[pypy-commit] pypy jitframe-on-heap: Update the comments.
arigo
noreply at buildbot.pypy.org
Mon Feb 25 20:08:06 CET 2013
Author: Armin Rigo <arigo at tunes.org>
Branch: jitframe-on-heap
Changeset: r61768:69eeabe5c271
Date: 2013-02-25 20:07 +0100
http://bitbucket.org/pypy/pypy/changeset/69eeabe5c271/
Log: Update the comments.
diff --git a/rpython/jit/backend/x86/regalloc.py b/rpython/jit/backend/x86/regalloc.py
--- a/rpython/jit/backend/x86/regalloc.py
+++ b/rpython/jit/backend/x86/regalloc.py
@@ -720,9 +720,8 @@
#
# - at least the non-callee-saved registers
#
- # - for shadowstack, we assume that any call can collect, and we
+ # - we assume that any call can collect, and we
# save also the callee-saved registers that contain GC pointers
- # XXX for asmgcc too for now.
#
# - for CALL_MAY_FORCE or CALL_ASSEMBLER, we have to save all regs
# anyway, in case we need to do cpu.force(). The issue is that
@@ -734,6 +733,9 @@
if not save_all_regs:
gcrootmap = self.assembler.cpu.gc_ll_descr.gcrootmap
# we save all the registers for shadowstack and asmgcc for now
+ # --- for asmgcc too: we can't say "register x is a gc ref"
+ # without distinguishing call sites, which we don't do any
+ # more for now.
if gcrootmap: # and gcrootmap.is_shadow_stack:
save_all_regs = 2
self.rm.before_call(force_store, save_all_regs=save_all_regs)
More information about the pypy-commit
mailing list