On Fri, Sep 09, 2016 at 09:54:17AM -0400, Frediano Ziglio wrote:
> > static int encode_frame(MJpegEncoder *encoder, const SpiceRect *src,
> > const SpiceBitmap *image, int top_down)
> > {
> > - SpiceChunks *chunks;
> > + RedChunkIterator it;
> > uint32_t image_stride;
> > - size_t offset;
> > - int i, chunk;
> > + unsigned int i;
> >
> > - chunks = image->data;
> > - offset = 0;
> > - chunk = 0;
> > image_stride = image->stride;
> > + red_chunk_iterator_init(&it, image->data);
> >
> > const int skip_lines = top_down ? src->top : image->y - (src->bottom -
> > 0);
> > for (i = 0; i < skip_lines; i++) {
> > - get_image_line(chunks, &offset, &chunk, image_stride);
> > + red_chunk_iterator_skip_bytes(&it, image_stride);
> > }
> >
> > const unsigned int stream_height = src->bottom - src->top;
> > const unsigned int stream_width = src->right - src->left;
> >
> > for (i = 0; i < stream_height; i++) {
> > - uint8_t *src_line = get_image_line(chunks, &offset, &chunk,
> > image_stride);
> > + uint8_t line_data[image_stride];
>
> Potentially 4gb on the stack.
> Beside this I think previously there was no allocation of a line
> and no memory copy.
> By definition a line have to be contained in a single chunk.
Avoiding that means exporting something similar to
red_chunk_iterator_get_data_one_chunk(). With RedChunkIterator, we
do not really care whether one line fits in one chunk or not. You say
"by definition", is there anything guaranteeing that? I agree the stack
usage, and then the memcpy should be fixed.
Christophe
Attachment:
signature.asc
Description: PGP signature
_______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/spice-devel