b73270aa752b6c8989c194eed996ab3c9..f5f1ea57cf433123cc9e2db925e40da76cbf1e3b 100644
--- a/sections/problem-definition.tex
+++ b/sections/problem-definition.tex
@@ -6,7 +6,7 @@
 		
 		\begin{itemize}
 			\item 
-				The client fetches/stores data in atomic units --- \emph{blocks} --- of size $B$ each
+				The client fetches/stores data in atomic units --- \emph{blocks} --- of size $B$ bytes each
 			\item 
 				Let $N$ be the working set --- number of distinct data blocks stored in ORAM
 		\end{itemize}