1 files changed, 17 insertions, 20 deletions
diff --git a/Source/Core/VideoBackends/OGL/TextureConverter.cpp b/Source/Core/VideoBackends/OGL/TextureConverter.cpp
index 8ea8cd5d8b..d1d4ae5477 100644
--- a/Source/Core/VideoBackends/OGL/TextureConverter.cpp
+++ b/Source/Core/VideoBackends/OGL/TextureConverter.cpp
@@ -242,34 +242,31 @@ static void EncodeToRamUsingShader(GLuint srcTexture, u8* destAddr, u32 dst_line
 
   int dstSize = dst_line_size * dstHeight;
 
-  if ((writeStride != dst_line_size) && (dstHeight > 1))
+  // When the dst_line_size and writeStride are the same, we could use glReadPixels directly to RAM.
+  // But instead we always copy the data via a PBO, because macOS inexplicably prefers this (most
+  // noticeably in the Super Mario Sunshine transition).
+  glBindBuffer(GL_PIXEL_PACK_BUFFER, s_PBO);
+  glBufferData(GL_PIXEL_PACK_BUFFER, dstSize, nullptr, GL_STREAM_READ);
+  glReadPixels(0, 0, (GLsizei)(dst_line_size / 4), (GLsizei)dstHeight, GL_BGRA, GL_UNSIGNED_BYTE,
+               nullptr);
+  u8* pbo = (u8*)glMapBufferRange(GL_PIXEL_PACK_BUFFER, 0, dstSize, GL_MAP_READ_BIT);
+
+  if (dst_line_size == writeStride)
+  {
+    memcpy(destAddr, pbo, dst_line_size * dstHeight);
+  }
+  else
   {
-    // writing to a texture of a different size
-    // also copy more then one block line, so the different strides matters
-    // copy into one pbo first, map this buffer, and then memcpy into GC memory
-    // in this way, we only have one vram->ram transfer, but maybe a bigger
-    // CPU overhead because of the pbo
-    glBindBuffer(GL_PIXEL_PACK_BUFFER, s_PBO);
-    glBufferData(GL_PIXEL_PACK_BUFFER, dstSize, nullptr, GL_STREAM_READ);
-    glReadPixels(0, 0, (GLsizei)(dst_line_size / 4), (GLsizei)dstHeight, GL_BGRA, GL_UNSIGNED_BYTE,
-                 nullptr);
-    u8* pbo = (u8*)glMapBufferRange(GL_PIXEL_PACK_BUFFER, 0, dstSize, GL_MAP_READ_BIT);
-
     for (size_t i = 0; i < dstHeight; ++i)
     {
       memcpy(destAddr, pbo, dst_line_size);
       pbo += dst_line_size;
       destAddr += writeStride;
     }
-
-    glUnmapBuffer(GL_PIXEL_PACK_BUFFER);
-    glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
-  }
-  else
-  {
-    glReadPixels(0, 0, (GLsizei)(dst_line_size / 4), (GLsizei)dstHeight, GL_BGRA, GL_UNSIGNED_BYTE,
-                 destAddr);
   }
+
+  glUnmapBuffer(GL_PIXEL_PACK_BUFFER);
+  glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
 }
 
 void EncodeToRamFromTexture(u8* dest_ptr, u32 format, u32 native_width, u32 bytes_per_row,