Test versions of getVarint functions. The updates essentially utilize loop unrolling...

FossilOrigin-Name: 682dc24dbe82d0326377e27c5ff97db3499873b0

diff --git a/manifest b/manifest
index 6271f7af8307469afe5e9ecf88807e8e9c613542..8a8f12e61f16175047574a7f26b5f76dde62b4f0 100644 (file)
--- a/manifest
+++ b/manifest
@@ -1,5 +1,5 @@
-C Add\scomment\sto\sspeculate\swhen\ssetting\sjournal_mode=OFF\son\sVACUUM\sdoes\snot\nhelp\sperformance.\s\sNo\schanges\sto\scode.\s(CVS\s5071)
-D 2008-04-30T16:38:23
+C Test\sversions\sof\sgetVarint\sfunctions.\sThe\supdates\sessentially\sutilize\sloop\sunrolling\sand\ssome\sshifting/anding\stricks\sto\sminimize\sthe\snumber\sof\slogical\soperations\srequired.\s(CVS\s5072)
+D 2008-05-01T02:47:04
 F Makefile.arm-wince-mingw32ce-gcc ac5f7b2cef0cd850d6f755ba6ee4ab961b1fadf7
 F Makefile.in 25b3282a4ac39388632c2fb0e044ff494d490952
 F Makefile.linux-gcc d53183f4aa6a9192d249731c90dbdffbd2c68654
@@ -167,7 +167,7 @@ F src/tokenize.c 8d77af8584cf027dc21375f0efa5818cb303c995
 F src/trigger.c 9bd3b6fa0beff4a02d262c96466f752ec15a7fc3
 F src/update.c 2d7143b9014e955509cc4f323f9a9584fb898f34
 F src/utf.c 8c94fa10efc78c2568d08d436acc59df4df7191b
-F src/util.c a3907b05dcc3720a6d71bb39e61d67b0d994b51f
+F src/util.c 43a77ab79275991b819428ded8ac8dc868604ac7
 F src/vacuum.c c3b2b70677f874102b8753bf494c232e777f3998
 F src/vdbe.c 26964ba7ed76d2a1c52747d601aaf2dc5b09b651
 F src/vdbe.h bfd84bda447f39cb599302c7ec85067dae20453c
@@ -633,7 +633,7 @@ F www/tclsqlite.tcl 8be95ee6dba05eabcd27a9d91331c803f2ce2130
 F www/vdbe.tcl 87a31ace769f20d3627a64fa1fade7fed47b90d0
 F www/version3.tcl 890248cf7b70e60c383b0e84d77d5132b3ead42b
 F www/whentouse.tcl fc46eae081251c3c181bd79c5faef8195d7991a5
-P aa59974ec15508d69c5b65ab89ec7bc32690018c
-R 75f14acd3f2bae6b9a939b5711d86e5b
-U drh
-Z 924c040e6c27c40e097c580fa3feeda2
+P 9c8b4babb2222a5e0c2ecf7a116b7df90084c81d
+R 329903c79b08ceeef301067fa9d79cfe
+U shane
+Z 34ccf6f71b1f112c4e41d623ea9e6a2c

diff --git a/manifest.uuid b/manifest.uuid
index 340a72eb2e9cdc2ec4b4ea194ba4466fce1de481..ac800c6027f8ac7c0b44bc1643b82d025cb20109 100644 (file)
--- a/manifest.uuid
+++ b/manifest.uuid
@@ -1 +1 @@
-9c8b4babb2222a5e0c2ecf7a116b7df90084c81d
\ No newline at end of file
+682dc24dbe82d0326377e27c5ff97db3499873b0
\ No newline at end of file

diff --git a/src/util.c b/src/util.c
index 1aa375bdf9a96204872d015e680668191d56372d..46dc5cb531605d7be8ec255f7a6f1608fd219184 100644 (file)
--- a/src/util.c
+++ b/src/util.c
@@ -14,7 +14,7 @@
 ** This file contains functions for allocating memory, comparing
 ** strings, and stuff like that.
 **
-** $Id: util.c,v 1.224 2008/04/28 17:41:31 shane Exp $
+** $Id: util.c,v 1.225 2008/05/01 02:47:04 shane Exp $
 */
 #include "sqliteInt.h"
 #include <stdarg.h>
@@ -533,41 +533,158 @@ int sqlite3PutVarint32(unsigned char *p, u32 v){
 ** Return the number of bytes read.  The value is stored in *v.
 */
 int sqlite3GetVarint(const unsigned char *p, u64 *v){
-  u32 x;
-  u64 x64;
-  int n;
-  unsigned char c;
-  if( ((c = p[0]) & 0x80)==0 ){
-    *v = c;
+  u32 a,b,s;
+
+  a = *p;
+  // a: p0 (unmasked)
+  if (!(a&0x80))
+  {
+    *v = a;
     return 1;
   }
-  x = c & 0x7f;
-  if( ((c = p[1]) & 0x80)==0 ){
-    *v = (x<<7) | c;
+
+  p++;
+  b = *p;
+  // b: p1 (unmasked)
+  if (!(b&0x80))
+  {
+    a &= 0x7f;
+    a = a<<7;
+    a |= b;
+    *v = a;
     return 2;
   }
-  x = (x<<7) | (c&0x7f);
-  if( ((c = p[2]) & 0x80)==0 ){
-    *v = (x<<7) | c;
+
+  p++;
+  a = a<<14;
+  a |= *p;
+  // a: p0<<14 | p2 (unmasked)
+  if (!(a&0x80))
+  {
+    a &= (0x7f<<14)|(0x7f);
+    b &= 0x7f;
+    b = b<<7;
+    a |= b;
+    *v = a;
     return 3;
   }
-  x = (x<<7) | (c&0x7f);
-  if( ((c = p[3]) & 0x80)==0 ){
-    *v = (x<<7) | c;
+
+  // CSE1 from below
+  a &= (0x7f<<14)|(0x7f);
+  p++;
+  b = b<<14;
+  b |= *p;
+  // b: p1<<14 | p3 (unmasked)
+  if (!(b&0x80))
+  {
+    b &= (0x7f<<14)|(0x7f);
+    // moved CSE1 up
+    // a &= (0x7f<<14)|(0x7f);
+    a = a<<7;
+    a |= b;
+    *v = a;
     return 4;
   }
-  x64 = (x<<7) | (c&0x7f);
-  n = 4;
-  do{
-    c = p[n++];
-    if( n==9 ){
-      x64 = (x64<<8) | c;
-      break;
-    }
-    x64 = (x64<<7) | (c&0x7f);
-  }while( (c & 0x80)!=0 );
-  *v = x64;
-  return n;
+
+  // a: p0<<14 | p2 (masked)
+  // b: p1<<14 | p3 (unmasked)
+  // 1:save off p0<<21 | p1<<14 | p2<<7 | p3 (masked)
+  // moved CSE1 up
+  // a &= (0x7f<<14)|(0x7f);
+  b &= (0x7f<<14)|(0x7f);
+  s = a;
+  // s: p0<<14 | p2 (masked)
+
+  p++;
+  a = a<<14;
+  a |= *p;
+  // a: p0<<28 | p2<<14 | p4 (unmasked)
+  if (!(a&0x80))
+  {
+    // we can skip these cause they were (effectively) done above in calc'ing s
+    // a &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+    // b &= (0x7f<<14)|(0x7f);
+    b = b<<7;
+    a |= b;
+    s = s>>18;
+    *v = ((u64)s)<<32 | a;
+    return 5;
+  }
+
+  // 2:save off p0<<21 | p1<<14 | p2<<7 | p3 (masked)
+  s = s<<7;
+  s |= b;
+  // s: p0<<21 | p1<<14 | p2<<7 | p3 (masked)
+
+  p++;
+  b = b<<14;
+  b |= *p;
+  // b: p1<<28 | p3<<14 | p5 (unmasked)
+  if (!(b&0x80))
+  {
+    // we can skip this cause it was (effectively) done above in calc'ing s
+    // b &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+    a &= (0x7f<<14)|(0x7f);
+    a = a<<7;
+    a |= b;
+    s = s>>18;
+    *v = ((u64)s)<<32 | a;
+    return 6;
+  }
+
+  p++;
+  a = a<<14;
+  a |= *p;
+  // a: p2<<28 | p4<<14 | p6 (unmasked)
+  if (!(a&0x80))
+  {
+    a &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+    b &= (0x7f<<14)|(0x7f);
+    b = b<<7;
+    a |= b;
+    s = s>>11;
+    *v = ((u64)s)<<32 | a;
+    return 7;
+  }
+
+  // CSE2 from below
+  a &= (0x7f<<14)|(0x7f);
+  p++;
+  b = b<<14;
+  b |= *p;
+  // b: p3<<28 | p5<<14 | p7 (unmasked)
+  if (!(b&0x80))
+  {
+    b &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+    // moved CSE2 up
+    // a &= (0x7f<<14)|(0x7f);
+    a = a<<7;
+    a |= b;
+    s = s>>4;
+    *v = ((u64)s)<<32 | a;
+    return 8;
+  }
+
+  p++;
+  a = a<<15;
+  a |= *p;
+  // a: p4<<29 | p6<<15 | p8 (unmasked)
+
+  // moved CSE2 up
+  // a &= (0x7f<<29)|(0x7f<<15)|(0xff);
+  b &= (0x7f<<14)|(0x7f);
+  b = b<<8;
+  a |= b;
+
+  s = s<<4;
+  b = p[-4];
+  b &= 0x7f;
+  b = b>>3;
+  s |= b;
+
+  *v = ((u64)s)<<32 | a;
+
+  return 9;
 }
 
 /*
@@ -578,27 +695,117 @@ int sqlite3GetVarint(const unsigned char *p, u64 *v){
 ** this function assumes the single-byte case has already been handled.
 */
 int sqlite3GetVarint32(const unsigned char *p, u32 *v){
-  u32 x;
-  int n;
-  unsigned char c;
+  u32 a,b;
+
+  a = *p;
+  // a: p0 (unmasked)
 #ifndef getVarint32
-  if( ((signed char*)p)[0]>=0 ){
-    *v = p[0];
+  if (!(a&0x80))
+  {
+    *v = a;
     return 1;
   }
 #endif
-  x = p[0] & 0x7f;
-  if( ((signed char*)p)[1]>=0 ){
-    *v = (x<<7) | p[1];
+
+  p++;
+  b = *p;
+  // b: p1 (unmasked)
+  if (!(b&0x80))
+  {
+    a &= 0x7f;
+    a = a<<7;
+    *v = a | b;
     return 2;
   }
-  x = (x<<7) | (p[1] & 0x7f);
-  n = 2;
-  do{
-    x = (x<<7) | ((c = p[n++])&0x7f);
-  }while( (c & 0x80)!=0 && n<9 );
-  *v = x;
-  return n;
+
+  p++;
+  a = a<<14;
+  a |= *p;
+  // a: p0<<14 | p2 (unmasked)
+  if (!(a&0x80))
+  {
+    a &= (0x7f<<14)|(0x7f);
+    b &= 0x7f;
+    b = b<<7;
+    *v = a | b;
+    return 3;
+  }
+
+  p++;
+  b = b<<14;
+  b |= *p;
+  // b: p1<<14 | p3 (unmasked)
+  if (!(b&0x80))
+  {
+    b &= (0x7f<<14)|(0x7f);
+    a &= (0x7f<<14)|(0x7f);
+    a = a<<7;
+    *v = a | b;
+    return 4;
+  }
+
+  p++;
+  a = a<<14;
+  a |= *p;
+  // a: p0<<28 | p2<<14 | p4 (unmasked)
+  if (!(a&0x80))
+  {
+    a &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+    b &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+    b = b<<7;
+    *v = a | b;
+    return 5;
+  }
+
+  p++;
+  b = b<<14;
+  b |= *p;
+  // b: p1<<28 | p3<<14 | p5 (unmasked)
+  if (!(b&0x80))
+  {
+    b &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+    a &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+    a = a<<7;
+    *v = a | b;
+    return 6;
+  }
+
+  p++;
+  a = a<<14;
+  a |= *p;
+  // a: p2<<28 | p4<<14 | p6 (unmasked)
+  if (!(a&0x80))
+  {
+    a &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+    b &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+    b = b<<7;
+    *v = a | b;
+    return 7;
+  }
+
+  p++;
+  b = b<<14;
+  b |= *p;
+  // b: p3<<28 | p5<<14 | p7 (unmasked)
+  if (!(b&0x80))
+  {
+    b &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+    a &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+    a = a<<7;
+    *v = a | b;
+    return 8;
+  }
+
+  p++;
+  a = a<<14;
+  a |= *p;
+  // a: p4<<28 | p6<<14 | p8 (unmasked)
+
+  a &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+  b &= (0x7f<<28)|(0x7f<<14)|(0x7f);
+  b = b<<7;
+  *v = a | b;
+  return 9;
 }
 
 /*