Re: [XEN PATCH 0/4] xen: rework compat headers generation

[Anthony PERARD <anthony.perard@xxxxxxxxxx>]

On Wed, Jun 01, 2022 at 05:17:36PM +0000, Andrew Cooper wrote:
> On 01/06/2022 17:59, Anthony PERARD wrote:
> > Patch series available in this git branch:
> > https://xenbits.xen.org/git-http/people/aperard/xen-unstable.git 
> > br.build-system-xen-include-rework-v1
> >
> > Hi,
> >
> > This patch series is about 2 improvement. First one is to use $(if_changed, 
> > )
> > in "include/Makefile" to make the generation of the compat headers less 
> > verbose
> > and to have the command line part of the decision to rebuild the headers.
> > Second one is to replace one slow script by a much faster one, and save time
> > when generating the headers.
> >
> > Thanks.
> >
> > Anthony PERARD (4):
> >   build: xen/include: use if_changed
> >   build: set PERL
> >   build: replace get-fields.sh by a perl script
> >   build: remove auto.conf prerequisite from compat/xlat.h target
> >
> >  xen/Makefile                 |   1 +
> >  xen/include/Makefile         | 106 ++++---
> >  xen/tools/compat-xlat-header | 539 +++++++++++++++++++++++++++++++++++
> >  xen/tools/get-fields.sh      | 528 ----------------------------------
> 
> Excellent.  I was planning to ask you about this.  (I also need to
> refreshing my half series cleaning up other bits of the build.)
> 
> One trivial observation is that it would probably be nicer to name the
> script with a .pl extension.

Sound fine, I don't think it matter much here.

> Any numbers on what the speedup in patch 3 is?

Yes, on my machine when doing a full build, with `ccache` enabled, it
saves about 1.17 seconds (out of ~17s), and without ccache, it saves
about 2.0 seconds (out of ~37s).

Without ccache:

before:
    $ for i in `seq 4`; do time make -j$(nproc) -s O=build 2>/dev/null 
>/dev/null; rm -r build; done
    make --no-print-directory -j$(nproc) -s O=build > /dev/null  244.98s user 
29.24s system 683% cpu 40.146 total
    make --no-print-directory -j$(nproc) -s O=build > /dev/null  47.05s user 
11.50s system 332% cpu 17.610 total
    make --no-print-directory -j$(nproc) -s O=build > /dev/null  47.35s user 
11.22s system 330% cpu 17.733 total
    make --no-print-directory -j$(nproc) -s O=build > /dev/null  47.31s user 
11.23s system 333% cpu 17.577 total

after:
    $ for i in `seq 4`; do time make -j$(nproc) -s O=build 
2>/dev/null>/dev/null; rm -r build; done
    make --no-print-directory -j$(nproc) -s O=build 2> /dev/null > /dev/null  
237.28s user 25.97s system 667% cpu 39.456 total
    make --no-print-directory -j$(nproc) -s O=build 2> /dev/null > /dev/null  
37.60s user 8.20s system 282% cpu 16.214 total
    make --no-print-directory -j$(nproc) -s O=build 2> /dev/null > /dev/null  
37.95s user 8.67s system 279% cpu 16.651 total
    make --no-print-directory -j$(nproc) -s O=build 2> /dev/null > /dev/null  
38.02s user 8.40s system 280% cpu 16.545 total

And without ccache:

before:
    $ for i in `seq 4`; do time make -j$(nproc) -s O=build 
2>/dev/null>/dev/null; rm -r build; done
    make --no-print-directory -j$(nproc) -s O=build 2> /dev/null > /dev/null  
206.37s user 22.19s system 640% cpu 35.695 total
    make --no-print-directory -j$(nproc) -s O=build 2> /dev/null > /dev/null  
221.45s user 22.26s system 667% cpu 36.537 total
    make --no-print-directory -j$(nproc) -s O=build 2> /dev/null > /dev/null  
233.95s user 23.80s system 686% cpu 37.518 total
    make --no-print-directory -j$(nproc) -s O=build 2> /dev/null > /dev/null  
234.27s user 23.83s system 680% cpu 37.923 total

after:
    $ for i in `seq 4`; do time make -j$(nproc) -s O=build 
2>/dev/null>/dev/null; rm -r build; done
    make --no-print-directory -j$(nproc) -s O=build 2> /dev/null > /dev/null  
198.62s user 18.64s system 642% cpu 33.841 total
    make --no-print-directory -j$(nproc) -s O=build 2> /dev/null > /dev/null  
202.91s user 19.46s system 655% cpu 33.912 total
    make --no-print-directory -j$(nproc) -s O=build 2> /dev/null > /dev/null  
224.42s user 20.89s system 680% cpu 36.025 total
    make --no-print-directory -j$(nproc) -s O=build 2> /dev/null > /dev/null  
222.93s user 21.29s system 683% cpu 35.708 total


> Are the generated compat headers identical before and after this
> series?  If yes, I'm very tempted to ack and commit it straight away.

Yes, the headers are identical. Hopefully, I've managed to check with
all compat headers enabled, since that depends on .config.

Cheers,

-- 
Anthony PERARD