* [parisc-linux] Any good rules of pratice about branching insn into the delay slot of another one?
@ 2006-04-02 11:46 Joel Soete
2006-04-02 15:20 ` [parisc-linux] " Randolph Chung
0 siblings, 1 reply; 2+ messages in thread
From: Joel Soete @ 2006-04-02 11:46 UTC (permalink / raw)
To: parisc-linux; +Cc: John David Anglin
Hello all,
This question about latest test related to some glibc clone().
This patch:
# diff -Nau clone.S.orig2 clone.S.New4
--- clone.S.orig2 2006-03-24 09:11:27.000000000 +0000
+++ clone.S.New4 2006-03-31 17:42:25.000000000 +0000
@@ -45,7 +45,31 @@
.text
ENTRY(__clone)
+ /* this ENTRY() macro do: stw rp, -20(sp)
+ and "FIXME: I have no idea how profiling works on hppa." */
+ /* Sanity check arguments. */
+ comib,= 0, %arg0, .Larg_error /* no NULL function pointers */
+ nop
+ comib,<>,n 0, %arg1, .Lno_arg_error /* no NULL stack pointers */
+ nop
+
+.Larg_error:
+ /* Create a stack frame */
+/* stwm %r3, 64(%sp) */
+ ldo 64(%sp), %sp
+ /* Set errno */
+ bl __syscall_error, %rp
+ ldi EINVAL, %arg0
+
+ /* Return after setting errno */
+ /* Restore rp from ENTRY() */
+ ldw -84(%sp), %rp
+ bv %r0(%rp)
+/* ldwm -64(%sp), %r3 */
+ ldo -64(%sp), %sp
+
+.Lno_arg_error:
/* Save the fn ptr and arg on the new stack. */
stwm %arg0, 64(%arg1)
stw %arg3, -60(%arg1)
@@ -75,8 +99,10 @@
ldi -4096, %r1
comclr,>>= %r1, %ret0, %r0 /* Note: unsigned compare. */
b,n .Lerror
+ nop
comib,=,n 0, %ret0, thread_start
+ nop /* to avoid bv in the delay slot? */
/* Successful return from the parent
No need to restore the PIC register,
@@ -94,10 +120,12 @@
#endif
/* Set errno */
copy %ret0, %r3
- b __syscall_error
+ bl __syscall_error, %rp
sub %r0, %ret0, %arg0
copy %r3, %ret0
/* Return after setting errno */
+ /* Restore rp from ENTRY() */
+ ldw -84(%sp), %rp
bv %r0(%rp)
ldwm -64(%sp), %r3
====<>====
seems to works seems to be ok:
# grep Err glibc-2.3.6-4.3.1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/localedata/sort-test.out] Error 1
make[2]: *** [localedata/tests] Error 2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/math/test-float.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/math/test-double.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/math/test-idouble.out] Error 1
make[2]: *** [math/tests] Error 2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/posix/tst-rxspencer.out] Error 139
make[3]: [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/posix/annexc.out] Error 1 (ignored)
make[2]: *** [posix/tests] Error 2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/linuxthreads/tst-attr1.out] Error 1
make[2]: *** [linuxthreads/tests] Error 2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/rt/tst-timer.out] Error 137
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/rt/tst-aio4.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/rt/tst-timer4.out] Error 1
make[2]: *** [rt/tests] Error 2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/elf/tst-execstack-needed.out] Error 1
make[2]: *** [elf/tests] Error 2
make[1]: *** [check] Error 2
(well doesn't seems to introduce regression in test)
OTOH if I remove the nop I added in delay slot of 'b,n .Lerror' and 'comib,=,n 0, %ret0, thread_start' i.e. rm hunk:
@@ -75,8 +99,10 @@
ldi -4096, %r1
comclr,>>= %r1, %ret0, %r0 /* Note: unsigned compare. */
b,n .Lerror
+ nop
comib,=,n 0, %ret0, thread_start
+ nop /* to avoid bv in the delay slot? */
/* Successful return from the parent
No need to restore the PIC register,
test became very regressive:
# grep Err glibc-2.3.6-4.3.2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/iconvdata/tst-iconv4.out] Error 1
make[2]: *** [iconvdata/tests] Error 2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/localedata/sort-test.out] Error 1
make[2]: *** [localedata/tests] Error 2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/math/test-float.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/math/test-double.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/math/test-idouble.out] Error 1
make[2]: *** [math/tests] Error 2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/stdio-common/tst-rndseek.out] Error 1
make[2]: *** [stdio-common/tests] Error 2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/libio/bug-mmap-fflush.out] Error 1
make[2]: *** [libio/tests] Error 2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/posix/tst-chmod.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/posix/bug-regex24.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/posix/tst-regex2.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/posix/tst-rxspencer.out] Error 139
make[3]: [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/posix/annexc.out] Error 1 (ignored)
make[2]: *** [posix/tests] Error 2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/linuxthreads/tst-attr1.out] Error 1
make[2]: *** [linuxthreads/tests] Error 2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/rt/tst-clock.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/rt/tst-timer.out] Error 139
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/rt/tst-aio4.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/rt/tst-aio5.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/rt/tst-aio6.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/rt/tst-timer3.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/rt/tst-timer4.out] Error 1
make[2]: *** [rt/tests] Error 2
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/elf/tst-tls13.out] Error 1
make[3]: *** [/CAD/parisc-linux/Dpkg/dpkg-work/glibc-2.3.6/build-tree/hppa-libc/elf/tst-execstack-prog.out] Error 1
make[2]: *** [elf/tests] Error 2
make[1]: *** [check] Error 2
Thanks in advance for advise,
Joel
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux
^ permalink raw reply [flat|nested] 2+ messages in thread
* [parisc-linux] Re: Any good rules of pratice about branching insn into the delay slot of another one?
2006-04-02 11:46 [parisc-linux] Any good rules of pratice about branching insn into the delay slot of another one? Joel Soete
@ 2006-04-02 15:20 ` Randolph Chung
0 siblings, 0 replies; 2+ messages in thread
From: Randolph Chung @ 2006-04-02 15:20 UTC (permalink / raw)
To: Joel Soete; +Cc: John David Anglin, parisc-linux
> This question about latest test related to some glibc clone().
why don't you try to step through the code with gdb (using stepi) and
see the sequence of insns that get executed in the different cases?
randolph
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2006-04-02 15:20 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2006-04-02 11:46 [parisc-linux] Any good rules of pratice about branching insn into the delay slot of another one? Joel Soete
2006-04-02 15:20 ` [parisc-linux] " Randolph Chung
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.