radeon_accel.c

x.org上有关ati系列显卡最新驱动

    stop.flush = 0;

    i = 0;
    do {
	ret = drmCommandWrite(info->drmFD, DRM_RADEON_CP_STOP, &stop,
			      sizeof(drmRadeonCPStop));
    } while (ret && errno == EBUSY && i++ < RADEON_IDLE_RETRY);

    if (ret == 0) {
	return 0;
    } else if (errno != EBUSY) {
	return -errno;
    }

    stop.idle = 0;

    if (drmCommandWrite(info->drmFD, DRM_RADEON_CP_STOP,
			&stop, sizeof(drmRadeonCPStop))) {
	return -errno;
    } else {
	return 0;
    }
}

/* Get an indirect buffer for the CP 2D acceleration commands  */
drmBufPtr RADEONCPGetBuffer(ScrnInfoPtr pScrn)
{
    RADEONInfoPtr  info = RADEONPTR(pScrn);
    drmDMAReq      dma;
    drmBufPtr      buf = NULL;
    int            indx = 0;
    int            size = 0;
    int            i = 0;
    int            ret;

#if 0
    /* FIXME: pScrn->pScreen has not been initialized when this is first
     * called from RADEONSelectBuffer via RADEONDRICPInit.  We could use
     * the screen index from pScrn, which is initialized, and then get
     * the screen from screenInfo.screens[index], but that is a hack.
     */
    dma.context = DRIGetContext(pScrn->pScreen);
#else
    /* This is the X server's context */
    dma.context = 0x00000001;
#endif

    dma.send_count    = 0;
    dma.send_list     = NULL;
    dma.send_sizes    = NULL;
    dma.flags         = 0;
    dma.request_count = 1;
    dma.request_size  = RADEON_BUFFER_SIZE;
    dma.request_list  = &indx;
    dma.request_sizes = &size;
    dma.granted_count = 0;

    while (1) {
	do {
	    ret = drmDMA(info->drmFD, &dma);
	    if (ret && ret != -EBUSY) {
		xf86DrvMsg(pScrn->scrnIndex, X_ERROR,
			   "%s: CP GetBuffer %d\n", __FUNCTION__, ret);
	    }
	} while ((ret == -EBUSY) && (i++ < RADEON_TIMEOUT));

	if (ret == 0) {
	    buf = &info->buffers->list[indx];
	    buf->used = 0;
	    if (RADEON_VERBOSE) {
		xf86DrvMsg(pScrn->scrnIndex, X_INFO,
			   "   GetBuffer returning %d %p\n",
			   buf->idx, buf->address);
	    }
	    return buf;
	}

	xf86DrvMsg(pScrn->scrnIndex, X_ERROR,
		   "GetBuffer timed out, resetting engine...\n");
	RADEONEngineReset(pScrn);
	RADEONEngineRestore(pScrn);

	/* Always restart the engine when doing CP 2D acceleration */
	RADEONCP_RESET(pScrn, info);
	RADEONCP_START(pScrn, info);
    }
}

/* Flush the indirect buffer to the kernel for submission to the card */
void RADEONCPFlushIndirect(ScrnInfoPtr pScrn, int discard)
{
    RADEONInfoPtr      info   = RADEONPTR(pScrn);
    drmBufPtr          buffer = info->indirectBuffer;
    int                start  = info->indirectStart;
    drmRadeonIndirect  indirect;

    if (!buffer) return;
    if (start == buffer->used && !discard) return;

    if (RADEON_VERBOSE) {
	xf86DrvMsg(pScrn->scrnIndex, X_INFO, "Flushing buffer %d\n",
		   buffer->idx);
    }

    indirect.idx     = buffer->idx;
    indirect.start   = start;
    indirect.end     = buffer->used;
    indirect.discard = discard;

    drmCommandWriteRead(info->drmFD, DRM_RADEON_INDIRECT,
			&indirect, sizeof(drmRadeonIndirect));

    if (discard) {
	info->indirectBuffer = RADEONCPGetBuffer(pScrn);
	info->indirectStart  = 0;
    } else {
	/* Start on a double word boundary */
	info->indirectStart  = buffer->used = (buffer->used + 7) & ~7;
	if (RADEON_VERBOSE) {
	    xf86DrvMsg(pScrn->scrnIndex, X_INFO, "   Starting at %d\n",
		       info->indirectStart);
	}
    }
}

/* Flush and release the indirect buffer */
void RADEONCPReleaseIndirect(ScrnInfoPtr pScrn)
{
    RADEONInfoPtr      info   = RADEONPTR(pScrn);
    drmBufPtr          buffer = info->indirectBuffer;
    int                start  = info->indirectStart;
    drmRadeonIndirect  indirect;

    info->indirectBuffer = NULL;
    info->indirectStart  = 0;

    if (!buffer) return;

    if (RADEON_VERBOSE) {
	xf86DrvMsg(pScrn->scrnIndex, X_INFO, "Releasing buffer %d\n",
		   buffer->idx);
    }

    indirect.idx     = buffer->idx;
    indirect.start   = start;
    indirect.end     = buffer->used;
    indirect.discard = 1;

    drmCommandWriteRead(info->drmFD, DRM_RADEON_INDIRECT,
			&indirect, sizeof(drmRadeonIndirect));
}

/** \brief Calculate HostDataBlit parameters from pointer and pitch
 *
 * This is a helper for the trivial HostDataBlit users that don't need to worry
 * about tiling etc.
 */
void
RADEONHostDataParams(ScrnInfoPtr pScrn, CARD8 *dst, CARD32 pitch, int cpp,
		     CARD32 *dstPitchOff, int *x, int *y)
{
    RADEONInfoPtr info = RADEONPTR( pScrn );
    CARD32 dstOffs = dst - info->FB + info->fbLocation;

    *dstPitchOff = pitch << 16 | (dstOffs & ~RADEON_BUFFER_ALIGN) >> 10;
    *y = ( dstOffs & RADEON_BUFFER_ALIGN ) / pitch;
    *x = ( ( dstOffs & RADEON_BUFFER_ALIGN ) - ( *y * pitch ) ) / cpp;
}

/* Set up a hostdata blit to transfer data from system memory to the
 * framebuffer. Returns the address where the data can be written to and sets
 * the dstPitch and hpass variables as required.
 */
CARD8*
RADEONHostDataBlit(
    ScrnInfoPtr pScrn,
    unsigned int cpp,
    unsigned int w,
    CARD32 dstPitchOff,
    CARD32 *bufPitch,
    int x,
    int *y,
    unsigned int *h,
    unsigned int *hpass
){
    RADEONInfoPtr info = RADEONPTR( pScrn );
    CARD32 format, dwords;
    CARD8 *ret;
    RING_LOCALS;

    if ( *h == 0 )
    {
	return NULL;
    }

    switch ( cpp )
    {
    case 4:
	format = RADEON_GMC_DST_32BPP;
	*bufPitch = 4 * w;
	break;
    case 2:
	format = RADEON_GMC_DST_16BPP;
	*bufPitch = 2 * ((w + 1) & ~1);
	break;
    case 1:
	format = RADEON_GMC_DST_8BPP_CI;
	*bufPitch = (w + 3) & ~3;
	break;
    default:
	xf86DrvMsg( pScrn->scrnIndex, X_ERROR,
		    "%s: Unsupported cpp %d!\n", __func__, cpp );
	return NULL;
    }

#if X_BYTE_ORDER == X_BIG_ENDIAN
    /* Swap doesn't work on R300 and later, it's handled during the
     * copy to ind. buffer pass
     */
    if (info->ChipFamily < CHIP_FAMILY_R300) {
        BEGIN_RING(2);
	if (cpp == 2)
	    OUT_RING_REG(RADEON_RBBM_GUICNTL,
			 RADEON_HOST_DATA_SWAP_HDW);
	else if (cpp == 1)
	    OUT_RING_REG(RADEON_RBBM_GUICNTL,
			 RADEON_HOST_DATA_SWAP_32BIT);
	else
	    OUT_RING_REG(RADEON_RBBM_GUICNTL,
			 RADEON_HOST_DATA_SWAP_NONE);
	ADVANCE_RING();
    }
#endif

    /*RADEON_PURGE_CACHE();
      RADEON_WAIT_UNTIL_IDLE();*/

    *hpass = min( *h, ( ( RADEON_BUFFER_SIZE - 10 * 4 ) / *bufPitch ) );
    dwords = *hpass * *bufPitch / 4;

    BEGIN_RING( dwords + 10 );
    OUT_RING( CP_PACKET3( RADEON_CP_PACKET3_CNTL_HOSTDATA_BLT, dwords + 10 - 2 ) );
    OUT_RING( RADEON_GMC_DST_PITCH_OFFSET_CNTL
	    | RADEON_GMC_DST_CLIPPING
	    | RADEON_GMC_BRUSH_NONE
	    | format
	    | RADEON_GMC_SRC_DATATYPE_COLOR
	    | RADEON_ROP3_S
	    | RADEON_DP_SRC_SOURCE_HOST_DATA
	    | RADEON_GMC_CLR_CMP_CNTL_DIS
	    | RADEON_GMC_WR_MSK_DIS );
    OUT_RING( dstPitchOff );
    OUT_RING( (*y << 16) | x );
    OUT_RING( ((*y + *hpass) << 16) | (x + w) );
    OUT_RING( 0xffffffff );
    OUT_RING( 0xffffffff );
    OUT_RING( *y << 16 | x );
    OUT_RING( *hpass << 16 | (*bufPitch / cpp) );
    OUT_RING( dwords );

    ret = ( CARD8* )&__head[__count];

    __count += dwords;
    ADVANCE_RING();

    *y += *hpass;
    *h -= *hpass;

    return ret;
}

void RADEONCopySwap(CARD8 *dst, CARD8 *src, unsigned int size, int swap)
{
    switch(swap) {
    case RADEON_HOST_DATA_SWAP_HDW:
        {
	    unsigned int *d = (unsigned int *)dst;
	    unsigned int *s = (unsigned int *)src;
	    unsigned int nwords = size >> 2;

	    for (; nwords > 0; --nwords, ++d, ++s)
		*d = ((*s & 0xffff) << 16) | ((*s >> 16) & 0xffff);
	    return;
        }
    case RADEON_HOST_DATA_SWAP_32BIT:
        {
	    unsigned int *d = (unsigned int *)dst;
	    unsigned int *s = (unsigned int *)src;
	    unsigned int nwords = size >> 2;

	    for (; nwords > 0; --nwords, ++d, ++s)
#ifdef __powerpc__
		asm volatile("stwbrx %0,0,%1" : : "r" (*s), "r" (d));
#else
		*d = ((*s >> 24) & 0xff) | ((*s >> 8) & 0xff00)
			| ((*s & 0xff00) << 8) | ((*s & 0xff) << 24);
#endif
	    return;
        }
    case RADEON_HOST_DATA_SWAP_16BIT:
        {
	    unsigned short *d = (unsigned short *)dst;
	    unsigned short *s = (unsigned short *)src;
	    unsigned int nwords = size >> 1;

	    for (; nwords > 0; --nwords, ++d, ++s)
#ifdef __powerpc__
		asm volatile("stwbrx %0,0,%1" : : "r" (*s), "r" (d));
#else
	        *d = ((*s >> 24) & 0xff) | ((*s >> 8) & 0xff00)
			| ((*s & 0xff00) << 8) | ((*s & 0xff) << 24);
#endif
	    return;
	}
    }
    if (src != dst)
	    memmove(dst, src, size);
}

/* Copies a single pass worth of data for a hostdata blit set up by
 * RADEONHostDataBlit().
 */
void
RADEONHostDataBlitCopyPass(
    ScrnInfoPtr pScrn,
    unsigned int cpp,
    CARD8 *dst,
    CARD8 *src,
    unsigned int hpass,
    unsigned int dstPitch,
    unsigned int srcPitch
){

#if X_BYTE_ORDER == X_BIG_ENDIAN
    RADEONInfoPtr info = RADEONPTR( pScrn );
#endif

    /* RADEONHostDataBlitCopy can return NULL ! */
    if( (dst==NULL) || (src==NULL)) return;

    if ( dstPitch == srcPitch )
    {
#if X_BYTE_ORDER == X_BIG_ENDIAN
        if (info->ChipFamily >= CHIP_FAMILY_R300) {
	    switch(cpp) {
	    case 1:
		RADEONCopySwap(dst, src, hpass * dstPitch,
			       RADEON_HOST_DATA_SWAP_32BIT);
		return;
	    case 2:
	        RADEONCopySwap(dst, src, hpass * dstPitch,
			       RADEON_HOST_DATA_SWAP_HDW);
		return;
	    }
	}
#endif
	memcpy( dst, src, hpass * dstPitch );
    }
    else
    {
	unsigned int minPitch = min( dstPitch, srcPitch );
	while ( hpass-- )
	{
#if X_BYTE_ORDER == X_BIG_ENDIAN
            if (info->ChipFamily >= CHIP_FAMILY_R300) {
		switch(cpp) {
		case 1:
		    RADEONCopySwap(dst, src, minPitch,
				   RADEON_HOST_DATA_SWAP_32BIT);
		    goto next;
		case 2:
	            RADEONCopySwap(dst, src, minPitch,
				   RADEON_HOST_DATA_SWAP_HDW);
		    goto next;
		}
	    }
#endif
	    memcpy( dst, src, minPitch );
#if X_BYTE_ORDER == X_BIG_ENDIAN
	next:
#endif
	    src += srcPitch;
	    dst += dstPitch;
	}
    }
}

#endif

Bool RADEONAccelInit(ScreenPtr pScreen)
{
    ScrnInfoPtr    pScrn = xf86Screens[pScreen->myNum];
    RADEONInfoPtr  info  = RADEONPTR(pScrn);

#ifdef USE_EXA
    if (info->useEXA) {
# ifdef XF86DRI
	if (info->directRenderingEnabled) {
	    if (!RADEONDrawInitCP(pScreen))
		return FALSE;
	} else
# endif /* XF86DRI */
	{
	    if (!RADEONDrawInitMMIO(pScreen))
		return FALSE;
	}
    }
#endif /* USE_EXA */
#ifdef USE_XAA
    if (!info->useEXA) {
	XAAInfoRecPtr  a;

	if (!(a = info->accel = XAACreateInfoRec())) {
	    xf86DrvMsg(pScrn->scrnIndex, X_ERROR, "XAACreateInfoRec Error\n");
	    return FALSE;
	}

#ifdef XF86DRI
	if (info->directRenderingEnabled)
	    RADEONAccelInitCP(pScreen, a);
	else
#endif /* XF86DRI */
	    RADEONAccelInitMMIO(pScreen, a);

	RADEONEngineInit(pScrn);

	if (!XAAInit(pScreen, a)) {
	    xf86DrvMsg(pScrn->scrnIndex, X_ERROR, "XAAInit Error\n");
	    return FALSE;
	}
    }
#endif /* USE_XAA */
    return TRUE;
}

void RADEONInit3DEngine(ScrnInfoPtr pScrn)
{
    RADEONInfoPtr info = RADEONPTR (pScrn);

#ifdef XF86DRI
    if (info->directRenderingEnabled) {
	RADEONSAREAPrivPtr pSAREAPriv;

	pSAREAPriv = DRIGetSAREAPrivate(pScrn->pScreen);
	pSAREAPriv->ctxOwner = DRIGetContext(pScrn->pScreen);
	RADEONInit3DEngineCP(pScrn);
    } else
#endif
	RADEONInit3DEngineMMIO(pScrn);

    info->XInited3D = TRUE;
}