gcc/libgomp/testsuite/libgomp.c++/scan-16.C

155 lines
2.6 KiB
C++
Raw Normal View History

omp-low.c (struct omp_context): Add for_simd_scan_phase member. * omp-low.c (struct omp_context): Add for_simd_scan_phase member. (maybe_lookup_ctx): Add forward declaration. (omp_find_scan): Likewise. Walk into body of simd if composited with worksharing loop. (scan_omp_simd_scan): New function. (scan_omp_1_stmt): Call it. (lower_rec_simd_input_clauses): Don't create rvar nor rvar2 if ctx->for_simd_scan_phase. (lower_rec_input_clauses): Do much less work for inscan reductions in ctx->for_simd_scan_phase is_simd regions. (lower_omp_scan): Set is_simd also on simd constructs composited with worksharing loop, unless ctx->for_simd_scan_phase. Never emit a sorry message. Don't change GIMPLE_OMP_SCAN stmts into nops and emit their body after in simd constructs composited with worksharing loop. (lower_omp_for_scan): Handle worksharing loop composited with simd. * c-c++-common/gomp/scan-4.c: Don't expect sorry message. * testsuite/libgomp.c/scan-11.c: New test. * testsuite/libgomp.c/scan-12.c: New test. * testsuite/libgomp.c/scan-13.c: New test. * testsuite/libgomp.c/scan-14.c: New test. * testsuite/libgomp.c/scan-15.c: New test. * testsuite/libgomp.c/scan-16.c: New test. * testsuite/libgomp.c/scan-17.c: New test. * testsuite/libgomp.c/scan-18.c: New test. * testsuite/libgomp.c++/scan-9.C: New test. * testsuite/libgomp.c++/scan-10.C: New test. * testsuite/libgomp.c++/scan-11.C: New test. * testsuite/libgomp.c++/scan-12.C: New test. * testsuite/libgomp.c++/scan-13.C: New test. * testsuite/libgomp.c++/scan-14.C: New test. * testsuite/libgomp.c++/scan-15.C: New test. * testsuite/libgomp.c++/scan-16.C: New test. From-SVN: r273157
2019-07-06 09:53:48 +02:00
// { dg-require-effective-target size32plus }
vectorizer: Fix up -fsimd-cost-model= handling > * testsuite/libgomp.c++/scan-10.C: Add option -fvect-cost-model=cheap. I don't think this is the right thing to do. This just means that at some point between 2013 when -fsimd-cost-model has been introduced and now -fsimd-cost-model= option at least partially stopped working properly. As documented, -fsimd-cost-model= overrides the -fvect-cost-model= setting for OpenMP simd loops (loop->force_vectorize is true) if specified differently from default. In tree-vectorizer.h we have: static inline bool unlimited_cost_model (loop_p loop) { if (loop != NULL && loop->force_vectorize && flag_simd_cost_model != VECT_COST_MODEL_DEFAULT) return flag_simd_cost_model == VECT_COST_MODEL_UNLIMITED; return (flag_vect_cost_model == VECT_COST_MODEL_UNLIMITED); } and use it in various places, but we also just use flag_vect_cost_model in lots of places (and in one spot use flag_simd_cost_model, not sure if we are sure it is a force_vectorize loop or what). So, IMHO we should change the above inline function to loop_cost_model and let it return the cost model and then just reimplement unlimited_cost_model as return loop_cost_model (loop) == VECT_COST_MODEL_UNLIMITED; and then adjust the direct uses of the flag and revert these changes. 2021-10-12 Jakub Jelinek <jakub@redhat.com> gcc/ * tree-vectorizer.h (loop_cost_model): New function. (unlimited_cost_model): Use it. * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model call instead of flag_vect_cost_model. * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise. (vect_prune_runtime_alias_test_list): Likewise. Also use it instead of flag_simd_cost_model. gcc/testsuite/ * gcc.dg/gomp/simd-2.c: Remove option -fvect-cost-model=cheap. * gcc.dg/gomp/simd-3.c: Likewise. libgomp/ * testsuite/libgomp.c/scan-11.c: Remove option -fvect-cost-model=cheap. * testsuite/libgomp.c/scan-12.c: Likewise. * testsuite/libgomp.c/scan-13.c: Likewise. * testsuite/libgomp.c/scan-14.c: Likewise. * testsuite/libgomp.c/scan-15.c: Likewise. * testsuite/libgomp.c/scan-16.c: Likewise. * testsuite/libgomp.c/scan-17.c: Likewise. * testsuite/libgomp.c/scan-18.c: Likewise. * testsuite/libgomp.c/scan-19.c: Likewise. * testsuite/libgomp.c/scan-20.c: Likewise. * testsuite/libgomp.c/scan-21.c: Likewise. * testsuite/libgomp.c/scan-22.c: Likewise. * testsuite/libgomp.c++/scan-9.C: Likewise. * testsuite/libgomp.c++/scan-10.C: Likewise. * testsuite/libgomp.c++/scan-11.C: Likewise. * testsuite/libgomp.c++/scan-12.C: Likewise. * testsuite/libgomp.c++/scan-13.C: Likewise. * testsuite/libgomp.c++/scan-14.C: Likewise. * testsuite/libgomp.c++/scan-15.C: Likewise. * testsuite/libgomp.c++/scan-16.C: Likewise.
2021-10-12 09:28:10 +02:00
// { dg-additional-options "-O2 -fopenmp -fdump-tree-vect-details" }
// { dg-additional-options "-msse2" { target sse2_runtime } }
omp-low.c (struct omp_context): Add for_simd_scan_phase member. * omp-low.c (struct omp_context): Add for_simd_scan_phase member. (maybe_lookup_ctx): Add forward declaration. (omp_find_scan): Likewise. Walk into body of simd if composited with worksharing loop. (scan_omp_simd_scan): New function. (scan_omp_1_stmt): Call it. (lower_rec_simd_input_clauses): Don't create rvar nor rvar2 if ctx->for_simd_scan_phase. (lower_rec_input_clauses): Do much less work for inscan reductions in ctx->for_simd_scan_phase is_simd regions. (lower_omp_scan): Set is_simd also on simd constructs composited with worksharing loop, unless ctx->for_simd_scan_phase. Never emit a sorry message. Don't change GIMPLE_OMP_SCAN stmts into nops and emit their body after in simd constructs composited with worksharing loop. (lower_omp_for_scan): Handle worksharing loop composited with simd. * c-c++-common/gomp/scan-4.c: Don't expect sorry message. * testsuite/libgomp.c/scan-11.c: New test. * testsuite/libgomp.c/scan-12.c: New test. * testsuite/libgomp.c/scan-13.c: New test. * testsuite/libgomp.c/scan-14.c: New test. * testsuite/libgomp.c/scan-15.c: New test. * testsuite/libgomp.c/scan-16.c: New test. * testsuite/libgomp.c/scan-17.c: New test. * testsuite/libgomp.c/scan-18.c: New test. * testsuite/libgomp.c++/scan-9.C: New test. * testsuite/libgomp.c++/scan-10.C: New test. * testsuite/libgomp.c++/scan-11.C: New test. * testsuite/libgomp.c++/scan-12.C: New test. * testsuite/libgomp.c++/scan-13.C: New test. * testsuite/libgomp.c++/scan-14.C: New test. * testsuite/libgomp.c++/scan-15.C: New test. * testsuite/libgomp.c++/scan-16.C: New test. From-SVN: r273157
2019-07-06 09:53:48 +02:00
// { dg-additional-options "-mavx" { target avx_runtime } }
// { dg-final { scan-tree-dump-times "vectorized \[2-6] loops" 2 "vect" { target sse2_runtime } } }
omp-low.c (struct omp_context): Add for_simd_scan_phase member. * omp-low.c (struct omp_context): Add for_simd_scan_phase member. (maybe_lookup_ctx): Add forward declaration. (omp_find_scan): Likewise. Walk into body of simd if composited with worksharing loop. (scan_omp_simd_scan): New function. (scan_omp_1_stmt): Call it. (lower_rec_simd_input_clauses): Don't create rvar nor rvar2 if ctx->for_simd_scan_phase. (lower_rec_input_clauses): Do much less work for inscan reductions in ctx->for_simd_scan_phase is_simd regions. (lower_omp_scan): Set is_simd also on simd constructs composited with worksharing loop, unless ctx->for_simd_scan_phase. Never emit a sorry message. Don't change GIMPLE_OMP_SCAN stmts into nops and emit their body after in simd constructs composited with worksharing loop. (lower_omp_for_scan): Handle worksharing loop composited with simd. * c-c++-common/gomp/scan-4.c: Don't expect sorry message. * testsuite/libgomp.c/scan-11.c: New test. * testsuite/libgomp.c/scan-12.c: New test. * testsuite/libgomp.c/scan-13.c: New test. * testsuite/libgomp.c/scan-14.c: New test. * testsuite/libgomp.c/scan-15.c: New test. * testsuite/libgomp.c/scan-16.c: New test. * testsuite/libgomp.c/scan-17.c: New test. * testsuite/libgomp.c/scan-18.c: New test. * testsuite/libgomp.c++/scan-9.C: New test. * testsuite/libgomp.c++/scan-10.C: New test. * testsuite/libgomp.c++/scan-11.C: New test. * testsuite/libgomp.c++/scan-12.C: New test. * testsuite/libgomp.c++/scan-13.C: New test. * testsuite/libgomp.c++/scan-14.C: New test. * testsuite/libgomp.c++/scan-15.C: New test. * testsuite/libgomp.c++/scan-16.C: New test. From-SVN: r273157
2019-07-06 09:53:48 +02:00
extern "C" void abort ();
struct S {
inline S ();
inline ~S ();
inline S (const S &);
inline S & operator= (const S &);
int s;
};
S::S () : s (0)
{
}
S::~S ()
{
}
S::S (const S &x)
{
s = x.s;
}
S &
S::operator= (const S &x)
{
s = x.s;
return *this;
}
static inline void
ini (S &x)
{
x.s = 0;
}
S r, a[1024], b[1024];
#pragma omp declare reduction (+: S: omp_out.s += omp_in.s)
#pragma omp declare reduction (plus: S: omp_out.s += omp_in.s) initializer (ini (omp_priv))
__attribute__((noipa)) void
foo (S *a, S *b, S &r)
{
#pragma omp for simd simdlen (1) reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r.s += a[i].s;
}
}
__attribute__((noipa)) S
bar (void)
{
S s;
#pragma omp parallel
#pragma omp for simd if (0) reduction (inscan, plus:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s.s += 2 * a[i].s;
}
return s;
}
__attribute__((noipa)) void
baz (S *a, S *b, S &r)
{
#pragma omp parallel for simd reduction (inscan, +:r)
for (int i = 0; i < 1024; i++)
{
b[i] = r;
#pragma omp scan exclusive(r)
r.s += a[i].s;
}
}
__attribute__((noipa)) S
qux (void)
{
S s;
#pragma omp parallel for simd reduction (inscan, plus:s)
for (int i = 0; i < 1024; i++)
{
b[i] = s;
#pragma omp scan exclusive(s)
s.s += 2 * a[i].s;
}
return s;
}
int
main ()
{
S s;
for (int i = 0; i < 1024; ++i)
{
a[i].s = i;
b[i].s = -1;
asm ("" : "+g" (i));
}
#pragma omp parallel
foo (a, b, r);
if (r.s != 1024 * 1023 / 2)
abort ();
for (int i = 0; i < 1024; ++i)
{
if (b[i].s != s.s)
abort ();
else
b[i].s = 25;
s.s += i;
}
if (bar ().s != 1024 * 1023)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i].s != s.s)
abort ();
s.s += 2 * i;
}
r.s = 0;
baz (a, b, r);
if (r.s != 1024 * 1023 / 2)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i].s != s.s)
abort ();
else
b[i].s = 25;
s.s += i;
}
if (qux ().s != 1024 * 1023)
abort ();
s.s = 0;
for (int i = 0; i < 1024; ++i)
{
if (b[i].s != s.s)
abort ();
s.s += 2 * i;
}
}