radeon_swtcl.c

Mesa is an open-source implementation of the OpenGL specification - a system for rendering interacti

/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_swtcl.c,v 1.6 2003/05/06 23:52:08 daenzer Exp $ */
/**************************************************************************

Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
                     VA Linux Systems Inc., Fremont, California.

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, sublicense, 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 above copyright notice and this permission notice (including the
next paragraph) shall be included in all copies or substantial
portions of the Software.

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 NONINFRINGEMENT.
IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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.

**************************************************************************/

/*
 * Authors:
 *   Keith Whitwell <[e-mail]>
 */

#include "glheader.h"
#include "mtypes.h"
#include "colormac.h"
#include "enums.h"
#include "imports.h"
#include "macros.h"

#include "swrast_setup/swrast_setup.h"
#include "math/m_translate.h"
#include "tnl/tnl.h"
#include "tnl/t_context.h"
#include "tnl/t_pipeline.h"

#include "radeon_context.h"
#include "radeon_ioctl.h"
#include "radeon_state.h"
#include "radeon_swtcl.h"
#include "radeon_tcl.h"


static void flush_last_swtcl_prim( radeonContextPtr rmesa  );

/* R100: xyzw, c0, c1/fog, stq[0..2]  = 4+1+1+3*3 = 15  right? */
/* R200: xyzw, c0, c1/fog, strq[0..5] = 4+1+1+4*6 = 30 */
#define RADEON_MAX_TNL_VERTEX_SIZE (15 * sizeof(GLfloat))	/* for mesa _tnl stage */

/***********************************************************************
 *                         Initialization 
 ***********************************************************************/

#define EMIT_ATTR( ATTR, STYLE, F0 )					\
do {									\
   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = (ATTR);	\
   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = (STYLE);	\
   rmesa->swtcl.vertex_attr_count++;					\
   fmt_0 |= F0;								\
} while (0)

#define EMIT_PAD( N )							\
do {									\
   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = 0;		\
   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = EMIT_PAD;	\
   rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].offset = (N);		\
   rmesa->swtcl.vertex_attr_count++;					\
} while (0)

static GLuint radeon_cp_vc_frmts[3][2] =
{
   { RADEON_CP_VC_FRMT_ST0, RADEON_CP_VC_FRMT_ST0 | RADEON_CP_VC_FRMT_Q0 },
   { RADEON_CP_VC_FRMT_ST1, RADEON_CP_VC_FRMT_ST1 | RADEON_CP_VC_FRMT_Q1 },
   { RADEON_CP_VC_FRMT_ST2, RADEON_CP_VC_FRMT_ST2 | RADEON_CP_VC_FRMT_Q2 },
};

static void radeonSetVertexFormat( GLcontext *ctx )
{
   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
   TNLcontext *tnl = TNL_CONTEXT(ctx);
   struct vertex_buffer *VB = &tnl->vb;
   DECLARE_RENDERINPUTS(index_bitset);
   int fmt_0 = 0;
   int offset = 0;

   RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset );

   /* Important:
    */
   if ( VB->NdcPtr != NULL ) {
      VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr;
   }
   else {
      VB->AttribPtr[VERT_ATTRIB_POS] = VB->ClipPtr;
   }

   assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL );
   rmesa->swtcl.vertex_attr_count = 0;

   /* EMIT_ATTR's must be in order as they tell t_vertex.c how to
    * build up a hardware vertex.
    */
   if ( !rmesa->swtcl.needproj ||
        RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) {	/* for projtex */
      EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F, 
		 RADEON_CP_VC_FRMT_XY |	RADEON_CP_VC_FRMT_Z | RADEON_CP_VC_FRMT_W0 );
      offset = 4;
   }
   else {
      EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_3F, 
		 RADEON_CP_VC_FRMT_XY |	RADEON_CP_VC_FRMT_Z );
      offset = 3;
   }

   rmesa->swtcl.coloroffset = offset;
#if MESA_LITTLE_ENDIAN 
   EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_RGBA, 
	      RADEON_CP_VC_FRMT_PKCOLOR );
#else
   EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_ABGR,
	      RADEON_CP_VC_FRMT_PKCOLOR );
#endif
   offset += 1;

   rmesa->swtcl.specoffset = 0;
   if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 ) ||
       RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) {

#if MESA_LITTLE_ENDIAN 
      if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) {
	 rmesa->swtcl.specoffset = offset;
	 EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_3UB_3F_RGB,
	 	    RADEON_CP_VC_FRMT_PKSPEC );
      }
      else {
	 EMIT_PAD( 3 );
      }

      if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) {
	 EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1UB_1F,
	 	    RADEON_CP_VC_FRMT_PKSPEC );
      }
      else {
	 EMIT_PAD( 1 );
      }
#else
      if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) {
	 EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1UB_1F,
	 	    RADEON_CP_VC_FRMT_PKSPEC );
      }
      else {
	 EMIT_PAD( 1 );
      }

      if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) {
	 rmesa->swtcl.specoffset = offset;
	 EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_3UB_3F_BGR,
	 	    RADEON_CP_VC_FRMT_PKSPEC );
      }
      else {
	 EMIT_PAD( 3 );
      }
#endif
   }

   if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) {
      int i;

      for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
	 if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) {
	    GLuint sz = VB->TexCoordPtr[i]->size;

	    switch (sz) {
	    case 1:
	    case 2:
	       EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_2F,
			  radeon_cp_vc_frmts[i][0] );
	       break;
	    case 3:
	    case 4:
	       if (ctx->Texture.Unit[i]._ReallyEnabled & (TEXTURE_CUBE_BIT) ) {
		  EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_3F,
			     radeon_cp_vc_frmts[i][1] );
	       } else {
		  EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_3F_XYW,
			     radeon_cp_vc_frmts[i][1] );
	       }
	       break;
	    default:
	       continue;
	    };
	 }
      }
   }

   if (!RENDERINPUTS_EQUAL( rmesa->tnl_index_bitset, index_bitset ) ||
	fmt_0 != rmesa->swtcl.vertex_format) {
      RADEON_NEWPRIM(rmesa);
      rmesa->swtcl.vertex_format = fmt_0;
      rmesa->swtcl.vertex_size =
	  _tnl_install_attrs( ctx,
			      rmesa->swtcl.vertex_attrs, 
			      rmesa->swtcl.vertex_attr_count,
			      NULL, 0 );
      rmesa->swtcl.vertex_size /= 4;
      RENDERINPUTS_COPY( rmesa->tnl_index_bitset, index_bitset );
      if (RADEON_DEBUG & DEBUG_VERTS)
	 fprintf( stderr, "%s: vertex_size= %d floats\n",
		  __FUNCTION__, rmesa->swtcl.vertex_size);
   }
}


static void radeonRenderStart( GLcontext *ctx )
{
   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );

   radeonSetVertexFormat( ctx );
   
   if (rmesa->dma.flush != 0 && 
       rmesa->dma.flush != flush_last_swtcl_prim)
      rmesa->dma.flush( rmesa );
}


/**
 * Set vertex state for SW TCL.  The primary purpose of this function is to
 * determine in advance whether or not the hardware can / should do the
 * projection divide or Mesa should do it.
 */
void radeonChooseVertexState( GLcontext *ctx )
{
   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
   TNLcontext *tnl = TNL_CONTEXT(ctx);

   GLuint se_coord_fmt = rmesa->hw.set.cmd[SET_SE_COORDFMT];
   
   se_coord_fmt &= ~(RADEON_VTX_XY_PRE_MULT_1_OVER_W0 |
		     RADEON_VTX_Z_PRE_MULT_1_OVER_W0 |
		     RADEON_VTX_W0_IS_NOT_1_OVER_W0);

   /* We must ensure that we don't do _tnl_need_projected_coords while in a
    * rasterization fallback.  As this function will be called again when we
    * leave a rasterization fallback, we can just skip it for now.
    */
   if (rmesa->Fallback != 0)
      return;

   /* HW perspective divide is a win, but tiny vertex formats are a
    * bigger one.
    */

   if ((!RENDERINPUTS_TEST_RANGE( tnl->render_inputs_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX ) &&
       !RENDERINPUTS_TEST( tnl->render_inputs_bitset, _TNL_ATTRIB_COLOR1 ))
       || (ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED))) {
      rmesa->swtcl.needproj = GL_TRUE;
      se_coord_fmt |= (RADEON_VTX_XY_PRE_MULT_1_OVER_W0 |
		      RADEON_VTX_Z_PRE_MULT_1_OVER_W0);
   }
   else {
      rmesa->swtcl.needproj = GL_FALSE;
      se_coord_fmt |= (RADEON_VTX_W0_IS_NOT_1_OVER_W0);
   }

   _tnl_need_projected_coords( ctx, rmesa->swtcl.needproj );

   if ( se_coord_fmt != rmesa->hw.set.cmd[SET_SE_COORDFMT] ) {
      RADEON_STATECHANGE( rmesa, set );
      rmesa->hw.set.cmd[SET_SE_COORDFMT] = se_coord_fmt;
   }
}


/* Flush vertices in the current dma region.
 */
static void flush_last_swtcl_prim( radeonContextPtr rmesa  )
{
   if (RADEON_DEBUG & DEBUG_IOCTL)
      fprintf(stderr, "%s\n", __FUNCTION__);

   rmesa->dma.flush = NULL;

   if (rmesa->dma.current.buf) {
      struct radeon_dma_region *current = &rmesa->dma.current;
      GLuint current_offset = (rmesa->radeonScreen->gart_buffer_offset +
			       current->buf->buf->idx * RADEON_BUFFER_SIZE + 
			       current->start);

      assert (!(rmesa->swtcl.hw_primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND));

      assert (current->start + 
	      rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
	      current->ptr);

      if (rmesa->dma.current.start != rmesa->dma.current.ptr) {
	 radeonEnsureCmdBufSpace( rmesa, VERT_AOS_BUFSZ +
			          rmesa->hw.max_state_size + VBUF_BUFSZ );

	 radeonEmitVertexAOS( rmesa,
			      rmesa->swtcl.vertex_size,
			      current_offset);

	 radeonEmitVbufPrim( rmesa,
			     rmesa->swtcl.vertex_format,
			     rmesa->swtcl.hw_primitive,
			     rmesa->swtcl.numverts);
      }

      rmesa->swtcl.numverts = 0;
      current->start = current->ptr;
   }
}


/* Alloc space in the current dma region.
 */
static INLINE void *
radeonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize )
{
   GLuint bytes = vsize * nverts;

   if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) 
      radeonRefillCurrentDmaRegion( rmesa );

   if (!rmesa->dma.flush) {
      rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
      rmesa->dma.flush = flush_last_swtcl_prim;
   }

   assert( vsize == rmesa->swtcl.vertex_size * 4 );
   assert( rmesa->dma.flush == flush_last_swtcl_prim );
   assert (rmesa->dma.current.start + 
	   rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 ==
	   rmesa->dma.current.ptr);


   {
      GLubyte *head = (GLubyte *)(rmesa->dma.current.address + rmesa->dma.current.ptr);
      rmesa->dma.current.ptr += bytes;
      rmesa->swtcl.numverts += nverts;
      return head;
   }

}


/*
 * Render unclipped vertex buffers by emitting vertices directly to
 * dma buffers.  Use strip/fan hardware primitives where possible.
 * Try to simulate missing primitives with indexed vertices.
 */
#define HAVE_POINTS      1
#define HAVE_LINES       1
#define HAVE_LINE_STRIPS 1
#define HAVE_TRIANGLES   1
#define HAVE_TRI_STRIPS  1
#define HAVE_TRI_STRIP_1 0
#define HAVE_TRI_FANS    1
#define HAVE_QUADS       0
#define HAVE_QUAD_STRIPS 0
#define HAVE_POLYGONS    0
/* \todo: is it possible to make "ELTS" work with t_vertex code ? */
#define HAVE_ELTS        0

static const GLuint hw_prim[GL_POLYGON+1] = {
   RADEON_CP_VC_CNTL_PRIM_TYPE_POINT,
   RADEON_CP_VC_CNTL_PRIM_TYPE_LINE,
   0,
   RADEON_CP_VC_CNTL_PRIM_TYPE_LINE_STRIP,
   RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_LIST,
   RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_STRIP,
   RADEON_CP_VC_CNTL_PRIM_TYPE_TRI_FAN,
   0,
   0,
   0
};

static INLINE void
radeonDmaPrimitive( radeonContextPtr rmesa, GLenum prim )
{
   RADEON_NEWPRIM( rmesa );
   rmesa->swtcl.hw_primitive = hw_prim[prim];
   assert(rmesa->dma.current.ptr == rmesa->dma.current.start);
}

#define LOCAL_VARS radeonContextPtr rmesa = RADEON_CONTEXT(ctx)
#define INIT( prim ) radeonDmaPrimitive( rmesa, prim )
#define FLUSH()  RADEON_NEWPRIM( rmesa )
#define GET_CURRENT_VB_MAX_VERTS() \
  (((int)rmesa->dma.current.end - (int)rmesa->dma.current.ptr) / (rmesa->swtcl.vertex_size*4))
#define GET_SUBSEQUENT_VB_MAX_VERTS() \
  ((RADEON_BUFFER_SIZE) / (rmesa->swtcl.vertex_size*4))
#define ALLOC_VERTS( nr ) \
  radeonAllocDmaLowVerts( rmesa, nr, rmesa->swtcl.vertex_size * 4 )
#define EMIT_VERTS( ctx, j, nr, buf ) \
  _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf)

#define TAG(x) radeon_dma_##x
#include "tnl_dd/t_dd_dmatmp.h"


/**********************************************************************/
/*                          Render pipeline stage                     */
/**********************************************************************/


static GLboolean radeon_run_render( GLcontext *ctx,
				    struct tnl_pipeline_stage *stage )
{
   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
   TNLcontext *tnl = TNL_CONTEXT(ctx);
   struct vertex_buffer *VB = &tnl->vb;
   tnl_render_func *tab = TAG(render_tab_verts);
   GLuint i;

   if (rmesa->swtcl.indexed_verts.buf) 
      RELEASE_ELT_VERTS();
   	
   if (rmesa->swtcl.RenderIndex != 0 ||   
       !radeon_dma_validate_render( ctx, VB ))
      return GL_TRUE;		

   tnl->Driver.Render.Start( ctx );

   for (i = 0 ; i < VB->PrimitiveCount ; i++)
   {
      GLuint prim = VB->Primitive[i].mode;
      GLuint start = VB->Primitive[i].start;
      GLuint length = VB->Primitive[i].count;

      if (!length)
	 continue;

      if (RADEON_DEBUG & DEBUG_PRIMS)