perf tools: Don't clone maps from parent when synthesizing forks
[linux/fpc-iii.git] / drivers / media / platform / ti-vpe / sc.c
blobe9273b7137829a6865b5c558ed815834355393d5
1 /*
2 * Scaler library
4 * Copyright (c) 2013 Texas Instruments Inc.
6 * David Griego, <dagriego@biglakesoftware.com>
7 * Dale Farnsworth, <dale@farnsworth.org>
8 * Archit Taneja, <archit@ti.com>
10 * This program is free software; you can redistribute it and/or modify it
11 * under the terms of the GNU General Public License version 2 as published by
12 * the Free Software Foundation.
15 #include <linux/err.h>
16 #include <linux/io.h>
17 #include <linux/module.h>
18 #include <linux/platform_device.h>
19 #include <linux/slab.h>
21 #include "sc.h"
22 #include "sc_coeff.h"
24 void sc_dump_regs(struct sc_data *sc)
26 struct device *dev = &sc->pdev->dev;
28 #define DUMPREG(r) dev_dbg(dev, "%-35s %08x\n", #r, \
29 ioread32(sc->base + CFG_##r))
31 dev_dbg(dev, "SC Registers @ %pa:\n", &sc->res->start);
33 DUMPREG(SC0);
34 DUMPREG(SC1);
35 DUMPREG(SC2);
36 DUMPREG(SC3);
37 DUMPREG(SC4);
38 DUMPREG(SC5);
39 DUMPREG(SC6);
40 DUMPREG(SC8);
41 DUMPREG(SC9);
42 DUMPREG(SC10);
43 DUMPREG(SC11);
44 DUMPREG(SC12);
45 DUMPREG(SC13);
46 DUMPREG(SC17);
47 DUMPREG(SC18);
48 DUMPREG(SC19);
49 DUMPREG(SC20);
50 DUMPREG(SC21);
51 DUMPREG(SC22);
52 DUMPREG(SC23);
53 DUMPREG(SC24);
54 DUMPREG(SC25);
56 #undef DUMPREG
58 EXPORT_SYMBOL(sc_dump_regs);
61 * set the horizontal scaler coefficients according to the ratio of output to
62 * input widths, after accounting for up to two levels of decimation
64 void sc_set_hs_coeffs(struct sc_data *sc, void *addr, unsigned int src_w,
65 unsigned int dst_w)
67 int sixteenths;
68 int idx;
69 int i, j;
70 u16 *coeff_h = addr;
71 const u16 *cp;
73 if (dst_w > src_w) {
74 idx = HS_UP_SCALE;
75 } else {
76 if ((dst_w << 1) < src_w)
77 dst_w <<= 1; /* first level decimation */
78 if ((dst_w << 1) < src_w)
79 dst_w <<= 1; /* second level decimation */
81 if (dst_w == src_w) {
82 idx = HS_LE_16_16_SCALE;
83 } else {
84 sixteenths = (dst_w << 4) / src_w;
85 if (sixteenths < 8)
86 sixteenths = 8;
87 idx = HS_LT_9_16_SCALE + sixteenths - 8;
91 cp = scaler_hs_coeffs[idx];
93 for (i = 0; i < SC_NUM_PHASES * 2; i++) {
94 for (j = 0; j < SC_H_NUM_TAPS; j++)
95 *coeff_h++ = *cp++;
97 * for each phase, the scaler expects space for 8 coefficients
98 * in it's memory. For the horizontal scaler, we copy the first
99 * 7 coefficients and skip the last slot to move to the next
100 * row to hold coefficients for the next phase
102 coeff_h += SC_NUM_TAPS_MEM_ALIGN - SC_H_NUM_TAPS;
105 sc->load_coeff_h = true;
107 EXPORT_SYMBOL(sc_set_hs_coeffs);
110 * set the vertical scaler coefficients according to the ratio of output to
111 * input heights
113 void sc_set_vs_coeffs(struct sc_data *sc, void *addr, unsigned int src_h,
114 unsigned int dst_h)
116 int sixteenths;
117 int idx;
118 int i, j;
119 u16 *coeff_v = addr;
120 const u16 *cp;
122 if (dst_h > src_h) {
123 idx = VS_UP_SCALE;
124 } else if (dst_h == src_h) {
125 idx = VS_1_TO_1_SCALE;
126 } else {
127 sixteenths = (dst_h << 4) / src_h;
128 if (sixteenths < 8)
129 sixteenths = 8;
130 idx = VS_LT_9_16_SCALE + sixteenths - 8;
133 cp = scaler_vs_coeffs[idx];
135 for (i = 0; i < SC_NUM_PHASES * 2; i++) {
136 for (j = 0; j < SC_V_NUM_TAPS; j++)
137 *coeff_v++ = *cp++;
139 * for the vertical scaler, we copy the first 5 coefficients and
140 * skip the last 3 slots to move to the next row to hold
141 * coefficients for the next phase
143 coeff_v += SC_NUM_TAPS_MEM_ALIGN - SC_V_NUM_TAPS;
146 sc->load_coeff_v = true;
148 EXPORT_SYMBOL(sc_set_vs_coeffs);
150 void sc_config_scaler(struct sc_data *sc, u32 *sc_reg0, u32 *sc_reg8,
151 u32 *sc_reg17, unsigned int src_w, unsigned int src_h,
152 unsigned int dst_w, unsigned int dst_h)
154 struct device *dev = &sc->pdev->dev;
155 u32 val;
156 int dcm_x, dcm_shift;
157 bool use_rav;
158 unsigned long lltmp;
159 u32 lin_acc_inc, lin_acc_inc_u;
160 u32 col_acc_offset;
161 u16 factor = 0;
162 int row_acc_init_rav = 0, row_acc_init_rav_b = 0;
163 u32 row_acc_inc = 0, row_acc_offset = 0, row_acc_offset_b = 0;
165 * location of SC register in payload memory with respect to the first
166 * register in the mmr address data block
168 u32 *sc_reg9 = sc_reg8 + 1;
169 u32 *sc_reg12 = sc_reg8 + 4;
170 u32 *sc_reg13 = sc_reg8 + 5;
171 u32 *sc_reg24 = sc_reg17 + 7;
173 val = sc_reg0[0];
175 /* clear all the features(they may get enabled elsewhere later) */
176 val &= ~(CFG_SELFGEN_FID | CFG_TRIM | CFG_ENABLE_SIN2_VER_INTP |
177 CFG_INTERLACE_I | CFG_DCM_4X | CFG_DCM_2X | CFG_AUTO_HS |
178 CFG_ENABLE_EV | CFG_USE_RAV | CFG_INVT_FID | CFG_SC_BYPASS |
179 CFG_INTERLACE_O | CFG_Y_PK_EN | CFG_HP_BYPASS | CFG_LINEAR);
181 if (src_w == dst_w && src_h == dst_h) {
182 val |= CFG_SC_BYPASS;
183 sc_reg0[0] = val;
184 return;
187 /* we only support linear scaling for now */
188 val |= CFG_LINEAR;
190 /* configure horizontal scaler */
192 /* enable 2X or 4X decimation */
193 dcm_x = src_w / dst_w;
194 if (dcm_x > 4) {
195 val |= CFG_DCM_4X;
196 dcm_shift = 2;
197 } else if (dcm_x > 2) {
198 val |= CFG_DCM_2X;
199 dcm_shift = 1;
200 } else {
201 dcm_shift = 0;
204 lltmp = dst_w - 1;
205 lin_acc_inc = div64_u64(((u64)(src_w >> dcm_shift) - 1) << 24, lltmp);
206 lin_acc_inc_u = 0;
207 col_acc_offset = 0;
209 dev_dbg(dev, "hs config: src_w = %d, dst_w = %d, decimation = %s, lin_acc_inc = %08x\n",
210 src_w, dst_w, dcm_shift == 2 ? "4x" :
211 (dcm_shift == 1 ? "2x" : "none"), lin_acc_inc);
213 /* configure vertical scaler */
215 /* use RAV for vertical scaler if vertical downscaling is > 4x */
216 if (dst_h < (src_h >> 2)) {
217 use_rav = true;
218 val |= CFG_USE_RAV;
219 } else {
220 use_rav = false;
223 if (use_rav) {
224 /* use RAV */
225 factor = (u16) ((dst_h << 10) / src_h);
227 row_acc_init_rav = factor + ((1 + factor) >> 1);
228 if (row_acc_init_rav >= 1024)
229 row_acc_init_rav -= 1024;
231 row_acc_init_rav_b = row_acc_init_rav +
232 (1 + (row_acc_init_rav >> 1)) -
233 (1024 >> 1);
235 if (row_acc_init_rav_b < 0) {
236 row_acc_init_rav_b += row_acc_init_rav;
237 row_acc_init_rav *= 2;
240 dev_dbg(dev, "vs config(RAV): src_h = %d, dst_h = %d, factor = %d, acc_init = %08x, acc_init_b = %08x\n",
241 src_h, dst_h, factor, row_acc_init_rav,
242 row_acc_init_rav_b);
243 } else {
244 /* use polyphase */
245 row_acc_inc = ((src_h - 1) << 16) / (dst_h - 1);
246 row_acc_offset = 0;
247 row_acc_offset_b = 0;
249 dev_dbg(dev, "vs config(POLY): src_h = %d, dst_h = %d,row_acc_inc = %08x\n",
250 src_h, dst_h, row_acc_inc);
254 sc_reg0[0] = val;
255 sc_reg0[1] = row_acc_inc;
256 sc_reg0[2] = row_acc_offset;
257 sc_reg0[3] = row_acc_offset_b;
259 sc_reg0[4] = ((lin_acc_inc_u & CFG_LIN_ACC_INC_U_MASK) <<
260 CFG_LIN_ACC_INC_U_SHIFT) | (dst_w << CFG_TAR_W_SHIFT) |
261 (dst_h << CFG_TAR_H_SHIFT);
263 sc_reg0[5] = (src_w << CFG_SRC_W_SHIFT) | (src_h << CFG_SRC_H_SHIFT);
265 sc_reg0[6] = (row_acc_init_rav_b << CFG_ROW_ACC_INIT_RAV_B_SHIFT) |
266 (row_acc_init_rav << CFG_ROW_ACC_INIT_RAV_SHIFT);
268 *sc_reg9 = lin_acc_inc;
270 *sc_reg12 = col_acc_offset << CFG_COL_ACC_OFFSET_SHIFT;
272 *sc_reg13 = factor;
274 *sc_reg24 = (src_w << CFG_ORG_W_SHIFT) | (src_h << CFG_ORG_H_SHIFT);
276 EXPORT_SYMBOL(sc_config_scaler);
278 struct sc_data *sc_create(struct platform_device *pdev, const char *res_name)
280 struct sc_data *sc;
282 dev_dbg(&pdev->dev, "sc_create\n");
284 sc = devm_kzalloc(&pdev->dev, sizeof(*sc), GFP_KERNEL);
285 if (!sc) {
286 dev_err(&pdev->dev, "couldn't alloc sc_data\n");
287 return ERR_PTR(-ENOMEM);
290 sc->pdev = pdev;
292 sc->res = platform_get_resource_byname(pdev, IORESOURCE_MEM, res_name);
293 if (!sc->res) {
294 dev_err(&pdev->dev, "missing '%s' platform resources data\n",
295 res_name);
296 return ERR_PTR(-ENODEV);
299 sc->base = devm_ioremap_resource(&pdev->dev, sc->res);
300 if (IS_ERR(sc->base)) {
301 dev_err(&pdev->dev, "failed to ioremap\n");
302 return ERR_CAST(sc->base);
305 return sc;
307 EXPORT_SYMBOL(sc_create);
309 MODULE_DESCRIPTION("TI VIP/VPE Scaler");
310 MODULE_AUTHOR("Texas Instruments Inc.");
311 MODULE_LICENSE("GPL v2");