https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69252
Bug ID: 69252
Summary: [4.9/5/6 Regression] gcc.dg/vect/vect-iv-9.c FAILs
with -Os -fmodulo-sched -fmodulo-sched-allow-regmoves
-fsched-pressure
Product: gcc
Version: 6.0
Status: UNCONFIRMED
Keywords: wrong-code
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: zsojka at seznam dot cz
Target Milestone: ---
Host: x86_64-pc-linux-gnu
Target: powerpc64-unknown-linux-gnu
Created attachment 37319
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=37319&action=edit
reduced testcase
Output (qemu userspace emulation used):
$ powerpc64-unknown-linux-gnu-gcc -v
Using built-in specs.
COLLECT_GCC=/repo/gcc-trunk/binary-trunk-232261-checking-yes-rtl-df-nographite-powerpc64/bin/powerpc64-unknown-linux-gnu-gcc
COLLECT_LTO_WRAPPER=/repo/gcc-trunk/binary-trunk-232261-checking-yes-rtl-df-nographite-powerpc64/bin/../libexec/gcc/powerpc64-unknown-linux-gnu/6.0.0/lto-wrapper
Target: powerpc64-unknown-linux-gnu
Configured with: /repo/gcc-trunk//configure --enable-languages=c,c++
--enable-checking=yes,rtl,df --without-cloog --without-ppl --without-isl
--build=x86_64-pc-linux-gnu --host=x86_64-pc-linux-gnu
--target=powerpc64-unknown-linux-gnu
--with-ld=/usr/bin/powerpc64-unknown-linux-gnu-ld
--with-as=/usr/bin/powerpc64-unknown-linux-gnu-as
--with-sysroot=/usr/powerpc64-unknown-linux-gnu --disable-multilib
--disable-libstdcxx-pch
--prefix=/repo/gcc-trunk//binary-trunk-232261-checking-yes-rtl-df-nographite-powerpc64
Thread model: posix
gcc version 6.0.0 20160112 (experimental) (GCC)
$ powerpc64-unknown-linux-gnu-gcc -Os -fmodulo-sched
-fmodulo-sched-allow-regmoves -fsched-pressure testcase.c
$ ./a.out
qemu: uncaught target signal 6 (Aborted) - core dumped
Aborted
Only powerpc64 seems to be affected.
The wrong code seems to be:
.L.main1:
li 9,0
addis 5,2,.LC1@toc@ha # gpr load fusion, type long
ld 5,.LC1@toc@l(5)
li 10,24
mtctr 10
sldi 6,9,2
lwzx 10,5,6
addi 9,9,1
sldi 6,9,2
add 7,10,8
mr 8,9
...
r8 is used uninitialised here (it has value 25 set by the caller).
r8 always stores the value of r9 from the previous iteration, but here, it is
unset.
Something like:
@@ -21,6 +21,7 @@
li 10,24
mtctr 10
sldi 6,9,2
+ mr 8,9
lwzx 10,5,6
addi 9,9,1
sldi 6,9,2
would fix the example.
The code emitted without -fmodulo-sched-allow-regmoves is much smaller (the
partial unrolling is not done).
Tested revisions:
trunk r232261 - FAIL
5-branch r232254 - FAIL
4_9-branch r232213 - FAIL
4_8-branch r224828 - FAIL
4_7-branch r211571 - OK
4_6-branch r197894 - OK
4_5-branch r189152 - OK
