mirror of https://github.com/qt/qtbase.git
b347d48704
In the case of QString, the vast majority of searches are case-
sensitive, so by inlining we make the user code call qustrchr() directly
instead of QtPrivate::findString(). In the case of QByteArray, the call
is to memchr(), which being a compiler intrinsic, may itself be inlined
or even just resolved at compile time.
In both cases, a great deal of searches use from=0.
Benchmark for QByteArray; before:
8.83207052 nsecs per iteration
22.01568546 CPU cycles per iteration, 2.49 GHz
60.00000331 instructions per iteration, 2.725 instr/cycle
21.00000281 branch instructions per iteration, 2.38 G/sec
After:
6.42561493 nsecs per iteration
16.01623130 CPU cycles per iteration, 2.49 GHz
49.00000261 instructions per iteration, 3.059 instr/cycle
18.00000211 branch instructions per iteration, 2.8 G/sec
This shaves 6 cycles in the execution and 11 instructions (3 of which
were branches), slightly improving the IPC raito, for the QByteArray
case. For QByteArrayView, there are 2 fewer instructions (1 a branch),
but the number of cycles is the same at 16.
Task-number: QTBUG-119750
Change-Id: Ica7a43f6147b49c187ccfffd179e2204ebb6a348
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
|
||
|---|---|---|
| .github/workflows | ||
| LICENSES | ||
| bin | ||
| cmake | ||
| coin | ||
| config.tests | ||
| dist | ||
| doc | ||
| examples | ||
| lib | ||
| libexec | ||
| mkspecs | ||
| qmake | ||
| src | ||
| tests | ||
| util | ||
| .cmake.conf | ||
| .gitattributes | ||
| .gitignore | ||
| .lgtm.yml | ||
| .tag | ||
| CMakeLists.txt | ||
| config_help.txt | ||
| configure | ||
| configure.bat | ||
| configure.cmake | ||
| dependencies.yaml | ||
| qt_cmdline.cmake | ||