This is the mail archive of the newlib@sourceware.cygnus.com mailing list for the newlib project.


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

illegal code for SH-1 cpu


Hello!

I've compiled newlib-1.8.1 for use with a SH-7032 board (SH-1 cpu kernel):

./configure -v --target=sh-coff --prefix=/usr/local/sh-coff
--host=i586-unknown-linux --norecursion

Debugging some initial test programs (I'm not very experienced with gcc
used for embedded systems) I got illegal opcode messages and found that
memset.S from newlib/libc/machine/sh indeed uses dt <reg> commands which
are only available for SH-2 and better cpu kernels.

My workaround was a substitution of the dt opcodes:

!
! Fast SH memset
!
! by Toshiyasu Morita (tm@netcom.com)
!
! Entry: r4: destination pointer
!        r5: fill value
!        r6: byte count
!
! Exit:  r0-r3: trashed
!

#include "asm.h"

ENTRY(memset)
	mov	r4,r3	! Save return value

	mov	r6,r0	! Check explicitly for zero
	cmp/eq	#0,r0
	bt	L_exit

	mov	#12,r0	! Check for small number of bytes
	cmp/gt	r6,r0
	bt	L_store_byte_loop

	neg	r4,r0	! Align destination
	add	#4,r0
	and	#3,r0
	tst	r0,r0
	bt	L_dup_bytes
	.balignw 4,0x0009
L_align_loop:
	mov.b	r5,@r4
	add	#-1,r6
	add	#1,r4
	add     #-1,r0  ! Substitute for dt r0
	cmp/eq  #0,r0   ! Substitute for dt r0
	bf	L_align_loop

L_dup_bytes:
	extu.b	r5,r5	! Duplicate bytes across longword
	swap.b	r5,r0
	or	r0,r5
	swap.w	r5,r0
	or	r0,r5

	mov	r6,r2	! Calculate number of double longwords
	shlr2	r2
	shlr	r2

	.balignw 4,0x0009
L_store_long_loop:
	mov.l	r5,@r4	! Store double longs to memory
        add     #-1,r2  ! Substitute for dt r2
	tst     r2,r2   ! Substitute for dt r2
	mov.l	r5,@(4,r4)
	add	#8,r4
	bf	L_store_long_loop

	mov	#7,r0
	and	r0,r6
	tst	r6,r6
	bt	L_exit
	.balignw 4,0x0009
L_store_byte_loop:
	mov.b	r5,@r4	! Store bytes to memory
	add	#1,r4
	add     #-1,r6  ! Substitute for dt r6
	tst     r6,r6   ! Substitute for dt r6
	bf	L_store_byte_loop

L_exit:
	rts
	mov	r3,r0

Thank you very much for this really good development system!

Yours,

Jens Decker


-- 

___________________________________________________________

* Jens Decker                                             *
* Universitaet Regensburg                                 *
* Institut fuer Physikalische und Theoretische Chemie     *
* Lehrst. Prof. Dick                                      *
* Universitaetstr. 31                                     *
* 93059 Regensburg                                        *
*                                                         *
* Tel: +49-941/943-4473, -4458 Fax: -4488  Room: 22.2.14  *
* Email: Jens.Decker@chemie.uni-regensburg.de             *
* WWW (coming up): http://                                *
* rchsg41.chemie.uni-regensburg.de/group/Jens.Decker.html *
___________________________________________________________