This is the mail archive of the ecos-bugs@sourceware.org mailing list for the eCos project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[Bug 1001275] Cortex-M (armV7) architecture endian instructions /Applied on lwIP


Please do not reply to this email. Use the web interface provided at:
http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001275

--- Comment #11 from Ilija Kocho <ilijak@siva.com.mk> 2011-08-22 14:23:22 BST ---
(In reply to comment #9)
> (In reply to comment #8)
> > Sergei
> > 
> > I did testing and I have working code, but I would postpone patch
> > generation because of the question below.
> > 
> > (In reply to comment #6)
> > > (In reply to comment #5)
> > > > (In reply to comment #4)
> > > > Personally I would like CYGARC_SWAPxx(), but since CYG_SWAPxx() are
> > > > already present in hal_endian.h let's use them. It will propagate REV
> > > > instructions in other parts of code that may benefit (i guess it is
> > > > safe).
> > 
> > hal_endian.h has a provision CYGBLD_HAL_ENDIAN_H for include of
> > architecture specific CYGARC_SWAPxx(). This is what I put in
> > hal_cortexm.cdl:
> > 
> >     define_proc {
> >         puts $::cdl_system_header "#define CYGBLD_HAL_ENDIAN_H \
> >                 <cyg/hal/cortexm_endian.h>"
> >     }
> > 
> 
> I like it, this is a great find!
> 
> If I understood you correctly you again returned back to CYGARC prefix
> for macros. If it is not a typo then I would add below those definitions
> (CYGARC_SWAPxx macros) in ``cortexm_endian.h'' the same check and far
> definitions (overrides)

It is a my mistake:( sorry, CYG_SWAPxx() it should be.

[snip (as it is consequence to my error)]

> 
> and add such CDL option in Cortex-M config file. Well, used name of the
> option is for demo only. Thus, users can rid fast inlines if they want.
> 
> > Now cortexm_endian.h is a new header that contains the CYGARC_SWAPxx()
> > macros and corresponding inline functions. I added this file in order
> > to avoid include of complete hal_arch.h
> > 

Here is cortexm_endian.h snippet:

#include <cyg/hal/cortexm_regs.h>

static __inline__ cyg_uint32 cyg_hal_swap32(cyg_uint32 original)
{
    cyg_uint32 swapped;
    CYGARC_REV(swapped, original);
    return swapped;
}

static __inline__ cyg_uint16 cyg_hal_swap16(cyg_uint16 original)
{
    cyg_uint16 swapped;
    CYGARC_REV16(swapped, original);
    return swapped;
}

#define CYG_SWAP32(__val) cyg_hal_swap32(__val)
#define CYG_SWAP16(__val) cyg_hal_swap16(__val)

> > For me it is OK to add this header in arch/include but there may be a
> > better proposal. Please comment.
> > 
> > Ilija
> 
> So, with your proposal we can leave hal_arch.h without any changes. Did
> you mean it? I read so and I guess that we have to manage the definition
> of LWIP_PLATFORM_BYTESWAP from configtool. In ideal world this value must
> be defined from lwip config file, e.g.

I can see 2 options:

Option 1. We include cortexm_endian.h in hal_arch.h (that's what i have
tested).
here are all changes to hal_arch.h:

diff -u -r1.3 hal_arch.h
--- hal_arch.h    13 Feb 2009 17:04:18 -0000    1.3
+++ hal_arch.h    22 Aug 2011 12:45:24 -0000
@@ -55,6 +55,9 @@

 #include <cyg/hal/var_arch.h>

+#include <cyg/hal/cortexm_regs.h>
+#include <cyg/hal/cortexm_endian.h>
+ 


Option 2. Probably we can include hal_endian.h in lwipots.h (not tested).

My vote is for Option 1. since it would propagate cortexm_endian.h everywhere
(for better, or for worse) Cortex-M is used.

> 
> to add into lwip_net.cdl:
> 
> cdl_option CYGBLD_LWIP_HTONS_HTONL_INLINED {
>   default_value 0
> }
> 

In either case we don't need CDL. Following is what I have tested (a patch to
your input) :

> to add into lwipopts.h:
> 
> #if CYGPKG_LWIP_HTONS_HTONL_INLINED
-#if CYGPKG_LWIP_HTONS_HTONL_INLINED

> # include <cyg/hal/hal_endian.h>
+#ifdef CYG_SWAP32
> # define LWIP_PLATFORM_BYTESWAP 1
> # define LWIP_PLATFORM_HTONS(__val) CYG_SWAP16(__val)
> # define LWIP_PLATFORM_HTONL(__val) CYG_SWAP32(__val)
> #endif
> 
> What do you think?

If you agree with adding cortexm_endian.h and #include it in hal_arch.h I could
submit patches for review.

Ilija

-- 
Configure bugmail: http://bugs.ecos.sourceware.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]