1/*
2 * Mesa 3-D graphics library
3 *
4 * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
5 * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
6 *
7 * Permission is hereby granted, free of charge, to any person obtaining a
8 * copy of this software and associated documentation files (the "Software"),
9 * to deal in the Software without restriction, including without limitation
10 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
11 * and/or sell copies of the Software, and to permit persons to whom the
12 * Software is furnished to do so, subject to the following conditions:
13 *
14 * The above copyright notice and this permission notice shall be included
15 * in all copies or substantial portions of the Software.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
18 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
21 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
22 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
23 * OTHER DEALINGS IN THE SOFTWARE.
24 */
25
26
27/**
28 * \file swrast/s_span.c
29 * \brief Span processing functions used by all rasterization functions.
30 * This is where all the per-fragment tests are performed
31 * \author Brian Paul
32 */
33
34#include "c99_math.h"
35#include "main/errors.h"
36#include "main/glheader.h"
37#include "main/format_pack.h"
38#include "main/format_unpack.h"
39#include "main/macros.h"
40
41#include "main/image.h"
42#include "main/samplerobj.h"
43#include "main/state.h"
44#include "main/stencil.h"
45#include "main/teximage.h"
46
47#include "s_atifragshader.h"
48#include "s_alpha.h"
49#include "s_blend.h"
50#include "s_context.h"
51#include "s_depth.h"
52#include "s_fog.h"
53#include "s_logic.h"
54#include "s_masking.h"
55#include "s_fragprog.h"
56#include "s_span.h"
57#include "s_stencil.h"
58#include "s_texcombine.h"
59
60#include <stdbool.h>
61
62/**
63 * Set default fragment attributes for the span using the
64 * current raster values.  Used prior to glDraw/CopyPixels
65 * and glBitmap.
66 */
67void
68_swrast_span_default_attribs(struct gl_context *ctx, SWspan *span)
69{
70   GLchan r, g, b, a;
71   /* Z*/
72   {
73      const GLfloat depthMax = ctx->DrawBuffer->_DepthMaxF;
74      if (ctx->DrawBuffer->Visual.depthBits <= 16)
75         span->z = FloatToFixed(ctx->Current.RasterPos[2] * depthMax + 0.5F);
76      else {
77         GLfloat tmpf = ctx->Current.RasterPos[2] * depthMax;
78         tmpf = MIN2(tmpf, depthMax);
79         span->z = (GLint)tmpf;
80      }
81      span->zStep = 0;
82      span->interpMask |= SPAN_Z;
83   }
84
85   /* W (for perspective correction) */
86   span->attrStart[VARYING_SLOT_POS][3] = 1.0;
87   span->attrStepX[VARYING_SLOT_POS][3] = 0.0;
88   span->attrStepY[VARYING_SLOT_POS][3] = 0.0;
89
90   /* primary color, or color index */
91   UNCLAMPED_FLOAT_TO_CHAN(r, ctx->Current.RasterColor[0]);
92   UNCLAMPED_FLOAT_TO_CHAN(g, ctx->Current.RasterColor[1]);
93   UNCLAMPED_FLOAT_TO_CHAN(b, ctx->Current.RasterColor[2]);
94   UNCLAMPED_FLOAT_TO_CHAN(a, ctx->Current.RasterColor[3]);
95#if CHAN_TYPE == GL_FLOAT
96   span->red = r;
97   span->green = g;
98   span->blue = b;
99   span->alpha = a;
100#else
101   span->red   = IntToFixed(r);
102   span->green = IntToFixed(g);
103   span->blue  = IntToFixed(b);
104   span->alpha = IntToFixed(a);
105#endif
106   span->redStep = 0;
107   span->greenStep = 0;
108   span->blueStep = 0;
109   span->alphaStep = 0;
110   span->interpMask |= SPAN_RGBA;
111
112   COPY_4V(span->attrStart[VARYING_SLOT_COL0], ctx->Current.RasterColor);
113   ASSIGN_4V(span->attrStepX[VARYING_SLOT_COL0], 0.0, 0.0, 0.0, 0.0);
114   ASSIGN_4V(span->attrStepY[VARYING_SLOT_COL0], 0.0, 0.0, 0.0, 0.0);
115
116   /* Secondary color */
117   if (ctx->Light.Enabled || ctx->Fog.ColorSumEnabled)
118   {
119      COPY_4V(span->attrStart[VARYING_SLOT_COL1], ctx->Current.RasterSecondaryColor);
120      ASSIGN_4V(span->attrStepX[VARYING_SLOT_COL1], 0.0, 0.0, 0.0, 0.0);
121      ASSIGN_4V(span->attrStepY[VARYING_SLOT_COL1], 0.0, 0.0, 0.0, 0.0);
122   }
123
124   /* fog */
125   {
126      const SWcontext *swrast = SWRAST_CONTEXT(ctx);
127      GLfloat fogVal; /* a coord or a blend factor */
128      if (swrast->_PreferPixelFog) {
129         /* fog blend factors will be computed from fog coordinates per pixel */
130         fogVal = ctx->Current.RasterDistance;
131      }
132      else {
133         /* fog blend factor should be computed from fogcoord now */
134         fogVal = _swrast_z_to_fogfactor(ctx, ctx->Current.RasterDistance);
135      }
136      span->attrStart[VARYING_SLOT_FOGC][0] = fogVal;
137      span->attrStepX[VARYING_SLOT_FOGC][0] = 0.0;
138      span->attrStepY[VARYING_SLOT_FOGC][0] = 0.0;
139   }
140
141   /* texcoords */
142   {
143      GLuint i;
144      for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
145         const GLuint attr = VARYING_SLOT_TEX0 + i;
146         const GLfloat *tc = ctx->Current.RasterTexCoords[i];
147         if (_swrast_use_fragment_program(ctx) ||
148             _mesa_ati_fragment_shader_enabled(ctx)) {
149            COPY_4V(span->attrStart[attr], tc);
150         }
151         else if (tc[3] > 0.0F) {
152            /* use (s/q, t/q, r/q, 1) */
153            span->attrStart[attr][0] = tc[0] / tc[3];
154            span->attrStart[attr][1] = tc[1] / tc[3];
155            span->attrStart[attr][2] = tc[2] / tc[3];
156            span->attrStart[attr][3] = 1.0;
157         }
158         else {
159            ASSIGN_4V(span->attrStart[attr], 0.0F, 0.0F, 0.0F, 1.0F);
160         }
161         ASSIGN_4V(span->attrStepX[attr], 0.0F, 0.0F, 0.0F, 0.0F);
162         ASSIGN_4V(span->attrStepY[attr], 0.0F, 0.0F, 0.0F, 0.0F);
163      }
164   }
165}
166
167
168/**
169 * Interpolate the active attributes (and'd with attrMask) to
170 * fill in span->array->attribs[].
171 * Perspective correction will be done.  The point/line/triangle function
172 * should have computed attrStart/Step values for VARYING_SLOT_POS[3]!
173 */
174static inline void
175interpolate_active_attribs(struct gl_context *ctx, SWspan *span,
176                           GLbitfield64 attrMask)
177{
178   const SWcontext *swrast = SWRAST_CONTEXT(ctx);
179
180   /*
181    * Don't overwrite existing array values, such as colors that may have
182    * been produced by glDraw/CopyPixels.
183    */
184   attrMask &= ~span->arrayAttribs;
185
186   ATTRIB_LOOP_BEGIN
187      if (attrMask & BITFIELD64_BIT(attr)) {
188         const GLfloat dwdx = span->attrStepX[VARYING_SLOT_POS][3];
189         GLfloat w = span->attrStart[VARYING_SLOT_POS][3];
190         const GLfloat dv0dx = span->attrStepX[attr][0];
191         const GLfloat dv1dx = span->attrStepX[attr][1];
192         const GLfloat dv2dx = span->attrStepX[attr][2];
193         const GLfloat dv3dx = span->attrStepX[attr][3];
194         GLfloat v0 = span->attrStart[attr][0] + span->leftClip * dv0dx;
195         GLfloat v1 = span->attrStart[attr][1] + span->leftClip * dv1dx;
196         GLfloat v2 = span->attrStart[attr][2] + span->leftClip * dv2dx;
197         GLfloat v3 = span->attrStart[attr][3] + span->leftClip * dv3dx;
198         GLuint k;
199         for (k = 0; k < span->end; k++) {
200            const GLfloat invW = 1.0f / w;
201            span->array->attribs[attr][k][0] = v0 * invW;
202            span->array->attribs[attr][k][1] = v1 * invW;
203            span->array->attribs[attr][k][2] = v2 * invW;
204            span->array->attribs[attr][k][3] = v3 * invW;
205            v0 += dv0dx;
206            v1 += dv1dx;
207            v2 += dv2dx;
208            v3 += dv3dx;
209            w += dwdx;
210         }
211         assert((span->arrayAttribs & BITFIELD64_BIT(attr)) == 0);
212         span->arrayAttribs |= BITFIELD64_BIT(attr);
213      }
214   ATTRIB_LOOP_END
215}
216
217
218/**
219 * Interpolate primary colors to fill in the span->array->rgba8 (or rgb16)
220 * color array.
221 */
222static inline void
223interpolate_int_colors(struct gl_context *ctx, SWspan *span)
224{
225#if CHAN_BITS != 32
226   const GLuint n = span->end;
227   GLuint i;
228
229   assert(!(span->arrayMask & SPAN_RGBA));
230#endif
231
232   switch (span->array->ChanType) {
233#if CHAN_BITS != 32
234   case GL_UNSIGNED_BYTE:
235      {
236         GLubyte (*rgba)[4] = span->array->rgba8;
237         if (span->interpMask & SPAN_FLAT) {
238            GLubyte color[4];
239            color[RCOMP] = FixedToInt(span->red);
240            color[GCOMP] = FixedToInt(span->green);
241            color[BCOMP] = FixedToInt(span->blue);
242            color[ACOMP] = FixedToInt(span->alpha);
243            for (i = 0; i < n; i++) {
244               COPY_4UBV(rgba[i], color);
245            }
246         }
247         else {
248            GLfixed r = span->red;
249            GLfixed g = span->green;
250            GLfixed b = span->blue;
251            GLfixed a = span->alpha;
252            GLint dr = span->redStep;
253            GLint dg = span->greenStep;
254            GLint db = span->blueStep;
255            GLint da = span->alphaStep;
256            for (i = 0; i < n; i++) {
257               rgba[i][RCOMP] = FixedToChan(r);
258               rgba[i][GCOMP] = FixedToChan(g);
259               rgba[i][BCOMP] = FixedToChan(b);
260               rgba[i][ACOMP] = FixedToChan(a);
261               r += dr;
262               g += dg;
263               b += db;
264               a += da;
265            }
266         }
267      }
268      break;
269   case GL_UNSIGNED_SHORT:
270      {
271         GLushort (*rgba)[4] = span->array->rgba16;
272         if (span->interpMask & SPAN_FLAT) {
273            GLushort color[4];
274            color[RCOMP] = FixedToInt(span->red);
275            color[GCOMP] = FixedToInt(span->green);
276            color[BCOMP] = FixedToInt(span->blue);
277            color[ACOMP] = FixedToInt(span->alpha);
278            for (i = 0; i < n; i++) {
279               COPY_4V(rgba[i], color);
280            }
281         }
282         else {
283            GLushort (*rgba)[4] = span->array->rgba16;
284            GLfixed r, g, b, a;
285            GLint dr, dg, db, da;
286            r = span->red;
287            g = span->green;
288            b = span->blue;
289            a = span->alpha;
290            dr = span->redStep;
291            dg = span->greenStep;
292            db = span->blueStep;
293            da = span->alphaStep;
294            for (i = 0; i < n; i++) {
295               rgba[i][RCOMP] = FixedToChan(r);
296               rgba[i][GCOMP] = FixedToChan(g);
297               rgba[i][BCOMP] = FixedToChan(b);
298               rgba[i][ACOMP] = FixedToChan(a);
299               r += dr;
300               g += dg;
301               b += db;
302               a += da;
303            }
304         }
305      }
306      break;
307#endif
308   case GL_FLOAT:
309      interpolate_active_attribs(ctx, span, VARYING_BIT_COL0);
310      break;
311   default:
312      _mesa_problem(ctx, "bad datatype 0x%x in interpolate_int_colors",
313                    span->array->ChanType);
314   }
315   span->arrayMask |= SPAN_RGBA;
316}
317
318
319/**
320 * Populate the VARYING_SLOT_COL0 array.
321 */
322static inline void
323interpolate_float_colors(SWspan *span)
324{
325   GLfloat (*col0)[4] = span->array->attribs[VARYING_SLOT_COL0];
326   const GLuint n = span->end;
327   GLuint i;
328
329   assert(!(span->arrayAttribs & VARYING_BIT_COL0));
330
331   if (span->arrayMask & SPAN_RGBA) {
332      /* convert array of int colors */
333      for (i = 0; i < n; i++) {
334         col0[i][0] = UBYTE_TO_FLOAT(span->array->rgba8[i][0]);
335         col0[i][1] = UBYTE_TO_FLOAT(span->array->rgba8[i][1]);
336         col0[i][2] = UBYTE_TO_FLOAT(span->array->rgba8[i][2]);
337         col0[i][3] = UBYTE_TO_FLOAT(span->array->rgba8[i][3]);
338      }
339   }
340   else {
341      /* interpolate red/green/blue/alpha to get float colors */
342      assert(span->interpMask & SPAN_RGBA);
343      if (span->interpMask & SPAN_FLAT) {
344         GLfloat r = FixedToFloat(span->red);
345         GLfloat g = FixedToFloat(span->green);
346         GLfloat b = FixedToFloat(span->blue);
347         GLfloat a = FixedToFloat(span->alpha);
348         for (i = 0; i < n; i++) {
349            ASSIGN_4V(col0[i], r, g, b, a);
350         }
351      }
352      else {
353         GLfloat r = FixedToFloat(span->red);
354         GLfloat g = FixedToFloat(span->green);
355         GLfloat b = FixedToFloat(span->blue);
356         GLfloat a = FixedToFloat(span->alpha);
357         GLfloat dr = FixedToFloat(span->redStep);
358         GLfloat dg = FixedToFloat(span->greenStep);
359         GLfloat db = FixedToFloat(span->blueStep);
360         GLfloat da = FixedToFloat(span->alphaStep);
361         for (i = 0; i < n; i++) {
362            col0[i][0] = r;
363            col0[i][1] = g;
364            col0[i][2] = b;
365            col0[i][3] = a;
366            r += dr;
367            g += dg;
368            b += db;
369            a += da;
370         }
371      }
372   }
373
374   span->arrayAttribs |= VARYING_BIT_COL0;
375   span->array->ChanType = GL_FLOAT;
376}
377
378
379
380/**
381 * Fill in the span.zArray array from the span->z, zStep values.
382 */
383void
384_swrast_span_interpolate_z( const struct gl_context *ctx, SWspan *span )
385{
386   const GLuint n = span->end;
387   GLuint i;
388
389   assert(!(span->arrayMask & SPAN_Z));
390
391   if (ctx->DrawBuffer->Visual.depthBits <= 16) {
392      GLfixed zval = span->z;
393      GLuint *z = span->array->z;
394      for (i = 0; i < n; i++) {
395         z[i] = FixedToInt(zval);
396         zval += span->zStep;
397      }
398   }
399   else {
400      /* Deep Z buffer, no fixed->int shift */
401      GLuint zval = span->z;
402      GLuint *z = span->array->z;
403      for (i = 0; i < n; i++) {
404         z[i] = zval;
405         zval += span->zStep;
406      }
407   }
408   span->interpMask &= ~SPAN_Z;
409   span->arrayMask |= SPAN_Z;
410}
411
412
413/**
414 * Compute mipmap LOD from partial derivatives.
415 * This the ideal solution, as given in the OpenGL spec.
416 */
417GLfloat
418_swrast_compute_lambda(GLfloat dsdx, GLfloat dsdy, GLfloat dtdx, GLfloat dtdy,
419                       GLfloat dqdx, GLfloat dqdy, GLfloat texW, GLfloat texH,
420                       GLfloat s, GLfloat t, GLfloat q, GLfloat invQ)
421{
422   GLfloat dudx = texW * ((s + dsdx) / (q + dqdx) - s * invQ);
423   GLfloat dvdx = texH * ((t + dtdx) / (q + dqdx) - t * invQ);
424   GLfloat dudy = texW * ((s + dsdy) / (q + dqdy) - s * invQ);
425   GLfloat dvdy = texH * ((t + dtdy) / (q + dqdy) - t * invQ);
426   GLfloat x = sqrtf(dudx * dudx + dvdx * dvdx);
427   GLfloat y = sqrtf(dudy * dudy + dvdy * dvdy);
428   GLfloat rho = MAX2(x, y);
429   GLfloat lambda = log2f(rho);
430   return lambda;
431}
432
433
434/**
435 * Compute mipmap LOD from partial derivatives.
436 * This is a faster approximation than above function.
437 */
438#if 0
439GLfloat
440_swrast_compute_lambda(GLfloat dsdx, GLfloat dsdy, GLfloat dtdx, GLfloat dtdy,
441                     GLfloat dqdx, GLfloat dqdy, GLfloat texW, GLfloat texH,
442                     GLfloat s, GLfloat t, GLfloat q, GLfloat invQ)
443{
444   GLfloat dsdx2 = (s + dsdx) / (q + dqdx) - s * invQ;
445   GLfloat dtdx2 = (t + dtdx) / (q + dqdx) - t * invQ;
446   GLfloat dsdy2 = (s + dsdy) / (q + dqdy) - s * invQ;
447   GLfloat dtdy2 = (t + dtdy) / (q + dqdy) - t * invQ;
448   GLfloat maxU, maxV, rho, lambda;
449   dsdx2 = fabsf(dsdx2);
450   dsdy2 = fabsf(dsdy2);
451   dtdx2 = fabsf(dtdx2);
452   dtdy2 = fabsf(dtdy2);
453   maxU = MAX2(dsdx2, dsdy2) * texW;
454   maxV = MAX2(dtdx2, dtdy2) * texH;
455   rho = MAX2(maxU, maxV);
456   lambda = logf2(rho);
457   return lambda;
458}
459#endif
460
461
462/**
463 * Fill in the span.array->attrib[VARYING_SLOT_TEXn] arrays from the
464 * using the attrStart/Step values.
465 *
466 * This function only used during fixed-function fragment processing.
467 *
468 * Note: in the places where we divide by Q (or mult by invQ) we're
469 * really doing two things: perspective correction and texcoord
470 * projection.  Remember, for texcoord (s,t,r,q) we need to index
471 * texels with (s/q, t/q, r/q).
472 */
473static void
474interpolate_texcoords(struct gl_context *ctx, SWspan *span)
475{
476   const GLuint maxUnit
477      = (ctx->Texture._EnabledCoordUnits > 1) ? ctx->Const.MaxTextureUnits : 1;
478   GLuint u;
479
480   /* XXX CoordUnits vs. ImageUnits */
481   for (u = 0; u < maxUnit; u++) {
482      if (ctx->Texture._EnabledCoordUnits & (1 << u)) {
483         const GLuint attr = VARYING_SLOT_TEX0 + u;
484         const struct gl_texture_object *obj = ctx->Texture.Unit[u]._Current;
485         GLfloat texW, texH;
486         GLboolean needLambda;
487         GLfloat (*texcoord)[4] = span->array->attribs[attr];
488         GLfloat *lambda = span->array->lambda[u];
489         const GLfloat dsdx = span->attrStepX[attr][0];
490         const GLfloat dsdy = span->attrStepY[attr][0];
491         const GLfloat dtdx = span->attrStepX[attr][1];
492         const GLfloat dtdy = span->attrStepY[attr][1];
493         const GLfloat drdx = span->attrStepX[attr][2];
494         const GLfloat dqdx = span->attrStepX[attr][3];
495         const GLfloat dqdy = span->attrStepY[attr][3];
496         GLfloat s = span->attrStart[attr][0] + span->leftClip * dsdx;
497         GLfloat t = span->attrStart[attr][1] + span->leftClip * dtdx;
498         GLfloat r = span->attrStart[attr][2] + span->leftClip * drdx;
499         GLfloat q = span->attrStart[attr][3] + span->leftClip * dqdx;
500
501         if (obj) {
502            const struct gl_texture_image *img = _mesa_base_tex_image(obj);
503            const struct swrast_texture_image *swImg =
504               swrast_texture_image_const(img);
505            const struct gl_sampler_object *samp = _mesa_get_samplerobj(ctx, u);
506
507            needLambda = (samp->Attrib.MinFilter != samp->Attrib.MagFilter)
508               || _swrast_use_fragment_program(ctx);
509            /* LOD is calculated directly in the ansiotropic filter, we can
510             * skip the normal lambda function as the result is ignored.
511             */
512            if (samp->Attrib.MaxAnisotropy > 1.0F &&
513                samp->Attrib.MinFilter == GL_LINEAR_MIPMAP_LINEAR) {
514               needLambda = GL_FALSE;
515            }
516            texW = swImg->WidthScale;
517            texH = swImg->HeightScale;
518         }
519         else {
520            /* using a fragment program */
521            texW = 1.0;
522            texH = 1.0;
523            needLambda = GL_FALSE;
524         }
525
526         if (needLambda) {
527            GLuint i;
528            if (_swrast_use_fragment_program(ctx)
529                || _mesa_ati_fragment_shader_enabled(ctx)) {
530               /* do perspective correction but don't divide s, t, r by q */
531               const GLfloat dwdx = span->attrStepX[VARYING_SLOT_POS][3];
532               GLfloat w = span->attrStart[VARYING_SLOT_POS][3] + span->leftClip * dwdx;
533               for (i = 0; i < span->end; i++) {
534                  const GLfloat invW = 1.0F / w;
535                  texcoord[i][0] = s * invW;
536                  texcoord[i][1] = t * invW;
537                  texcoord[i][2] = r * invW;
538                  texcoord[i][3] = q * invW;
539                  lambda[i] = _swrast_compute_lambda(dsdx, dsdy, dtdx, dtdy,
540                                                     dqdx, dqdy, texW, texH,
541                                                     s, t, q, invW);
542                  s += dsdx;
543                  t += dtdx;
544                  r += drdx;
545                  q += dqdx;
546                  w += dwdx;
547               }
548            }
549            else {
550               for (i = 0; i < span->end; i++) {
551                  const GLfloat invQ = (q == 0.0F) ? 1.0F : (1.0F / q);
552                  texcoord[i][0] = s * invQ;
553                  texcoord[i][1] = t * invQ;
554                  texcoord[i][2] = r * invQ;
555                  texcoord[i][3] = q;
556                  lambda[i] = _swrast_compute_lambda(dsdx, dsdy, dtdx, dtdy,
557                                                     dqdx, dqdy, texW, texH,
558                                                     s, t, q, invQ);
559                  s += dsdx;
560                  t += dtdx;
561                  r += drdx;
562                  q += dqdx;
563               }
564            }
565            span->arrayMask |= SPAN_LAMBDA;
566         }
567         else {
568            GLuint i;
569            if (_swrast_use_fragment_program(ctx) ||
570                _mesa_ati_fragment_shader_enabled(ctx)) {
571               /* do perspective correction but don't divide s, t, r by q */
572               const GLfloat dwdx = span->attrStepX[VARYING_SLOT_POS][3];
573               GLfloat w = span->attrStart[VARYING_SLOT_POS][3] + span->leftClip * dwdx;
574               for (i = 0; i < span->end; i++) {
575                  const GLfloat invW = 1.0F / w;
576                  texcoord[i][0] = s * invW;
577                  texcoord[i][1] = t * invW;
578                  texcoord[i][2] = r * invW;
579                  texcoord[i][3] = q * invW;
580                  lambda[i] = 0.0;
581                  s += dsdx;
582                  t += dtdx;
583                  r += drdx;
584                  q += dqdx;
585                  w += dwdx;
586               }
587            }
588            else if (dqdx == 0.0F) {
589               /* Ortho projection or polygon's parallel to window X axis */
590               const GLfloat invQ = (q == 0.0F) ? 1.0F : (1.0F / q);
591               for (i = 0; i < span->end; i++) {
592                  texcoord[i][0] = s * invQ;
593                  texcoord[i][1] = t * invQ;
594                  texcoord[i][2] = r * invQ;
595                  texcoord[i][3] = q;
596                  lambda[i] = 0.0;
597                  s += dsdx;
598                  t += dtdx;
599                  r += drdx;
600               }
601            }
602            else {
603               for (i = 0; i < span->end; i++) {
604                  const GLfloat invQ = (q == 0.0F) ? 1.0F : (1.0F / q);
605                  texcoord[i][0] = s * invQ;
606                  texcoord[i][1] = t * invQ;
607                  texcoord[i][2] = r * invQ;
608                  texcoord[i][3] = q;
609                  lambda[i] = 0.0;
610                  s += dsdx;
611                  t += dtdx;
612                  r += drdx;
613                  q += dqdx;
614               }
615            }
616         } /* lambda */
617      } /* if */
618   } /* for */
619}
620
621
622/**
623 * Fill in the arrays->attribs[VARYING_SLOT_POS] array.
624 */
625static inline void
626interpolate_wpos(struct gl_context *ctx, SWspan *span)
627{
628   GLfloat (*wpos)[4] = span->array->attribs[VARYING_SLOT_POS];
629   GLuint i;
630   const GLfloat zScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
631   GLfloat w, dw;
632
633   if (span->arrayMask & SPAN_XY) {
634      for (i = 0; i < span->end; i++) {
635         wpos[i][0] = (GLfloat) span->array->x[i];
636         wpos[i][1] = (GLfloat) span->array->y[i];
637      }
638   }
639   else {
640      for (i = 0; i < span->end; i++) {
641         wpos[i][0] = (GLfloat) span->x + i;
642         wpos[i][1] = (GLfloat) span->y;
643      }
644   }
645
646   dw = span->attrStepX[VARYING_SLOT_POS][3];
647   w = span->attrStart[VARYING_SLOT_POS][3] + span->leftClip * dw;
648   for (i = 0; i < span->end; i++) {
649      wpos[i][2] = (GLfloat) span->array->z[i] * zScale;
650      wpos[i][3] = w;
651      w += dw;
652   }
653}
654
655
656/**
657 * Apply the current polygon stipple pattern to a span of pixels.
658 */
659static inline void
660stipple_polygon_span(struct gl_context *ctx, SWspan *span)
661{
662   GLubyte *mask = span->array->mask;
663
664   assert(ctx->Polygon.StippleFlag);
665
666   if (span->arrayMask & SPAN_XY) {
667      /* arrays of x/y pixel coords */
668      GLuint i;
669      for (i = 0; i < span->end; i++) {
670         const GLint col = span->array->x[i] % 32;
671         const GLint row = span->array->y[i] % 32;
672         const GLuint stipple = ctx->PolygonStipple[row];
673         if (((1 << col) & stipple) == 0) {
674            mask[i] = 0;
675         }
676      }
677   }
678   else {
679      /* horizontal span of pixels */
680      const GLuint highBit = 1 << 31;
681      const GLuint stipple = ctx->PolygonStipple[span->y % 32];
682      GLuint i, m = highBit >> (GLuint) (span->x % 32);
683      for (i = 0; i < span->end; i++) {
684         if ((m & stipple) == 0) {
685            mask[i] = 0;
686         }
687         m = m >> 1;
688         if (m == 0) {
689            m = highBit;
690         }
691      }
692   }
693   span->writeAll = GL_FALSE;
694}
695
696
697/**
698 * Clip a pixel span to the current buffer/window boundaries:
699 * DrawBuffer->_Xmin, _Xmax, _Ymin, _Ymax.  This will accomplish
700 * window clipping and scissoring.
701 * Return:   GL_TRUE   some pixels still visible
702 *           GL_FALSE  nothing visible
703 */
704static inline GLuint
705clip_span( struct gl_context *ctx, SWspan *span )
706{
707   const GLint xmin = ctx->DrawBuffer->_Xmin;
708   const GLint xmax = ctx->DrawBuffer->_Xmax;
709   const GLint ymin = ctx->DrawBuffer->_Ymin;
710   const GLint ymax = ctx->DrawBuffer->_Ymax;
711
712   span->leftClip = 0;
713
714   if (span->arrayMask & SPAN_XY) {
715      /* arrays of x/y pixel coords */
716      const GLint *x = span->array->x;
717      const GLint *y = span->array->y;
718      const GLint n = span->end;
719      GLubyte *mask = span->array->mask;
720      GLint i;
721      GLuint passed = 0;
722      if (span->arrayMask & SPAN_MASK) {
723         /* note: using & intead of && to reduce branches */
724         for (i = 0; i < n; i++) {
725            mask[i] &= (x[i] >= xmin) & (x[i] < xmax)
726                     & (y[i] >= ymin) & (y[i] < ymax);
727            passed += mask[i];
728         }
729      }
730      else {
731         /* note: using & intead of && to reduce branches */
732         for (i = 0; i < n; i++) {
733            mask[i] = (x[i] >= xmin) & (x[i] < xmax)
734                    & (y[i] >= ymin) & (y[i] < ymax);
735            passed += mask[i];
736         }
737      }
738      return passed > 0;
739   }
740   else {
741      /* horizontal span of pixels */
742      const GLint x = span->x;
743      const GLint y = span->y;
744      GLint n = span->end;
745
746      /* Trivial rejection tests */
747      if (y < ymin || y >= ymax || x + n <= xmin || x >= xmax) {
748         span->end = 0;
749         return GL_FALSE;  /* all pixels clipped */
750      }
751
752      /* Clip to right */
753      if (x + n > xmax) {
754         assert(x < xmax);
755         n = span->end = xmax - x;
756      }
757
758      /* Clip to the left */
759      if (x < xmin) {
760         const GLint leftClip = xmin - x;
761         GLuint i;
762
763         assert(leftClip > 0);
764         assert(x + n > xmin);
765
766         /* Clip 'leftClip' pixels from the left side.
767          * The span->leftClip field will be applied when we interpolate
768          * fragment attributes.
769          * For arrays of values, shift them left.
770          */
771         for (i = 0; i < VARYING_SLOT_MAX; i++) {
772            if (span->interpMask & (1u << i)) {
773               GLuint j;
774               for (j = 0; j < 4; j++) {
775                  span->attrStart[i][j] += leftClip * span->attrStepX[i][j];
776               }
777            }
778         }
779
780         span->red += leftClip * span->redStep;
781         span->green += leftClip * span->greenStep;
782         span->blue += leftClip * span->blueStep;
783         span->alpha += leftClip * span->alphaStep;
784         span->index += leftClip * span->indexStep;
785         span->z += leftClip * span->zStep;
786         span->intTex[0] += leftClip * span->intTexStep[0];
787         span->intTex[1] += leftClip * span->intTexStep[1];
788
789#define SHIFT_ARRAY(ARRAY, SHIFT, LEN) \
790         memmove(ARRAY, ARRAY + (SHIFT), (LEN) * sizeof(ARRAY[0]))
791
792         for (i = 0; i < VARYING_SLOT_MAX; i++) {
793            if (span->arrayAttribs & BITFIELD64_BIT(i)) {
794               /* shift array elements left by 'leftClip' */
795               SHIFT_ARRAY(span->array->attribs[i], leftClip, n - leftClip);
796            }
797         }
798
799         SHIFT_ARRAY(span->array->mask, leftClip, n - leftClip);
800         SHIFT_ARRAY(span->array->rgba8, leftClip, n - leftClip);
801         SHIFT_ARRAY(span->array->rgba16, leftClip, n - leftClip);
802         SHIFT_ARRAY(span->array->x, leftClip, n - leftClip);
803         SHIFT_ARRAY(span->array->y, leftClip, n - leftClip);
804         SHIFT_ARRAY(span->array->z, leftClip, n - leftClip);
805         SHIFT_ARRAY(span->array->index, leftClip, n - leftClip);
806         for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
807            SHIFT_ARRAY(span->array->lambda[i], leftClip, n - leftClip);
808         }
809         SHIFT_ARRAY(span->array->coverage, leftClip, n - leftClip);
810
811#undef SHIFT_ARRAY
812
813         span->leftClip = leftClip;
814         span->x = xmin;
815         span->end -= leftClip;
816         span->writeAll = GL_FALSE;
817      }
818
819      assert(span->x >= xmin);
820      assert(span->x + span->end <= xmax);
821      assert(span->y >= ymin);
822      assert(span->y < ymax);
823
824      return GL_TRUE;  /* some pixels visible */
825   }
826}
827
828
829/**
830 * Add specular colors to primary colors.
831 * Only called during fixed-function operation.
832 * Result is float color array (VARYING_SLOT_COL0).
833 */
834static inline void
835add_specular(struct gl_context *ctx, SWspan *span)
836{
837   const SWcontext *swrast = SWRAST_CONTEXT(ctx);
838   const GLubyte *mask = span->array->mask;
839   GLfloat (*col0)[4] = span->array->attribs[VARYING_SLOT_COL0];
840   GLfloat (*col1)[4] = span->array->attribs[VARYING_SLOT_COL1];
841   GLuint i;
842
843   assert(!_swrast_use_fragment_program(ctx));
844   assert(span->arrayMask & SPAN_RGBA);
845   assert(swrast->_ActiveAttribMask & VARYING_BIT_COL1);
846   (void) swrast; /* silence warning */
847
848   if (span->array->ChanType == GL_FLOAT) {
849      if ((span->arrayAttribs & VARYING_BIT_COL0) == 0) {
850         interpolate_active_attribs(ctx, span, VARYING_BIT_COL0);
851      }
852   }
853   else {
854      /* need float colors */
855      if ((span->arrayAttribs & VARYING_BIT_COL0) == 0) {
856         interpolate_float_colors(span);
857      }
858   }
859
860   if ((span->arrayAttribs & VARYING_BIT_COL1) == 0) {
861      /* XXX could avoid this and interpolate COL1 in the loop below */
862      interpolate_active_attribs(ctx, span, VARYING_BIT_COL1);
863   }
864
865   assert(span->arrayAttribs & VARYING_BIT_COL0);
866   assert(span->arrayAttribs & VARYING_BIT_COL1);
867
868   for (i = 0; i < span->end; i++) {
869      if (mask[i]) {
870         col0[i][0] += col1[i][0];
871         col0[i][1] += col1[i][1];
872         col0[i][2] += col1[i][2];
873      }
874   }
875
876   span->array->ChanType = GL_FLOAT;
877}
878
879
880/**
881 * Apply antialiasing coverage value to alpha values.
882 */
883static inline void
884apply_aa_coverage(SWspan *span)
885{
886   const GLfloat *coverage = span->array->coverage;
887   GLuint i;
888   if (span->array->ChanType == GL_UNSIGNED_BYTE) {
889      GLubyte (*rgba)[4] = span->array->rgba8;
890      for (i = 0; i < span->end; i++) {
891         const GLfloat a = rgba[i][ACOMP] * coverage[i];
892         rgba[i][ACOMP] = (GLubyte) CLAMP(a, 0.0F, 255.0F);
893         assert(coverage[i] >= 0.0F);
894         assert(coverage[i] <= 1.0F);
895      }
896   }
897   else if (span->array->ChanType == GL_UNSIGNED_SHORT) {
898      GLushort (*rgba)[4] = span->array->rgba16;
899      for (i = 0; i < span->end; i++) {
900         const GLfloat a = rgba[i][ACOMP] * coverage[i];
901         rgba[i][ACOMP] = (GLushort) CLAMP(a, 0.0F, 65535.0F);
902      }
903   }
904   else {
905      GLfloat (*rgba)[4] = span->array->attribs[VARYING_SLOT_COL0];
906      for (i = 0; i < span->end; i++) {
907         rgba[i][ACOMP] = rgba[i][ACOMP] * coverage[i];
908         /* clamp later */
909      }
910   }
911}
912
913
914/**
915 * Clamp span's float colors to [0,1]
916 */
917static inline void
918clamp_colors(SWspan *span)
919{
920   GLfloat (*rgba)[4] = span->array->attribs[VARYING_SLOT_COL0];
921   GLuint i;
922   assert(span->array->ChanType == GL_FLOAT);
923   for (i = 0; i < span->end; i++) {
924      rgba[i][RCOMP] = CLAMP(rgba[i][RCOMP], 0.0F, 1.0F);
925      rgba[i][GCOMP] = CLAMP(rgba[i][GCOMP], 0.0F, 1.0F);
926      rgba[i][BCOMP] = CLAMP(rgba[i][BCOMP], 0.0F, 1.0F);
927      rgba[i][ACOMP] = CLAMP(rgba[i][ACOMP], 0.0F, 1.0F);
928   }
929}
930
931
932/**
933 * Convert the span's color arrays to the given type.
934 * The only way 'output' can be greater than zero is when we have a fragment
935 * program that writes to gl_FragData[1] or higher.
936 * \param output  which fragment program color output is being processed
937 */
938static inline void
939convert_color_type(SWspan *span, GLenum srcType, GLenum newType, GLuint output)
940{
941   GLvoid *src, *dst;
942
943   if (output > 0 || srcType == GL_FLOAT) {
944      src = span->array->attribs[VARYING_SLOT_COL0 + output];
945      span->array->ChanType = GL_FLOAT;
946   }
947   else if (srcType == GL_UNSIGNED_BYTE) {
948      src = span->array->rgba8;
949   }
950   else {
951      assert(srcType == GL_UNSIGNED_SHORT);
952      src = span->array->rgba16;
953   }
954
955   if (newType == GL_UNSIGNED_BYTE) {
956      dst = span->array->rgba8;
957   }
958   else if (newType == GL_UNSIGNED_SHORT) {
959      dst = span->array->rgba16;
960   }
961   else {
962      dst = span->array->attribs[VARYING_SLOT_COL0];
963   }
964
965   _mesa_convert_colors(span->array->ChanType, src,
966                        newType, dst,
967                        span->end, span->array->mask);
968
969   span->array->ChanType = newType;
970   span->array->rgba = dst;
971}
972
973
974
975/**
976 * Apply fragment shader, fragment program or normal texturing to span.
977 */
978static inline void
979shade_texture_span(struct gl_context *ctx, SWspan *span)
980{
981   if (_swrast_use_fragment_program(ctx) ||
982       _mesa_ati_fragment_shader_enabled(ctx)) {
983      /* programmable shading */
984      if (span->primitive == GL_BITMAP && span->array->ChanType != GL_FLOAT) {
985         convert_color_type(span, span->array->ChanType, GL_FLOAT, 0);
986      }
987      else {
988         span->array->rgba = (void *) span->array->attribs[VARYING_SLOT_COL0];
989      }
990
991      if (span->primitive != GL_POINT ||
992	  (span->interpMask & SPAN_RGBA) ||
993	  ctx->Point.PointSprite) {
994         /* for single-pixel points, we populated the arrays already */
995         interpolate_active_attribs(ctx, span, ~0);
996      }
997      span->array->ChanType = GL_FLOAT;
998
999      if (!(span->arrayMask & SPAN_Z))
1000         _swrast_span_interpolate_z (ctx, span);
1001
1002#if 0
1003      if (inputsRead & VARYING_BIT_POS)
1004#else
1005      /* XXX always interpolate wpos so that DDX/DDY work */
1006#endif
1007         interpolate_wpos(ctx, span);
1008
1009      /* Run fragment program/shader now */
1010      if (_swrast_use_fragment_program(ctx)) {
1011         _swrast_exec_fragment_program(ctx, span);
1012      }
1013      else {
1014         assert(_mesa_ati_fragment_shader_enabled(ctx));
1015         _swrast_exec_fragment_shader(ctx, span);
1016      }
1017   }
1018   else if (ctx->Texture._EnabledCoordUnits) {
1019      /* conventional texturing */
1020
1021#if CHAN_BITS == 32
1022      if ((span->arrayAttribs & VARYING_BIT_COL0) == 0) {
1023         interpolate_int_colors(ctx, span);
1024      }
1025#else
1026      if (!(span->arrayMask & SPAN_RGBA))
1027         interpolate_int_colors(ctx, span);
1028#endif
1029      if ((span->arrayAttribs & VARYING_BITS_TEX_ANY) == 0x0)
1030         interpolate_texcoords(ctx, span);
1031
1032      _swrast_texture_span(ctx, span);
1033   }
1034}
1035
1036
1037/** Put colors at x/y locations into a renderbuffer */
1038static void
1039put_values(struct gl_context *ctx, struct gl_renderbuffer *rb,
1040           GLenum datatype,
1041           GLuint count, const GLint x[], const GLint y[],
1042           const void *values, const GLubyte *mask)
1043{
1044   struct swrast_renderbuffer *srb = swrast_renderbuffer(rb);
1045   GLuint i;
1046
1047   for (i = 0; i < count; i++) {
1048      if (mask[i]) {
1049         if (datatype == GL_UNSIGNED_BYTE) {
1050            util_format_write_4ub(rb->Format,
1051                                  (uint8_t *)values + 4 * i, 0,
1052                                  srb->Map, srb->RowStride,
1053                                  x[i], y[i], 1, 1);
1054         }
1055         else {
1056            assert(datatype == GL_FLOAT);
1057            util_format_write_4(rb->Format,
1058                                (float *)values + 4 * i, 0,
1059                                srb->Map, srb->RowStride,
1060                                x[i], y[i], 1, 1);
1061         }
1062      }
1063   }
1064}
1065
1066
1067/** Put row of colors into renderbuffer */
1068void
1069_swrast_put_row(struct gl_context *ctx, struct gl_renderbuffer *rb,
1070                GLenum datatype,
1071                GLuint count, GLint x, GLint y,
1072                const void *values, const GLubyte *mask)
1073{
1074   GLubyte *dst = _swrast_pixel_address(rb, x, y);
1075
1076   if (!mask) {
1077      if (datatype == GL_UNSIGNED_BYTE) {
1078         _mesa_pack_ubyte_rgba_row(rb->Format, count, values, dst);
1079      }
1080      else {
1081         assert(datatype == GL_FLOAT);
1082         _mesa_pack_float_rgba_row(rb->Format, count,
1083                                   (const GLfloat (*)[4]) values, dst);
1084      }
1085   }
1086   else {
1087      const GLuint bpp = _mesa_get_format_bytes(rb->Format);
1088      GLuint i, runLen, runStart;
1089      /* We can't pass a 'mask' array to the _mesa_pack_rgba_row() functions
1090       * so look for runs where mask=1...
1091       */
1092      runLen = runStart = 0;
1093      for (i = 0; i < count; i++) {
1094         if (mask[i]) {
1095            if (runLen == 0)
1096               runStart = i;
1097            runLen++;
1098         }
1099
1100         if (!mask[i] || i == count - 1) {
1101            /* might be the end of a run of pixels */
1102            if (runLen > 0) {
1103               if (datatype == GL_UNSIGNED_BYTE) {
1104                  _mesa_pack_ubyte_rgba_row(rb->Format, runLen,
1105                                     (uint8_t *)values + runStart,
1106                                     dst + runStart * bpp);
1107               }
1108               else {
1109                  assert(datatype == GL_FLOAT);
1110                  _mesa_pack_float_rgba_row(rb->Format, runLen,
1111                                   (const GLfloat (*)[4]) values + runStart,
1112                                   dst + runStart * bpp);
1113               }
1114               runLen = 0;
1115            }
1116         }
1117      }
1118   }
1119}
1120
1121
1122
1123/**
1124 * Apply all the per-fragment operations to a span.
1125 * This now includes texturing (_swrast_write_texture_span() is history).
1126 * This function may modify any of the array values in the span.
1127 * span->interpMask and span->arrayMask may be changed but will be restored
1128 * to their original values before returning.
1129 */
1130void
1131_swrast_write_rgba_span( struct gl_context *ctx, SWspan *span)
1132{
1133   const SWcontext *swrast = SWRAST_CONTEXT(ctx);
1134   const GLbitfield origInterpMask = span->interpMask;
1135   const GLbitfield origArrayMask = span->arrayMask;
1136   const GLbitfield64 origArrayAttribs = span->arrayAttribs;
1137   const GLenum origChanType = span->array->ChanType;
1138   void * const origRgba = span->array->rgba;
1139   const GLboolean shader = (_swrast_use_fragment_program(ctx)
1140                             || _mesa_ati_fragment_shader_enabled(ctx));
1141   const GLboolean shaderOrTexture = shader || ctx->Texture._EnabledCoordUnits;
1142   struct gl_framebuffer *fb = ctx->DrawBuffer;
1143
1144   /*
1145   printf("%s()  interp 0x%x  array 0x%x\n", __func__,
1146          span->interpMask, span->arrayMask);
1147   */
1148
1149   assert(span->primitive == GL_POINT ||
1150          span->primitive == GL_LINE ||
1151	  span->primitive == GL_POLYGON ||
1152          span->primitive == GL_BITMAP);
1153
1154   /* Fragment write masks */
1155   if (span->arrayMask & SPAN_MASK) {
1156      /* mask was initialized by caller, probably glBitmap */
1157      span->writeAll = GL_FALSE;
1158   }
1159   else {
1160      memset(span->array->mask, 1, span->end);
1161      span->writeAll = GL_TRUE;
1162   }
1163
1164   /* Clip to window/scissor box */
1165   if (!clip_span(ctx, span)) {
1166      return;
1167   }
1168
1169   assert(span->end <= SWRAST_MAX_WIDTH);
1170
1171   /* Depth bounds test */
1172   if (ctx->Depth.BoundsTest && fb->Visual.depthBits > 0) {
1173      if (!_swrast_depth_bounds_test(ctx, span)) {
1174         return;
1175      }
1176   }
1177
1178#ifdef DEBUG
1179   /* Make sure all fragments are within window bounds */
1180   if (span->arrayMask & SPAN_XY) {
1181      /* array of pixel locations */
1182      GLuint i;
1183      for (i = 0; i < span->end; i++) {
1184         if (span->array->mask[i]) {
1185            assert(span->array->x[i] >= fb->_Xmin);
1186            assert(span->array->x[i] < fb->_Xmax);
1187            assert(span->array->y[i] >= fb->_Ymin);
1188            assert(span->array->y[i] < fb->_Ymax);
1189         }
1190      }
1191   }
1192#endif
1193
1194   /* Polygon Stippling */
1195   if (ctx->Polygon.StippleFlag && span->primitive == GL_POLYGON) {
1196      stipple_polygon_span(ctx, span);
1197   }
1198
1199   /* This is the normal place to compute the fragment color/Z
1200    * from texturing or shading.
1201    */
1202   if (shaderOrTexture && !swrast->_DeferredTexture) {
1203      shade_texture_span(ctx, span);
1204   }
1205
1206   /* Do the alpha test */
1207   if (ctx->Color.AlphaEnabled) {
1208      if (!_swrast_alpha_test(ctx, span)) {
1209         /* all fragments failed test */
1210         goto end;
1211      }
1212   }
1213
1214   /* Stencil and Z testing */
1215   if (_mesa_stencil_is_enabled(ctx) || ctx->Depth.Test) {
1216      if (!(span->arrayMask & SPAN_Z))
1217         _swrast_span_interpolate_z(ctx, span);
1218
1219      if (ctx->Transform.DepthClampNear && ctx->Transform.DepthClampFar)
1220	 _swrast_depth_clamp_span(ctx, span);
1221
1222      if (_mesa_stencil_is_enabled(ctx)) {
1223         /* Combined Z/stencil tests */
1224         if (!_swrast_stencil_and_ztest_span(ctx, span)) {
1225            /* all fragments failed test */
1226            goto end;
1227         }
1228      }
1229      else if (fb->Visual.depthBits > 0) {
1230         /* Just regular depth testing */
1231         assert(ctx->Depth.Test);
1232         assert(span->arrayMask & SPAN_Z);
1233         if (!_swrast_depth_test_span(ctx, span)) {
1234            /* all fragments failed test */
1235            goto end;
1236         }
1237      }
1238   }
1239
1240   if (ctx->Query.CurrentOcclusionObject) {
1241      /* update count of 'passed' fragments */
1242      struct gl_query_object *q = ctx->Query.CurrentOcclusionObject;
1243      GLuint i;
1244      for (i = 0; i < span->end; i++)
1245         q->Result += span->array->mask[i];
1246   }
1247
1248   /* We had to wait until now to check for glColorMask(0,0,0,0) because of
1249    * the occlusion test.
1250    */
1251   if (fb->_NumColorDrawBuffers == 1 &&
1252       !GET_COLORMASK(ctx->Color.ColorMask, 0)) {
1253      /* no colors to write */
1254      goto end;
1255   }
1256
1257   /* If we were able to defer fragment color computation to now, there's
1258    * a good chance that many fragments will have already been killed by
1259    * Z/stencil testing.
1260    */
1261   if (shaderOrTexture && swrast->_DeferredTexture) {
1262      shade_texture_span(ctx, span);
1263   }
1264
1265#if CHAN_BITS == 32
1266   if ((span->arrayAttribs & VARYING_BIT_COL0) == 0) {
1267      interpolate_active_attribs(ctx, span, VARYING_BIT_COL0);
1268   }
1269#else
1270   if ((span->arrayMask & SPAN_RGBA) == 0) {
1271      interpolate_int_colors(ctx, span);
1272   }
1273#endif
1274
1275   assert(span->arrayMask & SPAN_RGBA);
1276
1277   if (span->primitive == GL_BITMAP || !swrast->SpecularVertexAdd) {
1278      /* Add primary and specular (diffuse + specular) colors */
1279      if (!shader) {
1280         if (ctx->Fog.ColorSumEnabled ||
1281             (ctx->Light.Enabled &&
1282              ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR)) {
1283            add_specular(ctx, span);
1284         }
1285      }
1286   }
1287
1288   /* Fog */
1289   if (swrast->_FogEnabled) {
1290      _swrast_fog_rgba_span(ctx, span);
1291   }
1292
1293   /* Antialias coverage application */
1294   if (span->arrayMask & SPAN_COVERAGE) {
1295      apply_aa_coverage(span);
1296   }
1297
1298   /* Clamp color/alpha values over the range [0.0, 1.0] before storage */
1299   if (ctx->Color.ClampFragmentColor == GL_TRUE &&
1300       span->array->ChanType == GL_FLOAT) {
1301      clamp_colors(span);
1302   }
1303
1304   /*
1305    * Write to renderbuffers.
1306    * Depending on glDrawBuffer() state and the which color outputs are
1307    * written by the fragment shader, we may either replicate one color to
1308    * all renderbuffers or write a different color to each renderbuffer.
1309    * multiFragOutputs=TRUE for the later case.
1310    */
1311   {
1312      const GLuint numBuffers = fb->_NumColorDrawBuffers;
1313      const struct gl_program *fp = ctx->FragmentProgram._Current;
1314      const GLboolean multiFragOutputs =
1315         _swrast_use_fragment_program(ctx)
1316         && fp->info.outputs_written >= (1 << FRAG_RESULT_DATA0);
1317      /* Save srcColorType because convert_color_type() can change it */
1318      const GLenum srcColorType = span->array->ChanType;
1319      GLuint buf;
1320
1321      for (buf = 0; buf < numBuffers; buf++) {
1322         struct gl_renderbuffer *rb = fb->_ColorDrawBuffers[buf];
1323
1324         /* color[fragOutput] will be written to buffer[buf] */
1325
1326         if (rb) {
1327            /* re-use one of the attribute array buffers for rgbaSave */
1328            GLchan (*rgbaSave)[4] = (GLchan (*)[4]) span->array->attribs[0];
1329            struct swrast_renderbuffer *srb = swrast_renderbuffer(rb);
1330            const GLenum dstColorType = srb->ColorType;
1331
1332            assert(dstColorType == GL_UNSIGNED_BYTE ||
1333                   dstColorType == GL_FLOAT);
1334
1335            /* set span->array->rgba to colors for renderbuffer's datatype */
1336            if (srcColorType != dstColorType) {
1337               convert_color_type(span, srcColorType, dstColorType,
1338                                  multiFragOutputs ? buf : 0);
1339            }
1340            else {
1341               if (srcColorType == GL_UNSIGNED_BYTE) {
1342                  span->array->rgba = span->array->rgba8;
1343               }
1344               else {
1345                  span->array->rgba = (void *)
1346                     span->array->attribs[VARYING_SLOT_COL0];
1347               }
1348            }
1349
1350            if (!multiFragOutputs && numBuffers > 1) {
1351               /* save colors for second, third renderbuffer writes */
1352               memcpy(rgbaSave, span->array->rgba,
1353                      4 * span->end * sizeof(GLchan));
1354            }
1355
1356            assert(rb->_BaseFormat == GL_RGBA ||
1357                   rb->_BaseFormat == GL_RGB ||
1358                   rb->_BaseFormat == GL_RED ||
1359                   rb->_BaseFormat == GL_RG ||
1360		   rb->_BaseFormat == GL_ALPHA);
1361
1362            if (ctx->Color.ColorLogicOpEnabled) {
1363               _swrast_logicop_rgba_span(ctx, rb, span);
1364            }
1365            else if ((ctx->Color.BlendEnabled >> buf) & 1) {
1366               _swrast_blend_span(ctx, rb, span);
1367            }
1368
1369            if (GET_COLORMASK(ctx->Color.ColorMask, buf) != 0xf) {
1370               _swrast_mask_rgba_span(ctx, rb, span, buf);
1371            }
1372
1373            if (span->arrayMask & SPAN_XY) {
1374               /* array of pixel coords */
1375               put_values(ctx, rb,
1376                          span->array->ChanType, span->end,
1377                          span->array->x, span->array->y,
1378                          span->array->rgba, span->array->mask);
1379            }
1380            else {
1381               /* horizontal run of pixels */
1382               _swrast_put_row(ctx, rb,
1383                               span->array->ChanType,
1384                               span->end, span->x, span->y,
1385                               span->array->rgba,
1386                               span->writeAll ? NULL: span->array->mask);
1387            }
1388
1389            if (!multiFragOutputs && numBuffers > 1) {
1390               /* restore original span values */
1391               memcpy(span->array->rgba, rgbaSave,
1392                      4 * span->end * sizeof(GLchan));
1393            }
1394
1395         } /* if rb */
1396      } /* for buf */
1397   }
1398
1399end:
1400   /* restore these values before returning */
1401   span->interpMask = origInterpMask;
1402   span->arrayMask = origArrayMask;
1403   span->arrayAttribs = origArrayAttribs;
1404   span->array->ChanType = origChanType;
1405   span->array->rgba = origRgba;
1406}
1407
1408
1409/**
1410 * Read float RGBA pixels from a renderbuffer.  Clipping will be done to
1411 * prevent reading ouside the buffer's boundaries.
1412 * \param rgba  the returned colors
1413 */
1414void
1415_swrast_read_rgba_span( struct gl_context *ctx, struct gl_renderbuffer *rb,
1416                        GLuint n, GLint x, GLint y,
1417                        GLvoid *rgba)
1418{
1419   struct swrast_renderbuffer *srb = swrast_renderbuffer(rb);
1420   GLenum dstType = GL_FLOAT;
1421   const GLint bufWidth = (GLint) rb->Width;
1422   const GLint bufHeight = (GLint) rb->Height;
1423
1424   if (y < 0 || y >= bufHeight || x + (GLint) n < 0 || x >= bufWidth) {
1425      /* completely above, below, or right */
1426      /* XXX maybe leave rgba values undefined? */
1427      memset(rgba, 0, 4 * n * sizeof(GLchan));
1428   }
1429   else {
1430      GLint skip, length;
1431      GLubyte *src;
1432
1433      if (x < 0) {
1434         /* left edge clipping */
1435         skip = -x;
1436         length = (GLint) n - skip;
1437         if (length < 0) {
1438            /* completely left of window */
1439            return;
1440         }
1441         if (length > bufWidth) {
1442            length = bufWidth;
1443         }
1444      }
1445      else if ((GLint) (x + n) > bufWidth) {
1446         /* right edge clipping */
1447         skip = 0;
1448         length = bufWidth - x;
1449         if (length < 0) {
1450            /* completely to right of window */
1451            return;
1452         }
1453      }
1454      else {
1455         /* no clipping */
1456         skip = 0;
1457         length = (GLint) n;
1458      }
1459
1460      assert(rb);
1461      assert(rb->_BaseFormat == GL_RGBA ||
1462	     rb->_BaseFormat == GL_RGB ||
1463	     rb->_BaseFormat == GL_RG ||
1464	     rb->_BaseFormat == GL_RED ||
1465	     rb->_BaseFormat == GL_LUMINANCE ||
1466	     rb->_BaseFormat == GL_INTENSITY ||
1467	     rb->_BaseFormat == GL_LUMINANCE_ALPHA ||
1468	     rb->_BaseFormat == GL_ALPHA);
1469
1470      assert(srb->Map);
1471      (void) srb; /* silence unused var warning */
1472
1473      src = _swrast_pixel_address(rb, x + skip, y);
1474
1475      if (dstType == GL_UNSIGNED_BYTE) {
1476         _mesa_unpack_ubyte_rgba_row(rb->Format, length, src,
1477                                     (GLubyte (*)[4]) rgba + skip);
1478      }
1479      else if (dstType == GL_FLOAT) {
1480         _mesa_unpack_rgba_row(rb->Format, length, src,
1481                               (GLfloat (*)[4]) rgba + skip);
1482      }
1483      else {
1484         _mesa_problem(ctx, "unexpected type in _swrast_read_rgba_span()");
1485      }
1486   }
1487}
1488
1489
1490/**
1491 * Get colors at x/y positions with clipping.
1492 * \param type  type of values to return
1493 */
1494static void
1495get_values(struct gl_context *ctx, struct gl_renderbuffer *rb,
1496           GLuint count, const GLint x[], const GLint y[],
1497           void *values, GLenum type)
1498{
1499   GLuint i;
1500
1501   for (i = 0; i < count; i++) {
1502      if (x[i] >= 0 && y[i] >= 0 &&
1503	  x[i] < (GLint) rb->Width && y[i] < (GLint) rb->Height) {
1504         /* inside */
1505         const GLubyte *src = _swrast_pixel_address(rb, x[i], y[i]);
1506
1507         if (type == GL_UNSIGNED_BYTE) {
1508            _mesa_unpack_ubyte_rgba_row(rb->Format, 1, src,
1509                                        (GLubyte (*)[4]) values + i);
1510         }
1511         else if (type == GL_FLOAT) {
1512            _mesa_unpack_rgba_row(rb->Format, 1, src,
1513                                  (GLfloat (*)[4]) values + i);
1514         }
1515         else {
1516            _mesa_problem(ctx, "unexpected type in get_values()");
1517         }
1518      }
1519   }
1520}
1521
1522
1523/**
1524 * Get row of colors with clipping.
1525 * \param type  type of values to return
1526 */
1527static void
1528get_row(struct gl_context *ctx, struct gl_renderbuffer *rb,
1529        GLuint count, GLint x, GLint y,
1530        GLvoid *values, GLenum type)
1531{
1532   GLint skip = 0;
1533   GLubyte *src;
1534
1535   if (y < 0 || y >= (GLint) rb->Height)
1536      return; /* above or below */
1537
1538   if (x + (GLint) count <= 0 || x >= (GLint) rb->Width)
1539      return; /* entirely left or right */
1540
1541   if (x + count > rb->Width) {
1542      /* right clip */
1543      GLint clip = x + count - rb->Width;
1544      count -= clip;
1545   }
1546
1547   if (x < 0) {
1548      /* left clip */
1549      skip = -x;
1550      x = 0;
1551      count -= skip;
1552   }
1553
1554   src = _swrast_pixel_address(rb, x, y);
1555
1556   if (type == GL_UNSIGNED_BYTE) {
1557      _mesa_unpack_ubyte_rgba_row(rb->Format, count, src,
1558                                  (GLubyte (*)[4]) values + skip);
1559   }
1560   else if (type == GL_FLOAT) {
1561      _mesa_unpack_rgba_row(rb->Format, count, src,
1562                            (GLfloat (*)[4]) values + skip);
1563   }
1564   else {
1565      _mesa_problem(ctx, "unexpected type in get_row()");
1566   }
1567}
1568
1569
1570/**
1571 * Get RGBA pixels from the given renderbuffer.
1572 * Used by blending, logicop and masking functions.
1573 * \return pointer to the colors we read.
1574 */
1575void *
1576_swrast_get_dest_rgba(struct gl_context *ctx, struct gl_renderbuffer *rb,
1577                      SWspan *span)
1578{
1579   void *rbPixels;
1580
1581   /* Point rbPixels to a temporary space */
1582   rbPixels = span->array->attribs[VARYING_SLOT_MAX - 1];
1583
1584   /* Get destination values from renderbuffer */
1585   if (span->arrayMask & SPAN_XY) {
1586      get_values(ctx, rb, span->end, span->array->x, span->array->y,
1587                 rbPixels, span->array->ChanType);
1588   }
1589   else {
1590      get_row(ctx, rb, span->end, span->x, span->y,
1591              rbPixels, span->array->ChanType);
1592   }
1593
1594   return rbPixels;
1595}
1596