From 46d931b3dd31cbba7c3355ada63f155aa24a4e2b Mon Sep 17 00:00:00 2001
From: Richard Sandiford <richard.sandiford@arm.com>
Date: Thu, 30 May 2024 16:17:58 +0100
Subject: ira: Fix go_through_subreg offset calculation [PR115281]

go_through_subreg used:

  else if (!can_div_trunc_p (SUBREG_BYTE (x),
			     REGMODE_NATURAL_SIZE (GET_MODE (x)), offset))

to calculate the register offset for a pseudo subreg x.  In the blessed
days before poly-int, this was:

    *offset = (SUBREG_BYTE (x) / REGMODE_NATURAL_SIZE (GET_MODE (x)));

But I think this is testing the wrong natural size.  If we exclude
paradoxical subregs (which will get an offset of zero regardless),
it's the inner register that is being split, so it should be the
inner register's natural size that we use.

This matters in the testcase because we have an SFmode lowpart
subreg into the last of three variable-sized vectors.  The
SUBREG_BYTE is therefore equal to the size of two variable-sized
vectors.  Dividing by the vector size gives a register offset of 2,
as expected, but dividing by the size of a scalar FPR would give
a variable offset.

I think something similar could happen for fixed-size targets if
REGMODE_NATURAL_SIZE is different for vectors and integers (say),
although that case would trade an ICE for an incorrect offset.

gcc/
	PR rtl-optimization/115281
	* ira-conflicts.cc (go_through_subreg): Use the natural size of
	the inner mode rather than the outer mode.

gcc/testsuite/
	PR rtl-optimization/115281
	* gfortran.dg/pr115281.f90: New test.
---
 gcc/ira-conflicts.cc                   |  3 ++-
 gcc/testsuite/gfortran.dg/pr115281.f90 | 39 ++++++++++++++++++++++++++++++++++
 2 files changed, 41 insertions(+), 1 deletion(-)
 create mode 100644 gcc/testsuite/gfortran.dg/pr115281.f90

(limited to 'gcc')

diff --git a/gcc/ira-conflicts.cc b/gcc/ira-conflicts.cc
index 83274c5..15ac42d 100644
--- a/gcc/ira-conflicts.cc
+++ b/gcc/ira-conflicts.cc
@@ -227,8 +227,9 @@ go_through_subreg (rtx x, int *offset)
   if (REGNO (reg) < FIRST_PSEUDO_REGISTER)
     *offset = subreg_regno_offset (REGNO (reg), GET_MODE (reg),
 				   SUBREG_BYTE (x), GET_MODE (x));
+  /* The offset is always 0 for paradoxical subregs.  */
   else if (!can_div_trunc_p (SUBREG_BYTE (x),
-			     REGMODE_NATURAL_SIZE (GET_MODE (x)), offset))
+			     REGMODE_NATURAL_SIZE (GET_MODE (reg)), offset))
     /* Checked by validate_subreg.  We must know at compile time which
        inner hard registers are being accessed.  */
     gcc_unreachable ();
diff --git a/gcc/testsuite/gfortran.dg/pr115281.f90 b/gcc/testsuite/gfortran.dg/pr115281.f90
new file mode 100644
index 0000000..80aa822
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/pr115281.f90
@@ -0,0 +1,39 @@
+! { dg-options "-O3" }
+! { dg-additional-options "-mcpu=neoverse-v1" { target aarch64*-*-* } }
+
+SUBROUTINE fn0(ma, mb, nt)
+  CHARACTER ca
+  REAL r0(ma)
+  INTEGER i0(mb)
+  REAL r1(3,mb)
+  REAL r2(3,mb)
+  REAL r3(3,3)
+  zero=0.0
+  do na = 1, nt
+     nt = i0(na)
+     do l = 1, 3
+        r1 (l, na) =   r0 (nt)
+        r2(l, na) = zero
+     enddo
+  enddo
+  if (ca  .ne.'z') then
+     do j = 1, 3
+        do i = 1, 3
+           r4  = zero
+        enddo
+     enddo
+     do na = 1, nt
+        do k =  1, 3
+           do l = 1, 3
+              do m = 1, 3
+                 r3 = r4 * v
+              enddo
+           enddo
+        enddo
+     do i = 1, 3
+           do k = 1, ifn (r3)
+           enddo
+        enddo
+     enddo
+     endif
+END
-- 
cgit v1.1