/* @(#) Do a complicated compound operation for benchmarking the threading * @(#) system. Input should be a large LABQ image, output is a large sRGB * @(#) image. * @(#) * @(#) Usage: * @(#) * @(#) int im_benchmark( IMAGE *in, IMAGE *out ) * @(#) * @(#) Returns 0 on sucess and -1 on error. * @(#) * * 6/10/06 * - hacked in * 27/11/06 * - added im_benchmarkn() */ /* This file is part of VIPS. VIPS is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. You should have received a copy of the GNU Lesser General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ /* These files are distributed with VIPS - http://www.vips.ecs.soton.ac.uk */ #ifdef HAVE_CONFIG_H #include #endif /*HAVE_CONFIG_H*/ #include #include #include #include #include #ifdef WITH_DMALLOC #include #endif /*WITH_DMALLOC*/ /* VIPS SMP benchmark ------------------ This is adapted from the system used to generate images for POD: http://cima.ng-london.org.uk/~john/POD Images from a 10k by 10k studio digital camera are colour processed, resized, cropped and sharpened. The original POD script was written in nip (see below). This operation is a reimplementation in vanilla C to make it easier to run (and less fragile!). This thing was originally processing images off a remote server over a 100mbit network. No attempt was made to make it quick (there was no point): you could make it a lot faster very easily. ------ benchmark in nip2 ----------- #!/home/john/vips/bin/nip2 -s // get command-line arguments image_path = argv?1; crop_id = parse_pint argv?2; crop_left = parse_pint argv?3; crop_top = parse_pint argv?4; crop_width = parse_pint argv?5; crop_height = parse_pint argv?6; width = parse_pint argv?7; height = parse_pint argv?8; sharp = parse_pint argv?9; // scale down by this much to undo photographic's relativisation darken = Vector [1.18, 1, 1]; // fudge factor in XYZ to get a match under NGC lights on uv-durable paper white_point_adjust = Vector [1.06, 1, 1.01]; // brighten by this in XYZ to get relative colorimetry brighten = 1.5; // blacks down by this much in LAB blacks_down = Vector [-2, 0, 0]; // sharpen params for 400, 300, 200 and 150 dpi // just change the size of the area we search sharpen_params_table = [ [ 11, 2.5, 40, 20, 0.5, 1.5 ], [ 7, 2.5, 40, 20, 0.5, 1.5 ], [ 5, 2.5, 40, 20, 0.5, 1.5 ], [ 3, 2.5, 40, 20, 0.5, 1.5 ] ]; // convert D65 XYZ to D50 XYZ D652D50 = recomb D652D50_direct; stage_crop in = extract_area crop_left crop_top crop_width crop_height in, crop_id != 0 = in; // fit within a width / height stage_shrink image = image, factor > 1; // never upscale = resize factor factor Interpolate.BILINEAR image { hfactor = width / get_width image; vfactor = height / get_height image; factor = min_pair hfactor vfactor; } // unphotoize, go to xyz, convert to D50, adjust white point, back to lab stage_colour in = if in?0 > 99 then Vector [100, 0, 0] else in''' { // back to absolute in' = in / darken; xyz = colour_transform_to Image_type.XYZ in'; xyz' = D652D50 xyz * white_point_adjust * brighten; in'' = colour_transform_to Image_type.LAB xyz'; // shadows down in''' = in'' + blacks_down; } stage_sharp in = (sharpen params?0 params?1 params?2 params?3 params?4 params?5 @ colour_transform_to Image_type.LABQ) in { params = sharpen_params_table?sharp; } // This was: // // stage_srgb in // = (icc_export 8 "$VIPSHOME/share/nip2/data/sRGB.icm" 1 @ // colour_transform_to Image_type.LABQ) in; // // but that uses lcms which is single-threaded. So for this benchmark, we use // VIPS's own ->sRGB converter, which is less accurate but does thread. stage_srgb in = colour_transform_to Image_type.sRGB in; main = (get_image @ stage_srgb @ stage_sharp @ stage_colour @ stage_shrink @ stage_crop @ colour_transform_to Image_type.LAB @ Image_file) image_path; ------ benchmark in nip2 ----------- */ /* The main part of the benchmark ... transform labq to labq. Chain several of * these together to get a CPU-bound operation. */ static int benchmark( IMAGE *in, IMAGE *out ) { IMAGE *t[18]; double one[3] = { 1.0, 1.0, 1.0 }; double zero[3] = { 0.0, 0.0, 0.0 }; double darken[3] = { 1.0 / 1.18, 1.0, 1.0 }; double whitepoint[3] = { 1.06, 1.0, 1.01 }; double shadow[3] = { -2, 0, 0 }; double white[3] = { 100, 0, 0 }; DOUBLEMASK *d652d50 = im_create_dmaskv( "d652d50", 3, 3, 1.13529, -0.0604663, -0.0606321, 0.0975399, 0.935024, -0.0256156, -0.0336428, 0.0414702, 0.994135 ); im_add_close_callback( out, (im_callback_fn) im_free_dmask, d652d50, NULL ); return( /* Set of descriptors for this operation. */ im_open_local_array( out, t, 18, "im_benchmark", "p" ) || /* Unpack to float. */ im_LabQ2Lab( in, t[0] ) || /* Crop 100 pixels off all edges. */ im_extract_area( t[0], t[1], 100, 100, t[0]->Xsize - 200, t[0]->Ysize - 200 ) || /* Shrink by 10%, bilinear interp. */ im_affine( t[1], t[2], 0.9, 0, 0, 0.9, 0, 0, 0, 0, t[1]->Xsize * 0.9, t[1]->Ysize * 0.9 ) || /* Find L ~= 100 areas (white surround). */ im_extract_band( t[2], t[3], 0 ) || im_moreconst( t[3], t[4], 99 ) || /* Adjust white point and shadows. */ im_lintra_vec( 3, darken, t[2], zero, t[5] ) || im_Lab2XYZ( t[5], t[6] ) || im_recomb( t[6], t[7], d652d50 ) || im_lintra_vec( 3, whitepoint, t[7], zero, t[8] ) || im_lintra( 1.5, t[8], 0.0, t[9] ) || im_XYZ2Lab( t[9], t[10] ) || im_lintra_vec( 3, one, t[10], shadow, t[11] ) || /* Make a solid white image. */ im_black( t[12], t[4]->Xsize, t[4]->Ysize, 3 ) || im_lintra_vec( 3, zero, t[12], white, t[13] ) || /* Reattach border. */ im_ifthenelse( t[4], t[13], t[11], t[14] ) || /* Sharpen. */ im_Lab2LabQ( t[14], t[15] ) || im_sharpen( t[15], out, 11, 2.5, 40, 20, 0.5, 1.5 ) ); } /* Chain n benchmarks together to get a CPU-bound operation. */ int im_benchmarkn( IMAGE *in, IMAGE *out, int n ) { IMAGE *t[2]; if( n == 0 ) /* To sRGB. */ return( im_LabQ2disp( in, out, im_col_displays( 7 ) ) ); else return( im_open_local_array( out, t, 2, "benchmarkn", "p" ) || benchmark( in, t[0] ) || /* Expand back to the original size again ... * benchmark does a 200 pixel crop plus a 10% shrink, * so if we chain many of them together the image gets * too small. */ im_affine( t[0], t[1], (double) in->Xsize / t[0]->Xsize, 0, 0, (double) in->Ysize / t[0]->Ysize, 0, 0, 0, 0, in->Xsize, in->Ysize ) || im_benchmarkn( t[1], out, n - 1 ) ); } int im_benchmark2( IMAGE *in, double *out ) { IMAGE *t; return( !(t = im_open_local( in, "benchmarkn", "p" )) || im_benchmarkn( in, t, 1 ) || im_avg( t, out ) ); }