From 8ba79f39a9a9d1f0a68a89102e7ae5cbd7220dc2 Mon Sep 17 00:00:00 2001
From: nindanaoto <30498364+nindanaoto@users.noreply.github.com>
Date: Wed, 28 Apr 2021 11:48:38 +0900
Subject: [PATCH] Update README.md

---
 README.md | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/README.md b/README.md
index 8e1732c..468dd40 100644
--- a/README.md
+++ b/README.md
@@ -2,8 +2,10 @@
 CUDA-accelerated Torus Fully Homomorphic Encryption Library. This fork is maintained as a sub project of Virtual Secure Platform.
 
 v1.0_beta -- release on Mar/14/2018 original [cuFHE](https://github.com/vernamlab/cuFHE)
+
 v1 -- kvsp v29 compatible version.
-v2 -- refctored & reduced shared memory usage (depends on TFHEpp for parameter set select, runnable on old GPUs like GTX 1060Ti but slow)
+
+v2 -- refctored & reduced shared memory usage (depends on [TFHEpp](https://github.com/virtualsecureplatform/TFHEpp) for parameter set select, runnable on old GPUs like GTX 1060Ti but slow)
 
 ## What is cuFHE?
 The cuFHE library is an open-source library for Fully Homomorphic Encryption (FHE) on CUDA-enabled GPUs. It implements the TFHE scheme [CGGI16][CGGI17] proposed by Chillotti et al. in CUDA C++. Compared to the [TFHE lib](https://github.com/tfhe/tfhe) which reports the fastest gate-by-gate bootstrapping performance on CPUs, the cuFHE library yields almost same performance per SM. Since GPU has a lot of SMs (128 in A100), cuFHE gives better performace if there are enough number of parallely evaluable tasks. The cuFHE library benefits greatly from an improved CUDA implementation of the number-theoretic transform (NTT) proposed in the [cuHE library](https://github.com/vernamlab/cuHE) [Dai15] by Dai and Sunar.