* [PATCH] Fold VEC_[LR]SHIFT_EXPR (PR tree-optimization/57051)
@ 2013-04-26 0:32 Jakub Jelinek
2013-04-26 11:06 ` Richard Biener
2013-05-03 13:18 ` Jakub Jelinek
0 siblings, 2 replies; 6+ messages in thread
From: Jakub Jelinek @ 2013-04-26 0:32 UTC (permalink / raw)
To: Richard Biener; +Cc: gcc-patches
Hi!
This patch adds folding of constant arguments v>> and v<<, which helps to
optimize the testcase from the PR back into constant store after vectorized
loop is unrolled.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2013-04-25 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/57051
* fold-const.c (const_binop): Handle VEC_LSHIFT_EXPR
and VEC_RSHIFT_EXPR if shift count is a multiple of element
bitsize.
--- gcc/fold-const.c.jj 2013-04-12 10:16:25.000000000 +0200
+++ gcc/fold-const.c 2013-04-24 12:37:11.789122719 +0200
@@ -1380,17 +1380,42 @@ const_binop (enum tree_code code, tree a
int count = TYPE_VECTOR_SUBPARTS (type), i;
tree *elts = XALLOCAVEC (tree, count);
- for (i = 0; i < count; i++)
+ if (code == VEC_LSHIFT_EXPR
+ || code == VEC_RSHIFT_EXPR)
{
- tree elem1 = VECTOR_CST_ELT (arg1, i);
-
- elts[i] = const_binop (code, elem1, arg2);
+ if (!host_integerp (arg2, 1))
+ return NULL_TREE;
- /* It is possible that const_binop cannot handle the given
- code and return NULL_TREE */
- if (elts[i] == NULL_TREE)
+ unsigned HOST_WIDE_INT shiftc = tree_low_cst (arg2, 1);
+ unsigned HOST_WIDE_INT outerc = tree_low_cst (TYPE_SIZE (type), 1);
+ unsigned HOST_WIDE_INT innerc
+ = tree_low_cst (TYPE_SIZE (TREE_TYPE (type)), 1);
+ if (shiftc >= outerc || (shiftc % innerc) != 0)
return NULL_TREE;
+ int offset = shiftc / innerc;
+ if (code == VEC_LSHIFT_EXPR)
+ offset = -offset;
+ tree zero = build_zero_cst (TREE_TYPE (type));
+ for (i = 0; i < count; i++)
+ {
+ if (i + offset < 0 || i + offset >= count)
+ elts[i] = zero;
+ else
+ elts[i] = VECTOR_CST_ELT (arg1, i + offset);
+ }
}
+ else
+ for (i = 0; i < count; i++)
+ {
+ tree elem1 = VECTOR_CST_ELT (arg1, i);
+
+ elts[i] = const_binop (code, elem1, arg2);
+
+ /* It is possible that const_binop cannot handle the given
+ code and return NULL_TREE */
+ if (elts[i] == NULL_TREE)
+ return NULL_TREE;
+ }
return build_vector (type, elts);
}
Jakub
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] Fold VEC_[LR]SHIFT_EXPR (PR tree-optimization/57051)
2013-04-26 0:32 [PATCH] Fold VEC_[LR]SHIFT_EXPR (PR tree-optimization/57051) Jakub Jelinek
@ 2013-04-26 11:06 ` Richard Biener
2013-05-03 13:18 ` Jakub Jelinek
1 sibling, 0 replies; 6+ messages in thread
From: Richard Biener @ 2013-04-26 11:06 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: gcc-patches
On Thu, 25 Apr 2013, Jakub Jelinek wrote:
> Hi!
>
> This patch adds folding of constant arguments v>> and v<<, which helps to
> optimize the testcase from the PR back into constant store after vectorized
> loop is unrolled.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
Ok.
Thanks,
Richard.
> 2013-04-25 Jakub Jelinek <jakub@redhat.com>
>
> PR tree-optimization/57051
> * fold-const.c (const_binop): Handle VEC_LSHIFT_EXPR
> and VEC_RSHIFT_EXPR if shift count is a multiple of element
> bitsize.
>
> --- gcc/fold-const.c.jj 2013-04-12 10:16:25.000000000 +0200
> +++ gcc/fold-const.c 2013-04-24 12:37:11.789122719 +0200
> @@ -1380,17 +1380,42 @@ const_binop (enum tree_code code, tree a
> int count = TYPE_VECTOR_SUBPARTS (type), i;
> tree *elts = XALLOCAVEC (tree, count);
>
> - for (i = 0; i < count; i++)
> + if (code == VEC_LSHIFT_EXPR
> + || code == VEC_RSHIFT_EXPR)
> {
> - tree elem1 = VECTOR_CST_ELT (arg1, i);
> -
> - elts[i] = const_binop (code, elem1, arg2);
> + if (!host_integerp (arg2, 1))
> + return NULL_TREE;
>
> - /* It is possible that const_binop cannot handle the given
> - code and return NULL_TREE */
> - if (elts[i] == NULL_TREE)
> + unsigned HOST_WIDE_INT shiftc = tree_low_cst (arg2, 1);
> + unsigned HOST_WIDE_INT outerc = tree_low_cst (TYPE_SIZE (type), 1);
> + unsigned HOST_WIDE_INT innerc
> + = tree_low_cst (TYPE_SIZE (TREE_TYPE (type)), 1);
> + if (shiftc >= outerc || (shiftc % innerc) != 0)
> return NULL_TREE;
> + int offset = shiftc / innerc;
> + if (code == VEC_LSHIFT_EXPR)
> + offset = -offset;
> + tree zero = build_zero_cst (TREE_TYPE (type));
> + for (i = 0; i < count; i++)
> + {
> + if (i + offset < 0 || i + offset >= count)
> + elts[i] = zero;
> + else
> + elts[i] = VECTOR_CST_ELT (arg1, i + offset);
> + }
> }
> + else
> + for (i = 0; i < count; i++)
> + {
> + tree elem1 = VECTOR_CST_ELT (arg1, i);
> +
> + elts[i] = const_binop (code, elem1, arg2);
> +
> + /* It is possible that const_binop cannot handle the given
> + code and return NULL_TREE */
> + if (elts[i] == NULL_TREE)
> + return NULL_TREE;
> + }
>
> return build_vector (type, elts);
> }
>
> Jakub
>
>
--
Richard Biener <rguenther@suse.de>
SUSE / SUSE Labs
SUSE LINUX Products GmbH - Nuernberg - AG Nuernberg - HRB 16746
GF: Jeff Hawn, Jennifer Guild, Felix Imend
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] Fold VEC_[LR]SHIFT_EXPR (PR tree-optimization/57051)
2013-04-26 0:32 [PATCH] Fold VEC_[LR]SHIFT_EXPR (PR tree-optimization/57051) Jakub Jelinek
2013-04-26 11:06 ` Richard Biener
@ 2013-05-03 13:18 ` Jakub Jelinek
2013-05-16 18:01 ` Mikael Pettersson
1 sibling, 1 reply; 6+ messages in thread
From: Jakub Jelinek @ 2013-05-03 13:18 UTC (permalink / raw)
To: Richard Biener; +Cc: gcc-patches
On Thu, Apr 25, 2013 at 11:47:02PM +0200, Jakub Jelinek wrote:
> This patch adds folding of constant arguments v>> and v<<, which helps to
> optimize the testcase from the PR back into constant store after vectorized
> loop is unrolled.
As this fixes a regression on the 4.8 branch, I've backported it (and
minimal prerequisite for that) to 4.8 branch too.
As the non-whole vector shifts VECTOR_CST by INTEGER_CST don't have any
testcase showing a regression, I've left those out (trunk has instead
of that else return NULL_TREE; code to handle those).
2013-05-03 Jakub Jelinek <jakub@redhat.com>
Backported from mainline
2013-04-26 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/57051
* fold-const.c (const_binop): Handle VEC_LSHIFT_EXPR
and VEC_RSHIFT_EXPR if shift count is a multiple of element
bitsize.
2013-04-12 Marc Glisse <marc.glisse@inria.fr>
* fold-const.c (fold_binary_loc): Call const_binop also for mixed
vector-scalar operations.
--- gcc/fold-const.c (revision 198579)
+++ gcc/fold-const.c (working copy)
@@ -1366,6 +1366,44 @@ const_binop (enum tree_code code, tree a
return build_vector (type, elts);
}
+
+ /* Shifts allow a scalar offset for a vector. */
+ if (TREE_CODE (arg1) == VECTOR_CST
+ && TREE_CODE (arg2) == INTEGER_CST)
+ {
+ tree type = TREE_TYPE (arg1);
+ int count = TYPE_VECTOR_SUBPARTS (type), i;
+ tree *elts = XALLOCAVEC (tree, count);
+
+ if (code == VEC_LSHIFT_EXPR
+ || code == VEC_RSHIFT_EXPR)
+ {
+ if (!host_integerp (arg2, 1))
+ return NULL_TREE;
+
+ unsigned HOST_WIDE_INT shiftc = tree_low_cst (arg2, 1);
+ unsigned HOST_WIDE_INT outerc = tree_low_cst (TYPE_SIZE (type), 1);
+ unsigned HOST_WIDE_INT innerc
+ = tree_low_cst (TYPE_SIZE (TREE_TYPE (type)), 1);
+ if (shiftc >= outerc || (shiftc % innerc) != 0)
+ return NULL_TREE;
+ int offset = shiftc / innerc;
+ if (code == VEC_LSHIFT_EXPR)
+ offset = -offset;
+ tree zero = build_zero_cst (TREE_TYPE (type));
+ for (i = 0; i < count; i++)
+ {
+ if (i + offset < 0 || i + offset >= count)
+ elts[i] = zero;
+ else
+ elts[i] = VECTOR_CST_ELT (arg1, i + offset);
+ }
+ }
+ else
+ return NULL_TREE;
+
+ return build_vector (type, elts);
+ }
return NULL_TREE;
}
@@ -9862,7 +9900,8 @@ fold_binary_loc (location_t loc,
|| (TREE_CODE (arg0) == FIXED_CST && TREE_CODE (arg1) == FIXED_CST)
|| (TREE_CODE (arg0) == FIXED_CST && TREE_CODE (arg1) == INTEGER_CST)
|| (TREE_CODE (arg0) == COMPLEX_CST && TREE_CODE (arg1) == COMPLEX_CST)
- || (TREE_CODE (arg0) == VECTOR_CST && TREE_CODE (arg1) == VECTOR_CST))
+ || (TREE_CODE (arg0) == VECTOR_CST && TREE_CODE (arg1) == VECTOR_CST)
+ || (TREE_CODE (arg0) == VECTOR_CST && TREE_CODE (arg1) == INTEGER_CST))
{
if (kind == tcc_binary)
{
Jakub
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] Fold VEC_[LR]SHIFT_EXPR (PR tree-optimization/57051)
2013-05-03 13:18 ` Jakub Jelinek
@ 2013-05-16 18:01 ` Mikael Pettersson
2013-05-17 6:48 ` [PATCH] Fix VEC_[LR]SHIFT_EXPR folding for big-endian " Jakub Jelinek
0 siblings, 1 reply; 6+ messages in thread
From: Mikael Pettersson @ 2013-05-16 18:01 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: Richard Biener, gcc-patches
Jakub Jelinek writes:
> On Thu, Apr 25, 2013 at 11:47:02PM +0200, Jakub Jelinek wrote:
> > This patch adds folding of constant arguments v>> and v<<, which helps to
> > optimize the testcase from the PR back into constant store after vectorized
> > loop is unrolled.
>
> As this fixes a regression on the 4.8 branch, I've backported it (and
> minimal prerequisite for that) to 4.8 branch too.
Unfortunately this patch makes gcc.dg/vect/no-scevccp-outer-{7,13}.c fail
on powerpc64-linux:
+FAIL: gcc.dg/vect/no-scevccp-outer-13.c execution test
+FAIL: gcc.dg/vect/no-scevccp-outer-7.c execution test
which is a regression from 4.8-20130502. Reverting r198580 fixes it.
The same FAILs also occur on trunk.
/Mikael
^ permalink raw reply [flat|nested] 6+ messages in thread* [PATCH] Fix VEC_[LR]SHIFT_EXPR folding for big-endian (PR tree-optimization/57051)
2013-05-16 18:01 ` Mikael Pettersson
@ 2013-05-17 6:48 ` Jakub Jelinek
2013-05-17 8:02 ` Richard Biener
0 siblings, 1 reply; 6+ messages in thread
From: Jakub Jelinek @ 2013-05-17 6:48 UTC (permalink / raw)
To: Richard Biener, Mikael Pettersson; +Cc: gcc-patches
On Thu, May 16, 2013 at 07:59:00PM +0200, Mikael Pettersson wrote:
> Jakub Jelinek writes:
> > On Thu, Apr 25, 2013 at 11:47:02PM +0200, Jakub Jelinek wrote:
> > > This patch adds folding of constant arguments v>> and v<<, which helps to
> > > optimize the testcase from the PR back into constant store after vectorized
> > > loop is unrolled.
> >
> > As this fixes a regression on the 4.8 branch, I've backported it (and
> > minimal prerequisite for that) to 4.8 branch too.
>
> Unfortunately this patch makes gcc.dg/vect/no-scevccp-outer-{7,13}.c fail
> on powerpc64-linux:
>
> +FAIL: gcc.dg/vect/no-scevccp-outer-13.c execution test
> +FAIL: gcc.dg/vect/no-scevccp-outer-7.c execution test
>
> which is a regression from 4.8-20130502. Reverting r198580 fixes it.
>
> The same FAILs also occur on trunk.
Ah right, I was confused by the fact that VEC_RSHIFT_EXPR is used
not just on little endian targets, but on big endian as well
(VEC_LSHIFT_EXPR is never emitted), but the important spot is
when extracting the scalar result from the vector:
if (BYTES_BIG_ENDIAN)
bitpos = size_binop (MULT_EXPR,
bitsize_int (TYPE_VECTOR_SUBPARTS (vectype) - 1),
TYPE_SIZE (scalar_type));
else
bitpos = bitsize_zero_node;
Fixed thusly, ok for trunk/4.8?
2013-05-17 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/57051
* fold-const.c (const_binop) <case VEC_LSHIFT_EXPR,
case VEC_RSHIFT_EXPR>: Fix BYTES_BIG_ENDIAN handling.
--- gcc/fold-const.c.jj 2013-05-16 12:36:28.000000000 +0200
+++ gcc/fold-const.c 2013-05-17 08:38:12.575117676 +0200
@@ -1393,7 +1393,7 @@ const_binop (enum tree_code code, tree a
if (shiftc >= outerc || (shiftc % innerc) != 0)
return NULL_TREE;
int offset = shiftc / innerc;
- if (code == VEC_LSHIFT_EXPR)
+ if ((code == VEC_RSHIFT_EXPR) ^ (!BYTES_BIG_ENDIAN))
offset = -offset;
tree zero = build_zero_cst (TREE_TYPE (type));
for (i = 0; i < count; i++)
Jakub
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] Fix VEC_[LR]SHIFT_EXPR folding for big-endian (PR tree-optimization/57051)
2013-05-17 6:48 ` [PATCH] Fix VEC_[LR]SHIFT_EXPR folding for big-endian " Jakub Jelinek
@ 2013-05-17 8:02 ` Richard Biener
0 siblings, 0 replies; 6+ messages in thread
From: Richard Biener @ 2013-05-17 8:02 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: Mikael Pettersson, gcc-patches
On Fri, 17 May 2013, Jakub Jelinek wrote:
> On Thu, May 16, 2013 at 07:59:00PM +0200, Mikael Pettersson wrote:
> > Jakub Jelinek writes:
> > > On Thu, Apr 25, 2013 at 11:47:02PM +0200, Jakub Jelinek wrote:
> > > > This patch adds folding of constant arguments v>> and v<<, which helps to
> > > > optimize the testcase from the PR back into constant store after vectorized
> > > > loop is unrolled.
> > >
> > > As this fixes a regression on the 4.8 branch, I've backported it (and
> > > minimal prerequisite for that) to 4.8 branch too.
> >
> > Unfortunately this patch makes gcc.dg/vect/no-scevccp-outer-{7,13}.c fail
> > on powerpc64-linux:
> >
> > +FAIL: gcc.dg/vect/no-scevccp-outer-13.c execution test
> > +FAIL: gcc.dg/vect/no-scevccp-outer-7.c execution test
> >
> > which is a regression from 4.8-20130502. Reverting r198580 fixes it.
> >
> > The same FAILs also occur on trunk.
>
> Ah right, I was confused by the fact that VEC_RSHIFT_EXPR is used
> not just on little endian targets, but on big endian as well
> (VEC_LSHIFT_EXPR is never emitted), but the important spot is
> when extracting the scalar result from the vector:
>
> if (BYTES_BIG_ENDIAN)
> bitpos = size_binop (MULT_EXPR,
> bitsize_int (TYPE_VECTOR_SUBPARTS (vectype) - 1),
> TYPE_SIZE (scalar_type));
> else
> bitpos = bitsize_zero_node;
>
> Fixed thusly, ok for trunk/4.8?
Ok with a comment in front of (code == VEC_RSHIFT_EXPR) ^
(!BYTES_BIG_ENDIAN)
Thanks,
Richard.
> 2013-05-17 Jakub Jelinek <jakub@redhat.com>
>
> PR tree-optimization/57051
> * fold-const.c (const_binop) <case VEC_LSHIFT_EXPR,
> case VEC_RSHIFT_EXPR>: Fix BYTES_BIG_ENDIAN handling.
>
> --- gcc/fold-const.c.jj 2013-05-16 12:36:28.000000000 +0200
> +++ gcc/fold-const.c 2013-05-17 08:38:12.575117676 +0200
> @@ -1393,7 +1393,7 @@ const_binop (enum tree_code code, tree a
> if (shiftc >= outerc || (shiftc % innerc) != 0)
> return NULL_TREE;
> int offset = shiftc / innerc;
> - if (code == VEC_LSHIFT_EXPR)
> + if ((code == VEC_RSHIFT_EXPR) ^ (!BYTES_BIG_ENDIAN))
> offset = -offset;
> tree zero = build_zero_cst (TREE_TYPE (type));
> for (i = 0; i < count; i++)
>
>
> Jakub
>
>
--
Richard Biener <rguenther@suse.de>
SUSE / SUSE Labs
SUSE LINUX Products GmbH - Nuernberg - AG Nuernberg - HRB 16746
GF: Jeff Hawn, Jennifer Guild, Felix Imend
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2013-05-17 8:02 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2013-04-26 0:32 [PATCH] Fold VEC_[LR]SHIFT_EXPR (PR tree-optimization/57051) Jakub Jelinek
2013-04-26 11:06 ` Richard Biener
2013-05-03 13:18 ` Jakub Jelinek
2013-05-16 18:01 ` Mikael Pettersson
2013-05-17 6:48 ` [PATCH] Fix VEC_[LR]SHIFT_EXPR folding for big-endian " Jakub Jelinek
2013-05-17 8:02 ` Richard Biener
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).