Am 10.02.2017 um 17:04 schrieb Jeff King:
On Fri, Feb 10, 2017 at 04:49:02PM +0100, Johannes Schindelin wrote:
I think this is only half the story. A heavy-sha1 workload is faster,
which is good. But one of the original reasons to prefer blk-sha1 (at
least on Linux) is that resolving libcrypto.so symbols takes a
non-trivial amount of time. I just timed it again, and it seems to be
consistently 1ms slower to run "git rev-parse --git-dir" on my machine
(from the top-level of a repo).
1ms is mostly irrelevant, but it adds up on scripted workloads that
start a lot of git processes.
You know my answer to that. If scripting slows things down, we should
avoid it in production code. As it is, scripting slows us down. Therefore
I work slowly but steadily to get rid of scripting where it hurts most.
Well, yes. My question is more "what does it look like on normal Git
workloads?". Are you trading off an optimization for your giant 450MB
index workload (which _also_ could be fixed by trying do the slow
operation less, rather than micro-optimizing it) in a way that hurts
people working with more normal sized repos?
For instance, "make BLK_SHA1=Yes test" is measurably faster for me than
"make BLK_SHA1= test".
The patch does add a new runtime dependency on libcrypto.dll in my
environment. I would be surprised if it does not also with your modern
build tools.
I haven't had time to compare test suite runtimes.
I'm open to the argument that it doesn't matter in practice for normal
git users. But it's not _just_ scripting. It depends on the user's
pattern of invoking git commands (and how expensive the symbol
resolution is on their system).
It can be argued that in normal interactive use, it is hard to notice
that another DLL is loaded. Don't forget, though, that on Windows it is
not only the pure time to resolve the entry points, but also that
typically virus scanners inspect every executable file that is loaded,
which adds another share of time.
I'll use the patch for daily work for a while to see whether it hurts.
-- Hannes