Better mem*() stuff
This commit is contained in:
parent
b04ead6130
commit
5938aa9cfb
|
@ -14,19 +14,6 @@
|
|||
//
|
||||
static const char digits[36] =
|
||||
"0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ";
|
||||
|
||||
#if defined(_NEED_UTOA) || defined(_NEED_ULTOA)
|
||||
#define _S unsigned
|
||||
#else
|
||||
#define _S
|
||||
#endif
|
||||
|
||||
#if defined (_NEED_ITOA) || defined(_NEED_UTOA)
|
||||
#define _T int
|
||||
#else
|
||||
#define _T long
|
||||
#endif
|
||||
|
||||
//
|
||||
// Integer to string in any base between 2 and 36 (included)
|
||||
//
|
||||
|
|
|
@ -14,49 +14,34 @@
|
|||
// memset() family //
|
||||
//------------------------------------------//
|
||||
|
||||
//
|
||||
// Set "qwords"-many aligned qwords starting from ptr to val
|
||||
//
|
||||
static inline void *_memset_internal(void *ptr, ulong uval, size_t qwords)
|
||||
{
|
||||
size_t n;
|
||||
ulong *uptr = (ulong *)ptr;
|
||||
|
||||
// aligned memory write
|
||||
for (n = 0; n < qwords; n++) {
|
||||
*uptr++ = uval;
|
||||
}
|
||||
|
||||
return ptr;
|
||||
}
|
||||
|
||||
//
|
||||
// Set "bytes"-many bytes starting from ptr to val
|
||||
//
|
||||
void *memset(void *ptr, int val, size_t bytes)
|
||||
void *memsetb(void *ptr, int val, size_t bytes)
|
||||
{
|
||||
uchar *uptr = (uchar *)ptr;
|
||||
const size_t qwords = bytes/QWORD_SIZE;
|
||||
|
||||
// get rid of everything after the first byte
|
||||
val = val & 0xFF;
|
||||
|
||||
// deal with bytes before start of the first aligned qword
|
||||
while (((ulong)uptr % QWORD_ALIGN) > 0 && bytes--) {
|
||||
*uptr++ = (uchar)val;
|
||||
}
|
||||
|
||||
// move qword by qword
|
||||
if (qwords) {
|
||||
// we're qword-aligned now
|
||||
if (bytes > QWORD_SIZE) {
|
||||
const ulong uval = ((ulong)val << 56) | ((ulong)val << 48)
|
||||
| ((ulong)val << 40) | ((ulong)val << 32)
|
||||
| ((ulong)val << 24) | ((ulong)val << 16)
|
||||
| ((ulong)val << 8) | ((ulong)val);
|
||||
|
||||
_memset_internal(uptr, uval, qwords);
|
||||
ulong *uqptr = (ulong *)ptr;
|
||||
|
||||
uptr = (uchar *) ( (ulong)uptr / (qwords * QWORD_SIZE) );
|
||||
bytes %= QWORD_SIZE;
|
||||
// move qword by qword
|
||||
while (bytes > QWORD_SIZE) {
|
||||
*uqptr++ = uval;
|
||||
bytes -= QWORD_SIZE;
|
||||
}
|
||||
|
||||
uptr = (uchar *)(ulong)uqptr;
|
||||
}
|
||||
|
||||
// deal with what's left
|
||||
|
@ -74,10 +59,7 @@ void *memsetw(void *ptr, int val, size_t words)
|
|||
{
|
||||
ushort *uptr = (ushort *)ptr;
|
||||
|
||||
// get rid of everything after the first word
|
||||
val = val & 0xFFFF;
|
||||
|
||||
// can we do this an aligned way?
|
||||
// can't we do this an aligned way?
|
||||
if unlikely (((ulong)uptr % WORD_ALIGN) > 0) {
|
||||
// no, we can't align ourselves
|
||||
while (words--) {
|
||||
|
@ -93,17 +75,20 @@ void *memsetw(void *ptr, int val, size_t words)
|
|||
*uptr++ = (ushort)val;
|
||||
}
|
||||
|
||||
const size_t qwords = (words * WORD_SIZE)/QWORD_SIZE;
|
||||
|
||||
// move qword by qword
|
||||
if (qwords) {
|
||||
// we're aligned for sure
|
||||
if (words > QWORDS_TO_WORDS(1)) {
|
||||
const ulong uval = ((ulong)val << 48) | ((ulong)val << 32)
|
||||
| ((ulong)val << 16) | ((ulong)val);
|
||||
|
||||
_memset_internal(uptr, uval, qwords);
|
||||
ulong *uqptr = (ulong *)uptr;
|
||||
|
||||
uptr += qwords * QWORD_SIZE / WORD_SIZE;
|
||||
words %= QWORD_SIZE / WORD_SIZE;
|
||||
// move qword by qword
|
||||
while (words > QWORDS_TO_WORDS(1)) {
|
||||
words -= QWORDS_TO_WORDS(1);
|
||||
*uqptr++ = uval;
|
||||
}
|
||||
|
||||
uptr = (ushort *)(ulong)uqptr;
|
||||
}
|
||||
|
||||
// deal with what's left
|
||||
|
@ -131,23 +116,79 @@ void *memsetd(void *ptr, int val, size_t dwords)
|
|||
//
|
||||
void *memsetq(void *ptr, long val, size_t qwords)
|
||||
{
|
||||
return _memset_internal(ptr, (ulong)val, qwords);
|
||||
ulong *uptr = (ulong *)ptr;
|
||||
|
||||
while (qwords--) *uptr++ = (ulong)val;
|
||||
|
||||
return ptr;
|
||||
}
|
||||
|
||||
//
|
||||
// Set "bytes"-many bytes starting from ptr to 0
|
||||
//
|
||||
// WARNING
|
||||
// Assume "bytes" is large, for small sizes
|
||||
// use memset(ptr, 0, bytes) directly
|
||||
//
|
||||
void *memzero(void *ptr, size_t bytes)
|
||||
{
|
||||
// is direct aligned access possible? (is "unlikely" good here?)
|
||||
if unlikely (bytes % QWORD_SIZE && (ulong)ptr % QWORD_ALIGN) {
|
||||
return _memset_internal(ptr, (ulong)0, bytes/QWORD_SIZE);
|
||||
}
|
||||
|
||||
if unlikely (bytes % WORD_SIZE && (ulong)ptr % WORD_ALIGN) {
|
||||
return memsetw(ptr, (int)0, bytes/WORD_SIZE);
|
||||
}
|
||||
|
||||
return memset(ptr, 0, bytes);
|
||||
return memsetb(ptr, 0, bytes);
|
||||
}
|
||||
|
||||
|
||||
//
|
||||
// Copy "bytes"-many bytes of src to dst
|
||||
// Does not deal with overlapping blocks (memmove's job)
|
||||
//
|
||||
void *memcpy(void *dst, const void *src, size_t bytes)
|
||||
{
|
||||
const ulong *usrc = (const ulong *)src;
|
||||
ulong *udst = (ulong *)dst;
|
||||
|
||||
if unlikely (bytes == 0) return dst;
|
||||
|
||||
// can we align them both at once?
|
||||
if unlikely ((ulong)src % WORD_ALIGN == 1
|
||||
&& (ulong)dst % WORD_ALIGN == 1) {
|
||||
const uchar *ubsrc = (const uchar *)usrc;
|
||||
uchar *ubdst = (uchar *)udst;
|
||||
|
||||
*ubdst++ = *ubsrc++;
|
||||
bytes--;
|
||||
|
||||
udst = (ulong *)ubdst;
|
||||
usrc = (ulong *)ubsrc;
|
||||
}
|
||||
|
||||
const ushort *uwsrc = (const ushort *)usrc;
|
||||
ushort *uwdst = (ushort *)udst;
|
||||
|
||||
// align either dst or src for qword access
|
||||
while ((ulong)dst % QWORD_ALIGN > 0
|
||||
&& (ulong)src % QWORD_ALIGN > 0
|
||||
&& bytes > WORD_SIZE) {
|
||||
|
||||
*uwdst++ = *uwsrc++;
|
||||
bytes -= WORD_SIZE;
|
||||
}
|
||||
|
||||
udst = (ulong *)uwdst;
|
||||
usrc = (ulong *)uwsrc;
|
||||
|
||||
// should be most of the job
|
||||
while (bytes > QWORD_SIZE) {
|
||||
*udst++ = *usrc++;
|
||||
bytes -= QWORD_SIZE;
|
||||
}
|
||||
|
||||
const uchar *ubsrc = (const uchar *)usrc;
|
||||
ushort *ubdst = (ushort *)udst;
|
||||
|
||||
// deal with what's left
|
||||
while (bytes--) {
|
||||
*ubdst ++ = *ubsrc++;
|
||||
}
|
||||
|
||||
return dst;
|
||||
}
|
||||
|
||||
|
|
|
@ -1,37 +0,0 @@
|
|||
//----------------------------------------------------------------------------//
|
||||
// GNU GPL OS/K //
|
||||
// //
|
||||
// Authors: spectral` //
|
||||
// NeoX //
|
||||
// //
|
||||
// Desc: mem*() functions, suboptimal edition //
|
||||
//----------------------------------------------------------------------------//
|
||||
|
||||
#include <kaleid.h>
|
||||
|
||||
//
|
||||
// Set "bytes"-many bytes starting from ptr to val
|
||||
//
|
||||
void *memset(void *ptr, int val, size_t bytes)
|
||||
{
|
||||
uchar uval = val & 0xFF;
|
||||
uchar *uptr = (uchar *)ptr;
|
||||
|
||||
while (bytes--) *uptr++ = uval;
|
||||
|
||||
return ptr;
|
||||
}
|
||||
|
||||
//
|
||||
// Set "bytes"-many bytes starting from ptr to 0
|
||||
//
|
||||
void *memzero(void *ptr, size_t bytes)
|
||||
{
|
||||
uchar *uptr = (uchar *)ptr;
|
||||
|
||||
while (bytes--) *uptr++ = 0;
|
||||
|
||||
return ptr;
|
||||
}
|
||||
|
||||
|
|
@ -41,6 +41,22 @@
|
|||
# define LONG_BIT QWORD_BIT
|
||||
#endif
|
||||
|
||||
#ifndef DATA_SHIFTS_BLOCK
|
||||
#define DATA_SHIFTS_BLOCK
|
||||
# define BYTES_TO_WORDS(B) ((B) >> 1)
|
||||
# define BYTES_TO_DWORDS(B) ((B) >> 2)
|
||||
# define BYTES_TO_QWORDS(B) ((B) >> 3)
|
||||
# define WORDS_TO_BYTES(W) ((W) << 1)
|
||||
# define WORDS_TO_DWORDS(W) ((W) >> 1)
|
||||
# define WORDS_TO_QWORDS(W) ((W) >> 2)
|
||||
# define DWORDS_TO_BYTES(D) ((D) << 2)
|
||||
# define DWORDS_TO_WORDS(D) ((D) << 1)
|
||||
# define DWORDS_TO_QWORDS(D) ((D) >> 1)
|
||||
# define QWORDS_TO_BYTES(Q) ((Q) << 3)
|
||||
# define QWORDS_TO_WORDS(Q) ((Q) << 2)
|
||||
# define QWORDS_TO_DWORDS(Q) ((Q) << 1)
|
||||
#endif
|
||||
|
||||
//------------------------------------------//
|
||||
// Numeric data limits //
|
||||
//------------------------------------------//
|
||||
|
|
Loading…
Reference in New Issue