binutils-gdb/gas/NOTES

143 lines
5.8 KiB
Plaintext

-*- text -*-
PORTING:
Sorry, no description of the interfaces is written up yet. Look at existing
back ends and work from there.
New hosts: If your host system has a strange header file setup, create a
config/ho-foo.h file for it and include the appropriate header files or
definitions there. If your host has a broken compiler, or some broken macros
in header files, create a host-specific file and repair the damage there.
(See, for example, ho-rs6000.h. The "assert" macro on that system doesn't work
right, and a flag is set to rewrite an expression in tc-m68k.c that the native
compiler mis-compiles.)
New target formats: Look at the BFD_ASSEMBLER code. The a.out code might be a
fair example. There are no "good" examples yet, unfortunately, nor any good
documentation of the changes.
New target processors: Check first to see if the BFD_ASSEMBLER interface is
supported by the file format code you need to use.
New environments: ???
DOCUMENTATION:
The internals of gas need documenting.
The documentation should also contain a "Trouble" section similar to gcc's
manual: real bugs, common problems, incompatibilities, etc.
Anyone want to offer to maintain a man page?
BFD CONVERSION:
The "#ifdef BFD_ASSEMBLER" code is on its way in; the "#ifndef BFD_ASSEMBLER"
code is on its way out. The new code uses BFD data structures, and calls BFD
for anything that needs to be written to the output file. The old code did all
the writing itself, or in a couple of cases, used BFD as a slightly higher
level than stdio (i.e., bfd_seek, bfd_write -- these are not the preferred
interface).
Because of this, some of this code is messy. Lots of ifdef's, and the
non-BFD_ASSEMBLER version often has multiple conditional tests inside it for
various processors or formats. As the various targets get converted over,
these will gradually go away.
As of the moment I'm editing this file, only the "sun4" and "decstation-bsd"
targets can really use the BFD code. Other back ends still need merging or
touching up.
TO DO:
Remove DONTDEF code, commented-out code.
Eliminate, as much as possible, anything not in config that is conditionalized
on a CPU, format, or environment.
Merge COFF support into one version, supporting all the pseudo-ops used in
either versions now, but using BFD for high-level operations. (See second
following item.) Currently there are two versions (plus the new BFD code),
which support different features, and are used on different targets.
Convert remaining a.out/b.out targets to using the BFD_ASSEMBLER code by
default.
Finish conversion to using BFD for all object file writing. (This is the
BFD_ASSEMBLER code, not BFD or BFD_HEADERS.) VMS might be the tough one here,
since there's no BFD support for it at all yet. Eliminate the old code. Some
of this can be done target by target, so doing a target where the CPU or
format already supports BFD_ASSEMBLER mode may be easiest.
Fix lots of uses of empty strings to use null pointers. Will improve
efficiency, and should make code clearer too.
Clean up comments; lots of 'em are one previous maintainer griping about
another previous maintainer, unrelated to the code. (And with no names,
they're not so fun to read. :-)
For sparc: "call 0" becomes "jmpl %g0,%l7", and similarly for absolute
addresses in -4096...4095. (Solaris assembler does this. No
relocation required, no absolute symbol needed.) For addresses
outside the range, for COFF, keep generating an absolute symbol to use
for relocs.
Get Steve to document H8/500 stuff (and others).
Improve test suite. Incorporate more reported net bugs, and non-confidential
Cygnus customer bugs, and anything else.
Add support for i386/i486 16-bit mode, so operating system initialization code
doesn't require a separate assembler nor lots of `.byte' directives.
See if it's more maintainable (and not too much of a performance loss) to use
a yacc grammar for parsing input. The lexer will have to be flexible, and the
grammar will have to contain any construct used on any platform, but it may be
easier to maintain, instead of having code in most of the back ends.
PIC support.
Torbjorn Granlund <tege@cygnus.com> writes, regarding alpha .align:
Please make sure the .align directive works as in digital's assembler.
They fill the space with a sequence of "bis $31,$31,$31;ldq_u $31,0($30)"
since these two instructions can dual-issue. Since .align is ued a lot by
gcc, it is an important optimization.
Torbjorn Granlund <tege@cygnus.com> writes, regarding i386/i486/pentium:
In a new publication from Intel, "Optimization for Intel's 32 bit
Processors", they recommended code alignment on a 16 byte boundary if that
requires less than 8 bytes of fill instructions. The Pentium is not
affected by such alignment, the 386 wants alignment on a 4 byte boundary.
It is the 486 that is most helped by large alignment.
Recommended nop instructions:
1 byte: 90 xchg %eax,%eax
2 bytes: 8b c0 movl %eax,%eax
3 bytes: 8d 76 00 leal 0(%esi),%esi
4 bytes: 8d 74 26 00 leal 0(%esi),%esi
5 bytes: 8b c0 8d 76 00 movl %eax,%eax; leal 0(%esi),%esi
6 bytes: 8d b6 00 00 00 00 leal 0(%esi),%esi
7 bytes: 8d b4 26 00 00 00 00 leal 0(%esi),%esi
Note that `leal 0(%esi),%esi' has a few different encodings...
There are faster instructions for certain lengths, that are not true nops.
If you can determine that a register and the condition code is dead (by
scanning forwards for a register that is written before it is read, and
similar for cc) you can use a `incl reg' for a 3 times faster 1 cycle nop...
(From old "NOTES" file to-do list, not really reviewed:)
fix relocation types for i860, perhaps by adding a ref pointer to fixS?
remove the ifdef's from fx_callj tests?
space tighten sparc alignment?
md_ => tc_
share b.out with a.out.