Re: [9fans] too good to pass up (SRB Comments)
- From: "Douglas A. Gwyn" <DAGwyn@xxxxxxxx>
- Date: Wed, 3 May 2006 08:40:46 GMT
Richard Miller wrote:
I took the easy way out and rewrote the allocator to make sure it
extended the address space before the memory was used, rather than
relying on recoverable address faults.
Reminds me of a change I made to the PDP-11 C code generator
to test the new top-of-stack after setting up the stack
pointer in function prologues. The idea is that if a trap
was needed for the kernel to extend the process's allocation
of RAM, it was much better to do it in a controlled manner
than to have it happen in the middle of some complex machine
instruction (I think floating-point was the big problem, due
to the FPU running asynchronously to the integer ALU). You
may recall that a lot of code was used in some PDP-11 Unix
kernels to figure out how to correctly restart faulted
instructions; the lesson to be learned is not to get into
that situation.
.
- Prev by Date: Re: [9fans] too good to pass up (SRB Comments)
- Next by Date: Re: [9fans] USB trouble
- Previous by thread: Re: [9fans] too good to pass up (SRB Comments)
- Next by thread: Re: [9fans] ipconfig and ppp
- Index(es):
Relevant Pages
|