https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91858
Bug ID: 91858
Summary: [9/10 Regression] Compile time hog w/ complex float
trigonometric functions
Product: gcc
Version: 10.0
Status: UNCONFIRMED
Keywords: compile-time-hog
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: asolokha at gmx dot com
Target Milestone: ---
Target: x86_64-unknown-linux-gnu
g++-10.0.0-alpha20190915 snapshot (r275736) linked against GNU MP 6.1.2 takes
indefinite time when compiling the following testcase, reduced from
test/std/numerics/complex.number/complex.transcendentals/tan.pass.cpp from the
libcxx 9.0.0 test suite, w/ -O1:
template <typename> class yx;
template <typename wk> void
operator== (yx<wk>, yx<wk>);
template <> struct yx<float> {
yx (_Complex float ph) : lu{ph} { }
yx (float c0 = 0.0f, float m5 = 0.0f) : lu{c0, m5} { }
_Complex float lu;
};
template <class um> void
al (yx<um> qo)
{
_Complex float xy = qo.lu, ny = __builtin_ctanf (xy);
yx<float> k2 = ny, oi (1);
k2 == oi;
}
void
al ()
{
al (yx<float> (10000, 10000));
}
% timeout 10 g++-10.0.0-alpha20190915 -O1 -c ahpihctw.cpp
zsh: exit 124 timeout 10 g++-10.0.0-alpha20190915 -O1 -c ahpihctw.cpp
Some perf top entries:
51.41% libgmp.so.10.3.2 [.] __gmpn_mul_basecase
15.23% libgmp.so.10.3.2 [.] __gmpn_sub_n
11.09% libgmp.so.10.3.2 [.] __gmpn_add_n
10.50% libgmp.so.10.3.2 [.] __gmpn_sqr_basecase
6.00% libgmp.so.10.3.2 [.] __gmpn_rsblsh1_n
1.97% libgmp.so.10.3.2 [.] 0x00000000000468e3
1.84% libgmp.so.10.3.2 [.] __gmpn_lshiftc
0.69% libgmp.so.10.3.2 [.] __gmpn_copyi
0.53% libgmp.so.10.3.2 [.] __gmpn_lshift
0.36% libgmp.so.10.3.2 [.] __gmpn_rsh1sub_n
0.27% libgmp.so.10.3.2 [.] __gmpn_bdiv_dbm1c
0.03% libgmp.so.10.3.2 [.] 0x00000000000462fb
0.03% libgmp.so.10.3.2 [.] 0x000000000003c251
0.03% libgmp.so.10.3.2 [.] __gmpn_rsh1add_n
I believe this was already reported somewhere but I failed to find the PR.
