X-Git-Url: http://repo.macrolet.net/gitweb/?a=blobdiff_plain;f=doc%2Finternals%2Fcalling-convention.texinfo;h=f390521edfd1c16d97a9f62714ffaf036141794c;hb=f961fda4e87e79aacc6c61172303106d3e9522de;hp=bf5b2f37a385b04ebf0e5cf2edc5fcbc19027ea1;hpb=2973941cf214a5ea274cd0381a651ce0e1a7fab2;p=sbcl.git

diff --git a/doc/internals/calling-convention.texinfo b/doc/internals/calling-convention.texinfo
index bf5b2f3..f390521 100644
--- a/doc/internals/calling-convention.texinfo
+++ b/doc/internals/calling-convention.texinfo
@@ -119,7 +119,39 @@ frame to include sufficient space for its local variables, after
 possibly converting any @code{&rest} arguments to a proper list.
 
 The above scheme was changed in 1.0.27 on x86 and x86-64 by swapping
-the old frame pointer and the return address.
+the old frame pointer with the return address and making EBP point two
+words later:
+
+On x86/x86-64 the stack now looks like this (stack grows downwards):
+
+@verbatim
+----------
+RETURN PC
+----------
+OLD FP
+---------- <- FP points here
+EMPTY SLOT
+----------
+FIRST ARG
+----------
+@end verbatim
+
+just as if the function had been CALLed and upon entry executed the
+standard prologue: PUSH EBP; MOV EBP, ESP. On other architectures the
+stack looks like this (stack grows upwards):
+
+@verbatim
+----------
+FIRST ARG
+----------
+EMPTY SLOT
+----------
+RETURN PC
+----------
+OLD FP
+---------- <- FP points here
+@end verbatim
+
 
 @node Unknown-Values Returns
 @comment  node-name,  next,  previous,  up
@@ -180,9 +212,11 @@ the case of an entry point for a full call).
 @comment  node-name,  next,  previous,  up
 @section Additional Notes
 
-The low-hanging fruit here is going to be changing every call and
-return to use @code{CALL} and @code{RETURN} instructions instead of
-@code{JMP} instructions.
+The low-hanging fruit is going to be changing every call and return to
+use @code{CALL} and @code{RETURN} instructions instead of @code{JMP}
+instructions which is partly done on x86oids: a trampoline is
+@code{CALL}ed and that @code{JMP}s to the target which is sufficient
+to negate (most of?) the penalty.
 
 A more involved change would be to reduce the number of argument
 passing registers from three to two, which may be beneficial in terms