mirrors/gstreamer
mirrors/gstreamer
1
1
Fork 0
mirror of https://gitlab.freedesktop.org/gstreamer/gstreamer.git synced 2024-11-27 04:01:08 +00:00
Code Issues 1 Projects Releases Wiki Activity

cairooverlay: Optimize premultiplication/unpremultiplication loops

Browse source 
Pull in video frame fields into local variables. Without this the
compiler must assume that they could've changed on every use and read
them from memory again.

This reduces the inner loop from 6 memory reads per pixels to 4, and the
number of writes stays at 3.
This commit is contained in:
Sebastian Dröge 2018-12-07 19:09:30 +02:00 committed by Sebastian Dröge
parent c50be8f146
commit 172d14acef
1 changed files with 36 additions and 16 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

52
ext/cairo/gstcairooverlay.c
View file

@ -230,12 +230,17 @@ static void
gst_video_overlay_rectangle_premultiply_0 (GstVideoFrame * frame)
{
int i, j;
for (j = 0; j < GST_VIDEO_FRAME_HEIGHT (frame); ++j) {
int width = GST_VIDEO_FRAME_WIDTH (frame);
int height = GST_VIDEO_FRAME_HEIGHT (frame);
int stride = GST_VIDEO_FRAME_PLANE_STRIDE (frame, 0);
guint8 *data = GST_VIDEO_FRAME_PLANE_DATA (frame, 0);
for (j = 0; j < height; ++j) {
guint8 *line;
line = GST_VIDEO_FRAME_PLANE_DATA (frame, 0);
line += GST_VIDEO_FRAME_PLANE_STRIDE (frame, 0) * j;
for (i = 0; i < GST_VIDEO_FRAME_WIDTH (frame); ++i) {
line = data;
line += stride * j;
for (i = 0; i < width; ++i) {
int a = line[0];
line[1] = line[1] * a / 255;
line[2] = line[2] * a / 255;
@ -250,12 +255,17 @@ static void
gst_video_overlay_rectangle_premultiply_3 (GstVideoFrame * frame)
{
int i, j;
for (j = 0; j < GST_VIDEO_FRAME_HEIGHT (frame); ++j) {
int width = GST_VIDEO_FRAME_WIDTH (frame);
int height = GST_VIDEO_FRAME_HEIGHT (frame);
int stride = GST_VIDEO_FRAME_PLANE_STRIDE (frame, 0);
guint8 *data = GST_VIDEO_FRAME_PLANE_DATA (frame, 0);
for (j = 0; j < height; ++j) {
guint8 *line;
line = GST_VIDEO_FRAME_PLANE_DATA (frame, 0);
line += GST_VIDEO_FRAME_PLANE_STRIDE (frame, 0) * j;
for (i = 0; i < GST_VIDEO_FRAME_WIDTH (frame); ++i) {
line = data;
line += stride * j;
for (i = 0; i < width; ++i) {
int a = line[3];
line[0] = line[0] * a / 255;
line[1] = line[1] * a / 255;
@ -290,12 +300,17 @@ static void
gst_video_overlay_rectangle_unpremultiply_0 (GstVideoFrame * frame)
{
int i, j;
for (j = 0; j < GST_VIDEO_FRAME_HEIGHT (frame); ++j) {
int width = GST_VIDEO_FRAME_WIDTH (frame);
int height = GST_VIDEO_FRAME_HEIGHT (frame);
int stride = GST_VIDEO_FRAME_PLANE_STRIDE (frame, 0);
guint8 *data = GST_VIDEO_FRAME_PLANE_DATA (frame, 0);
for (j = 0; j < height; ++j) {
guint8 *line;
line = GST_VIDEO_FRAME_PLANE_DATA (frame, 0);
line += GST_VIDEO_FRAME_PLANE_STRIDE (frame, 0) * j;
for (i = 0; i < GST_VIDEO_FRAME_WIDTH (frame); ++i) {
line = data;
line += stride * j;
for (i = 0; i < width; ++i) {
int a = line[0];
if (a) {
line[1] = MIN ((line[1] * 255 + a / 2) / a, 255);
@ -312,12 +327,17 @@ static void
gst_video_overlay_rectangle_unpremultiply_3 (GstVideoFrame * frame)
{
int i, j;
for (j = 0; j < GST_VIDEO_FRAME_HEIGHT (frame); ++j) {
int width = GST_VIDEO_FRAME_WIDTH (frame);
int height = GST_VIDEO_FRAME_HEIGHT (frame);
int stride = GST_VIDEO_FRAME_PLANE_STRIDE (frame, 0);
guint8 *data = GST_VIDEO_FRAME_PLANE_DATA (frame, 0);
for (j = 0; j < height; ++j) {
guint8 *line;
line = GST_VIDEO_FRAME_PLANE_DATA (frame, 0);
line += GST_VIDEO_FRAME_PLANE_STRIDE (frame, 0) * j;
for (i = 0; i < GST_VIDEO_FRAME_WIDTH (frame); ++i) {
line = data;
line += stride * j;
for (i = 0; i < width; ++i) {
int a = line[3];
if (a) {
line[0] = MIN ((line[0] * 255 + a / 2) / a, 255);