Exponential distribution hypothesis testing I want exponential distribution hypothesis testing with MATLAB. Is any test or method for that?
 A: Use the technology of the 1960s rather than the 1930s: Draw a quantile-quantile plot. 
An exponential distribution has a simple quantile function. In practice, rank the data $y_{(1)} \le \dots \le y_{(n)}$ and calculate $-\ln[1 - (i - 0.5)/n]/\bar y$ for $i = 1(1)n$. The plot of the corresponding ordered series should be straight with slope $1$ and intercept $0$. 
If a reference case is needed to judge whether small deviations from linearity are acceptable, then draw plots for several simulated samples of the same size from an exponential with the same mean and see if the plot for the real data can be identified in a line-up of suspects. 
This has the signal advantage over almost any numerical significance test that if the fit is poor, you get to see where and why it is poor. 
EDIT: Dennis's comment that the graphical analysis and a significance test can be complementary is diplomatic, even though it leaves open the question of what to do if they contradict each other. 
A specific objection to Kolmogorov-Smirnov here is the suspicion that it works rather better at telling distributions apart that differ in their middles, which for an exponential is arguably the wrong way round. More on that in Miller, R.G. 1986. Beyond ANOVA. New York: John Wiley, p.14
At a deeper level, a quantile-quantile plot from a single dataset is a sample statistic; it is the result of manipulations of a sample; only hidebound tradition insists that "sample statistic" means a numeric scalar serving as summary. Similarly a sampling distribution of quantile-quantile plots from simulations can serve as reference distribution. The idea has attracted much attention over the last few years, but was already explicit in Shewhart, W.A. 1931. Economic control of quality of manufactured product. New York: Van Nostrand. (Shewhart's title downplays the general statistical content of his monograph.) 
A: You can use, for example, two-sample Kolmogorov-Smirnov test with kstest2. (If the other distribution is also available as a sample. If it is a prespecified distribution (e.g. exponential with a priori known parameter) you can use kstest.)
I just noted from your comment that the sample size is very large, so you'll likely get a significant difference. (Which, however may not be important for you, as the test will have very high power, detecting even minuscule deviations.) Taking that into account, you might want to check a QQ-plot with
PD = ProbDistUnivParam('Exponential','mu',3)
qqplot(a,PD)

(following the example in the comment).
