spline系リサイズを使用したときのmisalign accessの可能性があったのを修正。( #553 )

rigaya · Feb 7, 2024 · f2df8d3 · f2df8d3
1 parent 9da82d7
commit f2df8d3
Show file tree

Hide file tree

Showing 2 changed files with 9 additions and 8 deletions.
diff --git a/NVEncCore/NVEncFilterResize.cu b/NVEncCore/NVEncFilterResize.cu
@@ -302,13 +302,14 @@ __inline__ __device__
 float factor_spline(const float x_raw, const float *psCopyFactor) {
     const float x = fabs(x_raw);
     if (x >= (float)radius) return 0.0f;
-    const float4 weight = ((const float4 *)psCopyFactor)[min((int)x, radius - 1)];
+    const int pos = min((int)x, radius - 1);
+    const float *psWeight = psCopyFactor + min((int)x, radius - 1) * 4;
     //重みを計算
-    float w = weight.w;
-    w += x * weight.z;
+    float w = psWeight[3];
+    w += x * psWeight[2];
     const float x2 = x * x;
-    w += x2 * weight.y;
-    w += x2 * x * weight.x;
+    w += x2 * psWeight[1];
+    w += x2 * x * psWeight[0];
     return w;
 }
 

diff --git a/NVEncCore/rgy_version.h b/NVEncCore/rgy_version.h
@@ -29,9 +29,9 @@
 #ifndef __RGY_CONFIG_H__
 #define __RGY_CONFIG_H__
 
-#define VER_FILEVERSION              0,7,42,0
-#define VER_STR_FILEVERSION          "7.42beta1"
-#define VER_STR_FILEVERSION_TCHAR _T("7.42beta1")
+#define VER_FILEVERSION              0,7,42,1
+#define VER_STR_FILEVERSION          "7.42beta2"
+#define VER_STR_FILEVERSION_TCHAR _T("7.42beta2")
 
 #ifdef _M_IX86
 #define BUILD_ARCH_STR _T("x86")