1/* 2 * Copyright (C) 2012 The Android Open Source Project 3 * 4 * Licensed under the Apache License, Version 2.0 (the "License"); 5 * you may not use this file except in compliance with the License. 6 * You may obtain a copy of the License at 7 * 8 * http://www.apache.org/licenses/LICENSE-2.0 9 * 10 * Unless required by applicable law or agreed to in writing, software 11 * distributed under the License is distributed on an "AS IS" BASIS, 12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 13 * See the License for the specific language governing permissions and 14 * limitations under the License. 15 */ 16 17 18#include "rsdCore.h" 19#include "rsdIntrinsics.h" 20#include "rsdAllocation.h" 21 22#include "rsdIntrinsicInlines.h" 23 24using namespace android; 25using namespace android::renderscript; 26 27struct ConvolveParams { 28 float fp[16]; 29 short ip[16]; 30 bool use3x3; 31 bool useDot; 32}; 33 34static void ColorMatrix_SetVar(const Context *dc, const Script *script, void * intrinsicData, 35 uint32_t slot, void *data, size_t dataLength) { 36 ConvolveParams *cp = (ConvolveParams *)intrinsicData; 37 38 rsAssert(slot == 0); 39 memcpy (cp->fp, data, dataLength); 40 for(int ct=0; ct < 16; ct++) { 41 cp->ip[ct] = (short)(cp->fp[ct] * 255.f + 0.5f); 42 } 43 44 if ((cp->ip[3] == 0) && (cp->ip[7] == 0) && (cp->ip[11] == 0) && 45 (cp->ip[12] == 0) && (cp->ip[13] == 0) && (cp->ip[14] == 0) && 46 (cp->ip[15] == 255)) { 47 cp->use3x3 = true; 48 49 if ((cp->ip[0] == cp->ip[1]) && (cp->ip[0] == cp->ip[2]) && 50 (cp->ip[4] == cp->ip[5]) && (cp->ip[4] == cp->ip[6]) && 51 (cp->ip[8] == cp->ip[9]) && (cp->ip[8] == cp->ip[10])) { 52 cp->useDot = true; 53 } 54 } 55} 56 57extern "C" void rsdIntrinsicColorMatrix4x4_K(void *dst, const void *src, const short *coef, uint32_t count); 58extern "C" void rsdIntrinsicColorMatrix3x3_K(void *dst, const void *src, const short *coef, uint32_t count); 59extern "C" void rsdIntrinsicColorMatrixDot_K(void *dst, const void *src, const short *coef, uint32_t count); 60 61static void One(const RsForEachStubParamStruct *p, uchar4 *out, 62 const uchar4 *py, const float* coeff) { 63 float4 i = convert_float4(py[0]); 64 65 float4 sum; 66 sum.x = i.x * coeff[0] + 67 i.y * coeff[4] + 68 i.z * coeff[8] + 69 i.w * coeff[12]; 70 sum.y = i.x * coeff[1] + 71 i.y * coeff[5] + 72 i.z * coeff[9] + 73 i.w * coeff[13]; 74 sum.z = i.x * coeff[2] + 75 i.y * coeff[6] + 76 i.z * coeff[10] + 77 i.w * coeff[14]; 78 sum.w = i.x * coeff[3] + 79 i.y * coeff[7] + 80 i.z * coeff[11] + 81 i.w * coeff[15]; 82 83 sum.x = sum.x < 0 ? 0 : (sum.x > 255 ? 255 : sum.x); 84 sum.y = sum.y < 0 ? 0 : (sum.y > 255 ? 255 : sum.y); 85 sum.z = sum.z < 0 ? 0 : (sum.z > 255 ? 255 : sum.z); 86 sum.w = sum.w < 0 ? 0 : (sum.w > 255 ? 255 : sum.w); 87 88 *out = convert_uchar4(sum); 89} 90 91static void ColorMatrix_uchar4(const RsForEachStubParamStruct *p, 92 uint32_t xstart, uint32_t xend, 93 uint32_t instep, uint32_t outstep) { 94 ConvolveParams *cp = (ConvolveParams *)p->usr; 95 uchar4 *out = (uchar4 *)p->out; 96 uchar4 *in = (uchar4 *)p->in; 97 uint32_t x1 = xstart; 98 uint32_t x2 = xend; 99 100 in += xstart; 101 out += xstart; 102 103 if(x2 > x1) { 104#if defined(ARCH_ARM_HAVE_NEON) 105 int32_t len = (x2 - x1) >> 2; 106 if(len > 0) { 107 if (cp->use3x3) { 108 if (cp->useDot) { 109 rsdIntrinsicColorMatrixDot_K(out, in, cp->ip, len); 110 } else { 111 rsdIntrinsicColorMatrix3x3_K(out, in, cp->ip, len); 112 } 113 } else { 114 rsdIntrinsicColorMatrix4x4_K(out, in, cp->ip, len); 115 } 116 x1 += len << 2; 117 out += len << 2; 118 in += len << 2; 119 } 120#endif 121 122 while(x1 != x2) { 123 One(p, out++, in++, cp->fp); 124 x1++; 125 } 126 } 127} 128 129void * rsdIntrinsic_InitColorMatrix(const android::renderscript::Context *dc, 130 android::renderscript::Script *script, 131 RsdIntriniscFuncs_t *funcs) { 132 133 script->mHal.info.exportedVariableCount = 1; 134 funcs->setVar = ColorMatrix_SetVar; 135 funcs->root = ColorMatrix_uchar4; 136 137 ConvolveParams *cp = (ConvolveParams *)calloc(1, sizeof(ConvolveParams)); 138 cp->fp[0] = 1.f; 139 cp->fp[5] = 1.f; 140 cp->fp[10] = 1.f; 141 cp->fp[15] = 1.f; 142 for(int ct=0; ct < 16; ct++) { 143 cp->ip[ct] = (short)(cp->fp[ct] * 255.f + 0.5f); 144 } 145 return cp; 146} 147 148 149