using anonymous mmap with fixed address on ppc

[ruben]

Using an anonymous mmap on ppc to allocate data on certain fixed memory addresses seems to only work for memory addresses in the range 0x3000000 - 0x7FFFFFFF.
On x86 on the other hand i can use a much wider address range, far beyond 0x80000000.
Does anyone know the reason for this ?

[gnufreax]

Let me see AMD K7 and G4?
On RISC Architectures like PPC, SPARC,MIPS you never will see such high memory adress numbers in hex than on CISC Processors and if you still like other´s thing an P4 or K7 or even K8 is an RISC based or RISC like processor that`s an old MYTH because they are only CISC Processors wich emulate or even use RISC Ops hardware encoding but the truth is THEY ARE CISC BASED! NOT RISC or MIPS or even VLIW like an Transmeta Efficeon perhaps just to understand the REAL DIFFERENCE BETWEEN RISC and CISC just read this qote from ArsTechnica´s great text RISC Vs. CISC...

[ruben]

It's about a G3 (750FX) and a Pentium 4. I do know the basic difference between CISC and RISC. And maybe i'm wrong, but AFAIK both a G3 and a Pentium4 use 32bit memory addressing, which essentially gives you a range of 4Gigabyte (0x00000000 - 0xFFFFFFFF). Each process on a linux system gets (in a way) its own 4Gigabyte range. I'm not an expert on this, but some of that space is reserved by the kernel and can't be used for heap space. I've read before that with adjusting some parameters in the kernel you can give the process about 3.5Giga space that it can use for its own (heap,...). Now i just wonder why it seems that on the G3 the space that a process gets for its own is a lot less than on an x86... or why i can't use more of the process space with mmap.