From 5d3ebb71cbf9c6ed0ce04230a387485fa57e71ed Mon Sep 17 00:00:00 2001 From: Jan Hubicka Date: Sun, 5 Feb 2017 18:16:52 +0100 Subject: [PATCH] re PR tree-optimization/79347 (vect_do_peeling is messing up profile) PR tree-ssa/79347 * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add ELSE_PROB. * cfgloopmanip.h (loop_version): Update prototype. * modulo-sched.c (sms_schedule): Update call of loop_version. * tree-if-conv.c(version_loop_for_if_conversion): Likewise. * tree-parloops.c (gen_parallel_loop): Likewise. * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise. * tree-ssa-loop-split.c (split_loop): Likewise. * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise. * tree-vect-loop-manip.c (vect_loop_versioning): Likewise. * gcc.dg/tree-ssa/ifc-10.c: Match for profile mismatches. * gcc.dg/tree-ssa/ifc-11.c: Match for profile mismatches. * gcc.dg/tree-ssa/ifc-12.c: Match for profile mismatches. * gcc.dg/tree-ssa/ifc-20040816-1.c: Match for profile mismatches. * gcc.dg/tree-ssa/ifc-20040816-2.c: Match for profile mismatches. * gcc.dg/tree-ssa/ifc-5.c: Match for profile mismatches. * gcc.dg/tree-ssa/ifc-8.c: Match for profile mismatches. * gcc.dg/tree-ssa/ifc-9.c: Match for profile mismatches. * gcc.dg/tree-ssa/ifc-cd.c: Match for profile mismatches. * gcc.dg/tree-ssa/ifc-pr56541.c: Match for profile mismatches. * gcc.dg/tree-ssa/ifc-pr68583.c: Match for profile mismatches. * gcc.dg/tree-ssa/ifc-pr69489-1.c: Match for profile mismatches. * gcc.dg/tree-ssa/ifc-pr69489-2.c: Match for profile mismatches. From-SVN: r245196 --- gcc/ChangeLog | 14 ++++++++++++++ gcc/cfgloopmanip.c | 14 +++++++++----- gcc/cfgloopmanip.h | 3 ++- gcc/modulo-sched.c | 3 ++- gcc/testsuite/ChangeLog | 17 +++++++++++++++++ gcc/testsuite/gcc.dg/tree-ssa/ifc-10.c | 10 +++++++++- gcc/testsuite/gcc.dg/tree-ssa/ifc-11.c | 10 +++++++++- gcc/testsuite/gcc.dg/tree-ssa/ifc-12.c | 11 ++++++++++- gcc/testsuite/gcc.dg/tree-ssa/ifc-20040816-1.c | 9 ++++++++- gcc/testsuite/gcc.dg/tree-ssa/ifc-20040816-2.c | 10 +++++++++- gcc/testsuite/gcc.dg/tree-ssa/ifc-5.c | 9 ++++++++- gcc/testsuite/gcc.dg/tree-ssa/ifc-8.c | 10 +++++++++- gcc/testsuite/gcc.dg/tree-ssa/ifc-9.c | 9 ++++++++- gcc/testsuite/gcc.dg/tree-ssa/ifc-cd.c | 10 +++++++++- gcc/testsuite/gcc.dg/tree-ssa/ifc-pr56541.c | 10 +++++++++- gcc/testsuite/gcc.dg/tree-ssa/ifc-pr68583.c | 9 ++++++++- gcc/testsuite/gcc.dg/tree-ssa/ifc-pr69489-1.c | 10 +++++++++- gcc/testsuite/gcc.dg/tree-ssa/ifc-pr69489-2.c | 9 ++++++++- gcc/tree-if-conv.c | 10 ++++++++-- gcc/tree-parloops.c | 3 ++- gcc/tree-ssa-loop-manip.c | 3 ++- gcc/tree-ssa-loop-split.c | 3 ++- gcc/tree-ssa-loop-unswitch.c | 2 +- gcc/tree-vect-loop-manip.c | 6 ++++-- 24 files changed, 176 insertions(+), 28 deletions(-) diff --git a/gcc/ChangeLog b/gcc/ChangeLog index 22a8102f299..a8c6be100e2 100644 --- a/gcc/ChangeLog +++ b/gcc/ChangeLog @@ -1,3 +1,17 @@ +2017-02-05 Jan Hubicka + + PR tree-ssa/79347 + * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add + ELSE_PROB. + * cfgloopmanip.h (loop_version): Update prototype. + * modulo-sched.c (sms_schedule): Update call of loop_version. + * tree-if-conv.c(version_loop_for_if_conversion): Likewise. + * tree-parloops.c (gen_parallel_loop): Likewise. + * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise. + * tree-ssa-loop-split.c (split_loop): Likewise. + * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise. + * tree-vect-loop-manip.c (vect_loop_versioning): Likewise. + 2017-02-05 Martin Liska PR bootstrap/78985 diff --git a/gcc/cfgloopmanip.c b/gcc/cfgloopmanip.c index 71fbe6b56e4..a0e3e30720e 100644 --- a/gcc/cfgloopmanip.c +++ b/gcc/cfgloopmanip.c @@ -1645,11 +1645,14 @@ force_single_succ_latches (void) | +---------> [second_head] - THEN_PROB is the probability of then branch of the condition. */ + THEN_PROB is the probability of then branch of the condition. + ELSE_PROB is the probability of else branch. Note that they may be both + REG_BR_PROB_BASE when condition is IFN_LOOP_VECTORIZED. */ static basic_block lv_adjust_loop_entry_edge (basic_block first_head, basic_block second_head, - edge e, void *cond_expr, unsigned then_prob) + edge e, void *cond_expr, unsigned then_prob, + unsigned else_prob) { basic_block new_head = NULL; edge e1; @@ -1668,7 +1671,7 @@ lv_adjust_loop_entry_edge (basic_block first_head, basic_block second_head, e1 = make_edge (new_head, first_head, current_ir_type () == IR_GIMPLE ? EDGE_TRUE_VALUE : 0); e1->probability = then_prob; - e->probability = REG_BR_PROB_BASE - then_prob; + e->probability = else_prob; e1->count = apply_probability (e->count, e1->probability); e->count = apply_probability (e->count, e->probability); @@ -1701,7 +1704,8 @@ lv_adjust_loop_entry_edge (basic_block first_head, basic_block second_head, struct loop * loop_version (struct loop *loop, void *cond_expr, basic_block *condition_bb, - unsigned then_prob, unsigned then_scale, unsigned else_scale, + unsigned then_prob, unsigned else_prob, + unsigned then_scale, unsigned else_scale, bool place_after) { basic_block first_head, second_head; @@ -1732,7 +1736,7 @@ loop_version (struct loop *loop, /* Split loop entry edge and insert new block with cond expr. */ cond_bb = lv_adjust_loop_entry_edge (first_head, second_head, - entry, cond_expr, then_prob); + entry, cond_expr, then_prob, else_prob); if (condition_bb) *condition_bb = cond_bb; diff --git a/gcc/cfgloopmanip.h b/gcc/cfgloopmanip.h index 004becfccb1..7ad3bf5d7d4 100644 --- a/gcc/cfgloopmanip.h +++ b/gcc/cfgloopmanip.h @@ -58,6 +58,7 @@ basic_block create_preheader (struct loop *, int); extern void create_preheaders (int); extern void force_single_succ_latches (void); struct loop * loop_version (struct loop *, void *, - basic_block *, unsigned, unsigned, unsigned, bool); + basic_block *, unsigned, unsigned, + unsigned, unsigned, bool); #endif /* GCC_CFGLOOPMANIP_H */ diff --git a/gcc/modulo-sched.c b/gcc/modulo-sched.c index 619ea04c2e7..270adc5a200 100644 --- a/gcc/modulo-sched.c +++ b/gcc/modulo-sched.c @@ -1713,7 +1713,8 @@ sms_schedule (void) * REG_BR_PROB_BASE) / 100; loop_version (loop, comp_rtx, &condition_bb, - prob, prob, REG_BR_PROB_BASE - prob, + prob, REG_BR_PROB_BASE - prob, + prob, REG_BR_PROB_BASE - prob, true); } diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog index 898f55a44a3..3214eef3bda 100644 --- a/gcc/testsuite/ChangeLog +++ b/gcc/testsuite/ChangeLog @@ -1,3 +1,20 @@ +2017-02-05 Jan Hubicka + + PR tree-ssa/79347 + * gcc.dg/tree-ssa/ifc-10.c: Match for profile mismatches. + * gcc.dg/tree-ssa/ifc-11.c: Match for profile mismatches. + * gcc.dg/tree-ssa/ifc-12.c: Match for profile mismatches. + * gcc.dg/tree-ssa/ifc-20040816-1.c: Match for profile mismatches. + * gcc.dg/tree-ssa/ifc-20040816-2.c: Match for profile mismatches. + * gcc.dg/tree-ssa/ifc-5.c: Match for profile mismatches. + * gcc.dg/tree-ssa/ifc-8.c: Match for profile mismatches. + * gcc.dg/tree-ssa/ifc-9.c: Match for profile mismatches. + * gcc.dg/tree-ssa/ifc-cd.c: Match for profile mismatches. + * gcc.dg/tree-ssa/ifc-pr56541.c: Match for profile mismatches. + * gcc.dg/tree-ssa/ifc-pr68583.c: Match for profile mismatches. + * gcc.dg/tree-ssa/ifc-pr69489-1.c: Match for profile mismatches. + * gcc.dg/tree-ssa/ifc-pr69489-2.c: Match for profile mismatches. + 2017-02-05 Andre Vehreschild PR fortran/79344 diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ifc-10.c b/gcc/testsuite/gcc.dg/tree-ssa/ifc-10.c index fbac0099446..4097145eba6 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/ifc-10.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/ifc-10.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-Ofast -fno-split-loops -fdump-tree-ifcvt-stats" } */ +/* { dg-options "-Ofast -fno-split-loops -fdump-tree-ifcvt-stats-blocks-details" } */ /* { dg-require-visibility "" } */ int b[256] = {0}, y; @@ -20,3 +20,11 @@ int foo (int x, int n) } /* { dg-final { scan-tree-dump-times "Applying if-conversion" 1 "ifcvt" } } */ + +/* We insert into code + if (LOOP_VECTORIZED (...)) + which is folded by vectorizer. Both outgoing edges must have probability + 100% so the resulting profile match after folding. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of outgoing probabilities 200.0" 1 "ifcvt" } } */ +/* { dg-final { scan-tree-dump-times "Invalid sum of incoming frequencies" 1 "ifcvt" } } */ + diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ifc-11.c b/gcc/testsuite/gcc.dg/tree-ssa/ifc-11.c index bacf428ec03..a0333fbb28c 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/ifc-11.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/ifc-11.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-Ofast -fdump-tree-ifcvt-stats" } */ +/* { dg-options "-Ofast -fdump-tree-ifcvt-stats-blocks-details" } */ /* { dg-require-visibility "" } */ int a[1024] = {0.0}; @@ -18,3 +18,11 @@ int foo (float *x) } /* { dg-final { scan-tree-dump-times "Applying if-conversion" 1 "ifcvt" } } */ + +/* We insert into code + if (LOOP_VECTORIZED (...)) + which is folded by vectorizer. Both outgoing edges must have probability + 100% so the resulting profile match after folding. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of outgoing probabilities 200.0" 1 "ifcvt" } } */ +/* Sum is wrong here, but not enough for error to be reported. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of incoming frequencies" 0 "ifcvt" } } */ diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ifc-12.c b/gcc/testsuite/gcc.dg/tree-ssa/ifc-12.c index 89d42b4d6fd..535c1f0eb6c 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/ifc-12.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/ifc-12.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-Ofast -fdump-tree-ifcvt-stats" } */ +/* { dg-options "-Ofast -fdump-tree-ifcvt-stats-blocks-details" } */ /* { dg-require-visibility "" } */ struct st @@ -23,3 +23,12 @@ int foo (int x) return 0; } /* { dg-final { scan-tree-dump-times "Applying if-conversion" 1 "ifcvt" } } */ + +/* We insert into code + if (LOOP_VECTORIZED (...)) + which is folded by vectorizer. Both outgoing edges must have probability + 100% so the resulting profile match after folding. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of outgoing probabilities 200.0" 1 "ifcvt" } } */ +/* Sum is wrong here, but not enough for error to be reported. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of incoming frequencies" 0 "ifcvt" } } */ + diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ifc-20040816-1.c b/gcc/testsuite/gcc.dg/tree-ssa/ifc-20040816-1.c index 7a8e7a5a796..8badc762267 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/ifc-20040816-1.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/ifc-20040816-1.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-c -O2 -ftree-vectorize -fdump-tree-ifcvt-stats" { target *-*-* } } */ +/* { dg-options "-c -O2 -ftree-vectorize -fdump-tree-ifcvt-stats-blocks-details" { target *-*-* } } */ #include @@ -33,3 +33,10 @@ int main1 () /* { dg-final { scan-tree-dump-times "Applying if-conversion" 1 "ifcvt" } } */ + +/* We insert into code + if (LOOP_VECTORIZED (...)) + which is folded by vectorizer. Both outgoing edges must have probability + 100% so the resulting profile match after folding. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of outgoing probabilities 200.0" 1 "ifcvt" } } */ +/* { dg-final { scan-tree-dump-times "Invalid sum of incoming frequencies" 1 "ifcvt" } } */ diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ifc-20040816-2.c b/gcc/testsuite/gcc.dg/tree-ssa/ifc-20040816-2.c index b084e7b10a8..a517f6552e6 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/ifc-20040816-2.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/ifc-20040816-2.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-c -O2 -ftree-vectorize -fdump-tree-ifcvt-stats" { target *-*-* } } */ +/* { dg-options "-c -O2 -ftree-vectorize -fdump-tree-ifcvt-stats-blocks-details" { target *-*-* } } */ void foo(const int * __restrict__ zr_in, const int * __restrict__ zi_in, @@ -37,3 +37,11 @@ void foo(const int * __restrict__ zr_in, /* { dg-final { scan-tree-dump-times "Applying if-conversion" 1 "ifcvt" } } */ + +/* We insert into code + if (LOOP_VECTORIZED (...)) + which is folded by vectorizer. Both outgoing edges must have probability + 100% so the resulting profile match after folding. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of outgoing probabilities 200.0" 1 "ifcvt" } } */ +/* Sum is wrong here, but not enough for error to be reported. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of incoming frequencies" 0 "ifcvt" } } */ diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ifc-5.c b/gcc/testsuite/gcc.dg/tree-ssa/ifc-5.c index 875d2d3541f..58260dd878b 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/ifc-5.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/ifc-5.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-c -O2 -ftree-vectorize -fdump-tree-ifcvt-stats" { target *-*-* } } */ +/* { dg-options "-c -O2 -ftree-vectorize -fdump-tree-ifcvt-stats-blocks-details" { target *-*-* } } */ void dct_unquantize_h263_inter_c (short *block, int n, int qscale, int nCoeffs) @@ -21,3 +21,10 @@ dct_unquantize_h263_inter_c (short *block, int n, int qscale, int nCoeffs) } /* { dg-final { scan-tree-dump-times "Applying if-conversion" 1 "ifcvt" } } */ + +/* We insert into code + if (LOOP_VECTORIZED (...)) + which is folded by vectorizer. Both outgoing edges must have probability + 100% so the resulting profile match after folding. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of outgoing probabilities 200.0" 1 "ifcvt" } } */ +/* { dg-final { scan-tree-dump-times "Invalid sum of incoming frequencies" 1 "ifcvt" } } */ diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ifc-8.c b/gcc/testsuite/gcc.dg/tree-ssa/ifc-8.c index 81a4075cc29..6c26c209212 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/ifc-8.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/ifc-8.c @@ -1,6 +1,6 @@ /* { dg-do compile } */ -/* { dg-options "-Ofast -fdump-tree-ifcvt-details -ftree-loop-if-convert-stores" } */ +/* { dg-options "-Ofast -fdump-tree-ifcvt-details-blocks -ftree-loop-if-convert-stores" } */ /* { dg-require-visibility "" } */ #define LEN 4096 @@ -16,3 +16,11 @@ void test () } /* { dg-final { scan-tree-dump-times "Applying if-conversion" 1 "ifcvt" } } */ + +/* We insert into code + if (LOOP_VECTORIZED (...)) + which is folded by vectorizer. Both outgoing edges must have probability + 100% so the resulting profile match after folding. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of outgoing probabilities 200.0" 1 "ifcvt" } } */ +/* Sum is wrong here, but not enough for error to be reported. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of incoming frequencies" 0 "ifcvt" } } */ diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ifc-9.c b/gcc/testsuite/gcc.dg/tree-ssa/ifc-9.c index 338ea2545d2..789cb6ae23a 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/ifc-9.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/ifc-9.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-Ofast -fno-split-loops -fdump-tree-ifcvt-stats" } */ +/* { dg-options "-Ofast -fno-split-loops -fdump-tree-ifcvt-stats-blocks-details" } */ /* { dg-require-visibility "" } */ extern int b[256], y; @@ -20,3 +20,10 @@ int foo (int x, int n) } /* { dg-final { scan-tree-dump-times "Applying if-conversion" 1 "ifcvt" } } */ + +/* We insert into code + if (LOOP_VECTORIZED (...)) + which is folded by vectorizer. Both outgoing edges must have probability + 100% so the resulting profile match after folding. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of outgoing probabilities 200.0" 1 "ifcvt" } } */ +/* { dg-final { scan-tree-dump-times "Invalid sum of incoming frequencies" 1 "ifcvt" } } */ diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ifc-cd.c b/gcc/testsuite/gcc.dg/tree-ssa/ifc-cd.c index 29e2559e42b..11e142af321 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/ifc-cd.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/ifc-cd.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-O3 -fdump-tree-ifcvt-details -ftree-loop-if-convert-stores" } */ +/* { dg-options "-O3 -fdump-tree-ifcvt-details-blocks-details -ftree-loop-if-convert-stores" } */ void foo (int *x1, int *x2, int *x3, int *x4, int *y) { @@ -26,3 +26,11 @@ void foo (int *x1, int *x2, int *x3, int *x4, int *y) } /* { dg-final { scan-tree-dump-times "Use predicate of bb" 4 "ifcvt" } } */ + +/* We insert into code + if (LOOP_VECTORIZED (...)) + which is folded by vectorizer. Both outgoing edges must have probability + 100% so the resulting profile match after folding. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of outgoing probabilities 200.0" 1 "ifcvt" } } */ +/* Sum is wrong here, but not enough for error to be reported. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of incoming frequencies" 0 "ifcvt" } } */ diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr56541.c b/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr56541.c index a52bacacedf..9682fbc15df 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr56541.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr56541.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-O3 -fdump-tree-ifcvt-stats" } */ +/* { dg-options "-O3 -fdump-tree-ifcvt-stats-blocks-details" } */ float a,b,c,d; @@ -23,3 +23,11 @@ void foo() } /* { dg-final { scan-tree-dump-times "Applying if-conversion" 1 "ifcvt" } } */ + +/* We insert into code + if (LOOP_VECTORIZED (...)) + which is folded by vectorizer. Both outgoing edges must have probability + 100% so the resulting profile match after folding. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of outgoing probabilities 200.0" 1 "ifcvt" } } */ +/* Sum is wrong here, but not enough for error to be reported. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of incoming frequencies" 0 "ifcvt" } } */ diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr68583.c b/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr68583.c index be89e0855c3..b128deb4a21 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr68583.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr68583.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-O3 -fdump-tree-ifcvt" } */ +/* { dg-options "-O3 -fdump-tree-ifcvt-blocks-details" } */ void foo (long *a) { @@ -21,3 +21,10 @@ void foo (long *a) } /* { dg-final { scan-tree-dump "Applying if-conversion" "ifcvt" } } */ +/* We insert into code + if (LOOP_VECTORIZED (...)) + which is folded by vectorizer. Both outgoing edges must have probability + 100% so the resulting profile match after folding. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of outgoing probabilities 200.0" 1 "ifcvt" } } */ +/* { dg-final { scan-tree-dump-times "Invalid sum of incoming frequencies" 1 "ifcvt" } } */ + diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr69489-1.c b/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr69489-1.c index 4131afcbdff..3ba7de5e6a5 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr69489-1.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr69489-1.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-O2 -ftree-vectorize -fdump-tree-ifcvt-stats" { target *-*-* } } */ +/* { dg-options "-O2 -ftree-vectorize -fdump-tree-ifcvt-stats-blocks-details" { target *-*-* } } */ void foo (int a[], int b[]) { @@ -14,3 +14,11 @@ void foo (int a[], int b[]) } /* { dg-final { scan-tree-dump-times "Applying if-conversion" 1 "ifcvt" } } */ + +/* We insert into code + if (LOOP_VECTORIZED (...)) + which is folded by vectorizer. Both outgoing edges must have probability + 100% so the resulting profile match after folding. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of outgoing probabilities 200.0" 1 "ifcvt" } } */ +/* Sum is wrong here, but not enough for error to be reported. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of incoming frequencies" 0 "ifcvt" } } */ diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr69489-2.c b/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr69489-2.c index 6661f1a6c1b..07589fd7928 100644 --- a/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr69489-2.c +++ b/gcc/testsuite/gcc.dg/tree-ssa/ifc-pr69489-2.c @@ -1,5 +1,5 @@ /* { dg-do compile } */ -/* { dg-options "-S -O2 -ftree-vectorize -fdump-tree-ifcvt-stats" { target *-*-* } } */ +/* { dg-options "-S -O2 -ftree-vectorize -fdump-tree-ifcvt-stats-blocks-details" { target *-*-* } } */ double foo (const char *u, const char *v, long n) @@ -15,3 +15,10 @@ foo (const char *u, const char *v, long n) } /* { dg-final { scan-tree-dump-times "Applying if-conversion" 1 "ifcvt" } } */ + +/* We insert into code + if (LOOP_VECTORIZED (...)) + which is folded by vectorizer. Both outgoing edges must have probability + 100% so the resulting profile match after folding. */ +/* { dg-final { scan-tree-dump-times "Invalid sum of outgoing probabilities 200.0" 1 "ifcvt" } } */ +/* { dg-final { scan-tree-dump-times "Invalid sum of incoming frequencies" 1 "ifcvt" } } */ diff --git a/gcc/tree-if-conv.c b/gcc/tree-if-conv.c index c12987e55de..eb637952e22 100644 --- a/gcc/tree-if-conv.c +++ b/gcc/tree-if-conv.c @@ -2533,7 +2533,11 @@ combine_blocks (struct loop *loop) will be if-converted, the new copy of the loop will not, and the LOOP_VECTORIZED internal call will be guarding which loop to execute. The vectorizer pass will fold this - internal call into either true or false. */ + internal call into either true or false. + + Note that this function intentionally invalidates profile. Both edges + out of LOOP_VECTORIZED must have 100% probability so the profile remains + consistent after the condition is folded in the vectorizer. */ static struct loop * version_loop_for_if_conversion (struct loop *loop) @@ -2557,9 +2561,11 @@ version_loop_for_if_conversion (struct loop *loop) saved_preds[i] = ifc_bbs[i]->aux; initialize_original_copy_tables (); + /* At this point we invalidate porfile confistency until IFN_LOOP_VECTORIZED + is re-merged in the vectorizer. */ new_loop = loop_version (loop, cond, &cond_bb, REG_BR_PROB_BASE, REG_BR_PROB_BASE, - REG_BR_PROB_BASE, true); + REG_BR_PROB_BASE, REG_BR_PROB_BASE, true); free_original_copy_tables (); for (unsigned i = 0; i < save_length; i++) diff --git a/gcc/tree-parloops.c b/gcc/tree-parloops.c index 83aa8308def..739301140f0 100644 --- a/gcc/tree-parloops.c +++ b/gcc/tree-parloops.c @@ -2353,7 +2353,8 @@ gen_parallel_loop (struct loop *loop, /* We assume that the loop usually iterates a lot. */ prob = 4 * REG_BR_PROB_BASE / 5; loop_version (loop, many_iterations_cond, NULL, - prob, prob, REG_BR_PROB_BASE - prob, true); + prob, REG_BR_PROB_BASE - prob, + prob, REG_BR_PROB_BASE - prob, true); update_ssa (TODO_update_ssa); free_original_copy_tables (); } diff --git a/gcc/tree-ssa-loop-manip.c b/gcc/tree-ssa-loop-manip.c index 39fb48454ae..43df29c5415 100644 --- a/gcc/tree-ssa-loop-manip.c +++ b/gcc/tree-ssa-loop-manip.c @@ -1202,7 +1202,8 @@ tree_transform_and_unroll_loop (struct loop *loop, unsigned factor, scale_rest = REG_BR_PROB_BASE; new_loop = loop_version (loop, enter_main_cond, NULL, - prob_entry, scale_unrolled, scale_rest, true); + prob_entry, REG_BR_PROB_BASE - prob_entry, + scale_unrolled, scale_rest, true); gcc_assert (new_loop != NULL); update_ssa (TODO_update_ssa); diff --git a/gcc/tree-ssa-loop-split.c b/gcc/tree-ssa-loop-split.c index f104b2621c0..39bffc4ef17 100644 --- a/gcc/tree-ssa-loop-split.c +++ b/gcc/tree-ssa-loop-split.c @@ -562,7 +562,8 @@ split_loop (struct loop *loop1, struct tree_niter_desc *niter) basic_block cond_bb; struct loop *loop2 = loop_version (loop1, cond, &cond_bb, REG_BR_PROB_BASE, REG_BR_PROB_BASE, - REG_BR_PROB_BASE, true); + REG_BR_PROB_BASE, REG_BR_PROB_BASE, + true); gcc_assert (loop2); update_ssa (TODO_update_ssa); diff --git a/gcc/tree-ssa-loop-unswitch.c b/gcc/tree-ssa-loop-unswitch.c index 4ef3a6bf80a..1d0305a2a8b 100644 --- a/gcc/tree-ssa-loop-unswitch.c +++ b/gcc/tree-ssa-loop-unswitch.c @@ -493,7 +493,7 @@ tree_unswitch_loop (struct loop *loop, extract_true_false_edges_from_block (unswitch_on, &edge_true, &edge_false); prob_true = edge_true->probability; return loop_version (loop, unshare_expr (cond), - NULL, prob_true, prob_true, + NULL, prob_true, REG_BR_PROB_BASE - prob_true, prob_true, REG_BR_PROB_BASE - prob_true, false); } diff --git a/gcc/tree-vect-loop-manip.c b/gcc/tree-vect-loop-manip.c index 935cd75e554..f29449c37fd 100644 --- a/gcc/tree-vect-loop-manip.c +++ b/gcc/tree-vect-loop-manip.c @@ -2319,7 +2319,8 @@ vect_loop_versioning (loop_vec_info loop_vinfo, /* We don't want to scale SCALAR_LOOP's frequencies, we need to scale LOOP's frequencies instead. */ - nloop = loop_version (scalar_loop, cond_expr, &condition_bb, prob, + nloop = loop_version (scalar_loop, cond_expr, &condition_bb, + prob, REG_BR_PROB_BASE - prob, REG_BR_PROB_BASE, REG_BR_PROB_BASE - prob, true); scale_loop_frequencies (loop, prob, REG_BR_PROB_BASE); /* CONDITION_BB was created above SCALAR_LOOP's preheader, @@ -2348,7 +2349,8 @@ vect_loop_versioning (loop_vec_info loop_vinfo, } else nloop = loop_version (loop, cond_expr, &condition_bb, - prob, prob, REG_BR_PROB_BASE - prob, true); + prob, REG_BR_PROB_BASE - prob, + prob, REG_BR_PROB_BASE - prob, true); if (version_niter) { -- 2.30.2