r300: Improve texture layout calculations

The texture layout calculations for mipmapped cubemaps used to be completely
wrong, since the GPU expects images to be grouped by miplevel instead of by
face number.

This has been fixed now, though the memory layout is still slightly incorrect
for the smaller miplevels. Unfortunately, the docs are lacking in that area.

1 files changed, 1 insertions, 1 deletions

diff --git a/src/mesa/drivers/dri/r300/r300_texmem.c b/src/mesa/drivers/dri/r300/r300_texmem.c
index e0fda86ac8b..69847a4022d 100644
--- a/src/mesa/drivers/dri/r300/r300_texmem.c
+++ b/src/mesa/drivers/dri/r300/r300_texmem.c
@@ -349,7 +349,7 @@ static void r300UploadSubImage(r300ContextPtr rmesa, r300TexObjPtr t,
 	imageWidth = texImage->Width;
 	imageHeight = texImage->Height;
 
-	offset = t->bufAddr + t->base.totalSize / 6 * face;
+	offset = t->bufAddr;
 
 	if (RADEON_DEBUG & (DEBUG_TEXTURE | DEBUG_IOCTL)) {
 		GLint imageX = 0;