[PATCH] Fold VIEW_CONVERT_EXPR <type, STRING_CST> generated by Fortran FE a lot (PR target/35366)

[Jakub Jelinek <jakub at redhat dot com>]

Hi!

Fortran FE generates VIEW_CONVERT_EXPR of STRING_CST quite a lot, but
fold-const isn't able to fold it, so GCC generates very inefficient
and (in some cases also buggy, as can be seen on the equiv_7.f90
on powerpc64-darwin with -m64 -Os) code.

I've bootstrapped/regtested this on x86_64-linux.

Unfortrunately, it causes 2 Fortran testsuite failures, hollerith.f90
and transfer_simplify_4.f90, both at all optimization levels.
I believe the tests are just invalid though.  The first one does:
logical l
l = 4Ho wo
and expects all the 32 bits preserved, but that is certainly against
the semantics of BOOLEAN_TYPE and from quick skimming of the Fortran
standard also LOGICAL type.  BOOLEAN_TYPE has just two values, false
and true (similarly for LOGICAL .false. and .true.) and so the folder
IMHO correctly folds this into
l = .true.
(4Ho wo is non-zero).  The transfer_simplify_4.f90 testcase transfers an
integer into logical and back and expects again all the 32-bits to be
preserved.

Fortran folks, can you please look at these 2 testcases and say whether
they are valid Fortran or just undefined behavior?

2008-11-11  Jakub Jelinek  <jakub@redhat.com>

	PR target/35366
	* fold-const.c (native_encode_string): New function.
	(native_encode_expr): Use it for STRING_CST.

	* gfortran.dg/hollerith.f90: Don't assume a 32-bit value
	stored into logical variable will be preserved.
	* gfortran.dg/transfer_simplify_4.f90: Removed.

--- gcc/fold-const.c.jj	2008-10-29 18:49:06.000000000 +0100
+++ gcc/fold-const.c	2008-11-11 11:41:34.000000000 +0100
@@ -7315,6 +7315,37 @@ native_encode_vector (const_tree expr, u
 }
 
 
+/* Subroutine of native_encode_expr.  Encode the STRING_CST
+   specified by EXPR into the buffer PTR of length LEN bytes.
+   Return the number of bytes placed in the buffer, or zero
+   upon failure.  */
+
+static int
+native_encode_string (const_tree expr, unsigned char *ptr, int len)
+{
+  tree type = TREE_TYPE (expr);
+  HOST_WIDE_INT total_bytes;
+
+  if (TREE_CODE (type) != ARRAY_TYPE
+      || TREE_CODE (TREE_TYPE (type)) != INTEGER_TYPE
+      || GET_MODE_BITSIZE (TYPE_MODE (TREE_TYPE (type))) != BITS_PER_UNIT
+      || !host_integerp (TYPE_SIZE_UNIT (type), 0))
+    return 0;
+  total_bytes = tree_low_cst (TYPE_SIZE_UNIT (type), 0);
+  if (total_bytes > len)
+    return 0;
+  if (TREE_STRING_LENGTH (expr) < total_bytes)
+    {
+      memcpy (ptr, TREE_STRING_POINTER (expr), TREE_STRING_LENGTH (expr));
+      memset (ptr + TREE_STRING_LENGTH (expr), 0,
+	      total_bytes - TREE_STRING_LENGTH (expr));
+    }
+  else
+    memcpy (ptr, TREE_STRING_POINTER (expr), total_bytes);
+  return total_bytes;
+}
+
+
 /* Subroutine of fold_view_convert_expr.  Encode the INTEGER_CST,
    REAL_CST, COMPLEX_CST or VECTOR_CST specified by EXPR into the
    buffer PTR of length LEN bytes.  Return the number of bytes
@@ -7337,6 +7368,9 @@ native_encode_expr (const_tree expr, uns
     case VECTOR_CST:
       return native_encode_vector (expr, ptr, len);
 
+    case STRING_CST:
+      return native_encode_string (expr, ptr, len);
+
     default:
       return 0;
     }
--- gcc/testsuite/gfortran.dg/hollerith.f90.jj2	2008-09-30 16:56:06.000000000 +0200
+++ gcc/testsuite/gfortran.dg/hollerith.f90	2008-11-11 13:52:38.000000000 +0100
@@ -8,7 +8,7 @@ character z1(4)
 character*4 z2(2,2)
 character*80 line
 integer i
-logical l
+integer j
 real r
 character*8 c
 
@@ -20,15 +20,15 @@ data z2/4h(i7),'xxxx','xxxx','xxxx'/
 
 z2 (1,2) = 4h(i8)
 i = 4hHell
-l = 4Ho wo
+j = 4Ho wo
 r = 4Hrld! 
-write (line, '(3A4)') i, l, r
+write (line, '(3A4)') i, j, r
 if (line .ne. 'Hello world!') call abort
 i = 2Hab
 r = 2Hab
-l = 2Hab
+j = 2Hab
 c = 2Hab
-write (line, '(3A4, 8A)') i, l, r, c
+write (line, '(3A4, 8A)') i, j, r, c
 if (line .ne. 'ab  ab  ab  ab      ') call abort
 
 write(line, '(4A8, "!")' ) x
--- gcc/testsuite/gfortran.dg/transfer_simplify_4.f90	2008-09-30 16:56:06.000000000 +0200
+++ gcc/testsuite/gfortran.dg/transfer_simplify_4.f90	2008-09-30 10:18:57.740011359 +0200
@@ -1,30 +0,0 @@
-! { dg-do run }
-! { dg-options "-O0" }
-! Tests that the in-memory representation of a transferred variable
-! propagates properly.
-!
-  implicit none
-
-  integer, parameter :: ip1 = 42
-  logical, parameter :: ap1 = transfer(ip1, .true.)
-  integer, parameter :: ip2 = transfer(ap1, 0)
-
-  logical :: a
-  integer :: i
-  
-  i = transfer(transfer(ip1, .true.), 0)
-  if (i .ne. ip1) call abort ()
-
-  i = transfer(ap1, 0)
-  if (i .ne. ip1) call abort ()
-  
-  a = transfer(ip1, .true.)
-  i = transfer(a, 0)
-  if (i .ne. ip1) call abort ()
-
-  i = ip1
-  a = transfer(i, .true.)
-  i = transfer(a, 0)
-  if (i .ne. ip1) call abort ()
-
-end

	Jakub