diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index b71f1ad6196..0390e7a58b8 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,9 @@
+Mon Apr 27 18:23:51 1998  J"orn Rennecke <amylaar@cygnus.co.uk>
+
+        * sh.c (sh_expand_prologue, sh_expand_epilogue):
+        If TARGET_DOUBLE_ALIGN, preserve 64 bit stack alignment.
+        * sh.h (STACK_BOUNDARY): Likewise.
+
 Mon Apr 27 17:22:48 1998  J"orn Rennecke <amylaar@cygnus.co.uk>
 
 	* sh.h (LEGITIMIZE_RELOAD_ADDRESS): Define.
diff --git a/gcc/config/sh/sh.c b/gcc/config/sh/sh.c
index 437164c924a..03f6b8a5f5a 100644
--- a/gcc/config/sh/sh.c
+++ b/gcc/config/sh/sh.c
@@ -3467,6 +3467,7 @@ sh_expand_prologue ()
   int live_regs_mask;
   int d, i;
   int live_regs_mask2;
+  int double_align = 0;
 
   /* We have pretend args if we had an object sent partially in registers
      and partially on the stack, e.g. a large structure.  */
@@ -3505,7 +3506,11 @@ sh_expand_prologue ()
   live_regs_mask = calc_live_regs (&d, &live_regs_mask2);
   push_regs (live_regs_mask, live_regs_mask2);
 
-  output_stack_adjust (-get_frame_size (), stack_pointer_rtx, 3);
+  if (TARGET_ALIGN_DOUBLE && d & 1)
+    double_align = 4;
+
+  output_stack_adjust (-get_frame_size () - double_align,
+		       stack_pointer_rtx, 3);
 
   if (frame_pointer_needed)
     emit_insn (gen_movsi (frame_pointer_rtx, stack_pointer_rtx));
@@ -3518,10 +3523,16 @@ sh_expand_epilogue ()
   int d, i;
 
   int live_regs_mask2;
+  int frame_size = get_frame_size ();
+
+  live_regs_mask = calc_live_regs (&d, &live_regs_mask2);
+
+  if (TARGET_ALIGN_DOUBLE && d & 1)
+    frame_size += 4;
 
   if (frame_pointer_needed)
     {
-      output_stack_adjust (get_frame_size (), frame_pointer_rtx, 7);
+      output_stack_adjust (frame_size, frame_pointer_rtx, 7);
 
       /* We must avoid moving the stack pointer adjustment past code
 	 which reads from the local frame, else an interrupt could
@@ -3530,14 +3541,14 @@ sh_expand_epilogue ()
       emit_insn (gen_blockage ());
       emit_insn (gen_movsi (stack_pointer_rtx, frame_pointer_rtx));
     }
-  else if (get_frame_size ())
+  else if (frame_size)
     {
       /* We must avoid moving the stack pointer adjustment past code
 	 which reads from the local frame, else an interrupt could
 	 occur after the SP adjustment and clobber data in the local
 	 frame.  */
       emit_insn (gen_blockage ());
-      output_stack_adjust (get_frame_size (), stack_pointer_rtx, 7);
+      output_stack_adjust (frame_size, stack_pointer_rtx, 7);
     }
 
   /* Pop all the registers.  */
diff --git a/gcc/config/sh/sh.h b/gcc/config/sh/sh.h
index b4e921c1c0b..dce9acb1acd 100644
--- a/gcc/config/sh/sh.h
+++ b/gcc/config/sh/sh.h
@@ -231,7 +231,7 @@ do {								\
 #define PARM_BOUNDARY  	32
 
 /* Boundary (in *bits*) on which stack pointer should be aligned.  */
-#define STACK_BOUNDARY  32
+#define STACK_BOUNDARY  BIGGEST_ALIGNMENT
 
 /* The log (base 2) of the cache line size, in bytes.  Processors prior to
    SH3 have no actual cache, but they fetch code in chunks of 4 bytes.  */
diff --git a/gcc/invoke.texi b/gcc/invoke.texi
index 52e2f8de332..a7c9c2ea48b 100644
--- a/gcc/invoke.texi
+++ b/gcc/invoke.texi
@@ -371,7 +371,7 @@ in the following sections.
 -mrelax  -mh -ms -mint32  -malign-300
 
 @emph{SH Options}
--m1  -m2  -m3  -m3e  -mb  -ml  -mrelax
+-m1  -m2  -m3  -m3e  -mb  -ml  -mdalign -mrelax
 
 @emph{System V Options}
 -Qy  -Qn  -YP,@var{paths}  -Ym,@var{dir}
@@ -5373,6 +5373,11 @@ Compile code for the processor in big endian mode.
 @item -ml
 Compile code for the processor in little endian mode.
 
+@item -mdalign
+Align doubles at 64 bit boundaries.  Note that this changes the calling
+conventions, and thus some functions from the standard C library will
+not work unless you recompile it first with -mdalign.
+
 @item -mrelax
 Shorten some address references at link time, when possible; uses the
 linker option @samp{-relax}.