Probably not helpful, but I tend to rely on the jvm optimisations and just
-server. I figured this is an area where a little knowledge is a dangerous
thing.

At the very least I would have a realistic benchmark suite to prove to
myself that these gains were worth it. In my experience the performance
bottleneck is always in design.

Just my 2p, although with my self-confessed ignoramus status it is more
like 0.5p :).
On 14 May 2015 21:46, "Lee Spector" <lspec...@hampshire.edu> wrote:

> I'd like to get more guidance on how to specify :jvm-opts for maximum
> performance. I've received some help on this topic from people on this list
> in the past (thanks!), but I'm pretty sure that I'm still not doing things
> right.
>
> I'm almost always interested in maximum performance for long-running,
> compute-intensive and memory-intensive processes, almost never caring much
> at all about startup time or anything else.
>
> I also run my code on different machines, with different numbers of cores
> and amounts of memory, and would prefer to be able to put something in my
> project.clj that will do something reasonable regardless of what machine
> it's running on.
>
> I run my code with "lein run", and I'd like whatever I need to "run fast"
> to be in project.clj.
>
> I don't know a lot about JVM options, and I've tried to figure out what to
> specify for :jvm-opts by asking questions here and web searches, but I'm
> not at all confident that I'm doing it right yet. And because my systems
> are also stochastic, it's not easy for me to do timing tests on the options
> I've tried.
>
> I think there are options relevant to memory and also garbage collection
> and maybe also compilation... and what else? I wish there was a simple
> switch to get maximum performance of the sort I've outlined here (or at
> least a reasonable stab at it), but I gather that there isn't.
>
> Anyway, here's what I've been using recently, which just deals with memory
> and GC (and maybe not in the best way):
>
>   :jvm-opts ~(let [mem-to-use (long (* (.getTotalPhysicalMemorySize
>
>  (java.lang.management.ManagementFactory/getOperatingSystemMXBean))
>                                        0.8))]
>                [(str "-Xmx" mem-to-use)
>                 (str "-Xms" mem-to-use)
>                 "-XX:+UseG1GC"])
>
> After seeing Alex's post I thought that maybe I should add "-server", as
> follows:
>
>   :jvm-opts ~(let [mem-to-use (long (* (.getTotalPhysicalMemorySize
>
>  (java.lang.management.ManagementFactory/getOperatingSystemMXBean))
>                                        0.8))]
>                [(str "-Xmx" mem-to-use)
>                 (str "-Xms" mem-to-use)
>                 "-XX:+UseG1GC"
>                 "-server"])
>
> Is that right? Does it make sense? What does "-server" do? Also, should I
> also be using "^:replace"?
>
> I've looked in
> https://github.com/technomancy/leiningen/blob/master/sample.project.clj
> in hopes that this would say more about this stuff, but it doesn't say
> anything about -server or ^:replace.
>
> Looking into the compilation options, it looks from
> https://github.com/technomancy/leiningen/wiki/Faster that I should be
> specifying:
>
> :jvm-opts ^:replace []
>
> This is also familiar to me from some earlier discussions. But how would I
> combine this with the memory/GC/server(?) options above?
>
> A guess would be that maybe I should do this:
>
>   :jvm-opts ^:replace
>               ~(let [mem-to-use (long (* (.getTotalPhysicalMemorySize
>
>  (java.lang.management.ManagementFactory/getOperatingSystemMXBean))
>                                        0.8))]
>                [(str "-Xmx" mem-to-use)
>                 (str "-Xms" mem-to-use)
>                 "-XX:+UseG1GC"
>                 "-server"
>                 "-XX:-TieredCompilation"])
>
> Note that this guess involves changing a + to a - in the last line, which
> was suggested for the opposite purpose at
> https://github.com/technomancy/leiningen/wiki/Faster -- but I don't know
> if it's legitimate.
>
> Is this a reasonable thing to do to get maximum performance for
> long-running, compute-intensive and memory-intensive processes?
>
> Is the tiered compilation thing maybe already done by including "-server"?
>
> I'm probably at least somewhat confused about several different issues
> here...
>
> Any help or pointers would be appreciated.
>
> Thanks,
>
>  -Lee
>
>
>
>
> > On May 14, 2015, at 3:47 PM, Alex Miller <a...@puredanger.com> wrote:
> >
> > Gah. Meant in project.clj:
> >
> > :jvm-opts ^:replace ["-server"]  ;; maybe also set max heap with
> "-Xmx1g" in there
> >
>
> --
> You received this message because you are subscribed to the Google
> Groups "Clojure" group.
> To post to this group, send email to clojure@googlegroups.com
> Note that posts from new members are moderated - please be patient with
> your first post.
> To unsubscribe from this group, send email to
> clojure+unsubscr...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/clojure?hl=en
> ---
> You received this message because you are subscribed to the Google Groups
> "Clojure" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to clojure+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>

-- 
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your 
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to