xref: /third_party/mesa3d/src/gallium/drivers/llvmpipe/lp_linear_interp.c

/**************************************************************************
 *
 * Copyright 2010-2021 VMware, Inc.
 * All Rights Reserved.
 *
 * 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, sub license, 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 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 NON-INFRINGEMENT. IN NO EVENT SHALL
 * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS 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.
 *
 * The above copyright notice and this permission notice (including the
 * next paragraph) shall be included in all copies or substantial portions
 * of the Software.
 *
 **************************************************************************/


#include "pipe/p_config.h"

#include "util/u_math.h"
#include "util/u_cpu_detect.h"
#include "util/u_pack_color.h"
#include "util/u_rect.h"
#include "util/u_sse.h"

#include "lp_jit.h"
#include "lp_rast.h"
#include "lp_debug.h"
#include "lp_state_fs.h"
#include "lp_linear_priv.h"


#if defined(PIPE_ARCH_SSE)

#define FIXED15_ONE 0x7fff

/* Translate floating point value to 1.15 unsigned fixed-point.
 */
static inline ushort
float_to_ufixed_1_15(float f)
{
   return CLAMP((unsigned)(f * (float)FIXED15_ONE), 0, FIXED15_ONE);
}


/* Translate floating point value to 1.15 signed fixed-point.
 */
static inline int16_t
float_to_sfixed_1_15(float f)
{
   return CLAMP((signed)(f * (float)FIXED15_ONE), -FIXED15_ONE, FIXED15_ONE);
}


/* Interpolate in 1.15 space, but produce a packed row of 0.8 values.
 */
static const uint32_t *
interp_0_8(struct lp_linear_elem *elem)
{
   struct lp_linear_interp *interp = (struct lp_linear_interp *)elem;
   uint32_t *row = interp->row;
   __m128i a0 = interp->a0;
   const __m128i dadx = interp->dadx;
   const int width = (interp->width + 3) & ~3;

   for (int i = 0; i < width; i += 4) {
      __m128i l = _mm_srai_epi16(a0, 7); // l = a0 >> 7
      a0 = _mm_add_epi16(a0, dadx);      // a0 += dadx

      __m128i h = _mm_srai_epi16(a0, 7); // h = a0 >> 7
      a0 = _mm_add_epi16(a0, dadx);      // a0 += dadx

      // pack l[0..7] and h[0..7] as 16 bytes
      *(__m128i *)&row[i] =  _mm_packus_epi16(l, h);
   }

   // advance to next row
   interp->a0 = _mm_add_epi16(interp->a0, interp->dady);
   return interp->row;
}

static const uint32_t *
interp_noop(struct lp_linear_elem *elem)
{
   struct lp_linear_interp *interp = (struct lp_linear_interp *)elem;
   return interp->row;
}


static const uint32_t *
interp_check(struct lp_linear_elem *elem)
{
   struct lp_linear_interp *interp = (struct lp_linear_interp *)elem;
   interp->row[0] = 1;
   return interp->row;
}

/* Not quite a noop - we use row[0] to track whether this gets called
 * or not, so we can optimize which interpolants we care about.
 */
void
lp_linear_init_noop_interp(struct lp_linear_interp *interp)
{
   interp->row[0] = 0;
   interp->base.fetch = interp_check;
}

boolean
lp_linear_init_interp(struct lp_linear_interp *interp,
                      int x, int y, int width, int height,
                      unsigned usage_mask,
                      boolean perspective,
                      float oow,
                      const float *a0,
                      const float *dadx,
                      const float *dady)
{
   float s0[4];
   float dsdx[4];
   float dsdy[4];
   int16_t s0_fp[8];
   int16_t dsdx_fp[4];
   int16_t dsdy_fp[4];

   /* Zero coefficients to avoid using uninitialised values */
   memset(s0, 0, sizeof(s0));
   memset(dsdx, 0, sizeof(dsdx));
   memset(dsdy, 0, sizeof(dsdy));
   memset(s0_fp, 0, sizeof(s0_fp));
   memset(dsdx_fp, 0, sizeof(dsdx_fp));
   memset(dsdy_fp, 0, sizeof(dsdy_fp));

   if (perspective && oow != 1.0f) {
      for (unsigned j = 0; j < 4; j++) {
         if (usage_mask & (1<<j)) {
            s0[j]   =   a0[j] * oow;
            dsdx[j] = dadx[j] * oow;
            dsdy[j] = dady[j] * oow;
         }
      }
   } else {
      for (unsigned j = 0; j < 4; j++) {
         if (usage_mask & (1<<j)) {
            s0[j]   =   a0[j];
            dsdx[j] = dadx[j];
            dsdy[j] = dady[j];
         }
      }
   }

   s0[0] += x * dsdx[0] + y * dsdy[0];
   s0[1] += x * dsdx[1] + y * dsdy[1];
   s0[2] += x * dsdx[2] + y * dsdy[2];
   s0[3] += x * dsdx[3] + y * dsdy[3];

   /* XXX: lift all of this into the rectangle setup code.
    *
    * For rectangles with linear shaders, at setup time:
    *    - if w is constant (else mark as non-fastpath)
    *        - premultiply perspective interpolants by w
    *        - set w = 1 in position
    *   - check all interpolants for min/max 0..1 (else mark as
    *          non-fastpath)
    */
   for (unsigned j = 0; j < 4; j++) {
      if (usage_mask & (1<<j)) {
         // compute texcoords at rect corners
         float a = s0[j];
         float b = s0[j] + (width  - 1) * dsdx[j];
         float c = s0[j] + (height - 1) * dsdy[j];
         float d = s0[j] + (height - 1) * dsdy[j] + (width - 1) * dsdx[j];

         if (MIN4(a,b,c,d) < 0.0)
            FAIL("min < 0.0"); // out of bounds

         if (MAX4(a,b,c,d) > 1.0)
            FAIL("max > 1.0"); // out of bounds

         dsdx_fp[j]   = float_to_sfixed_1_15(dsdx[j]);
         dsdy_fp[j]   = float_to_sfixed_1_15(dsdy[j]);

         s0_fp[j]     = float_to_ufixed_1_15(s0[j]);  // first pixel
         s0_fp[j + 4] = s0_fp[j] + dsdx_fp[j];        // second pixel

         dsdx_fp[j] *= 2;
      }
   }

   interp->width = align(width, 4);
   /* RGBA->BGRA swizzle here */
   interp->a0    = _mm_setr_epi16(s0_fp[2], s0_fp[1], s0_fp[0], s0_fp[3],
                                  s0_fp[6], s0_fp[5], s0_fp[4], s0_fp[7]);

   interp->dadx  = _mm_setr_epi16(dsdx_fp[2], dsdx_fp[1], dsdx_fp[0], dsdx_fp[3],
                                  dsdx_fp[2], dsdx_fp[1], dsdx_fp[0], dsdx_fp[3]);

   interp->dady  = _mm_setr_epi16(dsdy_fp[2], dsdy_fp[1], dsdy_fp[0], dsdy_fp[3],
                                  dsdy_fp[2], dsdy_fp[1], dsdy_fp[0], dsdy_fp[3]);

   /* If the value is y-invariant, eagerly calculate it here and then
    * always return the precalculated value.
    */
   if (dsdy[0] == 0 &&
       dsdy[1] == 0 &&
       dsdy[2] == 0 &&
       dsdy[3] == 0) {
      interp_0_8(&interp->base);
      interp->base.fetch = interp_noop;
   } else {
      interp->base.fetch = interp_0_8;
   }

   return TRUE;
}

#else
boolean
lp_linear_init_interp(struct lp_linear_interp *interp,
                      int x, int y, int width, int height,
                      unsigned usage_mask,
                      boolean perspective,
                      float oow,
                      const float *a0,
                      const float *dadx,
                      const float *dady)
{
   return FALSE;
}
#endif