Saving your string: Never adding an endpoint

In fact, I have a lot to say:

 

The test shows that the higher the number of tests, the lower the efficiency. This should be because the memory allocation becomes the main time overhead.

 

Because the public server is too fast, the number of tests on the server is 10 or 100 times that on my PC (otherwise, the running time is less than 1 second !). Based on the above, this may be due to the limited improvement in GCC efficiency.

 

The efficiency of vs2008 was much higher than I previously mentioned. I made it into a debug version. Some versions are used for around 1/3 of the unavailable versions !!

 

Vc6 has a version that is about 20% slower than vs2008, but both parties use it honestly. vs2008 takes about 170% of vc6. Oh, my God, didn't I block the reference count of vc6!

 

Stlport is useful. If no version is available, the efficiency improvement will be quite dazzling. If yes, it will still be able to increase by more than 30%.

 

I am wronged by the Intel compiler ICC, and the efficiency of ICC V11 is much higher than that of V10. compared with vc6, ICC V10 provides a space increase of less than 20%. If stlport is used, the efficiency of ICC is not as high as that of vc6, and there is little difference between ICC and vc6 in some versions. the ICC V11 has significantly increased the level. Compared with the vs2008 version, ICC can squeeze out a 20% improvement in efficiency. if the parallel support option is enabled, the effect on my dual-core machine is even more shocking. (HoweverProgramThe file size is a bit large...). I guess the vc6 Runtime Library is somewhere incompatible with some optimized binary of ICC V11, which eventually causes ICC V11 to give up its compatibility.