diff --git a/TODO b/TODO index 31b87df5..918a4e5d 100644 --- a/TODO +++ b/TODO @@ -1,3 +1,38 @@ +- how odd, why is resize much faster? + + $ time vips resize wtc.tif small.tif 0.25 + memory: high-water mark 138.09 MB + real 0m0.430s + user 0m2.700s + sys 0m0.172s + $ time vipsthumbnail wtc.tif -o small.tif -s 2500 + memory: high-water mark 210.05 MB + real 0m0.959s + user 0m9.984s + sys 0m0.184s + + about the same speed for jpg + + $ time vips resize big.jpg[shrink=8] small.jpg 0.6666 + memory: high-water mark 47.21 MB + real 0m1.320s + user 0m2.216s + sys 0m0.060s + $ time vipsthumbnail big.jpg -s 2500 -o small.jpg + memory: high-water mark 58.65 MB + real 0m1.360s + user 0m2.392s + sys 0m0.100s + + seems to be because vipsthumbnail is not int shrinking enough: + + $ vipsthumbnail wtc.tif -o small.tif -s 2500 --vips-info + info: vipsthumbnail: integer shrink by 1 + info: vipsthumbnail: residual scale by 0.266752 + + should int shrink by 2! vips resize is doing this right, what's the + difference? + - are the mosaic functions calling vips_fastcor()? it must be very slow diff --git a/doc/libvips-docs.xml b/doc/libvips-docs.xml index 8732aa0d..4b6ebb80 100644 --- a/doc/libvips-docs.xml +++ b/doc/libvips-docs.xml @@ -9,7 +9,7 @@ VIPS Reference Manual - For VIPS 8.0.2. + For VIPS 8.1.0. The latest version of this documentation can be found on the VIPS website.