There are two open gcc bugzilla bugs for complex divide. 59714 complex divide is surprising on targets with FMA 60646 investigate improved complex division algorithms
Either might be appropriate. 59714 started my investigation, and my solution resolves that issue, while my solution seems closer to the request of 60646, except 60646 seems to target the middle-end instead of libgcc. Anyway, once I created the tarball and gzipped it, I found it shrank to 100Kbytes. Rather than have further delays, I've attached the gzip file to this email. If someone has trouble unzipping, I can email the .tar file to them directly (500Kbytes). - patrick mcgehearty On 1/25/2021 2:48 PM, Joseph Myers wrote:
On Mon, 25 Jan 2021, Patrick McGehearty wrote:On 1/21/2021 5:04 PM, Joseph Myers wrote:On Fri, 18 Dec 2020, Patrick McGehearty via Gcc-patches wrote:TEST DataI'd still like to see the test data / code used to produce the accuracy and performance results made available somewhere (presumably with a link then being provided in the commit message).I've packaged up the code/scripts/README and some data in a directory. I had it ready last time and meant to provide it, but it slipped my mind while I was preparing the patch. Total size is under a megabyte, so moving it around should not be a problem. Is this sort of development support code usually kept somewhere in the gcc tree or should it be placed elsewhere, like in my personal github directory?I think that sort of thing (when it's too big to just attach to the email with the patch - if it can be attached to the email, or to a relevant open bug in Bugzilla, that's the simplest thing to do) would typically go somewhere else rather than in the GCC tree.
cdiv_measure.tar.gz
Description: GNU Zip compressed data