DXR is a code search and navigation tool aimed at making sense of large projects. It supports full-text and regex searches as well as structural queries.

Mercurial (fddffdeab170)

VCS Links

Line Code
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269
/* vim: set ts=8 sw=8 noexpandtab: */
//  qcms
//  Copyright (C) 2009 Mozilla Corporation
//  Copyright (C) 1998-2007 Marti Maria
//
// Permission is hereby granted, free of charge, to any person obtaining
// a copy of this software and associated documentation files (the "Software"),
// to deal in the Software without restriction, including without limitation
// the rights to use, copy, modify, merge, publish, distribute, sublicense,
// and/or sell copies of the Software, and to permit persons to whom the Software
// is furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
// EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO
// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
// NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
// LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
// OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
// WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

#include <altivec.h>

#include "qcmsint.h"

#define FLOATSCALE (float)(PRECACHE_OUTPUT_SIZE)
#define CLAMPMAXVAL (((float) (PRECACHE_OUTPUT_SIZE - 1)) / PRECACHE_OUTPUT_SIZE)
static const ALIGN float floatScaleX4 = FLOATSCALE;
static const ALIGN float clampMaxValueX4 = CLAMPMAXVAL;

inline vector float load_aligned_float(float *dataPtr)
{
	vector float data = vec_lde(0, dataPtr);
	vector unsigned char moveToStart = vec_lvsl(0, dataPtr);
	return vec_perm(data, data, moveToStart);
}

void qcms_transform_data_rgb_out_lut_altivec(qcms_transform *transform,
                                             unsigned char *src,
                                             unsigned char *dest,
                                             size_t length)
{
	unsigned int i;
	float (*mat)[4] = transform->matrix;
	char input_back[32];
	/* Ensure we have a buffer that's 16 byte aligned regardless of the original
	 * stack alignment. We can't use __attribute__((aligned(16))) or __declspec(align(32))
	 * because they don't work on stack variables. gcc 4.4 does do the right thing
	 * on x86 but that's too new for us right now. For more info: gcc bug #16660 */
	float const *input = (float*)(((uintptr_t)&input_back[16]) & ~0xf);
	/* share input and output locations to save having to keep the
 	 * locations in separate registers */
	uint32_t const *output = (uint32_t*)input;

	/* deref *transform now to avoid it in loop */
	const float *igtbl_r = transform->input_gamma_table_r;
	const float *igtbl_g = transform->input_gamma_table_g;
	const float *igtbl_b = transform->input_gamma_table_b;

	/* deref *transform now to avoid it in loop */
	const uint8_t *otdata_r = &transform->output_table_r->data[0];
	const uint8_t *otdata_g = &transform->output_table_g->data[0];
	const uint8_t *otdata_b = &transform->output_table_b->data[0];

	/* input matrix values never change */
	const vector float mat0 = vec_ldl(0, (vector float*)mat[0]);
	const vector float mat1 = vec_ldl(0, (vector float*)mat[1]);
	const vector float mat2 = vec_ldl(0, (vector float*)mat[2]);

	/* these values don't change, either */
	const vector float max = vec_splat(vec_lde(0, (float*)&clampMaxValueX4), 0);
	const vector float min = (vector float)vec_splat_u32(0);
	const vector float scale = vec_splat(vec_lde(0, (float*)&floatScaleX4), 0);

	/* working variables */
	vector float vec_r, vec_g, vec_b, result;

	/* CYA */
	if (!length)
		return;

	/* one pixel is handled outside of the loop */
	length--;

	/* setup for transforming 1st pixel */
	vec_r = load_aligned_float((float*)&igtbl_r[src[0]]);
	vec_g = load_aligned_float((float*)&igtbl_r[src[1]]);
	vec_b = load_aligned_float((float*)&igtbl_r[src[2]]);
	src += 3;

	/* transform all but final pixel */

	for (i=0; i<length; i++)
	{
		/* position values from gamma tables */
		vec_r = vec_splat(vec_r, 0);
		vec_g = vec_splat(vec_g, 0);
		vec_b = vec_splat(vec_b, 0);

		/* gamma * matrix */
		vec_r = vec_madd(vec_r, mat0, min);
		vec_g = vec_madd(vec_g, mat1, min);
		vec_b = vec_madd(vec_b, mat2, min);

		/* crunch, crunch, crunch */
		vec_r = vec_add(vec_r, vec_add(vec_g, vec_b));
		vec_r = vec_max(min, vec_r);
		vec_r = vec_min(max, vec_r);
		result = vec_madd(vec_r, scale, min);

		/* store calc'd output tables indices */
		vec_st(vec_ctu(vec_round(result), 0), 0, (vector unsigned int*)output);

		/* load for next loop while store completes */
		vec_r = load_aligned_float((float*)&igtbl_r[src[0]]);
		vec_g = load_aligned_float((float*)&igtbl_r[src[1]]);
		vec_b = load_aligned_float((float*)&igtbl_r[src[2]]);
		src += 3;

		/* use calc'd indices to output RGB values */
		dest[0] = otdata_r[output[0]];
		dest[1] = otdata_g[output[1]];
		dest[2] = otdata_b[output[2]];
		dest += 3;
	}

	/* handle final (maybe only) pixel */

	vec_r = vec_splat(vec_r, 0);
	vec_g = vec_splat(vec_g, 0);
	vec_b = vec_splat(vec_b, 0);

	vec_r = vec_madd(vec_r, mat0, min);
	vec_g = vec_madd(vec_g, mat1, min);
	vec_b = vec_madd(vec_b, mat2, min);

	vec_r = vec_add(vec_r, vec_add(vec_g, vec_b));
	vec_r = vec_max(min, vec_r);
	vec_r = vec_min(max, vec_r);
	result = vec_madd(vec_r, scale, min);

	vec_st(vec_ctu(vec_round(result),0),0,(vector unsigned int*)output);

	dest[0] = otdata_r[output[0]];
	dest[1] = otdata_g[output[1]];
	dest[2] = otdata_b[output[2]];
}

void qcms_transform_data_rgba_out_lut_altivec(qcms_transform *transform,
                                              unsigned char *src,
                                              unsigned char *dest,
                                              size_t length)
{
	unsigned int i;
	float (*mat)[4] = transform->matrix;
	char input_back[32];
	/* Ensure we have a buffer that's 16 byte aligned regardless of the original
	 * stack alignment. We can't use __attribute__((aligned(16))) or __declspec(align(32))
	 * because they don't work on stack variables. gcc 4.4 does do the right thing
	 * on x86 but that's too new for us right now. For more info: gcc bug #16660 */
	float const *input = (float*)(((uintptr_t)&input_back[16]) & ~0xf);
	/* share input and output locations to save having to keep the
	 * locations in separate registers */
	uint32_t const *output = (uint32_t*)input;

	/* deref *transform now to avoid it in loop */
	const float *igtbl_r = transform->input_gamma_table_r;
	const float *igtbl_g = transform->input_gamma_table_g;
	const float *igtbl_b = transform->input_gamma_table_b;

	/* deref *transform now to avoid it in loop */
	const uint8_t *otdata_r = &transform->output_table_r->data[0];
	const uint8_t *otdata_g = &transform->output_table_g->data[0];
	const uint8_t *otdata_b = &transform->output_table_b->data[0];

	/* input matrix values never change */
	const vector float mat0 = vec_ldl(0, (vector float*)mat[0]);
	const vector float mat1 = vec_ldl(0, (vector float*)mat[1]);
	const vector float mat2 = vec_ldl(0, (vector float*)mat[2]);

	/* these values don't change, either */
	const vector float max = vec_splat(vec_lde(0, (float*)&clampMaxValueX4), 0);
	const vector float min = (vector float)vec_splat_u32(0);
	const vector float scale = vec_splat(vec_lde(0, (float*)&floatScaleX4), 0);

	/* working variables */
	vector float vec_r, vec_g, vec_b, result;
	unsigned char alpha;

	/* CYA */
	if (!length)
		return;

	/* one pixel is handled outside of the loop */
	length--;

	/* setup for transforming 1st pixel */
	vec_r = load_aligned_float((float*)&igtbl_r[src[0]]);
	vec_g = load_aligned_float((float*)&igtbl_r[src[1]]);
	vec_b = load_aligned_float((float*)&igtbl_r[src[2]]);
	alpha = src[3];
	src += 4;

	/* transform all but final pixel */

	for (i=0; i<length; i++)
	{
		/* position values from gamma tables */
		vec_r = vec_splat(vec_r, 0);
		vec_g = vec_splat(vec_g, 0);
		vec_b = vec_splat(vec_b, 0);

		/* gamma * matrix */
		vec_r = vec_madd(vec_r, mat0, min);
		vec_g = vec_madd(vec_g, mat1, min);
		vec_b = vec_madd(vec_b, mat2, min);

		/* store alpha for this pixel; load alpha for next */
		dest[3] = alpha;
		alpha = src[3];

		/* crunch, crunch, crunch */
		vec_r = vec_add(vec_r, vec_add(vec_g, vec_b));
		vec_r = vec_max(min, vec_r);
		vec_r = vec_min(max, vec_r);
		result = vec_madd(vec_r, scale, min);

		/* store calc'd output tables indices */
		vec_st(vec_ctu(vec_round(result), 0), 0, (vector unsigned int*)output);

		/* load gamma values for next loop while store completes */
		vec_r = load_aligned_float((float*)&igtbl_r[src[0]]);
		vec_g = load_aligned_float((float*)&igtbl_r[src[1]]);
		vec_b = load_aligned_float((float*)&igtbl_r[src[2]]);
		src += 4;

		/* use calc'd indices to output RGB values */
		dest[0] = otdata_r[output[0]];
		dest[1] = otdata_g[output[1]];
		dest[2] = otdata_b[output[2]];
		dest += 4;
	}

	/* handle final (maybe only) pixel */

	vec_r = vec_splat(vec_r, 0);
	vec_g = vec_splat(vec_g, 0);
	vec_b = vec_splat(vec_b, 0);

	vec_r = vec_madd(vec_r, mat0, min);
	vec_g = vec_madd(vec_g, mat1, min);
	vec_b = vec_madd(vec_b, mat2, min);

	dest[3] = alpha;

	vec_r = vec_add(vec_r, vec_add(vec_g, vec_b));
	vec_r = vec_max(min, vec_r);
	vec_r = vec_min(max, vec_r);
	result = vec_madd(vec_r, scale, min);

	vec_st(vec_ctu(vec_round(result), 0), 0, (vector unsigned int*)output);

	dest[0] = otdata_r[output[0]];
	dest[1] = otdata_g[output[1]];
	dest[2] = otdata_b[output[2]];
}