* Makefile.am (lib_a_SOURCES, SH64 case): Add memcpy.S, memset.S

	and strcpy.S.
	* Makefile.in: Regenerate.
	* asm.h (_ENTRY): Set SH5 alignment to 8 bytes.
	(SHHI, SHLO): Define.
	* memcpy.S: Add code for SH5.
	* memset.S: Likewise.
	* strcpy.S: Likewise.

1 files changed, 109 insertions, 23 deletions

diff --git a/newlib/libc/machine/sh/strcpy.S b/newlib/libc/machine/sh/strcpy.S
index 63aa7e75b..aab421007 100644
--- a/newlib/libc/machine/sh/strcpy.S
+++ b/newlib/libc/machine/sh/strcpy.S
@@ -1,41 +1,127 @@
-! Entry: r4: destination
-!        r5: source
-! Exit:  r0: destination
-!        r1-r2,r5: clobbered
+! Entry: arg0: destination
+!        arg1: source
+! Exit:  result: destination
+!
+! SH5 code Copyright 2002 SuperH Ltd.
 
 #include "asm.h"
 
 ENTRY(strcpy)
-	mov     r4,r0
-	or      r5,r0
+
+#if __SHMEDIA__
+
+	pta/l shortstring,tr1
+	ldlo.q r3,0,r4
+	ptabs r18,tr4
+	shlli r3,3,r7
+	add r2,r63,r0
+	mcmpeq.b r4,r63,r6
+	SHHI r6,r7,r6
+	bnei/u r6,0,tr1 // shortstring
+	pta/l no_lddst, tr2
+	ori r3,-8,r23
+	sub r2, r23, r0
+	sub r3, r2, r21
+	addi r21, 8, r20
+	ldx.q r0, r21, r5
+	pta/l loop, tr0
+	ori r2,-8,r22
+	mcmpeq.b r5, r63, r6
+	bge/u r22, r23, tr2 // no_lddst
+
+	// r22 < r23 ;  Need to do a load from the destination.
+	ldlo.q r2, 0, r9
+	movi -1, r8
+	SHLO r8, r7, r8
+	mcmv r4, r8, r9
+	stlo.q r2, 0, r9
+	beqi/l r6, 0, tr0 // loop
+
+	add r5, r63, r4
+	blink tr1, r63 // shortstring
+no_lddst:
+	stlo.q r2, 0, r4
+	SHHI r4, r7, r4
+	sthi.q r0, -1, r4
+	beqi/l r6, 0, tr0 // loop
+
+	add r5, r63, r4
+shortstring:
+#ifndef LITTLE_ENDIAN
+	pta/l shortstring2,tr1
+	byterev r4,r4
+#endif
+shortstring2:
+	st.b r0,0,r4
+	andi r4,0xff,r5
+	shlri r4,8,r4
+	addi r0,1,r0
+	bnei/l r5,0,tr1
+	blink tr4,r63 // return
+	
+	.balign 8
+loop:
+	stlo.q r0, 0, r5
+	ldx.q r0, r20, r4
+	addi r0, 16, r0
+	sthi.q r0, -9, r5
+	mcmpeq.b r4, r63, r6
+	bnei/u r6, 0, tr1 // shortstring
+	ldx.q r0, r21, r5
+	stlo.q r0, -8, r4
+	sthi.q r0, -1, r4
+	mcmpeq.b r5, r63, r6
+	beqi/l r6, 0, tr0 // loop
+
+	add r5, r63, r4
+	blink tr1, r63 // shortstring
+
+#else /* ! __SHMEDIA__, i.e. SH 1..4 / SHcompact */
+
+#ifdef __SH5__
+#define DST r2
+#define SRC r3
+#define TMP r4
+#define RESULT R2
+!        r0,r1,r3,r4: clobbered
+#else
+#define DST r4
+#define SRC r5
+#define TMP r2
+#define RESULT r0
+!        r1-r2,r5: clobbered
+#endif
+	mov     DST,r0
+	or      SRC,r0
 	tst	#3,r0
-	SL(bf, L_setup_char_loop, mov r4,r0)
-	mov.l   @r5+,r1
-	mov     #0,r2
-	cmp/str r2,r1
-	SL(bt, Longword_loop_end, sub r5,r0)
+	SL(bf, L_setup_char_loop, mov DST,r0)
+	mov.l   @SRC+,r1
+	mov     #0,TMP
+	cmp/str TMP,r1
+	SL(bt, Longword_loop_end, sub SRC,r0)
 	.align  2
 Longword_loop:
-	mov.l   r1,@(r0,r5)
-	mov.l   @r5+,r1
-	cmp/str r2,r1
+	mov.l   r1,@(r0,SRC)
+	mov.l   @SRC+,r1
+	cmp/str TMP,r1
 	bt      Longword_loop_end
-	mov.l   r1,@(r0,r5)
-	mov.l   @r5+,r1
-	cmp/str r2,r1
+	mov.l   r1,@(r0,SRC)
+	mov.l   @SRC+,r1
+	cmp/str TMP,r1
 	bf      Longword_loop
 Longword_loop_end:
-	add	#-4,r5
+	add	#-4,SRC
 	add	#3,r0
 	.align  2
 L_char_loop:
-	mov.b	@r5+,r1
+	mov.b	@SRC+,r1
 L_char_loop_start:
 	tst	r1,r1
-	SL(bf, L_char_loop, mov.b r1,@(r0,r5))
+	SL(bf, L_char_loop, mov.b r1,@(r0,SRC))
 	rts
-	mov r4,r0
+	mov DST,RESULT
 L_setup_char_loop:
-	mov.b	@r5+,r1
+	mov.b	@SRC+,r1
 	bra L_char_loop_start
-	sub r5,r0
+	sub SRC,r0
+#endif /* ! __SHMEDIA__ */