From debbugs-submit-bounces@debbugs.gnu.org Fri Dec 10 17:42:52 2021 Received: (at 52001) by debbugs.gnu.org; 10 Dec 2021 22:42:52 +0000 Received: from localhost ([127.0.0.1]:48142 helo=debbugs.gnu.org) by debbugs.gnu.org with esmtp (Exim 4.84_2) (envelope-from ) id 1mvobL-00066d-SP for submit@debbugs.gnu.org; Fri, 10 Dec 2021 17:42:52 -0500 Received: from eggs.gnu.org ([209.51.188.92]:60006) by debbugs.gnu.org with esmtp (Exim 4.84_2) (envelope-from ) id 1mvobK-00066P-2K for 52001@debbugs.gnu.org; Fri, 10 Dec 2021 17:42:50 -0500 Received: from [2001:470:142:3::e] (port=51922 helo=fencepost.gnu.org) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1mvobE-0004i0-E3; Fri, 10 Dec 2021 17:42:44 -0500 DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=gnu.org; s=fencepost-gnu-org; h=MIME-Version:In-Reply-To:Date:References:Subject:To: From; bh=BbdwOCUcRQp5p3tLzq7/AB4svYMD5TqZxIspD5zf2vA=; b=kpPOuZhTXmmV/dQHovd4 yyod/xA4zvyiFhtBfwFfHLvFEob4d59P3slA5pYrZK/TRBqMbhrfaYVhckkwcKvIpI+OsV+AJO2Pd 9NKbGoBA9N50lOJknZVuqfzN27wer+8ishuZFLdoYyF8cuqjaUlvkgEG7WcQcr+0oW8ukfXcuAV1H NDJtWTDmiEFNsuB3vVYfcXQ1o/3U50R9rMW40gFmt1LCAgmkxojTPLjQHPTW/6dhLVh9QOU5Ld8+l b6mB2WZfCEiwuSduhH0jLf0hgML3RC4Rzo75aXnTrFdoqNwhH+mklDRNs35mWWBLBVcd3gF9vAMo+ tvo2rpUgykPx+w==; Received: from 91-160-117-201.subs.proxad.net ([91.160.117.201]:49556 helo=ribbon) by fencepost.gnu.org with esmtpsa (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1mvobE-0000pm-CD; Fri, 10 Dec 2021 17:42:44 -0500 From: =?utf-8?Q?Ludovic_Court=C3=A8s?= To: Ricardo Wurmus Subject: Re: bug#52001: numpy CPU dispatch probably prevents builds of python-numpy from being reproducible References: <87y25ihm3e.fsf@elephly.net> Date: Fri, 10 Dec 2021 23:42:41 +0100 In-Reply-To: <87y25ihm3e.fsf@elephly.net> (Ricardo Wurmus's message of "Sat, 20 Nov 2021 15:36:57 +0000") Message-ID: <87y24s5bfi.fsf@gnu.org> User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/27.2 (gnu/linux) MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable X-Spam-Score: -2.3 (--) X-Debbugs-Envelope-To: 52001 Cc: 52001@debbugs.gnu.org, Maxime Devos X-BeenThere: debbugs-submit@debbugs.gnu.org X-Mailman-Version: 2.1.18 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: debbugs-submit-bounces@debbugs.gnu.org Sender: "Debbugs-submit" X-Spam-Score: -3.3 (---) Hi, Ricardo Wurmus skribis: > Numpy accelerates operations with native code. Early in the build it > probes the CPU / compiler for supported instruction extensions and > then proceeds to build extensions for different features. I=E2=80=99m > guessing that this won=E2=80=99t work when building on older CPUs, so the= =20 > results of two builds on different CPUs may differ. I don=E2=80=99t see anything in the build log that suggests it might be bui= lding for a specific ISA extension (on x86_64). On the contrary, it seems to do the right thing by dispatching to the right optimized implementation at run time, in numpy/core/src/umath. Apparently this directory contains templates for home-made function multi-versioning, with variants and dispatch code generated by numpy/core/code_generators/generate_umath.py. Besides, I see the same .pyc difference as Maxime, but I don=E2=80=99t thin= k it relates to ISA extensions. Ludo=E2=80=99.