Trac June 2016

trac@lists.einsteintoolkit.org

1 participants
119 discussions

[Einstein Toolkit] #1865: Automatically start SystemTopology
by Einstein Toolkit 03 Oct '16

03 Oct '16

#1865: Automatically start SystemTopology ---------------------------------+------------------------------------------ Reporter: dradice@… | Owner: eschnett Type: enhancement | Status: new Priority: major | Milestone: Component: Carpet | Version: development version Keywords: | ---------------------------------+------------------------------------------ Carpet used to load hwloc automatically and that would set thread affinities. Now this functionality is in the SystemTopology thorn, which is not automatically activated. This change could result in a significant performance regression on some systems (see discussion in #1850). Would it make sense to activate SystemTopology automatically? -- Ticket URL: <https://trac.einsteintoolkit.org/ticket/1865> Einstein Toolkit <http://einsteintoolkit.org> The Einstein Toolkit

1 14

[Einstein Toolkit] #1932: ML_BSSN: other_timelevels Parameter Not Respected
by Einstein Toolkit 03 Oct '16

03 Oct '16

#1932: ML_BSSN: other_timelevels Parameter Not Respected --------------------------------+------------------------------------------- Reporter: zachetie@… | Owner: Type: defect | Status: new Priority: major | Milestone: Component: Other | Version: development version Keywords: | --------------------------------+------------------------------------------- At the Jun 27, 2016 ET telecon, we found the following bug in McLachlan/ML_BSSN: Inside the ET_2016_05 ML_BSSN/schedule.ccl, you'll notice the following lines: STORAGE: ML_Ham[timelevels] STORAGE: ML_mom[timelevels] STORAGE: ML_cons_detg[timelevels] STORAGE: ML_cons_Gamma[timelevels] STORAGE: ML_cons_traceA[timelevels] in all of these lines, "timelevels" should be replaced by "other_timelevels". This should result in significantly increased memory usage in the latest ML_BSSN (possibly at the 10-20% level), particularly in vacuum evolutions. Related to this problem, I noticed that in a previous version of McLachlan (2015_05, where the above issue does not exist), all constraints are being stored in checkpoint files, despite having only one timelevel. ML_BSSN_Helper is supposed to overwrite the ML_BSSN/interface.ccl request to set the Checkpoint="no" tag. -- Ticket URL: <https://trac.einsteintoolkit.org/ticket/1932> Einstein Toolkit <http://einsteintoolkit.org> The Einstein Toolkit

1 10

[Einstein Toolkit] #1873: Parameter file parser cannot handle empty files
by Einstein Toolkit 25 Sep '16

25 Sep '16

#1873: Parameter file parser cannot handle empty files ----------------------+----------------------------------------------------- Reporter: eschnett | Owner: Type: defect | Status: new Priority: minor | Milestone: Component: Cactus | Version: development version Keywords: | ----------------------+----------------------------------------------------- I used a parameter file that was empty except for comments: {{{ # ActiveThorns = "PUGH" # ActiveThorns = "MultiThreading" # ActiveThorns = "FunHPC" }}} Cactus complais about this: {{{ Activating thorn Cactus...Success -> active implementation Cactus WARNING level 0 from host process 0 while executing schedule bin (none), routine (no thorn)::(no routine) in thorn cactus, file multithreading.par:1: -> ERROR IN PARAMETER FILE:In rule 'file::skipper' Line=3, Column=84 # ActiveThorns = "PUGH" # ActiveThorns = "MultiThreading" # ActiveThorns = "FunHPC" }}} -- Ticket URL: <https://trac.einsteintoolkit.org/ticket/1873> Einstein Toolkit <http://einsteintoolkit.org> The Einstein Toolkit

1 2

[Einstein Toolkit] #1816: segfaults on 64bit systems when build with c99
by Einstein Toolkit 17 Aug '16

17 Aug '16

#1816: segfaults on 64bit systems when build with c99 --------------------------------+------------------------------------------- Reporter: physik@… | Owner: Type: defect | Status: new Priority: unset | Milestone: Component: Cactus | Version: ET_2014_05 Keywords: | --------------------------------+------------------------------------------- I recently encountered a segfault in ScheduleInterface.c, more precisely in the function static int CCTKi_ScheduleCallFunction(void *function, t_attribute *attribute, t_sched_data *data) { /* find the timer for this function and this schedule bin */ t_timer *timer = attribute->timers; while (timer && strcmp(timer->schedule_bin, data->schedule_bin)) { timer = timer->next; } Running in the debugger revealed that timer->schedule_bin pointed to an invalid address. Curiously, it had the top 33 (33 not a typo) bits all set. Taking the lowest 32 bits gave a valid address which pointed to a reasonable string "CCTK_INITIAL". This suggests a 32/64 bit issue. The pointer timer->schedule_bin seems to be initialized in the same function using strdup: timer->schedule_bin = strdup (where); strdup is not part of the c99 standard, but only Posix. Compiling with gcc --std=c99 means it is not defined in <string.h>. This means the compiler treats the occurrence of strdup as an implicit function declaration, and assumes it returns int. Thus, it will do an implicit conversion of the result from int to char*. If the highest bit of the int was set, this resulted in a 64 bit pointer with all 32 high bits set (I checked with a small test code). When the address returned by the actual strdup code linked from glibc has the top 33 bits zero, the conversion yields the correct results. Therefore, the problem is hard to reproduce, it only occurred with a test case almost exhausting my workstations memory, but frustratingly not small tests. After this, I also found compiler warnings for ScheduleInterface.c of the type implicit declaration of function ‘strdup’ [-Wimplicit-function- declaration] and assignment makes pointer from integer without a cast [-Wint-conversion] Switching from --std=c99 to --std=gnu99 fixed the problem for now. However, this is a bug that might affect many users since the code compiles with --std=c99 and the compiler warnings are hidden within the thousands of other compiler warnings the ET code generates. Also, a quick grep revealed many occurrences of strdup, although some of them where redefined as Util_Strdup. The rest might lead to segfaults on 64 bit systems with std=c99. My findings concern the Wheeler release, I haven't had time to check the development version. -- Ticket URL: <https://trac.einsteintoolkit.org/ticket/1816> Einstein Toolkit <http://einsteintoolkit.org> The Einstein Toolkit

1 10

[Einstein Toolkit] #1926: remove CACHELINE_BYTES and CACHE_SIZE from flesh
by Einstein Toolkit 11 Jul '16

11 Jul '16

#1926: remove CACHELINE_BYTES and CACHE_SIZE from flesh -------------------------+-------------------------------------------------- Reporter: knarf | Owner: Type: enhancement | Status: new Priority: optional | Milestone: Component: Cactus | Version: development version Keywords: | -------------------------+-------------------------------------------------- Nothing in Cactus uses these variables, and in some cases they are best guesses only. Also, as of 2fb3e32532bbe70121559d582d96b3c300c95276 (Erik Schnetter, Thu Jan 15 12:57:04 2015 -0500), Cactus (the flesh) lost it's ability to overwrite these variables via parameters. Thus, this is a little clean-up within the flesh. Pull request: https://bitbucket.org/cactuscode/cactus/pull-requests/25 /remove-cacheline_bytes-and-cache_size-from/diff -- Ticket URL: <https://trac.einsteintoolkit.org/ticket/1926> Einstein Toolkit <http://einsteintoolkit.org> The Einstein Toolkit

1 3

[Einstein Toolkit] #1791: Allow aligning the interior of grid functions in looping macros
by Einstein Toolkit 28 Jun '16

28 Jun '16

#1791: Allow aligning the interior of grid functions in looping macros -------------------------+-------------------------------------------------- Reporter: eschnett | Owner: Type: enhancement | Status: new Priority: minor | Milestone: Component: Cactus | Version: development version Keywords: | -------------------------+-------------------------------------------------- Currently, when grid functions are aligned, Cactus expects their origin to be aligned. These changes update the looping macros to allow aligning the interior of grid functions instead. Whether and how grid functions are aligned is still determined by the driver -- this only makes it possible to still use the looping macros in this case. Implemented in <https://bitbucket.org/cactuscode/cactus/pull-request/16 /allow-aligning-the-interior-of-grid/diff> and <https://bitbucket.org/cactuscode/cactustest/pull-request/1/allow- aligning-the-interior-of-grid/diff>. -- Ticket URL: <https://trac.einsteintoolkit.org/ticket/1791> Einstein Toolkit <http://einsteintoolkit.org> The Einstein Toolkit

1 6

[Einstein Toolkit] #1913: Cactus shouldn't use 'u' within ARFLAGS
by Einstein Toolkit 28 Jun '16

28 Jun '16

#1913: Cactus shouldn't use 'u' within ARFLAGS --------------------+------------------------------------------------------- Reporter: knarf | Owner: Type: defect | Status: new Priority: major | Milestone: ET_2016_11 Component: Cactus | Version: development version Keywords: | --------------------+------------------------------------------------------- configure.in specifies 'u' (among other flags) in ARFLAGS, but shouldn't, as can lead to build failures and is in general not a good default. See #1902 for a discussion (where it was fixed using a workaround in PETSc). -- Ticket URL: <https://trac.einsteintoolkit.org/ticket/1913> Einstein Toolkit <http://einsteintoolkit.org> The Einstein Toolkit

1 1

[Einstein Toolkit] #1858: CarpetLib: "balanced" recomposition fix
by Einstein Toolkit 28 Jun '16

28 Jun '16

#1858: CarpetLib: "balanced" recomposition fix ----------------------+----------------------------------------------------- Reporter: knarf | Owner: eschnett Type: defect | Status: new Priority: optional | Milestone: Component: Carpet | Version: development version Keywords: | ----------------------+----------------------------------------------------- This pull request fixes some things that I believe are broken for this algorithm. I am not claiming that it now works correctly, but it works 'more correct' this way. I came across these issues while looking at some of the implemented recomposition algos, but will not use this particular one in the future. However, I think it might be useful for someone else, so here it is. https://bitbucket.org/eschnett/carpet/pull-requests/10/recompose-balance- partial-fix/diff -- Ticket URL: <https://trac.einsteintoolkit.org/ticket/1858> Einstein Toolkit <http://einsteintoolkit.org> The Einstein Toolkit

1 2

[Einstein Toolkit] #1657: ExternalThorns/pciutils ignores most option list options
by Einstein Toolkit 28 Jun '16

28 Jun '16

#1657: ExternalThorns/pciutils ignores most option list options -----------------------------------+---------------------------------------- Reporter: rhaas | Owner: Type: defect | Status: new Priority: major | Milestone: Component: EinsteinToolkit thorn | Version: development version Keywords: pciutils | -----------------------------------+---------------------------------------- It does not contain a configure script and its Makefile hard-codes the compiler and linker to be gcc. This is an issue if LDFLAGS (or CFLAGS possibly) contain {{{-openmp}}} like they do when using the intel compiler. Also using mkl could (should) be done with the {{{-mkl}}} switch to the intel compiler but gcc (as used in pciutils) will naturally not accept this. -- Ticket URL: <https://trac.einsteintoolkit.org/ticket/1657> Einstein Toolkit <http://einsteintoolkit.org> The Einstein Toolkit

1 4

[Einstein Toolkit] #1924: piraha contains code for expanding arbitrary variables
by Einstein Toolkit 27 Jun '16

27 Jun '16

#1924: piraha contains code for expanding arbitrary variables -------------------------+-------------------------------------------------- Reporter: rhaas | Owner: Type: enhancement | Status: new Priority: optional | Milestone: Component: Cactus | Version: development version Keywords: Piraha | -------------------------+-------------------------------------------------- line 369 of piraha's Call.cc contains code: {{{ nm_iter iter = variables.find(gr->group(0)->substring()); if(iter != variables.end()) { ret = iter->second; } }}} that would seem to allow arbitrary variable names to be expanded to values. However piraha does not let me assign values to these variables anymore as the only allowed assignments are "ActiveThorns = ..." and "thorn::par = ...". So either there is code to support no longer allowed operations or there is a lack of documentation on how to set variable values for later use by {{{$varname}}}. -- Ticket URL: <https://trac.einsteintoolkit.org/ticket/1924> Einstein Toolkit <http://einsteintoolkit.org> The Einstein Toolkit

1 4

← Newer
1
...
6
7
8
9
10
11
12
Older →

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

Trac June 2016