no SSE2 toggle on plugin

allow to compile withou SSE2

4 files changed, 18 insertions, 5 deletions

diff --git a/plugins/fast_float/include/lcms2_fast_float.h b/plugins/fast_float/include/lcms2_fast_float.h
index 1dbf134..4b96800 100644
--- a/plugins/fast_float/include/lcms2_fast_float.h
+++ b/plugins/fast_float/include/lcms2_fast_float.h
@@ -32,6 +32,12 @@ extern "C" {
 
 #define LCMS2_FAST_FLOAT_VERSION   1200
 
+// Configuration toggles
+
+// Uncomment this if yout target platform can deal with SSE2
+// #define CMS_DONT_USE_SSE2 1
+
+
 // The one and only plug-in entry point. To install this plugin in your code
 // you need to place this in some initialization place:
 //
diff --git a/plugins/fast_float/src/fast_16_tethra.c b/plugins/fast_float/src/fast_16_tethra.c
index 6808fd3..9252d93 100644
--- a/plugins/fast_float/src/fast_16_tethra.c
+++ b/plugins/fast_float/src/fast_16_tethra.c
@@ -334,6 +334,9 @@ cmsBool Optimize16BitRGBTransform(_cmsTransformFn* TransformFn,
     // Only real 16 bits
     if (T_BIT15(*InputFormat) != 0 || T_BIT15(*OutputFormat) != 0) return FALSE;
 
+	// Swap endian is not supported
+    if (T_ENDIAN16(*InputFormat) != 0 || T_ENDIAN16(*OutputFormat) != 0) return FALSE;
+
     // Only on input RGB
     if (T_COLORSPACE(*InputFormat)  != PT_RGB) return FALSE;
     
diff --git a/plugins/fast_float/src/fast_8_matsh_sse.c b/plugins/fast_float/src/fast_8_matsh_sse.c
index 233ffce..ddffd0d 100644
--- a/plugins/fast_float/src/fast_8_matsh_sse.c
+++ b/plugins/fast_float/src/fast_8_matsh_sse.c
@@ -24,6 +24,8 @@
 #include "fast_float_internal.h"
 
 
+#ifndef CMS_DONT_USE_SSE2
+
 #ifdef _MSC_VER
 #include <intrin.h>
 #else
@@ -229,7 +231,7 @@ void MatShaperXform8SSE(struct _cmstransform_struct *CMMcargo,
                   aout = (cmsUInt8Number*)Output + DestStartingOrder[3] + strideOut;
            
            /**
-           * Prefectch
+           * Prefetch
            */
            __m128 rvector = _mm_set1_ps(p->Shaper1R[*rin]);
            __m128 gvector = _mm_set1_ps(p->Shaper1G[*gin]);
@@ -336,7 +338,7 @@ cmsBool Optimize8MatrixShaperSSE(_cmsTransformFn* TransformFn,
     // Check for SSE2 support
     if (!(IsSSE2Available())) return FALSE;
 
-    // Only works on RGB to RGB and gray to gray 
+    // Only works on 3 to 3, probably RGB 
     if ( !( (T_CHANNELS(*InputFormat) == 3 && T_CHANNELS(*OutputFormat) == 3) ) ) return FALSE;
                    
     // Only works on 8 bit input
@@ -396,8 +398,8 @@ cmsBool Optimize8MatrixShaperSSE(_cmsTransformFn* TransformFn,
         _cmsStageToneCurvesData* mpeC1 = (_cmsStageToneCurvesData*) cmsStageData(Curve1);
         _cmsStageToneCurvesData* mpeC2 = (_cmsStageToneCurvesData*) cmsStageData(Curve2);
                 
-        // In this particular optimization, caché does not help as it takes more time to deal with 
-        // the caché that with the pixel handling
+        // In this particular optimization, cache does not help as it takes more time to deal with 
+        // the cache that with the pixel handling
         *dwFlags |= cmsFLAGS_NOCACHE;
   
 
@@ -414,4 +416,5 @@ cmsBool Optimize8MatrixShaperSSE(_cmsTransformFn* TransformFn,
     return TRUE;
 }
 
+#endif
 
diff --git a/plugins/fast_float/src/fast_float_sup.c b/plugins/fast_float/src/fast_float_sup.c
index 4b25705..e4bf794 100644
--- a/plugins/fast_float/src/fast_float_sup.c
+++ b/plugins/fast_float/src/fast_float_sup.c
@@ -40,9 +40,10 @@ cmsBool Floating_Point_Transforms_Dispatcher(_cmsTransformFn* TransformFn,
        // Try to optimize by joining curves
        if (Optimize8ByJoiningCurves(TransformFn, UserData, FreeUserData, Lut, InputFormat, OutputFormat, dwFlags)) return TRUE;
 
+#ifndef CMS_DONT_USE_SSE2
        // Try to use SSE2 to optimize as a set of curves plus a matrix plus a set of curves
        if (Optimize8MatrixShaperSSE(TransformFn, UserData, FreeUserData, Lut, InputFormat, OutputFormat, dwFlags)) return TRUE;
-
+#endif
        // Try to optimize as a set of curves plus a matrix plus a set of curves
        if (Optimize8MatrixShaper(TransformFn, UserData, FreeUserData, Lut, InputFormat, OutputFormat, dwFlags)) return TRUE;