Re: Do not trust ALIGN

All above discussion simply shows that RISC compilers (and probably
processors) are not mature enough to challenge x86 when it comes to produce
simple and accurate code in 'tricky' situations.

It's all to do with silicon area and power consumption; the logic
necessary to handle misaligned accesses automatically in hardware takes
up extra silicon area, consumes power, and is not necessary unless you
want to retain backwards compatibility, such as in the x86 family.

RISC designs not only have reduced instruction sets, but try and
eliminate unnecessary hardware - there's no point in carrying the power
consumption for a bit of hardware that's rarely used and can be
replaced by the compiler ensuring aligned accesses.

The discussion is really about how the compiler is generating code in
the odd 'tricky' situation that it has been asked to handle; it can do
it efficiently but you need to know what you're doing to make it happen
- embedded programming is all about knowing what you're doing and
Usenet is a good place to pass on that information.

I am sorry, however, that the discussion has hijacked the original
thread on assembler coding and in the future will post any "similarly"
points to a new thread.

Andrew.

.

Relevant Pages

  • Re: Better device probe values
    ... > actually somewhat tricky, on x86 for PCI-PCI you have ACPI, PCIBIOS, MPTable, ... Well, on x86 we have 3 different pci busses, but they aren't in any ...
    (freebsd-arch)
  • Re: Better device probe values
    ... > actually somewhat tricky, on x86 for PCI-PCI you have ACPI, PCIBIOS, MPTable, ... Well, on x86 we have 3 different pci busses, but they aren't in any ...
    (freebsd-arch)