lattice / quda

QUDA is a library for performing calculations in lattice QCD on GPUs.
https://lattice.github.io/quda
Other
292 stars 99 forks source link

add resIncreaseTotal counter to solvers #159

Closed mathiaswagner closed 10 years ago

mathiaswagner commented 10 years ago

Currently resIncrease only counts subsequent increases of the L2 norm. That does not catch alternating increase / decrease scenarios.

mathiaswagner commented 10 years ago

Need to think about how we want to handle that for the multi-shift as the reliable_shift that triggers the increase is not constant.

maddyscientist commented 10 years ago

Have an array to store the number of increases for each shift?


This email message is for the sole use of the intended recipient(s) and may contain confidential information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by

reply email and destroy all copies of the original message.

mathiaswagner commented 10 years ago

Yes, either that or reset resIncreaseTotal when reliable_shift changes. Just wanted to remind me whether this makes sense or there is a more 'wanted' way of stopping the solver. Probably the array is fine as the check is just there to stop the solver in case the resIncrease does not work and it is not worth spending a lot of time with this.

On 14.10.2014, at 12:14, mikeaclark notifications@github.com<mailto:notifications@github.com> wrote:

Have an array to store the number of increases for each shift?


This email message is for the sole use of the intended recipient(s) and may contain confidential information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by

reply email and destroy all copies of the original message.

— Reply to this email directly or view it on GitHubhttps://github.com/lattice/quda/issues/159#issuecomment-59072143.


Mathias Wagner Department of Physics SW 117 - Indiana University Bloomington, IN 47405 email: mathwagn@indiana.edumailto:mathwagn@indiana.edu

maddyscientist commented 10 years ago

Exactly, just get it done and move on.

On Oct 14, 2014, at 9:21, "Mathias Wagner" notifications@github.com<mailto:notifications@github.com> wrote:

Yes, either that or reset resIncreaseTotal when reliable_shift changes. Just wanted to remind me whether this makes sense or there is a more 'wanted' way of stopping the solver. Probably the array is fine as the check is just there to stop the solver in case the resIncrease does not work and it is not worth spending a lot of time with this.

On 14.10.2014, at 12:14, mikeaclark notifications@github.com<mailto:notifications@github.commailto:notifications@github.com> wrote:

Have an array to store the number of increases for each shift?


This email message is for the sole use of the intended recipient(s) and may contain confidential information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by

reply email and destroy all copies of the original message.

— Reply to this email directly or view it on GitHubhttps://github.com/lattice/quda/issues/159#issuecomment-59072143.


Mathias Wagner Department of Physics SW 117 - Indiana University Bloomington, IN 47405 email: mathwagn@indiana.edumailto:mathwagn@indiana.edumailto:mathwagn@indiana.edu

— Reply to this email directly or view it on GitHubhttps://github.com/lattice/quda/issues/159#issuecomment-59073252.

mathiaswagner commented 10 years ago

preconCG still uses a hard coded maxResIncrease=0. Is that intended?

jpfoley commented 10 years ago

In principle, that can be changed. preconCG isn't used in production code that that was for testing purposes.

On 10/14/2014 01:00 PM, Mathias Wagner wrote:

preconCG still uses a hard coded |maxResIncrease=0|. Is that intended?

— Reply to this email directly or view it on GitHub https://github.com/lattice/quda/issues/159#issuecomment-59079411.

maddyscientist commented 10 years ago

If a variable is present in the interface, as a general rule of thumb, we should never hard code it…


This email message is for the sole use of the intended recipient(s) and may contain confidential information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by

reply email and destroy all copies of the original message.

mathiaswagner commented 10 years ago

Added in b51689e654e4ff75c517653ef198aee764fb365c