/* This happens rather often with DTD9/st. */
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_pushbuf *push = nv->pushbuf;
NOUVEAU_DRV_STAT(nv->screen, constbuf_upload_count, 1);
NOUVEAU_DRV_STAT(nv->screen, constbuf_upload_bytes, words * 4);
assert(!(offset & 3));
size = align(size, 0x100);
BEGIN_NVC0(push, NVC0_3D(CB_SIZE), 3);
PUSH_DATA (push, size);
PUSH_DATAh(push, bo->offset + base);
PUSH_DATA (push, bo->offset + base);
while (words) {
unsigned nr = PUSH_AVAIL(push);
nr = MIN2(nr, words);
nr = MIN2(nr, NV04_PFIFO_MAX_PACKET_LEN - 1);
PUSH_SPACE(push, nr + 2);
PUSH_REFN (push, bo, NOUVEAU_BO_WR | domain);
BEGIN_1IC0(push, NVC0_3D(CB_POS), nr + 1);
PUSH_DATA (push, offset);
PUSH_DATAp(push, data, nr);
words -= nr;
data += nr;
offset += nr * 4;
}
}
static void
nvc0_render_condition(struct pipe_context *pipe,
struct pipe_query *pq,
boolean condition, uint mode)
{
struct nvc0_context *nvc0 = nvc0_context(pipe);
struct nouveau_pushbuf *push = nvc0->base.pushbuf;
struct nvc0_query *q;
uint32_t cond;
boolean negated = FALSE;
boolean wait =
mode != PIPE_RENDER_COND_NO_WAIT &&
mode != PIPE_RENDER_COND_BY_REGION_NO_WAIT;
nvc0->cond_query = pq;
nvc0->cond_cond = condition;
nvc0->cond_mode = mode;
if (!pq) {
PUSH_SPACE(push, 1);
IMMED_NVC0(push, NVC0_3D(COND_MODE), NVC0_3D_COND_MODE_ALWAYS);
return;
}
q = nvc0_query(pq);
/* NOTE: comparison of 2 queries only works if both have completed */
switch (q->type) {
case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
cond = negated ? NVC0_3D_COND_MODE_EQUAL :
NVC0_3D_COND_MODE_NOT_EQUAL;
wait = TRUE;
break;
case PIPE_QUERY_OCCLUSION_COUNTER:
case PIPE_QUERY_OCCLUSION_PREDICATE:
if (likely(!negated)) {
if (unlikely(q->nesting))
cond = wait ? NVC0_3D_COND_MODE_NOT_EQUAL :
NVC0_3D_COND_MODE_ALWAYS;
else
cond = NVC0_3D_COND_MODE_RES_NON_ZERO;
} else {
cond = wait ? NVC0_3D_COND_MODE_EQUAL : NVC0_3D_COND_MODE_ALWAYS;
}
break;
default:
assert(!"render condition query not a predicate");
mode = NVC0_3D_COND_MODE_ALWAYS;
break;
}
if (wait)
nvc0_query_fifo_wait(push, pq);
PUSH_SPACE(push, 4);
PUSH_REFN (push, q->bo, NOUVEAU_BO_GART | NOUVEAU_BO_RD);
BEGIN_NVC0(push, NVC0_3D(COND_ADDRESS_HIGH), 3);
PUSH_DATAh(push, q->bo->offset + q->offset);
PUSH_DATA (push, q->bo->offset + q->offset);
PUSH_DATA (push, cond);
}
void
nv50_query_pushbuf_submit(struct nouveau_pushbuf *push,
struct pipe_query *pq, unsigned result_offset)
{
struct nv50_query *q = nv50_query(pq);
/* XXX: does this exist ? */
#define NV50_IB_ENTRY_1_NO_PREFETCH (0 << (31 - 8))
PUSH_REFN(push, q->bo, NOUVEAU_BO_RD | NOUVEAU_BO_GART);
nouveau_pushbuf_space(push, 0, 0, 1);
nouveau_pushbuf_data(push, q->bo, q->offset + result_offset, 4 |
NV50_IB_ENTRY_1_NO_PREFETCH);
}
void
nv84_query_fifo_wait(struct nouveau_pushbuf *push, struct pipe_query *pq)
{
struct nv50_query *q = nv50_query(pq);
unsigned offset = q->offset;
PUSH_SPACE(push, 5);
PUSH_REFN (push, q->bo, NOUVEAU_BO_GART | NOUVEAU_BO_RD);
BEGIN_NV04(push, SUBC_3D(NV84_SUBCHAN_SEMAPHORE_ADDRESS_HIGH), 4);
PUSH_DATAh(push, q->bo->offset + offset);
PUSH_DATA (push, q->bo->offset + offset);
PUSH_DATA (push, q->sequence);
PUSH_DATA (push, NV84_SUBCHAN_SEMAPHORE_TRIGGER_ACQUIRE_EQUAL);
}
static void
nv50_query_get(struct nouveau_pushbuf *push, struct nv50_query *q,
unsigned offset, uint32_t get)
{
offset += q->offset;
PUSH_SPACE(push, 5);
PUSH_REFN (push, q->bo, NOUVEAU_BO_GART | NOUVEAU_BO_WR);
BEGIN_NV04(push, NV50_3D(QUERY_ADDRESS_HIGH), 4);
PUSH_DATAh(push, q->bo->offset + offset);
PUSH_DATA (push, q->bo->offset + offset);
PUSH_DATA (push, q->sequence);
PUSH_DATA (push, get);
}
static void
nvc0_clear_render_target(struct pipe_context *pipe,
struct pipe_surface *dst,
const union pipe_color_union *color,
unsigned dstx, unsigned dsty,
unsigned width, unsigned height)
{
struct nvc0_context *nvc0 = nvc0_context(pipe);
struct nouveau_pushbuf *push = nvc0->base.pushbuf;
struct nv50_surface *sf = nv50_surface(dst);
struct nv04_resource *res = nv04_resource(sf->base.texture);
unsigned z;
if (!PUSH_SPACE(push, 32 + sf->depth))
return;
PUSH_REFN (push, res->bo, res->domain | NOUVEAU_BO_WR);
BEGIN_NVC0(push, NVC0_3D(CLEAR_COLOR(0)), 4);
PUSH_DATAf(push, color->f[0]);
PUSH_DATAf(push, color->f[1]);
PUSH_DATAf(push, color->f[2]);
PUSH_DATAf(push, color->f[3]);
BEGIN_NVC0(push, NVC0_3D(SCREEN_SCISSOR_HORIZ), 2);
PUSH_DATA (push, ( width << 16) | dstx);
PUSH_DATA (push, (height << 16) | dsty);
BEGIN_NVC0(push, NVC0_3D(RT_CONTROL), 1);
PUSH_DATA (push, 1);
BEGIN_NVC0(push, NVC0_3D(RT_ADDRESS_HIGH(0)), 9);
PUSH_DATAh(push, res->address + sf->offset);
PUSH_DATA (push, res->address + sf->offset);
if (likely(nouveau_bo_memtype(res->bo))) {
struct nv50_miptree *mt = nv50_miptree(dst->texture);
PUSH_DATA(push, sf->width);
PUSH_DATA(push, sf->height);
PUSH_DATA(push, nvc0_format_table[dst->format].rt);
PUSH_DATA(push, (mt->layout_3d << 16) |
mt->level[sf->base.u.tex.level].tile_mode);
PUSH_DATA(push, dst->u.tex.first_layer + sf->depth);
PUSH_DATA(push, mt->layer_stride >> 2);
PUSH_DATA(push, dst->u.tex.first_layer);
} else {
if (res->base.target == PIPE_BUFFER) {
void
nvc0_query_fifo_wait(struct nouveau_pushbuf *push, struct pipe_query *pq)
{
struct nvc0_query *q = nvc0_query(pq);
unsigned offset = q->offset;
if (q->type == PIPE_QUERY_SO_OVERFLOW_PREDICATE) offset += 0x20;
PUSH_SPACE(push, 5);
PUSH_REFN (push, q->bo, NOUVEAU_BO_GART | NOUVEAU_BO_RD);
BEGIN_NVC0(push, SUBC_3D(NV84_SUBCHAN_SEMAPHORE_ADDRESS_HIGH), 4);
PUSH_DATAh(push, q->bo->offset + offset);
PUSH_DATA (push, q->bo->offset + offset);
PUSH_DATA (push, q->sequence);
PUSH_DATA (push, (1 << 12) |
NV84_SUBCHAN_SEMAPHORE_TRIGGER_ACQUIRE_EQUAL);
}
static void
nv50_render_condition(struct pipe_context *pipe,
struct pipe_query *pq,
boolean condition, uint mode)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_pushbuf *push = nv50->base.pushbuf;
struct nv50_query *q;
uint32_t cond;
boolean wait =
mode != PIPE_RENDER_COND_NO_WAIT &&
mode != PIPE_RENDER_COND_BY_REGION_NO_WAIT;
if (!pq) {
cond = NV50_3D_COND_MODE_ALWAYS;
}
else {
q = nv50_query(pq);
/* NOTE: comparison of 2 queries only works if both have completed */
switch (q->type) {
case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
cond = condition ? NV50_3D_COND_MODE_EQUAL :
NV50_3D_COND_MODE_NOT_EQUAL;
wait = TRUE;
break;
case PIPE_QUERY_OCCLUSION_COUNTER:
case PIPE_QUERY_OCCLUSION_PREDICATE:
if (likely(!condition)) {
/* XXX: Placeholder, handle nesting here if available */
if (unlikely(false))
cond = wait ? NV50_3D_COND_MODE_NOT_EQUAL :
NV50_3D_COND_MODE_ALWAYS;
else
cond = NV50_3D_COND_MODE_RES_NON_ZERO;
} else {
cond = wait ? NV50_3D_COND_MODE_EQUAL : NV50_3D_COND_MODE_ALWAYS;
}
break;
default:
assert(!"render condition query not a predicate");
cond = NV50_3D_COND_MODE_ALWAYS;
break;
}
}
nv50->cond_query = pq;
nv50->cond_cond = condition;
nv50->cond_condmode = cond;
nv50->cond_mode = mode;
if (!pq) {
PUSH_SPACE(push, 2);
BEGIN_NV04(push, NV50_3D(COND_MODE), 1);
PUSH_DATA (push, cond);
return;
}
PUSH_SPACE(push, 9);
if (wait) {
BEGIN_NV04(push, SUBC_3D(NV50_GRAPH_SERIALIZE), 1);
PUSH_DATA (push, 0);
}
PUSH_REFN (push, q->bo, NOUVEAU_BO_GART | NOUVEAU_BO_RD);
BEGIN_NV04(push, NV50_3D(COND_ADDRESS_HIGH), 3);
PUSH_DATAh(push, q->bo->offset + q->offset);
PUSH_DATA (push, q->bo->offset + q->offset);
PUSH_DATA (push, cond);
BEGIN_NV04(push, NV50_2D(COND_ADDRESS_HIGH), 2);
PUSH_DATAh(push, q->bo->offset + q->offset);
PUSH_DATA (push, q->bo->offset + q->offset);
}
void
nvc0_launch_grid(struct pipe_context *pipe, const struct pipe_grid_info *info)
{
struct nvc0_context *nvc0 = nvc0_context(pipe);
struct nouveau_pushbuf *push = nvc0->base.pushbuf;
struct nvc0_program *cp = nvc0->compprog;
int ret;
ret = !nvc0_state_validate_cp(nvc0, ~0);
if (ret) {
NOUVEAU_ERR("Failed to launch grid !\n");
return;
}
nvc0_compute_upload_input(nvc0, info);
BEGIN_NVC0(push, NVC0_CP(CP_START_ID), 1);
PUSH_DATA (push, nvc0_program_symbol_offset(cp, info->pc));
BEGIN_NVC0(push, NVC0_CP(LOCAL_POS_ALLOC), 3);
PUSH_DATA (push, (cp->hdr[1] & 0xfffff0) + align(cp->cp.lmem_size, 0x10));
PUSH_DATA (push, 0);
PUSH_DATA (push, 0x800); /* WARP_CSTACK_SIZE */
BEGIN_NVC0(push, NVC0_CP(SHARED_SIZE), 3);
PUSH_DATA (push, align(cp->cp.smem_size, 0x100));
PUSH_DATA (push, info->block[0] * info->block[1] * info->block[2]);
PUSH_DATA (push, cp->num_barriers);
BEGIN_NVC0(push, NVC0_CP(CP_GPR_ALLOC), 1);
PUSH_DATA (push, cp->num_gprs);
/* launch preliminary setup */
BEGIN_NVC0(push, NVC0_CP(GRIDID), 1);
PUSH_DATA (push, 0x1);
BEGIN_NVC0(push, SUBC_CP(0x036c), 1);
PUSH_DATA (push, 0);
BEGIN_NVC0(push, NVC0_CP(FLUSH), 1);
PUSH_DATA (push, NVC0_COMPUTE_FLUSH_GLOBAL | NVC0_COMPUTE_FLUSH_UNK8);
/* block setup */
BEGIN_NVC0(push, NVC0_CP(BLOCKDIM_YX), 2);
PUSH_DATA (push, (info->block[1] << 16) | info->block[0]);
PUSH_DATA (push, info->block[2]);
if (unlikely(info->indirect)) {
struct nv04_resource *res = nv04_resource(info->indirect);
uint32_t offset = res->offset + info->indirect_offset;
unsigned macro = NVC0_CP_MACRO_LAUNCH_GRID_INDIRECT;
nouveau_pushbuf_space(push, 16, 0, 1);
PUSH_REFN(push, res->bo, NOUVEAU_BO_RD | res->domain);
PUSH_DATA(push, NVC0_FIFO_PKHDR_1I(1, macro, 3));
nouveau_pushbuf_data(push, res->bo, offset,
NVC0_IB_ENTRY_1_NO_PREFETCH | 3 * 4);
} else {
/* grid setup */
BEGIN_NVC0(push, NVC0_CP(GRIDDIM_YX), 2);
PUSH_DATA (push, (info->grid[1] << 16) | info->grid[0]);
PUSH_DATA (push, info->grid[2]);
/* kernel launching */
BEGIN_NVC0(push, NVC0_CP(COMPUTE_BEGIN), 1);
PUSH_DATA (push, 0);
BEGIN_NVC0(push, SUBC_CP(0x0a08), 1);
PUSH_DATA (push, 0);
BEGIN_NVC0(push, NVC0_CP(LAUNCH), 1);
PUSH_DATA (push, 0x1000);
BEGIN_NVC0(push, NVC0_CP(COMPUTE_END), 1);
PUSH_DATA (push, 0);
BEGIN_NVC0(push, SUBC_CP(0x0360), 1);
PUSH_DATA (push, 0x1);
}
/* TODO: Not sure if this is really necessary. */
nvc0_compute_invalidate_surfaces(nvc0, 5);
nouveau_bufctx_reset(nvc0->bufctx_cp, NVC0_BIND_CP_SUF);
nvc0->dirty_cp |= NVC0_NEW_CP_SURFACES;
nvc0->images_dirty[5] |= nvc0->images_valid[5];
}
struct pipe_video_decoder *
nvc0_create_decoder(struct pipe_context *context,
enum pipe_video_profile profile,
enum pipe_video_entrypoint entrypoint,
enum pipe_video_chroma_format chroma_format,
unsigned width, unsigned height, unsigned max_references,
bool chunked_decode)
{
struct nouveau_screen *screen = &((struct nvc0_context *)context)->screen->base;
struct nvc0_decoder *dec;
struct nouveau_pushbuf **push;
union nouveau_bo_config cfg;
bool kepler = screen->device->chipset >= 0xe0;
cfg.nvc0.tile_mode = 0x10;
cfg.nvc0.memtype = 0xfe;
int ret, i;
uint32_t codec = 1, ppp_codec = 3;
uint32_t timeout;
u32 tmp_size = 0;
if (getenv("XVMC_VL"))
return vl_create_decoder(context, profile, entrypoint,
chroma_format, width, height,
max_references, chunked_decode);
if (entrypoint != PIPE_VIDEO_ENTRYPOINT_BITSTREAM) {
debug_printf("%x\n", entrypoint);
return NULL;
}
dec = CALLOC_STRUCT(nvc0_decoder);
if (!dec)
return NULL;
dec->client = screen->client;
if (!kepler) {
dec->bsp_idx = 5;
dec->vp_idx = 6;
dec->ppp_idx = 7;
} else {
dec->bsp_idx = 2;
dec->vp_idx = 2;
dec->ppp_idx = 2;
}
for (i = 0; i < 3; ++i)
if (i && !kepler) {
dec->channel[i] = dec->channel[0];
dec->pushbuf[i] = dec->pushbuf[0];
} else {
void *data;
u32 size;
struct nvc0_fifo nvc0_args = {};
struct nve0_fifo nve0_args = {};
if (!kepler) {
size = sizeof(nvc0_args);
data = &nvc0_args;
} else {
unsigned engine[] = {
NVE0_FIFO_ENGINE_BSP,
NVE0_FIFO_ENGINE_VP,
NVE0_FIFO_ENGINE_PPP
};
nve0_args.engine = engine[i];
size = sizeof(nve0_args);
data = &nve0_args;
}
ret = nouveau_object_new(&screen->device->object, 0,
NOUVEAU_FIFO_CHANNEL_CLASS,
data, size, &dec->channel[i]);
if (!ret)
ret = nouveau_pushbuf_new(screen->client, dec->channel[i], 4,
32 * 1024, true, &dec->pushbuf[i]);
if (ret)
break;
}
push = dec->pushbuf;
if (!kepler) {
if (!ret)
ret = nouveau_object_new(dec->channel[0], 0x390b1, 0x90b1, NULL, 0, &dec->bsp);
if (!ret)
ret = nouveau_object_new(dec->channel[1], 0x190b2, 0x90b2, NULL, 0, &dec->vp);
if (!ret)
ret = nouveau_object_new(dec->channel[2], 0x290b3, 0x90b3, NULL, 0, &dec->ppp);
} else {
if (!ret)
ret = nouveau_object_new(dec->channel[0], 0x95b1, 0x95b1, NULL, 0, &dec->bsp);
if (!ret)
ret = nouveau_object_new(dec->channel[1], 0x95b2, 0x95b2, NULL, 0, &dec->vp);
if (!ret)
ret = nouveau_object_new(dec->channel[2], 0x90b3, 0x90b3, NULL, 0, &dec->ppp);
}
if (ret)
goto fail;
BEGIN_NVC0(push[0], SUBC_BSP(NV01_SUBCHAN_OBJECT), 1);
PUSH_DATA (push[0], dec->bsp->handle);
BEGIN_NVC0(push[1], SUBC_VP(NV01_SUBCHAN_OBJECT), 1);
PUSH_DATA (push[1], dec->vp->handle);
BEGIN_NVC0(push[2], SUBC_PPP(NV01_SUBCHAN_OBJECT), 1);
PUSH_DATA (push[2], dec->ppp->handle);
dec->base.context = context;
dec->base.profile = profile;
dec->base.entrypoint = entrypoint;
dec->base.chroma_format = chroma_format;
dec->base.width = width;
dec->base.height = height;
dec->base.max_references = max_references;
dec->base.destroy = nvc0_decoder_destroy;
dec->base.flush = nvc0_decoder_flush;
dec->base.decode_bitstream = nvc0_decoder_decode_bitstream;
dec->base.begin_frame = nvc0_decoder_begin_frame;
dec->base.end_frame = nvc0_decoder_end_frame;
for (i = 0; i < NVC0_VIDEO_QDEPTH && !ret; ++i)
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
0, 1 << 20, &cfg, &dec->bsp_bo[i]);
if (!ret)
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
0x100, 4 << 20, &cfg, &dec->inter_bo[0]);
if (!ret) {
if (!kepler)
nouveau_bo_ref(dec->inter_bo[0], &dec->inter_bo[1]);
else
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
0x100, dec->inter_bo[0]->size, &cfg,
&dec->inter_bo[1]);
}
if (ret)
goto fail;
switch (u_reduce_video_profile(profile)) {
case PIPE_VIDEO_CODEC_MPEG12: {
codec = 1;
assert(max_references <= 2);
break;
}
case PIPE_VIDEO_CODEC_MPEG4: {
codec = 4;
tmp_size = mb(height)*16 * mb(width)*16;
assert(max_references <= 2);
break;
}
case PIPE_VIDEO_CODEC_VC1: {
ppp_codec = codec = 2;
tmp_size = mb(height)*16 * mb(width)*16;
assert(max_references <= 2);
break;
}
case PIPE_VIDEO_CODEC_MPEG4_AVC: {
codec = 3;
dec->tmp_stride = 16 * mb_half(width) * nvc0_video_align(height) * 3 / 2;
tmp_size = dec->tmp_stride * (max_references + 1);
assert(max_references <= 16);
break;
}
default:
fprintf(stderr, "invalid codec\n");
goto fail;
}
if (screen->device->chipset < 0xd0) {
int fd;
char path[PATH_MAX];
ssize_t r;
uint32_t *end, endval;
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
0x4000, &cfg, &dec->fw_bo);
if (!ret)
ret = nouveau_bo_map(dec->fw_bo, NOUVEAU_BO_WR, dec->client);
if (ret)
goto fail;
nvc0_video_getpath(profile, path);
fd = open(path, O_RDONLY | O_CLOEXEC);
if (fd < 0) {
fprintf(stderr, "opening firmware file %s failed: %m\n", path);
goto fw_fail;
}
r = read(fd, dec->fw_bo->map, 0x4000);
close(fd);
if (r < 0) {
fprintf(stderr, "reading firmware file %s failed: %m\n", path);
goto fw_fail;
}
if (r == 0x4000) {
fprintf(stderr, "firmware file %s too large!\n", path);
goto fw_fail;
}
if (r & 0xff) {
fprintf(stderr, "firmware file %s wrong size!\n", path);
goto fw_fail;
}
end = dec->fw_bo->map + r - 4;
endval = *end;
while (endval == *end)
end--;
r = (intptr_t)end - (intptr_t)dec->fw_bo->map + 4;
switch (u_reduce_video_profile(profile)) {
case PIPE_VIDEO_CODEC_MPEG12: {
assert((r & 0xff) == 0xe0);
dec->fw_sizes = (0x2e0<<16) | (r - 0x2e0);
break;
}
case PIPE_VIDEO_CODEC_MPEG4: {
assert((r & 0xff) == 0xe0);
dec->fw_sizes = (0x2e0<<16) | (r - 0x2e0);
break;
}
case PIPE_VIDEO_CODEC_VC1: {
assert((r & 0xff) == 0xac);
dec->fw_sizes = (0x3ac<<16) | (r - 0x3ac);
break;
}
case PIPE_VIDEO_CODEC_MPEG4_AVC: {
assert((r & 0xff) == 0x70);
dec->fw_sizes = (0x370<<16) | (r - 0x370);
break;
}
default:
goto fw_fail;
}
munmap(dec->fw_bo->map, dec->fw_bo->size);
dec->fw_bo->map = NULL;
}
if (codec != 3) {
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
0x400, &cfg, &dec->bitplane_bo);
if (ret)
goto fail;
}
dec->ref_stride = mb(width)*16 * (mb_half(height)*32 + nvc0_video_align(height)/2);
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
dec->ref_stride * (max_references+2) + tmp_size,
&cfg, &dec->ref_bo);
if (ret)
goto fail;
timeout = 0;
BEGIN_NVC0(push[0], SUBC_BSP(0x200), 2);
PUSH_DATA (push[0], codec);
PUSH_DATA (push[0], timeout);
BEGIN_NVC0(push[1], SUBC_VP(0x200), 2);
PUSH_DATA (push[1], codec);
PUSH_DATA (push[1], timeout);
BEGIN_NVC0(push[2], SUBC_PPP(0x200), 2);
PUSH_DATA (push[2], ppp_codec);
PUSH_DATA (push[2], timeout);
++dec->fence_seq;
#if NVC0_DEBUG_FENCE
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_GART|NOUVEAU_BO_MAP,
0, 0x1000, NULL, &dec->fence_bo);
if (ret)
goto fail;
nouveau_bo_map(dec->fence_bo, NOUVEAU_BO_RDWR, screen->client);
dec->fence_map = dec->fence_bo->map;
dec->fence_map[0] = dec->fence_map[4] = dec->fence_map[8] = 0;
dec->comm = (struct comm *)(dec->fence_map + (COMM_OFFSET/sizeof(*dec->fence_map)));
/* So lets test if the fence is working? */
nouveau_pushbuf_space(push[0], 6, 1, 0);
PUSH_REFN (push[0], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
BEGIN_NVC0(push[0], SUBC_BSP(0x240), 3);
PUSH_DATAh(push[0], dec->fence_bo->offset);
PUSH_DATA (push[0], dec->fence_bo->offset);
PUSH_DATA (push[0], dec->fence_seq);
BEGIN_NVC0(push[0], SUBC_BSP(0x304), 1);
PUSH_DATA (push[0], 0);
PUSH_KICK (push[0]);
nouveau_pushbuf_space(push[1], 6, 1, 0);
PUSH_REFN (push[1], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
BEGIN_NVC0(push[1], SUBC_VP(0x240), 3);
PUSH_DATAh(push[1], (dec->fence_bo->offset + 0x10));
PUSH_DATA (push[1], (dec->fence_bo->offset + 0x10));
PUSH_DATA (push[1], dec->fence_seq);
BEGIN_NVC0(push[1], SUBC_VP(0x304), 1);
PUSH_DATA (push[1], 0);
PUSH_KICK (push[1]);
nouveau_pushbuf_space(push[2], 6, 1, 0);
PUSH_REFN (push[2], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
BEGIN_NVC0(push[2], SUBC_PPP(0x240), 3);
PUSH_DATAh(push[2], (dec->fence_bo->offset + 0x20));
PUSH_DATA (push[2], (dec->fence_bo->offset + 0x20));
PUSH_DATA (push[2], dec->fence_seq);
BEGIN_NVC0(push[2], SUBC_PPP(0x304), 1);
PUSH_DATA (push[2], 0);
PUSH_KICK (push[2]);
usleep(100);
while (dec->fence_seq > dec->fence_map[0] ||
dec->fence_seq > dec->fence_map[4] ||
dec->fence_seq > dec->fence_map[8]) {
debug_printf("%u: %u %u %u\n", dec->fence_seq, dec->fence_map[0], dec->fence_map[4], dec->fence_map[8]);
usleep(100);
}
debug_printf("%u: %u %u %u\n", dec->fence_seq, dec->fence_map[0], dec->fence_map[4], dec->fence_map[8]);
#endif
return &dec->base;
fw_fail:
debug_printf("Cannot create decoder without firmware..\n");
nvc0_decoder_destroy(&dec->base);
return NULL;
fail:
debug_printf("Creation failed: %s (%i)\n", strerror(-ret), ret);
nvc0_decoder_destroy(&dec->base);
return NULL;
}
static void
nv50_render_condition(struct pipe_context *pipe,
struct pipe_query *pq,
boolean condition, enum pipe_render_cond_flag mode)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_pushbuf *push = nv50->base.pushbuf;
struct nv50_query *q = nv50_query(pq);
struct nv50_hw_query *hq = nv50_hw_query(q);
uint32_t cond;
bool wait =
mode != PIPE_RENDER_COND_NO_WAIT &&
mode != PIPE_RENDER_COND_BY_REGION_NO_WAIT;
if (!pq) {
cond = NV50_3D_COND_MODE_ALWAYS;
}
else {
/* NOTE: comparison of 2 queries only works if both have completed */
switch (q->type) {
case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
cond = condition ? NV50_3D_COND_MODE_EQUAL :
NV50_3D_COND_MODE_NOT_EQUAL;
wait = true;
break;
case PIPE_QUERY_OCCLUSION_COUNTER:
case PIPE_QUERY_OCCLUSION_PREDICATE:
case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
if (hq->state == NV50_HW_QUERY_STATE_READY)
wait = true;
if (likely(!condition)) {
cond = wait ? NV50_3D_COND_MODE_NOT_EQUAL : NV50_3D_COND_MODE_ALWAYS;
} else {
cond = wait ? NV50_3D_COND_MODE_EQUAL : NV50_3D_COND_MODE_ALWAYS;
}
break;
default:
assert(!"render condition query not a predicate");
cond = NV50_3D_COND_MODE_ALWAYS;
break;
}
}
nv50->cond_query = pq;
nv50->cond_cond = condition;
nv50->cond_condmode = cond;
nv50->cond_mode = mode;
if (!pq) {
PUSH_SPACE(push, 2);
BEGIN_NV04(push, NV50_3D(COND_MODE), 1);
PUSH_DATA (push, cond);
return;
}
PUSH_SPACE(push, 9);
if (wait && hq->state != NV50_HW_QUERY_STATE_READY) {
BEGIN_NV04(push, SUBC_3D(NV50_GRAPH_SERIALIZE), 1);
PUSH_DATA (push, 0);
}
PUSH_REFN (push, hq->bo, NOUVEAU_BO_GART | NOUVEAU_BO_RD);
BEGIN_NV04(push, NV50_3D(COND_ADDRESS_HIGH), 3);
PUSH_DATAh(push, hq->bo->offset + hq->offset);
PUSH_DATA (push, hq->bo->offset + hq->offset);
PUSH_DATA (push, cond);
BEGIN_NV04(push, NV50_2D(COND_ADDRESS_HIGH), 2);
PUSH_DATAh(push, hq->bo->offset + hq->offset);
PUSH_DATA (push, hq->bo->offset + hq->offset);
}
struct pipe_video_codec *
nvc0_create_decoder(struct pipe_context *context,
const struct pipe_video_codec *templ)
{
struct nouveau_screen *screen = &((struct nvc0_context *)context)->screen->base;
struct nouveau_vp3_decoder *dec;
struct nouveau_pushbuf **push;
union nouveau_bo_config cfg;
bool kepler = screen->device->chipset >= 0xe0;
cfg.nvc0.tile_mode = 0x10;
cfg.nvc0.memtype = 0xfe;
int ret, i;
uint32_t codec = 1, ppp_codec = 3;
uint32_t timeout;
u32 tmp_size = 0;
if (getenv("XVMC_VL"))
return vl_create_decoder(context, templ);
if (templ->entrypoint != PIPE_VIDEO_ENTRYPOINT_BITSTREAM) {
debug_printf("%x\n", templ->entrypoint);
return NULL;
}
dec = CALLOC_STRUCT(nouveau_vp3_decoder);
if (!dec)
return NULL;
dec->client = screen->client;
dec->base = *templ;
nouveau_vp3_decoder_init_common(&dec->base);
if (!kepler) {
dec->bsp_idx = 5;
dec->vp_idx = 6;
dec->ppp_idx = 7;
} else {
dec->bsp_idx = 2;
dec->vp_idx = 2;
dec->ppp_idx = 2;
}
for (i = 0; i < 3; ++i)
if (i && !kepler) {
dec->channel[i] = dec->channel[0];
dec->pushbuf[i] = dec->pushbuf[0];
} else {
void *data;
u32 size;
struct nvc0_fifo nvc0_args = {};
struct nve0_fifo nve0_args = {};
if (!kepler) {
size = sizeof(nvc0_args);
data = &nvc0_args;
} else {
unsigned engine[] = {
NVE0_FIFO_ENGINE_BSP,
NVE0_FIFO_ENGINE_VP,
NVE0_FIFO_ENGINE_PPP
};
nve0_args.engine = engine[i];
size = sizeof(nve0_args);
data = &nve0_args;
}
ret = nouveau_object_new(&screen->device->object, 0,
NOUVEAU_FIFO_CHANNEL_CLASS,
data, size, &dec->channel[i]);
if (!ret)
ret = nouveau_pushbuf_new(screen->client, dec->channel[i], 4,
32 * 1024, true, &dec->pushbuf[i]);
if (ret)
break;
}
push = dec->pushbuf;
if (!kepler) {
if (!ret)
ret = nouveau_object_new(dec->channel[0], 0x390b1, 0x90b1, NULL, 0, &dec->bsp);
if (!ret)
ret = nouveau_object_new(dec->channel[1], 0x190b2, 0x90b2, NULL, 0, &dec->vp);
if (!ret)
ret = nouveau_object_new(dec->channel[2], 0x290b3, 0x90b3, NULL, 0, &dec->ppp);
} else {
if (!ret)
ret = nouveau_object_new(dec->channel[0], 0x95b1, 0x95b1, NULL, 0, &dec->bsp);
if (!ret)
ret = nouveau_object_new(dec->channel[1], 0x95b2, 0x95b2, NULL, 0, &dec->vp);
if (!ret)
ret = nouveau_object_new(dec->channel[2], 0x90b3, 0x90b3, NULL, 0, &dec->ppp);
}
if (ret)
goto fail;
BEGIN_NVC0(push[0], SUBC_BSP(NV01_SUBCHAN_OBJECT), 1);
PUSH_DATA (push[0], dec->bsp->handle);
BEGIN_NVC0(push[1], SUBC_VP(NV01_SUBCHAN_OBJECT), 1);
PUSH_DATA (push[1], dec->vp->handle);
BEGIN_NVC0(push[2], SUBC_PPP(NV01_SUBCHAN_OBJECT), 1);
PUSH_DATA (push[2], dec->ppp->handle);
dec->base.context = context;
dec->base.decode_bitstream = nvc0_decoder_decode_bitstream;
for (i = 0; i < NOUVEAU_VP3_VIDEO_QDEPTH && !ret; ++i)
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
0, 1 << 20, &cfg, &dec->bsp_bo[i]);
if (!ret)
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
0x100, 4 << 20, &cfg, &dec->inter_bo[0]);
if (!ret) {
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM,
0x100, dec->inter_bo[0]->size, &cfg,
&dec->inter_bo[1]);
}
if (ret)
goto fail;
switch (u_reduce_video_profile(templ->profile)) {
case PIPE_VIDEO_FORMAT_MPEG12: {
codec = 1;
assert(templ->max_references <= 2);
break;
}
case PIPE_VIDEO_FORMAT_MPEG4: {
codec = 4;
tmp_size = mb(templ->height)*16 * mb(templ->width)*16;
assert(templ->max_references <= 2);
break;
}
case PIPE_VIDEO_FORMAT_VC1: {
ppp_codec = codec = 2;
tmp_size = mb(templ->height)*16 * mb(templ->width)*16;
assert(templ->max_references <= 2);
break;
}
case PIPE_VIDEO_FORMAT_MPEG4_AVC: {
codec = 3;
dec->tmp_stride = 16 * mb_half(templ->width) * nouveau_vp3_video_align(templ->height) * 3 / 2;
tmp_size = dec->tmp_stride * (templ->max_references + 1);
assert(templ->max_references <= 16);
break;
}
default:
fprintf(stderr, "invalid codec\n");
goto fail;
}
if (screen->device->chipset < 0xd0) {
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
0x4000, &cfg, &dec->fw_bo);
if (ret)
goto fail;
ret = nouveau_vp3_load_firmware(dec, templ->profile, screen->device->chipset);
if (ret)
goto fw_fail;
}
if (codec != 3) {
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
0x400, &cfg, &dec->bitplane_bo);
if (ret)
goto fail;
}
dec->ref_stride = mb(templ->width)*16 * (mb_half(templ->height)*32 + nouveau_vp3_video_align(templ->height)/2);
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_VRAM, 0,
dec->ref_stride * (templ->max_references+2) + tmp_size,
&cfg, &dec->ref_bo);
if (ret)
goto fail;
timeout = 0;
BEGIN_NVC0(push[0], SUBC_BSP(0x200), 2);
PUSH_DATA (push[0], codec);
PUSH_DATA (push[0], timeout);
BEGIN_NVC0(push[1], SUBC_VP(0x200), 2);
PUSH_DATA (push[1], codec);
PUSH_DATA (push[1], timeout);
BEGIN_NVC0(push[2], SUBC_PPP(0x200), 2);
PUSH_DATA (push[2], ppp_codec);
PUSH_DATA (push[2], timeout);
++dec->fence_seq;
#if NOUVEAU_VP3_DEBUG_FENCE
ret = nouveau_bo_new(screen->device, NOUVEAU_BO_GART|NOUVEAU_BO_MAP,
0, 0x1000, NULL, &dec->fence_bo);
if (ret)
goto fail;
nouveau_bo_map(dec->fence_bo, NOUVEAU_BO_RDWR, screen->client);
dec->fence_map = dec->fence_bo->map;
dec->fence_map[0] = dec->fence_map[4] = dec->fence_map[8] = 0;
dec->comm = (struct comm *)(dec->fence_map + (COMM_OFFSET/sizeof(*dec->fence_map)));
/* So lets test if the fence is working? */
nouveau_pushbuf_space(push[0], 6, 1, 0);
PUSH_REFN (push[0], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
BEGIN_NVC0(push[0], SUBC_BSP(0x240), 3);
PUSH_DATAh(push[0], dec->fence_bo->offset);
PUSH_DATA (push[0], dec->fence_bo->offset);
PUSH_DATA (push[0], dec->fence_seq);
BEGIN_NVC0(push[0], SUBC_BSP(0x304), 1);
PUSH_DATA (push[0], 0);
PUSH_KICK (push[0]);
nouveau_pushbuf_space(push[1], 6, 1, 0);
PUSH_REFN (push[1], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
BEGIN_NVC0(push[1], SUBC_VP(0x240), 3);
PUSH_DATAh(push[1], (dec->fence_bo->offset + 0x10));
PUSH_DATA (push[1], (dec->fence_bo->offset + 0x10));
PUSH_DATA (push[1], dec->fence_seq);
BEGIN_NVC0(push[1], SUBC_VP(0x304), 1);
PUSH_DATA (push[1], 0);
PUSH_KICK (push[1]);
nouveau_pushbuf_space(push[2], 6, 1, 0);
PUSH_REFN (push[2], dec->fence_bo, NOUVEAU_BO_GART|NOUVEAU_BO_RDWR);
BEGIN_NVC0(push[2], SUBC_PPP(0x240), 3);
PUSH_DATAh(push[2], (dec->fence_bo->offset + 0x20));
PUSH_DATA (push[2], (dec->fence_bo->offset + 0x20));
PUSH_DATA (push[2], dec->fence_seq);
BEGIN_NVC0(push[2], SUBC_PPP(0x304), 1);
PUSH_DATA (push[2], 0);
PUSH_KICK (push[2]);
usleep(100);
while (dec->fence_seq > dec->fence_map[0] ||
dec->fence_seq > dec->fence_map[4] ||
dec->fence_seq > dec->fence_map[8]) {
debug_printf("%u: %u %u %u\n", dec->fence_seq, dec->fence_map[0], dec->fence_map[4], dec->fence_map[8]);
usleep(100);
}
debug_printf("%u: %u %u %u\n", dec->fence_seq, dec->fence_map[0], dec->fence_map[4], dec->fence_map[8]);
#endif
return &dec->base;
fw_fail:
debug_printf("Cannot create decoder without firmware..\n");
dec->base.destroy(&dec->base);
return NULL;
fail:
debug_printf("Creation failed: %s (%i)\n", strerror(-ret), ret);
dec->base.destroy(&dec->base);
return NULL;
}
