Setup a 'proper' stackframe in translated blocks. This doesn't affect performance noticeably,
but does ensure that
a) The stack is aligned correctly on OS X with no extra effort, and
b) We can't mess up the stack and crash that way anymore.
Replace all PUSH/POP instructions (outside of prologue/epilogue) with ESP-rel moves to stack
local variables.
Finally merge ia32mac and ia32abi together, since they're pretty much the same now anyway (and
thereby simplifying maintenance a good deal)
src/sh4/ia32abi.h
src/sh4/ia32mac.h
src/sh4/ia64abi.h
src/sh4/sh4trans.h
src/sh4/sh4x86.in
src/sh4/x86op.h
but does ensure that
a) The stack is aligned correctly on OS X with no extra effort, and
b) We can't mess up the stack and crash that way anymore.
Replace all PUSH/POP instructions (outside of prologue/epilogue) with ESP-rel moves to stack
local variables.
Finally merge ia32mac and ia32abi together, since they're pretty much the same now anyway (and
thereby simplifying maintenance a good deal)
src/sh4/ia32abi.h
src/sh4/ia32mac.h
src/sh4/ia64abi.h
src/sh4/sh4trans.h
src/sh4/sh4x86.in
src/sh4/x86op.h
Make HAVE_FASTCALL dependent on whether the compiler supports the attribute,
rather than on whether we're making an optimized build - in any case there's
no good reason to disable it for debug builds.
acinclude.m4
aclocal.m4
configure.in
rather than on whether we're making an optimized build - in any case there's
no good reason to disable it for debug builds.
acinclude.m4
aclocal.m4
configure.in
Disable the generational translation cache - I've got no evidence that it
actually helps performance, and it simplifies things to get rid of it (in
particular, translated code doesn't have to worry about being moved now).
src/dreamcast.h
src/sh4/xltcache.c
src/test/testxlt.c
actually helps performance, and it simplifies things to get rid of it (in
particular, translated code doesn't have to worry about being moved now).
src/dreamcast.h
src/sh4/xltcache.c
src/test/testxlt.c
15 years agoOnly call finish_rendering() for texture renders - workaround bug in apple/intel drivers
Only call finish_rendering() for texture renders - workaround bug in apple/intel drivers
src/pvr2/pvr2.c
src/pvr2/pvr2.c
.