Fix yasm warnings in libjpeg-turbo.

google.patch

 Index: jmorecfg.h
 ===================================================================
 --- jmorecfg.h  (revision 64575)
 +++ jmorecfg.h  (working copy)
 @@ -159,8 +159,10 @@
  /* 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;
@@ skipping 27 matching lines @@
 +/* 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 73971)
-@@ -0,0 +1,115 @@
++++ jpeglibmangler.h    (revision 78340)
+@@ -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.
 +
@@ skipping 91 matching lines @@
 +#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
-+#define jpeg_std_message_table chromium_jpeg_std_message_table
-+#define jpeg_natural_order chromium_jpeg_natural_order
 +
 +#endif  // THIRD_PARTY_LIBJPEG_TURBO_JPEGLIBMANGLER_H_
 Index: jdhuff.c
 ===================================================================
 --- jdhuff.c    (revision 64575)
 +++ jdhuff.c    (working copy)
 @@ -658,7 +658,7 @@
        symbol |= GET_BITS(1); \
        size++; \
      } \
@@ skipping 121 matching lines @@
         align   16
 -       global  EXTN(jsimd_fdct_float_3dnow)
 +       global  EXTN(jsimd_fdct_float_3dnow) PRIVATE
  
  EXTN(jsimd_fdct_float_3dnow):
         push    ebp
 Index: simd/jcclrss2-64.asm
 ===================================================================
 --- simd/jcclrss2-64.asm        (revision 64575)
 +++ simd/jcclrss2-64.asm        (working copy)
-@@ -39,7 +39,7 @@
+@@ -17,8 +17,12 @@
+ %include "jcolsamp.inc"
+ 
+ ; --------------------------------------------------------------------------
++%ifndef NEED_SECTION
++%define NEED_SECTION
+        SECTION SEG_TEXT
+        BITS    64
++%endif
++
+ ;
+ ; Convert some rows of samples to the output colorspace.
+ ;
+@@ -39,7 +43,7 @@
  
         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
 ===================================================================
@@ skipping 257 matching lines @@
         align   16
 -       global  EXTN(jsimd_h2v2_downsample_sse2)
 +       global  EXTN(jsimd_h2v2_downsample_sse2) PRIVATE
  
  EXTN(jsimd_h2v2_downsample_sse2):
         push    ebp
 Index: simd/jdmrgmmx.asm
 ===================================================================
 --- simd/jdmrgmmx.asm   (revision 64575)
 +++ simd/jdmrgmmx.asm   (working copy)
-@@ -42,7 +42,7 @@
+@@ -19,8 +19,12 @@
+ %include "jcolsamp.inc"
+ 
+ ; --------------------------------------------------------------------------
++%ifndef NEED_SECTION
++%define NEED_SECTION
+        SECTION SEG_TEXT
+        BITS    32
++%endif
++
+ ;
+ ; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+ ;
+@@ -42,7 +46,7 @@
  %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
-@@ -411,7 +411,7 @@
+@@ -411,7 +415,7 @@
  %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
 ===================================================================
@@ skipping 107 matching lines @@
         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 64575)
 +++ simd/jdmrgss2.asm   (working copy)
-@@ -42,7 +42,7 @@
+@@ -19,8 +19,12 @@
+ %include "jcolsamp.inc"
+                                
+ ; --------------------------------------------------------------------------
++%ifndef NEED_SECTION
++%define NEED_SECTION
+        SECTION SEG_TEXT
+        BITS    32
++%endif
++
+ ;
+ ; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+ ;
+@@ -42,7 +46,7 @@
  %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
-@@ -309,6 +309,41 @@
+@@ -309,6 +313,41 @@
         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
@@ skipping 15 matching lines @@
 +       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
++       jz      short .endcolumn
 +       mov     BYTE [edi], al
 +%else
         mov     eax,ecx
         xor     ecx, byte 0x0F
         shl     ecx, 2
-@@ -348,6 +383,7 @@
+@@ -348,6 +387,7 @@
         por     xmmE,xmmC
  .adj0: ; ----------------
         maskmovdqu xmmA,xmmE                    ; movntdqu XMMWORD [edi], xmmA
 +%endif ; STRICT_MEMORY_ACCESS ; ---------------
  
  %else ; RGB_PIXELSIZE == 4 ; -----------
  
-@@ -436,6 +472,22 @@
+@@ -436,6 +476,22 @@
         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
++       jz      short .endcolumn
 +       movd    DWORD [edi], xmmA
 +%else
         cmp     ecx, byte SIZEOF_XMMWORD/16
         jb      short .endcolumn
         mov     eax,ecx
-@@ -475,6 +527,7 @@
+@@ -475,6 +531,7 @@
         por     xmmE,xmmG
  .adj0: ; ----------------
         maskmovdqu xmmA,xmmE                    ; movntdqu XMMWORD [edi], xmmA
 +%endif ; STRICT_MEMORY_ACCESS ; ---------------
  
  %endif ; RGB_PIXELSIZE ; ---------------
  
-@@ -509,7 +562,7 @@
+@@ -509,7 +566,7 @@
  %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
 ===================================================================
@@ skipping 229 matching lines @@
         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/jdclrss2-64.asm
 ===================================================================
 --- simd/jdclrss2-64.asm        (revision 64575)
 +++ simd/jdclrss2-64.asm        (working copy)
-@@ -41,7 +41,7 @@
+@@ -20,8 +20,12 @@
+ %include "jcolsamp.inc"
+                                
+ ; --------------------------------------------------------------------------
++%ifndef NEED_SECTION
++%define NEED_SECTION
+        SECTION SEG_TEXT
+        BITS    64
++%endif
++
+ ;
+ ; Convert some rows of samples to the output colorspace.
+ ;
+@@ -41,7 +45,7 @@
  %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
-@@ -292,6 +292,41 @@
+@@ -292,6 +296,41 @@
         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
@@ skipping 21 matching lines @@
 +.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
-@@ -331,6 +366,7 @@
+@@ -331,6 +370,7 @@
         por     xmmE,xmmC
  .adj0: ; ----------------
         maskmovdqu xmmA,xmmE                    ; movntdqu XMMWORD [rdi], xmmA
 +%endif ; STRICT_MEMORY_ACCESS ; ---------------
  
  %else ; RGB_PIXELSIZE == 4 ; -----------
  
-@@ -415,6 +451,22 @@
+@@ -415,6 +455,22 @@
         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
-@@ -454,6 +506,7 @@
+@@ -454,6 +510,7 @@
         por     xmmE,xmmG
  .adj0: ; ----------------
         maskmovdqu xmmA,xmmE                    ; movntdqu XMMWORD [rdi], xmmA
 +%endif ; STRICT_MEMORY_ACCESS ; ---------------
  
  %endif ; RGB_PIXELSIZE ; ---------------
  
 Index: simd/jcclrmmx.asm
 ===================================================================
 --- simd/jcclrmmx.asm   (revision 64575)
 +++ simd/jcclrmmx.asm   (working copy)
-@@ -42,7 +42,7 @@
+@@ -19,8 +19,12 @@
+ %include "jcolsamp.inc"
+ 
+ ; --------------------------------------------------------------------------
++%ifndef NEED_SECTION
++%define NEED_SECTION
+        SECTION SEG_TEXT
+        BITS    32
++%endif
++
+ ;
+ ; Convert some rows of samples to the output colorspace.
+ ;
+@@ -42,7 +46,7 @@
  %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/jimmxred.asm
 ===================================================================
@@ skipping 23 matching lines @@
         align   16
 -       global  EXTN(jsimd_idct_2x2_mmx)
 +       global  EXTN(jsimd_idct_2x2_mmx) PRIVATE
  
  EXTN(jsimd_idct_2x2_mmx):
         push    ebp
 Index: simd/jdclrmmx.asm
 ===================================================================
 --- simd/jdclrmmx.asm   (revision 64575)
 +++ simd/jdclrmmx.asm   (working copy)
-@@ -42,7 +42,7 @@
+@@ -19,8 +19,12 @@
+ %include "jcolsamp.inc"
+ 
+ ; --------------------------------------------------------------------------
++%ifndef NEED_SECTION
++%define NEED_SECTION
+        SECTION SEG_TEXT
+        BITS    32
++%endif
++
+ ;
+ ; Convert some rows of samples to the output colorspace.
+ ;
+@@ -42,7 +46,7 @@
  %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/jfsseflt.asm
 ===================================================================
@@ skipping 14 matching lines @@
         align   16
 -       global  EXTN(jsimd_fdct_float_sse)
 +       global  EXTN(jsimd_fdct_float_sse) PRIVATE
  
  EXTN(jsimd_fdct_float_sse):
         push    ebp
 Index: simd/jsimdext.inc
 ===================================================================
 --- simd/jsimdext.inc   (revision 64575)
 +++ simd/jsimdext.inc   (working copy)
+@@ -38,16 +38,16 @@
+ 
+ ; -- segment definition --
+ ;
+-%define SEG_TEXT    .text  align=16 public use32 class=CODE
+-%define SEG_CONST   .rdata align=16 public use32 class=CONST
++%define SEG_TEXT    .text  align=16 ; public use32 class=CODE
++%define SEG_CONST   .rdata align=16 ; public use32 class=CONST
+ 
+ %elifdef WIN64 ; ----(nasm -fwin64 -DWIN64 ...)--------
+ ; * Microsoft Visual C++
+ 
+ ; -- segment definition --
+ ;
+-%define SEG_TEXT    .text  align=16 public use64 class=CODE
+-%define SEG_CONST   .rdata align=16 public use64 class=CONST
++%define SEG_TEXT    .text  align=16 ; public use64 class=CODE
++%define SEG_CONST   .rdata align=16 ; public use64 class=CONST
+ %ifdef MSVC
+ %define EXTN(name)  name                       ; foo() -> foo
+ %endif
 @@ -369,4 +369,14 @@
  ;
  %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/jdmrgss2-64.asm
 ===================================================================
 --- simd/jdmrgss2-64.asm        (revision 64575)
 +++ simd/jdmrgss2-64.asm        (working copy)
-@@ -41,7 +41,7 @@
+@@ -20,8 +20,12 @@
+ %include "jcolsamp.inc"
+                                
+ ; --------------------------------------------------------------------------
++%ifndef NEED_SECTION
++%define NEED_SECTION
+        SECTION SEG_TEXT
+        BITS    64
++%endif
++
+ ;
+ ; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+ ;
+@@ -41,7 +45,7 @@
  %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
-@@ -296,6 +296,41 @@
+@@ -296,6 +300,41 @@
         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
@@ skipping 15 matching lines @@
 +       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
++       jz      short .endcolumn
 +       mov     BYTE [rdi], al
 +%else
         mov     rax,rcx
         xor     rcx, byte 0x0F
         shl     rcx, 2
-@@ -335,6 +370,7 @@
+@@ -335,6 +374,7 @@
         por     xmmE,xmmC
  .adj0: ; ----------------
         maskmovdqu xmmA,xmmE                    ; movntdqu XMMWORD [edi], xmmA
 +%endif ; STRICT_MEMORY_ACCESS ; ---------------
  
  %else ; RGB_PIXELSIZE == 4 ; -----------
  
-@@ -422,6 +458,22 @@
+@@ -422,6 +462,22 @@
         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
++       jz      short .endcolumn
 +       movd    DWORD [rdi], xmmA
 +%else
         cmp     rcx, byte SIZEOF_XMMWORD/16
         jb      near .endcolumn
         mov     rax,rcx
-@@ -461,6 +513,7 @@
+@@ -461,6 +517,7 @@
         por     xmmE,xmmG
  .adj0: ; ----------------
         maskmovdqu xmmA,xmmE                    ; movntdqu XMMWORD [edi], xmmA
 +%endif ; STRICT_MEMORY_ACCESS ; ---------------
  
  %endif ; RGB_PIXELSIZE ; ---------------
  
-@@ -492,7 +545,7 @@
+@@ -492,7 +549,7 @@
  ; 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/jccolss2.asm
 ===================================================================
@@ skipping 75 matching lines @@
         align   16
 -       global  EXTN(jsimd_quantize_sse2)
 +       global  EXTN(jsimd_quantize_sse2) PRIVATE
  
  EXTN(jsimd_quantize_sse2):
         push    rbp
 Index: simd/jcclrss2.asm
 ===================================================================
 --- simd/jcclrss2.asm   (revision 64575)
 +++ simd/jcclrss2.asm   (working copy)
-@@ -40,7 +40,7 @@
+@@ -16,8 +16,12 @@
+ %include "jcolsamp.inc"
+ 
+ ; --------------------------------------------------------------------------
++%ifndef NEED_SECTION
++%define NEED_SECTION
+        SECTION SEG_TEXT
+        BITS    32
++%endif
++
+ ;
+ ; Convert some rows of samples to the output colorspace.
+ ;
+@@ -40,7 +44,7 @@
  
         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
 ===================================================================
@@ skipping 23 matching lines @@
         align   16
 -       global  EXTN(jsimd_idct_2x2_sse2)
 +       global  EXTN(jsimd_idct_2x2_sse2) PRIVATE
  
  EXTN(jsimd_idct_2x2_sse2):
         push    ebp
 Index: simd/jdclrss2.asm
 ===================================================================
 --- simd/jdclrss2.asm   (revision 64575)
 +++ simd/jdclrss2.asm   (working copy)
-@@ -42,7 +42,7 @@
+@@ -19,8 +19,12 @@
+ %include "jcolsamp.inc"
+                                
+ ; --------------------------------------------------------------------------
++%ifndef NEED_SECTION
++%define NEED_SECTION
+        SECTION SEG_TEXT
+        BITS    32
++%endif
++
+ ;
+ ; Convert some rows of samples to the output colorspace.
+ ;
+@@ -42,7 +46,7 @@
  %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
-@@ -304,6 +304,41 @@
+@@ -304,6 +308,41 @@
         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
@@ skipping 21 matching lines @@
 +.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
-@@ -343,6 +378,7 @@
+@@ -343,6 +382,7 @@
         por     xmmE,xmmC
  .adj0: ; ----------------
         maskmovdqu xmmA,xmmE                    ; movntdqu XMMWORD [edi], xmmA
 +%endif ; STRICT_MEMORY_ACCESS ; ---------------
  
  %else ; RGB_PIXELSIZE == 4 ; -----------
  
-@@ -428,6 +464,22 @@
+@@ -428,6 +468,22 @@
         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
-@@ -467,6 +519,7 @@
+@@ -467,6 +523,7 @@
         por     xmmE,xmmG
  .adj0: ; ----------------
         maskmovdqu xmmA,xmmE                    ; movntdqu XMMWORD [edi], xmmA
 +%endif ; STRICT_MEMORY_ACCESS ; ---------------
  
  %endif ; RGB_PIXELSIZE ; ---------------
  
 Index: simd/jdmerss2.asm
 ===================================================================
 --- simd/jdmerss2.asm   (revision 64575)
@@ skipping 88 matching lines @@
  
 @@ -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