fill in some missing history on Racket's slice of lightning
This commit is contained in:
parent
69b9ff1117
commit
8d6870a7c3
|
@ -1,5 +1,6 @@
|
|||
|
||||
This variant of GNU Lightning version 1.2 is modified for Racket:
|
||||
This variant of GNU Lightning version 1.2 is modified for Racket in
|
||||
many ways, including the following:
|
||||
|
||||
* For PowerPC, branches can be generated in long-jump mode, so that
|
||||
patched addresses are not limited to a 2^16 or 2^26 difference from
|
||||
|
@ -18,6 +19,9 @@ This variant of GNU Lightning version 1.2 is modified for Racket:
|
|||
support has many limitations, such as allowing no more than 3
|
||||
function arguments.
|
||||
|
||||
* For i386 and x86_64, "tiny" and "short" (on x86_64) jumps are
|
||||
supported, along the lines of supporting long jumps on PowerPC.
|
||||
|
||||
* "core-common.h" adds a `jit_code' field into the `x' union, instead
|
||||
of taking the address of `pc' and casting it to a `jit_code*',
|
||||
because the latter violates C99 aliasing rules.
|
||||
|
@ -31,6 +35,12 @@ This variant of GNU Lightning version 1.2 is modified for Racket:
|
|||
On x86, there are aso _fppush variants that ignore the provided
|
||||
registers and work in terms of the FP stack, instead.
|
||||
|
||||
* The ARM support was backported from Paulo César Pereira de
|
||||
Andrade's original (pre-lightning-2.0) implementation. The register
|
||||
assignment was changed, jump-length modes added, and so on. The
|
||||
instruction-set choices are made statically based on pre-processor
|
||||
definitions.
|
||||
|
||||
----------------------------------------
|
||||
|
||||
GNU Lightning authors:
|
||||
|
@ -38,6 +48,7 @@ GNU Lightning authors:
|
|||
Paolo Bonzini <bonzini@gnu.org>
|
||||
i386 and PPC assemblers by Ian Piumarta <piumarta@inria.fr>
|
||||
Major PPC contributions by Laurent Michel <ldm@thorgal.homelinux.org>
|
||||
ARM assembler by Paulo César Pereira de Andrade
|
||||
|
||||
----------------------------------------
|
||||
|
||||
|
|
Loading…
Reference in New Issue
Block a user