------- Comment #27 from yanov at il dot ibm dot com 2006-01-13 13:26 ------- I've checked compiling the code in comment 24 with -mcpu=power5, compiled without problems. Compiling with -mcpu=power4 (or without -mcpu flag) did produce the ICE as can be seen below.
gcc was configured as follows: ../gcc/configure --enable-threads=posix --prefix=/Develop/mainline-yanov/install/ --enable-shared --enable-__cxa_atexit --host=powerpc-suse-linux --build=powerpc-suse-linux --target=powerpc-suse-linux --enable-targets=powerpc64-suse-linux --enable-biarch --enable-checking --enable-libgcj --with-system-zlib --enable-languages=c,c++,fortran [EMAIL PROTECTED]:~/smstests/sms_tests> /Develop/mainline-yanov/install/bin/gcc -fmod ulo-sched -c -O smaller_comb.c smaller_comb.c: In function foo: smaller_comb.c:9: error: wrong amount of branch edges after conditional jump 9, n_branch=0 smaller_comb.c:9: internal compiler error: verify_flow_info failed Please submit a full bug report, with preprocessed source if appropriate. See <URL:http://gcc.gnu.org/bugs.html> for instructions. [EMAIL PROTECTED]:~/smstests/sms_tests> [EMAIL PROTECTED]:~/smstests/sms_tests> [EMAIL PROTECTED]:~/smstests/sms_tests> /Develop/mainline-yanov/install/bin/gcc -fmod ulo-sched -c -O -mcpu=power5 smaller_comb.c [EMAIL PROTECTED]:~/smstests/sms_tests> [EMAIL PROTECTED]:~/smstests/sms_tests> [EMAIL PROTECTED]:~/smstests/sms_tests> /Develop/mainline-yanov/install/bin/gcc -fmod ulo-sched -c -O -mcpu=power4 smaller_comb.c smaller_comb.c: In function foo: smaller_comb.c:9: error: wrong amount of branch edges after conditional jump 9, n_branch=0 smaller_comb.c:9: internal compiler error: verify_flow_info failed Please submit a full bug report, with preprocessed source if appropriate. See <URL:http://gcc.gnu.org/bugs.html> for instructions. Vladimir (In reply to comment #26) > Subject: Re: [4.1/4.2 Regression] internal compiler error: verify_flow_info > failed > > > ------- Comment #25 from steven at gcc dot gnu dot org 2006-01-12 22:56 > > ------- > > I can not reproduce this ICE. We really need some way to reproduce the > > problem > > that Mustafa's patch was supposed to fix. > > This is still present in the trunk as of today. I can reproduce it using > the testcase in comment #2 on hppa-unknown-linux-gnu. This is the jump_insn: > > (gdb) p debug_rtx (insn) > (jump_insn 21 19 31 2 (parallel [ > (set (pc) > (if_then_else (eq (reg:SI 28 %r28) > (const_int 0 [0x0])) > (label_ref 31) > (pc))) > (set (reg/v:SI 3 %r3 [orig:94 call_result ] [94]) > (reg:SI 28 %r28)) > ]) 227 {*pa.md:8779} (insn_list:REG_DEP_TRUE 19 (nil)) > (expr_list:REG_DEAD (reg:SI 28 %r28) > (expr_list:REG_BR_PROB (const_int 10000 [0x2710]) > (nil)))) > > n_branch == 0. > > Did you try the testcase in #24 "without" Mustafa's patch to cfgrtl.c? > > Dave > -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24626