On Jun 8 20:07, Achim Gratz wrote: > Corinna Vinschen writes: > > I'm still not quite sure what problem this solves. > > Nothing earth-shattering, I just intensely dislike those long hex > strings. But thanks for asking, I just had a closer look at the > real-world results and it turns out that there is an effect beyond > (questionable, I know) aesthetical considerations. > > > Is it faster than plain SHA512 checksums? > > I'd say it's a noticeable speed improvement, depending on where you get > that file from. There are 17146 checksums in the 32bit setup.ini at the > moment. That makes for a saving of 720132 bytes with Base64-URL vs. Hex > encoding. I have another patch in testing that allows to use compressed > setup.ini (and xz compression) from "local" directories and since the > normal install is across the network, the reduced number of bytes to > push over the slow connection is making quite an impact. But I was more > surprised to see that it is actually noticeably faster even when run > locally from an SSD.
I'm certainly not de-valuing your patch, but that makes me suspicious. The impact of reading from a local SSD should be rather low. That reminds me of the buffer size problem slowing down SHA512 evaluation, see git diff 0c7564..8648b05. greping setup sources, it seems setup is using smaller buffers in more places, e.g. compress_bz::read uses 4K buffer compress_gz::read 16K io_stream::copy 16K YY_INPUT 8K Especially YY_READ_BUF_SIZE defined as 8K might be a culprit here. Can you speed up reading the ini file noticably by defining YY_READ_BUF_SIZE to 65536 in inilex.ll? If so, is the impact of using base64 still an issue, performance-wise? > The changes in setup are really small. Upset doesn't need to change, at > least not immediately. Ok. Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Maintainer cygwin AT cygwin DOT com Red Hat
pgpLSzUPs6npc.pgp
Description: PGP signature