1 /****************************************************************************

     2 **

     3 ** Copyright (C) 2009 Nokia Corporation and/or its subsidiary(-ies).

     4 ** All rights reserved.

     5 ** Contact: Nokia Corporation (qt-info@nokia.com)

     6 **

     7 ** This file is part of the QtGui module of the Qt Toolkit.

     8 **

     9 ** $QT_BEGIN_LICENSE:LGPL$

    10 ** No Commercial Usage

    11 ** This file contains pre-release code and may not be distributed.

    12 ** You may use this file in accordance with the terms and conditions

    13 ** contained in the Technology Preview License Agreement accompanying

    14 ** this package.

    15 **

    16 ** GNU Lesser General Public License Usage

    17 ** Alternatively, this file may be used under the terms of the GNU Lesser

    18 ** General Public License version 2.1 as published by the Free Software

    19 ** Foundation and appearing in the file LICENSE.LGPL included in the

    20 ** packaging of this file.  Please review the following information to

    21 ** ensure the GNU Lesser General Public License version 2.1 requirements

    22 ** will be met: http://www.gnu.org/licenses/old-licenses/lgpl-2.1.html.

    23 **

    24 ** In addition, as a special exception, Nokia gives you certain additional

    25 ** rights.  These rights are described in the Nokia Qt LGPL Exception

    26 ** version 1.1, included in the file LGPL_EXCEPTION.txt in this package.

    27 **

    28 ** If you have questions regarding the use of this file, please contact

    29 ** Nokia at qt-info@nokia.com.

    30 **

    31 **

    32 **

    33 **

    34 **

    35 **

    36 **

    37 **

    38 ** $QT_END_LICENSE$

    39 **

    40 ****************************************************************************/

    41 

    42 #include <private/qdrawhelper_p.h>

    43 

    44 #ifdef QT_HAVE_NEON

    45 

    46 #include <private/qdrawhelper_neon_p.h>

    47 #include <arm_neon.h>

    48 

    49 QT_BEGIN_NAMESPACE

    50 

    51 static inline int16x8_t qvdiv_255_s16(int16x8_t x, int16x8_t half)

    52 {

    53     // result = (x + (x >> 8) + 0x80) >> 8

    54 

    55     const int16x8_t temp = vshrq_n_s16(x, 8); // x >> 8

    56     const int16x8_t sum_part = vaddq_s16(x, half); // x + 0x80

    57     const int16x8_t sum = vaddq_s16(temp, sum_part);

    58 

    59     return vreinterpretq_s16_u16(vshrq_n_u16(vreinterpretq_u16_s16(sum), 8));

    60 }

    61 

    62 static inline int16x8_t qvbyte_mul_s16(int16x8_t x, int16x8_t alpha, int16x8_t half)

    63 {

    64     // t = qRound(x * alpha / 255.0)

    65 

    66     const int16x8_t t = vmulq_s16(x, alpha); // t

    67     return qvdiv_255_s16(t, half);

    68 }

    69 

    70 static inline int16x8_t qvinterpolate_pixel_255(int16x8_t x, int16x8_t a, int16x8_t y, int16x8_t b, int16x8_t half)

    71 {

    72     // t = x * a + y * b

    73 

    74     const int16x8_t ta = vmulq_s16(x, a);

    75     const int16x8_t tb = vmulq_s16(y, b);

    76 

    77     return qvdiv_255_s16(vaddq_s16(ta, tb), half);

    78 }

    79 

    80 static inline int16x8_t qvsource_over_s16(int16x8_t src16, int16x8_t dst16, int16x8_t half, int16x8_t full)

    81 {

    82     const int16x4_t alpha16_high = vdup_lane_s16(vget_high_s16(src16), 3);

    83     const int16x4_t alpha16_low = vdup_lane_s16(vget_low_s16(src16), 3);

    84 

    85     const int16x8_t alpha16 = vsubq_s16(full, vcombine_s16(alpha16_low, alpha16_high));

    86 

    87     return vaddq_s16(src16, qvbyte_mul_s16(dst16, alpha16, half));

    88 }

    89 

    90 void qt_blend_argb32_on_argb32_neon(uchar *destPixels, int dbpl,

    91                                     const uchar *srcPixels, int sbpl,

    92                                     int w, int h,

    93                                     int const_alpha)

    94 {

    95     const uint *src = (const uint *) srcPixels;

    96     uint *dst = (uint *) destPixels;

    97     int16x8_t half = vdupq_n_s16(0x80);

    98     int16x8_t full = vdupq_n_s16(0xff);

    99     if (const_alpha == 256) {

   100         for (int y = 0; y < h; ++y) {

   101             int x = 0;

   102             for (; x < w-3; x += 4) {

   103                 int32x4_t src32 = vld1q_s32((int32_t *)&src[x]);

   104                 if ((src[x] & src[x+1] & src[x+2] & src[x+3]) >= 0xff000000) {

   105                     // all opaque

   106                     vst1q_s32((int32_t *)&dst[x], src32);

   107                 } else if (src[x] | src[x+1] | src[x+2] | src[x+3]) {

   108                     int32x4_t dst32 = vld1q_s32((int32_t *)&dst[x]);

   109 

   110                     const uint8x16_t src8 = vreinterpretq_u8_s32(src32);

   111                     const uint8x16_t dst8 = vreinterpretq_u8_s32(dst32);

   112 

   113                     const uint8x8_t src8_low = vget_low_u8(src8);

   114                     const uint8x8_t dst8_low = vget_low_u8(dst8);

   115 

   116                     const uint8x8_t src8_high = vget_high_u8(src8);

   117                     const uint8x8_t dst8_high = vget_high_u8(dst8);

   118 

   119                     const int16x8_t src16_low = vreinterpretq_s16_u16(vmovl_u8(src8_low));

   120                     const int16x8_t dst16_low = vreinterpretq_s16_u16(vmovl_u8(dst8_low));

   121 

   122                     const int16x8_t src16_high = vreinterpretq_s16_u16(vmovl_u8(src8_high));

   123                     const int16x8_t dst16_high = vreinterpretq_s16_u16(vmovl_u8(dst8_high));

   124 

   125                     const int16x8_t result16_low = qvsource_over_s16(src16_low, dst16_low, half, full);

   126                     const int16x8_t result16_high = qvsource_over_s16(src16_high, dst16_high, half, full);

   127 

   128                     const int32x2_t result32_low = vreinterpret_s32_s8(vmovn_s16(result16_low));

   129                     const int32x2_t result32_high = vreinterpret_s32_s8(vmovn_s16(result16_high));

   130 

   131                     vst1q_s32((int32_t *)&dst[x], vcombine_s32(result32_low, result32_high));

   132                 }

   133             }

   134             for (; x<w; ++x) {

   135                 uint s = src[x];

   136                 if (s >= 0xff000000)

   137                     dst[x] = s;

   138                 else if (s != 0)

   139                     dst[x] = s + BYTE_MUL(dst[x], qAlpha(~s));

   140             }

   141             dst = (quint32 *)(((uchar *) dst) + dbpl);

   142             src = (const quint32 *)(((const uchar *) src) + sbpl);

   143         }

   144     } else if (const_alpha != 0) {

   145         const_alpha = (const_alpha * 255) >> 8;

   146         int16x8_t const_alpha16 = vdupq_n_s16(const_alpha);

   147         for (int y = 0; y < h; ++y) {

   148             int x = 0;

   149             for (; x < w-3; x += 4) {

   150                 if (src[x] | src[x+1] | src[x+2] | src[x+3]) {

   151                     int32x4_t src32 = vld1q_s32((int32_t *)&src[x]);

   152                     int32x4_t dst32 = vld1q_s32((int32_t *)&dst[x]);

   153 

   154                     const uint8x16_t src8 = vreinterpretq_u8_s32(src32);

   155                     const uint8x16_t dst8 = vreinterpretq_u8_s32(dst32);

   156 

   157                     const uint8x8_t src8_low = vget_low_u8(src8);

   158                     const uint8x8_t dst8_low = vget_low_u8(dst8);

   159 

   160                     const uint8x8_t src8_high = vget_high_u8(src8);

   161                     const uint8x8_t dst8_high = vget_high_u8(dst8);

   162 

   163                     const int16x8_t src16_low = vreinterpretq_s16_u16(vmovl_u8(src8_low));

   164                     const int16x8_t dst16_low = vreinterpretq_s16_u16(vmovl_u8(dst8_low));

   165 

   166                     const int16x8_t src16_high = vreinterpretq_s16_u16(vmovl_u8(src8_high));

   167                     const int16x8_t dst16_high = vreinterpretq_s16_u16(vmovl_u8(dst8_high));

   168 

   169                     const int16x8_t srcalpha16_low = qvbyte_mul_s16(src16_low, const_alpha16, half);

   170                     const int16x8_t srcalpha16_high = qvbyte_mul_s16(src16_high, const_alpha16, half);

   171 

   172                     const int16x8_t result16_low = qvsource_over_s16(srcalpha16_low, dst16_low, half, full);

   173                     const int16x8_t result16_high = qvsource_over_s16(srcalpha16_high, dst16_high, half, full);

   174 

   175                     const int32x2_t result32_low = vreinterpret_s32_s8(vmovn_s16(result16_low));

   176                     const int32x2_t result32_high = vreinterpret_s32_s8(vmovn_s16(result16_high));

   177 

   178                     vst1q_s32((int32_t *)&dst[x], vcombine_s32(result32_low, result32_high));

   179                 }

   180             }

   181             for (; x<w; ++x) {

   182                 uint s = src[x];

   183                 if (s != 0) {

   184                     s = BYTE_MUL(s, const_alpha);

   185                     dst[x] = s + BYTE_MUL(dst[x], qAlpha(~s));

   186                 }

   187             }

   188             dst = (quint32 *)(((uchar *) dst) + dbpl);

   189             src = (const quint32 *)(((const uchar *) src) + sbpl);

   190         }

   191     }

   192 }

   193 

   194 // qblendfunctions.cpp

   195 void qt_blend_rgb32_on_rgb32(uchar *destPixels, int dbpl,

   196                              const uchar *srcPixels, int sbpl,

   197                              int w, int h,

   198                              int const_alpha);

   199 

   200 void qt_blend_rgb32_on_rgb32_neon(uchar *destPixels, int dbpl,

   201                                   const uchar *srcPixels, int sbpl,

   202                                   int w, int h,

   203                                   int const_alpha)

   204 {

   205     if (const_alpha != 256) {

   206         if (const_alpha != 0) {

   207             const uint *src = (const uint *) srcPixels;

   208             uint *dst = (uint *) destPixels;

   209             int16x8_t half = vdupq_n_s16(0x80);

   210             const_alpha = (const_alpha * 255) >> 8;

   211             int one_minus_const_alpha = 255 - const_alpha;

   212             int16x8_t const_alpha16 = vdupq_n_s16(const_alpha);

   213             int16x8_t one_minus_const_alpha16 = vdupq_n_s16(255 - const_alpha);

   214             for (int y = 0; y < h; ++y) {

   215                 int x = 0;

   216                 for (; x < w-3; x += 4) {

   217                     int32x4_t src32 = vld1q_s32((int32_t *)&src[x]);

   218                     int32x4_t dst32 = vld1q_s32((int32_t *)&dst[x]);

   219 

   220                     const uint8x16_t src8 = vreinterpretq_u8_s32(src32);

   221                     const uint8x16_t dst8 = vreinterpretq_u8_s32(dst32);

   222 

   223                     const uint8x8_t src8_low = vget_low_u8(src8);

   224                     const uint8x8_t dst8_low = vget_low_u8(dst8);

   225 

   226                     const uint8x8_t src8_high = vget_high_u8(src8);

   227                     const uint8x8_t dst8_high = vget_high_u8(dst8);

   228 

   229                     const int16x8_t src16_low = vreinterpretq_s16_u16(vmovl_u8(src8_low));

   230                     const int16x8_t dst16_low = vreinterpretq_s16_u16(vmovl_u8(dst8_low));

   231 

   232                     const int16x8_t src16_high = vreinterpretq_s16_u16(vmovl_u8(src8_high));

   233                     const int16x8_t dst16_high = vreinterpretq_s16_u16(vmovl_u8(dst8_high));

   234 

   235                     const int16x8_t result16_low = qvinterpolate_pixel_255(src16_low, const_alpha16, dst16_low, one_minus_const_alpha16, half);

   236                     const int16x8_t result16_high = qvinterpolate_pixel_255(src16_high, const_alpha16, dst16_high, one_minus_const_alpha16, half);

   237 

   238                     const int32x2_t result32_low = vreinterpret_s32_s8(vmovn_s16(result16_low));

   239                     const int32x2_t result32_high = vreinterpret_s32_s8(vmovn_s16(result16_high));

   240 

   241                     vst1q_s32((int32_t *)&dst[x], vcombine_s32(result32_low, result32_high));

   242                 }

   243                 for (; x<w; ++x) {

   244                     uint s = src[x];

   245                     s = BYTE_MUL(s, const_alpha);

   246                     dst[x] = INTERPOLATE_PIXEL_255(src[x], const_alpha, dst[x], one_minus_const_alpha);

   247                 }

   248                 dst = (quint32 *)(((uchar *) dst) + dbpl);

   249                 src = (const quint32 *)(((const uchar *) src) + sbpl);

   250             }

   251         }

   252     } else {

   253         qt_blend_rgb32_on_rgb32(destPixels, dbpl, srcPixels, sbpl, w, h, const_alpha);

   254     }

   255 }

   256 

   257 QT_END_NAMESPACE

   258 

   259 #endif // QT_HAVE_NEON

   260