Make arm926ejs use -march=armv5t to avoid problems with EABI
RFC PATCH 1/1 Make arm926ejs use -march=armv5t to avoid problems with EABI
Make arm926ejs use -march=armv5t to avoid problems with EABI
Using -march=armv5t instead of armv5te allows Marvell Kirkwood-based boards to boot with the EABI changes introduced in commit f772acf8a584067033eff1e231fcd1fb3a00d3d9
---
This allows me to build with -mabi=aapcs-linux again. I still haven't found out what exactly causes the issues I had reported here
http://www.mail-archive.com/u-boot at lists.denx.de/msg20517.html
but with this patch it works fine again. Disassembling the binary, I see that ldrd/strd instructions are gone (as expected), although I don't know if that is the issue.
cpu/arm926ejs/config.mk | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
cpu/arm926ejs/config.mk +1 -1
@@ -24,7 +24,7 @@ PLATFORM_RELFLAGS += -fno-strict-aliasing -fno-common -ffixed-r8 \ -msoft-float -PLATFORM_CPPFLAGS += -march=armv5te +PLATFORM_CPPFLAGS += -march=armv5t # ========================================================================= # # Supply options according to compiler version
Signature · 1 line
1.6.0.4
RFC PATCH 1/1 Make arm926ejs use -march=armv5t to avoid problems with EABI
Quoted 9 lines · Simon Kagstrom
Make arm926ejs use -march=armv5t to avoid problems with EABI
Using -march=armv5t instead of armv5te allows Marvell Kirkwood-based boards to boot with the EABI changes introduced in commit
f772acf8a584067033eff1e231fcd1fb3a00d3d9Signed-off-by Simon Kagstrom <[email protected]>---
Quoted 8 lines
This allows me to build with -mabi=aapcs-linux again. I still haven't found out what exactly causes the issues I had reported here
http://www.mail-archive.com/u-boot at lists.denx.de/msg20517.htmlbut with this patch it works fine again. Disassembling the binary, I see that ldrd/strd instructions are gone (as expected), although I don't know if that is the issue.
it's wired because ldrd and strd are for 64bit data access and will be faster than ldm/stm
and the "e" is explecitly specity to active the dsp assembly to improve the data transfert speed
Best Regards,
J.
RFC PATCH 1/1 Make arm926ejs use -march=armv5t to avoid problems with EABI
In reply to Simon KagstromQuoted 5 lines · On Fri, 4 Sep 2009 11:27:19 +0200 Simon Kagstrom…
Make arm926ejs use -march=armv5t to avoid problems with EABI
Using -march=armv5t instead of armv5te allows Marvell Kirkwood-based boards to boot with the EABI changes introduced in commit
f772acf8a584067033eff1e231fcd1fb3a00d3d9
Well, further testing (with ubifs) has shown that this doesn't actually solve the problem - only making it disappear for some cases.
So don't apply this patch for now.
// Simon
Quoted 10 lines
cpu/arm926ejs/config.mk +1
-1@@ -24,7 +24,7 @@ PLATFORM_RELFLAGS += -fno-strict-aliasing -fno-common -ffixed-r8 \ -msoft-float -PLATFORM_CPPFLAGS += -march=armv5te +PLATFORM_CPPFLAGS += -march=armv5t
RFC PATCH 1/1 Make arm926ejs use -march=armv5t to avoid problems with EABI
Simon,
Quoted 8 lines · On Monday 14 September 2009 11:02:15 Simon Kagstrom wrote:
Quoted 5 lines
Make arm926ejs use -march=armv5t to avoid problems with EABI
Using -march=armv5t instead of armv5te allows Marvell Kirkwood-based boards to boot with the EABI changes introduced in commit
f772acf8a584067033eff1e231fcd1fb3a00d3d9Well, further testing (with ubifs) has shown that this doesn't actually solve the problem - only making it disappear for some cases.
Did you make any progress with this problem? And further insights here?
Thanks.
Cheers,
Stefan
--
DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-0 Fax: (+49)-8142-66989-80 Email: office at denx.de
RFC PATCH 1/1 Make arm926ejs use -march=armv5t to avoid problems with EABI
Quoted 11 lines · On Thu, 24 Sep 2009 14:36:47 +0200 Stefan Roese <[email protected]> wrote:
Quoted 8 lines · On Monday 14 September 2009 11:02:15 Simon Kagstrom wrote:
Quoted 5 lines
Make arm926ejs use -march=armv5t to avoid problems with EABI
Using -march=armv5t instead of armv5te allows Marvell Kirkwood-based boards to boot with the EABI changes introduced in commit
f772acf8a584067033eff1e231fcd1fb3a00d3d9Well, further testing (with ubifs) has shown that this doesn't actually solve the problem - only making it disappear for some cases.
Did you make any progress with this problem? And further insights here?
Not really. I currently "solve" the problem by using the patch below, i.e., simply switching to the ABI before the change for arm926ejs. With this, it works fine for me, but I understand that this maybe isn't a good fix to the problem.
// Simon