src/gallium/drivers/vc4/vc4_tiling_lt.c

   1 /*
   2  * Copyright © 2017 Broadcom
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice (including the next
  12  * paragraph) shall be included in all copies or substantial portions of the
  13  * Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  21  * IN THE SOFTWARE.
  22  */
  23
  24 /** @file vc4_tiling_lt.c
  25  *
  26  * Helper functions from vc4_tiling.c that will be compiled for using NEON
  27  * assembly or not.
  28  */
  29
  30 #include <string.h>
  31 #include "pipe/p_state.h"
  32 #include "vc4_tiling.h"
  33
  34 /** Returns the stride in bytes of a 64-byte microtile. */
  35 static uint32_t
  36 vc4_utile_stride(int cpp)
  37 {
  38         switch (cpp) {
  39         case 1:
  40                 return 8;
  41         case 2:
  42         case 4:
  43         case 8:
  44                 return 16;
  45         default:
  46                 unreachable("bad cpp");
  47         }
  48 }
  49
  50 static void
  51 vc4_load_utile(void *dst, void *src, uint32_t dst_stride, uint32_t cpp)
  52 {
  53         uint32_t src_stride = vc4_utile_stride(cpp);
  54
  55         for (uint32_t src_offset = 0; src_offset < 64; src_offset += src_stride) {
  56                 memcpy(dst, src + src_offset, src_stride);
  57                 dst += dst_stride;
  58         }
  59 }
  60
  61 static void
  62 vc4_store_utile(void *dst, void *src, uint32_t src_stride, uint32_t cpp)
  63 {
  64         uint32_t dst_stride = vc4_utile_stride(cpp);
  65
  66         for (uint32_t dst_offset = 0; dst_offset < 64; dst_offset += dst_stride) {
  67                 memcpy(dst + dst_offset, src, dst_stride);
  68                 src += src_stride;
  69         }
  70 }
  71
  72 void
  73 vc4_load_lt_image(void *dst, uint32_t dst_stride,
  74                   void *src, uint32_t src_stride,
  75                   int cpp, const struct pipe_box *box)
  76 {
  77         uint32_t utile_w = vc4_utile_width(cpp);
  78         uint32_t utile_h = vc4_utile_height(cpp);
  79         uint32_t xstart = box->x;
  80         uint32_t ystart = box->y;
  81
  82         for (uint32_t y = 0; y < box->height; y += utile_h) {
  83                 for (int x = 0; x < box->width; x += utile_w) {
  84                         vc4_load_utile(dst + (dst_stride * y +
  85                                               x * cpp),
  86                                        src + ((ystart + y) * src_stride +
  87                                               (xstart + x) * 64 / utile_w),
  88                                        dst_stride, cpp);
  89                 }
  90         }
  91 }
  92
  93 void
  94 vc4_store_lt_image(void *dst, uint32_t dst_stride,
  95                    void *src, uint32_t src_stride,
  96                    int cpp, const struct pipe_box *box)
  97 {
  98         uint32_t utile_w = vc4_utile_width(cpp);
  99         uint32_t utile_h = vc4_utile_height(cpp);
 100         uint32_t xstart = box->x;
 101         uint32_t ystart = box->y;
 102
 103         for (uint32_t y = 0; y < box->height; y += utile_h) {
 104                 for (int x = 0; x < box->width; x += utile_w) {
 105                         vc4_store_utile(dst + ((ystart + y) * dst_stride +
 106                                                (xstart + x) * 64 / utile_w),
 107                                         src + (src_stride * y +
 108                                                x * cpp),
 109                                         src_stride, cpp);
 110                 }
 111         }
 112 }