This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [Patch, AArch64] Fix shuffle for big-endian.
- From: Tejas Belagod <tbelagod at arm dot com>
- To: Richard Henderson <rth at redhat dot com>
- Cc: "gcc-patches at gcc dot gnu dot org" <gcc-patches at gcc dot gnu dot org>, Marcus Shawcroft <Marcus dot Shawcroft at arm dot com>
- Date: Wed, 02 Apr 2014 15:16:11 +0100
- Subject: Re: [Patch, AArch64] Fix shuffle for big-endian.
- Authentication-results: sourceware.org; auth=none
- References: <53077F31 dot 8070003 at arm dot com> <533078AC dot 3020307 at redhat dot com>
Richard Henderson wrote:
On 02/21/2014 08:30 AM, Tejas Belagod wrote:
+ /* If two vectors, we end up with a wierd mixed-endian mode on NEON. */
+ if (BYTES_BIG_ENDIAN)
+ {
+ if (!d->one_vector_p && d->perm[i] & nunits)
+ {
+ /* Extract the offset. */
+ elt = d->perm[i] & (nunits - 1);
+ /* Reverse the top half. */
+ elt = nunits - 1 - elt;
+ /* Offset it by the bottom half. */
+ elt += nunits;
+ }
+ else
+ elt = nunits - 1 - d->perm[i];
+ }
Isn't this just
elt = d->perm[i] ^ (nunits - 1);
all the time? I.e. invert the index within the word,
but leave the word index (nunits) unchanged.
Here is a revised patch. OK for stage-1?
Thanks
Tejas.
2014-04-02 Tejas Belagod <tejas.belagod@yahoo.com>
gcc/
* config/aarch64/aarch64.c (aarch64_evpc_tbl): Reverse order of elements
for big-endian.diff --git a/gcc/config/aarch64/aarch64.c b/gcc/config/aarch64/aarch64.c
index e839539..d30b79c 100644
--- a/gcc/config/aarch64/aarch64.c
+++ b/gcc/config/aarch64/aarch64.c
@@ -8129,7 +8129,15 @@ aarch64_evpc_tbl (struct expand_vec_perm_d *d)
return false;
for (i = 0; i < nelt; ++i)
- rperm[i] = GEN_INT (d->perm[i]);
+ {
+ int nunits = GET_MODE_NUNITS (vmode);
+
+ /* If big-endian and two vectors we end up with a wierd mixed-endian
+ mode on NEON. Reverse the index within each word but not the word
+ itself. */
+ rperm[i] = GEN_INT (BYTES_BIG_ENDIAN ? d->perm[i] ^ (nunits - 1)
+ : d->perm[i]);
+ }
sel = gen_rtx_CONST_VECTOR (vmode, gen_rtvec_v (nelt, rperm));
sel = force_reg (vmode, sel);