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

S/390: Define vectorization_cost hook

Browse Source 
Define the vectorization_cost hook.  The only change right now
compared to the default implementation is the reduced costs for
unaligned loads/stores.  This is supposed to prevent unnecessary loop
peeling performed to reach better alignments.

Further tuning of this hook is required.

-Andreas-

gcc/ChangeLog:

2016-12-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>

	* gcc/config/s390/s390.c (s390_builtin_vectorization_cost): New
	function.
	(TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST): Define target
	macro.

gcc/testsuite/ChangeLog:

2016-12-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>

	* gcc.target/s390/vector/vec-nopeel-1.c: New test.

From-SVN: r243158
This commit is contained in:
Andreas Krebbel 2016-12-02 08:25:27 +00:00 committed by Andreas Krebbel
parent 9ad49cdb53
commit 7f5fc63362
4 changed files with 65 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
gcc/ChangeLog
View File

@ -1,3 +1,10 @@
2016-12-02 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
* gcc/config/s390/s390.c (s390_builtin_vectorization_cost): New
function.
(TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST): Define target
macro.
2016-12-02 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
* config/s390/vector.md (vec_halfhalf): New mode iterator.

37
gcc/config/s390/s390.c
View File

@ -3674,6 +3674,40 @@ s390_address_cost (rtx addr, machine_mode mode ATTRIBUTE_UNUSED,
return ad.indx? COSTS_N_INSNS (1) + 1 : COSTS_N_INSNS (1);
}
/* Implement targetm.vectorize.builtin_vectorization_cost. */
static int
s390_builtin_vectorization_cost (enum vect_cost_for_stmt type_of_cost,
tree vectype,
int misalign ATTRIBUTE_UNUSED)
{
switch (type_of_cost)
{
case scalar_stmt:
case scalar_load:
case scalar_store:
case vector_stmt:
case vector_load:
case vector_store:
case vec_to_scalar:
case scalar_to_vec:
case cond_branch_not_taken:
case vec_perm:
case vec_promote_demote:
case unaligned_load:
case unaligned_store:
return 1;
case cond_branch_taken:
return 3;
case vec_construct:
return TYPE_VECTOR_SUBPARTS (vectype) - 1;
default:
gcc_unreachable ();
}
}
/* If OP is a SYMBOL_REF of a thread-local symbol, return its TLS mode,
otherwise return 0. */
@ -15428,6 +15462,9 @@ s390_excess_precision (enum excess_precision_type type)
#define TARGET_REGISTER_MOVE_COST s390_register_move_cost
#undef TARGET_MEMORY_MOVE_COST
#define TARGET_MEMORY_MOVE_COST s390_memory_move_cost
#undef TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST
#define TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST \
s390_builtin_vectorization_cost
#undef TARGET_MACHINE_DEPENDENT_REORG
#define TARGET_MACHINE_DEPENDENT_REORG s390_reorg

4
gcc/testsuite/ChangeLog
View File

@ -1,3 +1,7 @@
2016-12-02 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
* gcc.target/s390/vector/vec-nopeel-1.c: New test.
2016-12-02 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
* gcc.target/s390/zvector/vec-cmp-2.c: New test.

17
gcc/testsuite/gcc.target/s390/vector/vec-nopeel-1.c Normal file
View File

@ -0,0 +1,17 @@
/* { dg-do compile } */
/* { dg-options "-O3 -mzarch -march=z13" } */
/* { dg-require-effective-target vector } */
int
foo (int * restrict a, int n)
{
int i, result = 0;
for (i = 0; i < n * 4; i++)
result += a[i];
return result;
}
/* We do NOT want this loop to get peeled. Without peeling no scalar
memory add should appear. */
/* { dg-final { scan-assembler-not "\ta\t" } } */