patch-2.1.9 linux/arch/sparc/lib/memset.S

• Lines: 167
• Date: Sat Nov 9 10:12:05 1996
• Orig file: v2.1.8/linux/arch/sparc/lib/memset.S
• Orig date: Thu Jan 1 02:00:00 1970

diff -u --recursive --new-file v2.1.8/linux/arch/sparc/lib/memset.S linux/arch/sparc/lib/memset.S
@@ -0,0 +1,166 @@
+/* linux/arch/sparc/lib/memset.S: Sparc optimized memset and bzero code
+ * Hand optimized from GNU libc's memset
+ * Copyright (C) 1991,1996 Free Software Foundation
+ * Copyright (C) 1996 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
+ * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
+ */
+
+#include <asm/cprefix.h>
+#include <asm/ptrace.h>
+
+#define HANDLE_UNALIGNED 1
+
+	/* Store 64 bytes at (BASE + OFFSET) using value SOURCE. */
+#define ZERO_BIG_BLOCK(base, offset, source)    \
+	std	source, [base + offset + 0x00]; \
+	std	source, [base + offset + 0x08]; \
+	std	source, [base + offset + 0x10]; \
+	std	source, [base + offset + 0x18]; \
+	std	source, [base + offset + 0x20]; \
+	std	source, [base + offset + 0x28]; \
+	std	source, [base + offset + 0x30]; \
+	std	source, [base + offset + 0x38];
+
+#define ZERO_LAST_BLOCKS(base, offset, source)	\
+	std	source, [base - offset - 0x38]; \
+	std	source, [base - offset - 0x30]; \
+	std	source, [base - offset - 0x28]; \
+	std	source, [base - offset - 0x20]; \
+	std	source, [base - offset - 0x18]; \
+	std	source, [base - offset - 0x10]; \
+	std	source, [base - offset - 0x08]; \
+	std	source, [base - offset - 0x00];
+
+	.text
+	.align 4
+
+	.globl	C_LABEL(__bzero), C_LABEL(__memset), C_LABEL(memset)
+C_LABEL(__memset):
+C_LABEL(memset):
+	and	%o1, 0xff, %g3
+	sll	%g3, 8, %g2
+	or	%g3, %g2, %g3
+	sll	%g3, 16, %g2
+	or	%g3, %g2, %g3
+	b	1f
+	 mov	%o2, %o1
+
+#if HANDLE_UNALIGNED
+/* As this is highly unprobable, we optimize the other case (4 aligned)
+ * Define HANDLE_UNALIGNED to 0, if all the alignment work is done by
+ * the trap. Then we have to hope nobody will memset something unaligned
+ * with large counts, as this would lead to a lot of traps...
+ */
+3:
+	cmp	%o2, 3
+	be	2f
+	 stb	%g3, [%o0]
+
+	cmp	%o2, 2
+	be	2f
+	 stb	%g3, [%o0 + 0x01]
+
+	stb	%g3, [%o0 + 0x02]
+2:
+	sub	%o2, 4, %o2
+	add	%o1, %o2, %o1
+	b	4f
+	 sub	%o0, %o2, %o0
+#endif /* HANDLE_UNALIGNED */
+
+	.globl	C_LABEL(__clear_user)
+C_LABEL(__clear_user):
+	st	%o0, [%g6 + THREAD_EX_ADDR]
+	ld	[%g6 + THREAD_EX_COUNT], %g1
+	set	clear_user_failure, %g2
+	add	%g1, 1, %g1
+	st	%o7, [%g6 + THREAD_EX_PC]
+	st	%g1, [%g6 + THREAD_EX_COUNT]
+	call	C_LABEL(__bzero)
+	 st	%g2, [%g6 + THREAD_EX_EXPC]
+
+clear_user_success:
+	ldd	[%g6 + THREAD_EX_COUNT], %g2
+	mov	0, %o0
+	sub	%g2, 1, %g1
+	jmpl	%g3 + 0x8, %g0
+	 st	%g1, [%g6 + THREAD_EX_COUNT]
+
+clear_user_failure:
+	jmpl	%g3 + 0x8, %g0
+	 mov	%g2, %o0
+
+C_LABEL(__bzero):
+	mov	%g0, %g3
+1:
+	cmp	%o1, 7
+	bleu	7f
+	 mov	%o0, %g1
+
+#if HANDLE_UNALIGNED
+	andcc	%o0, 3, %o2
+	bne	3b
+#endif /* HANDLE_UNALIGNED */
+4:
+	 andcc	%o0, 4, %g0
+
+	be	2f
+	 mov	%g3, %g2
+
+	st	%g3, [%o0]
+	sub	%o1, 4, %o1
+	add	%o0, 4, %o0
+2:
+	andcc	%o1, 0xffffff80, %o3	! Now everything is 8 aligned and o1 is len to run
+	be	9f
+	 andcc	%o1, 0x78, %o2
+4:
+	ZERO_BIG_BLOCK(%o0, 0x00, %g2)
+	subcc	%o3, 128, %o3
+	ZERO_BIG_BLOCK(%o0, 0x40, %g2)
+	bne	4b
+	 add	%o0, 128, %o0
+
+	orcc	%o2, %g0, %g0
+9:
+	be	6f
+	 andcc	%o1, 7, %o1
+
+	srl	%o2, 1, %o3
+	set	bzero_table + 64, %o4
+	sub	%o4, %o3, %o4
+	jmp	%o4
+	 add	%o0, %o2, %o0
+
+bzero_table:
+	ZERO_LAST_BLOCKS(%o0, 0x48, %g2)
+	ZERO_LAST_BLOCKS(%o0, 0x08, %g2)
+
+6:
+	be	8f
+	 andcc	%o1, 4, %g0
+
+	be	1f
+	 andcc	%o1, 2, %g0
+
+	st	%g3, [%o0]
+	add	%o0, 4, %o0
+1:
+	be	1f
+	 andcc	%o1, 1, %g0
+
+	sth	%g3, [%o0]
+	add	%o0, 2, %o0
+1:
+	bne,a	8f
+	 stb	%g3, [%o0]
+8:
+	retl
+	 mov	%g1,%o0
+
+/* Don't care about alignment here. It is highly 
+ * unprobable and at most two traps may happen
+ */
+7:
+	b	6b
+	 orcc	%o1, 0, %g0