>One question to consider is do we want to also > support something similar to Linux's > "/proc/cpuinfo"?
"something similar" might be up for discussion, but it's been said before that unlike Linux, /proc on Solaris is _not_ a dumping ground for every random piece of system info (worse yet, presented in text format, which may be human readable, but may not always be easy to parse correctly). However, other pseudo filesystems, separate from /proc, have appeared for various purposes. I think even Linux has started to create separate filesystems for some of these things, recognizing that their /proc was seriously overloaded with too many different bits of functionality. I imagine a number of things to consider before a detailed proposal could be made. Consider first the hierarchy of a NUMA system; add to that possibly multiple CPU chips local to one another, multiple cores per chip, multiple threads per core, sub-core thread groups, different breakdowns of components that could be in use simultaneously on a multi-thread core (affects scheduling for best throughput), various cache arrangements (I don't even know enough to describe this, but I'm sure there are a _lot_ of possible variations), software or hardware-supported (para)virtualization, etc. Not to mention differences in which instructions are actually implemented in hardware (POPC is in software on a lot of chips, I think), optional instructions or instruction sets, differences in instruction timing (if you want to go down to that level); and of course the obvious (and often over-valued) clock frequency (as modified by power management, etc). And who knows what else that I forgot... I don't know whether a filesystem-like presentation would be flexible enough to handle all that. Perhaps some well-designed data structures and functions to access them in different ways are what's really needed? But before going crazy, there's also the question not only what is there that _could_ be described now or in the forseeable future, but how much of that would it be _useful_ to describe? Aside from simply documenting hardware configuration, how much might some program running on a system need to know to make best use of the system, or alternatively to balance its needs with a desire to be a good citizen and not take unreasonable advantage? -- This messages posted from opensolaris.org _______________________________________________ observability-discuss mailing list [EMAIL PROTECTED] _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
