[Spice-devel] [PATCH 4/6] LZ4: Fix the row alignment when it is not on a 32bit boundary
Javier Celaya
javier.celaya at flexvm.es
Mon Feb 2 01:11:32 PST 2015
Hello
El Viernes, 30 de enero de 2015 11:36:08 Christophe Fergeau escribió:
> On Wed, Jan 28, 2015 at 11:49:47AM +0100, Javier Celaya wrote:
> > Fix the row alignment for 16/24 bpp images when it is not in a 32bit
> > boundary. This is needed for 16bpp images when the width is an odd
> > number, and for the future support of 24bpp images.
> > ---
> > @@ -620,6 +623,21 @@ static pixman_image_t *canvas_get_lz4(CanvasBase
> > *canvas, SpiceImage *image, int>
> > data += enc_size;
> >
> > } while (data < data_end);
> >
> > + if (stride_abs > stride_encoded) {
> > + // Fix the row alignment
> > + int row;
> > + uint32_t *dest_aligned, *dest_misaligned;
> > + dest = (uint8_t *)pixman_image_get_data(surface);
> > + if (!top_down) {
> > + dest -= (stride_abs * (height - 1));
> > + }
> > + for (row = height - 1; row > 0; --row) {
> > + dest_aligned = (uint32_t *)(dest + stride_abs*row);
> > + dest_misaligned = (uint32_t*)(dest + stride_encoded*row);
> > + memmove(dest_aligned, dest_misaligned, stride_encoded);
> > + }
> > + }
> > +
>
> Looking at common/canvas_utils.c:surface_create, we don't seem to align
> rows when top_down is TRUE, it seems this code will need to be more
> careful in this case :(
>
> Christophe
IMHO, rows should be aligned whether the image is top-down or bottom-up. In
common/canvas_utils.c:surface_create, pixman_image_create_bits is called with
an aligned stride for bottom-up images. Pixman checks it in pixman-bits-
image.c, line 996, in function create_bits_image_internal, which is called by
pixman_image_create_bits:
http://cgit.freedesktop.org/pixman/tree/pixman/pixman-bits-image.c#n996
But for top-down images, pixman_image_create_bits is called with a null buffer
and a zero stride. Then, pixman allocates the buffer and computes an aligned
stride by itself, in function create_bits:
http://cgit.freedesktop.org/pixman/tree/pixman/pixman-bits-image.c#n898
More information about the Spice-devel
mailing list