OpenE2K/gcc
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Make the vectoriser drop to strided accesses for stores with gaps

Browse Source 
We could vectorise:

     for (...)
       {
         a[0] = ...;
         a[1] = ...;
         a[2] = ...;
         a[3] = ...;
         a += stride;
       }

(including the case when stride == 8) but not:

     for (...)
       {
         a[0] = ...;
         a[1] = ...;
         a[2] = ...;
         a[3] = ...;
         a += 8;
       }

(where the stride is always 8).  The former was treated as a "grouped
and strided" store, while the latter was treated as a grouped store
with gaps, which we don't support.

This patch makes us treat groups of stores with gaps at the end as
strided groups too.  I tried to go through all uses of STMT_VINFO_STRIDED_P
and all vector uses of DR_STEP to see whether there were any hard-baked
assumptions, but couldn't see any.  I wondered whether we should relax:

  /* We do not have to consider dependences between accesses that belong
     to the same group, unless the stride could be smaller than the
     group size.  */
  if (DR_GROUP_FIRST_ELEMENT (stmtinfo_a)
      && (DR_GROUP_FIRST_ELEMENT (stmtinfo_a)
          == DR_GROUP_FIRST_ELEMENT (stmtinfo_b))
      && !STMT_VINFO_STRIDED_P (stmtinfo_a))
    return false;

for cases in which the step is constant and the absolute step is known
to be greater than the group size, but data dependence analysis should
already return chrec_known for those cases.

The new test is a version of vect-avg-15.c with the variable step
replaced by a constant one.

A natural follow-on would be to do the same for groups with gaps in
the middle:

          /* Check that the distance between two accesses is equal to the type
             size. Otherwise, we have gaps.  */
          diff = (TREE_INT_CST_LOW (DR_INIT (data_ref))
                  - TREE_INT_CST_LOW (prev_init)) / type_size;
          if (diff != 1)
            {
              [...]
              if (DR_IS_WRITE (data_ref))
                {
                  if (dump_enabled_p ())
                    dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
                                     "interleaved store with gaps\n");
                  return false;
                }

But I think we should do that separately and see what the fallout
from this change is first.

2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
	* tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
	grouped stores with gaps to a strided group.

gcc/testsuite/
	* gcc.dg/vect/vect-avg-16.c: New test.
	* gcc.dg/vect/slp-37.c: Expect the loop to be vectorized.
	* gcc.dg/vect/vect-strided-u8-i8-gap4.c,
	* gcc.dg/vect/vect-strided-u8-i8-gap4-big-array.c: Likewise for
	the second loop in main1.

From-SVN: r263772
This commit is contained in:
Richard Sandiford 2018-08-22 12:59:08 +00:00 committed by Richard Sandiford
parent 3ad3b3ac8c
commit 203942b8af
7 changed files with 77 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
gcc/ChangeLog
View File

@ -1,3 +1,8 @@
2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
* tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
grouped stores with gaps to a strided group.
2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
* tree-vect-stmts.c (get_group_load_store_type)

8
gcc/testsuite/ChangeLog
View File

@ -1,3 +1,11 @@
2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
* gcc.dg/vect/vect-avg-16.c: New test.
* gcc.dg/vect/slp-37.c: Expect the loop to be vectorized.
* gcc.dg/vect/vect-strided-u8-i8-gap4.c,
* gcc.dg/vect/vect-strided-u8-i8-gap4-big-array.c: Likewise for
the second loop in main1.
2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
* gcc.dg/lto/pr85248_0.c (test_alias):

9
gcc/testsuite/gcc.dg/vect/slp-37.c
View File

@ -17,8 +17,8 @@ foo1 (s1 *arr)
int i;
s1 *ptr = arr;
/* Different constant types - not SLPable. The group size is not power of 2,
interleaving is not supported either. */
/* Vectorized as a strided SLP pair of accesses to <a, b> and a single
strided access to c. */
for (i = 0; i < N; i++)
{
ptr->a = 6;
@ -58,6 +58,5 @@ int main (void)
return 0;
}
/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 0 "vect" } } */
/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 "vect" } } */
/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" } } */
/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "vect" } } */

52
gcc/testsuite/gcc.dg/vect/vect-avg-16.c Normal file
View File

@ -0,0 +1,52 @@
/* { dg-additional-options "-O3" } */
/* { dg-require-effective-target vect_int } */
#include "tree-vect.h"
#define N 80
void __attribute__ ((noipa))
f (signed char *restrict a, signed char *restrict b,
signed char *restrict c, int n)
{
for (int j = 0; j < n; ++j)
{
for (int i = 0; i < 16; ++i)
a[i] = (b[i] + c[i]) >> 1;
a += 20;
b += 20;
c += 20;
}
}
#define BASE1 -126
#define BASE2 -42
signed char a[N], b[N], c[N];
int
main (void)
{
check_vect ();
for (int i = 0; i < N; ++i)
{
a[i] = i;
b[i] = BASE1 + i * 3;
c[i] = BASE2 + i * 2;
asm volatile ("" ::: "memory");
}
f (a, b, c, N / 20);
for (int i = 0; i < N; ++i)
{
int d = (BASE1 + BASE2 + i * 5) >> 1;
if (a[i] != (i % 20 < 16 ? d : i))
__builtin_abort ();
}
return 0;
}
/* { dg-final { scan-tree-dump "vect_recog_average_pattern: detected" "vect" } } */
/* { dg-final { scan-tree-dump {\.AVG_FLOOR} "vect" { target vect_avg_qi } } } */
/* { dg-final { scan-tree-dump "Loop contains only SLP stmts" "vect" { target vect_avg_qi } } } */
/* { dg-final { scan-tree-dump-times "vectorized 1 loop" 1 "vect" { target vect_avg_qi } } } */

5
gcc/testsuite/gcc.dg/vect/vect-strided-u8-i8-gap4-big-array.c
View File

@ -55,7 +55,7 @@ main1 (s *arr)
}
ptr = arr;
/* Not vectorizable: gap in store. */
/* Vectorized as a strided SLP pair. */
for (i = 0; i < N; i++)
{
res[i].a = ptr->b;
@ -110,5 +110,4 @@ int main (void)
return 0;
}
/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { target vect_strided8 } } } */
/* { dg-final { scan-tree-dump-times "vectorized 2 loops" 1 "vect" { target vect_strided8 } } } */

5
gcc/testsuite/gcc.dg/vect/vect-strided-u8-i8-gap4.c
View File

@ -53,7 +53,7 @@ main1 (s *arr)
}
ptr = arr;
/* Not vectorizable: gap in store. */
/* Vectorized as a strided SLP pair. */
for (i = 0; i < N; i++)
{
res[i].a = ptr->b;
@ -97,5 +97,4 @@ int main (void)
return 0;
}
/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { target vect_strided8 } } } */
/* { dg-final { scan-tree-dump-times "vectorized 2 loops" 1 "vect" { target vect_strided8 } } } */

8
gcc/tree-vect-data-refs.c
View File

@ -2633,10 +2633,8 @@ vect_analyze_group_access_1 (dr_vec_info *dr_info)
if (groupsize != count
&& !DR_IS_READ (dr))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"interleaved store with gaps\n");
return false;
groupsize = count;
STMT_VINFO_STRIDED_P (stmt_info) = true;
}
/* If there is a gap after the last load in the group it is the
@ -2652,6 +2650,8 @@ vect_analyze_group_access_1 (dr_vec_info *dr_info)
"Detected interleaving ");
if (DR_IS_READ (dr))
dump_printf (MSG_NOTE, "load ");
else if (STMT_VINFO_STRIDED_P (stmt_info))
dump_printf (MSG_NOTE, "strided store ");
else
dump_printf (MSG_NOTE, "store ");
dump_printf (MSG_NOTE, "of size %u starting with ",