public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc r14-8645] tree-optimization/113670 - gather/scatter to/from hard registers
@ 2024-01-31  9:13 Richard Biener
  0 siblings, 0 replies; only message in thread
From: Richard Biener @ 2024-01-31  9:13 UTC (permalink / raw)
  To: gcc-cvs

https://gcc.gnu.org/g:924137b9012cee5603482242de08fbf0b2030f6a

commit r14-8645-g924137b9012cee5603482242de08fbf0b2030f6a
Author: Richard Biener <rguenther@suse.de>
Date:   Wed Jan 31 09:09:50 2024 +0100

    tree-optimization/113670 - gather/scatter to/from hard registers
    
    The following makes sure we're not taking the address of hard
    registers when vectorizing appearant gathers or scatters to/from
    them.
    
            PR tree-optimization/113670
            * tree-vect-data-refs.cc (vect_check_gather_scatter):
            Make sure we can take the address of the reference base.
    
            * gcc.target/i386/pr113670.c: New testcase.

Diff:
---
 gcc/testsuite/gcc.target/i386/pr113670.c | 16 ++++++++++++++++
 gcc/tree-vect-data-refs.cc               |  5 +++++
 2 files changed, 21 insertions(+)

diff --git a/gcc/testsuite/gcc.target/i386/pr113670.c b/gcc/testsuite/gcc.target/i386/pr113670.c
new file mode 100644
index 000000000000..8b9d3744fe2e
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr113670.c
@@ -0,0 +1,16 @@
+/* { dg-do compile } */
+/* { dg-options "-msse2 -O2 -fno-vect-cost-model" } */
+
+typedef float __attribute__ ((vector_size (16))) vec;
+typedef int __attribute__ ((vector_size (16))) ivec;
+ivec x;
+
+void
+test (void)
+{
+  register vec a asm("xmm3"), b asm("xmm4");
+  register ivec c asm("xmm5");
+  for (int i = 0; i < 4; i++)
+    c[i] = a[i] < b[i] ? -1 : 1;
+  x = c;
+}
diff --git a/gcc/tree-vect-data-refs.cc b/gcc/tree-vect-data-refs.cc
index f592aeb8028a..e6a3035064b2 100644
--- a/gcc/tree-vect-data-refs.cc
+++ b/gcc/tree-vect-data-refs.cc
@@ -4325,6 +4325,11 @@ vect_check_gather_scatter (stmt_vec_info stmt_info, loop_vec_info loop_vinfo,
   if (!multiple_p (pbitpos, BITS_PER_UNIT))
     return false;
 
+  /* We need to be able to form an address to the base which for example
+     isn't possible for hard registers.  */
+  if (may_be_nonaddressable_p (base))
+    return false;
+
   poly_int64 pbytepos = exact_div (pbitpos, BITS_PER_UNIT);
 
   if (TREE_CODE (base) == MEM_REF)

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2024-01-31  9:13 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-01-31  9:13 [gcc r14-8645] tree-optimization/113670 - gather/scatter to/from hard registers 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).