public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [committed] hppa: Improve handling of REG+D addresses when generating PA 2.0 code
@ 2024-03-18  0:47 John David Anglin
  0 siblings, 0 replies; only message in thread
From: John David Anglin @ 2024-03-18  0:47 UTC (permalink / raw)
  To: GCC Patches

[-- Attachment #1: Type: text/plain, Size: 6397 bytes --]

Tested on hppa-unknown-linux-gnu.  Committed to trunk.

Dave
---

hppa: Improve handling of REG+D addresses when generating PA 2.0 code

In looking at PR 112415, it became clear that improvements could be
made in the handling of loads and stores using REG+D addresses.  A
change in 2002 conflated two issues:

1) We can't generate insns with 14-bit displacements before reload
completes when generating PA 1.x code since floating-point loads and
stores only support 5-bit offsets in PA 1.x.

2) The GNU ELF 32-bit linker lacks relocation support for PA 2.0
floating point instructions with 14-bit displacements.  These
relocations affect instructions with symbolic references.

The result of the change was to block creation of PA 2.0 instructions
with 14-bit REG_D displacements for SImode, DImode, SFmode and DFmode
on the GNU linux target before reload.  This was unnecessary as these
instructions don't need relocation.

This change revise the INT14_OK_STRICT define to allow creation
of instructions with 14-bit REG+D addresses before reload when
generating PA 2.0 code.

2024-03-17  John David Anglin  <danglin@gcc.gnu.org>

gcc/ChangeLog:

	PR rtl-optimization/112415
	* config/pa/pa.cc (pa_emit_move_sequence): Revise condition
	for symbolic memory operands.
	(pa_legitimate_address_p): Revise LO_SUM condition.
	* config/pa/pa.h (INT14_OK_STRICT): Revise define.  Move
	comment about GNU linker to predicates.md.
	* config/pa/predicates.md (floating_point_store_memory_operand):
	Revise condition for symbolic memory operands.  Update
	comment.

diff --git a/gcc/config/pa/pa.cc b/gcc/config/pa/pa.cc
index 129289f8e62..d7666103de8 100644
--- a/gcc/config/pa/pa.cc
+++ b/gcc/config/pa/pa.cc
@@ -2039,7 +2039,8 @@ pa_emit_move_sequence (rtx *operands, machine_mode mode, rtx scratch_reg)
 	      op1 = replace_equiv_address (op1, scratch_reg);
 	    }
 	}
-      else if ((!INT14_OK_STRICT && symbolic_memory_operand (op1, VOIDmode))
+      else if (((TARGET_ELF32 || !TARGET_PA_20)
+		&& symbolic_memory_operand (op1, VOIDmode))
 	       || IS_LO_SUM_DLT_ADDR_P (XEXP (op1, 0))
 	       || IS_INDEX_ADDR_P (XEXP (op1, 0)))
 	{
@@ -2088,7 +2089,8 @@ pa_emit_move_sequence (rtx *operands, machine_mode mode, rtx scratch_reg)
 	      op0 = replace_equiv_address (op0, scratch_reg);
 	    }
 	}
-      else if ((!INT14_OK_STRICT && symbolic_memory_operand (op0, VOIDmode))
+      else if (((TARGET_ELF32 || !TARGET_PA_20)
+		&& symbolic_memory_operand (op0, VOIDmode))
 	       || IS_LO_SUM_DLT_ADDR_P (XEXP (op0, 0))
 	       || IS_INDEX_ADDR_P (XEXP (op0, 0)))
 	{
@@ -11032,18 +11040,22 @@ pa_legitimate_address_p (machine_mode mode, rtx x, bool strict, code_helper)
 	  && (strict ? STRICT_REG_OK_FOR_BASE_P (y)
 		     : REG_OK_FOR_BASE_P (y)))
 	{
+	  y = XEXP (x, 1);
+
 	  /* Needed for -fPIC */
 	  if (mode == Pmode
-	      && GET_CODE (XEXP (x, 1)) == UNSPEC)
+	      && GET_CODE (y) == UNSPEC)
 	    return true;
 
-	  if (!INT14_OK_STRICT
-	      && (strict || !(reload_in_progress || reload_completed))
+	  /* Before reload, we need support for 14-bit floating
+	     point loads and stores, and associated relocations.  */
+	  if ((TARGET_ELF32 || !INT14_OK_STRICT)
+	      && !reload_completed
 	      && mode != QImode
 	      && mode != HImode)
 	    return false;
 
-	  if (CONSTANT_P (XEXP (x, 1)))
+	  if (CONSTANT_P (y))
 	    return true;
 	}
       return false;
diff --git a/gcc/config/pa/pa.h b/gcc/config/pa/pa.h
index 7abaeae269e..403f16c5cb5 100644
--- a/gcc/config/pa/pa.h
+++ b/gcc/config/pa/pa.h
@@ -828,19 +828,8 @@ extern int may_call_alloca;
 
 /* Nonzero if 14-bit offsets can be used for all loads and stores.
    This is not possible when generating PA 1.x code as floating point
-   accesses only support 5-bit offsets.  Note that we do not forbid
-   the use of 14-bit offsets prior to reload.  Instead, we use secondary
-   reloads to fix REG+D memory addresses for floating-point accesses.
-
-   FIXME: the GNU ELF linker clobbers the LSB of the FP register number
-   in PA 2.0 floating-point insns with long displacements.  This is
-   because R_PARISC_DPREL14WR and other relocations like it are not
-   yet supported by GNU ld.  For now, we reject long displacements
-   on this target.  */
-
-#define INT14_OK_STRICT \
-  (TARGET_SOFT_FLOAT                                                   \
-   || (TARGET_PA_20 && !TARGET_ELF32 && !TARGET_ELF64))
+   accesses only support 5-bit offsets.  */
+#define INT14_OK_STRICT (TARGET_SOFT_FLOAT || TARGET_PA_20)
 
 /* The macros REG_OK_FOR..._P assume that the arg is a REG rtx
    and check its validity for a certain class.
diff --git a/gcc/config/pa/predicates.md b/gcc/config/pa/predicates.md
index 3bd3a7c212f..50dffa1138c 100644
--- a/gcc/config/pa/predicates.md
+++ b/gcc/config/pa/predicates.md
@@ -331,12 +331,16 @@
 	  && !IS_INDEX_ADDR_P (XEXP (op, 0)));
 })
 
-;; True iff the operand OP can be used as the destination operand of
-;; a floating point store.  This also implies the operand could be used as
+;; True iff the operand OP can be used as the destination operand of a
+;; floating point store.  This also implies the operand could be used as
 ;; the source operand of a floating point load.  LO_SUM DLT and indexed
-;; memory operands are not allowed.  Symbolic operands are accepted if
-;; INT14_OK_STRICT is true.  We accept reloading pseudos and other memory
-;; operands.
+;; memory operands are not allowed.  Symbolic operands are accepted for
+;; PA 2.0 when TARGET_ELF32 is not true.  We accept reloading pseudos
+;; and other memory; operands.
+
+;; FIXME: The GNU ELF32 linker clobbers the LSB of the FP register number
+;; in PA 2.0 {fldw,fstw} insns with long displacements.  This is because
+;; R_PARISC_DPREL14WR and other relocations like it are not supported.
 
 (define_predicate "floating_point_store_memory_operand"
   (match_code "reg,mem")
@@ -362,7 +366,8 @@
     return false;
 
   return ((reload_in_progress || memory_address_p (mode, XEXP (op, 0)))
-	  && (INT14_OK_STRICT || !symbolic_memory_operand (op, VOIDmode))
+	  && !((TARGET_ELF32 || !TARGET_PA_20)
+	       && symbolic_memory_operand (op, VOIDmode))
 	  && !IS_LO_SUM_DLT_ADDR_P (XEXP (op, 0))
 	  && !IS_INDEX_ADDR_P (XEXP (op, 0)));
 })

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

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

only message in thread, other threads:[~2024-03-18  0:47 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-03-18  0:47 [committed] hppa: Improve handling of REG+D addresses when generating PA 2.0 code John David Anglin

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).