void
nvc0_m2mf_copy_linear(struct nouveau_context *nv,
struct nouveau_bo *dst, unsigned dstoff, unsigned dstdom,
struct nouveau_bo *src, unsigned srcoff, unsigned srcdom,
unsigned size)
{
struct nouveau_channel *chan = nv->screen->channel;
while (size) {
unsigned bytes = MIN2(size, 1 << 17);
MARK_RING (chan, 11, 4);
BEGIN_RING(chan, RING_MF(OFFSET_OUT_HIGH), 2);
OUT_RELOCh(chan, dst, dstoff, dstdom | NOUVEAU_BO_WR);
OUT_RELOCl(chan, dst, dstoff, dstdom | NOUVEAU_BO_WR);
BEGIN_RING(chan, RING_MF(OFFSET_IN_HIGH), 2);
OUT_RELOCh(chan, src, srcoff, srcdom | NOUVEAU_BO_RD);
OUT_RELOCl(chan, src, srcoff, srcdom | NOUVEAU_BO_RD);
BEGIN_RING(chan, RING_MF(LINE_LENGTH_IN), 2);
OUT_RING (chan, bytes);
OUT_RING (chan, 1);
BEGIN_RING(chan, RING_MF(EXEC), 1);
OUT_RING (chan, (1 << NVC0_M2MF_EXEC_INC__SHIFT) |
NVC0_M2MF_EXEC_LINEAR_IN | NVC0_M2MF_EXEC_LINEAR_OUT);
srcoff += bytes;
dstoff += bytes;
size -= bytes;
}
}
static int
nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst)
{
struct nv50_miptree *mt = nv50_miptree(ps->texture);
struct nouveau_channel *chan = screen->eng2d->channel;
struct nouveau_grobj *eng2d = screen->eng2d;
struct nouveau_bo *bo = nv50_miptree(ps->texture)->base.bo;
int format, mthd = dst ? NV50_2D_DST_FORMAT : NV50_2D_SRC_FORMAT;
int flags = NOUVEAU_BO_VRAM | (dst ? NOUVEAU_BO_WR : NOUVEAU_BO_RD);
format = nv50_format(ps->format);
if (format < 0)
return 1;
if (!bo->tile_flags) {
BEGIN_RING(chan, eng2d, mthd, 2);
OUT_RING (chan, format);
OUT_RING (chan, 1);
BEGIN_RING(chan, eng2d, mthd + 0x14, 5);
OUT_RING (chan, mt->level[ps->level].pitch);
OUT_RING (chan, ps->width);
OUT_RING (chan, ps->height);
OUT_RELOCh(chan, bo, ps->offset, flags);
OUT_RELOCl(chan, bo, ps->offset, flags);
} else {
BEGIN_RING(chan, eng2d, mthd, 5);
OUT_RING (chan, format);
OUT_RING (chan, 0);
OUT_RING (chan, mt->level[ps->level].tile_mode << 4);
OUT_RING (chan, 1);
OUT_RING (chan, 0);
BEGIN_RING(chan, eng2d, mthd + 0x18, 4);
OUT_RING (chan, ps->width);
OUT_RING (chan, ps->height);
OUT_RELOCh(chan, bo, ps->offset, flags);
OUT_RELOCl(chan, bo, ps->offset, flags);
}
#if 0
if (dst) {
BEGIN_RING(chan, eng2d, NV50_2D_CLIP_X, 4);
OUT_RING (chan, 0);
OUT_RING (chan, 0);
OUT_RING (chan, surf->width);
OUT_RING (chan, surf->height);
}
#endif
return 0;
}
static void
nv50_render_condition(struct pipe_context *pipe,
struct pipe_query *pq, uint mode)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_channel *chan = nv50->screen->base.channel;
struct nouveau_grobj *tesla = nv50->screen->tesla;
struct nv50_query *q;
if (!pq) {
BEGIN_RING(chan, tesla, NV50TCL_COND_MODE, 1);
OUT_RING (chan, NV50TCL_COND_MODE_ALWAYS);
return;
}
q = nv50_query(pq);
if (mode == PIPE_RENDER_COND_WAIT ||
mode == PIPE_RENDER_COND_BY_REGION_WAIT) {
/* XXX: big fence, FIFO semaphore might be better */
BEGIN_RING(chan, tesla, 0x0110, 1);
OUT_RING (chan, 0);
}
BEGIN_RING(chan, tesla, NV50TCL_COND_ADDRESS_HIGH, 3);
OUT_RELOCh(chan, q->bo, 0, NOUVEAU_BO_GART | NOUVEAU_BO_RD);
OUT_RELOCl(chan, q->bo, 0, NOUVEAU_BO_GART | NOUVEAU_BO_RD);
OUT_RING (chan, NV50TCL_COND_MODE_RES);
}
static INLINE void
nvc0_program_update_context_state(struct nvc0_context *nvc0,
struct nvc0_program *prog, int stage)
{
struct nouveau_channel *chan = nvc0->screen->base.channel;
if (prog->hdr[1])
nvc0->state.tls_required |= 1 << stage;
else
nvc0->state.tls_required &= ~(1 << stage);
if (prog->immd_size) {
const unsigned rl = NOUVEAU_BO_VRAM | NOUVEAU_BO_RD;
BEGIN_RING(chan, RING_3D(CB_SIZE), 3);
/* NOTE: may overlap code of a different shader */
OUT_RING (chan, align(prog->immd_size, 0x100));
OUT_RELOCh(chan, nvc0->screen->text, prog->immd_base, rl);
OUT_RELOCl(chan, nvc0->screen->text, prog->immd_base, rl);
BEGIN_RING(chan, RING_3D(CB_BIND(stage)), 1);
OUT_RING (chan, (14 << 4) | 1);
nvc0->state.c14_bound |= 1 << stage;
} else
if (nvc0->state.c14_bound & (1 << stage)) {
BEGIN_RING(chan, RING_3D(CB_BIND(stage)), 1);
OUT_RING (chan, (14 << 4) | 0);
nvc0->state.c14_bound &= ~(1 << stage);
}
}
static void
nvc0_m2mf_push_rect(struct pipe_screen *pscreen,
const struct nv50_m2mf_rect *dst,
const void *data,
unsigned nblocksx, unsigned nblocksy)
{
struct nouveau_channel *chan;
const uint8_t *src = (const uint8_t *)data;
const int cpp = dst->cpp;
const int line_len = nblocksx * cpp;
int dy = dst->y;
assert(nouveau_bo_tile_layout(dst->bo));
BEGIN_RING(chan, RING_MF(TILING_MODE_OUT), 5);
OUT_RING (chan, dst->tile_mode);
OUT_RING (chan, dst->width * cpp);
OUT_RING (chan, dst->height);
OUT_RING (chan, dst->depth);
OUT_RING (chan, dst->z);
while (nblocksy) {
int line_count, words;
int size = MIN2(AVAIL_RING(chan), NV04_PFIFO_MAX_PACKET_LEN);
if (size < (12 + words)) {
FIRE_RING(chan);
continue;
}
line_count = (size * 4) / line_len;
words = (line_count * line_len + 3) / 4;
BEGIN_RING(chan, RING_MF(OFFSET_OUT_HIGH), 2);
OUT_RELOCh(chan, dst->bo, dst->base, dst->domain | NOUVEAU_BO_WR);
OUT_RELOCl(chan, dst->bo, dst->base, dst->domain | NOUVEAU_BO_WR);
BEGIN_RING(chan, RING_MF(TILING_POSITION_OUT_X), 2);
OUT_RING (chan, dst->x * cpp);
OUT_RING (chan, dy);
BEGIN_RING(chan, RING_MF(LINE_LENGTH_IN), 2);
OUT_RING (chan, line_len);
OUT_RING (chan, line_count);
BEGIN_RING(chan, RING_MF(EXEC), 1);
OUT_RING (chan, (1 << NVC0_M2MF_EXEC_INC__SHIFT) |
NVC0_M2MF_EXEC_PUSH | NVC0_M2MF_EXEC_LINEAR_IN);
BEGIN_RING_NI(chan, RING_MF(DATA), words);
OUT_RINGp (chan, src, words);
dy += line_count;
src += line_len * line_count;
nblocksy -= line_count;
}
}
static void
nv50_clear_depth_stencil(struct pipe_context *pipe,
struct pipe_surface *dst,
unsigned clear_flags,
double depth,
unsigned stencil,
unsigned dstx, unsigned dsty,
unsigned width, unsigned height)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nv50_screen *screen = nv50->screen;
struct nouveau_channel *chan = screen->base.channel;
struct nv50_miptree *mt = nv50_miptree(dst->texture);
struct nv50_surface *sf = nv50_surface(dst);
struct nouveau_bo *bo = mt->base.bo;
uint32_t mode = 0;
if (clear_flags & PIPE_CLEAR_DEPTH) {
BEGIN_RING(chan, RING_3D(CLEAR_DEPTH), 1);
OUT_RINGf (chan, depth);
mode |= NV50_3D_CLEAR_BUFFERS_Z;
}
if (clear_flags & PIPE_CLEAR_STENCIL) {
BEGIN_RING(chan, RING_3D(CLEAR_STENCIL), 1);
OUT_RING (chan, stencil & 0xff);
mode |= NV50_3D_CLEAR_BUFFERS_S;
}
if (MARK_RING(chan, 17, 2))
return;
BEGIN_RING(chan, RING_3D(ZETA_ADDRESS_HIGH), 5);
OUT_RELOCh(chan, bo, sf->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
OUT_RELOCl(chan, bo, sf->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
OUT_RING (chan, nv50_format_table[dst->format].rt);
OUT_RING (chan, mt->level[sf->base.u.tex.level].tile_mode << 4);
OUT_RING (chan, 0);
BEGIN_RING(chan, RING_3D(ZETA_ENABLE), 1);
OUT_RING (chan, 1);
BEGIN_RING(chan, RING_3D(ZETA_HORIZ), 3);
OUT_RING (chan, sf->width);
OUT_RING (chan, sf->height);
OUT_RING (chan, (1 << 16) | 1);
BEGIN_RING(chan, RING_3D(VIEWPORT_HORIZ(0)), 2);
OUT_RING (chan, (width << 16) | dstx);
OUT_RING (chan, (height << 16) | dsty);
BEGIN_RING(chan, RING_3D(CLEAR_BUFFERS), 1);
OUT_RING (chan, mode);
nv50->dirty |= NV50_NEW_FRAMEBUFFER;
}
static void
nv50_clear_render_target(struct pipe_context *pipe,
struct pipe_surface *dst,
const float *rgba,
unsigned dstx, unsigned dsty,
unsigned width, unsigned height)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nv50_screen *screen = nv50->screen;
struct nouveau_channel *chan = screen->base.channel;
struct nv50_miptree *mt = nv50_miptree(dst->texture);
struct nv50_surface *sf = nv50_surface(dst);
struct nouveau_bo *bo = mt->base.bo;
BEGIN_RING(chan, RING_3D(CLEAR_COLOR(0)), 4);
OUT_RINGf (chan, rgba[0]);
OUT_RINGf (chan, rgba[1]);
OUT_RINGf (chan, rgba[2]);
OUT_RINGf (chan, rgba[3]);
if (MARK_RING(chan, 18, 2))
return;
BEGIN_RING(chan, RING_3D(RT_CONTROL), 1);
OUT_RING (chan, 1);
BEGIN_RING(chan, RING_3D(RT_ADDRESS_HIGH(0)), 5);
OUT_RELOCh(chan, bo, sf->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
OUT_RELOCl(chan, bo, sf->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
OUT_RING (chan, nv50_format_table[dst->format].rt);
OUT_RING (chan, mt->level[sf->base.u.tex.level].tile_mode << 4);
OUT_RING (chan, 0);
BEGIN_RING(chan, RING_3D(RT_HORIZ(0)), 2);
OUT_RING (chan, sf->width);
OUT_RING (chan, sf->height);
BEGIN_RING(chan, RING_3D(RT_ARRAY_MODE), 1);
OUT_RING (chan, 1);
/* NOTE: only works with D3D clear flag (5097/0x143c bit 4) */
BEGIN_RING(chan, RING_3D(VIEWPORT_HORIZ(0)), 2);
OUT_RING (chan, (width << 16) | dstx);
OUT_RING (chan, (height << 16) | dsty);
BEGIN_RING(chan, RING_3D(CLEAR_BUFFERS), 1);
OUT_RING (chan, 0x3c);
nv50->dirty |= NV50_NEW_FRAMEBUFFER;
}
static void
nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_channel *chan = nv50->screen->base.channel;
struct nouveau_grobj *tesla = nv50->screen->tesla;
struct nv50_query *q = nv50_query(pq);
WAIT_RING (chan, 5);
BEGIN_RING(chan, tesla, NV50TCL_QUERY_ADDRESS_HIGH, 4);
OUT_RELOCh(chan, q->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
OUT_RELOCl(chan, q->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
OUT_RING (chan, 0x00000000);
OUT_RING (chan, 0x0100f002);
FIRE_RING (chan);
}
static void
nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_channel *chan = nv50->screen->nvws->channel;
struct nouveau_grobj *tesla = nv50->screen->tesla;
struct nv50_query *q = nv50_query(pq);
struct nouveau_bo *bo = nv50->screen->nvws->get_bo(q->buffer);
WAIT_RING (chan, 5);
BEGIN_RING(chan, tesla, 0x1b00, 4);
OUT_RELOCh(chan, bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
OUT_RELOCl(chan, bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
OUT_RING (chan, 0x00000000);
OUT_RING (chan, 0x0100f002);
FIRE_RING (chan);
}
static void
nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_channel *chan = nv50->screen->base.channel;
struct nouveau_grobj *tesla = nv50->screen->tesla;
struct nv50_query *q = nv50_query(pq);
MARK_RING (chan, 5, 2); /* flush on lack of space or relocs */
BEGIN_RING(chan, tesla, NV50TCL_QUERY_ADDRESS_HIGH, 4);
OUT_RELOCh(chan, q->bo, 0, NOUVEAU_BO_GART | NOUVEAU_BO_WR);
OUT_RELOCl(chan, q->bo, 0, NOUVEAU_BO_GART | NOUVEAU_BO_WR);
OUT_RING (chan, 0x00000000);
OUT_RING (chan, 0x0100f002);
BEGIN_RING(chan, tesla, NV50TCL_SAMPLECNT_ENABLE, 1);
OUT_RING (chan, 0);
}
void
nvc0_m2mf_push_linear(struct nouveau_context *nv,
struct nouveau_bo *dst, unsigned offset, unsigned domain,
unsigned size, void *data)
{
struct nouveau_channel *chan = nv->screen->channel;
uint32_t *src = (uint32_t *)data;
unsigned count = (size + 3) / 4;
MARK_RING (chan, 8, 2);
BEGIN_RING(chan, RING_MF(OFFSET_OUT_HIGH), 2);
OUT_RELOCh(chan, dst, offset, domain | NOUVEAU_BO_WR);
OUT_RELOCl(chan, dst, offset, domain | NOUVEAU_BO_WR);
BEGIN_RING(chan, RING_MF(LINE_LENGTH_IN), 2);
OUT_RING (chan, size);
OUT_RING (chan, 1);
BEGIN_RING(chan, RING_MF(EXEC), 1);
OUT_RING (chan, 0x100111);
while (count) {
unsigned nr = AVAIL_RING(chan);
if (nr < 9) {
FIRE_RING(chan);
nouveau_bo_validate(chan, dst, NOUVEAU_BO_WR);
continue;
}
nr = MIN2(count, nr - 1);
nr = MIN2(nr, NV04_PFIFO_MAX_PACKET_LEN);
BEGIN_RING_NI(chan, RING_MF(DATA), nr);
OUT_RINGp (chan, src, nr);
src += nr;
count -= nr;
}
}
static void
instance_step(struct nv50_context *nv50, struct instance *a)
{
struct nouveau_channel *chan = nv50->screen->tesla->channel;
struct nouveau_grobj *tesla = nv50->screen->tesla;
int i;
for (i = 0; i < nv50->vtxelt->num_elements; i++) {
if (!a[i].divisor)
continue;
BEGIN_RING(chan, tesla,
NV50TCL_VERTEX_ARRAY_START_HIGH(i), 2);
OUT_RELOCh(chan, a[i].bo, a[i].delta, NOUVEAU_BO_RD |
NOUVEAU_BO_VRAM | NOUVEAU_BO_GART);
OUT_RELOCl(chan, a[i].bo, a[i].delta, NOUVEAU_BO_RD |
NOUVEAU_BO_VRAM | NOUVEAU_BO_GART);
if (++a[i].step == a[i].divisor) {
a[i].step = 0;
a[i].delta += a[i].stride;
}
}
}
void
nvc0_m2mf_push_linear(struct nouveau_context *nv,
struct nouveau_bo *dst, unsigned offset, unsigned domain,
unsigned size, const void *data)
{
struct nouveau_channel *chan = nv->screen->channel;
uint32_t *src = (uint32_t *)data;
unsigned count = (size + 3) / 4;
while (count) {
unsigned nr;
MARK_RING (chan, 16, 2);
nr = AVAIL_RING(chan) - 9;
nr = MIN2(count, nr);
nr = MIN2(nr, NV04_PFIFO_MAX_PACKET_LEN);
BEGIN_RING(chan, RING_MF(OFFSET_OUT_HIGH), 2);
OUT_RELOCh(chan, dst, offset, domain | NOUVEAU_BO_WR);
OUT_RELOCl(chan, dst, offset, domain | NOUVEAU_BO_WR);
BEGIN_RING(chan, RING_MF(LINE_LENGTH_IN), 2);
OUT_RING (chan, nr * 4);
OUT_RING (chan, 1);
BEGIN_RING(chan, RING_MF(EXEC), 1);
OUT_RING (chan, 0x100111);
/* must not be interrupted (trap on QUERY fence, 0x50 works however) */
BEGIN_RING_NI(chan, RING_MF(DATA), nr);
OUT_RINGp (chan, src, nr);
count -= nr;
src += nr;
offset += nr * 4;
}
}
void
nvc0_cb_push(struct nouveau_context *nv,
struct nouveau_bo *bo, unsigned domain,
unsigned base, unsigned size,
unsigned offset, unsigned words, const uint32_t *data)
{
struct nouveau_channel *chan = nv->screen->channel;
assert(!(offset & 3));
size = align(size, 0x100);
MARK_RING (chan, 16, 2);
BEGIN_RING(chan, RING_3D(CB_SIZE), 3);
OUT_RING (chan, size);
OUT_RELOCh(chan, bo, base, domain | NOUVEAU_BO_WR);
OUT_RELOCl(chan, bo, base, domain | NOUVEAU_BO_WR);
while (words) {
unsigned nr = AVAIL_RING(chan);
nr = MIN2(nr, words);
nr = MIN2(nr, NV04_PFIFO_MAX_PACKET_LEN - 1);
BEGIN_RING_1I(chan, RING_3D(CB_POS), nr + 1);
OUT_RING (chan, offset);
OUT_RINGp (chan, data, nr);
words -= nr;
data += nr;
offset += nr * 4;
if (words) {
MARK_RING(chan, 6, 1);
nouveau_bo_validate(chan, bo, domain | NOUVEAU_BO_WR);
}
}
}
static inline Bool
NVAccelDownloadM2MF(PixmapPtr pspix, int x, int y, int w, int h,
char *dst, unsigned dst_pitch)
{
ScrnInfoPtr pScrn = xf86Screens[pspix->drawable.pScreen->myNum];
NVPtr pNv = NVPTR(pScrn);
struct nouveau_channel *chan = pNv->chan;
struct nouveau_grobj *m2mf = pNv->NvMemFormat;
struct nouveau_bo *bo = nouveau_pixmap_bo(pspix);
unsigned cpp = pspix->drawable.bitsPerPixel / 8;
unsigned line_len = w * cpp;
unsigned src_offset = 0, src_pitch = 0, linear = 0;
/* Maximum DMA transfer */
unsigned line_count = pNv->GART->size / line_len;
if (!nv50_style_tiled_pixmap(pspix)) {
linear = 1;
src_pitch = exaGetPixmapPitch(pspix);
src_offset += (y * src_pitch) + (x * cpp);
}
/* HW limitations */
if (line_count > 2047)
line_count = 2047;
while (h) {
int i;
char *src;
if (line_count > h)
line_count = h;
if (MARK_RING(chan, 32, 6))
return FALSE;
BEGIN_RING(chan, m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_DMA_BUFFER_IN, 2);
if (OUT_RELOCo(chan, bo, NOUVEAU_BO_GART | NOUVEAU_BO_VRAM |
NOUVEAU_BO_RD) ||
OUT_RELOCo(chan, pNv->GART, NOUVEAU_BO_GART |
NOUVEAU_BO_WR)) {
MARK_UNDO(chan);
return FALSE;
}
if (pNv->Architecture >= NV_ARCH_50) {
if (!linear) {
BEGIN_RING(chan, m2mf, NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_IN, 7);
OUT_RING (chan, 0);
OUT_RING (chan, bo->tile_mode << 4);
OUT_RING (chan, pspix->drawable.width * cpp);
OUT_RING (chan, pspix->drawable.height);
OUT_RING (chan, 1);
OUT_RING (chan, 0);
OUT_RING (chan, (y << 16) | (x * cpp));
} else {
BEGIN_RING(chan, m2mf, NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_IN, 1);
OUT_RING (chan, 1);
}
BEGIN_RING(chan, m2mf, NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_OUT, 1);
OUT_RING (chan, 1);
BEGIN_RING(chan, m2mf, NV50_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN_HIGH, 2);
if (OUT_RELOCh(chan, bo, src_offset, NOUVEAU_BO_GART |
NOUVEAU_BO_VRAM | NOUVEAU_BO_RD) ||
OUT_RELOCh(chan, pNv->GART, 0, NOUVEAU_BO_GART |
NOUVEAU_BO_WR)) {
MARK_UNDO(chan);
return FALSE;
}
}
BEGIN_RING(chan, m2mf,
NV04_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 8);
if (OUT_RELOCl(chan, bo, src_offset, NOUVEAU_BO_GART |
NOUVEAU_BO_VRAM | NOUVEAU_BO_RD) ||
OUT_RELOCl(chan, pNv->GART, 0, NOUVEAU_BO_GART |
NOUVEAU_BO_WR)) {
MARK_UNDO(chan);
return FALSE;
}
OUT_RING (chan, src_pitch);
OUT_RING (chan, line_len);
OUT_RING (chan, line_len);
OUT_RING (chan, line_count);
OUT_RING (chan, (1<<8)|1);
OUT_RING (chan, 0);
if (nouveau_bo_map(pNv->GART, NOUVEAU_BO_RD)) {
MARK_UNDO(chan);
return FALSE;
}
src = pNv->GART->map;
if (dst_pitch == line_len) {
memcpy(dst, src, dst_pitch * line_count);
dst += dst_pitch * line_count;
} else {
for (i = 0; i < line_count; i++) {
memcpy(dst, src, line_len);
src += line_len;
dst += dst_pitch;
}
}
nouveau_bo_unmap(pNv->GART);
if (linear)
src_offset += line_count * src_pitch;
h -= line_count;
y += line_count;
}
return TRUE;
}
static int
nv50_2d_texture_set(struct nouveau_channel *chan, int dst,
struct nv50_miptree *mt, unsigned level, unsigned layer)
{
struct nouveau_bo *bo = mt->base.bo;
uint32_t width, height, depth;
uint32_t format;
uint32_t mthd = dst ? NV50_2D_DST_FORMAT : NV50_2D_SRC_FORMAT;
uint32_t flags = mt->base.domain | (dst ? NOUVEAU_BO_WR : NOUVEAU_BO_RD);
uint32_t offset = mt->level[level].offset;
format = nv50_2d_format(mt->base.base.format);
if (!format) {
NOUVEAU_ERR("invalid/unsupported surface format: %s\n",
util_format_name(mt->base.base.format));
return 1;
}
width = u_minify(mt->base.base.width0, level) << mt->ms_x;
height = u_minify(mt->base.base.height0, level) << mt->ms_y;
offset = mt->level[level].offset;
if (!mt->layout_3d) {
offset += mt->layer_stride * layer;
depth = 1;
layer = 0;
} else {
depth = u_minify(mt->base.base.depth0, level);
}
if (!(bo->tile_flags & NOUVEAU_BO_TILE_LAYOUT_MASK)) {
BEGIN_RING(chan, RING_2D_(mthd), 2);
OUT_RING (chan, format);
OUT_RING (chan, 1);
BEGIN_RING(chan, RING_2D_(mthd + 0x14), 5);
OUT_RING (chan, mt->level[level].pitch);
OUT_RING (chan, width);
OUT_RING (chan, height);
OUT_RELOCh(chan, bo, offset, flags);
OUT_RELOCl(chan, bo, offset, flags);
} else {
BEGIN_RING(chan, RING_2D_(mthd), 5);
OUT_RING (chan, format);
OUT_RING (chan, 0);
OUT_RING (chan, mt->level[level].tile_mode << 4);
OUT_RING (chan, depth);
OUT_RING (chan, layer);
BEGIN_RING(chan, RING_2D_(mthd + 0x18), 4);
OUT_RING (chan, width);
OUT_RING (chan, height);
OUT_RELOCh(chan, bo, offset, flags);
OUT_RELOCl(chan, bo, offset, flags);
}
#if 0
if (dst) {
BEGIN_RING(chan, RING_2D_(NV50_2D_CLIP_X), 4);
OUT_RING (chan, 0);
OUT_RING (chan, 0);
OUT_RING (chan, width);
OUT_RING (chan, height);
}
#endif
return 0;
}
void
nv50_upload_sifc(struct nv50_context *nv50,
struct nouveau_bo *bo, unsigned dst_offset, unsigned reloc,
unsigned dst_format, int dst_w, int dst_h, int dst_pitch,
void *src, unsigned src_format, int src_pitch,
int x, int y, int w, int h, int cpp)
{
struct nouveau_channel *chan = nv50->screen->base.channel;
struct nouveau_grobj *eng2d = nv50->screen->eng2d;
unsigned line_dwords = (w * cpp + 3) / 4;
reloc |= NOUVEAU_BO_WR;
MARK_RING (chan, 32, 2); /* flush on lack of space or relocs */
if (bo->tile_flags) {
BEGIN_RING(chan, eng2d, NV50_2D_DST_FORMAT, 5);
OUT_RING (chan, dst_format);
OUT_RING (chan, 0);
OUT_RING (chan, bo->tile_mode << 4);
OUT_RING (chan, 1);
OUT_RING (chan, 0);
} else {
BEGIN_RING(chan, eng2d, NV50_2D_DST_FORMAT, 2);
OUT_RING (chan, dst_format);
OUT_RING (chan, 1);
BEGIN_RING(chan, eng2d, NV50_2D_DST_PITCH, 1);
OUT_RING (chan, dst_pitch);
}
BEGIN_RING(chan, eng2d, NV50_2D_DST_WIDTH, 4);
OUT_RING (chan, dst_w);
OUT_RING (chan, dst_h);
OUT_RELOCh(chan, bo, dst_offset, reloc);
OUT_RELOCl(chan, bo, dst_offset, reloc);
/* NV50_2D_OPERATION_SRCCOPY assumed already set */
BEGIN_RING(chan, eng2d, NV50_2D_SIFC_BITMAP_ENABLE, 2);
OUT_RING (chan, 0);
OUT_RING (chan, src_format);
BEGIN_RING(chan, eng2d, NV50_2D_SIFC_WIDTH, 10);
OUT_RING (chan, w);
OUT_RING (chan, h);
OUT_RING (chan, 0);
OUT_RING (chan, 1);
OUT_RING (chan, 0);
OUT_RING (chan, 1);
OUT_RING (chan, 0);
OUT_RING (chan, x);
OUT_RING (chan, 0);
OUT_RING (chan, y);
while (h--) {
const uint32_t *p = src;
unsigned count = line_dwords;
while (count) {
unsigned nr = MIN2(count, 1792);
if (AVAIL_RING(chan) <= nr) {
FIRE_RING (chan);
BEGIN_RING(chan, eng2d,
NV50_2D_DST_ADDRESS_HIGH, 2);
OUT_RELOCh(chan, bo, dst_offset, reloc);
OUT_RELOCl(chan, bo, dst_offset, reloc);
}
assert(AVAIL_RING(chan) > nr);
BEGIN_RING(chan, eng2d,
NV50_2D_SIFC_DATA | (2 << 29), nr);
OUT_RINGp (chan, p, nr);
p += nr;
count -= nr;
}
src = (uint8_t *) src + src_pitch;
}
}
static void
nv50_transfer_rect_m2mf(struct pipe_screen *pscreen,
struct nouveau_bo *src_bo, unsigned src_offset,
int src_pitch, unsigned src_tile_mode,
int sx, int sy, int sz, int sw, int sh, int sd,
struct nouveau_bo *dst_bo, unsigned dst_offset,
int dst_pitch, unsigned dst_tile_mode,
int dx, int dy, int dz, int dw, int dh, int dd,
int cpp, int width, int height,
unsigned src_reloc, unsigned dst_reloc)
{
struct nv50_screen *screen = nv50_screen(pscreen);
struct nouveau_channel *chan = screen->m2mf->channel;
struct nouveau_grobj *m2mf = screen->m2mf;
src_reloc |= NOUVEAU_BO_RD;
dst_reloc |= NOUVEAU_BO_WR;
WAIT_RING (chan, 14);
if (!src_bo->tile_flags) {
BEGIN_RING(chan, m2mf,
NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_IN, 1);
OUT_RING (chan, 1);
BEGIN_RING(chan, m2mf,
NV04_MEMORY_TO_MEMORY_FORMAT_PITCH_IN, 1);
OUT_RING (chan, src_pitch);
src_offset += (sy * src_pitch) + (sx * cpp);
} else {
BEGIN_RING(chan, m2mf,
NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_IN, 6);
OUT_RING (chan, 0);
OUT_RING (chan, src_tile_mode << 4);
OUT_RING (chan, sw * cpp);
OUT_RING (chan, sh);
OUT_RING (chan, sd);
OUT_RING (chan, sz); /* copying only 1 zslice per call */
}
if (!dst_bo->tile_flags) {
BEGIN_RING(chan, m2mf,
NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_OUT, 1);
OUT_RING (chan, 1);
BEGIN_RING(chan, m2mf,
NV04_MEMORY_TO_MEMORY_FORMAT_PITCH_OUT, 1);
OUT_RING (chan, dst_pitch);
dst_offset += (dy * dst_pitch) + (dx * cpp);
} else {
BEGIN_RING(chan, m2mf,
NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_OUT, 6);
OUT_RING (chan, 0);
OUT_RING (chan, dst_tile_mode << 4);
OUT_RING (chan, dw * cpp);
OUT_RING (chan, dh);
OUT_RING (chan, dd);
OUT_RING (chan, dz); /* copying only 1 zslice per call */
}
while (height) {
int line_count = height > 2047 ? 2047 : height;
MARK_RING (chan, 15, 4); /* flush on lack of space or relocs */
BEGIN_RING(chan, m2mf,
NV50_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN_HIGH, 2);
OUT_RELOCh(chan, src_bo, src_offset, src_reloc);
OUT_RELOCh(chan, dst_bo, dst_offset, dst_reloc);
BEGIN_RING(chan, m2mf,
NV04_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 2);
OUT_RELOCl(chan, src_bo, src_offset, src_reloc);
OUT_RELOCl(chan, dst_bo, dst_offset, dst_reloc);
if (src_bo->tile_flags) {
BEGIN_RING(chan, m2mf,
NV50_MEMORY_TO_MEMORY_FORMAT_TILING_POSITION_IN, 1);
OUT_RING (chan, (sy << 16) | (sx * cpp));
} else {
src_offset += (line_count * src_pitch);
}
if (dst_bo->tile_flags) {
BEGIN_RING(chan, m2mf,
NV50_MEMORY_TO_MEMORY_FORMAT_TILING_POSITION_OUT, 1);
OUT_RING (chan, (dy << 16) | (dx * cpp));
} else {
dst_offset += (line_count * dst_pitch);
}
BEGIN_RING(chan, m2mf,
NV04_MEMORY_TO_MEMORY_FORMAT_LINE_LENGTH_IN, 4);
OUT_RING (chan, width * cpp);
OUT_RING (chan, line_count);
OUT_RING (chan, 0x00000101);
OUT_RING (chan, 0);
FIRE_RING (chan);
height -= line_count;
sy += line_count;
dy += line_count;
}
}
void
nvc0_m2mf_transfer_rect(struct pipe_screen *pscreen,
const struct nv50_m2mf_rect *dst,
const struct nv50_m2mf_rect *src,
uint32_t nblocksx, uint32_t nblocksy)
{
struct nouveau_channel *chan = nouveau_screen(pscreen)->channel;
const int cpp = dst->cpp;
uint32_t src_ofst = src->base;
uint32_t dst_ofst = dst->base;
uint32_t height = nblocksy;
uint32_t sy = src->y;
uint32_t dy = dst->y;
uint32_t exec = (1 << 20);
assert(dst->cpp == src->cpp);
if (nouveau_bo_tile_layout(src->bo)) {
BEGIN_RING(chan, RING_MF(TILING_MODE_IN), 5);
OUT_RING (chan, src->tile_mode);
OUT_RING (chan, src->width * cpp);
OUT_RING (chan, src->height);
OUT_RING (chan, src->depth);
OUT_RING (chan, src->z);
} else {
src_ofst += src->y * src->pitch + src->x * cpp;
BEGIN_RING(chan, RING_MF(PITCH_IN), 1);
OUT_RING (chan, src->width * cpp);
exec |= NVC0_M2MF_EXEC_LINEAR_IN;
}
if (nouveau_bo_tile_layout(dst->bo)) {
BEGIN_RING(chan, RING_MF(TILING_MODE_OUT), 5);
OUT_RING (chan, dst->tile_mode);
OUT_RING (chan, dst->width * cpp);
OUT_RING (chan, dst->height);
OUT_RING (chan, dst->depth);
OUT_RING (chan, dst->z);
} else {
dst_ofst += dst->y * dst->pitch + dst->x * cpp;
BEGIN_RING(chan, RING_MF(PITCH_OUT), 1);
OUT_RING (chan, dst->width * cpp);
exec |= NVC0_M2MF_EXEC_LINEAR_OUT;
}
while (height) {
int line_count = height > 2047 ? 2047 : height;
MARK_RING (chan, 17, 4);
BEGIN_RING(chan, RING_MF(OFFSET_IN_HIGH), 2);
OUT_RELOCh(chan, src->bo, src_ofst, src->domain | NOUVEAU_BO_RD);
OUT_RELOCl(chan, src->bo, src_ofst, src->domain | NOUVEAU_BO_RD);
BEGIN_RING(chan, RING_MF(OFFSET_OUT_HIGH), 2);
OUT_RELOCh(chan, dst->bo, dst_ofst, dst->domain | NOUVEAU_BO_WR);
OUT_RELOCl(chan, dst->bo, dst_ofst, dst->domain | NOUVEAU_BO_WR);
if (!(exec & NVC0_M2MF_EXEC_LINEAR_IN)) {
BEGIN_RING(chan, RING_MF(TILING_POSITION_IN_X), 2);
OUT_RING (chan, src->x * cpp);
OUT_RING (chan, sy);
} else {
src_ofst += line_count * src->pitch;
}
if (!(exec & NVC0_M2MF_EXEC_LINEAR_OUT)) {
BEGIN_RING(chan, RING_MF(TILING_POSITION_OUT_X), 2);
OUT_RING (chan, dst->x * cpp);
OUT_RING (chan, dy);
} else {
dst_ofst += line_count * dst->pitch;
}
BEGIN_RING(chan, RING_MF(LINE_LENGTH_IN), 2);
OUT_RING (chan, nblocksx * cpp);
OUT_RING (chan, line_count);
BEGIN_RING(chan, RING_MF(EXEC), 1);
OUT_RING (chan, exec);
height -= line_count;
sy += line_count;
dy += line_count;
}
}
