{
struct nouveau_channel *chan = nvc0->screen->base.channel;
struct pipe_framebuffer_state *fb = &nvc0->framebuffer;
- struct nvc0_surface *sf = nvc0_surface(fb->zsbuf);
- struct nvc0_miptree *mt = nvc0_miptree(sf->base.texture);
+ struct nv50_surface *sf = nv50_surface(fb->zsbuf);
+ struct nv50_miptree *mt = nv50_miptree(sf->base.texture);
struct nouveau_bo *bo = mt->base.bo;
uint32_t size;
uint32_t offset = align(mt->total_size, 1 << 17);
struct nouveau_channel *chan = nvc0->screen->base.channel;
struct pipe_framebuffer_state *fb = &nvc0->framebuffer;
unsigned i;
+ unsigned ms_mode = NVC0_3D_MULTISAMPLE_MODE_MS1;
boolean serialize = FALSE;
nvc0_bufctx_reset(nvc0, NVC0_BUFCTX_FRAME);
MARK_RING(chan, 9 * fb->nr_cbufs, 2 * fb->nr_cbufs);
for (i = 0; i < fb->nr_cbufs; ++i) {
- struct nvc0_miptree *mt = nvc0_miptree(fb->cbufs[i]->texture);
- struct nvc0_surface *sf = nvc0_surface(fb->cbufs[i]);
- struct nouveau_bo *bo = mt->base.bo;
- uint32_t offset = sf->offset;
+ struct nv50_surface *sf = nv50_surface(fb->cbufs[i]);
+ struct nv04_resource *res = nv04_resource(sf->base.texture);
+ struct nouveau_bo *bo = res->bo;
+ uint32_t offset = sf->offset + res->offset;
BEGIN_RING(chan, RING_3D(RT_ADDRESS_HIGH(i)), 9);
- OUT_RELOCh(chan, bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR);
- OUT_RELOCl(chan, bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR);
- OUT_RING (chan, sf->width);
- OUT_RING (chan, sf->height);
- OUT_RING (chan, nvc0_format_table[sf->base.format].rt);
- OUT_RING (chan, (mt->layout_3d << 16) |
- mt->level[sf->base.u.tex.level].tile_mode);
- OUT_RING (chan, sf->base.u.tex.first_layer + sf->depth);
- OUT_RING (chan, mt->layer_stride >> 2);
- OUT_RING (chan, sf->base.u.tex.first_layer);
-
- if (mt->base.status & NOUVEAU_BUFFER_STATUS_GPU_READING)
+ OUT_RELOCh(chan, res->bo, offset, res->domain | NOUVEAU_BO_RDWR);
+ OUT_RELOCl(chan, res->bo, offset, res->domain | NOUVEAU_BO_RDWR);
+ if (likely(nouveau_bo_tile_layout(bo))) {
+ struct nv50_miptree *mt = nv50_miptree(sf->base.texture);
+
+ assert(sf->base.texture->target != PIPE_BUFFER);
+
+ OUT_RING(chan, sf->width);
+ OUT_RING(chan, sf->height);
+ OUT_RING(chan, nvc0_format_table[sf->base.format].rt);
+ OUT_RING(chan, (mt->layout_3d << 16) |
+ mt->level[sf->base.u.tex.level].tile_mode);
+ OUT_RING(chan, sf->base.u.tex.first_layer + sf->depth);
+ OUT_RING(chan, mt->layer_stride >> 2);
+ OUT_RING(chan, sf->base.u.tex.first_layer);
+
+ ms_mode = mt->ms_mode;
+ } else {
+ if (res->base.target == PIPE_BUFFER) {
+ OUT_RING(chan, 262144);
+ OUT_RING(chan, 1);
+ } else {
+ OUT_RING(chan, nv50_miptree(sf->base.texture)->level[0].pitch);
+ OUT_RING(chan, sf->height);
+ }
+ OUT_RING(chan, nvc0_format_table[sf->base.format].rt);
+ OUT_RING(chan, 1 << 12);
+ OUT_RING(chan, 1);
+ OUT_RING(chan, 0);
+ OUT_RING(chan, 0);
+
+ nvc0_resource_fence(res, NOUVEAU_BO_WR);
+
+ assert(!fb->zsbuf);
+ }
+
+ if (res->status & NOUVEAU_BUFFER_STATUS_GPU_READING)
serialize = TRUE;
- mt->base.status |= NOUVEAU_BUFFER_STATUS_GPU_WRITING;
- mt->base.status &= ~NOUVEAU_BUFFER_STATUS_GPU_READING;
+ res->status |= NOUVEAU_BUFFER_STATUS_GPU_WRITING;
+ res->status &= ~NOUVEAU_BUFFER_STATUS_GPU_READING;
/* only register for writing, otherwise we'd always serialize here */
- nvc0_bufctx_add_resident(nvc0, NVC0_BUFCTX_FRAME, &mt->base,
- NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+ nvc0_bufctx_add_resident(nvc0, NVC0_BUFCTX_FRAME, res,
+ res->domain | NOUVEAU_BO_WR);
}
if (fb->zsbuf) {
- struct nvc0_miptree *mt = nvc0_miptree(fb->zsbuf->texture);
- struct nvc0_surface *sf = nvc0_surface(fb->zsbuf);
+ struct nv50_miptree *mt = nv50_miptree(fb->zsbuf->texture);
+ struct nv50_surface *sf = nv50_surface(fb->zsbuf);
struct nouveau_bo *bo = mt->base.bo;
int unk = mt->base.base.target == PIPE_TEXTURE_2D;
uint32_t offset = sf->offset;
BEGIN_RING(chan, RING_3D(ZETA_BASE_LAYER), 1);
OUT_RING (chan, sf->base.u.tex.first_layer);
+ ms_mode = mt->ms_mode;
+
if (mt->base.status & NOUVEAU_BUFFER_STATUS_GPU_READING)
serialize = TRUE;
mt->base.status |= NOUVEAU_BUFFER_STATUS_GPU_WRITING;
OUT_RING (chan, 0);
}
+ IMMED_RING(chan, RING_3D(MULTISAMPLE_MODE), ms_mode);
+
if (serialize) {
BEGIN_RING(chan, RING_3D(SERIALIZE), 1);
OUT_RING (chan, 0);
OUT_RINGf (chan, zmax);
}
+static INLINE void
+nvc0_upload_uclip_planes(struct nvc0_context *nvc0)
+{
+ struct nouveau_channel *chan = nvc0->screen->base.channel;
+ struct nouveau_bo *bo = nvc0->screen->uniforms;
+
+ MARK_RING (chan, 6 + PIPE_MAX_CLIP_PLANES * 4, 2);
+ BEGIN_RING(chan, RING_3D(CB_SIZE), 3);
+ OUT_RING (chan, 256);
+ OUT_RELOCh(chan, bo, 5 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+ OUT_RELOCl(chan, bo, 5 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+ BEGIN_RING_1I(chan, RING_3D(CB_POS), PIPE_MAX_CLIP_PLANES * 4 + 1);
+ OUT_RING (chan, 0);
+ OUT_RINGp (chan, &nvc0->clip.ucp[0][0], PIPE_MAX_CLIP_PLANES * 4);
+}
+
+static INLINE void
+nvc0_check_program_ucps(struct nvc0_context *nvc0,
+ struct nvc0_program *vp, uint8_t mask)
+{
+ const unsigned n = util_logbase2(mask) + 1;
+
+ if (vp->vp.num_ucps >= n)
+ return;
+ nvc0_program_destroy(nvc0, vp);
+
+ vp->vp.num_ucps = n;
+ if (likely(vp == nvc0->vertprog))
+ nvc0_vertprog_validate(nvc0);
+ else
+ if (likely(vp == nvc0->gmtyprog))
+ nvc0_vertprog_validate(nvc0);
+ else
+ nvc0_tevlprog_validate(nvc0);
+}
+
static void
nvc0_validate_clip(struct nvc0_context *nvc0)
{
struct nouveau_channel *chan = nvc0->screen->base.channel;
- uint32_t clip;
-
- if (nvc0->clip.depth_clamp) {
- clip =
- NVC0_3D_VIEW_VOLUME_CLIP_CTRL_UNK1_UNK1 |
- NVC0_3D_VIEW_VOLUME_CLIP_CTRL_DEPTH_CLAMP_NEAR |
- NVC0_3D_VIEW_VOLUME_CLIP_CTRL_DEPTH_CLAMP_FAR |
- NVC0_3D_VIEW_VOLUME_CLIP_CTRL_UNK12_UNK2;
- } else {
- clip = NVC0_3D_VIEW_VOLUME_CLIP_CTRL_UNK1_UNK1;
+ struct nvc0_program *vp;
+ uint8_t clip_enable = nvc0->rast->pipe.clip_plane_enable;
+
+ if (nvc0->dirty & NVC0_NEW_CLIP)
+ nvc0_upload_uclip_planes(nvc0);
+
+ vp = nvc0->gmtyprog;
+ if (!vp) {
+ vp = nvc0->tevlprog;
+ if (!vp)
+ vp = nvc0->vertprog;
}
- BEGIN_RING(chan, RING_3D(VIEW_VOLUME_CLIP_CTRL), 1);
- OUT_RING (chan, clip);
-
- if (nvc0->clip.nr) {
- struct nouveau_bo *bo = nvc0->screen->uniforms;
-
- MARK_RING (chan, 6 + nvc0->clip.nr * 4, 2);
- BEGIN_RING(chan, RING_3D(CB_SIZE), 3);
- OUT_RING (chan, 256);
- OUT_RELOCh(chan, bo, 5 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
- OUT_RELOCl(chan, bo, 5 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
- BEGIN_RING_1I(chan, RING_3D(CB_POS), nvc0->clip.nr * 4 + 1);
- OUT_RING (chan, 0);
- OUT_RINGp (chan, &nvc0->clip.ucp[0][0], nvc0->clip.nr * 4);
-
- BEGIN_RING(chan, RING_3D(VP_CLIP_DISTANCE_ENABLE), 1);
- OUT_RING (chan, (1 << nvc0->clip.nr) - 1);
- } else {
- IMMED_RING(chan, RING_3D(VP_CLIP_DISTANCE_ENABLE), 0);
+ if (clip_enable && vp->vp.num_ucps < PIPE_MAX_CLIP_PLANES)
+ nvc0_check_program_ucps(nvc0, vp, clip_enable);
+
+ clip_enable &= vp->vp.clip_enable;
+
+ if (nvc0->state.clip_enable != clip_enable) {
+ nvc0->state.clip_enable = clip_enable;
+ IMMED_RING(chan, RING_3D(CLIP_DISTANCE_ENABLE), clip_enable);
+ }
+ if (nvc0->state.clip_mode != vp->vp.clip_mode) {
+ nvc0->state.clip_mode = vp->vp.clip_mode;
+ BEGIN_RING(chan, RING_3D(CLIP_DISTANCE_MODE), 1);
+ OUT_RING (chan, vp->vp.clip_mode);
}
}
while (nvc0->constbuf_dirty[s]) {
unsigned base = 0;
- unsigned offset = 0, words = 0;
+ unsigned words = 0;
boolean rebind = TRUE;
i = ffs(nvc0->constbuf_dirty[s]) - 1;
}
if (!nouveau_resource_mapped_by_gpu(&res->base)) {
- if (i == 0) {
+ if (i == 0 && (res->status & NOUVEAU_BUFFER_STATUS_USER_MEMORY)) {
base = s << 16;
bo = nvc0->screen->uniforms;
else
nvc0->state.uniform_buffer_bound[s] =
align(res->base.width0, 0x100);
+
+ words = res->base.width0 / 4;
} else {
+ nouveau_buffer_migrate(&nvc0->base, res, NOUVEAU_BO_VRAM);
bo = res->bo;
+ base = res->offset;
}
-#if 0
- nvc0_m2mf_push_linear(nvc0, bo, NOUVEAU_BO_VRAM,
- base, res->base.width0, res->data);
- BEGIN_RING(chan, RING_3D_(0x021c), 1);
- OUT_RING (chan, 0x1111);
-#else
- words = res->base.width0 / 4;
-#endif
} else {
bo = res->bo;
+ base = res->offset;
if (i == 0)
nvc0->state.uniform_buffer_bound[s] = 0;
}
OUT_RING (chan, (i << 4) | 1);
}
- while (words) {
- unsigned nr = AVAIL_RING(chan);
-
- if (nr < 16) {
- FIRE_RING(chan);
- continue;
- }
- nr = MIN2(MIN2(nr - 6, words), NV04_PFIFO_MAX_PACKET_LEN - 1);
-
- MARK_RING (chan, nr + 5, 2);
- BEGIN_RING(chan, RING_3D(CB_SIZE), 3);
- OUT_RING (chan, align(res->base.width0, 0x100));
- OUT_RELOCh(chan, bo, base, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
- OUT_RELOCl(chan, bo, base, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
- BEGIN_RING_1I(chan, RING_3D(CB_POS), nr + 1);
- OUT_RING (chan, offset);
- OUT_RINGp (chan, &res->data[offset], nr);
-
- offset += nr * 4;
- words -= nr;
- }
+ if (words)
+ nvc0_cb_push(&nvc0->base,
+ bo, NOUVEAU_BO_VRAM, base, res->base.width0,
+ 0, words, (const uint32_t *)res->data);
}
}
}
+static void
+nvc0_validate_sample_mask(struct nvc0_context *nvc0)
+{
+ struct nouveau_channel *chan = nvc0->screen->base.channel;
+
+ unsigned mask[4] =
+ {
+ nvc0->sample_mask & 0xffff,
+ nvc0->sample_mask & 0xffff,
+ nvc0->sample_mask & 0xffff,
+ nvc0->sample_mask & 0xffff
+ };
+
+ BEGIN_RING(chan, RING_3D(MSAA_MASK(0)), 4);
+ OUT_RING (chan, mask[0]);
+ OUT_RING (chan, mask[1]);
+ OUT_RING (chan, mask[2]);
+ OUT_RING (chan, mask[3]);
+ BEGIN_RING(chan, RING_3D(SAMPLE_SHADING), 1);
+ OUT_RING (chan, 0x01);
+}
+
static void
nvc0_validate_derived_1(struct nvc0_context *nvc0)
{
struct nouveau_channel *chan = nvc0->screen->base.channel;
boolean early_z;
+ boolean rasterizer_discard;
early_z = nvc0->fragprog->fp.early_z && !nvc0->zsa->pipe.alpha.enabled;
nvc0->state.early_z = early_z;
IMMED_RING(chan, RING_3D(EARLY_FRAGMENT_TESTS), early_z);
}
+
+ rasterizer_discard = (!nvc0->fragprog || !nvc0->fragprog->hdr[18]) &&
+ !nvc0->zsa->pipe.depth.enabled && !nvc0->zsa->pipe.stencil[0].enabled;
+ rasterizer_discard = rasterizer_discard ||
+ nvc0->rast->pipe.rasterizer_discard;
+
+ if (rasterizer_discard != nvc0->state.rasterizer_discard) {
+ nvc0->state.rasterizer_discard = rasterizer_discard;
+ IMMED_RING(chan, RING_3D(RASTERIZE_ENABLE), !rasterizer_discard);
+ }
}
static void
if (!ctx_to->blend)
ctx_to->dirty &= ~NVC0_NEW_BLEND;
if (!ctx_to->rast)
- ctx_to->dirty &= ~NVC0_NEW_RASTERIZER;
+ ctx_to->dirty &= ~(NVC0_NEW_RASTERIZER | NVC0_NEW_SCISSOR);
if (!ctx_to->zsa)
ctx_to->dirty &= ~NVC0_NEW_ZSA;
- ctx_to->screen->base.channel->user_private = ctx_to->screen->cur_ctx =
- ctx_to;
+ ctx_to->screen->cur_ctx = ctx_to;
}
static struct state_validate {
{ nvc0_validate_fb, NVC0_NEW_FRAMEBUFFER },
{ nvc0_validate_blend, NVC0_NEW_BLEND },
{ nvc0_validate_zsa, NVC0_NEW_ZSA },
+ { nvc0_validate_sample_mask, NVC0_NEW_SAMPLE_MASK },
{ nvc0_validate_rasterizer, NVC0_NEW_RASTERIZER },
{ nvc0_validate_blend_colour, NVC0_NEW_BLEND_COLOUR },
{ nvc0_validate_stencil_ref, NVC0_NEW_STENCIL_REF },
{ nvc0_validate_stipple, NVC0_NEW_STIPPLE },
{ nvc0_validate_scissor, NVC0_NEW_SCISSOR | NVC0_NEW_RASTERIZER },
{ nvc0_validate_viewport, NVC0_NEW_VIEWPORT },
- { nvc0_validate_clip, NVC0_NEW_CLIP },
{ nvc0_vertprog_validate, NVC0_NEW_VERTPROG },
{ nvc0_tctlprog_validate, NVC0_NEW_TCTLPROG },
{ nvc0_tevlprog_validate, NVC0_NEW_TEVLPROG },
{ nvc0_gmtyprog_validate, NVC0_NEW_GMTYPROG },
{ nvc0_fragprog_validate, NVC0_NEW_FRAGPROG },
- { nvc0_validate_derived_1, NVC0_NEW_FRAGPROG | NVC0_NEW_ZSA },
+ { nvc0_validate_derived_1, NVC0_NEW_FRAGPROG | NVC0_NEW_ZSA |
+ NVC0_NEW_RASTERIZER },
+ { nvc0_validate_clip, NVC0_NEW_CLIP | NVC0_NEW_RASTERIZER |
+ NVC0_NEW_VERTPROG |
+ NVC0_NEW_TEVLPROG |
+ NVC0_NEW_GMTYPROG },
{ nvc0_constbufs_validate, NVC0_NEW_CONSTBUF },
{ nvc0_validate_textures, NVC0_NEW_TEXTURES },
{ nvc0_validate_samplers, NVC0_NEW_SAMPLERS },
{ nvc0_vertex_arrays_validate, NVC0_NEW_VERTEX | NVC0_NEW_ARRAYS },
- { nvc0_tfb_validate, NVC0_NEW_TFB | NVC0_NEW_TFB_BUFFERS }
+ { nvc0_tfb_validate, NVC0_NEW_TFB_TARGETS | NVC0_NEW_GMTYPROG }
};
#define validate_list_len (sizeof(validate_list) / sizeof(validate_list[0]))
boolean
-nvc0_state_validate(struct nvc0_context *nvc0)
+nvc0_state_validate(struct nvc0_context *nvc0, uint32_t mask, unsigned words)
{
+ uint32_t state_mask;
unsigned i;
if (nvc0->screen->cur_ctx != nvc0)
nvc0_switch_pipe_context(nvc0);
- if (nvc0->dirty) {
+ state_mask = nvc0->dirty & mask;
+
+ if (state_mask) {
for (i = 0; i < validate_list_len; ++i) {
struct state_validate *validate = &validate_list[i];
- if (nvc0->dirty & validate->states)
+ if (state_mask & validate->states)
validate->func(nvc0);
}
- nvc0->dirty = 0;
+ nvc0->dirty &= ~state_mask;
}
+ MARK_RING(nvc0->screen->base.channel, words, 0);
+
nvc0_bufctx_emit_relocs(nvc0);
return TRUE;