https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89295
Bug ID: 89295
Summary: [9 regression] compilation of
gcc.dg-struct-layout-1/t001_x.c takes 30 times as long
after r268404
Product: gcc
Version: 9.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: rtl-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: seurer at gcc dot gnu.org
Target Milestone: ---
Executing on host: /home/seurer/gcc/build/gcc-trunk/gcc/xgcc
-B/home/seurer/gcc/build/gcc-trunk/gcc/ -fno-diagnostics-show-caret
-fno-diagnostics-show-line-numbers -fdiagnostics-color=never -w
-I/home/seurer/gcc/gcc-trunk/gcc/testsuite/gcc.dg/compat -Wno-abi -c -o
c_compat_x_tst.o
/home/seurer/gcc/build/gcc-trunk/gcc/testsuite/gcc5/gcc.dg-struct-layout-1//t001_x.c
(timeout = 300)
spawn -ignore SIGHUP /home/seurer/gcc/build/gcc-trunk/gcc/xgcc
-B/home/seurer/gcc/build/gcc-trunk/gcc/ -fno-diagnostics-show-caret
-fno-diagnostics-show-line-numbers -fdiagnostics-color=never -w
-I/home/seurer/gcc/gcc-trunk/gcc/testsuite/gcc.dg/compat -Wno-abi -c -o
c_compat_x_tst.o
/home/seurer/gcc/build/gcc-trunk/gcc/testsuite/gcc5/gcc.dg-struct-layout-1//t001_x.c
WARNING: program timed out
compiler exited with status 1
FAIL: tmpdir-gcc.dg-struct-layout-1/t001 c_compat_x_tst.o compile
I started noticing this occasional timeout issue with this test case recently
on some of our test machines. Note that the timeout value is 300 seconds when
run during a make check.
Trying the same test with some older revisions showed (when it is compiled
alone) that the compilation finishes in about 2s. With recent revisions it
takes 60+s.
This may not be the only test that slows down of course. "make -k check-gcc
RUNTESTFLAGS=struct-layout-1.exp" which runs many tests including the above
test takes about 4m30s at r268403 and about 7m at r268404.
Looking at the ftime-report outputs the big differences are in LRA
Before:
LRA non-specific : 0.25 ( 10%) 0.01 ( 4%) 0.24 ( 9%)
4168 kB ( 5%)
LRA virtuals elimination : 0.03 ( 1%) 0.00 ( 0%) 0.05 ( 2%)
355 kB ( 0%)
LRA reload inheritance : 0.03 ( 1%) 0.00 ( 0%) 0.03 ( 1%)
1618 kB ( 2%)
LRA create live ranges : 0.04 ( 2%) 0.00 ( 0%) 0.04 ( 1%)
538 kB ( 1%)
LRA hard reg assignment : 0.07 ( 3%) 0.00 ( 0%) 0.06 ( 2%)
0 kB ( 0%)
After:
LRA non-specific : 7.34 ( 12%) 0.04 ( 9%) 7.59 ( 13%)
14335 kB ( 13%)
LRA virtuals elimination : 26.07 ( 43%) 0.06 ( 13%) 25.26 ( 42%)
3613 kB ( 3%)
LRA reload inheritance : 0.04 ( 0%) 0.00 ( 0%) 0.00 ( 0%)
1618 kB ( 2%)
LRA create live ranges : 16.84 ( 28%) 0.06 ( 13%) 17.30 ( 29%)
1735 kB ( 2%)
LRA hard reg assignment : 7.53 ( 13%) 0.02 ( 4%) 7.72 ( 13%)
0 kB ( 0%)
Note: I tried this on x86 and did not see the same slowdown between r268403
and r268404.
The trigger revision was
------------------------------------------------------------------------
r268404 | vmakarov | 2019-01-30 15:49:23 -0600 (Wed, 30 Jan 2019) | 12 lines
2019-01-30 Vladimir Makarov <vmaka...@redhat.com>
PR rtl-optimization/87246
* lra-constraints.c (simplify_operand_subreg): Reload memory
in subreg if the address became invalid.
2019-01-30 Vladimir Makarov <vmaka...@redhat.com>
PR rtl-optimization/87246
* gcc.target/i386/pr87246.c: New.
