I have 2 comments/questions on your new 64-bit article:

1) Since the R10 and R11 registers are volatile across function calls, do you know why aren't they used to pass parameters as well?  If the compiler were using them for something else they'll have to be reloaded after calling the function anyway, so why not just use them to pass in parameters?  Is there some negative processor or cache effect I'm missing?

2) If you update or do another 64-bit article, you might want to consider also mentioning a couple of other things (or maybe not):
    - HANDLES are pointer sized
    - be careful of 0xFFFFFFFF - it may not always extend to -1
    - _WIN32 is defined for 64-bit builds (in other words, _WIN32 really means "this is not a Win16 build")
   
Thanks.
   
   

And how about that RIP register?  Does it point to where the process was executed, or where it was buried, or where your previous posting eulogized SoftICE?  Glad to see Intel has a sense of humour even if it's a bit tame  ^u^

Actually I'll have one or two serious comments on your article too, if I'll find time to read it carefully enough and write the questions.

Mr.Pietrek,

Why don't you update this blog more frequently?  I would like to read more what you write here!

I will read the article soon and be here back again with some comments.

I read your article with interest. Thanks.

It says that in the new 64-bit world, HANDLE is still 4 bytes.
(Search for "HANDLE are still 4 bytes" in the article.)

I believe that HANDLE is defined as "void *" and therefore
should be 64 bytes in Win64.