On Fri, 1 Jul 2022 09:13:52 GMT, Daniel Fuchs <dfu...@openjdk.org> wrote:
>> In case we read all bytes from an `InputStream` we don't need wrapping with >> `BufferedInputStream` as the bytes are not written into internal buffer. >> With removal of redundant buffer we save 8 kB of allocated memory. > >> In case we read all bytes from an InputStream we don't need wrapping with >> BufferedInputStream as the bytes are not written into internal buffer. With >> removal of redundant buffer we save 8 kB of allocated memory. > > I believe that in that case the bytes might get double buffered, depending on > whether the buffer used by `readAllBytes` is bigger than the buffer used by > `BufferedInputStream`, since IIAMNM `readAllBytes()` will call `read(byte[], > off, len)` - but I otherwise agree that regardless of that wrapping in a > `BufferedInputStream` is always wasteful in this case. Good find! @dfuch as of current implementation of `InputStream.readNBytes()` it relies on `InputStream.DEFAULT_BUFFER_SIZE` which is 8192 and equals to `DEFAULT_BUFFER_SIZE` in `BufferedInputStream`. ------------- PR: https://git.openjdk.org/jdk/pull/9343
