with the future coming of CPU's having mult core's (8,16,500) on a chip - is there any plans/thought of having a way to select/control how many cores a project runs on, when available before running other project's if the first one is not available?

with the future coming of CPU's having mult core's (8,16,500) on a chip - is there any plans/thought of having a way to select/control how many cores a project runs on, when available before running other project's if the first one is not available?

Do you mean as in this thread?

Or are you thinking about running one work unit on several cores at the same time?
(Which is still in the discussion stage here.)

with the future coming of CPU's having mult core's (8,16,500) on a chip - is there any plans/thought of having a way to select/control how many cores a project runs on, when available before running other project's if the first one is not available?

Do you mean as in this thread?

Or are you thinking about running one work unit on several cores at the same time?

I suppose he means limiting any specific project (with single-threaded tasks) to an independent number of cores, like: CPDN=1 (do not take it personally, Les ;-), Rosetta=2, Einstein=2, Seti=3, WCG=2, Superlink=1, uFluids=0.2, Hydrogen=1, etc.

A long lasting wish to be able to control the number of memory hogs launched in parallel or how many tasks some (often) mis-behaving app can launch concurrently. In addition to the Resource Share, which control the tasks distribution in other way.

Peter

with the future coming of CPU's having mult core's (8,16,500) on a chip - is there any plans/thought of having a way to select/control how many cores a project runs on, when available before running other project's if the first one is not available?

Do you mean as in this thread?

Or are you thinking about running one work unit on several cores at the same time?

I suppose he means limiting any specific project (with single-threaded tasks) to an independent number of cores, like: CPDN=1 (do not take it personally, Les ;-), Rosetta=2, Einstein=2, Seti=3, WCG=2, Superlink=1, uFluids=0.2, Hydrogen=1, etc.

A long lasting wish to be able to control the number of memory hogs launched in parallel or how many tasks some (often) mis-behaving app can launch concurrently. In addition to the Resource Share, which control the tasks distribution in other way.

Peter

my thinking was p1 runs on core 1 thru 10, p2 on 11 thru 20, etc w/ mult. copies of the application's going at the same time and if that p1 was down or not available that group of cores would run other projects until p1 is back up. was not thinking of one application running across mult cores.

my thinking was p1 runs on core 1 thru 10, p2 on 11 thru 20, etc w/ mult. copies of the application's going at the same time and if that p1 was down or not available that group of cores would run other projects until p1 is back up.

So, this could be solved by setting independent processor affinity mask for each project and using it when lauching projects' applications?

Peter

my thinking was p1 runs on core 1 thru 10, p2 on 11 thru 20, etc w/ mult. copies of the application's going at the same time and if that p1 was down or not available that group of cores would run other projects until p1 is back up.

So, this could be solved by setting independent processor affinity mask for each project and using it when lauching projects' applications?

No, tried it with Process Lasso 2 months ago. For a Quad, it ended up running 4 BOINC jobs of same make running on 1 core.

Off course. The client would have to set the affinity masks upon launching an app and simultaneously not spawn more apps from any project than the number of bits set in project's affinity mask. (And we are back to the same problem - to do it effectively, again the client would have to limit the number of any project's apps launched. The affinities is then the secondary thing, solvable by either the client or any external aplication.)

Peter