|
From: <jhr...@t-...> - 2003-10-18 12:41:35
|
Hello, when running one of my regression tests built with ICC 7.1 under the newly released valgrind I see the following assertion failure: ---------- ==6555== Memcheck, a.k.a. Valgrind, a memory error detector for x86-linux. ==6555== Copyright (C) 2002-2003, and GNU GPL'd, by Julian Seward. ==6555== Using valgrind-20031012, a program supervision framework for x86-linux. ==6555== Copyright (C) 2000-2003, and GNU GPL'd, by Julian Seward. ==6555== Estimated CPU clock rate is 1720 MHz ==6555== For more details, rerun with: -v ==6555== test_vector double, unbounded_array v1 = v2 = [3](1,2,3) v1.assign_temporary (v2) = [3](1,2,3) v1.swap (v2) = [3](1,2,3) [3](1,2,3) - v1 = [3](-1,-2,-3) conj (v1) = [3](1,2,3) v1 + v2 = [3](2,4,6) v1 - v2 = [3](0,0,0) element_prod (v1, v2) = [3](1,4,9) 1. * v1 = [3](1,2,3) N * v1 = [3](3,6,9) v1 * 1. = [3](1,2,3) v1 * N = [3](3,6,9) v2 += v1 = [3](2,4,6) v2 -= v1 = [3](1,2,3) v1 *= 1. = [3](1,2,3) v1 *= N = [3](3,6,9) sum (v1) = 6 norm_1 (v1) = 6 Memcheck: mc_translate.c:1108 (memcheck_instrument): Assertion `u_in->size == 4 || u_in->size == 16' failed. sched status: Thread 1: status = Runnable, associated_mx = 0x0, associated_cv = 0x0 ==6555== at 0x804A5B8: (within /usr/local/lib/boost_dev/boost/libs/numeric/ublas/test1/bin/test1/intel-linu x/release/runtime-link-dynamic/test1) ==6555== by 0x8067622: (within /usr/local/lib/boost_dev/boost/libs/numeric/ublas/test1/bin/test1/intel-linu x/release/runtime-link-dynamic/test1) ==6555== by 0x804A2AA: (within /usr/local/lib/boost_dev/boost/libs/numeric/ublas/test1/bin/test1/intel-linu x/release/runtime-link-dynamic/test1) ==6555== by 0x804A1F8: (within /usr/local/lib/boost_dev/boost/libs/numeric/ublas/test1/bin/test1/intel-linu x/release/runtime-link-dynamic/test1) Note: see also the FAQ.txt in the source distribution. It contains workarounds to several common problems. If that doesn't help, please report this bug to: js...@ac... In the bug report, send all the above text, the valgrind version, and what Linux distro you are using. Thanks. ---------- Thanks for any advice, Joerg P.S.: this is *not* a regression, as far as I can tell. P.P.S.: Running the test under SuSE 7.3, but this should be irrelevant. P.P.P.S.: Many thanks for improving the SSE support! |
|
From: Nicholas N. <nj...@ca...> - 2003-10-18 13:53:57
|
On Sat, 18 Oct 2003, Joerg Walter wrote: > when running one of my regression tests built with ICC 7.1 under the newly > released valgrind I see the following assertion failure: > > Memcheck: mc_translate.c:1108 (memcheck_instrument): Assertion `u_in->size > == 4 || u_in->size == 16' failed. It's a bug, change the line to this: sk_assert(u_in->size == 4 || u_in->size == 8 || u_in->size == 16); I've committed the change. Thanks for the report. N |
Dear Nicholas, you wrote: > > when running one of my regression tests built with ICC 7.1 under the newly > > released valgrind I see the following assertion failure: > > > > Memcheck: mc_translate.c:1108 (memcheck_instrument): Assertion `u_in->size > > == 4 || u_in->size == 16' failed. > > It's a bug, change the line to this: > > sk_assert(u_in->size == 4 || u_in->size == 8 || u_in->size == 16); Done. This fixed the original problem. Now it seems I'm again stuck with a missing instruction: ---------- ==1989== Memcheck, a.k.a. Valgrind, a memory error detector for x86-linux. ==1989== Copyright (C) 2002-2003, and GNU GPL'd, by Julian Seward. ==1989== Using valgrind-20031012, a program supervision framework for x86-linux. ==1989== Copyright (C) 2000-2003, and GNU GPL'd, by Julian Seward. ==1989== Estimated CPU clock rate is 1716 MHz ==1989== For more details, rerun with: -v ==1989== test_vector double, unbounded_array v1 = v2 = [3](1,2,3) v1.assign_temporary (v2) = [3](1,2,3) v1.swap (v2) = [3](1,2,3) [3](1,2,3) - v1 = [3](-1,-2,-3) conj (v1) = [3](1,2,3) v1 + v2 = [3](2,4,6) v1 - v2 = [3](0,0,0) element_prod (v1, v2) = [3](1,4,9) 1. * v1 = [3](1,2,3) N * v1 = [3](3,6,9) v1 * 1. = [3](1,2,3) v1 * N = [3](3,6,9) v2 += v1 = [3](2,4,6) v2 -= v1 = [3](1,2,3) v1 *= 1. = [3](1,2,3) v1 *= N = [3](3,6,9) sum (v1) = 6 norm_1 (v1) = 6 disInstr: unhandled instruction bytes: 0x66 0xF 0x2E 0x5 ==1989== Use of uninitialised value of size 4 ==1989== at 0x8050AB0: (within /usr/local/lib/boost_dev/boost/libs/numeric/ublas/test1/bin/test1/intel-linu x/release/runtime-link-dynamic/test1) ==1989== ==1989== Jump to the invalid address stated on the next line ==1989== at 0x0: ??? ==1989== Address 0x0 is not stack'd, malloc'd or free'd ---------- Thanks very much for your help, Joerg |
In message <002001c39581$91f9e8a0$010...@ms...>
jhr...@t-... (Joerg Walter) wrote:
> Done. This fixed the original problem. Now it seems I'm again stuck with a
> missing instruction:
[ snipped ]
> disInstr: unhandled instruction bytes: 0x66 0xF 0x2E 0x5
That's UCOMISD which looks like it is implemtented in the CVS head
version of valgrind.
Tom
--
Tom Hughes (th...@cy...)
Software Engineer, Cyberscience Corporation
http://www.cyberscience.com/
|
|
From: Dirk M. <dm...@gm...> - 2003-10-18 15:41:09
|
On Saturday 18 October 2003 13:28, Joerg Walter wrote: > when running one of my regression tests built with ICC 7.1 under the newly > released valgrind I see the following assertion failure: can you try CVS HEAD? this was recently fixed iirc. |
|
From: <jhr...@t-...> - 2003-10-19 13:38:38
|
----- Original Message ----- From: "Dirk Mueller" <dm...@gm...> To: <val...@li...> Sent: Saturday, October 18, 2003 3:17 PM Subject: Re: [Valgrind-users] Memcheck: mc_translate.c:1108 (memcheck_instrument): Assertion `u_in->size == 4 || u_in->size == 16' failed. > On Saturday 18 October 2003 13:28, Joerg Walter wrote: > > > when running one of my regression tests built with ICC 7.1 under the newly > > released valgrind I see the following assertion failure: > > can you try CVS HEAD? this was recently fixed iirc. I tried CVS HEAD and yes, the original problem is fixed on the trunk already. But both the fixed latest release (see Nicholas' hint) and trunk stumble across disInstr: unhandled instruction bytes: 0x66 0xF 0x2E 0x5 now. Thanks, Joerg |
|
From: Dirk M. <dm...@gm...> - 2003-10-18 17:05:30
Attachments:
ucode.diff
|
On Saturday 18 October 2003 18:20, Joerg Walter wrote: > already. But both the fixed latest release (see Nicholas' hint) and trunk > stumble across > > disInstr: unhandled instruction bytes: 0x66 0xF 0x2E 0x5 oh, thats simple. see patch. |
Dear Dirk, you wrote: > > But both the fixed latest release (see Nicholas' hint) and trunk > > stumble across > > > > disInstr: unhandled instruction bytes: 0x66 0xF 0x2E 0x5 > > oh, thats simple. If you say it ;-) > see patch. Tested against trunk and 20031012: works fine for me. Thank you very much, Joerg |