[5/7] arc: Add support for ARC-specific GCC

ARC support in GCC is not upstream yet...

Signed-off-by: Mischa Jonker <mjonker@synopsys.com>
---
 support/gnuconfig/config.sub                       |    4 +-
 toolchain/Config.in                                |    3 +-
 .../gcc/4.4.7-arc/fix_branch_out_of_range.patch    |   22 ++++++++++++++++++++
 toolchain/gcc/Config.in                            |   22 ++++++++++++-------
 toolchain/gcc/gcc-uclibc-4.x.mk                    |    4 +++
 5 files changed, 44 insertions(+), 11 deletions(-)
 create mode 100644 toolchain/gcc/4.4.7-arc/fix_branch_out_of_range.patch

Dear Mischa Jonker,

On Mon, 22 Apr 2013 13:37:29 +0200, Mischa Jonker wrote:
> ARC support in GCC is not upstream yet...
> 
> Signed-off-by: Mischa Jonker <mjonker@synopsys.com>
> ---
>  support/gnuconfig/config.sub                       |    4 +-
>  toolchain/Config.in                                |    3 +-
>  .../gcc/4.4.7-arc/fix_branch_out_of_range.patch    |   22 ++++++++++++++++++++
>  toolchain/gcc/Config.in                            |   22 ++++++++++++-------
>  toolchain/gcc/gcc-uclibc-4.x.mk                    |    4 +++
>  5 files changed, 44 insertions(+), 11 deletions(-)
>  create mode 100644 toolchain/gcc/4.4.7-arc/fix_branch_out_of_range.patch
> 
> diff --git a/support/gnuconfig/config.sub b/support/gnuconfig/config.sub
> index 8509813..b2958ad 100755
> --- a/support/gnuconfig/config.sub
> +++ b/support/gnuconfig/config.sub
> @@ -259,7 +259,7 @@ case $basic_machine in
>  	| alpha | alphaev[4-8] | alphaev56 | alphaev6[78] | alphapca5[67] \
>  	| alpha64 | alpha64ev[4-8] | alpha64ev56 | alpha64ev6[78] | alpha64pca5[67] \
>  	| am33_2.0 \
> -	| arc \
> +	| arc | arceb \
>  	| arm | arm[bl]e | arme[lb] | armv[2-8] | armv[3-8][lb] | armv7[arm] \
>  	| avr | avr32 \
>  	| be32 | be64 \
> @@ -372,7 +372,7 @@ case $basic_machine in
>  	| aarch64-* | aarch64_be-* \
>  	| alpha-* | alphaev[4-8]-* | alphaev56-* | alphaev6[78]-* \
>  	| alpha64-* | alpha64ev[4-8]-* | alpha64ev56-* | alpha64ev6[78]-* \
> -	| alphapca5[67]-* | alpha64pca5[67]-* | arc-* \
> +	| alphapca5[67]-* | alpha64pca5[67]-* | arc-* | arceb-* \
>  	| arm-*  | armbe-* | armle-* | armeb-* | armv*-* \
>  	| avr-* | avr32-* \
>  	| be32-* | be64-* \

This should be in a separate patch. Is this change upstream in the
gnuconfig project. We from time to time sync up with the upstream
variant, so if you want those changes to be preserved over time, they
need to be upstreamed.

> diff --git a/toolchain/Config.in b/toolchain/Config.in
> index fa78cf4..5d1c592 100644
> --- a/toolchain/Config.in
> +++ b/toolchain/Config.in
> @@ -22,7 +22,8 @@ config BR2_TOOLCHAIN_EXTERNAL
>  
>  config BR2_TOOLCHAIN_CTNG
>  	bool "Crosstool-NG toolchain"
> -	depends on !BR2_microblaze && !BR2_aarch64 && !BR2_xtensa
> +	depends on !BR2_microblaze && !BR2_aarch64 && !BR2_xtensa && \
> +		   !BR2_arc && !BR2_arceb
>  	select BR2_TOOLCHAIN_HAS_SHADOW_PASSWORDS
>  	help
>  	  Say 'y' if you want to generate the toolchain with crosstool-NG

This should also be part of a separate patch, ideally.

> diff --git a/toolchain/gcc/4.4.7-arc/fix_branch_out_of_range.patch b/toolchain/gcc/4.4.7-arc/fix_branch_out_of_range.patch
> new file mode 100644
> index 0000000..f042a71
> --- /dev/null
> +++ b/toolchain/gcc/4.4.7-arc/fix_branch_out_of_range.patch
> @@ -0,0 +1,22 @@
> +diff --git a/gcc/config/arc/arc.c b/gcc/config/arc/arc.c
> +index ff602c0..b3ca4c4 100644
> +--- a/gcc/config/arc/arc.c
> ++++ b/gcc/config/arc/arc.c
> +@@ -6565,7 +6565,7 @@ estimate required size increase).
> + 	      rtx *ccp = &XEXP (XVECEXP (pat, 0, 1), 0);
> + 
> + 	      offset = branch_dest (insn) - INSN_ADDRESSES (INSN_UID (insn));
> +-	      if ((offset >= -140 && offset < 140)
> ++	      if ((offset >= -120 && offset < 120)
> + 		  && rtx_equal_p (XEXP (op, 1), const0_rtx)
> +  		  && compact_register_operand (XEXP (op, 0), VOIDmode)
> + 		  && equality_comparison_operator (op, VOIDmode))
> +@@ -6687,7 +6687,7 @@ estimate required size increase).
> + 
> + 		  if (op0 != cmp0)
> +  		    cc_clob_rtx = gen_rtx_REG (CC_ZNmode, CC_REG);
> +- 		  else if ((offset >= -140 && offset < 140)
> ++ 		  else if ((offset >= -120 && offset < 120)
> + 			   && rtx_equal_p (op1, const0_rtx)
> + 			   && compact_register_operand (op0, VOIDmode)
> + 			   && (GET_CODE (op) == EQ

Patch lacks a description + Signed-off-by.

> diff --git a/toolchain/gcc/Config.in b/toolchain/gcc/Config.in
> index d7e8715..2396e8e 100644
> --- a/toolchain/gcc/Config.in
> +++ b/toolchain/gcc/Config.in
> @@ -9,44 +9,49 @@ choice
>  	prompt "GCC compiler Version"
>  	default BR2_GCC_VERSION_4_4_X if BR2_sparc_sparchfleon || BR2_sparc_sparchfleonv8 || BR2_sparc_sparcsfleon || BR2_sparc_sparcsfleonv8
>  	default BR2_GCC_VERSION_4_2_2_AVR32_2_1_5 if BR2_avr32
> +	default BR2_GCC_VERSION_4_4_7_ARC if BR2_arc || BR2_arceb
>  	default BR2_GCC_VERSION_4_7_X
>  	help
>  	  Select the version of gcc you wish to use.
>  
> +	config BR2_GCC_VERSION_4_4_7_ARC
> +		depends on BR2_arc || BR2_arceb
> +		bool "gcc 4.4.7-arc"
> +
>  	config BR2_GCC_VERSION_4_2_2_AVR32_2_1_5
>  	       depends on BR2_avr32
>  	       bool "gcc 4.2.2-avr32-2.1.5"
>  
>  	config BR2_GCC_VERSION_4_3_X
> -		depends on !BR2_avr32 && !BR2_sparc_sparchfleon && !BR2_sparc_sparchfleonv8 && !BR2_sparc_sparcsfleon && !BR2_sparc_sparcsfleonv8 && !BR2_cortex_a5 && !BR2_cortex_a8 && !BR2_cortex_a9 && !BR2_cortex_a15 && !BR2_x86_atom && !BR2_powerpc_e300c2 && !BR2_powerpc_e300c3 && !BR2_powerpc_e500mc && !BR2_powerpc_464 && !BR2_powerpc_464fp && !BR2_powerpc_476 && !BR2_powerpc_476fp && !BR2_fa526
> +		depends on !BR2_arc && !BR2_arceb && !BR2_avr32 && !BR2_sparc_sparchfleon && !BR2_sparc_sparchfleonv8 && !BR2_sparc_sparcsfleon && !BR2_sparc_sparcsfleonv8 && !BR2_cortex_a5 && !BR2_cortex_a8 && !BR2_cortex_a9 && !BR2_cortex_a15 && !BR2_x86_atom && !BR2_powerpc_e300c2 && !BR2_powerpc_e300c3 && !BR2_powerpc_e500mc && !BR2_powerpc_464 && !BR2_powerpc_464fp && !BR2_powerpc_476 && !BR2_powerpc_476fp && !BR2_fa526
>  		bool "gcc 4.3.x"
>  
>  	config BR2_GCC_VERSION_4_4_X
> -		depends on !BR2_avr32 && !BR2_cortex_a5 && !BR2_cortex_a15 && !BR2_x86_atom && !BR2_powerpc_476 && !BR2_powerpc_476fp && !BR2_fa526
> +		depends on !BR2_arc && !BR2_arceb && !BR2_avr32 && !BR2_cortex_a5 && !BR2_cortex_a15 && !BR2_x86_atom && !BR2_powerpc_476 && !BR2_powerpc_476fp && !BR2_fa526
>  		bool "gcc 4.4.x"
>  
>  	config BR2_GCC_VERSION_4_5_X
> -		depends on !BR2_avr32 && !BR2_cortex_a15 && !BR2_sparc_sparchfleon && !BR2_sparc_sparchfleonv8 && !BR2_sparc_sparcsfleon && !BR2_sparc_sparcsfleonv8 && !BR2_fa526
> +		depends on !BR2_arc && !BR2_arceb && !BR2_avr32 && !BR2_cortex_a15 && !BR2_sparc_sparchfleon && !BR2_sparc_sparchfleonv8 && !BR2_sparc_sparcsfleon && !BR2_sparc_sparcsfleonv8 && !BR2_fa526
>  		select BR2_GCC_NEEDS_MPC
>  		bool "gcc 4.5.x"
>  
>  	config BR2_GCC_VERSION_4_6_X
> -		depends on !BR2_avr32 && !BR2_sparc_sparchfleon && !BR2_sparc_sparchfleonv8 && !BR2_sparc_sparcsfleon && !BR2_sparc_sparcsfleonv8
> +		depends on !BR2_arc && !BR2_arceb && !BR2_avr32 && !BR2_sparc_sparchfleon && !BR2_sparc_sparchfleonv8 && !BR2_sparc_sparcsfleon && !BR2_sparc_sparcsfleonv8
>  		select BR2_GCC_NEEDS_MPC
>  		bool "gcc 4.6.x"
>  
>  	config BR2_GCC_VERSION_4_7_X
> -		depends on !BR2_avr32 && !BR2_sparc_sparchfleon && !BR2_sparc_sparchfleonv8 && !BR2_sparc_sparcsfleon && !BR2_sparc_sparcsfleonv8
> +		depends on !BR2_arc && !BR2_arceb && !BR2_avr32 && !BR2_sparc_sparchfleon && !BR2_sparc_sparchfleonv8 && !BR2_sparc_sparcsfleon && !BR2_sparc_sparcsfleonv8
>  		select BR2_GCC_NEEDS_MPC
>  		bool "gcc 4.7.x"
>  
>  	config BR2_GCC_VERSION_4_8_X
> -		depends on !BR2_avr32 && !BR2_sparc_sparchfleon && !BR2_sparc_sparchfleonv8 && !BR2_sparc_sparcsfleon && !BR2_sparc_sparcsfleonv8
> +		depends on !BR2_arc && !BR2_arceb && !BR2_avr32 && !BR2_sparc_sparchfleon && !BR2_sparc_sparchfleonv8 && !BR2_sparc_sparcsfleon && !BR2_sparc_sparcsfleonv8
>  		select BR2_GCC_NEEDS_MPC
>  		bool "gcc 4.8.x"
>  
>  	config BR2_GCC_VERSION_SNAP
> -		depends on !BR2_avr32 && !BR2_sparc_sparchfleon && !BR2_sparc_sparchfleonv8 && !BR2_sparc_sparcsfleon && !BR2_sparc_sparcsfleonv8
> +		depends on !BR2_arc && !BR2_arceb && !BR2_avr32 && !BR2_sparc_sparchfleon && !BR2_sparc_sparchfleonv8 && !BR2_sparc_sparcsfleon && !BR2_sparc_sparcsfleonv8
>  		select BR2_GCC_NEEDS_MPC
>  		bool "gcc snapshot"

Hum, not really nice. Maybe it's time to split the lines. Or even to
have something like "BR2_ARCH_HAS_GCC_4_8", "BR2_ARCH_HAS_GCC_4_5",
etc. But maybe that can be as follow-up cleanups.

>  endchoice
> @@ -74,6 +79,7 @@ config BR2_GCC_VERSION
>  	default "4.2.2-avr32-2.1.5" if BR2_GCC_VERSION_4_2_2_AVR32_2_1_5
>  	default "4.3.6"     if BR2_GCC_VERSION_4_3_X
>  	default "4.4.7"     if BR2_GCC_VERSION_4_4_X
> +	default "4.4.7-arc" if BR2_GCC_VERSION_4_4_7_ARC
>  	default "4.5.4"     if BR2_GCC_VERSION_4_5_X
>  	default "4.6.4"     if BR2_GCC_VERSION_4_6_X
>  	default "4.7.3"     if BR2_GCC_VERSION_4_7_X
> @@ -119,6 +125,6 @@ config BR2_GCC_ENABLE_TLS
>  
>  config BR2_GCC_ENABLE_OPENMP
>  	bool "Enable compiler OpenMP support"
> -	depends on !BR2_PTHREADS_NONE && !BR2_avr32
> +	depends on !BR2_PTHREADS_NONE && !BR2_avr32 && !BR2_arc && !BR2_arceb
>  	help
>  	  Enable OpenMP support for the compiler
> diff --git a/toolchain/gcc/gcc-uclibc-4.x.mk b/toolchain/gcc/gcc-uclibc-4.x.mk
> index fc11ad4..31afd48 100644
> --- a/toolchain/gcc/gcc-uclibc-4.x.mk
> +++ b/toolchain/gcc/gcc-uclibc-4.x.mk
> @@ -29,6 +29,8 @@ ifneq ($(GCC_SNAP_DATE),)
>   GCC_SITE:=ftp://gcc.gnu.org/pub/gcc/snapshots/$(GCC_SNAP_DATE)/
>  else ifeq ($(findstring avr32,$(GCC_VERSION)),avr32)
>   GCC_SITE:=ftp://www.at91.com/pub/buildroot/
> +else ifeq ($(findstring arc,$(GCC_VERSION)),arc)
> + GCC_SITE:=$(BR2_ARC_SITE)
>  else
>   GCC_SITE:=$(BR2_GNU_MIRROR:/=)/gcc/gcc-$(GCC_VERSION)
>  endif
> @@ -63,7 +65,9 @@ endif
>  
>  # Determine soft-float options
>  ifeq ($(BR2_SOFT_FLOAT),y)
> +ifeq ($(BR2_arc)$(BR2_arceb),)
>  SOFT_FLOAT_CONFIG_OPTION:=--with-float=soft
> +endif

Hum. Maybe this --with-float=soft should instead only be passed for the
architectures for which it makes sense (ARM? PowerPC?).

>  ifeq ($(BR2_arm)$(BR2_armeb),y) # only set float-abi for arm
>  TARGET_SOFT_FLOAT:=-mfloat-abi=soft
>  else

Thomas

>> diff --git a/support/gnuconfig/config.sub 
>> b/support/gnuconfig/config.sub index 8509813..b2958ad 100755
>> -	| arc \
>> +	| arc | arceb \

> This should be in a separate patch. Is this change upstream in the gnuconfig project. We from time to time sync up with the upstream variant, so if you want those changes to be preserved over time, they need to be upstreamed.

The change is in the upstream project now. Would you like me to change the gnuconfig patch in BuildRoot from 'add ARC to gnuconfig' to 'update gnuconfig from upstream'?

Mischa

Dear Mischa Jonker,

On Wed, 24 Apr 2013 12:44:20 +0000, Mischa Jonker wrote:

> > This should be in a separate patch. Is this change upstream in the
> > gnuconfig project. We from time to time sync up with the upstream
> > variant, so if you want those changes to be preserved over time,
> > they need to be upstreamed.
> 
> The change is in the upstream project now. Would you like me to
> change the gnuconfig patch in BuildRoot from 'add ARC to gnuconfig'
> to 'update gnuconfig from upstream'?

Yes, that'd be great.

Thanks,

Thomas

Hi Thomas,

> > The change is in the upstream project now. Would you like me to change 
> > the gnuconfig patch in BuildRoot from 'add ARC to gnuconfig'
> > to 'update gnuconfig from upstream'?

> Yes, that'd be great.

There are some patches with your name on it in support/gnuconfig/patches. 
These patches seem to be applied already on support/gnuconfig/*. I have 
updated these patches so that they apply cleanly on top of the new gnuconfig
(+small ARC-specific addition), but they still have your Signed-off-by tag. 

What's the proper procedure for signing off patches within a patch? 

Mischa

Mischa, All,

On Wed, Apr 24, 2013 at 03:19:14PM +0000, Mischa Jonker wrote:
> There are some patches with your name on it in support/gnuconfig/patches. 
> These patches seem to be applied already on support/gnuconfig/*. I have 
> updated these patches so that they apply cleanly on top of the new gnuconfig
> (+small ARC-specific addition), but they still have your Signed-off-by tag. 
> 
> What's the proper procedure for signing off patches within a patch? 

Two options:
  - update the patch, add your SoB-line
  - add a new patch with only your SoB-line

If your change adds a new semantic change (which is the case: add a new
architecture), then I'd favour the second option.

If your change is an addition (as a fix, or for completeness), then go
for option 1.

Regards,
Yann E. MORIN.