I was able to fool the libjpeg/SIMD build into thinking I had an i386 chip, but it still doesn't want to build the assembly routines:
make[1]: *** No rule to make target `jsimdcpu.lo', needed by `libjpeg.la'. Stop. make[1]: Leaving directory `/home/drc/src/tigervnc/common/jpeg' make: *** [all] Error 2 I was, however, able to port the SIMD code into the VirtualGL build environment and build it there using the version of libjpeg in the VirtualGL source tree. From this, I got some good benchmark results and was then able to use that as a platform to add the Huffman codec optimizations from TurboJPEG into jchuff.c. I am attaching both a spreadsheet of my findings as well as a proposed patch to jchuff.c. Performance results: Without modifications, I found libjpeg/SIMD to be about twice as fast as the unaccelerated libjpeg but still about 30% slower on compression than TurboJPEG/OML (the version of TurboJPEG that uses Open mediaLib Lite.) I suspected that the TurboJPEG Huffman codec was largely responsible for this performance disparity, so I ported the TurboJPEG Huffman codec optimizations to the libjpeg/SIMD Huffman codec. The resulting optimized libjpeg/SIMD met or exceeded the performance of TurboJPEG/OML in almost all cases. Conclusion: I don't think there is any need to bother with integrating Open mediaLib Lite. I think libjpeg/SIMD should work fine for our purposes. I would like to look into making it 64-bit compatible sooner rather than later, and of course I'd love to be able to actually build it without having to use my own custom build environment. Once I can get it to build properly, I'll add the pixel format extensions and proceed up the code stack from there to look at the holistic performance of TigerVNC vs. TurboVNC. Note regarding the patch: The enhancements borrowed some code from TurboJPEG/OML, so it was necessary to license the modifications to jchuff.c using the wxWidgets Library License. This license is fully GPL-compatible, however.
tigervncvsipp.ods
Description: Binary data
Index: common/jpeg/jchuff.c
===================================================================
--- common/jpeg/jchuff.c (revision 3656)
+++ common/jpeg/jchuff.c (working copy)
@@ -14,12 +14,34 @@
* permanent JPEG objects only upon successful completion of an MCU.
*/
+/* Modifications:
+ * Copyright (C)2007 Sun Microsystems, Inc.
+ * Copyright (C)2009 D. R. Commander
+ *
+ * This library is free software and may be redistributed and/or modified under
+ * the terms of the wxWindows Library License, Version 3.1 or (at your option)
+ * any later version. The full license is in the LICENSE.txt file included
+ * with this distribution.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * wxWindows Library License for more details.
+ */
+
#define JPEG_INTERNALS
#include "jinclude.h"
#include "jpeglib.h"
#include "jchuff.h" /* Declarations shared with jcphuff.c */
+static unsigned char jpeg_first_bit_table[65536];
+int jpeg_first_bit_table_init=0;
+
+#define CALC_FIRST_BIT(nbits, t) \
+ nbits = jpeg_first_bit_table[t&255]; \
+ if (t > 255) nbits = jpeg_first_bit_table[t>>8] + 8;
+
/* Expanded entropy encoder object for Huffman encoding.
*
* The savable_state subrecord contains fields that change within an MCU,
@@ -261,6 +283,15 @@
dtbl->ehufco[i] = huffcode[p];
dtbl->ehufsi[i] = huffsize[p];
}
+
+ if(!jpeg_first_bit_table_init) {
+ for(i = 0; i < 65536; i++) {
+ int bit = 0, val = i;
+ while (val) {val >>= 1; bit++;}
+ jpeg_first_bit_table[i] = bit;
+ }
+ jpeg_first_bit_table_init = 1;
+ }
}
@@ -297,148 +328,155 @@
* between calls, so 24 bits are sufficient.
*/
-inline
-LOCAL(boolean)
-emit_bits (working_state * state, unsigned int code, int size)
-/* Emit some bits; return TRUE if successful, FALSE if must suspend */
-{
- /* This routine is heavily used, so it's worth coding tightly. */
- register INT32 put_buffer = (INT32) code;
- register int put_bits = state->cur.put_bits;
+/***************************************************************/
- /* if size is 0, caller used an invalid Huffman table entry */
- if (size == 0)
- ERREXIT(state->cinfo, JERR_HUFF_MISSING_CODE);
+#define DUMP_BITS_(code, size) { \
+ put_bits += size; \
+ put_buffer = (put_buffer << size) | code; \
+ if (put_bits > 7) \
+ while(put_bits > 7) \
+ if (0xFF == (*buffer++ = put_buffer >> (put_bits -= 8))) \
+ *buffer++ = 0; \
+ }
- put_buffer &= (((INT32) 1)<<size) - 1; /* mask off any extra bits in code */
-
- put_bits += size; /* new number of bits in buffer */
-
- put_buffer <<= 24 - put_bits; /* align incoming bits */
+/***************************************************************/
- put_buffer |= state->cur.put_buffer; /* and merge with old buffer contents */
-
- while (put_bits >= 8) {
- int c = (int) ((put_buffer >> 16) & 0xFF);
-
- emit_byte(state, c, return FALSE);
- if (c == 0xFF) { /* need to stuff a zero byte? */
- emit_byte(state, 0, return FALSE);
- }
- put_buffer <<= 8;
- put_bits -= 8;
- }
+#define DUMP_BITS(code, size) { \
+ put_bits += size; \
+ put_buffer = (put_buffer << size) | code; \
+ if (put_bits > 15) { \
+ if (0xFF == (*buffer++ = put_buffer >> (put_bits -= 8))) \
+ *buffer++ = 0; \
+ if (0xFF == (*buffer++ = put_buffer >> (put_bits -= 8))) \
+ *buffer++ = 0; \
+ } \
+ }
- state->cur.put_buffer = put_buffer; /* update state variables */
- state->cur.put_bits = put_bits;
+/***************************************************************/
- return TRUE;
-}
+#define DUMP_SINGLE_VALUE(ht, codevalue) { \
+ size = ht->ehufsi[codevalue]; \
+ code = ht->ehufco[codevalue]; \
+ \
+ DUMP_BITS(code, size) \
+ }
+/***************************************************************/
+#define DUMP_VALUE(ht, codevalue, t, nbits) { \
+ size = ht->ehufsi[codevalue]; \
+ code = ht->ehufco[codevalue]; \
+ t &= ~(-1 << nbits); \
+ DUMP_BITS(code, size) \
+ DUMP_BITS(t, nbits) \
+ }
+
+/***************************************************************/
+
+
LOCAL(boolean)
flush_bits (working_state * state)
{
- if (! emit_bits(state, 0x7F, 7)) /* fill any partial byte with ones */
- return FALSE;
+ unsigned char *buffer;
+ int put_buffer, put_bits;
+
+ if ((state)->free_in_buffer < DCTSIZE2 * 2)
+ if (! dump_buffer(state)) return FALSE;
+
+ buffer = state->next_output_byte;
+ put_buffer = state->cur.put_buffer;
+ put_bits = state->cur.put_bits;
+
+ DUMP_BITS_(0x7F, 7)
+
state->cur.put_buffer = 0; /* and reset bit-buffer to empty */
state->cur.put_bits = 0;
+ state->free_in_buffer -= (buffer - state->next_output_byte);
+ state->next_output_byte = buffer;
+
+ if ((state)->free_in_buffer < DCTSIZE2 * 2)
+ if (! dump_buffer(state)) return FALSE;
+
return TRUE;
}
-
/* Encode a single block's worth of coefficients */
LOCAL(boolean)
encode_one_block (working_state * state, JCOEFPTR block, int last_dc_val,
c_derived_tbl *dctbl, c_derived_tbl *actbl)
{
- register int temp, temp2;
- register int nbits;
- register int k, r, i;
-
+ int temp, temp2;
+ int nbits;
+ int r, sflag, size, code;
+ unsigned char *buffer;
+ int put_buffer, put_bits;
+ int code_0xf0 = actbl->ehufco[0xf0], size_0xf0 = actbl->ehufsi[0xf0];
+
+ if ((state)->free_in_buffer < DCTSIZE2 * 2)
+ if (! dump_buffer(state)) return FALSE;
+
+ buffer = state->next_output_byte;
+ put_buffer = state->cur.put_buffer;
+ put_bits = state->cur.put_bits;
+
/* Encode the DC coefficient difference per section F.1.2.1 */
temp = temp2 = block[0] - last_dc_val;
- if (temp < 0) {
- temp = -temp; /* temp is abs value of input */
- /* For a negative input, want temp2 = bitwise complement of abs(input) */
- /* This code assumes we are on a two's complement machine */
- temp2--;
- }
-
- /* Find the number of bits needed for the magnitude of the coefficient */
- nbits = 0;
- while (temp) {
- nbits++;
- temp >>= 1;
- }
- /* Check for out-of-range coefficient values.
- * Since we're encoding a difference, the range limit is twice as much.
- */
- if (nbits > MAX_COEF_BITS+1)
- ERREXIT(state->cinfo, JERR_BAD_DCT_COEF);
-
- /* Emit the Huffman-coded symbol for the number of bits */
- if (! emit_bits(state, dctbl->ehufco[nbits], dctbl->ehufsi[nbits]))
- return FALSE;
+ sflag = temp >> 31;
+ temp -= ((temp + temp) & sflag);
+ temp2 += sflag;
+ CALC_FIRST_BIT(nbits, temp)
+ DUMP_VALUE(dctbl, nbits, temp2, nbits)
- /* Emit that number of bits of the value, if positive, */
- /* or the complement of its magnitude, if negative. */
- if (nbits) /* emit_bits rejects calls with size 0 */
- if (! emit_bits(state, (unsigned int) temp2, nbits))
- return FALSE;
-
/* Encode the AC coefficients per section F.1.2.2 */
r = 0; /* r = run length of zeros */
-
- for (k = 1; k < DCTSIZE2; k++) {
- if ((temp = block[jpeg_natural_order[k]]) == 0) {
- r++;
- } else {
- /* if run length > 15, must emit special run-length-16 codes (0xF0) */
- while (r > 15) {
- if (! emit_bits(state, actbl->ehufco[0xF0], actbl->ehufsi[0xF0]))
- return FALSE;
- r -= 16;
- }
- temp2 = temp;
- if (temp < 0) {
- temp = -temp; /* temp is abs value of input */
- /* This code assumes we are on a two's complement machine */
- temp2--;
- }
-
- /* Find the number of bits needed for the magnitude of the coefficient */
- nbits = 1; /* there must be at least one 1 bit */
- while ((temp >>= 1))
- nbits++;
- /* Check for out-of-range coefficient values */
- if (nbits > MAX_COEF_BITS)
- ERREXIT(state->cinfo, JERR_BAD_DCT_COEF);
-
- /* Emit Huffman symbol for run length / number of bits */
- i = (r << 4) + nbits;
- if (! emit_bits(state, actbl->ehufco[i], actbl->ehufsi[i]))
- return FALSE;
+#define innerloop(order) { \
+ temp2 = *(JCOEF*)((unsigned char*)block + order); \
+ if(temp2 == 0) r++; \
+ else { \
+ temp = (JCOEF)temp2; \
+ sflag = temp >> 31; \
+ temp = (temp ^ sflag) - sflag; \
+ temp2 += sflag; \
+ nbits = jpeg_first_bit_table[temp]; \
+ for(; r > 15; r -= 16) DUMP_BITS(code_0xf0, size_0xf0) \
+ sflag = (r << 4) + nbits; \
+ DUMP_VALUE(actbl, sflag, temp2, nbits) \
+ r = 0; \
+ }}
- /* Emit that number of bits of the value, if positive, */
- /* or the complement of its magnitude, if negative. */
- if (! emit_bits(state, (unsigned int) temp2, nbits))
- return FALSE;
-
- r = 0;
- }
- }
+ innerloop(2*1); innerloop(2*8); innerloop(2*16); innerloop(2*9);
+ innerloop(2*2); innerloop(2*3); innerloop(2*10); innerloop(2*17);
+ innerloop(2*24); innerloop(2*32); innerloop(2*25); innerloop(2*18);
+ innerloop(2*11); innerloop(2*4); innerloop(2*5); innerloop(2*12);
+ innerloop(2*19); innerloop(2*26); innerloop(2*33); innerloop(2*40);
+ innerloop(2*48); innerloop(2*41); innerloop(2*34); innerloop(2*27);
+ innerloop(2*20); innerloop(2*13); innerloop(2*6); innerloop(2*7);
+ innerloop(2*14); innerloop(2*21); innerloop(2*28); innerloop(2*35);
+ innerloop(2*42); innerloop(2*49); innerloop(2*56); innerloop(2*57);
+ innerloop(2*50); innerloop(2*43); innerloop(2*36); innerloop(2*29);
+ innerloop(2*22); innerloop(2*15); innerloop(2*23); innerloop(2*30);
+ innerloop(2*37); innerloop(2*44); innerloop(2*51); innerloop(2*58);
+ innerloop(2*59); innerloop(2*52); innerloop(2*45); innerloop(2*38);
+ innerloop(2*31); innerloop(2*39); innerloop(2*46); innerloop(2*53);
+ innerloop(2*60); innerloop(2*61); innerloop(2*54); innerloop(2*47);
+ innerloop(2*55); innerloop(2*62); innerloop(2*63);
/* If the last coef(s) were zero, emit an end-of-block code */
- if (r > 0)
- if (! emit_bits(state, actbl->ehufco[0], actbl->ehufsi[0]))
- return FALSE;
+ if (r > 0) DUMP_SINGLE_VALUE(actbl, 0x0)
+ state->cur.put_buffer = put_buffer;
+ state->cur.put_bits = put_bits;
+ state->free_in_buffer -= (buffer - state->next_output_byte);
+ state->next_output_byte = buffer;
+
+ if ((state)->free_in_buffer < DCTSIZE2 * 2)
+ if (! dump_buffer(state)) return FALSE;
+
return TRUE;
}
------------------------------------------------------------------------------ Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are powering Web 2.0 with engaging, cross-platform capabilities. Quickly and easily build your RIAs with Flex Builder, the Eclipse(TM)based development software that enables intelligent coding and step-through debugging. Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com
_______________________________________________ Tigervnc-devel mailing list Tigervnc-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tigervnc-devel
