[bfin] Add option -mfast-fp

[Jie Zhang <jzhang918 at gmail dot com>]

This is the patch for GCC 4.1 to add -mfast-fp for Blackfin port.GCC 
will link with a fast Blackfin floating-point library if this option is 
added in command line. That library is available from

http://blackfin.uclinux.org

Is it OK to commit?


Jie
	* config/bfin/bfin.opt (mfast-fp): Add.
	* config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Let
	libbffastfp override libgcc if -mfast-fp.
	* config/bfin/bfin.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
	* doc/invoke.texi (Option Summary): Mention -mfast-fp.
	(Blackfin Options): Document -mfast-fp.

Index: doc/invoke.texi
===================================================================
--- doc/invoke.texi	(revision 127728)
+++ doc/invoke.texi	(working copy)
@@ -450,7 +450,7 @@ Objective-C and Objective-C++ Dialects}.
 -mstack-check-l1  -mid-shared-library  -mno-id-shared-library @gol
 -mshared-library-id=@var{n}  -mleaf-id-shared-library @gol
 -mno-leaf-id-shared-library  -msep-data  -mno-sep-data  -mlong-calls @gol
--mno-long-calls}
+-mno-long-calls -mfast-fp}
 
 @emph{CRIS Options}
 @gccoptlist{-mcpu=@var{cpu}  -march=@var{cpu}  -mtune=@var{cpu} @gol
@@ -8612,6 +8612,12 @@ This feature is not enabled by default. 
 @option{-mno-long-calls} will restore the default behavior.  Note these
 switches have no effect on how the compiler generates code to handle
 function calls via function pointers.
+
+@item -mfast-fp
+@opindex mfast-fp
+Link with the fast floating-point library. This library relaxes some of
+the IEEE floating-point standard's rules for checking inputs against
+Not-a-Number (NAN), in the interest of performance.
 @end table
 
 @node CRIS Options
Index: config/bfin/bfin.opt
===================================================================
--- config/bfin/bfin.opt	(revision 127728)
+++ config/bfin/bfin.opt	(working copy)
@@ -64,6 +64,10 @@ mlong-calls
 Target Report Mask(LONG_CALLS)
 Avoid generating pc-relative calls; use indirection
 
+mfast-fp
+Target Report Mask(FAST_FP)
+Link with the fast floating-point library
+
 mfdpic
 Target Report Mask(FDPIC)
 Enable Function Descriptor PIC mode
Index: config/bfin/linux.h
===================================================================
--- config/bfin/linux.h	(revision 127728)
+++ config/bfin/linux.h	(working copy)
@@ -35,7 +35,8 @@ asm (TEXT_SECTION_ASM_OP);
 
 #undef LINK_GCC_C_SEQUENCE_SPEC
 #define LINK_GCC_C_SEQUENCE_SPEC \
-  "%{mfast-fp:-lbffastfp} %{static:--start-group} %G %L %{static:--end-group}%{!static:%G}"
+  "%{static:--start-group} %{mfast-fp:-lbffastfp} %G %L %{static:--end-group} \
+   %{!static:%{mfast-fp:-lbffastfp} %G}"
 
 #undef LINK_SPEC
 #define LINK_SPEC "\
Index: config/bfin/bfin.h
===================================================================
--- config/bfin/bfin.h	(revision 127728)
+++ config/bfin/bfin.h	(working copy)
@@ -89,9 +89,9 @@ extern int target_flags;
 # define SUBTARGET_DRIVER_SELF_SPECS
 #endif
 
-#define LINK_GCC_C_SEQUENCE_SPEC \
-  "%{mfdpic:%{!static: %L} %{static: %G %L %G}} \
-  %{!mfdpic:%G %L %G}"
+#define LINK_GCC_C_SEQUENCE_SPEC "\
+  %{mfast-fp:-lbffastfp} %G %L %{mfast-fp:-lbffastfp} %G \
+"
 
 /* A C string constant that tells the GCC driver program options to pass to
    the assembler.  It can also specify how to translate options you give to GNU