|
From: | Philippe Mathieu-Daudé |
Subject: | Re: [Qemu-devel] [PATCH 02/23] shippable: build using all available cpus |
Date: | Mon, 8 May 2017 11:42:14 -0300 |
User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Thunderbird/45.8.0 |
diff --git a/.shippable.yml b/.shippable.yml index 231c29b620..5170486ff9 100644 --- a/.shippable.yml +++ b/.shippable.yml @@ -20,4 +20,4 @@ build: ci: - unset CC - ./configure ${QEMU_CONFIGURE_OPTS} --target-list=${TARGET_LIST} - - make -j2 + - make -j$(getconf _NPROCESSORS_ONLN)Did you ever check what this generates on Shippable? AFAICT they only have one core for each docker image. I following the core+1 pattern although that may not make as much sense for low core counts.From the 'Job node info': $ sudo docker info Architecture: x86_64 CPUs: 2 Checking at 'ci' stage: $ echo "container proc:" `getconf _NPROCESSORS_ONLN` `getconf _NPROCESSORS_CONF` container proc: 2 2 I think the correct cores+1 pattern should instead be: make -j$(($(getconf _NPROCESSORS_ONLN) + 1)) So we stay building at full speed if Shippable eventually upgrade/allow to use more cores.
For what it's worth using this pattern builds last on average <8min vs >9min previously :)
[Prev in Thread] | Current Thread | [Next in Thread] |