Muli Ben-Yehuda wrote: > You are confusing *virtual* memory and *physical* memory. PAE has > nothing to do with virtual memory and everything to do with physical > memory. > I don't think I am. The simple truth of the matter is that it is not possible to access physical memory directly (at least, not unless you are in 16 bit real mod, in which case your constraints are much much harsher than 32 bit, they are more like 20 bit). As such, since a single linear pointer can only be 32 bit, you cannot simultaneously access more than 4GB of memory consecutively.
Yes, there are plenty tricks you can do. You can resurrect segmented addresses. You can juggle the physical memory around the virtual addresses. You can (as Linux does), allocate the different physical addresses to different contexts. Either way, this is not as simple as merely accessing the full 64GB as if they were one contiguous memory (which is what you could do with 64bit platform). >> I don't know the mechanics of PAE. Did it resurrect the dreaded segment >> registers? Is it a part of the MMU mapping? >> > > See above, or your favorite search engine :-) > Which roughly translates to "if you wish". Modern operating systems have made us used to not touching the segment registers (which is a good thing), so we are used to only using what Intel provides us in the "offset" section of the address. You could, however, use the segments to point to different entries in the PTE, and thus access two 4GB chunks "simultaneously". > More accurate would be to say that more physical memory requires more > work to keep track off, but you can trade-off computation and storage > when performing that work. > Yes, I guess you could, theoretically, do some level of memory-CPU trade off here. I don't see that it will actually bring you down to O(1) memory usage, since you do need to keep track of more than what's used and what's free, but this really delves into serious nitpicking, and I should stop here. > ... and my point was that this is a Linux design and implementation > issue, rather than a universal truth (unless you restate it as I did > above in terms of computation vs. storage overhead). > Ok. So let's agree that Linux requires more memory overhead the more physical memory is available, which results in the above mentioned trade off (more physical -> less virtual), which is why the 4/4 split patch was originally written. I am sure the difference was crucial to the understanding of my explanation about the difference between PAE and highmem. Happy? > Cheers, > Muli > To save you on the urge to use a magnifying glass to look for more inaccuracies and things which are not 102.5% correct, I misspelt a few of the words in this email. Wouldn't want to have a good nitpicking thread stop.... Shachar ================================================================= To unsubscribe, send mail to [EMAIL PROTECTED] with the word "unsubscribe" in the message body, e.g., run the command echo unsubscribe | mail [EMAIL PROTECTED]
