clear out gorp

26 files changed, 546 insertions, 41 deletions

diff --git a/simd/CMakeLists.txt b/simd/CMakeLists.txt
index 263579d..397a9f3 100755
--- a/simd/CMakeLists.txt
+++ b/simd/CMakeLists.txt
@@ -18,16 +18,16 @@ if(CMAKE_BUILD_TYPE STREQUAL "Debug"
 endif()
 
 if(SIMD_X86_64)
-  set(SIMD_BASENAMES jfsseflt-64 jccolss2-64 jdcolss2-64 jcsamss2-64
-    jdsamss2-64 jdmerss2-64 jcqnts2i-64 jfss2fst-64 jfss2int-64 jiss2red-64
-    jiss2int-64 jiss2fst-64 jcqnts2f-64 jiss2flt-64)
+  set(SIMD_BASENAMES jfsseflt-64 jccolss2-64 jdcolss2-64 jcgrass2-64
+    jcsamss2-64 jdsamss2-64 jdmerss2-64 jcqnts2i-64 jfss2fst-64 jfss2int-64
+    jiss2red-64 jiss2int-64 jiss2fst-64 jcqnts2f-64 jiss2flt-64)
   message(STATUS "Building x86_64 SIMD extensions")
 else()
-  set(SIMD_BASENAMES jsimdcpu jccolmmx jdcolmmx jcsammmx jdsammmx jdmermmx
-    jcqntmmx jfmmxfst jfmmxint jimmxred jimmxint jimmxfst jcqnt3dn jf3dnflt
-    ji3dnflt jcqntsse jfsseflt jisseflt jccolss2 jdcolss2 jcsamss2 jdsamss2
-    jdmerss2 jcqnts2i jfss2fst jfss2int jiss2red jiss2int jiss2fst jcqnts2f
-    jiss2flt)
+  set(SIMD_BASENAMES jsimdcpu jccolmmx jcgrammx jdcolmmx jcsammmx jdsammmx
+    jdmermmx jcqntmmx jfmmxfst jfmmxint jimmxred jimmxint jimmxfst jcqnt3dn
+    jf3dnflt ji3dnflt jcqntsse jfsseflt jisseflt jccolss2 jcgrass2 jdcolss2
+    jcsamss2 jdsamss2 jdmerss2 jcqnts2i jfss2fst jfss2int jiss2red jiss2int
+    jiss2fst jcqnts2f jiss2flt)
   message(STATUS "Building i386 SIMD extensions")
 endif()
 
@@ -48,6 +48,10 @@ foreach(file ${SIMD_BASENAMES})
     set(DEPFILE ${SIMD_SRC})
     string(REGEX REPLACE "mer" "mrg" DEPFILE ${DEPFILE})
   endif()
+  if(${file} MATCHES gra)
+    set(DEPFILE ${SIMD_SRC})
+    string(REGEX REPLACE "gra" "gry" DEPFILE ${DEPFILE})
+  endif()
   set(SIMD_OBJ ${OBJDIR}/${file}.obj)
   add_custom_command(OUTPUT ${SIMD_OBJ} DEPENDS ${SIMD_SRC} ${DEPFILE} *.inc
     COMMAND ${NASM} ${NAFLAGS} ${SIMD_SRC} -o${SIMD_OBJ})
diff --git a/simd/Makefile.am b/simd/Makefile.am
index 81c23af..fbba807 100644
--- a/simd/Makefile.am
+++ b/simd/Makefile.am
@@ -4,14 +4,14 @@ BUILT_SOURCES = jsimdcfg.inc
 
 EXTRA_DIST = nasm_lt.sh jcclrmmx.asm jcclrss2.asm jdclrmmx.asm jdclrss2.asm \
 	jdmrgmmx.asm jdmrgss2.asm jcclrss2-64.asm jdclrss2-64.asm \
-	jdmrgss2-64.asm CMakeLists.txt
+	jdmrgss2-64.asm jcgryss2-64.asm jcgrymmx.asm jcgryss2.asm CMakeLists.txt
 
 if SIMD_X86_64
 
 libsimd_la_SOURCES = jsimd_x86_64.c \
 	jsimd.h jsimdcfg.inc.h \
 	jsimdext.inc jcolsamp.inc jdct.inc \
-	jfsseflt-64.asm \
+	jfsseflt-64.asm jcgrass2-64.asm \
 	jccolss2-64.asm jdcolss2-64.asm \
 	jcsamss2-64.asm jdsamss2-64.asm jdmerss2-64.asm \
 	jcqnts2i-64.asm jfss2fst-64.asm jfss2int-64.asm \
@@ -20,6 +20,7 @@ libsimd_la_SOURCES = jsimd_x86_64.c \
 
 jccolss2-64.lo: jcclrss2-64.asm
 jdcolss2-64.lo: jdclrss2-64.asm
+jcgrass2-64.lo: jcgryss2-64.asm
 jdmerss2-64.lo: jdmrgss2-64.asm
 endif
 
@@ -29,26 +30,34 @@ libsimd_la_SOURCES = jsimd_i386.c \
 	jsimd.h jsimdcfg.inc.h \
 	jsimdext.inc jcolsamp.inc jdct.inc \
 	jsimdcpu.asm \
-	jccolmmx.asm jdcolmmx.asm \
+	jccolmmx.asm jdcolmmx.asm jcgrammx.asm \
 	jcsammmx.asm jdsammmx.asm jdmermmx.asm \
 	jcqntmmx.asm jfmmxfst.asm jfmmxint.asm \
 	jimmxred.asm jimmxint.asm jimmxfst.asm \
 	jcqnt3dn.asm jf3dnflt.asm ji3dnflt.asm \
 	jcqntsse.asm jfsseflt.asm jisseflt.asm \
-	jccolss2.asm jdcolss2.asm \
+	jccolss2.asm jdcolss2.asm jcgrass2.asm \
 	jcsamss2.asm jdsamss2.asm jdmerss2.asm \
 	jcqnts2i.asm jfss2fst.asm jfss2int.asm \
 	jiss2red.asm jiss2int.asm jiss2fst.asm \
 	jcqnts2f.asm jiss2flt.asm
 
 jccolmmx.lo: jcclrmmx.asm
+jcgrammx.lo: jcgrymmx.asm
 jccolss2.lo: jcclrss2.asm
+jcgrass2.lo: jcgryss2.asm
 jdcolmmx.lo: jdclrmmx.asm
 jdcolss2.lo: jdclrss2.asm
 jdmermmx.lo: jdmrgmmx.asm
 jdmerss2.lo: jdmrgss2.asm
 endif
 
+if SIMD_ARM
+
+libsimd_la_SOURCES = jsimd_arm.c jsimd_arm_neon.S
+
+endif
+
 AM_CPPFLAGS = -I$(top_srcdir) 
 
 .asm.lo:
diff --git a/simd/jcclrmmx.asm b/simd/jcclrmmx.asm
index b6b8912..e095253 100644
--- a/simd/jcclrmmx.asm
+++ b/simd/jcclrmmx.asm
@@ -19,8 +19,6 @@
 %include "jcolsamp.inc"
 
 ; --------------------------------------------------------------------------
-	SECTION	SEG_TEXT
-	BITS	32
 ;
 ; Convert some rows of samples to the output colorspace.
 ;
diff --git a/simd/jcclrss2-64.asm b/simd/jcclrss2-64.asm
index 8ca47aa..f5d6bed 100644
--- a/simd/jcclrss2-64.asm
+++ b/simd/jcclrss2-64.asm
@@ -17,8 +17,6 @@
 %include "jcolsamp.inc"
 
 ; --------------------------------------------------------------------------
-	SECTION	SEG_TEXT
-	BITS	64
 ;
 ; Convert some rows of samples to the output colorspace.
 ;
diff --git a/simd/jcclrss2.asm b/simd/jcclrss2.asm
index 8def718..517b705 100644
--- a/simd/jcclrss2.asm
+++ b/simd/jcclrss2.asm
@@ -16,8 +16,6 @@
 %include "jcolsamp.inc"
 
 ; --------------------------------------------------------------------------
-	SECTION	SEG_TEXT
-	BITS	32
 ;
 ; Convert some rows of samples to the output colorspace.
 ;
diff --git a/simd/jccolmmx.asm b/simd/jccolmmx.asm
index 5e7f3be..d0f0d79 100644
--- a/simd/jccolmmx.asm
+++ b/simd/jccolmmx.asm
@@ -51,6 +51,9 @@ PD_ONEHALF	times 2 dd  (1 << (SCALEBITS-1))
 	alignz	16
 
 ; --------------------------------------------------------------------------
+	SECTION	SEG_TEXT
+	BITS	32
+
 %include "jcclrmmx.asm"
 
 %undef RGB_RED
diff --git a/simd/jccolss2-64.asm b/simd/jccolss2-64.asm
index 64ee0ba..18de456 100644
--- a/simd/jccolss2-64.asm
+++ b/simd/jccolss2-64.asm
@@ -48,6 +48,9 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 	alignz	16
 
 ; --------------------------------------------------------------------------
+	SECTION	SEG_TEXT
+	BITS	64
+
 %include "jcclrss2-64.asm"
 
 %undef RGB_RED
diff --git a/simd/jccolss2.asm b/simd/jccolss2.asm
index 8d1f734..7acb59c 100644
--- a/simd/jccolss2.asm
+++ b/simd/jccolss2.asm
@@ -48,6 +48,9 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 	alignz	16
 
 ; --------------------------------------------------------------------------
+	SECTION	SEG_TEXT
+	BITS	32
+
 %include "jcclrss2.asm"
 
 %undef RGB_RED
diff --git a/simd/jdclrmmx.asm b/simd/jdclrmmx.asm
index 79772e0..1c255e8 100644
--- a/simd/jdclrmmx.asm
+++ b/simd/jdclrmmx.asm
@@ -19,8 +19,6 @@
 %include "jcolsamp.inc"
 
 ; --------------------------------------------------------------------------
-	SECTION	SEG_TEXT
-	BITS	32
 ;
 ; Convert some rows of samples to the output colorspace.
 ;
diff --git a/simd/jdclrss2-64.asm b/simd/jdclrss2-64.asm
index 4282bd2..fdb33a3 100644
--- a/simd/jdclrss2-64.asm
+++ b/simd/jdclrss2-64.asm
@@ -20,8 +20,6 @@
 %include "jcolsamp.inc"
 				
 ; --------------------------------------------------------------------------
-	SECTION	SEG_TEXT
-	BITS	64
 ;
 ; Convert some rows of samples to the output colorspace.
 ;
diff --git a/simd/jdclrss2.asm b/simd/jdclrss2.asm
index 865fa82..3059d7d 100644
--- a/simd/jdclrss2.asm
+++ b/simd/jdclrss2.asm
@@ -19,8 +19,6 @@
 %include "jcolsamp.inc"
 				
 ; --------------------------------------------------------------------------
-	SECTION	SEG_TEXT
-	BITS	32
 ;
 ; Convert some rows of samples to the output colorspace.
 ;
diff --git a/simd/jdcolmmx.asm b/simd/jdcolmmx.asm
index 58775e8..0834bab 100644
--- a/simd/jdcolmmx.asm
+++ b/simd/jdcolmmx.asm
@@ -48,6 +48,9 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 	alignz	16
 
 ; --------------------------------------------------------------------------
+	SECTION	SEG_TEXT
+	BITS	32
+
 %include "jdclrmmx.asm"
 
 %undef RGB_RED
diff --git a/simd/jdcolss2-64.asm b/simd/jdcolss2-64.asm
index 2e97d59..d14a28a 100644
--- a/simd/jdcolss2-64.asm
+++ b/simd/jdcolss2-64.asm
@@ -48,6 +48,9 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 	alignz	16
 
 ; --------------------------------------------------------------------------
+	SECTION	SEG_TEXT
+	BITS	64
+
 %include "jdclrss2-64.asm"
 
 %undef RGB_RED
diff --git a/simd/jdcolss2.asm b/simd/jdcolss2.asm
index 7ae985d..cab4dd0 100644
--- a/simd/jdcolss2.asm
+++ b/simd/jdcolss2.asm
@@ -48,6 +48,9 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 	alignz	16
 
 ; --------------------------------------------------------------------------
+	SECTION	SEG_TEXT
+	BITS	32
+
 %include "jdclrss2.asm"
 
 %undef RGB_RED
diff --git a/simd/jdmermmx.asm b/simd/jdmermmx.asm
index fd587fb..75baaa5 100644
--- a/simd/jdmermmx.asm
+++ b/simd/jdmermmx.asm
@@ -48,6 +48,9 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 	alignz	16
 
 ; --------------------------------------------------------------------------
+	SECTION	SEG_TEXT
+	BITS	32
+
 %include "jdmrgmmx.asm"
 
 %undef RGB_RED
diff --git a/simd/jdmerss2-64.asm b/simd/jdmerss2-64.asm
index 1f0b10f..a1fe963 100644
--- a/simd/jdmerss2-64.asm
+++ b/simd/jdmerss2-64.asm
@@ -48,6 +48,9 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 	alignz	16
 
 ; --------------------------------------------------------------------------
+	SECTION	SEG_TEXT
+	BITS	64
+
 %include "jdmrgss2-64.asm"
 
 %undef RGB_RED
diff --git a/simd/jdmerss2.asm b/simd/jdmerss2.asm
index 2294e0d..21881b4 100644
--- a/simd/jdmerss2.asm
+++ b/simd/jdmerss2.asm
@@ -48,6 +48,9 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 	alignz	16
 
 ; --------------------------------------------------------------------------
+	SECTION	SEG_TEXT
+	BITS	32
+
 %include "jdmrgss2.asm"
 
 %undef RGB_RED
diff --git a/simd/jdmrgmmx.asm b/simd/jdmrgmmx.asm
index b5777a3..d0800a7 100644
--- a/simd/jdmrgmmx.asm
+++ b/simd/jdmrgmmx.asm
@@ -19,8 +19,6 @@
 %include "jcolsamp.inc"
 
 ; --------------------------------------------------------------------------
-	SECTION	SEG_TEXT
-	BITS	32
 ;
 ; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
 ;
diff --git a/simd/jdmrgss2-64.asm b/simd/jdmrgss2-64.asm
index 121bb82..0c2503f 100644
--- a/simd/jdmrgss2-64.asm
+++ b/simd/jdmrgss2-64.asm
@@ -20,8 +20,6 @@
 %include "jcolsamp.inc"
 				
 ; --------------------------------------------------------------------------
-	SECTION	SEG_TEXT
-	BITS	64
 ;
 ; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
 ;
diff --git a/simd/jdmrgss2.asm b/simd/jdmrgss2.asm
index 99b7eb9..368ac3c 100644
--- a/simd/jdmrgss2.asm
+++ b/simd/jdmrgss2.asm
@@ -19,8 +19,6 @@
 %include "jcolsamp.inc"
 				
 ; --------------------------------------------------------------------------
-	SECTION	SEG_TEXT
-	BITS	32
 ;
 ; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
 ;
diff --git a/simd/jsimd.h b/simd/jsimd.h
index 89ac1b7..39a0867 100644
--- a/simd/jsimd.h
+++ b/simd/jsimd.h
@@ -2,6 +2,7 @@
  * simd/jsimd.h
  *
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright 2011 D. R. Commander
  * 
  * Based on the x86 SIMD extension for IJG JPEG library,
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
@@ -11,11 +12,12 @@
 
 /* Bitmask for supported acceleration methods */
 
-#define JSIMD_NONE    0x00
-#define JSIMD_MMX     0x01
-#define JSIMD_3DNOW   0x02
-#define JSIMD_SSE     0x04
-#define JSIMD_SSE2    0x08
+#define JSIMD_NONE       0x00
+#define JSIMD_MMX        0x01
+#define JSIMD_3DNOW      0x02
+#define JSIMD_SSE        0x04
+#define JSIMD_SSE2       0x08
+#define JSIMD_ARM_NEON   0x10
 
 /* Short forms of external names for systems with brain-damaged linkers. */
 
@@ -28,6 +30,13 @@
 #define jsimd_extbgrx_ycc_convert_mmx         jSEXTBGRXYCCM
 #define jsimd_extxbgr_ycc_convert_mmx         jSEXTXBGRYCCM
 #define jsimd_extxrgb_ycc_convert_mmx         jSEXTXRGBYCCM
+#define jsimd_rgb_gray_convert_mmx            jSRGBGRYM
+#define jsimd_extrgb_gray_convert_mmx         jSEXTRGBGRYM
+#define jsimd_extrgbx_gray_convert_mmx        jSEXTRGBXGRYM
+#define jsimd_extbgr_gray_convert_mmx         jSEXTBGRGRYM
+#define jsimd_extbgrx_gray_convert_mmx        jSEXTBGRXGRYM
+#define jsimd_extxbgr_gray_convert_mmx        jSEXTXBGRGRYM
+#define jsimd_extxrgb_gray_convert_mmx        jSEXTXRGBGRYM
 #define jsimd_ycc_rgb_convert_mmx             jSYCCRGBM
 #define jsimd_ycc_extrgb_convert_mmx          jSYCCEXTRGBM
 #define jsimd_ycc_extrgbx_convert_mmx         jSYCCEXTRGBXM
@@ -43,6 +52,14 @@
 #define jsimd_extbgrx_ycc_convert_sse2        jSEXTBGRXYCCS2
 #define jsimd_extxbgr_ycc_convert_sse2        jSEXTXBGRYCCS2
 #define jsimd_extxrgb_ycc_convert_sse2        jSEXTXRGBYCCS2
+#define jconst_rgb_gray_convert_sse2          jSCRGBGRYS2
+#define jsimd_rgb_gray_convert_sse2           jSRGBGRYS2
+#define jsimd_extrgb_gray_convert_sse2        jSEXTRGBGRYS2
+#define jsimd_extrgbx_gray_convert_sse2       jSEXTRGBXGRYS2
+#define jsimd_extbgr_gray_convert_sse2        jSEXTBGRGRYS2
+#define jsimd_extbgrx_gray_convert_sse2       jSEXTBGRXGRYS2
+#define jsimd_extxbgr_gray_convert_sse2       jSEXTXBGRGRYS2
+#define jsimd_extxrgb_gray_convert_sse2       jSEXTXRGBGRYS2
 #define jconst_ycc_rgb_convert_sse2           jSCYCCRGBS2
 #define jsimd_ycc_rgb_convert_sse2            jSYCCRGBS2
 #define jsimd_ycc_extrgb_convert_sse2         jSYCCEXTRGBS2
@@ -163,6 +180,35 @@ EXTERN(void) jsimd_extxrgb_ycc_convert_mmx
              JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
              JDIMENSION output_row, int num_rows));
 
+EXTERN(void) jsimd_rgb_gray_convert_mmx
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+EXTERN(void) jsimd_extrgb_gray_convert_mmx
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+EXTERN(void) jsimd_extrgbx_gray_convert_mmx
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+EXTERN(void) jsimd_extbgr_gray_convert_mmx
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+EXTERN(void) jsimd_extbgrx_gray_convert_mmx
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+EXTERN(void) jsimd_extxbgr_gray_convert_mmx
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+EXTERN(void) jsimd_extxrgb_gray_convert_mmx
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+
 EXTERN(void) jsimd_ycc_rgb_convert_mmx
         JPP((JDIMENSION out_width,
              JSAMPIMAGE input_buf, JDIMENSION input_row,
@@ -222,6 +268,36 @@ EXTERN(void) jsimd_extxrgb_ycc_convert_sse2
              JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
              JDIMENSION output_row, int num_rows));
 
+extern const int jconst_rgb_gray_convert_sse2[];
+EXTERN(void) jsimd_rgb_gray_convert_sse2
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+EXTERN(void) jsimd_extrgb_gray_convert_sse2
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+EXTERN(void) jsimd_extrgbx_gray_convert_sse2
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+EXTERN(void) jsimd_extbgr_gray_convert_sse2
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+EXTERN(void) jsimd_extbgrx_gray_convert_sse2
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+EXTERN(void) jsimd_extxbgr_gray_convert_sse2
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+EXTERN(void) jsimd_extxrgb_gray_convert_sse2
+        JPP((JDIMENSION img_width,
+             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+             JDIMENSION output_row, int num_rows));
+
 extern const int jconst_ycc_rgb_convert_sse2[];
 EXTERN(void) jsimd_ycc_rgb_convert_sse2
         JPP((JDIMENSION out_width,
@@ -252,6 +328,35 @@ EXTERN(void) jsimd_ycc_extxrgb_convert_sse2
              JSAMPIMAGE input_buf, JDIMENSION input_row,
              JSAMPARRAY output_buf, int num_rows));
 
+EXTERN(void) jsimd_ycc_rgb_convert_neon
+        JPP((JDIMENSION out_width,
+             JSAMPIMAGE input_buf, JDIMENSION input_row,
+             JSAMPARRAY output_buf, int num_rows));
+EXTERN(void) jsimd_ycc_extrgb_convert_neon
+        JPP((JDIMENSION out_width,
+             JSAMPIMAGE input_buf, JDIMENSION input_row,
+             JSAMPARRAY output_buf, int num_rows));
+EXTERN(void) jsimd_ycc_extrgbx_convert_neon
+        JPP((JDIMENSION out_width,
+             JSAMPIMAGE input_buf, JDIMENSION input_row,
+             JSAMPARRAY output_buf, int num_rows));
+EXTERN(void) jsimd_ycc_extbgr_convert_neon
+        JPP((JDIMENSION out_width,
+             JSAMPIMAGE input_buf, JDIMENSION input_row,
+             JSAMPARRAY output_buf, int num_rows));
+EXTERN(void) jsimd_ycc_extbgrx_convert_neon
+        JPP((JDIMENSION out_width,
+             JSAMPIMAGE input_buf, JDIMENSION input_row,
+             JSAMPARRAY output_buf, int num_rows));
+EXTERN(void) jsimd_ycc_extxbgr_convert_neon
+        JPP((JDIMENSION out_width,
+             JSAMPIMAGE input_buf, JDIMENSION input_row,
+             JSAMPARRAY output_buf, int num_rows));
+EXTERN(void) jsimd_ycc_extxrgb_convert_neon
+        JPP((JDIMENSION out_width,
+             JSAMPIMAGE input_buf, JDIMENSION input_row,
+             JSAMPARRAY output_buf, int num_rows));
+
 /* SIMD Downsample */
 EXTERN(void) jsimd_h2v2_downsample_mmx
         JPP((JDIMENSION image_width, int max_v_samp_factor,
@@ -464,6 +569,15 @@ EXTERN(void) jsimd_idct_4x4_sse2 JPP((void * dct_table,
                                       JSAMPARRAY output_buf,
                                       JDIMENSION output_col));
 
+EXTERN(void) jsimd_idct_2x2_neon JPP((void * dct_table,
+                                      JCOEFPTR coef_block,
+                                      JSAMPARRAY output_buf,
+                                      JDIMENSION output_col));
+EXTERN(void) jsimd_idct_4x4_neon JPP((void * dct_table,
+                                      JCOEFPTR coef_block,
+                                      JSAMPARRAY output_buf,
+                                      JDIMENSION output_col));
+
 /* SIMD Inverse DCT */
 EXTERN(void) jsimd_idct_islow_mmx JPP((void * dct_table,
                                        JCOEFPTR coef_block,
@@ -485,6 +599,11 @@ EXTERN(void) jsimd_idct_ifast_sse2 JPP((void * dct_table,
                                         JSAMPARRAY output_buf,
                                         JDIMENSION output_col));
 
+EXTERN(void) jsimd_idct_ifast_neon JPP((void * dct_table,
+                                        JCOEFPTR coef_block,
+                                        JSAMPARRAY output_buf,
+                                        JDIMENSION output_col));
+
 EXTERN(void) jsimd_idct_float_3dnow JPP((void * dct_table,
                                          JCOEFPTR coef_block,
                                          JSAMPARRAY output_buf,
diff --git a/simd/jsimd_arm.c b/simd/jsimd_arm.c
index a9d920c..9ed49fd 100644
--- a/simd/jsimd_arm.c
+++ b/simd/jsimd_arm.c
@@ -136,6 +136,7 @@ jsimd_can_rgb_ycc (void)
 {
   init_simd();
 
+<<<<<<< HEAD
   /* The code is optimised for these values only */
   if (BITS_IN_JSAMPLE != 8)
     return 0;
@@ -147,6 +148,8 @@ jsimd_can_rgb_ycc (void)
   if (simd_support & JSIMD_ARM_NEON)
     return 1;
 
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
   return 0;
 }
 
@@ -181,6 +184,7 @@ jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
                        JDIMENSION output_row, int num_rows)
 {
+<<<<<<< HEAD
   void (*neonfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
 
   switch(cinfo->in_color_space)
@@ -211,6 +215,8 @@ jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
   if (simd_support & JSIMD_ARM_NEON)
     neonfct(cinfo->image_width, input_buf,
         output_buf, output_row, num_rows);
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
 }
 
 GLOBAL(void)
@@ -386,6 +392,7 @@ jsimd_can_convsamp (void)
 {
   init_simd();
 
+<<<<<<< HEAD
   /* The code is optimised for these values only */
   if (DCTSIZE != 8)
     return 0;
@@ -399,6 +406,8 @@ jsimd_can_convsamp (void)
   if (simd_support & JSIMD_ARM_NEON)
     return 1;
 
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
   return 0;
 }
 
@@ -414,8 +423,11 @@ GLOBAL(void)
 jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
                 DCTELEM * workspace)
 {
+<<<<<<< HEAD
   if (simd_support & JSIMD_ARM_NEON)
     jsimd_convsamp_neon(sample_data, start_col, workspace);
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
 }
 
 GLOBAL(void)
@@ -437,6 +449,7 @@ jsimd_can_fdct_ifast (void)
 {
   init_simd();
 
+<<<<<<< HEAD
   /* The code is optimised for these values only */
   if (DCTSIZE != 8)
     return 0;
@@ -446,6 +459,8 @@ jsimd_can_fdct_ifast (void)
   if (simd_support & JSIMD_ARM_NEON)
     return 1;
 
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
   return 0;
 }
 
@@ -465,8 +480,11 @@ jsimd_fdct_islow (DCTELEM * data)
 GLOBAL(void)
 jsimd_fdct_ifast (DCTELEM * data)
 {
+<<<<<<< HEAD
   if (simd_support & JSIMD_ARM_NEON)
     jsimd_fdct_ifast_neon(data);
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
 }
 
 GLOBAL(void)
@@ -479,6 +497,7 @@ jsimd_can_quantize (void)
 {
   init_simd();
 
+<<<<<<< HEAD
   /* The code is optimised for these values only */
   if (DCTSIZE != 8)
     return 0;
@@ -490,6 +509,8 @@ jsimd_can_quantize (void)
   if (simd_support & JSIMD_ARM_NEON)
     return 1;
 
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
   return 0;
 }
 
@@ -505,8 +526,11 @@ GLOBAL(void)
 jsimd_quantize (JCOEFPTR coef_block, DCTELEM * divisors,
                 DCTELEM * workspace)
 {
+<<<<<<< HEAD
   if (simd_support & JSIMD_ARM_NEON)
     jsimd_quantize_neon(coef_block, divisors, workspace);
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
 }
 
 GLOBAL(void)
@@ -584,6 +608,7 @@ jsimd_can_idct_islow (void)
 {
   init_simd();
 
+<<<<<<< HEAD
   /* The code is optimised for these values only */
   if (DCTSIZE != 8)
     return 0;
@@ -599,6 +624,8 @@ jsimd_can_idct_islow (void)
   if (simd_support & JSIMD_ARM_NEON)
     return 1;
 
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
   return 0;
 }
 
@@ -640,8 +667,11 @@ jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr,
                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
                 JDIMENSION output_col)
 {
+<<<<<<< HEAD
   if ((simd_support & JSIMD_ARM_NEON))
     jsimd_idct_islow_neon(compptr->dct_table, coef_block, output_buf, output_col);
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
 }
 
 GLOBAL(void)
diff --git a/simd/jsimd_arm_neon.S b/simd/jsimd_arm_neon.S
index 9ef6efc..3f1cf9e 100644
--- a/simd/jsimd_arm_neon.S
+++ b/simd/jsimd_arm_neon.S
@@ -62,6 +62,7 @@ _\fname:
     vtrn.32 \x1, \x3
 .endm
 
+<<<<<<< HEAD
 #define CENTERJSAMPLE 128
 
 /*****************************************************************************/
@@ -536,6 +537,8 @@ asm_function jsimd_idct_islow_neon
     .unreq          ROW7R
 .endfunc
 
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
 /*****************************************************************************/
 
 /*
@@ -543,6 +546,7 @@ asm_function jsimd_idct_islow_neon
  *
  * This function contains a fast, not so accurate integer implementation of
  * the inverse DCT (Discrete Cosine Transform). It uses the same calculations
+<<<<<<< HEAD
  * and produces exactly the same output as IJG's original 'jpeg_idct_ifast'
  * function from jidctfst.c
  *
@@ -552,6 +556,12 @@ asm_function jsimd_idct_islow_neon
  * like "x * 1.082392200" have to be converted to "x * 0.082392200 + x",
  * which introduces an extra addition. Overall, there are 6 extra additions
  * per 1-D IDCT pass, totalling to 5 VQDMULH and 35 VADD/VSUB instructions.
+=======
+ * and produces exactly the same output as IJG's original 'jpeg_idct_fast'
+ * function from jidctfst.c
+ *
+ * TODO: a bit better instructions scheduling is needed.
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
  */
 
 #define XFIX_1_082392200 d0[0]
@@ -566,12 +576,70 @@ jsimd_idct_ifast_neon_consts:
     .short (473 * 128 - 256 * 128) /* XFIX_1_847759065 */
     .short (669 * 128 - 512 * 128) /* XFIX_2_613125930 */
 
+<<<<<<< HEAD
+=======
+/* 1-D IDCT helper macro */
+
+.macro idct_helper  x0, x1, x2, x3, x4, x5, x6, x7, \
+                    t10, t11, t12, t13, t14
+
+    vsub.s16        \t10, \x0, \x4
+    vadd.s16        \x4,  \x0, \x4
+    vswp.s16        \t10, \x0
+    vsub.s16        \t11, \x2, \x6
+    vadd.s16        \x6,  \x2, \x6
+    vswp.s16        \t11, \x2
+    vsub.s16        \t10, \x3, \x5
+    vadd.s16        \x5,  \x3, \x5
+    vswp.s16        \t10, \x3
+    vsub.s16        \t11, \x1, \x7
+    vadd.s16        \x7,  \x1, \x7
+    vswp.s16        \t11, \x1
+
+    vqdmulh.s16     \t13, \x2,  d0[1]
+    vadd.s16        \t12, \x3,  \x3
+    vadd.s16        \x2,  \x2,  \t13
+    vqdmulh.s16     \t13, \x3,  d0[3]
+    vsub.s16        \t10,  \x1, \x3
+    vadd.s16        \t12, \t12, \t13
+    vqdmulh.s16     \t13, \t10, d0[2]
+    vsub.s16        \t11, \x7,  \x5
+    vadd.s16        \t10, \t10, \t13
+    vqdmulh.s16     \t13, \t11, d0[1]
+    vadd.s16        \t11, \t11, \t13
+
+    vqdmulh.s16     \t13, \x1,  d0[0]
+    vsub.s16        \x2,  \x6,  \x2
+    vsub.s16        \t14, \x0,  \x2
+    vadd.s16        \x2,  \x0,  \x2
+    vadd.s16        \x0,  \x4,  \x6
+    vsub.s16        \x4,  \x4,  \x6
+    vadd.s16        \x1,  \x1,  \t13
+    vadd.s16        \t13, \x7,  \x5
+    vsub.s16        \t12, \t13, \t12
+    vsub.s16        \t12, \t12, \t10
+    vadd.s16        \t11, \t12, \t11
+    vsub.s16        \t10, \x1,  \t10
+    vadd.s16        \t10, \t10, \t11
+
+    vsub.s16        \x7,  \x0,  \t13
+    vadd.s16        \x0,  \x0,  \t13
+    vadd.s16        \x6,  \t14, \t12
+    vsub.s16        \x1,  \t14, \t12
+    vsub.s16        \x5,  \x2,  \t11
+    vadd.s16        \x2,  \x2,  \t11
+    vsub.s16        \x3,  \x4,  \t10
+    vadd.s16        \x4,  \x4,  \t10
+.endm
+
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
 asm_function jsimd_idct_ifast_neon
 
     DCT_TABLE       .req r0
     COEF_BLOCK      .req r1
     OUTPUT_BUF      .req r2
     OUTPUT_COL      .req r3
+<<<<<<< HEAD
     TMP1            .req r0
     TMP2            .req r1
     TMP3            .req r2
@@ -756,18 +824,117 @@ asm_function jsimd_idct_ifast_neon
       vst1.8          {d21}, [TMP2]
       vst1.8          {d22}, [TMP3]
       vst1.8          {d23}, [TMP4]
+=======
+    TMP             .req ip
+
+    vpush           {d8-d15}
+
+    /* Load constants */
+    adr             TMP, jsimd_idct_ifast_neon_consts
+    vld1.16         {d0}, [TMP, :64]
+
+    /* Load all COEF_BLOCK into NEON registers with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | d4      | d5
+     *   1 | d6      | d7
+     *   2 | d8      | d9
+     *   3 | d10     | d11
+     *   4 | d12     | d13
+     *   5 | d14     | d15
+     *   6 | d16     | d17
+     *   7 | d18     | d19
+     */
+    vld1.16         {d4, d5, d6, d7}, [COEF_BLOCK]!
+    vld1.16         {d8, d9, d10, d11}, [COEF_BLOCK]!
+    vld1.16         {d12, d13, d14, d15}, [COEF_BLOCK]!
+    vld1.16         {d16, d17, d18, d19}, [COEF_BLOCK]!
+    /* Dequantize */
+    vld1.16         {d20, d21, d22, d23}, [DCT_TABLE]!
+    vmul.s16        q2, q2, q10
+    vld1.16         {d24, d25, d26, d27}, [DCT_TABLE]!
+    vmul.s16        q3, q3, q11
+    vmul.s16        q4, q4, q12
+    vld1.16         {d28, d29, d30, d31}, [DCT_TABLE]!
+    vmul.s16        q5, q5, q13
+    vmul.s16        q6, q6, q14
+    vld1.16         {d20, d21, d22, d23}, [DCT_TABLE]!
+    vmul.s16        q7, q7, q15
+    vmul.s16        q8, q8, q10
+    vmul.s16        q9, q9, q11
+
+    /* Pass 1 */
+    idct_helper     q2, q3, q4, q5, q6, q7, q8, q9, q10, q11, q12, q13, q14
+    /* Transpose */
+    transpose_4x4   d4,  d6,  d8,  d10
+    transpose_4x4   d5,  d7,  d9,  d11
+    transpose_4x4   d12, d14, d16, d18
+    transpose_4x4   d13, d15, d17, d19
+    vswp            d12, d5
+    vswp            d14, d7
+    vswp            d16, d9
+    vswp            d18, d11
+
+    /* Pass 2 */
+    idct_helper     q2, q3, q4, q5, q6, q7, q8, q9, q10, q11, q12, q13, q14
+    /* Transpose */
+    transpose_4x4   d4,  d6,  d8,  d10
+    transpose_4x4   d5,  d7,  d9,  d11
+    transpose_4x4   d12, d14, d16, d18
+    transpose_4x4   d13, d15, d17, d19
+    vswp            d12, d5
+    vswp            d14, d7
+    vswp            d16, d9
+    vswp            d18, d11
+
+    /* Descale and range limit */
+    vmov.s16        q15, #(0x80 << 5)
+    vqadd.s16       q2, q2, q15
+    vqadd.s16       q3, q3, q15
+    vqadd.s16       q4, q4, q15
+    vqadd.s16       q5, q5, q15
+    vqadd.s16       q6, q6, q15
+    vqadd.s16       q7, q7, q15
+    vqadd.s16       q8, q8, q15
+    vqadd.s16       q9, q9, q15
+    vqshrun.s16     d4, q2, #5
+    vqshrun.s16     d6, q3, #5
+    vqshrun.s16     d8, q4, #5
+    vqshrun.s16     d10, q5, #5
+    vqshrun.s16     d12, q6, #5
+    vqshrun.s16     d14, q7, #5
+    vqshrun.s16     d16, q8, #5
+    vqshrun.s16     d18, q9, #5
+
+    /* Store results to the output buffer */
+    .irp            x, d4, d6, d8, d10, d12, d14, d16, d18
+    ldr             TMP, [OUTPUT_BUF], #4
+    add             TMP, TMP, OUTPUT_COL
+    vst1.8          {\x}, [TMP]!
+    .endr
+
+    vpop            {d8-d15}
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
     bx              lr
 
     .unreq          DCT_TABLE
     .unreq          COEF_BLOCK
     .unreq          OUTPUT_BUF
     .unreq          OUTPUT_COL
+<<<<<<< HEAD
     .unreq          TMP1
     .unreq          TMP2
     .unreq          TMP3
     .unreq          TMP4
 .endfunc
 
+=======
+    .unreq          TMP
+.endfunc
+
+.purgem idct_helper
+
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
 /*****************************************************************************/
 
 /*
@@ -1152,12 +1319,21 @@ asm_function jsimd_idct_2x2_neon
 
 .macro do_load size
     .if \size == 8
+<<<<<<< HEAD
         vld1.8  {d4}, [U, :64]!
         vld1.8  {d5}, [V, :64]!
         vld1.8  {d0}, [Y, :64]!
         pld     [U, #64]
         pld     [V, #64]
         pld     [Y, #64]
+=======
+        vld1.8  {d4}, [U]!
+        vld1.8  {d5}, [V]!
+        vld1.8  {d0}, [Y]!
+        pld     [Y, #64]
+        pld     [U, #64]
+        pld     [V, #64]
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
     .elseif \size == 4
         vld1.8  {d4[0]}, [U]!
         vld1.8  {d4[1]}, [U]!
@@ -1227,11 +1403,15 @@ asm_function jsimd_idct_2x2_neon
 
 .macro generate_jsimd_ycc_rgb_convert_neon colorid, bpp, r_offs, g_offs, b_offs
 
+<<<<<<< HEAD
 /*
  * 2 stage pipelined YCbCr->RGB conversion
  */
 
 .macro do_yuv_to_rgb_stage1
+=======
+.macro do_yuv_to_rgb
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
     vaddw.u8        q3, q1, d4     /* q3 = u - 128 */
     vaddw.u8        q4, q1, d5     /* q2 = v - 128 */
     vmull.s16       q10, d6, d1[1] /* multiply by -11277 */
@@ -1242,9 +1422,12 @@ asm_function jsimd_idct_2x2_neon
     vmull.s16       q13, d9, d1[0] /* multiply by 22971 */
     vmull.s16       q14, d6, d1[3] /* multiply by 29033 */
     vmull.s16       q15, d7, d1[3] /* multiply by 29033 */
+<<<<<<< HEAD
 .endm
 
 .macro do_yuv_to_rgb_stage2
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
     vrshrn.s32      d20, q10, #15
     vrshrn.s32      d21, q11, #15
     vrshrn.s32      d24, q12, #14
@@ -1259,6 +1442,7 @@ asm_function jsimd_idct_2x2_neon
     vqmovun.s16     d1\b_offs, q14
 .endm
 
+<<<<<<< HEAD
 .macro do_yuv_to_rgb_stage2_store_load_stage1
     vld1.8          {d4}, [U, :64]!
       vrshrn.s32      d20, q10, #15
@@ -1296,6 +1480,8 @@ asm_function jsimd_idct_2x2_neon
     do_yuv_to_rgb_stage2
 .endm
 
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
 /* Apple gas crashes on adrl, work around that by using adr.
  * But this requires a copy of these constants for each function.
  */
@@ -1356,6 +1542,7 @@ asm_function jsimd_ycc_\colorid\()_convert_neon
 
     /* Inner loop over pixels */
     subs            N, N, #8
+<<<<<<< HEAD
     blt             3f
     do_load         8
     do_yuv_to_rgb_stage1
@@ -1371,6 +1558,18 @@ asm_function jsimd_ycc_\colorid\()_convert_neon
     tst             N, #7
     beq             8f
 3:
+=======
+    blt             2f
+1:
+    do_load         8
+    do_yuv_to_rgb
+    do_store        \bpp, 8
+    subs            N, N, #8
+    bge             1b
+    tst             N, #7
+    beq             8f
+2:
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
     tst             N, #4
     beq             3f
     do_load         4
@@ -1418,9 +1617,12 @@ asm_function jsimd_ycc_\colorid\()_convert_neon
 .endfunc
 
 .purgem do_yuv_to_rgb
+<<<<<<< HEAD
 .purgem do_yuv_to_rgb_stage1
 .purgem do_yuv_to_rgb_stage2
 .purgem do_yuv_to_rgb_stage2_store_load_stage1
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
 
 .endm
 
@@ -1436,6 +1638,7 @@ generate_jsimd_ycc_rgb_convert_neon extxrgb, 32, 1, 2, 3
 .purgem do_store
 
 /*****************************************************************************/
+<<<<<<< HEAD
 
 /*
  * jsimd_extrgb_ycc_convert_neon
@@ -2031,3 +2234,5 @@ asm_function jsimd_quantize_neon
     .unreq          SHIFT
     .unreq          LOOP_COUNT
 .endfunc
+=======
+>>>>>>> 4a72099711359606b1fc10c1744057a6c568d5d8
diff --git a/simd/jsimd_i386.c b/simd/jsimd_i386.c
index d9bb774..f77c5ef 100644
--- a/simd/jsimd_i386.c
+++ b/simd/jsimd_i386.c
@@ -2,7 +2,7 @@
  * jsimd_i386.c
  *
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright 2009 D. R. Commander
+ * Copyright 2009-2011 D. R. Commander
  * 
  * Based on the x86 SIMD extension for IJG JPEG library,
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
@@ -84,6 +84,28 @@ jsimd_can_rgb_ycc (void)
 }
 
 GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
 jsimd_can_ycc_rgb (void)
 {
   init_simd();
@@ -155,6 +177,55 @@ jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
 }
 
 GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+  void (*mmxfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space)
+  {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_extrgb_gray_convert_sse2;
+      mmxfct=jsimd_extrgb_gray_convert_mmx;
+      break;
+    case JCS_EXT_RGBX:
+      sse2fct=jsimd_extrgbx_gray_convert_sse2;
+      mmxfct=jsimd_extrgbx_gray_convert_mmx;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_extbgr_gray_convert_sse2;
+      mmxfct=jsimd_extbgr_gray_convert_mmx;
+      break;
+    case JCS_EXT_BGRX:
+      sse2fct=jsimd_extbgrx_gray_convert_sse2;
+      mmxfct=jsimd_extbgrx_gray_convert_mmx;
+      break;
+    case JCS_EXT_XBGR:
+      sse2fct=jsimd_extxbgr_gray_convert_sse2;
+      mmxfct=jsimd_extxbgr_gray_convert_mmx;
+      break;
+    case JCS_EXT_XRGB:
+      sse2fct=jsimd_extxrgb_gray_convert_sse2;
+      mmxfct=jsimd_extxrgb_gray_convert_mmx;
+      break;
+    default:
+      sse2fct=jsimd_rgb_gray_convert_sse2;
+      mmxfct=jsimd_rgb_gray_convert_mmx;
+      break;
+  }
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2))
+    sse2fct(cinfo->image_width, input_buf,
+        output_buf, output_row, num_rows);
+  else if (simd_support & JSIMD_MMX)
+    mmxfct(cinfo->image_width, input_buf,
+        output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
 jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
                        JSAMPIMAGE input_buf, JDIMENSION input_row,
                        JSAMPARRAY output_buf, int num_rows)
diff --git a/simd/jsimd_x86_64.c b/simd/jsimd_x86_64.c
index 7659249..2951268 100644
--- a/simd/jsimd_x86_64.c
+++ b/simd/jsimd_x86_64.c
@@ -2,7 +2,7 @@
  * jsimd_x86_64.c
  *
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright 2009 D. R. Commander
+ * Copyright 2009-2011 D. R. Commander
  * 
  * Based on the x86 SIMD extension for IJG JPEG library,
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
@@ -47,6 +47,23 @@ jsimd_can_rgb_ycc (void)
 }
 
 GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (!IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2))
+    return 0;
+
+  return 1;
+}
+
+GLOBAL(int)
 jsimd_can_ycc_rgb (void)
 {
   /* The code is optimised for these values only */
@@ -99,6 +116,41 @@ jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
 }
 
 GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space)
+  {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_extrgb_gray_convert_sse2;
+      break;
+    case JCS_EXT_RGBX:
+      sse2fct=jsimd_extrgbx_gray_convert_sse2;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_extbgr_gray_convert_sse2;
+      break;
+    case JCS_EXT_BGRX:
+      sse2fct=jsimd_extbgrx_gray_convert_sse2;
+      break;
+    case JCS_EXT_XBGR:
+      sse2fct=jsimd_extxbgr_gray_convert_sse2;
+      break;
+    case JCS_EXT_XRGB:
+      sse2fct=jsimd_extxrgb_gray_convert_sse2;
+      break;
+    default:
+      sse2fct=jsimd_rgb_gray_convert_sse2;
+      break;
+  }
+
+  sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
 jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
                        JSAMPIMAGE input_buf, JDIMENSION input_row,
                        JSAMPARRAY output_buf, int num_rows)
diff --git a/simd/jsimdext.inc b/simd/jsimdext.inc
index c4297f9..635a931 100644
--- a/simd/jsimdext.inc
+++ b/simd/jsimdext.inc
@@ -38,16 +38,26 @@
 
 ; -- segment definition --
 ;
+%ifdef __YASM_VER__
+%define SEG_TEXT    .text  align=16
+%define SEG_CONST   .rdata align=16
+%else
 %define SEG_TEXT    .text  align=16 public use32 class=CODE
 %define SEG_CONST   .rdata align=16 public use32 class=CONST
+%endif
 
 %elifdef WIN64	; ----(nasm -fwin64 -DWIN64 ...)--------
 ; * Microsoft Visual C++
 
 ; -- segment definition --
 ;
+%ifdef __YASM_VER__
+%define SEG_TEXT    .text  align=16
+%define SEG_CONST   .rdata align=16
+%else
 %define SEG_TEXT    .text  align=16 public use64 class=CODE
 %define SEG_CONST   .rdata align=16 public use64 class=CONST
+%endif
 %define EXTN(name)  name			; foo() -> foo
 
 %elifdef OBJ32	; ----(nasm -fobj -DOBJ32 ...)----------
@@ -297,8 +307,6 @@ const_base:
 %ifdef WIN64
 
 %imacro collect_args 0
-	push r10
-	push r11
 	push r12
 	push r13
 	push r14
@@ -328,8 +336,6 @@ const_base:
 	pop r14
 	pop r13
 	pop r12
-	pop r11
-	pop r10
 %endmacro
 
 %else