radeon_state.c

linux-2.6.15.6

					tileoffset += depthpixperline >> 6;
				}
			} else if (dev_priv->microcode_version == UCODE_R200) {
				/* works for rv250. */
				/* find first macro tile (8x2 4x4 z-pixels on rv250) */
				tileoffset =
				    ((pbox[i].y1 >> 3) * depthpixperline +
				     pbox[i].x1) >> 5;
				nrtilesx =
				    (pbox[i].x2 >> 5) - (pbox[i].x1 >> 5);
				nrtilesy =
				    (pbox[i].y2 >> 3) - (pbox[i].y1 >> 3);
				for (j = 0; j <= nrtilesy; j++) {
					BEGIN_RING(4);
					OUT_RING(CP_PACKET3
						 (RADEON_3D_CLEAR_ZMASK, 2));
					/* first tile */
					/* judging by the first tile offset needed, could possibly
					   directly address/clear 4x4 tiles instead of 8x2 * 4x4
					   macro tiles, though would still need clear mask for
					   right/bottom if truely 4x4 granularity is desired ? */
					OUT_RING(tileoffset * 16);
					/* the number of tiles to clear */
					OUT_RING(nrtilesx + 1);
					/* clear mask : chooses the clearing pattern. */
					OUT_RING(clearmask);
					ADVANCE_RING();
					tileoffset += depthpixperline >> 5;
				}
			} else {	/* rv 100 */
				/* rv100 might not need 64 pix alignment, who knows */
				/* offsets are, hmm, weird */
				tileoffset =
				    ((pbox[i].y1 >> 4) * depthpixperline +
				     pbox[i].x1) >> 6;
				nrtilesx =
				    ((pbox[i].x2 & ~63) -
				     (pbox[i].x1 & ~63)) >> 4;
				nrtilesy =
				    (pbox[i].y2 >> 4) - (pbox[i].y1 >> 4);
				for (j = 0; j <= nrtilesy; j++) {
					BEGIN_RING(4);
					OUT_RING(CP_PACKET3
						 (RADEON_3D_CLEAR_ZMASK, 2));
					OUT_RING(tileoffset * 128);
					/* the number of tiles to clear */
					OUT_RING(nrtilesx + 4);
					/* clear mask : chooses the clearing pattern. */
					OUT_RING(clearmask);
					ADVANCE_RING();
					tileoffset += depthpixperline >> 6;
				}
			}
		}

		/* TODO don't always clear all hi-level z tiles */
		if ((dev_priv->flags & CHIP_HAS_HIERZ)
		    && (dev_priv->microcode_version == UCODE_R200)
		    && (flags & RADEON_USE_HIERZ))
			/* r100 and cards without hierarchical z-buffer have no high-level z-buffer */
			/* FIXME : the mask supposedly contains low-res z values. So can't set
			   just to the max (0xff? or actually 0x3fff?), need to take z clear
			   value into account? */
		{
			BEGIN_RING(4);
			OUT_RING(CP_PACKET3(RADEON_3D_CLEAR_HIZ, 2));
			OUT_RING(0x0);	/* First tile */
			OUT_RING(0x3cc0);
			OUT_RING((0xff << 22) | (0xff << 6) | 0x003f003f);
			ADVANCE_RING();
		}
	}

	/* We have to clear the depth and/or stencil buffers by
	 * rendering a quad into just those buffers.  Thus, we have to
	 * make sure the 3D engine is configured correctly.
	 */
	if ((dev_priv->microcode_version == UCODE_R200) &&
	    (flags & (RADEON_DEPTH | RADEON_STENCIL))) {

		int tempPP_CNTL;
		int tempRE_CNTL;
		int tempRB3D_CNTL;
		int tempRB3D_ZSTENCILCNTL;
		int tempRB3D_STENCILREFMASK;
		int tempRB3D_PLANEMASK;
		int tempSE_CNTL;
		int tempSE_VTE_CNTL;
		int tempSE_VTX_FMT_0;
		int tempSE_VTX_FMT_1;
		int tempSE_VAP_CNTL;
		int tempRE_AUX_SCISSOR_CNTL;

		tempPP_CNTL = 0;
		tempRE_CNTL = 0;

		tempRB3D_CNTL = depth_clear->rb3d_cntl;

		tempRB3D_ZSTENCILCNTL = depth_clear->rb3d_zstencilcntl;
		tempRB3D_STENCILREFMASK = 0x0;

		tempSE_CNTL = depth_clear->se_cntl;

		/* Disable TCL */

		tempSE_VAP_CNTL = (	/* SE_VAP_CNTL__FORCE_W_TO_ONE_MASK |  */
					  (0x9 <<
					   SE_VAP_CNTL__VF_MAX_VTX_NUM__SHIFT));

		tempRB3D_PLANEMASK = 0x0;

		tempRE_AUX_SCISSOR_CNTL = 0x0;

		tempSE_VTE_CNTL =
		    SE_VTE_CNTL__VTX_XY_FMT_MASK | SE_VTE_CNTL__VTX_Z_FMT_MASK;

		/* Vertex format (X, Y, Z, W) */
		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;
		}

		if (flags & RADEON_USE_COMP_ZBUF) {
			tempRB3D_ZSTENCILCNTL |= RADEON_Z_COMPRESSION_ENABLE |
			    RADEON_Z_DECOMPRESSION_ENABLE;
		}
		if (flags & RADEON_USE_HIERZ) {
			tempRB3D_ZSTENCILCNTL |= RADEON_Z_HIERARCHY_ENABLE;
		}

		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))) {

		int tempRB3D_ZSTENCILCNTL = depth_clear->rb3d_zstencilcntl;

		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;
		}

		if (flags & RADEON_USE_COMP_ZBUF) {
			tempRB3D_ZSTENCILCNTL |= RADEON_Z_COMPRESSION_ENABLE |
			    RADEON_Z_DECOMPRESSION_ENABLE;
		}
		if (flags & RADEON_USE_HIERZ) {
			tempRB3D_ZSTENCILCNTL |= RADEON_Z_HIERARCHY_ENABLE;
		}

		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, tempRB3D_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_radeon_private_t *dev_priv = dev->dev_private;