some constness stuff for buffer_0, an experimental optimization for

byte_copy and add buffer_putspace, a trivial shortcut.

Makefile

@@ -2,9 +2,9 @@ all: t byte.a fmt.a scan.a str.a uint.a open.a stralloc.a unix.a socket.a buffer
 
 VPATH=str:byte:fmt:scan:uint:open:stralloc:unix:socket:buffer:mmap
 
-CC=egcc
+CC=gcc
 #CFLAGS=-I. -pipe -Wall -Os -march=pentiumpro -fomit-frame-pointer -fschedule-insns2 -Wall
-CFLAGS=-I. -I../dietlibc/include -pipe -Wall -Os -march=athlon -mcpu=athlon -malign-functions=2 -fschedule-insns2 -g
+CFLAGS=-I. -I../dietlibc/include -pipe -Wall -Os -malign-functions=2 -fschedule-insns2 -g
 #CFLAGS=-I../dietlibc/include -I. -pipe -Wall -Os -march=pentiumpro -mcpu=athlon -fomit-frame-pointer -fschedule-insns2 -Wall
 #CFLAGS=-I../dietlibc/include -pipe -Os -march=pentiumpro -mcpu=pentiumpro -fomit-frame-pointer -fschedule-insns2 -Wall
 

buffer.h

@@ -25,6 +25,8 @@ extern int buffer_puts(buffer* b,const char* x);
 extern int buffer_putsalign(buffer* b,const char* x);
 extern int buffer_putsflush(buffer* b,const char* x);
 
+extern int buffer_putspace(buffer* b);
+
 #define buffer_PUTC(s,c) \
   ( ((s)->n != (s)->p) \
     ? ( (s)->x[(s)->p++] = (c), 0 ) \

buffer/buffer_0.c

@@ -1,7 +1,7 @@
 #include <unistd.h>
 #include "buffer.h"
 
-static int b0read(int fd,const char* buf, unsigned int len) {
+static int b0read(int fd,char* buf, unsigned int len) {
   if (buffer_flush(buffer_1)<0) return -1;
   return read(fd,buf,len);
 }

buffer/buffer_0small.c

@@ -1,7 +1,7 @@
 #include <unistd.h>
 #include "buffer.h"
 
-static int b0read(int fd,const char* buf, unsigned int len) {
+static int b0read(int fd,char* buf, unsigned int len) {
   if (buffer_flush(buffer_1small)<0) return -1;
   return read(fd,buf,len);
 }

buffer/buffer_putspace.c

@@ -0,0 +1,7 @@
+#include "str.h"
+#include "buffer.h"
+
+int buffer_putspace(buffer* b) {
+  static char space=' ';
+  return buffer_put(b,&space,1);
+}

byte/byte_copy.c

@@ -6,6 +6,21 @@ void byte_copy(void* out, unsigned int len, const void* in) {
   register char* s=out;
   register const char* t=in;
   register const char* u=in+len;
+  if (len>127) {
+    if (sizeof(unsigned long)>4) {	/* a good compiler should optimize this check away */
+      for (;(unsigned long)t&7;) {
+	if (t==u) break; *s=*t; ++s; ++t;
+      }
+    } else {
+      for (;(unsigned long)t&3;) {
+	if (t==u) break; *s=*t; ++s; ++t;
+      }
+    }
+    while (t+sizeof(long)<=u) {
+      *(unsigned long*)s=*(unsigned long*)t;
+      s+=sizeof(long); t+=sizeof(long);
+    }
+  }
   for (;;) {
     if (t==u) break; *s=*t; ++s; ++t;
     if (t==u) break; *s=*t; ++s; ++t;

t.c

@@ -14,6 +14,22 @@
      __asm__ __volatile__ ("rdtsc" : "=a" (low) : : "edx")
 
 int main(int argc,char* argv[]) {
+  buffer_putspace(buffer_1);
+  buffer_flush(buffer_1);
+#if 0
+  long a,b,c;
+  char buf[4096];
+  char buf2[4096];
+  memcpy(buf,buf2,4096);
+  byte_copy(buf,4096,buf2);
+  rdtscl(a);
+  memcpy(buf,buf2,4096);
+  rdtscl(b);
+  byte_copy(buf,4096,buf2);
+  rdtscl(c);
+  printf("memcpy: %d - byte_copy: %d\n",b-a,c-b);
+#endif
+#if 0
   char ip[16];
   int i;
   if ((i=scan_ip6(argv[1],ip))) {
@@ -21,6 +37,7 @@ int main(int argc,char* argv[]) {
     buf[fmt_ip6(buf,ip)]=0;
     puts(buf);
   }
+#endif
 #if 0
   char buf[100];
   strcpy(buf,"foobarbaz");