Best Solar Flood Lights,Solar Motion Flood Lights,Solar Sensor Security Light,Commercial Solar Flood Lights Jiangmen Liangtu Photoelectric Technology Co., Ltd. , https://www.liangtulight.com
Recently, the author passed the test of two rabbits that use the same set-top box of the RK3128 platform as the AnTu rabbit (AnTuTu V5.7), but got a disparity. Among them, China Mobile's tender test a RK3128 box total score of 24154, a brand RK3128 box total score of 14673, a difference of 9481 points! This is particularly evident in the single-threaded floating point, which actually differs by 5959 points on the two boxes.
The single-threaded capability can be reflected in actual performance under certain low power operation settings. Users need to know where their device performance limits are, and they also need to know how much performance they can use daily.
In the inherent sense, the increase in the number of CPU cores will be considered as an improvement in the overall performance. This is not unfounded. However, in actual use, many APP applications and games that we often come into contact with are not optimized for multi-core CPUs. In most cases, only one or two CPU cores are enabled.
Therefore, the single-core single-threaded performance is also considered to have an important impact on the security Bunny, and its test scores are also included in the scoring system, so that the new version of security Bunny test score scores increased again. 
RK3128 box security rabbit run comparison 
China Mobile Bid Testing RK3128 Box Security Bunny Running Points 
A certain brand RK3128 box security Bunny Run points
From the results, the author suspects that there may be some special treatment for libm (dynamic link library file) (the profile shows more calls to libm when running the single thread floating test), such as running libm for security Bunny The data input by the function when calculating, can directly return the result without calculation, so that it can be much faster in speed than actually calculated by the VFP instruction.
In order to further confirm whether this box has been cheated against AnTutu, the author tested the running scores of these two box Geek bench. The following focuses only on the floating point part. 
Geek bench test floating point score for RK3128 box 
China Mobile Bidding Test A RK3128 Box Geek Bench Testing Floating Point Score 
A brand RK3128 box Geek bench test floating point score
According to Geek bench's running test, there is no obvious difference in the floating point score between the RK3128 box and a certain brand RK3128 box from China Mobile's bidding test. The result of running the score in Table 1 can be seen clearly. Bunny APK made "special optimization", which affected the final run results.
In terms of performance, the same Cortex-A7 kernel SOC, if not cheated by technology under the same conditions, the single-threaded floating point index score results will not be much different. For this reason, in order to further verify, the author also found an additional Cortex-A7 box and the aforementioned two RK3128 boxes based on the same version of AnTutu (AnTuTu V5.7) and rerun the subtest. 
Cortex-A7 Box Security Bunny Running Test
From the results, the single-threaded floating-point index scores of the A brand A31S box and a certain brand RK3128 box are all at 800+, while the China Mobile bidding test RK3128 box has a single-threaded floating-point index score, which is 7 times that of the other two boxes. Moreover, it further verified the cheating of the security Bunny Run points.
In fact, after searching on the Internet, the author discovered that such cheating is not the first time. Some netizens have been propaganda for RK3288 to advertise the security of the Bunny to run at 43000+. They have done very detailed assessments and used lib injection for floating point operations. Ways to cheat.
It is understood that the early security Bunny version is through the direct call of libm pow, cos, sin and other mathematical functions to test, because the implementation of these functions in libm is open source, so vulnerable to targeted changes. For example, when the security Bunny is detected running, it is possible to return these functions directly to a specific result without actually calculating it. This can greatly save the test time and increase the score.
For the above considerations, the older version of the security rabbit has already implemented an implementation of these library functions in the libabenchmark.so. The so is an internal security test program, which can be effective to a certain extent. Avoiding the cheating problem caused by the vendor's modification of the open source libm.so does not mean that this completely solves the problem of cheating.
At present, there have been a number of more subtle injection methods for cheating, that is, when loading the so-called So rabbits, use the self-doing functions to replace the function in the Antutu library. To achieve the purpose of running high scores.
Sofa butler technology security Bunny evaluation is the most commonly used performance testing tool for domestic Android mobile phones and tablets. At present, it is also commonly used to evaluate the performance of Android set-top boxes, in order to obtain better score results on a specific test item, on top of this Cheating is also common.
March 28, 2021