https://osmocom.org/https://osmocom.org/favicon.ico?16647414092018-01-11T17:22:51ZOpen Source Mobile CommunicationsOsmoTRX - Bug #2826: osmo-trx: different output in convolve test between x86 vs neon vs neon-vfpv4https://osmocom.org/issues/2826?journal_id=71212018-01-11T17:22:51Zpespin
<ul><li><strong>Related to</strong> <i><a class="issue tracker-1 status-3 priority-3 priority-high3 closed" href="/issues/2721">Bug #2721</a>: OsmoTRX build verification job for arm</i> added</li></ul> OsmoTRX - Bug #2826: osmo-trx: different output in convolve test between x86 vs neon vs neon-vfpv4https://osmocom.org/issues/2826?journal_id=71542018-01-16T10:31:23Zpespin
<ul><li><strong>Related to</strong> <i><a class="issue tracker-1 status-3 priority-2 priority-default closed" href="/issues/2828">Bug #2828</a>: some osmo-trx tests failing in i586 OBS machine</i> added</li></ul> OsmoTRX - Bug #2826: osmo-trx: different output in convolve test between x86 vs neon vs neon-vfpv4https://osmocom.org/issues/2826?journal_id=71652018-01-16T23:01:46Zttsou
<ul></ul><p>I still need to look into the specific numbers as well as reproduce the issue locally, however, I suspect the issue relates to IEEE-754 floating point conformance.</p>
<p>The ARMv7 NEON instruction set is not IEEE-754 compliant, which can very well lead to inconsistent floating point output across builds.</p>
<p>An interesting question would be if the test fails on ARMv8 where the NEON instruction set is IEEE-754 compliant.</p>
<p>I will have more details shortly.</p> OsmoTRX - Bug #2826: osmo-trx: different output in convolve test between x86 vs neon vs neon-vfpv4https://osmocom.org/issues/2826?journal_id=71662018-01-17T00:01:40Zpespin
<ul></ul><p>Thanks for looking into it. If you have time and didn't see it yet, please read <a class="external" href="https://lists.osmocom.org/pipermail/openbsc/2018-January/011655.html">https://lists.osmocom.org/pipermail/openbsc/2018-January/011655.html</a></p> OsmoTRX - Bug #2826: osmo-trx: different output in convolve test between x86 vs neon vs neon-vfpv4https://osmocom.org/issues/2826?journal_id=114382018-09-18T16:02:03Zpespin
<ul></ul><p><a class="user active" href="https://osmocom.org/users/51">ttsou</a> friendly ping in case you forgot about this topic :-)</p> OsmoTRX - Bug #2826: osmo-trx: different output in convolve test between x86 vs neon vs neon-vfpv4https://osmocom.org/issues/2826?journal_id=122912018-10-20T19:26:01Zlaforge
<ul><li><strong>Assignee</strong> changed from <i>ttsou</i> to <i>tnt</i></li></ul> OsmoTRX - Bug #2826: osmo-trx: different output in convolve test between x86 vs neon vs neon-vfpv4https://osmocom.org/issues/2826?journal_id=126962018-11-26T14:32:39Zlaforge
<ul><li><strong>Priority</strong> changed from <i>Normal</i> to <i>High</i></li></ul> OsmoTRX - Bug #2826: osmo-trx: different output in convolve test between x86 vs neon vs neon-vfpv4https://osmocom.org/issues/2826?journal_id=129902018-12-20T13:23:57Ztnt
<ul><li><strong>Status</strong> changed from <i>New</i> to <i>In Progress</i></li></ul><p>Looking at that now ...</p>
<p>I mean direct exact text compare for such kind of test is never going to work because of the IEEE precision ... sometimes because of non-perfect compliant operation, or because we purposefully do a faster less precise operation, or because operations in different orders, ... and so in either case you convole_test should have the test vector and expected results built-in and do the compare internally tolerating small deltas (sometime like X% deviation + a small epsilon).</p>
<p>But looking at the logs posted, some of the deviations seems a bit large, so I'll be checking that first.</p> OsmoTRX - Bug #2826: osmo-trx: different output in convolve test between x86 vs neon vs neon-vfpv4https://osmocom.org/issues/2826?journal_id=129972018-12-21T15:55:07Ztnt
<ul></ul><p>Ok.</p>
<pre><code>- Removed unused / unsupported features from the codebase<br /> - Reworked the convole_test utility to be more useful and consistent across platform<br /> - That new test utility actually uncovered a bug in the neon vfp4 implementation, so fixed that.<br /> - Re-enabled convolve_test by default.</code></pre>
<p>All pushed on gerrit for review.</p> OsmoTRX - Bug #2826: osmo-trx: different output in convolve test between x86 vs neon vs neon-vfpv4https://osmocom.org/issues/2826?journal_id=132912019-02-08T16:31:31Ztnt
<ul><li><strong>Status</strong> changed from <i>In Progress</i> to <i>Resolved</i></li></ul>