I was curious to compare the performance of the STL string implementation versus ATL CString, using Visual Studio 2019, so I wrote some simple C++ benchmark code for this purpose.

I also added into the mix a custom string pool allocator based on some code from The Old New Thing blog, that I modified and bug-fixed. This allocator basically maintains a singly-linked list of chunks, and string memory is carved from each chunk just increasing a string pointer. When there isn’t enough memory in the current chunk to serve the allocation, a new chunk is allocated. The new chunk is safely linked to the previous chunk list, and the memory for the requested strings is carved from this new chunk. The linked list of chunks is traversed at destruction time to properly release the allocated memory blocks.

I measured the times to create and fill string vectors, and the times to sort the same vectors.

TL;DR: The STL string performance is great! You can improve creation times with a custom pool allocator.

These times are measured for vectors storing each kind of strings: STL’s wstring, ATL’s CString (i.e. CStringW in Unicode builds), and the strings created using the custom string pool allocator.

This is a sample run (executed on a Windows 10 64-bit Intel i7 PC):

String benchmark: STL vs. ATL vs. custom string pool allocator

As you can note, the best creation times are obtained with the custom string pool allocator. This was expected, as the allocation strategy of carving string memory from pre-allocated blocks is very efficient.

On the other hand, regarding the sorting times, STL and the custom pool strings perform very similarly.

ATL’s CString, which is based on CoW (Copy on Write), shows the worst times for both creation and sorting.

Benchmark Variation: Tiny Strings

It’s also possible to run the benchmark with short strings, triggering the SSO (compile the code #define’ing TEST_TINY_STRINGS).

String benchmark with tiny strings: STL vs. ATL vs. custom string pool allocator

As you can see in this case, thanks to the SSO, STL strings win by an important margin in both creation and sorting times.

Note: Making the Sorting Comparison Uniform

I’d also like to clarify that, to make the sorting comparison more uniform, I used custom comparator functions with std::sort, invoking wcscmp for each kind of strings. In fact, spelunking in the VS 2019 STL and ATL implementation code (Thank You, Step Into Specific menu), I found that std::wstring invokes wmemcmp (file <xstring>, line #235), while ATL::CString uses wcscmp (file <cstringt.h>, line #564). Probably std::wstring uses wmemcmp because wstring instances can contain embedded NULs.