v1.8 bug ? memory allocation crash in map calculation (phenix.refine, phenix.maps)

Phil Jeffrey

10 Jul 2012 10 Jul '12

9:46 p.m.

This is in F432 in a=b=c=340A at 2.85A resolution. Mac OSX 10.5.8, the machine has 8Gb memory of which something like 6Gb is hanging around begging to be used. 1,327,104,000 if that's in bytes doesn't seem all that excessive. On the other hand, 340*340*340= 39,419,600 so it's perhaps a little greedy. Limits set by shell: phil% limit cputime unlimited filesize unlimited datasize unlimited stacksize 65532 kbytes coredumpsize 0 kbytes memoryuse unlimited descriptors 256 memorylocked unlimited maxproc 266 phenix.refine does the same thing right at the end, which is kind of a shame since it takes quite a long time to run. phenix.maps does the same thing even if I ask it just to generate F,phase in MTZ format. [snip] Compute maps. Python(632) malloc: *** mmap(size=1327104000) failed (error code=12) *** error: can't allocate region *** set a breakpoint in malloc_error_break to debug show_stack(1): /usr/local/phenix-1.8-1069/cctbx_project/cctbx/miller/__init__.py(3067) double_step_filtration show_stack(2): /usr/local/phenix-1.8-1069/cctbx_project/mmtbx/maps/__init__.py(416) map_coefficients_from_fmodel show_stack(3): /usr/local/phenix-1.8-1069/cctbx_project/mmtbx/maps/__init__.py(543) __init__ show_stack(4): /usr/local/phenix-1.8-1069/build/mac-intel-osx/../../cctbx_project/mmtbx/command_line/maps.py(325) run show_stack(5): /usr/local/phenix-1.8-1069/build/mac-intel-osx/../../cctbx_project/mmtbx/command_line/maps.py(402) <module> Bus error (Python call stack above) This crash may be due to a problem in any imported Python module, including modules which are not part of the cctbx project. To disable the traps leading to this message, define these environment variables (e.g. assign the value 1): BOOST_ADAPTBX_FPE_DEFAULT BOOST_ADAPTBX_SIGNALS_DEFAULT This will NOT solve the problem, just mask it, but may allow you to proceed in case it is not critical.

Show replies by date

Nathaniel Echols

10 Jul 10 Jul

9:49 p.m.

Try turning off the filling of missing F-obs (fill_missing_f_obs=False). On Tue, Jul 10, 2012 at 2:46 PM, Phil Jeffrey wrote:

...

This is in F432 in a=b=c=340A at 2.85A resolution. Mac OSX 10.5.8, the machine has 8Gb memory of which something like 6Gb is hanging around begging to be used.

1,327,104,000 if that's in bytes doesn't seem all that excessive. On the other hand, 340*340*340= 39,419,600 so it's perhaps a little greedy.

Limits set by shell: phil% limit cputime unlimited filesize unlimited datasize unlimited stacksize 65532 kbytes coredumpsize 0 kbytes memoryuse unlimited descriptors 256 memorylocked unlimited maxproc 266

phenix.refine does the same thing right at the end, which is kind of a shame since it takes quite a long time to run. phenix.maps does the same thing even if I ask it just to generate F,phase in MTZ format.

[snip] Compute maps. Python(632) malloc: *** mmap(size=1327104000) failed (error code=12) *** error: can't allocate region *** set a breakpoint in malloc_error_break to debug show_stack(1): /usr/local/phenix-1.8-1069/cctbx_project/cctbx/miller/__init__.py(3067) double_step_filtration show_stack(2): /usr/local/phenix-1.8-1069/cctbx_project/mmtbx/maps/__init__.py(416) map_coefficients_from_fmodel show_stack(3): /usr/local/phenix-1.8-1069/cctbx_project/mmtbx/maps/__init__.py(543) __init__ show_stack(4): /usr/local/phenix-1.8-1069/build/mac-intel-osx/../../cctbx_project/mmtbx/command_line/maps.py(325) run show_stack(5): /usr/local/phenix-1.8-1069/build/mac-intel-osx/../../cctbx_project/mmtbx/command_line/maps.py(402) <module> Bus error (Python call stack above) This crash may be due to a problem in any imported Python module, including modules which are not part of the cctbx project. To disable the traps leading to this message, define these environment variables (e.g. assign the value 1): BOOST_ADAPTBX_FPE_DEFAULT BOOST_ADAPTBX_SIGNALS_DEFAULT This will NOT solve the problem, just mask it, but may allow you to proceed in case it is not critical.

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

Nathaniel Echols

10:22 p.m.

On Tue, Jul 10, 2012 at 2:46 PM, Phil Jeffrey wrote:

...

This is in F432 in a=b=c=340A at 2.85A resolution. Mac OSX 10.5.8, the machine has 8Gb memory of which something like 6Gb is hanging around begging to be used.

1,327,104,000 if that's in bytes doesn't seem all that excessive. On the other hand, 340*340*340= 39,419,600 so it's perhaps a little greedy.

Forgot to address this - in fact the formula for map size in bytes is something vaguely like this (I am totally guessing here): map_size = 8 * a * b * c / (d_min/resolution_factor)^3 resolution_factor is 0.25 in this context. The multiplier of 8 is because we're using double-precision floating point numbers, which are 64 bits each. This comes out to 868,250,664 bytes for your structure, but I'm probably missing some essential details, and I think the grid size will be adjusted to behave well in the FFT. It's hard to guess exactly why this crashes in your case; it doesn't seem like something that should run out of memory, but this particular routine may be a bit of a bottleneck. -Nat

Phil Jeffrey

11 Jul 11 Jul

10:24 p.m.

New subject: v1.8 bug ? memory allocation crash - apparently also in AutoBuild

I think phenix is not particularly fond of F432, because in AutoBuild I get: Python(6992) malloc: *** mmap(size=283115520) failed (error code=12) *** error: can't allocate region *** set a breakpoint in malloc_error_break to debug While a larger unit cell, the asymmetric unit is not huge, nor is the # of reflections. Phil Jeffrey Princeton

Nathaniel Echols

10:30 p.m.

New subject: v1.8 bug ? memory allocation crash - apparently also in AutoBuild

On Wed, Jul 11, 2012 at 3:24 PM, Phil Jeffrey wrote:

...

I think phenix is not particularly fond of F432, because in AutoBuild I get:

Python(6992) malloc: *** mmap(size=283115520) failed (error code=12) *** error: can't allocate region *** set a breakpoint in malloc_error_break to debug

While a larger unit cell, the asymmetric unit is not huge, nor is the # of reflections.

You might be able to circumvent this crash by setting this parameter: filled_2fofc_maps=False (which I hope does the same thing as my previous advice.) -Nat

Phil Jeffrey

11:05 p.m.

New subject: v1.8 bug ? memory allocation crash - apparently also in AutoBuild

I find that parameter under the model building GUI and verify its setting under View All Settings but it makes no difference. Don't see any other parameters involving "fill" set to True. Potentially informative last part of log: Possibly cresolve size 96 was too big for this machine ---------------------------------- TRIED SIZE OF 48 ...but not OK Last log file in AutoBuild_run_8_/LAST.LOG Last 20 lines of logfile: ---------------------------------- *************ERROR ENDING ******************* ******************************************************************************** Failed to carry out AutoBuild_multiple_models: COMMANDS USED: workdir AutoBuild_run_8_/TEMP0 outputdir AutoBuild_run_8_ freer_if_present hklin aniso_data_PHX.mtz labin FP=F120 SIGFP=SIGF120 hklout exptl_phases_and_amplitudes_for_density_modification_aniso.mtz solvent_content 0.52 no_build mask_cycles 1 minor_cycles 0 ha_file NONE keep_missing use_wang no_ha database 5 no_optimize_ncs spg_name_use F 4 3 2 min_z_value_rho -3.0 start_chain 1 8 start_chain 2 10 start_chain 3 105 start_chain 4 105 no_create_free use_all_for_test delta_phi 20.00 group_ca_length 1 group_length 2 n_random_frag 0 See overall log file and also resolve log file (in /Users/phil/Structures/Cog5-Cog7/native/from_Jun/phenix-tests/AutoBuild_run_8_/LAST.LOG) Note: It is possible that the required size of resolve needs more memory than is available ******************************************************************************** *************ERROR ENDING ******************* On 7/11/12 6:30 PM, Nathaniel Echols wrote:

...

On Wed, Jul 11, 2012 at 3:24 PM, Phil Jeffrey mailto:[email protected]> wrote:

I think phenix is not particularly fond of F432, because in AutoBuild I get:

Python(6992) malloc: *** mmap(size=283115520) failed (error code=12) *** error: can't allocate region *** set a breakpoint in malloc_error_break to debug

While a larger unit cell, the asymmetric unit is not huge, nor is the # of reflections.

You might be able to circumvent this crash by setting this parameter:

filled_2fofc_maps=False

(which I hope does the same thing as my previous advice.)

-Nat

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

Terwilliger, Thomas C

11:17 p.m.

New subject: v1.8 bug ? memory allocation crash - apparently also in AutoBuild

Hi Phil, Usually that error message means literally that the required resolve size is too big. In that space group it could be so even if the cell is not huge. Sometimes there is some other error however. You can probably tell by cutting the resolution way back and then if there is an error it is coming from something else (In that case please let me know!) All the best, Tom T ________________________________________ From: [email protected] [[email protected]] on behalf of Phil Jeffrey [[email protected]] Sent: Wednesday, July 11, 2012 5:05 PM To: PHENIX user mailing list Subject: Re: [phenixbb] v1.8 bug ? memory allocation crash - apparently also in AutoBuild I find that parameter under the model building GUI and verify its setting under View All Settings but it makes no difference. Don't see any other parameters involving "fill" set to True. Potentially informative last part of log: Possibly cresolve size 96 was too big for this machine ---------------------------------- TRIED SIZE OF 48 ...but not OK Last log file in AutoBuild_run_8_/LAST.LOG Last 20 lines of logfile: ---------------------------------- *************ERROR ENDING ******************* ******************************************************************************** Failed to carry out AutoBuild_multiple_models: COMMANDS USED: workdir AutoBuild_run_8_/TEMP0 outputdir AutoBuild_run_8_ freer_if_present hklin aniso_data_PHX.mtz labin FP=F120 SIGFP=SIGF120 hklout exptl_phases_and_amplitudes_for_density_modification_aniso.mtz solvent_content 0.52 no_build mask_cycles 1 minor_cycles 0 ha_file NONE keep_missing use_wang no_ha database 5 no_optimize_ncs spg_name_use F 4 3 2 min_z_value_rho -3.0 start_chain 1 8 start_chain 2 10 start_chain 3 105 start_chain 4 105 no_create_free use_all_for_test delta_phi 20.00 group_ca_length 1 group_length 2 n_random_frag 0 See overall log file and also resolve log file (in /Users/phil/Structures/Cog5-Cog7/native/from_Jun/phenix-tests/AutoBuild_run_8_/LAST.LOG) Note: It is possible that the required size of resolve needs more memory than is available ******************************************************************************** *************ERROR ENDING ******************* On 7/11/12 6:30 PM, Nathaniel Echols wrote:

...

On Wed, Jul 11, 2012 at 3:24 PM, Phil Jeffrey mailto:[email protected]> wrote:

I think phenix is not particularly fond of F432, because in AutoBuild I get:

Python(6992) malloc: *** mmap(size=283115520) failed (error code=12) *** error: can't allocate region *** set a breakpoint in malloc_error_break to debug

While a larger unit cell, the asymmetric unit is not huge, nor is the # of reflections.

You might be able to circumvent this crash by setting this parameter:

filled_2fofc_maps=False

(which I hope does the same thing as my previous advice.)

-Nat

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

Phil Jeffrey

12 Jul 12 Jul

4:15 p.m.

New subject: v1.8 bug ? memory allocation crash - apparently also in AutoBuild

Hi Tom, Even setting the high resolution limit in the Other Options form to 6.0, and with filled_2fofc_maps=false, it dies in the same place and interestingly with the same mmap memory allocation request - this seems invariant with the resolution parameter: Python(12184) malloc: *** mmap(size=283115520) failed (error code=12) *** error: can't allocate region *** set a breakpoint in malloc_error_break to debug The memory on the machine is not filling up while the process is running. Phil On 7/11/12 7:17 PM, Terwilliger, Thomas C wrote:

...

Hi Phil,

Usually that error message means literally that the required resolve size is too big. In that space group it could be so even if the cell is not huge. Sometimes there is some other error however. You can probably tell by cutting the resolution way back and then if there is an error it is coming from something else (In that case please let me know!)

All the best, Tom T

________________________________________ From: [email protected] [[email protected]] on behalf of Phil Jeffrey [[email protected]] Sent: Wednesday, July 11, 2012 5:05 PM To: PHENIX user mailing list Subject: Re: [phenixbb] v1.8 bug ? memory allocation crash - apparently also in AutoBuild

I find that parameter under the model building GUI and verify its setting under View All Settings but it makes no difference. Don't see any other parameters involving "fill" set to True.

Potentially informative last part of log:

Possibly cresolve size 96 was too big for this machine ---------------------------------- TRIED SIZE OF 48 ...but not OK Last log file in AutoBuild_run_8_/LAST.LOG Last 20 lines of logfile: ---------------------------------- *************ERROR ENDING *******************

******************************************************************************** Failed to carry out AutoBuild_multiple_models:

COMMANDS USED: workdir AutoBuild_run_8_/TEMP0 outputdir AutoBuild_run_8_ freer_if_present hklin aniso_data_PHX.mtz labin FP=F120 SIGFP=SIGF120 hklout exptl_phases_and_amplitudes_for_density_modification_aniso.mtz solvent_content 0.52 no_build mask_cycles 1 minor_cycles 0 ha_file NONE keep_missing use_wang no_ha database 5 no_optimize_ncs spg_name_use F 4 3 2 min_z_value_rho -3.0 start_chain 1 8 start_chain 2 10 start_chain 3 105 start_chain 4 105 no_create_free use_all_for_test delta_phi 20.00 group_ca_length 1 group_length 2 n_random_frag 0

See overall log file and also resolve log file (in /Users/phil/Structures/Cog5-Cog7/native/from_Jun/phenix-tests/AutoBuild_run_8_/LAST.LOG)

Note: It is possible that the required size of resolve needs more memory than is available ********************************************************************************

*************ERROR ENDING *******************

On 7/11/12 6:30 PM, Nathaniel Echols wrote:

...
On Wed, Jul 11, 2012 at 3:24 PM, Phil Jeffreymailto:[email protected]> wrote:

I think phenix is not particularly fond of F432, because in AutoBuild I get:

Python(6992) malloc: *** mmap(size=283115520) failed (error code=12) *** error: can't allocate region *** set a breakpoint in malloc_error_break to debug

While a larger unit cell, the asymmetric unit is not huge, nor is the # of reflections.

You might be able to circumvent this crash by setting this parameter:

filled_2fofc_maps=False

(which I hope does the same thing as my previous advice.)

-Nat

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

Terwilliger, Thomas C

4:17 p.m.

New subject: v1.8 bug ? memory allocation crash - apparently also in AutoBuild

Hi Phil, Hmm...sounds bad! In this case might it be possible to send me the required inputs off-list so that I can track down the problem? Thanks very much! -Tom T ________________________________________ From: Phil Jeffrey [[email protected]] Sent: Thursday, July 12, 2012 10:15 AM To: PHENIX user mailing list Cc: Terwilliger, Thomas C Subject: Re: [phenixbb] v1.8 bug ? memory allocation crash - apparently also in AutoBuild Hi Tom, Even setting the high resolution limit in the Other Options form to 6.0, and with filled_2fofc_maps=false, it dies in the same place and interestingly with the same mmap memory allocation request - this seems invariant with the resolution parameter: Python(12184) malloc: *** mmap(size=283115520) failed (error code=12) *** error: can't allocate region *** set a breakpoint in malloc_error_break to debug The memory on the machine is not filling up while the process is running. Phil On 7/11/12 7:17 PM, Terwilliger, Thomas C wrote:

...

Hi Phil,

Usually that error message means literally that the required resolve size is too big. In that space group it could be so even if the cell is not huge. Sometimes there is some other error however. You can probably tell by cutting the resolution way back and then if there is an error it is coming from something else (In that case please let me know!)

All the best, Tom T

________________________________________ From: [email protected] [[email protected]] on behalf of Phil Jeffrey [[email protected]] Sent: Wednesday, July 11, 2012 5:05 PM To: PHENIX user mailing list Subject: Re: [phenixbb] v1.8 bug ? memory allocation crash - apparently also in AutoBuild

I find that parameter under the model building GUI and verify its setting under View All Settings but it makes no difference. Don't see any other parameters involving "fill" set to True.

Potentially informative last part of log:

Possibly cresolve size 96 was too big for this machine ---------------------------------- TRIED SIZE OF 48 ...but not OK Last log file in AutoBuild_run_8_/LAST.LOG Last 20 lines of logfile: ---------------------------------- *************ERROR ENDING *******************

******************************************************************************** Failed to carry out AutoBuild_multiple_models:

COMMANDS USED: workdir AutoBuild_run_8_/TEMP0 outputdir AutoBuild_run_8_ freer_if_present hklin aniso_data_PHX.mtz labin FP=F120 SIGFP=SIGF120 hklout exptl_phases_and_amplitudes_for_density_modification_aniso.mtz solvent_content 0.52 no_build mask_cycles 1 minor_cycles 0 ha_file NONE keep_missing use_wang no_ha database 5 no_optimize_ncs spg_name_use F 4 3 2 min_z_value_rho -3.0 start_chain 1 8 start_chain 2 10 start_chain 3 105 start_chain 4 105 no_create_free use_all_for_test delta_phi 20.00 group_ca_length 1 group_length 2 n_random_frag 0

See overall log file and also resolve log file (in /Users/phil/Structures/Cog5-Cog7/native/from_Jun/phenix-tests/AutoBuild_run_8_/LAST.LOG)

Note: It is possible that the required size of resolve needs more memory than is available ********************************************************************************

*************ERROR ENDING *******************

On 7/11/12 6:30 PM, Nathaniel Echols wrote:

...
On Wed, Jul 11, 2012 at 3:24 PM, Phil Jeffreymailto:[email protected]> wrote:

I think phenix is not particularly fond of F432, because in AutoBuild I get:

Python(6992) malloc: *** mmap(size=283115520) failed (error code=12) *** error: can't allocate region *** set a breakpoint in malloc_error_break to debug

While a larger unit cell, the asymmetric unit is not huge, nor is the # of reflections.

You might be able to circumvent this crash by setting this parameter:

filled_2fofc_maps=False

(which I hope does the same thing as my previous advice.)

-Nat

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

4551

Age (days ago)

4553

Last active (days ago)

List overview

Download

8 comments

3 participants

participants (3)

Nathaniel Echols
Phil Jeffrey
Terwilliger, Thomas C