radeon_state.c

ati driver

		tempSE_VTX_FMT_0 =
			SE_VTX_FMT_0__VTX_Z0_PRESENT_MASK |
			SE_VTX_FMT_0__VTX_W0_PRESENT_MASK;
		tempSE_VTX_FMT_1 = 0x0;


		/* 
		 * Depth buffer specific enables 
		 */
		if (flags & RADEON_DEPTH) {
			/* Enable depth buffer */
			tempRB3D_CNTL |= RADEON_Z_ENABLE;
		} else {
			/* Disable depth buffer */
			tempRB3D_CNTL &= ~RADEON_Z_ENABLE;
		}

		/* 
		 * Stencil buffer specific enables
		 */
		if ( flags & RADEON_STENCIL ) {
			tempRB3D_CNTL |=  RADEON_STENCIL_ENABLE;
			tempRB3D_STENCILREFMASK = clear->depth_mask; 
		} else {
			tempRB3D_CNTL &= ~RADEON_STENCIL_ENABLE;
			tempRB3D_STENCILREFMASK = 0x00000000;
		}

		BEGIN_RING( 26 );
		RADEON_WAIT_UNTIL_2D_IDLE();

		OUT_RING_REG( RADEON_PP_CNTL, tempPP_CNTL );
		OUT_RING_REG( R200_RE_CNTL, tempRE_CNTL );
		OUT_RING_REG( RADEON_RB3D_CNTL, tempRB3D_CNTL );
		OUT_RING_REG( RADEON_RB3D_ZSTENCILCNTL,
			      tempRB3D_ZSTENCILCNTL );
		OUT_RING_REG( RADEON_RB3D_STENCILREFMASK, 
			      tempRB3D_STENCILREFMASK );
		OUT_RING_REG( RADEON_RB3D_PLANEMASK, tempRB3D_PLANEMASK );
		OUT_RING_REG( RADEON_SE_CNTL, tempSE_CNTL );
		OUT_RING_REG( R200_SE_VTE_CNTL, tempSE_VTE_CNTL );
		OUT_RING_REG( R200_SE_VTX_FMT_0, tempSE_VTX_FMT_0 );
		OUT_RING_REG( R200_SE_VTX_FMT_1, tempSE_VTX_FMT_1 );
		OUT_RING_REG( R200_SE_VAP_CNTL, tempSE_VAP_CNTL );
		OUT_RING_REG( R200_RE_AUX_SCISSOR_CNTL, 
			      tempRE_AUX_SCISSOR_CNTL );
		ADVANCE_RING();

		/* Make sure we restore the 3D state next time.
		 */
		dev_priv->sarea_priv->ctx_owner = 0;

		for ( i = 0 ; i < nbox ; i++ ) {
			
			/* Funny that this should be required -- 
			 *  sets top-left?
			 */
			radeon_emit_clip_rect( dev_priv,
					       &sarea_priv->boxes[i] );

			BEGIN_RING( 14 );
			OUT_RING( CP_PACKET3( R200_3D_DRAW_IMMD_2, 12 ) );
			OUT_RING( (RADEON_PRIM_TYPE_RECT_LIST |
				   RADEON_PRIM_WALK_RING |
				   (3 << RADEON_NUM_VERTICES_SHIFT)) );
			OUT_RING( depth_boxes[i].ui[CLEAR_X1] );
			OUT_RING( depth_boxes[i].ui[CLEAR_Y1] );
			OUT_RING( depth_boxes[i].ui[CLEAR_DEPTH] );
			OUT_RING( 0x3f800000 );
			OUT_RING( depth_boxes[i].ui[CLEAR_X1] );
			OUT_RING( depth_boxes[i].ui[CLEAR_Y2] );
			OUT_RING( depth_boxes[i].ui[CLEAR_DEPTH] );
			OUT_RING( 0x3f800000 );
			OUT_RING( depth_boxes[i].ui[CLEAR_X2] );
			OUT_RING( depth_boxes[i].ui[CLEAR_Y2] );
			OUT_RING( depth_boxes[i].ui[CLEAR_DEPTH] );
			OUT_RING( 0x3f800000 );
			ADVANCE_RING();
		}
	} 
	else if ( (flags & (RADEON_DEPTH | RADEON_STENCIL)) ) {

		rb3d_cntl = depth_clear->rb3d_cntl;

		if ( flags & RADEON_DEPTH ) {
			rb3d_cntl |=  RADEON_Z_ENABLE;
		} else {
			rb3d_cntl &= ~RADEON_Z_ENABLE;
		}

		if ( flags & RADEON_STENCIL ) {
			rb3d_cntl |=  RADEON_STENCIL_ENABLE;
			rb3d_stencilrefmask = clear->depth_mask; /* misnamed field */
		} else {
			rb3d_cntl &= ~RADEON_STENCIL_ENABLE;
			rb3d_stencilrefmask = 0x00000000;
		}

		BEGIN_RING( 13 );
		RADEON_WAIT_UNTIL_2D_IDLE();

		OUT_RING( CP_PACKET0( RADEON_PP_CNTL, 1 ) );
		OUT_RING( 0x00000000 );
		OUT_RING( rb3d_cntl );
		
		OUT_RING_REG( RADEON_RB3D_ZSTENCILCNTL,
			      depth_clear->rb3d_zstencilcntl );
		OUT_RING_REG( RADEON_RB3D_STENCILREFMASK,
			      rb3d_stencilrefmask );
		OUT_RING_REG( RADEON_RB3D_PLANEMASK,
			      0x00000000 );
		OUT_RING_REG( RADEON_SE_CNTL,
			      depth_clear->se_cntl );
		ADVANCE_RING();

		/* Make sure we restore the 3D state next time.
		 */
		dev_priv->sarea_priv->ctx_owner = 0;

		for ( i = 0 ; i < nbox ; i++ ) {
			
			/* Funny that this should be required -- 
			 *  sets top-left?
			 */
			radeon_emit_clip_rect( dev_priv,
					       &sarea_priv->boxes[i] );

			BEGIN_RING( 15 );

			OUT_RING( CP_PACKET3( RADEON_3D_DRAW_IMMD, 13 ) );
			OUT_RING( RADEON_VTX_Z_PRESENT |
				  RADEON_VTX_PKCOLOR_PRESENT);
			OUT_RING( (RADEON_PRIM_TYPE_RECT_LIST |
				   RADEON_PRIM_WALK_RING |
				   RADEON_MAOS_ENABLE |
				   RADEON_VTX_FMT_RADEON_MODE |
				   (3 << RADEON_NUM_VERTICES_SHIFT)) );


			OUT_RING( depth_boxes[i].ui[CLEAR_X1] );
			OUT_RING( depth_boxes[i].ui[CLEAR_Y1] );
			OUT_RING( depth_boxes[i].ui[CLEAR_DEPTH] );
			OUT_RING( 0x0 );

			OUT_RING( depth_boxes[i].ui[CLEAR_X1] );
			OUT_RING( depth_boxes[i].ui[CLEAR_Y2] );
			OUT_RING( depth_boxes[i].ui[CLEAR_DEPTH] );
			OUT_RING( 0x0 );

			OUT_RING( depth_boxes[i].ui[CLEAR_X2] );
			OUT_RING( depth_boxes[i].ui[CLEAR_Y2] );
			OUT_RING( depth_boxes[i].ui[CLEAR_DEPTH] );
			OUT_RING( 0x0 );

			ADVANCE_RING();
		}
	}

	/* Increment the clear counter.  The client-side 3D driver must
	 * wait on this value before performing the clear ioctl.  We
	 * need this because the card's so damned fast...
	 */
	dev_priv->sarea_priv->last_clear++;

	BEGIN_RING( 4 );

	RADEON_CLEAR_AGE( dev_priv->sarea_priv->last_clear );
	RADEON_WAIT_UNTIL_IDLE();

	ADVANCE_RING();
}

static void radeon_cp_dispatch_swap( drm_device_t *dev )
{
	drm_radeon_private_t *dev_priv = dev->dev_private;
	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
	int nbox = sarea_priv->nbox;
	drm_clip_rect_t *pbox = sarea_priv->boxes;
	int i;
	RING_LOCALS;
	DRM_DEBUG( "\n" );

	/* Do some trivial performance monitoring...
	 */
	if (dev_priv->do_boxes)
		radeon_cp_performance_boxes( dev_priv );


	/* Wait for the 3D stream to idle before dispatching the bitblt.
	 * This will prevent data corruption between the two streams.
	 */
	BEGIN_RING( 2 );

	RADEON_WAIT_UNTIL_3D_IDLE();

	ADVANCE_RING();

	for ( i = 0 ; i < nbox ; i++ ) {
		int x = pbox[i].x1;
		int y = pbox[i].y1;
		int w = pbox[i].x2 - x;
		int h = pbox[i].y2 - y;

		DRM_DEBUG( "dispatch swap %d,%d-%d,%d\n",
			   x, y, w, h );

		BEGIN_RING( 7 );

		OUT_RING( CP_PACKET3( RADEON_CNTL_BITBLT_MULTI, 5 ) );
		OUT_RING( RADEON_GMC_SRC_PITCH_OFFSET_CNTL |
			  RADEON_GMC_DST_PITCH_OFFSET_CNTL |
			  RADEON_GMC_BRUSH_NONE |
			  (dev_priv->color_fmt << 8) |
			  RADEON_GMC_SRC_DATATYPE_COLOR |
			  RADEON_ROP3_S |
			  RADEON_DP_SRC_SOURCE_MEMORY |
			  RADEON_GMC_CLR_CMP_CNTL_DIS |
			  RADEON_GMC_WR_MSK_DIS );
		
		/* Make this work even if front & back are flipped:
		 */
		if (dev_priv->current_page == 0) {
			OUT_RING( dev_priv->back_pitch_offset );
			OUT_RING( dev_priv->front_pitch_offset );
		} 
		else {
			OUT_RING( dev_priv->front_pitch_offset );
			OUT_RING( dev_priv->back_pitch_offset );
		}

		OUT_RING( (x << 16) | y );
		OUT_RING( (x << 16) | y );
		OUT_RING( (w << 16) | h );

		ADVANCE_RING();
	}

	/* Increment the frame counter.  The client-side 3D driver must
	 * throttle the framerate by waiting for this value before
	 * performing the swapbuffer ioctl.
	 */
	dev_priv->sarea_priv->last_frame++;

	BEGIN_RING( 4 );

	RADEON_FRAME_AGE( dev_priv->sarea_priv->last_frame );
	RADEON_WAIT_UNTIL_2D_IDLE();

	ADVANCE_RING();
}

static void radeon_cp_dispatch_flip( drm_device_t *dev )
{
	drm_radeon_private_t *dev_priv = dev->dev_private;
	drm_sarea_t *sarea = (drm_sarea_t *)dev_priv->sarea->handle;
	int offset = (dev_priv->current_page == 1)
		   ? dev_priv->front_offset : dev_priv->back_offset;
	RING_LOCALS;
	DRM_DEBUG( "%s: page=%d pfCurrentPage=%d\n", 
		__FUNCTION__, 
		dev_priv->current_page,
		dev_priv->sarea_priv->pfCurrentPage);

	/* Do some trivial performance monitoring...
	 */
	if (dev_priv->do_boxes) {
		dev_priv->stats.boxes |= RADEON_BOX_FLIP;
		radeon_cp_performance_boxes( dev_priv );
	}

	/* Update the frame offsets for both CRTCs
	 */
	BEGIN_RING( 6 );

	RADEON_WAIT_UNTIL_3D_IDLE();
	OUT_RING_REG( RADEON_CRTC_OFFSET, ( ( sarea->frame.y * dev_priv->front_pitch
					      + sarea->frame.x 
					      * ( dev_priv->color_fmt - 2 ) ) & ~7 )
					  + offset );
	OUT_RING_REG( RADEON_CRTC2_OFFSET, dev_priv->sarea_priv->crtc2_base
					   + offset );

	ADVANCE_RING();

	/* Increment the frame counter.  The client-side 3D driver must
	 * throttle the framerate by waiting for this value before
	 * performing the swapbuffer ioctl.
	 */
	dev_priv->sarea_priv->last_frame++;
	dev_priv->sarea_priv->pfCurrentPage = dev_priv->current_page =
					      1 - dev_priv->current_page;

	BEGIN_RING( 2 );

	RADEON_FRAME_AGE( dev_priv->sarea_priv->last_frame );

	ADVANCE_RING();
}

static int bad_prim_vertex_nr( int primitive, int nr )
{
	switch (primitive & RADEON_PRIM_TYPE_MASK) {
	case RADEON_PRIM_TYPE_NONE:
	case RADEON_PRIM_TYPE_POINT:
		return nr < 1;
	case RADEON_PRIM_TYPE_LINE:
		return (nr & 1) || nr == 0;
	case RADEON_PRIM_TYPE_LINE_STRIP:
		return nr < 2;
	case RADEON_PRIM_TYPE_TRI_LIST:
	case RADEON_PRIM_TYPE_3VRT_POINT_LIST:
	case RADEON_PRIM_TYPE_3VRT_LINE_LIST:
	case RADEON_PRIM_TYPE_RECT_LIST:
		return nr % 3 || nr == 0;
	case RADEON_PRIM_TYPE_TRI_FAN:
	case RADEON_PRIM_TYPE_TRI_STRIP:
		return nr < 3;
	default:
		return 1;
	}	
}



typedef struct {
	unsigned int start;
	unsigned int finish;
	unsigned int prim;
	unsigned int numverts;
	unsigned int offset;   
        unsigned int vc_format;
} drm_radeon_tcl_prim_t;

static void radeon_cp_dispatch_vertex( drm_device_t *dev,
				       drm_buf_t *buf,
				       drm_radeon_tcl_prim_t *prim,
				       drm_clip_rect_t *boxes,
				       int nbox )

{
	drm_radeon_private_t *dev_priv = dev->dev_private;
	drm_clip_rect_t box;
	int offset = dev_priv->agp_buffers_offset + buf->offset + prim->start;
	int numverts = (int)prim->numverts;
	int i = 0;
	RING_LOCALS;

	DRM_DEBUG("hwprim 0x%x vfmt 0x%x %d..%d %d verts\n",
		  prim->prim,
		  prim->vc_format,
		  prim->start,
		  prim->finish,
		  prim->numverts);

	if (bad_prim_vertex_nr( prim->prim, prim->numverts )) {
		DRM_ERROR( "bad prim %x numverts %d\n", 
			   prim->prim, prim->numverts );
		return;
	}

	do {
		/* Emit the next cliprect */
		if ( i < nbox ) {
			if (DRM_COPY_FROM_USER_UNCHECKED( &box, &boxes[i], sizeof(box) ))
				return;

			radeon_emit_clip_rect( dev_priv, &box );
		}

		/* Emit the vertex buffer rendering commands */
		BEGIN_RING( 5 );

		OUT_RING( CP_PACKET3( RADEON_3D_RNDR_GEN_INDX_PRIM, 3 ) );
		OUT_RING( offset );
		OUT_RING( numverts );
		OUT_RING( prim->vc_format );
		OUT_RING( prim->prim | RADEON_PRIM_WALK_LIST |
			  RADEON_COLOR_ORDER_RGBA |
			  RADEON_VTX_FMT_RADEON_MODE |
			  (numverts << RADEON_NUM_VERTICES_SHIFT) );

		ADVANCE_RING();

		i++;
	} while ( i < nbox );
}



static void radeon_cp_discard_buffer( drm_device_t *dev, drm_buf_t *buf )
{
	drm_radeon_private_t *dev_priv = dev->dev_private;
	drm_radeon_buf_priv_t *buf_priv = buf->dev_private;
	RING_LOCALS;

	buf_priv->age = ++dev_priv->sarea_priv->last_dispatch;

	/* Emit the vertex buffer age */
	BEGIN_RING( 2 );
	RADEON_DISPATCH_AGE( buf_priv->age );
	ADVANCE_RING();

	buf->pending = 1;
	buf->used = 0;
}

static void radeon_cp_dispatch_indirect( drm_device_t *dev,
					 drm_buf_t *buf,
					 int start, int end )
{
	drm_radeon_private_t *dev_priv = dev->dev_private;
	RING_LOCALS;
	DRM_DEBUG( "indirect: buf=%d s=0x%x e=0x%x\n",
		   buf->idx, start, end );

	if ( start != end ) {
		int offset = (dev_priv->agp_buffers_offset
			      + buf->offset + start);
		int dwords = (end - start + 3) / sizeof(u32);

		/* Indirect buffer data must be an even number of
		 * dwords, so if we've been given an odd number we must
		 * pad the data with a Type-2 CP packet.
		 */
		if ( dwords & 1 ) {
			u32 *data = (u32 *)
				((char *)dev_priv->buffers->handle
				 + buf->offset + start);
			data[dwords++] = RADEON_CP_PACKET2;
		}

		/* Fire off the indirect buffer */
		BEGIN_RING( 3 );

		OUT_RING( CP_PACKET0( RADEON_CP_IB_BASE, 1 ) );
		OUT_RING( offset );
		OUT_RING( dwords );

		ADVANCE_RING();
	}
}


static void radeon_cp_dispatch_indices( drm_device_t *dev,
					drm_buf_t *elt_buf,
					drm_radeon_tcl_prim_t *prim, 
					drm_clip_rect_t *boxes,
					int nbox )
{
	drm_radeon_private_t *dev_priv = dev->dev_private;
	drm_clip_rect_t box;
	int offset = dev_priv->agp_buffers_offset + prim->offset;
	u32 *data;
	int dwords;
	int i = 0;
	int start = prim->start + RADEON_INDEX_PRIM_OFFSET;
	int count = (prim->finish - start) / sizeof(u16);

	DRM_DEBUG("hwprim 0x%x vfmt 0x%x %d..%d offset: %x nr %d\n",
		  prim->prim,
		  prim->vc_format,
		  prim->start,
		  prim->finish,
		  prim->offset,
		  prim->numverts);

	if (bad_prim_vertex_nr( prim->prim, count )) {
		DRM_ERROR( "bad prim %x count %d\n", 
			   prim->prim, count );
		return;
	}


	if ( start >= prim->finish ||
	     (prim->start & 0x7) ) {
		DRM_ERROR( "buffer prim %d\n", prim->prim );
		return;
	}

	dwords = (prim->finish - prim->start + 3) / sizeof(u32);

	data = (u32 *)((char *)dev_priv->buffers->handle +
		       elt_buf->offset + prim->start);

	data[0] = CP_PACKET3( RADEON_3D_RNDR_GEN_INDX_PRIM, dwords-2 );
	data[1] = offset;
	data[2] = prim->numverts;
	data[3] = prim->vc_format;
	data[4] = (prim->prim |
		   RADEON_PRIM_WALK_IND |
		   RADEON_COLOR_ORDER_RGBA |
		   RADEON_VTX_FMT_RADEON_MODE |
		   (count << RADEON_NUM_VERTICES_SHIFT) );

	do {
		if ( i < nbox ) {
			if (DRM_COPY_FROM_USER_UNCHECKED( &box, &boxes[i], sizeof(box) ))
				return;
			
			radeon_emit_clip_rect( dev_priv, &box );
		}

		radeon_cp_dispatch_indirect( dev, elt_buf,
					     prim->start,
					     prim->finish );

		i++;
	} while ( i < nbox );

}

#define RADEON_MAX_TEXTURE_SIZE (RADEON_BUFFER_SIZE - 8 * sizeof(u32))

static int radeon_cp_dispatch_texture( drm_device_t *dev,
				       drm_radeon_texture_t *tex,
				       drm_radeon_tex_image_t *image )
{
	drm_radeon_private_t *dev_priv = dev->dev_private;
	drm_buf_t *buf;
	u32 format;
	u32 *buffer;
	const u8 *data;
	int size, dwords, tex_width, blit_width;
	u32 y, height;
	int i;
	RING_LOCALS;

	dev_priv->stats.boxes |= RADEON_BOX_TEXTURE_LOAD;

	/* Flush the pixel cache.  This ensures no pixel data gets mixed
	 * up with the texture data from the host data blit, otherwise
	 * part of the texture image may be corrupted.
	 */
	BEGIN_RING( 4 );
	RADEON_FLUSH_CACHE();
	RADEON_WAIT_UNTIL_IDLE();
	ADVANCE_RING();

#ifdef __BIG_ENDIAN
	/* The Mesa texture functions provide the data in little endian as the
	 * chip wants it, but we need to compensate for the fact that the CP
	 * ring gets byte-swapped
	 */
	BEGIN_RING( 2 );
	OUT_RING_REG( RADEON_RBBM_GUICNTL, RADEON_HOST_DATA_SWAP_32BIT );
	ADVANCE_RING();
#endif


	/* The compiler won't optimize away a division by a variable,
	 * even if the only legal values are powers of two.  Thus, we'll
	 * use a shift instead.