帮忙翻译一段英文论文，100分高分急求！

Question

The performance advantage of Reviser in comparison to
full recomputations is visible best when applied repeatedly.
When we assume that the static analysis is performed on a
continuous-integration server for every check-in into a version
control system, savings are accumulated over time as depicted
on the right sides of Figures 5{11. For instance, over the 25
most recent Soot revisions, the 75% savings introduced by
Reviser accumulate to over 11,000 seconds (3 hours).
In general, the performance of Reviser depends on the
number of jump functions that need to be recomputed. The
larger the impact of the code change is, the more edges are
aected. Clearing and repropagating over an edge takes
generally longer than the initial computation for that edge.
Performance gains are thus achieved by recomputing only (a
suciently precise overapproximation of) the aected nodes
which is usually a small subset of all nodes in the program
graph. In the worst case, all edges must be recomputed and
the algorithm degenerates to a recomputation with some
overhead, caused by changeset computation and wrapper
lookups as described in Section 4. This, for instance, happens
for the JUnit update from version 4.10 to version 4.11.

Answer 1

The performance advantage of Reviser in comparison to
full recomputations is visible best when applied repeatedly.
When we assume that the static analysis is performed on a
continuous-integration server for every check-in into a version
control system, savings are accumulated over time as depicted
on the right sides of Figures 5{11. For instance, over the 25
most recent Soot revisions, the 75% savings introduced by
Reviser accumulate to over 11,000 seconds (3 hours).
In general, the performance of Reviser depends on the
number of jump functions that need to be recomputed. The
larger the impact of the code change is, the more edges are
aected. Clearing and repropagating over an edge takes
generally longer than the initial computation for that edge.
Performance gains are thus achieved by recomputing only (a
suciently precise overapproximation of) the aected nodes
which is usually a small subset of all nodes in the program
graph. In the worst case, all edges must be recomputed and
the algorithm degenerates to a recomputation with some
overhead, caused by changeset computation and wrapper
lookups as described in Section 4. This, for instance, happens
for the JUnit update from version 4.10 to version 4.11.

比较修正的性能优势
全是最好的反复应用时可见recomputations。
当我们假定静态分析是在一个
每个旅客托运的一个版本持续集成服务器
控制系统，随着时间的积累储蓄所
右侧的数字5 { 11。例如，在25
最近修订了烟尘，节省75%
审校积累超过11000秒（3小时）。
在一般情况下，修正的性能取决于
跳跃，需要重新计算功能。
代码更改的影响越大,越边缘

aected。清算和repropagating优势需要
一般超过最初的计算优势。性能因此只通过验算(
suciently精确overapproximation aected节点
这是通常的一个小子集中的所有节点计划
图。在最坏的情况下,所有的边缘必须重新计算
重新计算一些算法退化
开销,由于变更集计算和包装
查找如第四节所述。例如,这发生
对于JUnit更新从4.10版到4.11版。

Answer 2

相比校订者的性能优势
全部重新计算可见最好反复应用。
当我们假定在一个执行静态分析
持续集成服务器每签入到版本
控制系统、储蓄积累随着时间的描述
右边的图5 { 11。例如,在25岁
最近的烟灰修订,75%的储蓄了
校订者积累超过11000秒(3个小时)。
一般来说,校订者取决于的性能
跳数的函数需要重新计算。的
更大的代码变化的影响,更多的边缘
aected。清算和repropagating优势需要
一般超过最初的计算优势。
性能因此只通过验算(
suciently精确overapproximation aected节点
这是通常的一个小子集中的所有节点计划
图。在最坏的情况下,所有的边缘必须重新计算
重新计算一些算法退化
开销,由于变更集计算和包装
查找如第四节所述。例如,这发生
对于JUnit更新从4.10版到4.11版。