re PR tree-optimization/44284 (vectorization does not work for short variable)

2010-05-27  Richard Guenther  <rguenther@suse.de>

	PR tree-optimization/44284
	* tree-vect-stmts.c (vectorizable_assignment): Handle
	sign-changing conversions as simple copy.

	* gcc.dg/vect/vect-118.c: New testcase.
	* gcc.dg/vect/bb-slp-20.c: Adjust.
	* gcc.dg/vect/no-section-anchors-vect-36.c: Likewise.
	* gcc.dg/vect/slp-9.c: Likewise.
	* gcc.dg/vect/slp-reduc-4.c: Likewise.
	* gcc.dg/vect/vect-10.c: Likewise.
	* gcc.dg/vect/vect-109.c: Likewise.
	* gcc.dg/vect/vect-12.c: Likewise.
	* gcc.dg/vect/vect-36.c: Likewise.
	* gcc.dg/vect/vect-7.c: Likewise.
	* gcc.dg/vect/vect-iv-8.c: Likewise.
	* gcc.dg/vect/vect-multitypes-10.c: Likewise.
	* gcc.dg/vect/vect-multitypes-13.c: Likewise.
	* gcc.dg/vect/vect-multitypes-14.c: Likewise.
	* gcc.dg/vect/vect-multitypes-15.c: Likewise.
	* gcc.dg/vect/vect-multitypes-7.c: Likewise.
	* gcc.dg/vect/vect-multitypes-8.c: Likewise.
	* gcc.dg/vect/vect-multitypes-9.c: Likewise.
	* gcc.dg/vect/vect-reduc-dot-s16b.c: Likewise.
	* gcc.dg/vect/vect-reduc-dot-s8a.c: Likewise.
	* gcc.dg/vect/vect-reduc-dot-s8b.c: Likewise.
	* gcc.dg/vect/vect-reduc-dot-u16b.c: Likewise.
	* gcc.dg/vect/vect-strided-a-u32-mult.c: Likewise.
	* gcc.dg/vect/vect-strided-u32-mult.c: Likewise.
	* gcc.dg/vect/vect-widen-mult-s16.c: Likewise.
	* gcc.dg/vect/vect-widen-mult-s8.c: Likewise.
	* gcc.dg/vect/vect-widen-mult-sum.c: Likewise.
	* gcc.dg/vect/vect-widen-mult-u16.c: Likewise.

From-SVN: r159920

gcc/ChangeLog

+2010-05-27  Richard Guenther  <rguenther@suse.de>
+
+	PR tree-optimization/44284
+	* tree-vect-stmts.c (vectorizable_assignment): Handle
+	sign-changing conversions as simple copy.
+
 2010-05-27  Maxim Kuvyrkov  <maxim@codesourcery.com>
 
 	* gthr-posix.h (pthread_cancel): Don't declare if compiling against

gcc/testsuite/ChangeLog

+2010-05-27  Richard Guenther  <rguenther@suse.de>
+
+	PR tree-optimization/44284
+	* gcc.dg/vect/vect-118.c: New testcase.
+	* gcc.dg/vect/bb-slp-20.c: Adjust.
+	* gcc.dg/vect/no-section-anchors-vect-36.c: Likewise.
+	* gcc.dg/vect/slp-9.c: Likewise.
+	* gcc.dg/vect/slp-reduc-4.c: Likewise.
+	* gcc.dg/vect/vect-10.c: Likewise.
+	* gcc.dg/vect/vect-109.c: Likewise.
+	* gcc.dg/vect/vect-12.c: Likewise.
+	* gcc.dg/vect/vect-36.c: Likewise.
+	* gcc.dg/vect/vect-7.c: Likewise.
+	* gcc.dg/vect/vect-iv-8.c: Likewise.
+	* gcc.dg/vect/vect-multitypes-10.c: Likewise.
+	* gcc.dg/vect/vect-multitypes-13.c: Likewise.
+	* gcc.dg/vect/vect-multitypes-14.c: Likewise.
+	* gcc.dg/vect/vect-multitypes-15.c: Likewise.
+	* gcc.dg/vect/vect-multitypes-7.c: Likewise.
+	* gcc.dg/vect/vect-multitypes-8.c: Likewise.
+	* gcc.dg/vect/vect-multitypes-9.c: Likewise.
+	* gcc.dg/vect/vect-reduc-dot-s16b.c: Likewise.
+	* gcc.dg/vect/vect-reduc-dot-s8a.c: Likewise.
+	* gcc.dg/vect/vect-reduc-dot-s8b.c: Likewise.
+	* gcc.dg/vect/vect-reduc-dot-u16b.c: Likewise.
+	* gcc.dg/vect/vect-strided-a-u32-mult.c: Likewise.
+	* gcc.dg/vect/vect-strided-u32-mult.c: Likewise.
+	* gcc.dg/vect/vect-widen-mult-s16.c: Likewise.
+	* gcc.dg/vect/vect-widen-mult-s8.c: Likewise.
+	* gcc.dg/vect/vect-widen-mult-sum.c: Likewise.
+	* gcc.dg/vect/vect-widen-mult-u16.c: Likewise.
+
 2010-05-27  Maxim Kuvyrkov  <maxim@codesourcery.com>
 
 	* gcc.dg/glibc-uclibc-1.c, gcc.dg/glibc-uclibc-2.c: Remove, no longer

gcc/testsuite/gcc.dg/vect/bb-slp-20.c

@@ -21,8 +21,7 @@ main1 (unsigned int x, unsigned int y)
   else
     y = x;
 
-  /* Two SLP instances in the basic block, only one is supported for now,
-     the second one contains type conversion.  */
+  /* Two SLP instances in the basic block.  */
   a0 = in[0] + 23;
   a1 = in[1] + 142;
   a2 = in[2] + 2;
@@ -63,6 +62,6 @@ int main (void)
 }
 
 /* { dg-final { scan-tree-dump-times "basic block vectorized using SLP" 1 "slp" { target vect_int_mult } } } */
-/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "slp" { target vect_int_mult } } } */
+/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 2 "slp" { target vect_int_mult } } } */
 /* { dg-final { cleanup-tree-dump "slp" } } */
   

gcc/testsuite/gcc.dg/vect/no-section-anchors-vect-36.c

@@ -18,6 +18,7 @@ int main1 ()
   for (i = 0; i < N; i++)
     {
       s.cb[i] = 3*i;
+      __asm__ volatile ("");
     }
 
   for (i = 0; i < N; i++)

gcc/testsuite/gcc.dg/vect/slp-9.c

@@ -29,6 +29,7 @@ int main (void)
   for (i=0; i<N; i++) {
     X[i] = i;
     Y[i] = 64-i;
+    __asm__ volatile ("");
   }
 
   foo1 (N);

gcc/testsuite/gcc.dg/vect/slp-reduc-4.c

@@ -48,7 +48,10 @@ int main (void)
   check_vect ();
 
   for (i = 0; i < N; i++)
+    {
       uc[i] = i+3;
+      __asm__ volatile ("");
+    }
 
   main1 (N/8, 123, 124, 125, 126, 127, 128, 129, 313);
   return 0;

gcc/testsuite/gcc.dg/vect/vect-10.c

@@ -12,8 +12,7 @@ int foo ()
   short b[N] = {0,3,6,9,12,15,18,21,24,27,30,33,36,39,42,45};
   short c[N] = {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15};
 
-
-  /* Not vectorizable yet (strided access pattern).  */
+  /* Strided access pattern.  */
   for (i = 0; i < N/2; i++)
     {
       a[i] = b[2*i+1] * c[2*i+1] - b[2*i] * c[2*i];
@@ -23,5 +22,5 @@ int foo ()
   return 0;
 }
 
-/* { dg-final { scan-tree-dump-times "vectorized 0 loops" 1 "vect" } } */
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { xfail { ! vect_extract_even_odd } } } } */
 /* { dg-final { cleanup-tree-dump "vect" } } */

gcc/testsuite/gcc.dg/vect/vect-109.c

@@ -72,7 +72,7 @@ int main (void)
   return 0;
 }
 
-/* { dg-final { scan-tree-dump-times "vectorized 0 loops" 2 "vect" } } */
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 2 "vect" } } */
 /* { dg-final { scan-tree-dump-times "not vectorized: unsupported unaligned store" 2 "vect" { xfail vect_hw_misalign } } } */
 /* { dg-final { scan-tree-dump-times "Vectorizing an unaligned access" 10 "vect" { target vect_hw_misalign } } } */
 /* { dg-final { cleanup-tree-dump "vect" } } */

gcc/testsuite/gcc.dg/vect/vect-118.c

+/* { dg-do compile } */
+/* { dg-require-effective-target vect_int } */
+
+void f(short * __restrict__ a, short * __restrict__ b, short * __restrict__ x)
+{
+  int i;
+  for (i=0;i<1024;i++)
+    x[i] = a[i] + b[i];
+}
+
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" } } */
+/* { dg-final { cleanup-tree-dump "vect" } } */

gcc/testsuite/gcc.dg/vect/vect-12.c

@@ -16,7 +16,7 @@ int main1 ()
   short sc[N] = {0,3,6,9,12,15,18,21,24,27,30,33,36,39,42,45};
   short sb[N] = {0,3,6,9,12,15,18,21,24,27,30,33,36,39,42,45};
 
-  /* Not vectorizable yet (multiple types with different nunits in vector).  */
+  /* Multiple types with different nunits in vector.  */
   for (i = 0; i < N; i++)
     {
       ia[i] = ib[i] + ic[i];
@@ -40,5 +40,5 @@ int main (void)
   return main1 ();
 }
 
-/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { xfail *-*-* } } } */
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" } } */
 /* { dg-final { cleanup-tree-dump "vect" } } */

gcc/testsuite/gcc.dg/vect/vect-36.c

@@ -18,6 +18,7 @@ int main1 ()
   for (i = 0; i < N; i++)
     {
       s.cb[i] = 3*i;
+      __asm__ volatile ("");
     }
 
   for (i = 0; i < N; i++)

gcc/testsuite/gcc.dg/vect/vect-7.c

@@ -47,7 +47,6 @@ int main (void)
   return main1 ();
 }
 
-/* Fails for 32-bit targets that don't vectorize PLUS.  */
-/* { dg-final { scan-tree-dump-times "vectorized 2 loops" 1 "vect" { xfail *-*-* } } } */
+/* { dg-final { scan-tree-dump-times "vectorized 2 loops" 1 "vect" } } */
 /* { dg-final { scan-tree-dump-times "Vectorizing an unaligned access" 0 "vect" } } */
 /* { dg-final { cleanup-tree-dump "vect" } } */

gcc/testsuite/gcc.dg/vect/vect-iv-8.c

@@ -39,9 +39,5 @@ int main (void)
   return main1 (3);
 } 
 
-/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { xfail *-*-* } } } */
-/* Fails to get vectorized due to a redundant cast. Once this is fixed, 
-   should be vectorized as follows:
-   dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { target { vect_pack_trunc && vect_unpack } } } 
-*/
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { target { vect_pack_trunc && vect_unpack } } } } */
 /* { dg-final { cleanup-tree-dump "vect" } } */

gcc/testsuite/gcc.dg/vect/vect-multitypes-10.c

@@ -43,6 +43,7 @@ int main (void)
   for (i=0; i<N; i++) {
     X[i] = 16-i;
     uX[i] = 16-i;
+    __asm__ volatile ("");
   }
 
   foo1 (N);

gcc/testsuite/gcc.dg/vect/vect-multitypes-13.c

@@ -39,6 +39,7 @@ int main (void)
   for (i=0; i<N; i++) {
     X[i] = 16-i;
     uX[i] = 16-i;
+    __asm__ volatile ("");
   }
 
   foo1 (N);

gcc/testsuite/gcc.dg/vect/vect-multitypes-14.c

@@ -31,6 +31,7 @@ int main (void)
   for (i=0; i<N; i++) {
     uX[i] = 16-i;
     uY[i] = 16-i;
+    __asm__ volatile ("");
   }
 
   foo1 (N);

gcc/testsuite/gcc.dg/vect/vect-multitypes-15.c

@@ -38,6 +38,7 @@ int main (void)
     X[i] = i;
     Y[i] = 64-i;
     CX[i] = i;
+    __asm__ volatile ("");
   }
 
   foo1 (N);

gcc/testsuite/gcc.dg/vect/vect-multitypes-7.c

@@ -39,6 +39,7 @@ int main (void)
     X[i] = i;
     Y[i] = 64-i;
     CX[i] = i;
+    __asm__ volatile ("");
   }
 
   foo1 (N);

gcc/testsuite/gcc.dg/vect/vect-multitypes-8.c

@@ -31,6 +31,7 @@ int main (void)
   for (i=0; i<N; i++) {
     uX[i] = 16-i;
     uY[i] = 16-i;
+    __asm__ volatile ("");
   }
 
   foo1 (N);

gcc/testsuite/gcc.dg/vect/vect-multitypes-9.c

@@ -39,6 +39,7 @@ int main (void)
   for (i=0; i<N; i++) {
     X[i] = 16-i;
     uX[i] = 16-i;
+    __asm__ volatile ("");
   }
 
   foo1 (N);

gcc/testsuite/gcc.dg/vect/vect-reduc-dot-s16a.c

@@ -39,6 +39,7 @@ main (void)
     {
       X[i] = i;
       Y[i] = N - i;
+      __asm__ volatile ("");
     }
 
   dot = foo (N);

gcc/testsuite/gcc.dg/vect/vect-reduc-dot-s16b.c

@@ -39,6 +39,7 @@ main (void)
     {
       X[i] = i;
       Y[i] = 64 - i;
+      __asm__ volatile ("");
     }
 
   dot = foo (N);

gcc/testsuite/gcc.dg/vect/vect-reduc-dot-s8a.c

@@ -40,6 +40,7 @@ int main (void)
   for (i=0; i<N; i++) {
     X[i] = i;
     Y[i] = 64-i;
+    __asm__ volatile ("");
   }
 
   dot1 = foo1 (N);

gcc/testsuite/gcc.dg/vect/vect-reduc-dot-s8b.c

@@ -47,6 +47,7 @@ int main (void)
   for (i=0; i<N; i++) {
     X[i] = i;
     Y[i] = 64-i;
+    __asm__ volatile ("");
   }
 
   dot2 = foo2 (N);

gcc/testsuite/gcc.dg/vect/vect-reduc-dot-s8c.c

@@ -32,6 +32,7 @@ int main (void)
   for (i=0; i<N; i++) {
     X[i] = i;
     Y[i] = 64-i;
+    __asm__ volatile ("");
   }
 
   dot3 = foo3 (N);

gcc/testsuite/gcc.dg/vect/vect-reduc-dot-u16b.c

@@ -37,6 +37,7 @@ int main (void)
   for (i=0; i<N; i++) {
     X[i] = i;
     Y[i] = 64-i;
+    __asm__ volatile ("");
   }
 
   dot2 = foo2 (N);

gcc/testsuite/gcc.dg/vect/vect-strided-a-u32-mult.c

@@ -26,8 +26,7 @@ main1 ()
       arr[i] = i;
       iarr[i].a = i;
       iarr[i].b = i * 3;
-      if (arr[i] == 178)
-         abort();
+      __asm__ volatile ("");
     }
 
   for (i = 0; i < N; i++)

gcc/testsuite/gcc.dg/vect/vect-strided-u32-mult.c

@@ -53,8 +53,7 @@ int main (void)
       arr[i] = i;
       iarr[i].a = i;
       iarr[i].b = i * 3;
-      if (arr[i] == 178)
-         abort();
+      __asm__ volatile ("");
     }
   main1 (arr, iarr); 
     

gcc/testsuite/gcc.dg/vect/vect-widen-mult-s16.c

@@ -28,6 +28,7 @@ int main (void)
   for (i=0; i<N; i++) {
     X[i] = i;
     Y[i] = 64-i;
+    __asm__ volatile ("");
   }
 
   foo1 (N);

gcc/testsuite/gcc.dg/vect/vect-widen-mult-s8.c

@@ -28,6 +28,7 @@ int main (void)
   for (i=0; i<N; i++) {
     X[i] = i;
     Y[i] = 64-i;
+    __asm__ volatile ("");
   }
 
   foo1 (N);

gcc/testsuite/gcc.dg/vect/vect-widen-mult-sum.c

@@ -30,6 +30,7 @@ int main (void)
 
   for (i=0; i<N; i++) {
     X[i] = 16-i; 
+    __asm__ volatile ("");
   }
 
   sum = main1 (X, 1, 16, N-1);

gcc/testsuite/gcc.dg/vect/vect-widen-mult-u16.c

@@ -30,6 +30,7 @@ int main (void)
   for (i=0; i<N; i++) {
     X[i] = i;
     Y[i] = 64-i;
+    __asm__ volatile ("");
   }
 
   foo1 (N);

gcc/tree-vect-stmts.c

@@ -1829,7 +1829,7 @@ vectorizable_assignment (gimple stmt, gimple_stmt_iterator *gsi,
   tree def;
   gimple def_stmt;
   enum vect_def_type dt[2] = {vect_unknown_def_type, vect_unknown_def_type};
-  int nunits = TYPE_VECTOR_SUBPARTS (vectype);
+  unsigned int nunits = TYPE_VECTOR_SUBPARTS (vectype);
   int ncopies;
   int i, j;
   VEC(tree,heap) *vec_oprnds = NULL;
@@ -1837,6 +1837,8 @@ vectorizable_assignment (gimple stmt, gimple_stmt_iterator *gsi,
   bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_info);
   gimple new_stmt = NULL;
   stmt_vec_info prev_stmt_info = NULL;
+  enum tree_code code;
+  tree vectype_in;
 
   /* Multiple types in SLP are handled by creating the appropriate number of
      vectorized stmts for each SLP node. Hence, NCOPIES is always 1 in
@@ -1862,19 +1864,31 @@ vectorizable_assignment (gimple stmt, gimple_stmt_iterator *gsi,
   if (TREE_CODE (scalar_dest) != SSA_NAME)
     return false;
 
+  code = gimple_assign_rhs_code (stmt);
   if (gimple_assign_single_p (stmt)
-      || gimple_assign_rhs_code (stmt) == PAREN_EXPR)
+      || code == PAREN_EXPR
+      || CONVERT_EXPR_CODE_P (code))
     op = gimple_assign_rhs1 (stmt);
   else
     return false;
 
-  if (!vect_is_simple_use (op, loop_vinfo, bb_vinfo, &def_stmt, &def, &dt[0]))
+  if (!vect_is_simple_use_1 (op, loop_vinfo, bb_vinfo,
+			     &def_stmt, &def, &dt[0], &vectype_in))
     {
       if (vect_print_dump_info (REPORT_DETAILS))
         fprintf (vect_dump, "use not simple.");
       return false;
     }
 
+  /* We can handle NOP_EXPR conversions that do not change the number
+     of elements or the vector size.  */
+  if (CONVERT_EXPR_CODE_P (code)
+      && (!vectype_in
+	  || TYPE_VECTOR_SUBPARTS (vectype_in) != nunits
+	  || (GET_MODE_SIZE (TYPE_MODE (vectype))
+	      != GET_MODE_SIZE (TYPE_MODE (vectype_in)))))
+    return false;
+
   if (!vec_stmt) /* transformation not required.  */
     {
       STMT_VINFO_TYPE (stmt_info) = assignment_vec_info_type;
@@ -1903,6 +1917,8 @@ vectorizable_assignment (gimple stmt, gimple_stmt_iterator *gsi,
       /* Arguments are ready. create the new vector stmt.  */
       for (i = 0; VEC_iterate (tree, vec_oprnds, i, vop); i++)
        {
+	 if (CONVERT_EXPR_CODE_P (code))
+	   vop = build1_stat (VIEW_CONVERT_EXPR, vectype, vop);
          new_stmt = gimple_build_assign (vec_dest, vop);
          new_temp = make_ssa_name (vec_dest, new_stmt);
          gimple_assign_set_lhs (new_stmt, new_temp);