Skip to content

R
riscv-gcc-1
Commit 5417e022 by Zdenek Dvorak
Options

tree-vectorizer.h (DR_MISALIGNMENT): Cast aux to integer. 

	* tree-vectorizer.h (DR_MISALIGNMENT): Cast aux to integer.
	(SET_DR_MISALIGNMENT): New.
	* tree-vect-analyze.c (vect_compute_data_ref_alignment,
	vect_update_misalignment_for_peel, vect_enhance_data_refs_alignment):
	Use SET_DR_MISALIGNMENT.
	* tree-predcom.c (split_data_refs_to_components): Cast dr->aux from
	pointer.
	* tree-data-ref.c (create_data_ref, compute_all_dependences,
	find_loop_nest): Export.
	* tree-data-ref.h (struct data_reference): Change aux field to pointer.
	(create_data_ref, compute_all_dependences, find_loop_nest): Declare.
	* tree-ssa-loop-prefetch.c: Include tree-data-ref.h.
	(L1_CACHE_SIZE_BYTES, L2_CACHE_SIZE_BYTES, NONTEMPORAL_FRACTION):
	New macros.
	(struct mem_ref): Add field reuse_distance.
	(find_or_create_group, record_ref): Use XNEW instead of xcalloc.
	Initialize reuse_distance field.
	(issue_prefetch_ref): Select temporality of prefetch according to
	reuse_distance.
	(volume_of_references, volume_of_dist_vector, add_subscript_strides,
	self_reuse_distance, determine_loop_nest_reuse): New functions.
	(loop_prefetch_arrays): Call determine_loop_nest_reuse.
	(tree_ssa_prefetch_arrays): Dump L2 cache size.
	* Makefile.in (tree-ssa-loop-prefetch.o): Add TREE_DATA_REF_H
	dependency.

	* gcc.dg/tree-ssa/prefetch-6.c: New test.

From-SVN: r125172
parent cd5ecab6
Showing with 114 additions and 22 deletions
gcc/ChangeLog
2007-05-29 Zdenek Dvorak <dvorakz@suse.cz>
* tree-vectorizer.h (DR_MISALIGNMENT): Cast aux to integer.
(SET_DR_MISALIGNMENT): New.
* tree-vect-analyze.c (vect_compute_data_ref_alignment,
vect_update_misalignment_for_peel, vect_enhance_data_refs_alignment):
Use SET_DR_MISALIGNMENT.
* tree-predcom.c (split_data_refs_to_components): Cast dr->aux from
pointer.
* tree-data-ref.c (create_data_ref, compute_all_dependences,
find_loop_nest): Export.
* tree-data-ref.h (struct data_reference): Change aux field to pointer.
(create_data_ref, compute_all_dependences, find_loop_nest): Declare.
* tree-ssa-loop-prefetch.c: Include tree-data-ref.h.
(L1_CACHE_SIZE_BYTES, L2_CACHE_SIZE_BYTES, NONTEMPORAL_FRACTION):
New macros.
(struct mem_ref): Add field reuse_distance.
(find_or_create_group, record_ref): Use XNEW instead of xcalloc.
Initialize reuse_distance field.
(issue_prefetch_ref): Select temporality of prefetch according to
reuse_distance.
(volume_of_references, volume_of_dist_vector, add_subscript_strides,
self_reuse_distance, determine_loop_nest_reuse): New functions.
(loop_prefetch_arrays): Call determine_loop_nest_reuse.
(tree_ssa_prefetch_arrays): Dump L2 cache size.
* Makefile.in (tree-ssa-loop-prefetch.o): Add TREE_DATA_REF_H
dependency.
2007-05-29 Daniel Berlin <dberlin@dberlin.org>
* tree-ssa-alias.c: Add aliasing overview.
2007-05-29 Zuxy Meng <zuxy.meng@gmail.com>
Danny Smith <dannysmith@users.sourceforge.net>
Danny Smith <dannysmith@users.sourceforge.net>
PR target/29498
* config/i386/t-crtfm: Compile crtfastmath.o with
......
gcc/Makefile.in
......@@ -2084,7 +2084,7 @@ tree-ssa-loop-prefetch.o: tree-ssa-loop-prefetch.c $(TREE_FLOW_H) $(CONFIG_H) \
output.h $(DIAGNOSTIC_H) $(TIMEVAR_H) $(TM_H) coretypes.h $(TREE_DUMP_H) \
tree-pass.h $(GGC_H) $(RECOG_H) insn-config.h $(HASHTAB_H) $(SCEV_H) \
$(CFGLOOP_H) $(PARAMS_H) langhooks.h $(BASIC_BLOCK_H) hard-reg-set.h \
tree-chrec.h toplev.h langhooks.h $(TREE_INLINE_H)
tree-chrec.h toplev.h langhooks.h $(TREE_INLINE_H) $(TREE_DATA_REF_H)
tree-predcom.o: tree-predcom.c $(CONFIG_H) $(SYSTEM_H) $(TREE_H) $(TM_P_H) \
$(CFGLOOP_H) $(TREE_FLOW_H) $(GGC_H) $(TREE_DATA_REF_H) $(SCEV_H) \
$(PARAMS_H) $(DIAGNOSTIC_H) tree-pass.h $(TM_H) coretypes.h tree-affine.h \
......
gcc/testsuite/ChangeLog
2007-05-29 Zdenek Dvorak <dvorakz@suse.cz>
* gcc.dg/tree-ssa/prefetch-6.c: New test.
2007-05-29 Tobias Schlter <tobi@gcc.gnu.org>
* gfortran.dg/sizeof.f90: New.
gcc/testsuite/gcc.dg/tree-ssa/prefetch-6.c 0 → 100644
/* { dg-do compile { target i?86-*-* x86_64-*-* } } */
/* { dg-require-effective-target ilp32 } */
/* { dg-options "-O2 -fprefetch-loop-arrays -march=athlon -msse2 -mfpmath=sse --param simultaneous-prefetches=100 -fdump-tree-aprefetch-details" } */
#define N 1000
#define K 900
double a[N][N];
double test(void)
{
unsigned i, j;
double sum = 0;
/* Here, we should use non-temporal prefetch instruction. */
for (i = 0; i < K; i++)
for (j = 0; j < K; j++)
sum += a[i][j];
/* Here, we should not use non-temporal prefetch instruction, since the
value of a[i+10][j] is reused in L2 cache. */
for (i = 0; i < K; i++)
for (j = 0; j < K; j++)
sum += a[i][j] * a[i + 10][j];
/* Here, we should use non-temporal prefetch instruction, since the
value of a[i+100][j] is too far to be reused in L2 cache. */
for (i = 0; i < K; i++)
for (j = 0; j < K; j++)
sum += a[i][j] * a[i + 100][j];
/* Here, temporal prefetches should be used, since the volume of the
memory accesses is smaller than L2 cache. */
for (i = 0; i < 100; i++)
for (j = 0; j < 100; j++)
sum += a[i][j] * a[i + 100][j];
/* Temporal prefetches should be used here (even though the accesses to
a[j][i] are independent, the same cache line is almost always hit
every N iterations). */
for (i = 0; i < N; i++)
for (j = 0; j < N; j++)
sum += a[j][i];
return sum;
}
/* { dg-final { scan-tree-dump-times "Issued prefetch" 5 "aprefetch" } } */
/* { dg-final { scan-tree-dump-times "Issued nontemporal prefetch" 3 "aprefetch" } } */
/* { dg-final { scan-assembler-times "prefetcht" 5 } } */
/* { dg-final { scan-assembler-times "prefetchnta" 3 } } */
/* { dg-final { cleanup-tree-dump "aprefetch" } } */
gcc/tree-data-ref.c
......@@ -771,7 +771,7 @@ free_data_ref (data_reference_p dr)
data_reference description of MEMREF. NEST is the outermost loop of the
loop nest in that the reference should be analysed. */
static struct data_reference *
struct data_reference *
create_data_ref (struct loop *nest, tree memref, tree stmt, bool is_read)
{
struct data_reference *dr;
......@@ -3843,7 +3843,7 @@ compute_self_dependence (struct data_dependence_relation *ddr)
COMPUTE_SELF_AND_RR is FALSE, don't compute read-read and self
relations. */
static void
void
compute_all_dependences (VEC (data_reference_p, heap) *datarefs,
VEC (ddr_p, heap) **dependence_relations,
VEC (loop_p, heap) *loop_nest,
......@@ -4055,7 +4055,7 @@ find_loop_nest_1 (struct loop *loop, VEC (loop_p, heap) **loop_nest)
contain the loops from the outermost to the innermost, as they will
appear in the classic distance vector. */
static bool
bool
find_loop_nest (struct loop *loop, VEC (loop_p, heap) **loop_nest)
{
VEC_safe_push (loop_p, heap, *loop_nest, loop);
......
gcc/tree-data-ref.h
......@@ -104,7 +104,7 @@ struct data_reference
tree ref;
/* Auxiliary info specific to a pass. */
int aux;
void *aux;
/* True when the data reference is in RHS of a stmt. */
bool is_read;
......@@ -320,7 +320,10 @@ extern void dump_data_dependence_direction (FILE *,
extern void free_dependence_relation (struct data_dependence_relation *);
extern void free_dependence_relations (VEC (ddr_p, heap) *);
extern void free_data_refs (VEC (data_reference_p, heap) *);
struct data_reference *create_data_ref (struct loop *, tree, tree, bool);
bool find_loop_nest (struct loop *, VEC (loop_p, heap) **);
void compute_all_dependences (VEC (data_reference_p, heap) *,
VEC (ddr_p, heap) **, VEC (loop_p, heap) *, bool);
/* Return the index of the variable VAR in the LOOP_NEST array. */
......
gcc/tree-predcom.c
......@@ -700,7 +700,7 @@ split_data_refs_to_components (struct loop *loop,
just fail. */
goto end;
}
dr->aux = i;
dr->aux = (void *) (size_t) i;
comp_father[i] = i;
comp_size[i] = 1;
}
......@@ -715,7 +715,7 @@ split_data_refs_to_components (struct loop *loop,
if (!suitable_reference_p (dr, &dummy))
{
ia = dr->aux;
ia = (unsigned) (size_t) dr->aux;
merge_comps (comp_father, comp_size, n, ia);
}
}
......@@ -729,8 +729,8 @@ split_data_refs_to_components (struct loop *loop,
dra = DDR_A (ddr);
drb = DDR_B (ddr);
ia = component_of (comp_father, dra->aux);
ib = component_of (comp_father, drb->aux);
ia = component_of (comp_father, (unsigned) (size_t) dra->aux);
ib = component_of (comp_father, (unsigned) (size_t) drb->aux);
if (ia == ib)
continue;
......@@ -749,7 +749,7 @@ split_data_refs_to_components (struct loop *loop,
bad = component_of (comp_father, n);
for (i = 0; VEC_iterate (data_reference_p, datarefs, i, dr); i++)
{
ia = dr->aux;
ia = (unsigned) (size_t) dr->aux;
ca = component_of (comp_father, ia);
if (ca == bad)
continue;
......
gcc/tree-vect-analyze.c
......@@ -1128,7 +1128,7 @@ vect_compute_data_ref_alignment (struct data_reference *dr)
fprintf (vect_dump, "vect_compute_data_ref_alignment:");
/* Initialize misalignment to unknown. */
DR_MISALIGNMENT (dr) = -1;
SET_DR_MISALIGNMENT (dr, -1);
misalign = DR_INIT (dr);
aligned_to = DR_ALIGNED_TO (dr);
......@@ -1198,7 +1198,7 @@ vect_compute_data_ref_alignment (struct data_reference *dr)
return false;
}
DR_MISALIGNMENT (dr) = TREE_INT_CST_LOW (misalign);
SET_DR_MISALIGNMENT (dr, TREE_INT_CST_LOW (misalign));
if (vect_print_dump_info (REPORT_DETAILS))
{
......@@ -1267,21 +1267,23 @@ vect_update_misalignment_for_peel (struct data_reference *dr,
continue;
gcc_assert (DR_MISALIGNMENT (dr) / dr_size ==
DR_MISALIGNMENT (dr_peel) / dr_peel_size);
DR_MISALIGNMENT (dr) = 0;
SET_DR_MISALIGNMENT (dr, 0);
return;
}
if (known_alignment_for_access_p (dr)
&& known_alignment_for_access_p (dr_peel))
{
DR_MISALIGNMENT (dr) += npeel * dr_size;
DR_MISALIGNMENT (dr) %= UNITS_PER_SIMD_WORD;
int misal = DR_MISALIGNMENT (dr);
misal += npeel * dr_size;
misal %= UNITS_PER_SIMD_WORD;
SET_DR_MISALIGNMENT (dr, misal);
return;
}
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "Setting misalignment to -1.");
DR_MISALIGNMENT (dr) = -1;
SET_DR_MISALIGNMENT (dr, -1);
}
......@@ -1577,7 +1579,7 @@ vect_enhance_data_refs_alignment (loop_vec_info loop_vinfo)
save_misalignment = DR_MISALIGNMENT (dr);
vect_update_misalignment_for_peel (dr, dr0, npeel);
supportable_dr_alignment = vect_supportable_dr_alignment (dr);
DR_MISALIGNMENT (dr) = save_misalignment;
SET_DR_MISALIGNMENT (dr, save_misalignment);
if (!supportable_dr_alignment)
{
......@@ -1601,7 +1603,7 @@ vect_enhance_data_refs_alignment (loop_vec_info loop_vinfo)
LOOP_VINFO_UNALIGNED_DR (loop_vinfo) = dr0;
LOOP_PEELING_FOR_ALIGNMENT (loop_vinfo) = DR_MISALIGNMENT (dr0);
DR_MISALIGNMENT (dr0) = 0;
SET_DR_MISALIGNMENT (dr0, 0);
if (vect_print_dump_info (REPORT_ALIGNMENT))
fprintf (vect_dump, "Alignment of access forced using peeling.");
......@@ -1702,7 +1704,7 @@ vect_enhance_data_refs_alignment (loop_vec_info loop_vinfo)
{
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
dr = STMT_VINFO_DATA_REF (stmt_info);
DR_MISALIGNMENT (dr) = 0;
SET_DR_MISALIGNMENT (dr, 0);
if (vect_print_dump_info (REPORT_ALIGNMENT))
fprintf (vect_dump, "Alignment of access forced using versioning.");
}
......
gcc/tree-vectorizer.h
......@@ -339,7 +339,8 @@ is_pattern_stmt_p (stmt_vec_info stmt_info)
/* Reflects actual alignment of first access in the vectorized loop,
taking into account peeling/versioning if applied. */
#define DR_MISALIGNMENT(DR) (DR)->aux
#define DR_MISALIGNMENT(DR) ((int) (size_t) (DR)->aux)
#define SET_DR_MISALIGNMENT(DR, VAL) ((DR)->aux = (void *) (size_t) (VAL))
static inline bool
aligned_access_p (struct data_reference *data_ref_info)
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment