From: Will Schmidt <will_schmidt@vnet.ibm.com>
To: Richard Biener <richard.guenther@gmail.com>,
Segher Boessenkool <segher@kernel.crashing.org>,
Bill Schmidt <wschmidt@linux.vnet.ibm.com>,
David Edelsohn <dje.gcc@gmail.com>
Cc: GCC Patches <gcc-patches@gcc.gnu.org>
Subject: [PATCH, rs6000] Early gimple folding of vec_mergeh and vec_mergel for float
Date: Tue, 07 Aug 2018 19:25:00 -0000 [thread overview]
Message-ID: <1533669898.4452.11.camel@brimstone.rchland.ibm.com> (raw)
Hi,
This adds support for gimple folding of vec_mergeh and vec_mergel
for float and double types. Support for the integral types is already
in-tree.
This change includes updates to the fold_mergehl_helper function to handle
creating the (integral) permute vector when the vector type non integer.
Relevant testcases are already in-tree.
OK for trunk?
Thanks,
-Will
[gcc]
2018-08-07 Will Schmidt <will_schmidt@vnet.ibm.com>
* config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
allow folding of mergeh() and mergel() for the float and double types.
(fold_mergehl_helper): Rework to handle building a permute tree
for float vectors.
diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c
index 6cb5c87..35c32be 100644
--- a/gcc/config/rs6000/rs6000.c
+++ b/gcc/config/rs6000/rs6000.c
@@ -15119,24 +15119,39 @@ fold_mergehl_helper (gimple_stmt_iterator *gsi, gimple *stmt, int use_high)
{
tree arg0 = gimple_call_arg (stmt, 0);
tree arg1 = gimple_call_arg (stmt, 1);
tree lhs = gimple_call_lhs (stmt);
tree lhs_type = TREE_TYPE (lhs);
- tree lhs_type_type = TREE_TYPE (lhs_type);
int n_elts = TYPE_VECTOR_SUBPARTS (lhs_type);
int midpoint = n_elts / 2;
int offset = 0;
if (use_high == 1)
offset = midpoint;
- tree_vector_builder elts (lhs_type, VECTOR_CST_NELTS (arg0), 1);
+ /* The permute_type will match the lhs for integral types. For double and
+ float types, the permute type needs to map to the V2 or V4 type that
+ matches size. */
+ tree permute_type;
+ if (INTEGRAL_TYPE_P (TREE_TYPE (lhs_type)))
+ permute_type = lhs_type;
+ else
+ if (TREE_TYPE (lhs_type) == TREE_TYPE (V2DF_type_node))
+ permute_type = V2DI_type_node;
+ else if (TREE_TYPE (lhs_type) == TREE_TYPE (V4SF_type_node))
+ permute_type = V4SI_type_node;
+ else
+ gcc_unreachable ();
+
+ tree_vector_builder elts (permute_type, VECTOR_CST_NELTS (arg0), 1);
for (int i = 0; i < midpoint; i++)
{
- elts.safe_push (build_int_cst (lhs_type_type, offset + i));
- elts.safe_push (build_int_cst (lhs_type_type, offset + n_elts + i));
+ elts.safe_push (build_int_cst (TREE_TYPE (permute_type),
+ offset + i));
+ elts.safe_push (build_int_cst (TREE_TYPE (permute_type),
+ offset + n_elts + i));
}
tree permute = elts.build ();
gimple *g = gimple_build_assign (lhs, VEC_PERM_EXPR, arg0, arg1, permute);
@@ -15757,18 +15772,22 @@ rs6000_gimple_fold_builtin (gimple_stmt_iterator *gsi)
case ALTIVEC_BUILTIN_VMRGLH:
case ALTIVEC_BUILTIN_VMRGLW:
case VSX_BUILTIN_XXMRGLW_4SI:
case ALTIVEC_BUILTIN_VMRGLB:
case VSX_BUILTIN_VEC_MERGEL_V2DI:
+ case VSX_BUILTIN_XXMRGLW_4SF:
+ case VSX_BUILTIN_VEC_MERGEL_V2DF:
fold_mergehl_helper (gsi, stmt, 1);
return true;
/* vec_mergeh (integrals). */
case ALTIVEC_BUILTIN_VMRGHH:
case ALTIVEC_BUILTIN_VMRGHW:
case VSX_BUILTIN_XXMRGHW_4SI:
case ALTIVEC_BUILTIN_VMRGHB:
case VSX_BUILTIN_VEC_MERGEH_V2DI:
+ case VSX_BUILTIN_XXMRGHW_4SF:
+ case VSX_BUILTIN_VEC_MERGEH_V2DF:
fold_mergehl_helper (gsi, stmt, 0);
return true;
/* d = vec_pack (a, b) */
case P8V_BUILTIN_VPKUDUM:
next reply other threads:[~2018-08-07 19:25 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-08-07 19:25 Will Schmidt [this message]
2018-08-07 22:59 ` Segher Boessenkool
2018-08-17 14:01 ` Richard Biener
2018-08-17 15:05 ` Will Schmidt
2018-08-17 15:16 ` Richard Biener
2018-08-20 21:40 ` [PATCH, rs6000] Improve TREE_TYPE comparisons in fold_mergehl_helper() Will Schmidt
2018-08-20 22:07 ` Segher Boessenkool
2018-08-21 8:10 ` Richard Biener
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1533669898.4452.11.camel@brimstone.rchland.ibm.com \
--to=will_schmidt@vnet.ibm.com \
--cc=dje.gcc@gmail.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=richard.guenther@gmail.com \
--cc=segher@kernel.crashing.org \
--cc=wschmidt@linux.vnet.ibm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).