cosmetics

avisynth/src/edge_detect.cpp

@@ -38,21 +38,21 @@ static inline void line_copy(float* dstp, const float* srcp, int width)
 }
 
 
-static inline __m128 mm_abs_ps(const __m128& val)
+static __forceinline __m128 mm_abs_ps(const __m128& val)
 {
     const __m128 mask = _mm_castsi128_ps(_mm_set1_epi32(0x7FFFFFFF));
     return _mm_and_ps(val, mask);
 }
 
 
-static inline __m128 mm_ivtsign_ps(const __m128& val)
+static __forceinline __m128 mm_ivtsign_ps(const __m128& val)
 {
     const __m128 mask = _mm_castsi128_ps(_mm_set1_epi32(0x80000000));
     return _mm_xor_ps(val, mask);
 }
 
 
-static inline __m128 mm_rcp_hq_ps(const __m128& xmm0)
+static __forceinline __m128 mm_rcp_hq_ps(const __m128& xmm0)
 {
     __m128 rcp = _mm_rcp_ps(xmm0);
     __m128 xmm1 =  _mm_mul_ps(_mm_mul_ps(xmm0, rcp), rcp);
@@ -61,7 +61,7 @@ static inline __m128 mm_rcp_hq_ps(const __m128& xmm0)
 }
 
 
-static inline __m128i mm_calc_dir(const __m128& gx, const __m128& gy)
+static __forceinline __m128i mm_calc_dir(const __m128& gx, const __m128& gy)
 {
     static const float t0225 = (float)(sqrt(2.0) - 1.0); // tan(pi/8)
     static const float t0675 = (float)(sqrt(2.0) + 1.0); // tan(3*pi/8)
@@ -132,13 +132,13 @@ void __stdcall TCannyM::standerd_operator(int width, int height)
 
                 gx = _mm_add_ps(_mm_mul_ps(gx, gx), _mm_mul_ps(gy, gy));
                 gx = _mm_sqrt_ps(gx);
-                _mm_store_ps(dstp + x + i * 4, gx);
+                _mm_stream_ps(dstp + x + i * 4, gx);
             }
 
             d[0] = _mm_packs_epi32(d[0], d[1]);
             d[1] = _mm_packs_epi32(d[2], d[3]);
             d[0] = _mm_packs_epi16(d[0], d[1]);
-            _mm_store_si128((__m128i*)(dir + x), d[0]);
+            _mm_stream_si128((__m128i*)(dir + x), d[0]);
         }
         dstp += frame_pitch;
         dir += frame_pitch;
@@ -191,13 +191,13 @@ void __stdcall TCannyM::sobel_operator(int width, int height)
 
                 gx = _mm_add_ps(_mm_mul_ps(gx, gx), _mm_mul_ps(gy, gy));
                 gx = _mm_sqrt_ps(gx);
-                _mm_store_ps(dstp + x + i * 4, gx);
+                _mm_stream_ps(dstp + x + i * 4, gx);
             }
 
             d[0] = _mm_packs_epi32(d[0], d[1]);
             d[1] = _mm_packs_epi32(d[2], d[3]);
             d[0] = _mm_packs_epi16(d[0], d[1]);
-            _mm_store_si128((__m128i*)(dir + x), d[0]);
+            _mm_stream_si128((__m128i*)(dir + x), d[0]);
         }
         dstp += frame_pitch;
         dir += frame_pitch;

avisynth/src/gaussian_blur.cpp

@@ -39,10 +39,10 @@ convert_to_float(int width, int height, const uint8_t* srcp, float* dstp,
             __m128i xmm1 = _mm_unpackhi_epi8(xmm0, zero);
             xmm0 = _mm_unpacklo_epi8(xmm0, zero);
 
-            _mm_store_ps(dstp + x, _mm_cvtepi32_ps(_mm_unpacklo_epi16(xmm0, zero)));
-            _mm_store_ps(dstp + x + 4, _mm_cvtepi32_ps(_mm_unpackhi_epi16(xmm0, zero)));
-            _mm_store_ps(dstp + x + 8, _mm_cvtepi32_ps(_mm_unpacklo_epi16(xmm1, zero)));
-            _mm_store_ps(dstp + x + 12, _mm_cvtepi32_ps(_mm_unpackhi_epi16(xmm1, zero)));
+            _mm_stream_ps(dstp + x, _mm_cvtepi32_ps(_mm_unpacklo_epi16(xmm0, zero)));
+            _mm_stream_ps(dstp + x + 4, _mm_cvtepi32_ps(_mm_unpackhi_epi16(xmm0, zero)));
+            _mm_stream_ps(dstp + x + 8, _mm_cvtepi32_ps(_mm_unpacklo_epi16(xmm1, zero)));
+            _mm_stream_ps(dstp + x + 12, _mm_cvtepi32_ps(_mm_unpackhi_epi16(xmm1, zero)));
         }
         srcp += src_pitch;
         dstp += dst_pitch;
@@ -74,7 +74,7 @@ horizontal_blur(float* srcp, const int radius, const int length, int width,
             __m128 xmm0 = _mm_loadu_ps(srcp + x + i);
             sum = _mm_add_ps(sum, _mm_mul_ps(xmm0, k));
         }
-        _mm_store_ps(dstp + x, sum);
+        _mm_stream_ps(dstp + x, sum);
     }
 }
 

avisynth/src/tcannymod.hpp

@@ -32,7 +32,7 @@
 #include "write_frame.h"
 
 
-#define TCANNY_M_VERSION "0.1.1"
+#define TCANNY_M_VERSION "0.2.0"
 
 #define GB_MAX_LENGTH 17
 

avisynth/src/write_frame.cpp

@@ -82,7 +82,7 @@ write_edge_direction(const float* edgep, const uint8_t* dir, float th_max,
             xmm0 = _mm_packs_epi16(_mm_packs_epi32(xmm0, xmm1), _mm_packs_epi32(xmm2, xmm3));
             xmm1 = _mm_load_si128((__m128i*)(dir + x));
             xmm0 = _mm_and_si128(xmm0, xmm1);
-            _mm_store_si128((__m128i*)(dstp + x), xmm0);
+            _mm_stream_si128((__m128i*)(dstp + x), xmm0);
         }
         edgep += frame_pitch;
         dir += frame_pitch;
@@ -104,7 +104,7 @@ write_binary_mask(const float* srcp, float th_max, int width, int height,
             __m128i xmm2 = _mm_castps_si128(_mm_cmpge_ps(_mm_load_ps(srcp + x + 8), tmax));
             __m128i xmm3 = _mm_castps_si128(_mm_cmpge_ps(_mm_load_ps(srcp + x + 12), tmax));
             xmm0 = _mm_packs_epi16(_mm_packs_epi32(xmm0, xmm1), _mm_packs_epi32(xmm2, xmm3));
-            _mm_store_si128((__m128i*)(dstp + x), xmm0);
+            _mm_stream_si128((__m128i*)(dstp + x), xmm0);
         }
         srcp += src_pitch;
         dstp += dst_pitch;