Ahh, I was using a debug build of lld. After making a release build, the link time drops to 1.1 seconds! It's definitely better than gold.

On 2/19/19 10:35 PM, Ioi Lam wrote:
Hi Martin,

Thanks for the info. I didn't know that this has been discussed so recently. Should have checked befor I posted.

Anyway, I built the latest lld from GIT, but it's slower than gold on my machine. Just linking libjvm.so is 5.8s vs 3.8s.

On 2/19/19 5:23 PM, Martin Buchholz wrote:
https://openjdk.markmail.org/thread/q3layufiyjivu42c

On Tue, Feb 19, 2019 at 5:03 PM Ioi Lam <ioi....@oracle.com <mailto:ioi....@oracle.com>> wrote:

    For the impatient engineers ....

    I did some measurement between the default GNU linker "ld" vs
    "ld.gold".
    I am trying to get the fastest rebuild time after I modify one cpp
    file.
    With gold, it's down to about 1/3 of the original time.


    slowdebug (~220MB libjvm.so)

                      recompile 1 cpp file  |  relink libjvm.so only
    ld:                   33 s                   25 s
    gold 1 thread:        16 s                    9 s
    gold 8 threads:       13 s                    6 s


    fastdebug (~360MB libjvm.so)


                      recompile 1 cpp file  |  relink libjvm.so only
    ld:                   35 s                    25 s
    gold 1 thread:        18 s                    10 s
    gold 8 threads:       15 s                     6 s


    Question: do we want to add built-in support for gold into the JDK
    makefiles?


    Notes:

    To choose gold, run configure with something like:

         --with-extra-ldflags='-fuse-ld=gold
    -Wl,--threads,--thread-count,8'

    I essentially do a "make hotspot" and then move the libjvm.so into
    a JDK
    image, instead of doing a full JDK image build.

    "make hotspot" makes a copy of libjvm.so (from
    support/modules_libs/java.base/server/ to jdk/lib/server/). I
    hacked the
    Makefile to make a hard link instead to avoid the unnecessary I/O.

    libjvm.so is built with --with-native-debug-symbols=internal to avoid
    the expensive invocations of objcopy and strip.


    My environment:

    I am using gcc7.3.0 on Ubuntu 16.04.5 on a 5 year old Dell Precision
    T7600 with dual socket Xeon E5-2665 @ 2.40GHz, 64GB RAM, Samsung
    840 PRO
    SSD. I suspect gold can run even faster, but my slow SSD is
    holding it back.

    ld version   = GNU ld (GNU Binutils) 2.30
    gold version = GNU gold (GNU Binutils for Ubuntu 2.26.1) 1.11

    (These are just the versions available to me on my machine, not
    necessarily the best)



Reply via email to