Index: jdmarker.c
===================================================================
--- jdmarker.c	(revision 677)
+++ jdmarker.c	(working copy)
@@ -940,7 +940,144 @@
   return TRUE;
 }
 
+#ifdef MOTION_JPEG_SUPPORTED
 
+/* The default Huffman tables used by motion JPEG frames. When a motion JPEG
+ * frame does not have DHT tables, we should use the huffman tables suggested by
+ * the JPEG standard. Each of these tables represents a member of the JHUFF_TBLS
+ * struct so we can just copy it to the according JHUFF_TBLS member.
+ */
+/* DC table 0 */
+LOCAL(const unsigned char) mjpg_dc0_bits[] = {
+  0x00, 0x01, 0x05, 0x01, 0x01, 0x01, 0x01, 0x01,
+  0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00
+};
+
+LOCAL(const unsigned char) mjpg_dc0_huffval[] = {
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
+  0x08, 0x09, 0x0A, 0x0B
+};
+
+/* DC table 1 */
+LOCAL(const unsigned char) mjpg_dc1_bits[] = {
+  0x00, 0x03, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01,
+  0x01, 0x01, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00
+};
+
+LOCAL(const unsigned char) mjpg_dc1_huffval[] = {
+  0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
+  0x08, 0x09, 0x0A, 0x0B
+};
+  
+/* AC table 0 */
+LOCAL(const unsigned char) mjpg_ac0_bits[] = {
+  0x00, 0x02, 0x01, 0x03, 0x03, 0x02, 0x04, 0x03,
+  0x05, 0x05, 0x04, 0x04, 0x00, 0x00, 0x01, 0x7D
+};
+
+LOCAL(const unsigned char) mjpg_ac0_huffval[] = {
+  0x01, 0x02, 0x03, 0x00, 0x04, 0x11, 0x05, 0x12,
+  0x21, 0x31, 0x41, 0x06, 0x13, 0x51, 0x61, 0x07,
+  0x22, 0x71, 0x14, 0x32, 0x81, 0x91, 0xA1, 0x08,
+  0x23, 0x42, 0xB1, 0xC1, 0x15, 0x52, 0xD1, 0xF0,
+  0x24, 0x33, 0x62, 0x72, 0x82, 0x09, 0x0A, 0x16,
+  0x17, 0x18, 0x19, 0x1A, 0x25, 0x26, 0x27, 0x28,
+  0x29, 0x2A, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39,
+  0x3A, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49,
+  0x4A, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59,
+  0x5A, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
+  0x6A, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
+  0x7A, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
+  0x8A, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
+  0x99, 0x9A, 0xA2, 0xA3, 0xA4, 0xA5, 0xA6, 0xA7,
+  0xA8, 0xA9, 0xAA, 0xB2, 0xB3, 0xB4, 0xB5, 0xB6,
+  0xB7, 0xB8, 0xB9, 0xBA, 0xC2, 0xC3, 0xC4, 0xC5,
+  0xC6, 0xC7, 0xC8, 0xC9, 0xCA, 0xD2, 0xD3, 0xD4,
+  0xD5, 0xD6, 0xD7, 0xD8, 0xD9, 0xDA, 0xE1, 0xE2,
+  0xE3, 0xE4, 0xE5, 0xE6, 0xE7, 0xE8, 0xE9, 0xEA,
+  0xF1, 0xF2, 0xF3, 0xF4, 0xF5, 0xF6, 0xF7, 0xF8,
+  0xF9, 0xFA
+};
+
+/* AC table 1 */
+LOCAL(const unsigned char) mjpg_ac1_bits[] = {
+  0x00, 0x02, 0x01, 0x02, 0x04, 0x04, 0x03, 0x04,
+  0x07, 0x05, 0x04, 0x04, 0x00, 0x01, 0x02, 0x77
+};
+
+LOCAL(const unsigned char) mjpg_ac1_huffval[] = {
+  0x00, 0x01, 0x02, 0x03, 0x11, 0x04, 0x05, 0x21,
+  0x31, 0x06, 0x12, 0x41, 0x51, 0x07, 0x61, 0x71,
+  0x13, 0x22, 0x32, 0x81, 0x08, 0x14, 0x42, 0x91,
+  0xA1, 0xB1, 0xC1, 0x09, 0x23, 0x33, 0x52, 0xF0,
+  0x15, 0x62, 0x72, 0xD1, 0x0A, 0x16, 0x24, 0x34,
+  0xE1, 0x25, 0xF1, 0x17, 0x18, 0x19, 0x1A, 0x26,
+  0x27, 0x28, 0x29, 0x2A, 0x35, 0x36, 0x37, 0x38,
+  0x39, 0x3A, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48,
+  0x49, 0x4A, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58,
+  0x59, 0x5A, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
+  0x69, 0x6A, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78,
+  0x79, 0x7A, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
+  0x88, 0x89, 0x8A, 0x92, 0x93, 0x94, 0x95, 0x96,
+  0x97, 0x98, 0x99, 0x9A, 0xA2, 0xA3, 0xA4, 0xA5,
+  0xA6, 0xA7, 0xA8, 0xA9, 0xAA, 0xB2, 0xB3, 0xB4,
+  0xB5, 0xB6, 0xB7, 0xB8, 0xB9, 0xBA, 0xC2, 0xC3,
+  0xC4, 0xC5, 0xC6, 0xC7, 0xC8, 0xC9, 0xCA, 0xD2,
+  0xD3, 0xD4, 0xD5, 0xD6, 0xD7, 0xD8, 0xD9, 0xDA,
+  0xE2, 0xE3, 0xE4, 0xE5, 0xE6, 0xE7, 0xE8, 0xE9,
+  0xEA, 0xF2, 0xF3, 0xF4, 0xF5, 0xF6, 0xF7, 0xF8,
+  0xF9, 0xFA
+};
+
+/* Loads the default Huffman tables used by motion JPEG frames. This function
+ * just copies the huffman tables suggested in the JPEG standard when we have
+ * not load them.
+ */
+LOCAL(void)
+mjpg_load_huff_tables (j_decompress_ptr cinfo)
+{
+  JHUFF_TBL *htblptr;
+
+  if (! cinfo->dc_huff_tbl_ptrs[0]) {
+    htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+    MEMZERO(htblptr, SIZEOF(JHUFF_TBL));
+    MEMCOPY(&htblptr->bits[1], mjpg_dc0_bits, SIZEOF(mjpg_dc0_bits));
+    MEMCOPY(&htblptr->huffval[0], mjpg_dc0_huffval, SIZEOF(mjpg_dc0_huffval));
+    cinfo->dc_huff_tbl_ptrs[0] = htblptr;
+  }
+
+  if (! cinfo->dc_huff_tbl_ptrs[1]) {
+    htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+    MEMZERO(htblptr, SIZEOF(JHUFF_TBL));
+    MEMCOPY(&htblptr->bits[1], mjpg_dc1_bits, SIZEOF(mjpg_dc1_bits));
+    MEMCOPY(&htblptr->huffval[0], mjpg_dc1_huffval, SIZEOF(mjpg_dc1_huffval));
+    cinfo->dc_huff_tbl_ptrs[1] = htblptr;
+  }
+
+  if (! cinfo->ac_huff_tbl_ptrs[0]) {
+    htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+    MEMZERO(htblptr, SIZEOF(JHUFF_TBL));
+    MEMCOPY(&htblptr->bits[1], mjpg_ac0_bits, SIZEOF(mjpg_ac0_bits));
+    MEMCOPY(&htblptr->huffval[0], mjpg_ac0_huffval, SIZEOF(mjpg_ac0_huffval));
+    cinfo->ac_huff_tbl_ptrs[0] = htblptr;
+  }
+
+  if (! cinfo->ac_huff_tbl_ptrs[1]) {
+    htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+    MEMZERO(htblptr, SIZEOF(JHUFF_TBL));
+    MEMCOPY(&htblptr->bits[1], mjpg_ac1_bits, SIZEOF(mjpg_ac1_bits));
+    MEMCOPY(&htblptr->huffval[0], mjpg_ac1_huffval, SIZEOF(mjpg_ac1_huffval));
+    cinfo->ac_huff_tbl_ptrs[1] = htblptr;
+  }
+}
+
+#else
+
+#define mjpg_load_huff_tables(cinfo)
+
+#endif /* MOTION_JPEG_SUPPORTED */
+
+
 /*
  * Read markers until SOS or EOI.
  *
@@ -1009,6 +1146,7 @@
       break;
 
     case M_SOS:
+      mjpg_load_huff_tables(cinfo);
       if (! get_sos(cinfo))
 	return JPEG_SUSPENDED;
       cinfo->unread_marker = 0;	/* processed the marker */
Index: jmorecfg.h
===================================================================
--- jmorecfg.h	(revision 677)
+++ jmorecfg.h	(working copy)
@@ -153,14 +153,18 @@
 /* INT16 must hold at least the values -32768..32767. */
 
 #ifndef XMD_H			/* X11/xmd.h correctly defines INT16 */
+#ifndef _BASETSD_H_		/* basetsd.h correctly defines INT32 */
 typedef short INT16;
 #endif
+#endif
 
 /* INT32 must hold at least signed 32-bit values. */
 
 #ifndef XMD_H			/* X11/xmd.h correctly defines INT32 */
+#ifndef _BASETSD_H_		/* basetsd.h correctly defines INT32 */
 typedef long INT32;
 #endif
+#endif
 
 /* Datatype used for image dimensions.  The JPEG standard only supports
  * images up to 64K*64K due to 16-bit fields in SOF markers.  Therefore
@@ -210,11 +214,13 @@
  * explicit coding is needed; see uses of the NEED_FAR_POINTERS symbol.
  */
 
+#ifndef FAR
 #ifdef NEED_FAR_POINTERS
 #define FAR  far
 #else
 #define FAR
 #endif
+#endif
 
 
 /*
Index: jpeglib.h
===================================================================
--- jpeglib.h	(revision 677)
+++ jpeglib.h	(working copy)
@@ -15,6 +15,10 @@
 #ifndef JPEGLIB_H
 #define JPEGLIB_H
 
+/* Begin chromium edits */
+#include "jpeglibmangler.h"
+/* End chromium edits */
+
 /*
  * First we include the configuration files that record how this
  * installation of the JPEG library is set up.  jconfig.h can be
Index: jpeglibmangler.h
===================================================================
--- jpeglibmangler.h	(revision 0)
+++ jpeglibmangler.h	(revision 0)
@@ -0,0 +1,113 @@
+// Copyright (c) 2009 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef THIRD_PARTY_LIBJPEG_TURBO_JPEGLIBMANGLER_H_
+#define THIRD_PARTY_LIBJPEG_TURBO_JPEGLIBMANGLER_H_
+
+// Mangle all externally visible function names so we can build our own libjpeg
+// without system libraries trying to use it.
+
+#define jpeg_make_c_derived_tbl chromium_jpeg_make_c_derived_tbl
+#define jpeg_gen_optimal_table chromium_jpeg_gen_optimal_table
+#define jpeg_make_d_derived_tbl chromium_jpeg_make_d_derived_tbl
+#define jpeg_fill_bit_buffer chromium_jpeg_fill_bit_buffer
+#define jpeg_huff_decode chromium_jpeg_huff_decode
+#define jpeg_fdct_islow chromium_jpeg_fdct_islow
+#define jpeg_fdct_ifast chromium_jpeg_fdct_ifast
+#define jpeg_fdct_float chromium_jpeg_fdct_float
+#define jpeg_idct_islow chromium_jpeg_idct_islow
+#define jpeg_idct_ifast chromium_jpeg_idct_ifast
+#define jpeg_idct_float chromium_jpeg_idct_float
+#define jpeg_idct_4x4 chromium_jpeg_idct_4x4
+#define jpeg_idct_2x2 chromium_jpeg_idct_2x2
+#define jpeg_idct_1x1 chromium_jpeg_idct_1x1
+#define jinit_compress_master chromium_jinit_compress_master
+#define jinit_c_master_control chromium_jinit_c_master_control
+#define jinit_c_main_controller chromium_jinit_c_main_controller
+#define jinit_c_prep_controller chromium_jinit_c_prep_controller
+#define jinit_c_coef_controller chromium_jinit_c_coef_controller
+#define jinit_color_converter chromium_jinit_color_converter
+#define jinit_downsampler chromium_jinit_downsampler
+#define jinit_forward_dct chromium_jinit_forward_dct
+#define jinit_huff_encoder chromium_jinit_huff_encoder
+#define jinit_phuff_encoder chromium_jinit_phuff_encoder
+#define jinit_marker_writer chromium_jinit_marker_writer
+#define jinit_master_decompress chromium_jinit_master_decompress
+#define jinit_d_main_controller chromium_jinit_d_main_controller
+#define jinit_d_coef_controller chromium_jinit_d_coef_controller
+#define jinit_d_post_controller chromium_jinit_d_post_controller
+#define jinit_input_controller chromium_jinit_input_controller
+#define jinit_marker_reader chromium_jinit_marker_reader
+#define jinit_huff_decoder chromium_jinit_huff_decoder
+#define jinit_phuff_decoder chromium_jinit_phuff_decoder
+#define jinit_inverse_dct chromium_jinit_inverse_dct
+#define jinit_upsampler chromium_jinit_upsampler
+#define jinit_color_deconverter chromium_jinit_color_deconverter
+#define jinit_1pass_quantizer chromium_jinit_1pass_quantizer
+#define jinit_2pass_quantizer chromium_jinit_2pass_quantizer
+#define jinit_merged_upsampler chromium_jinit_merged_upsampler
+#define jinit_memory_mgr chromium_jinit_memory_mgr
+#define jdiv_round_up chromium_jdiv_round_up
+#define jround_up chromium_jround_up
+#define jcopy_sample_rows chromium_jcopy_sample_rows
+#define jcopy_block_row chromium_jcopy_block_row
+#define jzero_far chromium_jzero_far
+#define jpeg_std_error chromium_jpeg_std_error
+#define jpeg_CreateCompress chromium_jpeg_CreateCompress
+#define jpeg_CreateDecompress chromium_jpeg_CreateDecompress
+#define jpeg_destroy_compress chromium_jpeg_destroy_compress
+#define jpeg_destroy_decompress chromium_jpeg_destroy_decompress
+#define jpeg_stdio_dest chromium_jpeg_stdio_dest
+#define jpeg_stdio_src chromium_jpeg_stdio_src
+#define jpeg_set_defaults chromium_jpeg_set_defaults
+#define jpeg_set_colorspace chromium_jpeg_set_colorspace
+#define jpeg_default_colorspace chromium_jpeg_default_colorspace
+#define jpeg_set_quality chromium_jpeg_set_quality
+#define jpeg_set_linear_quality chromium_jpeg_set_linear_quality
+#define jpeg_add_quant_table chromium_jpeg_add_quant_table
+#define jpeg_quality_scaling chromium_jpeg_quality_scaling
+#define jpeg_simple_progression chromium_jpeg_simple_progression
+#define jpeg_suppress_tables chromium_jpeg_suppress_tables
+#define jpeg_alloc_quant_table chromium_jpeg_alloc_quant_table
+#define jpeg_alloc_huff_table chromium_jpeg_alloc_huff_table
+#define jpeg_start_compress chromium_jpeg_start_compress
+#define jpeg_write_scanlines chromium_jpeg_write_scanlines
+#define jpeg_finish_compress chromium_jpeg_finish_compress
+#define jpeg_write_raw_data chromium_jpeg_write_raw_data
+#define jpeg_write_marker chromium_jpeg_write_marker
+#define jpeg_write_m_header chromium_jpeg_write_m_header
+#define jpeg_write_m_byte chromium_jpeg_write_m_byte
+#define jpeg_write_tables chromium_jpeg_write_tables
+#define jpeg_read_header chromium_jpeg_read_header
+#define jpeg_start_decompress chromium_jpeg_start_decompress
+#define jpeg_read_scanlines chromium_jpeg_read_scanlines
+#define jpeg_finish_decompress chromium_jpeg_finish_decompress
+#define jpeg_read_raw_data chromium_jpeg_read_raw_data
+#define jpeg_has_multiple_scans chromium_jpeg_has_multiple_scans
+#define jpeg_start_output chromium_jpeg_start_output
+#define jpeg_finish_output chromium_jpeg_finish_output
+#define jpeg_input_complete chromium_jpeg_input_complete
+#define jpeg_new_colormap chromium_jpeg_new_colormap
+#define jpeg_consume_input chromium_jpeg_consume_input
+#define jpeg_calc_output_dimensions chromium_jpeg_calc_output_dimensions
+#define jpeg_save_markers chromium_jpeg_save_markers
+#define jpeg_set_marker_processor chromium_jpeg_set_marker_processor
+#define jpeg_read_coefficients chromium_jpeg_read_coefficients
+#define jpeg_write_coefficients chromium_jpeg_write_coefficients
+#define jpeg_copy_critical_parameters chromium_jpeg_copy_critical_parameters
+#define jpeg_abort_compress chromium_jpeg_abort_compress
+#define jpeg_abort_decompress chromium_jpeg_abort_decompress
+#define jpeg_abort chromium_jpeg_abort
+#define jpeg_destroy chromium_jpeg_destroy
+#define jpeg_resync_to_restart chromium_jpeg_resync_to_restart
+#define jpeg_get_small chromium_jpeg_get_small
+#define jpeg_free_small chromium_jpeg_free_small
+#define jpeg_get_large chromium_jpeg_get_large
+#define jpeg_free_large chromium_jpeg_free_large
+#define jpeg_mem_available chromium_jpeg_mem_available
+#define jpeg_open_backing_store chromium_jpeg_open_backing_store
+#define jpeg_mem_init chromium_jpeg_mem_init
+#define jpeg_mem_term chromium_jpeg_mem_term
+
+#endif  // THIRD_PARTY_LIBJPEG_TURBO_JPEGLIBMANGLER_H_
Index: simd/jcgrass2-64.asm
===================================================================
--- simd/jcgrass2-64.asm	(revision 677)
+++ simd/jcgrass2-64.asm	(working copy)
@@ -30,7 +30,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_rgb_gray_convert_sse2)
+	global	EXTN(jconst_rgb_gray_convert_sse2) PRIVATE
 
 EXTN(jconst_rgb_gray_convert_sse2):
 
Index: simd/jiss2fst.asm
===================================================================
--- simd/jiss2fst.asm	(revision 677)
+++ simd/jiss2fst.asm	(working copy)
@@ -59,7 +59,7 @@
 %define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
 
 	alignz	16
-	global	EXTN(jconst_idct_ifast_sse2)
+	global	EXTN(jconst_idct_ifast_sse2) PRIVATE
 
 EXTN(jconst_idct_ifast_sse2):
 
@@ -92,7 +92,7 @@
 %define WK_NUM		2
 
 	align	16
-	global	EXTN(jsimd_idct_ifast_sse2)
+	global	EXTN(jsimd_idct_ifast_sse2) PRIVATE
 
 EXTN(jsimd_idct_ifast_sse2):
 	push	ebp
Index: simd/jcclrss2-64.asm
===================================================================
--- simd/jcclrss2-64.asm	(revision 677)
+++ simd/jcclrss2-64.asm	(working copy)
@@ -37,7 +37,7 @@
 
 	align	16
 
-	global	EXTN(jsimd_rgb_ycc_convert_sse2)
+	global	EXTN(jsimd_rgb_ycc_convert_sse2) PRIVATE
 
 EXTN(jsimd_rgb_ycc_convert_sse2):
 	push	rbp
Index: simd/jiss2red-64.asm
===================================================================
--- simd/jiss2red-64.asm	(revision 677)
+++ simd/jiss2red-64.asm	(working copy)
@@ -73,7 +73,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_idct_red_sse2)
+	global	EXTN(jconst_idct_red_sse2) PRIVATE
 
 EXTN(jconst_idct_red_sse2):
 
@@ -114,7 +114,7 @@
 %define WK_NUM		2
 
 	align	16
-	global	EXTN(jsimd_idct_4x4_sse2)
+	global	EXTN(jsimd_idct_4x4_sse2) PRIVATE
 
 EXTN(jsimd_idct_4x4_sse2):
 	push	rbp
@@ -413,7 +413,7 @@
 ; r13 = JDIMENSION output_col
 
 	align	16
-	global	EXTN(jsimd_idct_2x2_sse2)
+	global	EXTN(jsimd_idct_2x2_sse2) PRIVATE
 
 EXTN(jsimd_idct_2x2_sse2):
 	push	rbp
Index: simd/ji3dnflt.asm
===================================================================
--- simd/ji3dnflt.asm	(revision 677)
+++ simd/ji3dnflt.asm	(working copy)
@@ -27,7 +27,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_idct_float_3dnow)
+	global	EXTN(jconst_idct_float_3dnow) PRIVATE
 
 EXTN(jconst_idct_float_3dnow):
 
@@ -63,7 +63,7 @@
 					; FAST_FLOAT workspace[DCTSIZE2]
 
 	align	16
-	global	EXTN(jsimd_idct_float_3dnow)
+	global	EXTN(jsimd_idct_float_3dnow) PRIVATE
 
 EXTN(jsimd_idct_float_3dnow):
 	push	ebp
Index: simd/jsimdcpu.asm
===================================================================
--- simd/jsimdcpu.asm	(revision 677)
+++ simd/jsimdcpu.asm	(working copy)
@@ -29,7 +29,7 @@
 ;
 
 	align	16
-	global	EXTN(jpeg_simd_cpu_support)
+	global	EXTN(jpeg_simd_cpu_support) PRIVATE
 
 EXTN(jpeg_simd_cpu_support):
 	push	ebx
Index: simd/jdmerss2-64.asm
===================================================================
--- simd/jdmerss2-64.asm	(revision 677)
+++ simd/jdmerss2-64.asm	(working copy)
@@ -35,7 +35,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_merged_upsample_sse2)
+	global	EXTN(jconst_merged_upsample_sse2) PRIVATE
 
 EXTN(jconst_merged_upsample_sse2):
 
Index: simd/jdsammmx.asm
===================================================================
--- simd/jdsammmx.asm	(revision 677)
+++ simd/jdsammmx.asm	(working copy)
@@ -22,7 +22,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_fancy_upsample_mmx)
+	global	EXTN(jconst_fancy_upsample_mmx) PRIVATE
 
 EXTN(jconst_fancy_upsample_mmx):
 
@@ -58,7 +58,7 @@
 %define output_data_ptr(b)	(b)+20		; JSAMPARRAY * output_data_ptr
 
 	align	16
-	global	EXTN(jsimd_h2v1_fancy_upsample_mmx)
+	global	EXTN(jsimd_h2v1_fancy_upsample_mmx) PRIVATE
 
 EXTN(jsimd_h2v1_fancy_upsample_mmx):
 	push	ebp
@@ -216,7 +216,7 @@
 %define gotptr		wk(0)-SIZEOF_POINTER	; void * gotptr
 
 	align	16
-	global	EXTN(jsimd_h2v2_fancy_upsample_mmx)
+	global	EXTN(jsimd_h2v2_fancy_upsample_mmx) PRIVATE
 
 EXTN(jsimd_h2v2_fancy_upsample_mmx):
 	push	ebp
@@ -542,7 +542,7 @@
 %define output_data_ptr(b)	(b)+20		; JSAMPARRAY * output_data_ptr
 
 	align	16
-	global	EXTN(jsimd_h2v1_upsample_mmx)
+	global	EXTN(jsimd_h2v1_upsample_mmx) PRIVATE
 
 EXTN(jsimd_h2v1_upsample_mmx):
 	push	ebp
@@ -643,7 +643,7 @@
 %define output_data_ptr(b)	(b)+20		; JSAMPARRAY * output_data_ptr
 
 	align	16
-	global	EXTN(jsimd_h2v2_upsample_mmx)
+	global	EXTN(jsimd_h2v2_upsample_mmx) PRIVATE
 
 EXTN(jsimd_h2v2_upsample_mmx):
 	push	ebp
Index: simd/jdmrgmmx.asm
===================================================================
--- simd/jdmrgmmx.asm	(revision 677)
+++ simd/jdmrgmmx.asm	(working copy)
@@ -40,7 +40,7 @@
 %define gotptr		wk(0)-SIZEOF_POINTER	; void * gotptr
 
 	align	16
-	global	EXTN(jsimd_h2v1_merged_upsample_mmx)
+	global	EXTN(jsimd_h2v1_merged_upsample_mmx) PRIVATE
 
 EXTN(jsimd_h2v1_merged_upsample_mmx):
 	push	ebp
@@ -409,7 +409,7 @@
 %define output_buf(b)		(b)+20		; JSAMPARRAY output_buf
 
 	align	16
-	global	EXTN(jsimd_h2v2_merged_upsample_mmx)
+	global	EXTN(jsimd_h2v2_merged_upsample_mmx) PRIVATE
 
 EXTN(jsimd_h2v2_merged_upsample_mmx):
 	push	ebp
Index: simd/jdsamss2.asm
===================================================================
--- simd/jdsamss2.asm	(revision 677)
+++ simd/jdsamss2.asm	(working copy)
@@ -22,7 +22,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_fancy_upsample_sse2)
+	global	EXTN(jconst_fancy_upsample_sse2) PRIVATE
 
 EXTN(jconst_fancy_upsample_sse2):
 
@@ -58,7 +58,7 @@
 %define output_data_ptr(b)	(b)+20		; JSAMPARRAY * output_data_ptr
 
 	align	16
-	global	EXTN(jsimd_h2v1_fancy_upsample_sse2)
+	global	EXTN(jsimd_h2v1_fancy_upsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v1_fancy_upsample_sse2):
 	push	ebp
@@ -214,7 +214,7 @@
 %define gotptr		wk(0)-SIZEOF_POINTER	; void * gotptr
 
 	align	16
-	global	EXTN(jsimd_h2v2_fancy_upsample_sse2)
+	global	EXTN(jsimd_h2v2_fancy_upsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v2_fancy_upsample_sse2):
 	push	ebp
@@ -538,7 +538,7 @@
 %define output_data_ptr(b)	(b)+20		; JSAMPARRAY * output_data_ptr
 
 	align	16
-	global	EXTN(jsimd_h2v1_upsample_sse2)
+	global	EXTN(jsimd_h2v1_upsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v1_upsample_sse2):
 	push	ebp
@@ -637,7 +637,7 @@
 %define output_data_ptr(b)	(b)+20		; JSAMPARRAY * output_data_ptr
 
 	align	16
-	global	EXTN(jsimd_h2v2_upsample_sse2)
+	global	EXTN(jsimd_h2v2_upsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v2_upsample_sse2):
 	push	ebp
Index: simd/jiss2flt-64.asm
===================================================================
--- simd/jiss2flt-64.asm	(revision 677)
+++ simd/jiss2flt-64.asm	(working copy)
@@ -38,7 +38,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_idct_float_sse2)
+	global	EXTN(jconst_idct_float_sse2) PRIVATE
 
 EXTN(jconst_idct_float_sse2):
 
@@ -74,7 +74,7 @@
 					; FAST_FLOAT workspace[DCTSIZE2]
 
 	align	16
-	global	EXTN(jsimd_idct_float_sse2)
+	global	EXTN(jsimd_idct_float_sse2) PRIVATE
 
 EXTN(jsimd_idct_float_sse2):
 	push	rbp
Index: simd/jfss2int-64.asm
===================================================================
--- simd/jfss2int-64.asm	(revision 677)
+++ simd/jfss2int-64.asm	(working copy)
@@ -67,7 +67,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_fdct_islow_sse2)
+	global	EXTN(jconst_fdct_islow_sse2) PRIVATE
 
 EXTN(jconst_fdct_islow_sse2):
 
@@ -101,7 +101,7 @@
 %define WK_NUM		6
 
 	align	16
-	global	EXTN(jsimd_fdct_islow_sse2)
+	global	EXTN(jsimd_fdct_islow_sse2) PRIVATE
 
 EXTN(jsimd_fdct_islow_sse2):
 	push	rbp
Index: simd/jcqnts2f.asm
===================================================================
--- simd/jcqnts2f.asm	(revision 677)
+++ simd/jcqnts2f.asm	(working copy)
@@ -35,7 +35,7 @@
 %define workspace	ebp+16		; FAST_FLOAT * workspace
 
 	align	16
-	global	EXTN(jsimd_convsamp_float_sse2)
+	global	EXTN(jsimd_convsamp_float_sse2) PRIVATE
 
 EXTN(jsimd_convsamp_float_sse2):
 	push	ebp
@@ -115,7 +115,7 @@
 %define workspace	ebp+16		; FAST_FLOAT * workspace
 
 	align	16
-	global	EXTN(jsimd_quantize_float_sse2)
+	global	EXTN(jsimd_quantize_float_sse2) PRIVATE
 
 EXTN(jsimd_quantize_float_sse2):
 	push	ebp
Index: simd/jdmrgss2.asm
===================================================================
--- simd/jdmrgss2.asm	(revision 677)
+++ simd/jdmrgss2.asm	(working copy)
@@ -40,7 +40,7 @@
 %define gotptr		wk(0)-SIZEOF_POINTER	; void * gotptr
 
 	align	16
-	global	EXTN(jsimd_h2v1_merged_upsample_sse2)
+	global	EXTN(jsimd_h2v1_merged_upsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v1_merged_upsample_sse2):
 	push	ebp
@@ -307,6 +307,41 @@
 	movdqa	xmmA,xmmD
 	sub	ecx, byte SIZEOF_XMMWORD
 .column_st15:
+%ifdef STRICT_MEMORY_ACCESS
+	; Store the lower 8 bytes of xmmA to the output when it has enough
+	; space.
+	cmp	ecx, byte SIZEOF_MMWORD
+	jb	short .column_st7
+	movq	MMWORD [edi], xmmA
+	add	edi, byte SIZEOF_MMWORD
+	sub	ecx, byte SIZEOF_MMWORD
+	psrldq	xmmA, SIZEOF_MMWORD
+.column_st7:
+	; Store the lower 4 bytes of xmmA to the output when it has enough
+	; space.
+	cmp	ecx, byte SIZEOF_DWORD
+	jb	short .column_st3
+	movd	DWORD [edi], xmmA
+	add	edi, byte SIZEOF_DWORD
+	sub	ecx, byte SIZEOF_DWORD
+	psrldq	xmmA, SIZEOF_DWORD
+.column_st3:
+	; Store the lower 2 bytes of eax to the output when it has enough
+	; space.
+	movd	eax, xmmA
+	cmp	ecx, byte SIZEOF_WORD
+	jb	short .column_st1
+	mov	WORD [edi], ax
+	add	edi, byte SIZEOF_WORD
+	sub	ecx, byte SIZEOF_WORD
+	shr	eax, 16
+.column_st1:
+	; Store the lower 1 byte of eax to the output when it has enough
+	; space.
+	test	ecx, ecx
+	jz	short .endcolumn
+	mov	BYTE [edi], al
+%else
 	mov	eax,ecx
 	xor	ecx, byte 0x0F
 	shl	ecx, 2
@@ -346,6 +381,7 @@
 	por	xmmE,xmmC
 .adj0:	; ----------------
 	maskmovdqu xmmA,xmmE			; movntdqu XMMWORD [edi], xmmA
+%endif ; STRICT_MEMORY_ACCESS ; ---------------
 
 %else ; RGB_PIXELSIZE == 4 ; -----------
 
@@ -434,6 +470,22 @@
 	movdqa	xmmA,xmmD
 	sub	ecx, byte SIZEOF_XMMWORD/4
 .column_st15:
+%ifdef STRICT_MEMORY_ACCESS
+	; Store two pixels (8 bytes) of xmmA to the output when it has enough
+	; space.
+	cmp	ecx, byte SIZEOF_XMMWORD/8
+	jb	short .column_st7
+	movq	MMWORD [edi], xmmA
+	add	edi, byte SIZEOF_XMMWORD/8*4
+	sub	ecx, byte SIZEOF_XMMWORD/8
+	psrldq	xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+	; Store one pixel (4 bytes) of xmmA to the output when it has enough
+	; space.
+	test	ecx, ecx
+	jz	short .endcolumn
+	movd	DWORD [edi], xmmA
+%else
 	cmp	ecx, byte SIZEOF_XMMWORD/16
 	jb	short .endcolumn
 	mov	eax,ecx
@@ -473,6 +525,7 @@
 	por	xmmE,xmmG
 .adj0:	; ----------------
 	maskmovdqu xmmA,xmmE			; movntdqu XMMWORD [edi], xmmA
+%endif ; STRICT_MEMORY_ACCESS ; ---------------
 
 %endif ; RGB_PIXELSIZE ; ---------------
 
@@ -507,7 +560,7 @@
 %define output_buf(b)		(b)+20		; JSAMPARRAY output_buf
 
 	align	16
-	global	EXTN(jsimd_h2v2_merged_upsample_sse2)
+	global	EXTN(jsimd_h2v2_merged_upsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v2_merged_upsample_sse2):
 	push	ebp
Index: simd/jfmmxint.asm
===================================================================
--- simd/jfmmxint.asm	(revision 677)
+++ simd/jfmmxint.asm	(working copy)
@@ -66,7 +66,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_fdct_islow_mmx)
+	global	EXTN(jconst_fdct_islow_mmx) PRIVATE
 
 EXTN(jconst_fdct_islow_mmx):
 
@@ -101,7 +101,7 @@
 %define WK_NUM		2
 
 	align	16
-	global	EXTN(jsimd_fdct_islow_mmx)
+	global	EXTN(jsimd_fdct_islow_mmx) PRIVATE
 
 EXTN(jsimd_fdct_islow_mmx):
 	push	ebp
Index: simd/jcgryss2-64.asm
===================================================================
--- simd/jcgryss2-64.asm	(revision 677)
+++ simd/jcgryss2-64.asm	(working copy)
@@ -37,7 +37,7 @@
 
 	align	16
 
-	global	EXTN(jsimd_rgb_gray_convert_sse2)
+	global	EXTN(jsimd_rgb_gray_convert_sse2) PRIVATE
 
 EXTN(jsimd_rgb_gray_convert_sse2):
 	push	rbp
Index: simd/jcqnts2i.asm
===================================================================
--- simd/jcqnts2i.asm	(revision 677)
+++ simd/jcqnts2i.asm	(working copy)
@@ -35,7 +35,7 @@
 %define workspace	ebp+16		; DCTELEM * workspace
 
 	align	16
-	global	EXTN(jsimd_convsamp_sse2)
+	global	EXTN(jsimd_convsamp_sse2) PRIVATE
 
 EXTN(jsimd_convsamp_sse2):
 	push	ebp
@@ -117,7 +117,7 @@
 %define workspace	ebp+16		; DCTELEM * workspace
 
 	align	16
-	global	EXTN(jsimd_quantize_sse2)
+	global	EXTN(jsimd_quantize_sse2) PRIVATE
 
 EXTN(jsimd_quantize_sse2):
 	push	ebp
Index: simd/jiss2fst-64.asm
===================================================================
--- simd/jiss2fst-64.asm	(revision 677)
+++ simd/jiss2fst-64.asm	(working copy)
@@ -60,7 +60,7 @@
 %define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
 
 	alignz	16
-	global	EXTN(jconst_idct_ifast_sse2)
+	global	EXTN(jconst_idct_ifast_sse2) PRIVATE
 
 EXTN(jconst_idct_ifast_sse2):
 
@@ -93,7 +93,7 @@
 %define WK_NUM		2
 
 	align	16
-	global	EXTN(jsimd_idct_ifast_sse2)
+	global	EXTN(jsimd_idct_ifast_sse2) PRIVATE
 
 EXTN(jsimd_idct_ifast_sse2):
 	push	rbp
Index: simd/jiss2flt.asm
===================================================================
--- simd/jiss2flt.asm	(revision 677)
+++ simd/jiss2flt.asm	(working copy)
@@ -37,7 +37,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_idct_float_sse2)
+	global	EXTN(jconst_idct_float_sse2) PRIVATE
 
 EXTN(jconst_idct_float_sse2):
 
@@ -73,7 +73,7 @@
 					; FAST_FLOAT workspace[DCTSIZE2]
 
 	align	16
-	global	EXTN(jsimd_idct_float_sse2)
+	global	EXTN(jsimd_idct_float_sse2) PRIVATE
 
 EXTN(jsimd_idct_float_sse2):
 	push	ebp
Index: simd/jiss2int.asm
===================================================================
--- simd/jiss2int.asm	(revision 677)
+++ simd/jiss2int.asm	(working copy)
@@ -66,7 +66,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_idct_islow_sse2)
+	global	EXTN(jconst_idct_islow_sse2) PRIVATE
 
 EXTN(jconst_idct_islow_sse2):
 
@@ -105,7 +105,7 @@
 %define WK_NUM		12
 
 	align	16
-	global	EXTN(jsimd_idct_islow_sse2)
+	global	EXTN(jsimd_idct_islow_sse2) PRIVATE
 
 EXTN(jsimd_idct_islow_sse2):
 	push	ebp
Index: simd/jfsseflt-64.asm
===================================================================
--- simd/jfsseflt-64.asm	(revision 677)
+++ simd/jfsseflt-64.asm	(working copy)
@@ -38,7 +38,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_fdct_float_sse)
+	global	EXTN(jconst_fdct_float_sse) PRIVATE
 
 EXTN(jconst_fdct_float_sse):
 
@@ -65,7 +65,7 @@
 %define WK_NUM		2
 
 	align	16
-	global	EXTN(jsimd_fdct_float_sse)
+	global	EXTN(jsimd_fdct_float_sse) PRIVATE
 
 EXTN(jsimd_fdct_float_sse):
 	push	rbp
Index: simd/jccolss2-64.asm
===================================================================
--- simd/jccolss2-64.asm	(revision 677)
+++ simd/jccolss2-64.asm	(working copy)
@@ -34,7 +34,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_rgb_ycc_convert_sse2)
+	global	EXTN(jconst_rgb_ycc_convert_sse2) PRIVATE
 
 EXTN(jconst_rgb_ycc_convert_sse2):
 
Index: simd/jcsamss2-64.asm
===================================================================
--- simd/jcsamss2-64.asm	(revision 677)
+++ simd/jcsamss2-64.asm	(working copy)
@@ -41,7 +41,7 @@
 ; r15 = JSAMPARRAY output_data
 
 	align	16
-	global	EXTN(jsimd_h2v1_downsample_sse2)
+	global	EXTN(jsimd_h2v1_downsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v1_downsample_sse2):
 	push	rbp
@@ -185,7 +185,7 @@
 ; r15 = JSAMPARRAY output_data
 
 	align	16
-	global	EXTN(jsimd_h2v2_downsample_sse2)
+	global	EXTN(jsimd_h2v2_downsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v2_downsample_sse2):
 	push	rbp
Index: simd/jdclrss2-64.asm
===================================================================
--- simd/jdclrss2-64.asm	(revision 677)
+++ simd/jdclrss2-64.asm	(working copy)
@@ -39,7 +39,7 @@
 %define WK_NUM		2
 
 	align	16
-	global	EXTN(jsimd_ycc_rgb_convert_sse2)
+	global	EXTN(jsimd_ycc_rgb_convert_sse2) PRIVATE
 
 EXTN(jsimd_ycc_rgb_convert_sse2):
 	push	rbp
@@ -290,6 +290,41 @@
 	movdqa	xmmA,xmmD
 	sub	rcx, byte SIZEOF_XMMWORD
 .column_st15:
+%ifdef STRICT_MEMORY_ACCESS
+	; Store the lower 8 bytes of xmmA to the output when it has enough
+	; space.
+	cmp	rcx, byte SIZEOF_MMWORD
+	jb	short .column_st7
+	movq	MMWORD [rdi], xmmA
+	add	rdi, byte SIZEOF_MMWORD
+	sub	rcx, byte SIZEOF_MMWORD
+	psrldq	xmmA, SIZEOF_MMWORD
+.column_st7:
+	; Store the lower 4 bytes of xmmA to the output when it has enough
+	; space.
+	cmp	rcx, byte SIZEOF_DWORD
+	jb	short .column_st3
+	movd	DWORD [rdi], xmmA
+	add	rdi, byte SIZEOF_DWORD
+	sub	rcx, byte SIZEOF_DWORD
+	psrldq	xmmA, SIZEOF_DWORD
+.column_st3:
+	; Store the lower 2 bytes of rax to the output when it has enough
+	; space.
+	movd	rax, xmmA
+	cmp	rcx, byte SIZEOF_WORD
+	jb	short .column_st1
+	mov	WORD [rdi], ax
+	add	rdi, byte SIZEOF_WORD
+	sub	rcx, byte SIZEOF_WORD
+	shr	rax, 16
+.column_st1:
+	; Store the lower 1 byte of rax to the output when it has enough
+	; space.
+	test	rcx, rcx
+	jz	short .nextrow
+	mov	BYTE [rdi], al
+%else
 	mov	rax,rcx
 	xor	rcx, byte 0x0F
 	shl	rcx, 2
@@ -329,6 +364,7 @@
 	por	xmmE,xmmC
 .adj0:	; ----------------
 	maskmovdqu xmmA,xmmE			; movntdqu XMMWORD [rdi], xmmA
+%endif ; STRICT_MEMORY_ACCESS ; ---------------
 
 %else ; RGB_PIXELSIZE == 4 ; -----------
 
@@ -413,6 +449,22 @@
 	movdqa	xmmA,xmmD
 	sub	rcx, byte SIZEOF_XMMWORD/4
 .column_st15:
+%ifdef STRICT_MEMORY_ACCESS
+	; Store two pixels (8 bytes) of xmmA to the output when it has enough
+	; space.
+	cmp	rcx, byte SIZEOF_XMMWORD/8
+	jb	short .column_st7
+	movq	MMWORD [rdi], xmmA
+	add	rdi, byte SIZEOF_XMMWORD/8*4
+	sub	rcx, byte SIZEOF_XMMWORD/8
+	psrldq	xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+	; Store one pixel (4 bytes) of xmmA to the output when it has enough
+	; space.
+	test	rcx, rcx
+	jz	short .nextrow
+	movd	DWORD [rdi], xmmA
+%else
 	cmp	rcx, byte SIZEOF_XMMWORD/16
 	jb	near .nextrow
 	mov	rax,rcx
@@ -452,6 +504,7 @@
 	por	xmmE,xmmG
 .adj0:	; ----------------
 	maskmovdqu xmmA,xmmE			; movntdqu XMMWORD [rdi], xmmA
+%endif ; STRICT_MEMORY_ACCESS ; ---------------
 
 %endif ; RGB_PIXELSIZE ; ---------------
 
Index: simd/jdcolmmx.asm
===================================================================
--- simd/jdcolmmx.asm	(revision 677)
+++ simd/jdcolmmx.asm	(working copy)
@@ -35,7 +35,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_ycc_rgb_convert_mmx)
+	global	EXTN(jconst_ycc_rgb_convert_mmx) PRIVATE
 
 EXTN(jconst_ycc_rgb_convert_mmx):
 
Index: simd/jcclrmmx.asm
===================================================================
--- simd/jcclrmmx.asm	(revision 677)
+++ simd/jcclrmmx.asm	(working copy)
@@ -40,7 +40,7 @@
 %define gotptr		wk(0)-SIZEOF_POINTER	; void * gotptr
 
 	align	16
-	global	EXTN(jsimd_rgb_ycc_convert_mmx)
+	global	EXTN(jsimd_rgb_ycc_convert_mmx) PRIVATE
 
 EXTN(jsimd_rgb_ycc_convert_mmx):
 	push	ebp
Index: simd/jfsseflt.asm
===================================================================
--- simd/jfsseflt.asm	(revision 677)
+++ simd/jfsseflt.asm	(working copy)
@@ -37,7 +37,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_fdct_float_sse)
+	global	EXTN(jconst_fdct_float_sse) PRIVATE
 
 EXTN(jconst_fdct_float_sse):
 
@@ -65,7 +65,7 @@
 %define WK_NUM		2
 
 	align	16
-	global	EXTN(jsimd_fdct_float_sse)
+	global	EXTN(jsimd_fdct_float_sse) PRIVATE
 
 EXTN(jsimd_fdct_float_sse):
 	push	ebp
Index: simd/jdmrgss2-64.asm
===================================================================
--- simd/jdmrgss2-64.asm	(revision 677)
+++ simd/jdmrgss2-64.asm	(working copy)
@@ -39,7 +39,7 @@
 %define WK_NUM		3
 
 	align	16
-	global	EXTN(jsimd_h2v1_merged_upsample_sse2)
+	global	EXTN(jsimd_h2v1_merged_upsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v1_merged_upsample_sse2):
 	push	rbp
@@ -294,6 +294,41 @@
 	movdqa	xmmA,xmmD
 	sub	rcx, byte SIZEOF_XMMWORD
 .column_st15:
+%ifdef STRICT_MEMORY_ACCESS
+	; Store the lower 8 bytes of xmmA to the output when it has enough
+	; space.
+	cmp	rcx, byte SIZEOF_MMWORD
+	jb	short .column_st7
+	movq	MMWORD [rdi], xmmA
+	add	rdi, byte SIZEOF_MMWORD
+	sub	rcx, byte SIZEOF_MMWORD
+	psrldq	xmmA, SIZEOF_MMWORD
+.column_st7:
+	; Store the lower 4 bytes of xmmA to the output when it has enough
+	; space.
+	cmp	rcx, byte SIZEOF_DWORD
+	jb	short .column_st3
+	movd	DWORD [rdi], xmmA
+	add	rdi, byte SIZEOF_DWORD
+	sub	rcx, byte SIZEOF_DWORD
+	psrldq	xmmA, SIZEOF_DWORD
+.column_st3:
+	; Store the lower 2 bytes of rax to the output when it has enough
+	; space.
+	movd	rax, xmmA
+	cmp	rcx, byte SIZEOF_WORD
+	jb	short .column_st1
+	mov	WORD [rdi], ax
+	add	rdi, byte SIZEOF_WORD
+	sub	rcx, byte SIZEOF_WORD
+	shr	rax, 16
+.column_st1:
+	; Store the lower 1 byte of rax to the output when it has enough
+	; space.
+	test	rcx, rcx
+	jz	short .endcolumn
+	mov	BYTE [rdi], al
+%else
 	mov	rax,rcx
 	xor	rcx, byte 0x0F
 	shl	rcx, 2
@@ -333,6 +368,7 @@
 	por	xmmE,xmmC
 .adj0:	; ----------------
 	maskmovdqu xmmA,xmmE			; movntdqu XMMWORD [edi], xmmA
+%endif ; STRICT_MEMORY_ACCESS ; ---------------
 
 %else ; RGB_PIXELSIZE == 4 ; -----------
 
@@ -420,6 +456,22 @@
 	movdqa	xmmA,xmmD
 	sub	rcx, byte SIZEOF_XMMWORD/4
 .column_st15:
+%ifdef STRICT_MEMORY_ACCESS
+	; Store two pixels (8 bytes) of xmmA to the output when it has enough
+	; space.
+	cmp	rcx, byte SIZEOF_XMMWORD/8
+	jb	short .column_st7
+	movq	MMWORD [rdi], xmmA
+	add	rdi, byte SIZEOF_XMMWORD/8*4
+	sub	rcx, byte SIZEOF_XMMWORD/8
+	psrldq	xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+	; Store one pixel (4 bytes) of xmmA to the output when it has enough
+	; space.
+	test	rcx, rcx
+	jz	short .endcolumn
+	movd	DWORD [rdi], xmmA
+%else
 	cmp	rcx, byte SIZEOF_XMMWORD/16
 	jb	near .endcolumn
 	mov	rax,rcx
@@ -459,6 +511,7 @@
 	por	xmmE,xmmG
 .adj0:	; ----------------
 	maskmovdqu xmmA,xmmE			; movntdqu XMMWORD [edi], xmmA
+%endif ; STRICT_MEMORY_ACCESS ; ---------------
 
 %endif ; RGB_PIXELSIZE ; ---------------
 
@@ -490,7 +543,7 @@
 ; r13 = JSAMPARRAY output_buf
 
 	align	16
-	global	EXTN(jsimd_h2v2_merged_upsample_sse2)
+	global	EXTN(jsimd_h2v2_merged_upsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v2_merged_upsample_sse2):
 	push	rbp
Index: simd/jdcolss2.asm
===================================================================
--- simd/jdcolss2.asm	(revision 677)
+++ simd/jdcolss2.asm	(working copy)
@@ -35,7 +35,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_ycc_rgb_convert_sse2)
+	global	EXTN(jconst_ycc_rgb_convert_sse2) PRIVATE
 
 EXTN(jconst_ycc_rgb_convert_sse2):
 
Index: simd/jdmermmx.asm
===================================================================
--- simd/jdmermmx.asm	(revision 677)
+++ simd/jdmermmx.asm	(working copy)
@@ -35,7 +35,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_merged_upsample_mmx)
+	global	EXTN(jconst_merged_upsample_mmx) PRIVATE
 
 EXTN(jconst_merged_upsample_mmx):
 
Index: simd/jcclrss2.asm
===================================================================
--- simd/jcclrss2.asm	(revision 677)
+++ simd/jcclrss2.asm	(working copy)
@@ -38,7 +38,7 @@
 
 	align	16
 
-	global	EXTN(jsimd_rgb_ycc_convert_sse2)
+	global	EXTN(jsimd_rgb_ycc_convert_sse2) PRIVATE
 
 EXTN(jsimd_rgb_ycc_convert_sse2):
 	push	ebp
Index: simd/jiss2red.asm
===================================================================
--- simd/jiss2red.asm	(revision 677)
+++ simd/jiss2red.asm	(working copy)
@@ -72,7 +72,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_idct_red_sse2)
+	global	EXTN(jconst_idct_red_sse2) PRIVATE
 
 EXTN(jconst_idct_red_sse2):
 
@@ -113,7 +113,7 @@
 %define WK_NUM		2
 
 	align	16
-	global	EXTN(jsimd_idct_4x4_sse2)
+	global	EXTN(jsimd_idct_4x4_sse2) PRIVATE
 
 EXTN(jsimd_idct_4x4_sse2):
 	push	ebp
@@ -424,7 +424,7 @@
 %define output_col(b)	(b)+20		; JDIMENSION output_col
 
 	align	16
-	global	EXTN(jsimd_idct_2x2_sse2)
+	global	EXTN(jsimd_idct_2x2_sse2) PRIVATE
 
 EXTN(jsimd_idct_2x2_sse2):
 	push	ebp
Index: simd/jdmerss2.asm
===================================================================
--- simd/jdmerss2.asm	(revision 677)
+++ simd/jdmerss2.asm	(working copy)
@@ -35,7 +35,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_merged_upsample_sse2)
+	global	EXTN(jconst_merged_upsample_sse2) PRIVATE
 
 EXTN(jconst_merged_upsample_sse2):
 
Index: simd/jfss2fst-64.asm
===================================================================
--- simd/jfss2fst-64.asm	(revision 677)
+++ simd/jfss2fst-64.asm	(working copy)
@@ -53,7 +53,7 @@
 %define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
 
 	alignz	16
-	global	EXTN(jconst_fdct_ifast_sse2)
+	global	EXTN(jconst_fdct_ifast_sse2) PRIVATE
 
 EXTN(jconst_fdct_ifast_sse2):
 
@@ -80,7 +80,7 @@
 %define WK_NUM		2
 
 	align	16
-	global	EXTN(jsimd_fdct_ifast_sse2)
+	global	EXTN(jsimd_fdct_ifast_sse2) PRIVATE
 
 EXTN(jsimd_fdct_ifast_sse2):
 	push	rbp
Index: simd/jcqntmmx.asm
===================================================================
--- simd/jcqntmmx.asm	(revision 677)
+++ simd/jcqntmmx.asm	(working copy)
@@ -35,7 +35,7 @@
 %define workspace	ebp+16		; DCTELEM * workspace
 
 	align	16
-	global	EXTN(jsimd_convsamp_mmx)
+	global	EXTN(jsimd_convsamp_mmx) PRIVATE
 
 EXTN(jsimd_convsamp_mmx):
 	push	ebp
@@ -140,7 +140,7 @@
 %define workspace	ebp+16		; DCTELEM * workspace
 
 	align	16
-	global	EXTN(jsimd_quantize_mmx)
+	global	EXTN(jsimd_quantize_mmx) PRIVATE
 
 EXTN(jsimd_quantize_mmx):
 	push	ebp
Index: simd/jimmxfst.asm
===================================================================
--- simd/jimmxfst.asm	(revision 677)
+++ simd/jimmxfst.asm	(working copy)
@@ -59,7 +59,7 @@
 %define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
 
 	alignz	16
-	global	EXTN(jconst_idct_ifast_mmx)
+	global	EXTN(jconst_idct_ifast_mmx) PRIVATE
 
 EXTN(jconst_idct_ifast_mmx):
 
@@ -94,7 +94,7 @@
 					; JCOEF workspace[DCTSIZE2]
 
 	align	16
-	global	EXTN(jsimd_idct_ifast_mmx)
+	global	EXTN(jsimd_idct_ifast_mmx) PRIVATE
 
 EXTN(jsimd_idct_ifast_mmx):
 	push	ebp
Index: simd/jfss2fst.asm
===================================================================
--- simd/jfss2fst.asm	(revision 677)
+++ simd/jfss2fst.asm	(working copy)
@@ -52,7 +52,7 @@
 %define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
 
 	alignz	16
-	global	EXTN(jconst_fdct_ifast_sse2)
+	global	EXTN(jconst_fdct_ifast_sse2) PRIVATE
 
 EXTN(jconst_fdct_ifast_sse2):
 
@@ -80,7 +80,7 @@
 %define WK_NUM		2
 
 	align	16
-	global	EXTN(jsimd_fdct_ifast_sse2)
+	global	EXTN(jsimd_fdct_ifast_sse2) PRIVATE
 
 EXTN(jsimd_fdct_ifast_sse2):
 	push	ebp
Index: simd/jcgrammx.asm
===================================================================
--- simd/jcgrammx.asm	(revision 677)
+++ simd/jcgrammx.asm	(working copy)
@@ -33,7 +33,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_rgb_gray_convert_mmx)
+	global	EXTN(jconst_rgb_gray_convert_mmx) PRIVATE
 
 EXTN(jconst_rgb_gray_convert_mmx):
 
Index: simd/jdcolss2-64.asm
===================================================================
--- simd/jdcolss2-64.asm	(revision 677)
+++ simd/jdcolss2-64.asm	(working copy)
@@ -35,7 +35,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_ycc_rgb_convert_sse2)
+	global	EXTN(jconst_ycc_rgb_convert_sse2) PRIVATE
 
 EXTN(jconst_ycc_rgb_convert_sse2):
 
Index: simd/jf3dnflt.asm
===================================================================
--- simd/jf3dnflt.asm	(revision 677)
+++ simd/jf3dnflt.asm	(working copy)
@@ -27,7 +27,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_fdct_float_3dnow)
+	global	EXTN(jconst_fdct_float_3dnow) PRIVATE
 
 EXTN(jconst_fdct_float_3dnow):
 
@@ -55,7 +55,7 @@
 %define WK_NUM		2
 
 	align	16
-	global	EXTN(jsimd_fdct_float_3dnow)
+	global	EXTN(jsimd_fdct_float_3dnow) PRIVATE
 
 EXTN(jsimd_fdct_float_3dnow):
 	push	ebp
Index: simd/jdsamss2-64.asm
===================================================================
--- simd/jdsamss2-64.asm	(revision 677)
+++ simd/jdsamss2-64.asm	(working copy)
@@ -23,7 +23,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_fancy_upsample_sse2)
+	global	EXTN(jconst_fancy_upsample_sse2) PRIVATE
 
 EXTN(jconst_fancy_upsample_sse2):
 
@@ -59,7 +59,7 @@
 ; r13 = JSAMPARRAY * output_data_ptr
 
 	align	16
-	global	EXTN(jsimd_h2v1_fancy_upsample_sse2)
+	global	EXTN(jsimd_h2v1_fancy_upsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v1_fancy_upsample_sse2):
 	push	rbp
@@ -201,7 +201,7 @@
 %define WK_NUM		4
 
 	align	16
-	global	EXTN(jsimd_h2v2_fancy_upsample_sse2)
+	global	EXTN(jsimd_h2v2_fancy_upsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v2_fancy_upsample_sse2):
 	push	rbp
@@ -498,7 +498,7 @@
 ; r13 = JSAMPARRAY * output_data_ptr
 
 	align	16
-	global	EXTN(jsimd_h2v1_upsample_sse2)
+	global	EXTN(jsimd_h2v1_upsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v1_upsample_sse2):
 	push	rbp
@@ -587,7 +587,7 @@
 ; r13 = JSAMPARRAY * output_data_ptr
 
 	align	16
-	global	EXTN(jsimd_h2v2_upsample_sse2)
+	global	EXTN(jsimd_h2v2_upsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v2_upsample_sse2):
 	push	rbp
Index: simd/jcgrass2.asm
===================================================================
--- simd/jcgrass2.asm	(revision 677)
+++ simd/jcgrass2.asm	(working copy)
@@ -30,7 +30,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_rgb_gray_convert_sse2)
+	global	EXTN(jconst_rgb_gray_convert_sse2) PRIVATE
 
 EXTN(jconst_rgb_gray_convert_sse2):
 
Index: simd/jcsammmx.asm
===================================================================
--- simd/jcsammmx.asm	(revision 677)
+++ simd/jcsammmx.asm	(working copy)
@@ -40,7 +40,7 @@
 %define output_data(b)	(b)+28	; JSAMPARRAY output_data
 
 	align	16
-	global	EXTN(jsimd_h2v1_downsample_mmx)
+	global	EXTN(jsimd_h2v1_downsample_mmx) PRIVATE
 
 EXTN(jsimd_h2v1_downsample_mmx):
 	push	ebp
@@ -182,7 +182,7 @@
 %define output_data(b)	(b)+28	; JSAMPARRAY output_data
 
 	align	16
-	global	EXTN(jsimd_h2v2_downsample_mmx)
+	global	EXTN(jsimd_h2v2_downsample_mmx) PRIVATE
 
 EXTN(jsimd_h2v2_downsample_mmx):
 	push	ebp
Index: simd/jcqnts2f-64.asm
===================================================================
--- simd/jcqnts2f-64.asm	(revision 677)
+++ simd/jcqnts2f-64.asm	(working copy)
@@ -36,7 +36,7 @@
 ; r12 = FAST_FLOAT * workspace
 
 	align	16
-	global	EXTN(jsimd_convsamp_float_sse2)
+	global	EXTN(jsimd_convsamp_float_sse2) PRIVATE
 
 EXTN(jsimd_convsamp_float_sse2):
 	push	rbp
@@ -110,7 +110,7 @@
 ; r12 = FAST_FLOAT * workspace
 
 	align	16
-	global	EXTN(jsimd_quantize_float_sse2)
+	global	EXTN(jsimd_quantize_float_sse2) PRIVATE
 
 EXTN(jsimd_quantize_float_sse2):
 	push	rbp
Index: simd/jcqnt3dn.asm
===================================================================
--- simd/jcqnt3dn.asm	(revision 677)
+++ simd/jcqnt3dn.asm	(working copy)
@@ -35,7 +35,7 @@
 %define workspace	ebp+16		; FAST_FLOAT * workspace
 
 	align	16
-	global	EXTN(jsimd_convsamp_float_3dnow)
+	global	EXTN(jsimd_convsamp_float_3dnow) PRIVATE
 
 EXTN(jsimd_convsamp_float_3dnow):
 	push	ebp
@@ -138,7 +138,7 @@
 %define workspace	ebp+16		; FAST_FLOAT * workspace
 
 	align	16
-	global	EXTN(jsimd_quantize_float_3dnow)
+	global	EXTN(jsimd_quantize_float_3dnow) PRIVATE
 
 EXTN(jsimd_quantize_float_3dnow):
 	push	ebp
Index: simd/jcsamss2.asm
===================================================================
--- simd/jcsamss2.asm	(revision 677)
+++ simd/jcsamss2.asm	(working copy)
@@ -40,7 +40,7 @@
 %define output_data(b)	(b)+28		; JSAMPARRAY output_data
 
 	align	16
-	global	EXTN(jsimd_h2v1_downsample_sse2)
+	global	EXTN(jsimd_h2v1_downsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v1_downsample_sse2):
 	push	ebp
@@ -195,7 +195,7 @@
 %define output_data(b)	(b)+28	; JSAMPARRAY output_data
 
 	align	16
-	global	EXTN(jsimd_h2v2_downsample_sse2)
+	global	EXTN(jsimd_h2v2_downsample_sse2) PRIVATE
 
 EXTN(jsimd_h2v2_downsample_sse2):
 	push	ebp
Index: simd/jimmxint.asm
===================================================================
--- simd/jimmxint.asm	(revision 677)
+++ simd/jimmxint.asm	(working copy)
@@ -66,7 +66,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_idct_islow_mmx)
+	global	EXTN(jconst_idct_islow_mmx) PRIVATE
 
 EXTN(jconst_idct_islow_mmx):
 
@@ -107,7 +107,7 @@
 					; JCOEF workspace[DCTSIZE2]
 
 	align	16
-	global	EXTN(jsimd_idct_islow_mmx)
+	global	EXTN(jsimd_idct_islow_mmx) PRIVATE
 
 EXTN(jsimd_idct_islow_mmx):
 	push	ebp
Index: simd/jcgrymmx.asm
===================================================================
--- simd/jcgrymmx.asm	(revision 677)
+++ simd/jcgrymmx.asm	(working copy)
@@ -41,7 +41,7 @@
 %define gotptr		wk(0)-SIZEOF_POINTER	; void * gotptr
 
 	align	16
-	global	EXTN(jsimd_rgb_gray_convert_mmx)
+	global	EXTN(jsimd_rgb_gray_convert_mmx) PRIVATE
 
 EXTN(jsimd_rgb_gray_convert_mmx):
 	push	ebp
Index: simd/jfss2int.asm
===================================================================
--- simd/jfss2int.asm	(revision 677)
+++ simd/jfss2int.asm	(working copy)
@@ -66,7 +66,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_fdct_islow_sse2)
+	global	EXTN(jconst_fdct_islow_sse2) PRIVATE
 
 EXTN(jconst_fdct_islow_sse2):
 
@@ -101,7 +101,7 @@
 %define WK_NUM		6
 
 	align	16
-	global	EXTN(jsimd_fdct_islow_sse2)
+	global	EXTN(jsimd_fdct_islow_sse2) PRIVATE
 
 EXTN(jsimd_fdct_islow_sse2):
 	push	ebp
Index: simd/jcgryss2.asm
===================================================================
--- simd/jcgryss2.asm	(revision 677)
+++ simd/jcgryss2.asm	(working copy)
@@ -39,7 +39,7 @@
 
 	align	16
 
-	global	EXTN(jsimd_rgb_gray_convert_sse2)
+	global	EXTN(jsimd_rgb_gray_convert_sse2) PRIVATE
 
 EXTN(jsimd_rgb_gray_convert_sse2):
 	push	ebp
Index: simd/jccolmmx.asm
===================================================================
--- simd/jccolmmx.asm	(revision 677)
+++ simd/jccolmmx.asm	(working copy)
@@ -37,7 +37,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_rgb_ycc_convert_mmx)
+	global	EXTN(jconst_rgb_ycc_convert_mmx) PRIVATE
 
 EXTN(jconst_rgb_ycc_convert_mmx):
 
Index: simd/jimmxred.asm
===================================================================
--- simd/jimmxred.asm	(revision 677)
+++ simd/jimmxred.asm	(working copy)
@@ -72,7 +72,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_idct_red_mmx)
+	global	EXTN(jconst_idct_red_mmx) PRIVATE
 
 EXTN(jconst_idct_red_mmx):
 
@@ -115,7 +115,7 @@
 					; JCOEF workspace[DCTSIZE2]
 
 	align	16
-	global	EXTN(jsimd_idct_4x4_mmx)
+	global	EXTN(jsimd_idct_4x4_mmx) PRIVATE
 
 EXTN(jsimd_idct_4x4_mmx):
 	push	ebp
@@ -503,7 +503,7 @@
 %define output_col(b)	(b)+20		; JDIMENSION output_col
 
 	align	16
-	global	EXTN(jsimd_idct_2x2_mmx)
+	global	EXTN(jsimd_idct_2x2_mmx) PRIVATE
 
 EXTN(jsimd_idct_2x2_mmx):
 	push	ebp
Index: simd/jsimdext.inc
===================================================================
--- simd/jsimdext.inc	(revision 677)
+++ simd/jsimdext.inc	(working copy)
@@ -373,4 +373,14 @@
 ;
 %include "jsimdcfg.inc"
 
+; Begin chromium edits
+%ifdef MACHO ; ----(nasm -fmacho -DMACHO ...)--------
+%define PRIVATE :private_extern
+%elifdef ELF ; ----(nasm -felf[64] -DELF ...)------------
+%define PRIVATE :hidden
+%else
+%define PRIVATE
+%endif
+; End chromium edits
+
 ; --------------------------------------------------------------------------
Index: simd/jdclrmmx.asm
===================================================================
--- simd/jdclrmmx.asm	(revision 677)
+++ simd/jdclrmmx.asm	(working copy)
@@ -40,7 +40,7 @@
 %define gotptr		wk(0)-SIZEOF_POINTER	; void * gotptr
 
 	align	16
-	global	EXTN(jsimd_ycc_rgb_convert_mmx)
+	global	EXTN(jsimd_ycc_rgb_convert_mmx) PRIVATE
 
 EXTN(jsimd_ycc_rgb_convert_mmx):
 	push	ebp
Index: simd/jccolss2.asm
===================================================================
--- simd/jccolss2.asm	(revision 677)
+++ simd/jccolss2.asm	(working copy)
@@ -34,7 +34,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_rgb_ycc_convert_sse2)
+	global	EXTN(jconst_rgb_ycc_convert_sse2) PRIVATE
 
 EXTN(jconst_rgb_ycc_convert_sse2):
 
Index: simd/jisseflt.asm
===================================================================
--- simd/jisseflt.asm	(revision 677)
+++ simd/jisseflt.asm	(working copy)
@@ -37,7 +37,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_idct_float_sse)
+	global	EXTN(jconst_idct_float_sse) PRIVATE
 
 EXTN(jconst_idct_float_sse):
 
@@ -73,7 +73,7 @@
 					; FAST_FLOAT workspace[DCTSIZE2]
 
 	align	16
-	global	EXTN(jsimd_idct_float_sse)
+	global	EXTN(jsimd_idct_float_sse) PRIVATE
 
 EXTN(jsimd_idct_float_sse):
 	push	ebp
Index: simd/jcqnts2i-64.asm
===================================================================
--- simd/jcqnts2i-64.asm	(revision 677)
+++ simd/jcqnts2i-64.asm	(working copy)
@@ -36,7 +36,7 @@
 ; r12 = DCTELEM * workspace
 
 	align	16
-	global	EXTN(jsimd_convsamp_sse2)
+	global	EXTN(jsimd_convsamp_sse2) PRIVATE
 
 EXTN(jsimd_convsamp_sse2):
 	push	rbp
@@ -112,7 +112,7 @@
 ; r12 = DCTELEM * workspace
 
 	align	16
-	global	EXTN(jsimd_quantize_sse2)
+	global	EXTN(jsimd_quantize_sse2) PRIVATE
 
 EXTN(jsimd_quantize_sse2):
 	push	rbp
Index: simd/jdclrss2.asm
===================================================================
--- simd/jdclrss2.asm	(revision 677)
+++ simd/jdclrss2.asm	(working copy)
@@ -40,7 +40,7 @@
 %define gotptr		wk(0)-SIZEOF_POINTER	; void * gotptr
 
 	align	16
-	global	EXTN(jsimd_ycc_rgb_convert_sse2)
+	global	EXTN(jsimd_ycc_rgb_convert_sse2) PRIVATE
 
 EXTN(jsimd_ycc_rgb_convert_sse2):
 	push	ebp
@@ -302,6 +302,41 @@
 	movdqa	xmmA,xmmD
 	sub	ecx, byte SIZEOF_XMMWORD
 .column_st15:
+%ifdef STRICT_MEMORY_ACCESS
+	; Store the lower 8 bytes of xmmA to the output when it has enough
+	; space.
+	cmp	ecx, byte SIZEOF_MMWORD
+	jb	short .column_st7
+	movq	MMWORD [edi], xmmA
+	add	edi, byte SIZEOF_MMWORD
+	sub	ecx, byte SIZEOF_MMWORD
+	psrldq	xmmA, SIZEOF_MMWORD
+.column_st7:
+	; Store the lower 4 bytes of xmmA to the output when it has enough
+	; space.
+	cmp	ecx, byte SIZEOF_DWORD
+	jb	short .column_st3
+	movd	DWORD [edi], xmmA
+	add	edi, byte SIZEOF_DWORD
+	sub	ecx, byte SIZEOF_DWORD
+	psrldq	xmmA, SIZEOF_DWORD
+.column_st3:
+	; Store the lower 2 bytes of eax to the output when it has enough
+	; space.
+	movd	eax, xmmA
+	cmp	ecx, byte SIZEOF_WORD
+	jb	short .column_st1
+	mov	WORD [edi], ax
+	add	edi, byte SIZEOF_WORD
+	sub	ecx, byte SIZEOF_WORD
+	shr	eax, 16
+.column_st1:
+	; Store the lower 1 byte of eax to the output when it has enough
+	; space.
+	test	ecx, ecx
+	jz	short .nextrow
+	mov	BYTE [edi], al
+%else
 	mov	eax,ecx
 	xor	ecx, byte 0x0F
 	shl	ecx, 2
@@ -341,6 +376,7 @@
 	por	xmmE,xmmC
 .adj0:	; ----------------
 	maskmovdqu xmmA,xmmE			; movntdqu XMMWORD [edi], xmmA
+%endif ; STRICT_MEMORY_ACCESS ; ---------------
 
 %else ; RGB_PIXELSIZE == 4 ; -----------
 
@@ -426,6 +462,22 @@
 	movdqa	xmmA,xmmD
 	sub	ecx, byte SIZEOF_XMMWORD/4
 .column_st15:
+%ifdef STRICT_MEMORY_ACCESS
+	; Store two pixels (8 bytes) of xmmA to the output when it has enough
+	; space.
+	cmp	ecx, byte SIZEOF_XMMWORD/8
+	jb	short .column_st7
+	movq	MMWORD [edi], xmmA
+	add	edi, byte SIZEOF_XMMWORD/8*4
+	sub	ecx, byte SIZEOF_XMMWORD/8
+	psrldq	xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+	; Store one pixel (4 bytes) of xmmA to the output when it has enough
+	; space.
+	test	ecx, ecx
+	jz	short .nextrow
+	movd	DWORD [edi], xmmA
+%else
 	cmp	ecx, byte SIZEOF_XMMWORD/16
 	jb	short .nextrow
 	mov	eax,ecx
@@ -465,6 +517,7 @@
 	por	xmmE,xmmG
 .adj0:	; ----------------
 	maskmovdqu xmmA,xmmE			; movntdqu XMMWORD [edi], xmmA
+%endif ; STRICT_MEMORY_ACCESS ; ---------------
 
 %endif ; RGB_PIXELSIZE ; ---------------
 
Index: simd/jcqntsse.asm
===================================================================
--- simd/jcqntsse.asm	(revision 677)
+++ simd/jcqntsse.asm	(working copy)
@@ -35,7 +35,7 @@
 %define workspace	ebp+16		; FAST_FLOAT * workspace
 
 	align	16
-	global	EXTN(jsimd_convsamp_float_sse)
+	global	EXTN(jsimd_convsamp_float_sse) PRIVATE
 
 EXTN(jsimd_convsamp_float_sse):
 	push	ebp
@@ -138,7 +138,7 @@
 %define workspace	ebp+16		; FAST_FLOAT * workspace
 
 	align	16
-	global	EXTN(jsimd_quantize_float_sse)
+	global	EXTN(jsimd_quantize_float_sse) PRIVATE
 
 EXTN(jsimd_quantize_float_sse):
 	push	ebp
Index: simd/jiss2int-64.asm
===================================================================
--- simd/jiss2int-64.asm	(revision 677)
+++ simd/jiss2int-64.asm	(working copy)
@@ -67,7 +67,7 @@
 	SECTION	SEG_CONST
 
 	alignz	16
-	global	EXTN(jconst_idct_islow_sse2)
+	global	EXTN(jconst_idct_islow_sse2) PRIVATE
 
 EXTN(jconst_idct_islow_sse2):
 
@@ -106,7 +106,7 @@
 %define WK_NUM		12
 
 	align	16
-	global	EXTN(jsimd_idct_islow_sse2)
+	global	EXTN(jsimd_idct_islow_sse2) PRIVATE
 
 EXTN(jsimd_idct_islow_sse2):
 	push	rbp
Index: simd/jfmmxfst.asm
===================================================================
--- simd/jfmmxfst.asm	(revision 677)
+++ simd/jfmmxfst.asm	(working copy)
@@ -52,7 +52,7 @@
 %define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
 
 	alignz	16
-	global	EXTN(jconst_fdct_ifast_mmx)
+	global	EXTN(jconst_fdct_ifast_mmx) PRIVATE
 
 EXTN(jconst_fdct_ifast_mmx):
 
@@ -80,7 +80,7 @@
 %define WK_NUM		2
 
 	align	16
-	global	EXTN(jsimd_fdct_ifast_mmx)
+	global	EXTN(jsimd_fdct_ifast_mmx) PRIVATE
 
 EXTN(jsimd_fdct_ifast_mmx):
 	push	ebp