[gcc r12-8737] arm: correctly handle misaligned MEMs on MVE [PR105463]

[gcc r12-8737] arm: correctly handle misaligned MEMs on MVE [PR105463]

[Richard Earnshaw]

https://gcc.gnu.org/g:de1ba234311b935b1a38d512e57329d4b6e8354d

commit r12-8737-gde1ba234311b935b1a38d512e57329d4b6e8354d
Author: Richard Earnshaw <rearnsha@arm.com>
Date:   Wed May 11 13:08:40 2022 +0100

    arm: correctly handle misaligned MEMs on MVE [PR105463]
    
    Vector operations in MVE must be aligned to the element size, so if we
    are asked for a misaligned move in a wider mode we must recast it to a
    form suitable for the known alignment (larger elements have better
    address offset ranges, so there is some advantage to using wider
    element sizes if possible).  Whilst fixing this, also rework the
    predicates used for validating operands - the Neon predicates are
    not right for MVE.
    
    gcc/ChangeLog:
    
            PR target/105463
            * config/arm/mve.md (*movmisalign<mode>_mve_store): Use
            mve_memory_operand.
            (*movmisalign<mode>_mve_load): Likewise.
            * config/arm/vec-common.md (movmisalign<mode>): Convert to generator
            form...
            (@movmisalign<mode>): ... thus.  Use generic predicates and then
            rework operands if they are not valid.  For MVE rework to a
            narrower element size if the alignment is not high enough.
    
    (cherry picked from commit 6a116728e27c4da65d84483c0e75561a7479d4d5)

Diff:
---
 gcc/config/arm/mve.md        |  4 +-
 gcc/config/arm/vec-common.md | 90 ++++++++++++++++++++++++++++++++++----------
 2 files changed, 73 insertions(+), 21 deletions(-)

diff --git a/gcc/config/arm/mve.md b/gcc/config/arm/mve.md
index 369d7a79f6c..f16991c0a34 100644
--- a/gcc/config/arm/mve.md
+++ b/gcc/config/arm/mve.md
@@ -10462,7 +10462,7 @@
 )
 
 (define_insn "*movmisalign<mode>_mve_store"
-  [(set (match_operand:MVE_VLD_ST 0 "neon_permissive_struct_operand"	     "=Ux")
+  [(set (match_operand:MVE_VLD_ST 0 "mve_memory_operand"	     "=Ux")
 	(unspec:MVE_VLD_ST [(match_operand:MVE_VLD_ST 1 "s_register_operand" " w")]
 	 UNSPEC_MISALIGNED_ACCESS))]
   "((TARGET_HAVE_MVE && VALID_MVE_SI_MODE (<MODE>mode))
@@ -10475,7 +10475,7 @@
 
 (define_insn "*movmisalign<mode>_mve_load"
   [(set (match_operand:MVE_VLD_ST 0 "s_register_operand"				 "=w")
-	(unspec:MVE_VLD_ST [(match_operand:MVE_VLD_ST 1 "neon_permissive_struct_operand" " Ux")]
+	(unspec:MVE_VLD_ST [(match_operand:MVE_VLD_ST 1 "mve_memory_operand" " Ux")]
 	 UNSPEC_MISALIGNED_ACCESS))]
   "((TARGET_HAVE_MVE && VALID_MVE_SI_MODE (<MODE>mode))
     || (TARGET_HAVE_MVE_FLOAT && VALID_MVE_SF_MODE (<MODE>mode)))
diff --git a/gcc/config/arm/vec-common.md b/gcc/config/arm/vec-common.md
index fd878cba22d..1fd68f3ac43 100644
--- a/gcc/config/arm/vec-common.md
+++ b/gcc/config/arm/vec-common.md
@@ -280,29 +280,81 @@
   DONE;
 })
 
-(define_expand "movmisalign<mode>"
- [(set (match_operand:VDQ 0 "neon_perm_struct_or_reg_operand")
-	(unspec:VDQ [(match_operand:VDQ 1 "neon_perm_struct_or_reg_operand")]
+(define_expand "@movmisalign<mode>"
+ [(set (match_operand:VDQ 0 "nonimmediate_operand")
+	(unspec:VDQ [(match_operand:VDQ 1 "general_operand")]
 	 UNSPEC_MISALIGNED_ACCESS))]
  "ARM_HAVE_<MODE>_LDST && !BYTES_BIG_ENDIAN
   && unaligned_access && !TARGET_REALLY_IWMMXT"
 {
- rtx adjust_mem;
- /* This pattern is not permitted to fail during expansion: if both arguments
-    are non-registers (e.g. memory := constant, which can be created by the
-    auto-vectorizer), force operand 1 into a register.  */
- if (!s_register_operand (operands[0], <MODE>mode)
-     && !s_register_operand (operands[1], <MODE>mode))
-   operands[1] = force_reg (<MODE>mode, operands[1]);
-
- if (s_register_operand (operands[0], <MODE>mode))
-   adjust_mem = operands[1];
- else
-   adjust_mem = operands[0];
-
- /* Legitimize address.  */
- if (!neon_vector_mem_operand (adjust_mem, 2, true))
-   XEXP (adjust_mem, 0) = force_reg (Pmode, XEXP (adjust_mem, 0));
+  rtx *memloc;
+  bool for_store = false;
+  /* This pattern is not permitted to fail during expansion: if both arguments
+     are non-registers (e.g. memory := constant, which can be created by the
+     auto-vectorizer), force operand 1 into a register.  */
+  if (!s_register_operand (operands[0], <MODE>mode)
+      && !s_register_operand (operands[1], <MODE>mode))
+    operands[1] = force_reg (<MODE>mode, operands[1]);
+
+  if (s_register_operand (operands[0], <MODE>mode))
+    memloc = &operands[1];
+  else
+    {
+      memloc = &operands[0];
+      for_store = true;
+    }
+
+  /* For MVE, vector loads/stores must be aligned to the element size.  If the
+     alignment is less than that convert the load/store to a suitable mode.  */
+  if (TARGET_HAVE_MVE
+      && (MEM_ALIGN (*memloc)
+	  < GET_MODE_ALIGNMENT (GET_MODE_INNER (<MODE>mode))))
+    {
+      scalar_mode new_smode;
+      switch (MEM_ALIGN (*memloc))
+	{
+	case 64:
+	case 32:
+	  new_smode = SImode;
+	  break;
+	case 16:
+	  new_smode = HImode;
+	  break;
+	default:
+	  new_smode = QImode;
+	  break;
+	}
+      machine_mode new_mode
+	= mode_for_vector (new_smode,
+			   GET_MODE_SIZE (<MODE>mode)
+			   / GET_MODE_SIZE (new_smode)).require ();
+      rtx new_mem = adjust_address (*memloc, new_mode, 0);
+
+      if (!for_store)
+	{
+	  rtx reg = gen_reg_rtx (new_mode);
+	  emit_insn (gen_movmisalign (new_mode, reg, new_mem));
+	  emit_move_insn (operands[0], gen_lowpart (<MODE>mode, reg));
+	  DONE;
+	}
+      emit_insn (gen_movmisalign (new_mode, new_mem,
+				  gen_lowpart (new_mode, operands[1])));
+      DONE;
+    }
+
+  /* Legitimize address.  */
+  if ((TARGET_HAVE_MVE
+       && !mve_vector_mem_operand (<MODE>mode, XEXP (*memloc, 0), false))
+      || (!TARGET_HAVE_MVE
+	  && !neon_vector_mem_operand (*memloc, 2, false)))
+    {
+      rtx new_mem
+	= replace_equiv_address (*memloc,
+				 force_reg (Pmode, XEXP (*memloc, 0)),
+				 false);
+      gcc_assert (MEM_ALIGN (new_mem) == MEM_ALIGN (*memloc));
+      *memloc = new_mem;
+    }
 })
 
 (define_insn "mve_vshlq_<supf><mode>"