RFA: command line switches for QED series of MIPS processors.

RFA: command line switches for QED series of MIPS processors.

[Nick Clifton]

Hi Gavin, Hi Ulf,

  I would like to get approval to check in the patch below.  It adds
  command line switches for the QED series of MIPS processors to both
  GCC and GAS.  (These processors are r5000 variants).

Cheers
	Nick


2000-06-15  Nick Clifton  <nickc@cygnus.com>

	* config/mips/mips.c (override_options): Accept RM5200,
	RM5230, RM5231, RM5261, RM5721 and RM7000 as r5000 cpu
	variants.

	* invoke.texi: Document newly accepted cpu variants.

Index: gcc/config/mips/mips.c
===================================================================
RCS file: /cvs/gcc/egcs/gcc/config/mips/mips.c,v
retrieving revision 1.90
diff -p -r1.90 mips.c
*** mips.c	2000/06/14 22:47:56	1.90
--- mips.c	2000/06/15 19:01:43
*************** override_options ()
*** 4776,4782 ****
  	  break;
  	}
      }
- 
    else
      {
        const char *p = mips_cpu_string;
--- 4776,4781 ----
*************** override_options ()
*** 4851,4856 ****
--- 4850,4873 ----
  	  if (!strcmp (p, "orion"))
  	    mips_cpu = PROCESSOR_R4600;
  	  break;
+ 	  
+ 	case 'm':
+ 	case 'M':
+ 	  switch (atoi (p + 1))
+ 	    {
+ 	    case 5200:
+ 	    case 5230:
+ 	    case 5231:
+ 	    case 5261:
+ 	    case 5721:
+ 	    case 7000:
+ 	      target_flags |= MASK_MAD;
+ 	      mips_cpu = PROCESSOR_R5000;
+ 	      break;
+ 	      
+ 	    default:
+ 	      break;
+ 	    }
  	}
  
        if (seen_v
  
Index: gcc/invoke.texi
===================================================================
RCS file: /cvs/gcc/egcs/gcc/invoke.texi,v
retrieving revision 1.196
diff -p -r1.196 invoke.texi
*** invoke.texi	2000/06/09 16:26:21	1.196
--- invoke.texi	2000/06/15 19:01:43
*************** These @samp{-m} options are defined for 
*** 5656,5669 ****
  Assume the defaults for the machine type @var{cpu type} when scheduling
  instructions.  The choices for @var{cpu type} are @samp{r2000}, @samp{r3000},
  @samp{r3900}, @samp{r4000}, @samp{r4100}, @samp{r4300}, @samp{r4400},
! @samp{r4600}, @samp{r4650}, @samp{r5000}, @samp{r6000}, @samp{r8000},
! and @samp{orion}.  Additionally, the @samp{r2000}, @samp{r3000},
! @samp{r4000}, @samp{r5000}, and @samp{r6000} can be abbreviated as
! @samp{r2k} (or @samp{r2K}), @samp{r3k}, etc.  While picking a specific
! @var{cpu type} will schedule things appropriately for that particular
! chip, the compiler will not generate any code that does not meet level 1
! of the MIPS ISA (instruction set architecture) without a @samp{-mipsX}
! or @samp{-mabi} switch being used.
  
  @item -mips1
  Issue instructions from level 1 of the MIPS ISA.  This is the default.
--- 5656,5671 ----
  Assume the defaults for the machine type @var{cpu type} when scheduling
  instructions.  The choices for @var{cpu type} are @samp{r2000}, @samp{r3000},
  @samp{r3900}, @samp{r4000}, @samp{r4100}, @samp{r4300}, @samp{r4400},
! @samp{r4600}, @samp{r4650}, @samp{r5000}, @samp{rm5200}, @samp{rm5230},
! @samp{rm5231}, @samp{rm5261}, @samp{rm5721}, @samp{r6000},
! @samp{rm7000}, @samp{r8000}, and @samp{orion}.  Additionally, the
! @samp{r2000}, @samp{r3000}, @samp{r4000}, @samp{r5000}, and @samp{r6000}
! can be abbreviated as @samp{r2k} (or @samp{r2K}), @samp{r3k}, etc.
! While picking a specific @var{cpu type} will schedule things
! appropriately for that particular chip, the compiler will not generate
! any code that does not meet level 1 of the MIPS ISA (instruction set
! architecture) without a @samp{-mipsX} or @samp{-mabi} switch being
! used. 
  
  @item -mips1
  Issue instructions from level 1 of the MIPS ISA.  This is the default.


2000-06-15  Nick Clifton  <nickc@cygnus.com>

	* config/tc-mips.c (md_parse_option): Accept RM5200,RM5230,
	RM5231, RM5261, RM5721 and RM7000 as r5000 cpu variants.

	* doc/c-mips.texi: Document newly accepted cpu variants.

Index: gas/config/tc-mips.c
===================================================================
RCS file: /cvs/src//src/gas/config/tc-mips.c,v
retrieving revision 1.17
diff -p -r1.17 tc-mips.c
*** tc-mips.c	2000/06/09 00:00:03	1.17
--- tc-mips.c	2000/06/15 19:01:59
*************** md_parse_option (c, arg)
*** 9041,9048 ****
  		if (strcmp (p, "orion") == 0)
  		  mips_cpu = 4600;
  		break;
- 	      }
  
  	    if (sv
  		&& (mips_cpu != 4300
  		    && mips_cpu != 4100
--- 9041,9063 ----
  		if (strcmp (p, "orion") == 0)
  		  mips_cpu = 4600;
  		break;
  
+ 	      case 'm':
+ 	      case 'M':
+ 		switch (atoi (p + 1))
+ 		  {
+ 		  case 5200:
+ 		  case 5230:
+ 		  case 5231:
+ 		  case 5261:
+ 		  case 5721:
+ 		  case 7000:
+ 		    mips_cpu = 5000;
+ 		  default:
+ 		    break;
+ 		  }
+ 	      }
+ 	    
  	    if (sv
  		&& (mips_cpu != 4300
  		    && mips_cpu != 4100

Index: gas/doc/c-mips.texi
===================================================================
RCS file: /cvs/src//src/gas/doc/c-mips.texi,v
retrieving revision 1.4
diff -p -r1.4 c-mips.texi
*** c-mips.texi	2000/02/26 01:48:35	1.4
--- c-mips.texi	2000/06/15 19:01:59
*************** understood.  Valid @var{cpu} value are:
*** 132,138 ****
--- 132,144 ----
  4600,
  4650,
  5000,
+ rm5200,
+ rm5230,
+ rm5231,
+ rm5261,
+ rm5721,
  6000,
+ rm7000,
  8000,
  10000
  @end quotation

Re: RFA: command line switches for QED series of MIPS processors.

[John Vickers]

Hi.

Can you expand on the motivation for these patches ?

In itself, it doesn't seem to offer offers any new functionality.
Identical results would be obtained by describing the QED CPUs
as R5000 (from which they are derived).

Since you only add QED CPUs, and not e.g. (NEC) VR5432, VR5264
or (IDT) RC64574, RC64575, it looks - in itself - like a marketing
exercise commisioned by QED, which needlessly complicates the
cpu options.

If you were planning further patches to support QED-specific
instructions or scheduling, you would only need two options,
since RM7000[A-Z] have one CPU core, and RM5[27][367][01][A-Z]
have another (modulo process shrinks and bug fixes).

By the logic of this patch, the 68k ports should
recognise at least
6833[01234], 6834[019], 68360 as cpu32 (not to mention all
the 68360 variants), and recognise
6800[08], 68[EHS]C000 as 68000.

What is the policy here ?   IMNSHO, we should choose a canonical name
for the implementations of a given core, maybe the manufacturer's
name for it (cpu32, arm6), or the name of the most well known
implementation.

Regards,

John.

Nick Clifton wrote:
> 
> Hi Gavin, Hi Ulf,
> 
>   I would like to get approval to check in the patch below.  It adds
>   command line switches for the QED series of MIPS processors to both
>   GCC and GAS.  (These processors are r5000 variants).
> 
> Cheers
>         Nick
> 
> 2000-06-15  Nick Clifton  <nickc@cygnus.com>
> 
>         * config/mips/mips.c (override_options): Accept RM5200,
>         RM5230, RM5231, RM5261, RM5721 and RM7000 as r5000 cpu
>         variants.
> 
>         * invoke.texi: Document newly accepted cpu variants.
[...]
> +           {
> +           case 5200:
> +           case 5230:
> +           case 5231:
> +           case 5261:
> +           case 5721:
> +           case 7000:
> +             target_flags |= MASK_MAD;
> +             mips_cpu = PROCESSOR_R5000;
> +             break;

Re: RFA: command line switches for QED series of MIPS processors.

[Jeffrey A Law]

  In message < 394BE917.D4AD78A3@dial.pipex.com >you write:
  > By the logic of this patch, the 68k ports should
  > recognise at least
  > 6833[01234], 6834[019], 68360 as cpu32 (not to mention all
  > the 68360 variants), and recognise
  > 6800[08], 68[EHS]C000 as 68000.
  > 
  > What is the policy here ?   IMNSHO, we should choose a canonical name
  > for the implementations of a given core, maybe the manufacturer's
  > name for it (cpu32, arm6), or the name of the most well known
  > implementation.
There is no strict policy on this matter that I'm aware of.

jeff

Re: RFA: command line switches for QED series of MIPS processors.

[Jeffrey A Law]

  In message < 200006172152.OAA28095@elmo.cygnus.com >you write:
  > : Since you only add QED CPUs, and not e.g. (NEC) VR5432, VR5264
  > : or (IDT) RC64574, RC64575, it looks - in itself - like a marketing
  > : exercise commisioned by QED.
  > 
  > This is probably correct.  All I know is that as part of a contract
  > with QED we (Cygnus/Red Hat) were asked to add these command line
  > switches and to contribute them back to the net community.
QED wants their users to be able to supply flags for the QED chips and
has paid Cygnus/Red Hat to make such options available to QED.

We agreed to contribute those changes to the official tree; however, if the
changes are rejected by the community, then they will not go into the
official sources.

  > There may be more optimisations to follow, but if so, I am not
  > scheduled to do the work, so I do not know what might be involved.
I have no idea.

jeff

Re: RFA: command line switches for QED series of MIPS processors.

[Nick Clifton]

Hi John,

: Can you expand on the motivation for these patches ?

I am attempting to keep the GCC sources in synch with the internal
copy of the sources at Red Hat.

: Since you only add QED CPUs, and not e.g. (NEC) VR5432, VR5264
: or (IDT) RC64574, RC64575, it looks - in itself - like a marketing
: exercise commisioned by QED.

This is probably correct.  All I know is that as part of a contract
with QED we (Cygnus/Red Hat) were asked to add these command line
switches and to contribute them back to the net community.

There may be more optimisations to follow, but if so, I am not
scheduled to do the work, so I do not know what might be involved.

As for complicating the command line, I do not really think that is a
serious problem.  I would expect that users would prefer to type in
the exact name of the cpu that they are targetting (if they know it) or
the cannonical name (if they do not).

Cheers
	Nick