+#include "util/u_math.h"
+
#include "nvc0_context.h"
-#include "os/os_time.h"
+#if 0
static void
nvc0_validate_zcull(struct nvc0_context *nvc0)
{
- struct nouveau_channel *chan = nvc0->screen->base.channel;
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
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);
else
width = fb->width;
- MARK_RING (chan, 23, 4);
- BEGIN_RING(chan, RING_3D_(0x1590), 1); /* ZCULL_REGION_INDEX (bits 0x3f) */
- OUT_RING (chan, 0);
- BEGIN_RING(chan, RING_3D_(0x07e8), 2); /* ZCULL_ADDRESS_A_HIGH */
- OUT_RELOCh(chan, bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR);
- OUT_RELOCl(chan, bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR);
+ BEGIN_NVC0(push, NVC0_3D(ZCULL_REGION), 1);
+ PUSH_DATA (push, 0);
+ BEGIN_NVC0(push, NVC0_3D(ZCULL_ADDRESS_HIGH), 2);
+ PUSH_DATAh(push, bo->offset + offset);
+ PUSH_DATA (push, bo->offset + offset);
offset += 1 << 17;
- BEGIN_RING(chan, RING_3D_(0x07f0), 2); /* ZCULL_ADDRESS_B_HIGH */
- OUT_RELOCh(chan, bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR);
- OUT_RELOCl(chan, bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR);
- BEGIN_RING(chan, RING_3D_(0x07e0), 2);
- OUT_RING (chan, size);
- OUT_RING (chan, size >> 16);
- BEGIN_RING(chan, RING_3D_(0x15c8), 1); /* bits 0x3 */
- OUT_RING (chan, 2);
- BEGIN_RING(chan, RING_3D_(0x07c0), 4); /* ZCULL dimensions */
- OUT_RING (chan, width);
- OUT_RING (chan, height);
- OUT_RING (chan, 1);
- OUT_RING (chan, 0);
- BEGIN_RING(chan, RING_3D_(0x15fc), 2);
- OUT_RING (chan, 0); /* bits 0xffff */
- OUT_RING (chan, 0); /* bits 0xffff */
- BEGIN_RING(chan, RING_3D_(0x1958), 1);
- OUT_RING (chan, 0); /* bits ~0 */
+ BEGIN_NVC0(push, NVC0_3D(ZCULL_LIMIT_HIGH), 2);
+ PUSH_DATAh(push, bo->offset + offset);
+ PUSH_DATA (push, bo->offset + offset);
+ BEGIN_NVC0(push, SUBC_3D(0x07e0), 2);
+ PUSH_DATA (push, size);
+ PUSH_DATA (push, size >> 16);
+ BEGIN_NVC0(push, SUBC_3D(0x15c8), 1); /* bits 0x3 */
+ PUSH_DATA (push, 2);
+ BEGIN_NVC0(push, NVC0_3D(ZCULL_WIDTH), 4);
+ PUSH_DATA (push, width);
+ PUSH_DATA (push, height);
+ PUSH_DATA (push, 1);
+ PUSH_DATA (push, 0);
+ BEGIN_NVC0(push, NVC0_3D(ZCULL_WINDOW_OFFSET_X), 2);
+ PUSH_DATA (push, 0);
+ PUSH_DATA (push, 0);
+ BEGIN_NVC0(push, NVC0_3D(ZCULL_INVALIDATE), 1);
+ PUSH_DATA (push, 0);
}
+#endif
static void
nvc0_validate_fb(struct nvc0_context *nvc0)
{
- struct nouveau_channel *chan = nvc0->screen->base.channel;
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
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);
+ nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_FB);
- BEGIN_RING(chan, RING_3D(RT_CONTROL), 1);
- OUT_RING (chan, (076543210 << 4) | fb->nr_cbufs);
- BEGIN_RING(chan, RING_3D(SCREEN_SCISSOR_HORIZ), 2);
- OUT_RING (chan, fb->width << 16);
- OUT_RING (chan, fb->height << 16);
-
- MARK_RING(chan, 9 * fb->nr_cbufs, 2 * fb->nr_cbufs);
+ BEGIN_NVC0(push, NVC0_3D(RT_CONTROL), 1);
+ PUSH_DATA (push, (076543210 << 4) | fb->nr_cbufs);
+ BEGIN_NVC0(push, NVC0_3D(SCREEN_SCISSOR_HORIZ), 2);
+ PUSH_DATA (push, fb->width << 16);
+ PUSH_DATA (push, fb->height << 16);
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;
-
- 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)
+ struct nv50_surface *sf = nv50_surface(fb->cbufs[i]);
+ struct nv04_resource *res = nv04_resource(sf->base.texture);
+ struct nouveau_bo *bo = res->bo;
+
+ BEGIN_NVC0(push, NVC0_3D(RT_ADDRESS_HIGH(i)), 9);
+ PUSH_DATAh(push, res->address + sf->offset);
+ PUSH_DATA (push, res->address + sf->offset);
+ if (likely(nouveau_bo_memtype(bo))) {
+ struct nv50_miptree *mt = nv50_miptree(sf->base.texture);
+
+ assert(sf->base.texture->target != PIPE_BUFFER);
+
+ PUSH_DATA(push, sf->width);
+ PUSH_DATA(push, sf->height);
+ PUSH_DATA(push, nvc0_format_table[sf->base.format].rt);
+ PUSH_DATA(push, (mt->layout_3d << 16) |
+ mt->level[sf->base.u.tex.level].tile_mode);
+ PUSH_DATA(push, sf->base.u.tex.first_layer + sf->depth);
+ PUSH_DATA(push, mt->layer_stride >> 2);
+ PUSH_DATA(push, sf->base.u.tex.first_layer);
+
+ ms_mode = mt->ms_mode;
+ } else {
+ if (res->base.target == PIPE_BUFFER) {
+ PUSH_DATA(push, 262144);
+ PUSH_DATA(push, 1);
+ } else {
+ PUSH_DATA(push, nv50_miptree(sf->base.texture)->level[0].pitch);
+ PUSH_DATA(push, sf->height);
+ }
+ PUSH_DATA(push, nvc0_format_table[sf->base.format].rt);
+ PUSH_DATA(push, 1 << 12);
+ PUSH_DATA(push, 1);
+ PUSH_DATA(push, 0);
+ PUSH_DATA(push, 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;
- nvc0_bufctx_add_resident(nvc0, NVC0_BUFCTX_FRAME, &mt->base,
- NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR);
+ /* only register for writing, otherwise we'd always serialize here */
+ BCTX_REFN(nvc0->bufctx_3d, FB, res, WR);
}
if (fb->zsbuf) {
- struct nvc0_miptree *mt = nvc0_miptree(fb->zsbuf->texture);
- struct nvc0_surface *sf = nvc0_surface(fb->zsbuf);
- struct nouveau_bo *bo = mt->base.bo;
+ struct nv50_miptree *mt = nv50_miptree(fb->zsbuf->texture);
+ struct nv50_surface *sf = nv50_surface(fb->zsbuf);
int unk = mt->base.base.target == PIPE_TEXTURE_2D;
- uint32_t offset = sf->offset;
-
- MARK_RING (chan, 12, 2);
- BEGIN_RING(chan, RING_3D(ZETA_ADDRESS_HIGH), 5);
- 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, nvc0_format_table[fb->zsbuf->format].rt);
- OUT_RING (chan, mt->level[sf->base.u.tex.level].tile_mode);
- OUT_RING (chan, mt->layer_stride >> 2);
- 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, (unk << 16) |
+
+ BEGIN_NVC0(push, NVC0_3D(ZETA_ADDRESS_HIGH), 5);
+ PUSH_DATAh(push, mt->base.address + sf->offset);
+ PUSH_DATA (push, mt->base.address + sf->offset);
+ PUSH_DATA (push, nvc0_format_table[fb->zsbuf->format].rt);
+ PUSH_DATA (push, mt->level[sf->base.u.tex.level].tile_mode);
+ PUSH_DATA (push, mt->layer_stride >> 2);
+ BEGIN_NVC0(push, NVC0_3D(ZETA_ENABLE), 1);
+ PUSH_DATA (push, 1);
+ BEGIN_NVC0(push, NVC0_3D(ZETA_HORIZ), 3);
+ PUSH_DATA (push, sf->width);
+ PUSH_DATA (push, sf->height);
+ PUSH_DATA (push, (unk << 16) |
(sf->base.u.tex.first_layer + sf->depth));
- BEGIN_RING(chan, RING_3D(ZETA_BASE_LAYER), 1);
- OUT_RING (chan, sf->base.u.tex.first_layer);
+ BEGIN_NVC0(push, NVC0_3D(ZETA_BASE_LAYER), 1);
+ PUSH_DATA (push, 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;
mt->base.status &= ~NOUVEAU_BUFFER_STATUS_GPU_READING;
- nvc0_bufctx_add_resident(nvc0, NVC0_BUFCTX_FRAME, &mt->base,
- NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR);
+ BCTX_REFN(nvc0->bufctx_3d, FB, &mt->base, WR);
} else {
- BEGIN_RING(chan, RING_3D(ZETA_ENABLE), 1);
- OUT_RING (chan, 0);
+ BEGIN_NVC0(push, NVC0_3D(ZETA_ENABLE), 1);
+ PUSH_DATA (push, 0);
}
- if (serialize) {
- BEGIN_RING(chan, RING_3D(SERIALIZE), 1);
- OUT_RING (chan, 0);
- }
+ IMMED_NVC0(push, NVC0_3D(MULTISAMPLE_MODE), ms_mode);
+
+ if (serialize)
+ IMMED_NVC0(push, NVC0_3D(SERIALIZE), 0);
+
+ NOUVEAU_DRV_STAT(&nvc0->screen->base, gpu_serialize_count, serialize);
}
static void
nvc0_validate_blend_colour(struct nvc0_context *nvc0)
{
- struct nouveau_channel *chan = nvc0->screen->base.channel;
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
- BEGIN_RING(chan, RING_3D(BLEND_COLOR(0)), 4);
- OUT_RINGf (chan, nvc0->blend_colour.color[0]);
- OUT_RINGf (chan, nvc0->blend_colour.color[1]);
- OUT_RINGf (chan, nvc0->blend_colour.color[2]);
- OUT_RINGf (chan, nvc0->blend_colour.color[3]);
+ BEGIN_NVC0(push, NVC0_3D(BLEND_COLOR(0)), 4);
+ PUSH_DATAf(push, nvc0->blend_colour.color[0]);
+ PUSH_DATAf(push, nvc0->blend_colour.color[1]);
+ PUSH_DATAf(push, nvc0->blend_colour.color[2]);
+ PUSH_DATAf(push, nvc0->blend_colour.color[3]);
}
static void
nvc0_validate_stencil_ref(struct nvc0_context *nvc0)
{
- struct nouveau_channel *chan = nvc0->screen->base.channel;
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
+ const ubyte *ref = &nvc0->stencil_ref.ref_value[0];
- BEGIN_RING(chan, RING_3D(STENCIL_FRONT_FUNC_REF), 1);
- OUT_RING (chan, nvc0->stencil_ref.ref_value[0]);
- BEGIN_RING(chan, RING_3D(STENCIL_BACK_FUNC_REF), 1);
- OUT_RING (chan, nvc0->stencil_ref.ref_value[1]);
+ IMMED_NVC0(push, NVC0_3D(STENCIL_FRONT_FUNC_REF), ref[0]);
+ IMMED_NVC0(push, NVC0_3D(STENCIL_BACK_FUNC_REF), ref[1]);
}
static void
nvc0_validate_stipple(struct nvc0_context *nvc0)
{
- struct nouveau_channel *chan = nvc0->screen->base.channel;
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
unsigned i;
- BEGIN_RING(chan, RING_3D(POLYGON_STIPPLE_PATTERN(0)), 32);
+ BEGIN_NVC0(push, NVC0_3D(POLYGON_STIPPLE_PATTERN(0)), 32);
for (i = 0; i < 32; ++i)
- OUT_RING(chan, util_bswap32(nvc0->stipple.stipple[i]));
+ PUSH_DATA(push, util_bswap32(nvc0->stipple.stipple[i]));
}
static void
nvc0_validate_scissor(struct nvc0_context *nvc0)
{
- struct nouveau_channel *chan = nvc0->screen->base.channel;
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
struct pipe_scissor_state *s = &nvc0->scissor;
if (!(nvc0->dirty & NVC0_NEW_SCISSOR) &&
return;
nvc0->state.scissor = nvc0->rast->pipe.scissor;
- BEGIN_RING(chan, RING_3D(SCISSOR_HORIZ(0)), 2);
+ BEGIN_NVC0(push, NVC0_3D(SCISSOR_HORIZ(0)), 2);
if (nvc0->rast->pipe.scissor) {
- OUT_RING(chan, (s->maxx << 16) | s->minx);
- OUT_RING(chan, (s->maxy << 16) | s->miny);
+ PUSH_DATA(push, (s->maxx << 16) | s->minx);
+ PUSH_DATA(push, (s->maxy << 16) | s->miny);
} else {
- OUT_RING(chan, (0xffff << 16) | 0);
- OUT_RING(chan, (0xffff << 16) | 0);
+ PUSH_DATA(push, (0xffff << 16) | 0);
+ PUSH_DATA(push, (0xffff << 16) | 0);
}
}
static void
nvc0_validate_viewport(struct nvc0_context *nvc0)
{
- struct nouveau_channel *chan = nvc0->screen->base.channel;
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
struct pipe_viewport_state *vp = &nvc0->viewport;
int x, y, w, h;
float zmin, zmax;
- BEGIN_RING(chan, RING_3D(VIEWPORT_TRANSLATE_X(0)), 3);
- OUT_RINGf (chan, vp->translate[0]);
- OUT_RINGf (chan, vp->translate[1]);
- OUT_RINGf (chan, vp->translate[2]);
- BEGIN_RING(chan, RING_3D(VIEWPORT_SCALE_X(0)), 3);
- OUT_RINGf (chan, vp->scale[0]);
- OUT_RINGf (chan, vp->scale[1]);
- OUT_RINGf (chan, vp->scale[2]);
+ BEGIN_NVC0(push, NVC0_3D(VIEWPORT_TRANSLATE_X(0)), 3);
+ PUSH_DATAf(push, vp->translate[0]);
+ PUSH_DATAf(push, vp->translate[1]);
+ PUSH_DATAf(push, vp->translate[2]);
+ BEGIN_NVC0(push, NVC0_3D(VIEWPORT_SCALE_X(0)), 3);
+ PUSH_DATAf(push, vp->scale[0]);
+ PUSH_DATAf(push, vp->scale[1]);
+ PUSH_DATAf(push, vp->scale[2]);
/* now set the viewport rectangle to viewport dimensions for clipping */
- x = (int)(vp->translate[0] - fabsf(vp->scale[0]));
- y = (int)(vp->translate[1] - fabsf(vp->scale[1]));
- w = (int)fabsf(2.0f * vp->scale[0]);
- h = (int)fabsf(2.0f * vp->scale[1]);
+ x = util_iround(MAX2(0.0f, vp->translate[0] - fabsf(vp->scale[0])));
+ y = util_iround(MAX2(0.0f, vp->translate[1] - fabsf(vp->scale[1])));
+ w = util_iround(vp->translate[0] + fabsf(vp->scale[0])) - x;
+ h = util_iround(vp->translate[1] + fabsf(vp->scale[1])) - y;
+
zmin = vp->translate[2] - fabsf(vp->scale[2]);
zmax = vp->translate[2] + fabsf(vp->scale[2]);
- BEGIN_RING(chan, RING_3D(VIEWPORT_HORIZ(0)), 2);
- OUT_RING (chan, (w << 16) | x);
- OUT_RING (chan, (h << 16) | y);
- BEGIN_RING(chan, RING_3D(DEPTH_RANGE_NEAR(0)), 2);
- OUT_RINGf (chan, zmin);
- OUT_RINGf (chan, zmax);
+ BEGIN_NVC0(push, NVC0_3D(VIEWPORT_HORIZ(0)), 2);
+ PUSH_DATA (push, (w << 16) | x);
+ PUSH_DATA (push, (h << 16) | y);
+ BEGIN_NVC0(push, NVC0_3D(DEPTH_RANGE_NEAR(0)), 2);
+ PUSH_DATAf(push, zmin);
+ PUSH_DATAf(push, zmax);
+}
+
+static INLINE void
+nvc0_upload_uclip_planes(struct nvc0_context *nvc0, unsigned s)
+{
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
+ struct nouveau_bo *bo = nvc0->screen->uniform_bo;
+
+ BEGIN_NVC0(push, NVC0_3D(CB_SIZE), 3);
+ PUSH_DATA (push, 512);
+ PUSH_DATAh(push, bo->offset + (5 << 16) + (s << 9));
+ PUSH_DATA (push, bo->offset + (5 << 16) + (s << 9));
+ BEGIN_1IC0(push, NVC0_3D(CB_POS), PIPE_MAX_CLIP_PLANES * 4 + 1);
+ PUSH_DATA (push, 256);
+ PUSH_DATAp(push, &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;
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
+ struct nvc0_program *vp;
+ unsigned stage;
+ uint8_t clip_enable = nvc0->rast->pipe.clip_plane_enable;
+
+ if (nvc0->gmtyprog) {
+ stage = 3;
+ vp = nvc0->gmtyprog;
+ } else
+ if (nvc0->tevlprog) {
+ stage = 2;
+ vp = nvc0->tevlprog;
} else {
- clip = NVC0_3D_VIEW_VOLUME_CLIP_CTRL_UNK1_UNK1;
+ stage = 0;
+ vp = nvc0->vertprog;
}
- BEGIN_RING(chan, RING_3D(VIEW_VOLUME_CLIP_CTRL), 1);
- OUT_RING (chan, clip);
+ if (clip_enable && vp->vp.num_ucps < PIPE_MAX_CLIP_PLANES)
+ nvc0_check_program_ucps(nvc0, vp, clip_enable);
- if (nvc0->clip.nr) {
- struct nouveau_bo *bo = nvc0->screen->uniforms;
+ if (nvc0->dirty & (NVC0_NEW_CLIP | (NVC0_NEW_VERTPROG << stage)))
+ if (vp->vp.num_ucps > 0 && vp->vp.num_ucps <= PIPE_MAX_CLIP_PLANES)
+ nvc0_upload_uclip_planes(nvc0, stage);
- 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);
+ clip_enable &= vp->vp.clip_enable;
- 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 (nvc0->state.clip_enable != clip_enable) {
+ nvc0->state.clip_enable = clip_enable;
+ IMMED_NVC0(push, NVC0_3D(CLIP_DISTANCE_ENABLE), clip_enable);
+ }
+ if (nvc0->state.clip_mode != vp->vp.clip_mode) {
+ nvc0->state.clip_mode = vp->vp.clip_mode;
+ BEGIN_NVC0(push, NVC0_3D(CLIP_DISTANCE_MODE), 1);
+ PUSH_DATA (push, vp->vp.clip_mode);
}
}
static void
nvc0_validate_blend(struct nvc0_context *nvc0)
{
- struct nouveau_channel *chan = nvc0->screen->base.channel;
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
- WAIT_RING(chan, nvc0->blend->size);
- OUT_RINGp(chan, nvc0->blend->state, nvc0->blend->size);
+ PUSH_SPACE(push, nvc0->blend->size);
+ PUSH_DATAp(push, nvc0->blend->state, nvc0->blend->size);
}
static void
nvc0_validate_zsa(struct nvc0_context *nvc0)
{
- struct nouveau_channel *chan = nvc0->screen->base.channel;
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
- WAIT_RING(chan, nvc0->zsa->size);
- OUT_RINGp(chan, nvc0->zsa->state, nvc0->zsa->size);
+ PUSH_SPACE(push, nvc0->zsa->size);
+ PUSH_DATAp(push, nvc0->zsa->state, nvc0->zsa->size);
}
static void
nvc0_validate_rasterizer(struct nvc0_context *nvc0)
{
- struct nouveau_channel *chan = nvc0->screen->base.channel;
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
- WAIT_RING(chan, nvc0->rast->size);
- OUT_RINGp(chan, nvc0->rast->state, nvc0->rast->size);
+ PUSH_SPACE(push, nvc0->rast->size);
+ PUSH_DATAp(push, nvc0->rast->state, nvc0->rast->size);
}
static void
-nvc0_validate_sprite_coords(struct nvc0_context *nvc0)
+nvc0_constbufs_validate(struct nvc0_context *nvc0)
{
- struct nouveau_channel *chan = nvc0->screen->base.channel;
- uint32_t reg;
-
- if (nvc0->rast->pipe.sprite_coord_mode == PIPE_SPRITE_COORD_UPPER_LEFT)
- reg = NVC0_3D_POINT_COORD_REPLACE_COORD_ORIGIN_UPPER_LEFT;
- else
- reg = NVC0_3D_POINT_COORD_REPLACE_COORD_ORIGIN_LOWER_LEFT;
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
+ unsigned s;
- if (nvc0->rast->pipe.point_quad_rasterization) {
- uint32_t en = nvc0->rast->pipe.sprite_coord_enable;
+ for (s = 0; s < 5; ++s) {
+ while (nvc0->constbuf_dirty[s]) {
+ int i = ffs(nvc0->constbuf_dirty[s]) - 1;
+ nvc0->constbuf_dirty[s] &= ~(1 << i);
- while (en) {
- int i = ffs(en) - 1;
- en &= ~(1 << i);
- if (i >= 0 && i < 8)
- reg |= 8 << i;
+ if (nvc0->constbuf[s][i].user) {
+ struct nouveau_bo *bo = nvc0->screen->uniform_bo;
+ const unsigned base = s << 16;
+ const unsigned size = nvc0->constbuf[s][0].size;
+ assert(i == 0); /* we really only want OpenGL uniforms here */
+ assert(nvc0->constbuf[s][0].u.data);
+
+ if (nvc0->state.uniform_buffer_bound[s] < size) {
+ nvc0->state.uniform_buffer_bound[s] = align(size, 0x100);
+
+ BEGIN_NVC0(push, NVC0_3D(CB_SIZE), 3);
+ PUSH_DATA (push, nvc0->state.uniform_buffer_bound[s]);
+ PUSH_DATAh(push, bo->offset + base);
+ PUSH_DATA (push, bo->offset + base);
+ BEGIN_NVC0(push, NVC0_3D(CB_BIND(s)), 1);
+ PUSH_DATA (push, (0 << 4) | 1);
+ }
+ nvc0_cb_push(&nvc0->base, bo, NOUVEAU_BO_VRAM,
+ base, nvc0->state.uniform_buffer_bound[s],
+ 0, (size + 3) / 4,
+ nvc0->constbuf[s][0].u.data);
+ } else {
+ struct nv04_resource *res =
+ nv04_resource(nvc0->constbuf[s][i].u.buf);
+ if (res) {
+ BEGIN_NVC0(push, NVC0_3D(CB_SIZE), 3);
+ PUSH_DATA (push, nvc0->constbuf[s][i].size);
+ PUSH_DATAh(push, res->address + nvc0->constbuf[s][i].offset);
+ PUSH_DATA (push, res->address + nvc0->constbuf[s][i].offset);
+ BEGIN_NVC0(push, NVC0_3D(CB_BIND(s)), 1);
+ PUSH_DATA (push, (i << 4) | 1);
+
+ BCTX_REFN(nvc0->bufctx_3d, CB(s, i), res, RD);
+ } else {
+ BEGIN_NVC0(push, NVC0_3D(CB_BIND(s)), 1);
+ PUSH_DATA (push, (i << 4) | 0);
+ }
+ if (i == 0)
+ nvc0->state.uniform_buffer_bound[s] = 0;
+ }
}
}
-
- BEGIN_RING(chan, RING_3D(POINT_COORD_REPLACE), 1);
- OUT_RING (chan, reg);
}
static void
-nvc0_constbufs_validate(struct nvc0_context *nvc0)
+nvc0_validate_sample_mask(struct nvc0_context *nvc0)
{
- struct nouveau_channel *chan = nvc0->screen->base.channel;
- struct nouveau_bo *bo;
- unsigned s;
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
+
+ unsigned mask[4] =
+ {
+ nvc0->sample_mask & 0xffff,
+ nvc0->sample_mask & 0xffff,
+ nvc0->sample_mask & 0xffff,
+ nvc0->sample_mask & 0xffff
+ };
+
+ BEGIN_NVC0(push, NVC0_3D(MSAA_MASK(0)), 4);
+ PUSH_DATA (push, mask[0]);
+ PUSH_DATA (push, mask[1]);
+ PUSH_DATA (push, mask[2]);
+ PUSH_DATA (push, mask[3]);
+ BEGIN_NVC0(push, NVC0_3D(SAMPLE_SHADING), 1);
+ PUSH_DATA (push, 0x01);
+}
- for (s = 0; s < 5; ++s) {
- struct nv04_resource *res;
- int i;
+void
+nvc0_validate_global_residents(struct nvc0_context *nvc0,
+ struct nouveau_bufctx *bctx, int bin)
+{
+ unsigned i;
- while (nvc0->constbuf_dirty[s]) {
- unsigned base = 0;
- unsigned offset = 0, words = 0;
- boolean rebind = TRUE;
+ for (i = 0; i < nvc0->global_residents.size / sizeof(struct pipe_resource *);
+ ++i) {
+ struct pipe_resource *res = *util_dynarray_element(
+ &nvc0->global_residents, struct pipe_resource *, i);
+ if (res)
+ nvc0_add_resident(bctx, bin, nv04_resource(res), NOUVEAU_BO_RDWR);
+ }
+}
- i = ffs(nvc0->constbuf_dirty[s]) - 1;
- nvc0->constbuf_dirty[s] &= ~(1 << i);
+static void
+nvc0_validate_derived_1(struct nvc0_context *nvc0)
+{
+ struct nouveau_pushbuf *push = nvc0->base.pushbuf;
+ boolean rasterizer_discard;
- res = nv04_resource(nvc0->constbuf[s][i]);
- if (!res) {
- BEGIN_RING(chan, RING_3D(CB_BIND(s)), 1);
- OUT_RING (chan, (i << 4) | 0);
- if (i == 0)
- nvc0->state.uniform_buffer_bound[s] = 0;
- continue;
- }
+ if (nvc0->rast && nvc0->rast->pipe.rasterizer_discard) {
+ rasterizer_discard = TRUE;
+ } else {
+ boolean zs = nvc0->zsa &&
+ (nvc0->zsa->pipe.depth.enabled || nvc0->zsa->pipe.stencil[0].enabled);
+ rasterizer_discard = !zs &&
+ (!nvc0->fragprog || !nvc0->fragprog->hdr[18]);
+ }
- if (!nouveau_resource_mapped_by_gpu(&res->base)) {
- if (i == 0) {
- base = s << 16;
- bo = nvc0->screen->uniforms;
+ if (rasterizer_discard != nvc0->state.rasterizer_discard) {
+ nvc0->state.rasterizer_discard = rasterizer_discard;
+ IMMED_NVC0(push, NVC0_3D(RASTERIZE_ENABLE), !rasterizer_discard);
+ }
+}
- if (nvc0->state.uniform_buffer_bound[s] >= res->base.width0)
- rebind = FALSE;
- else
- nvc0->state.uniform_buffer_bound[s] =
- align(res->base.width0, 0x100);
- } else {
- bo = res->bo;
- }
-#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;
- if (i == 0)
- nvc0->state.uniform_buffer_bound[s] = 0;
- }
+static void
+nvc0_switch_pipe_context(struct nvc0_context *ctx_to)
+{
+ struct nvc0_context *ctx_from = ctx_to->screen->cur_ctx;
+ unsigned s;
- if (bo != nvc0->screen->uniforms)
- nvc0_bufctx_add_resident(nvc0, NVC0_BUFCTX_CONSTANT, res,
- NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
-
- if (rebind) {
- MARK_RING (chan, 4, 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(chan, RING_3D(CB_BIND(s)), 1);
- OUT_RING (chan, (i << 4) | 1);
- }
+ if (ctx_from)
+ ctx_to->state = ctx_from->state;
- while (words) {
- unsigned nr = AVAIL_RING(chan);
+ ctx_to->dirty = ~0;
- 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;
- }
- }
+ for (s = 0; s < 5; ++s) {
+ ctx_to->samplers_dirty[s] = ~0;
+ ctx_to->textures_dirty[s] = ~0;
}
+
+ if (!ctx_to->vertex)
+ ctx_to->dirty &= ~(NVC0_NEW_VERTEX | NVC0_NEW_ARRAYS);
+ if (!ctx_to->idxbuf.buffer)
+ ctx_to->dirty &= ~NVC0_NEW_IDXBUF;
+
+ if (!ctx_to->vertprog)
+ ctx_to->dirty &= ~NVC0_NEW_VERTPROG;
+ if (!ctx_to->fragprog)
+ ctx_to->dirty &= ~NVC0_NEW_FRAGPROG;
+
+ if (!ctx_to->blend)
+ ctx_to->dirty &= ~NVC0_NEW_BLEND;
+ if (!ctx_to->rast)
+ ctx_to->dirty &= ~(NVC0_NEW_RASTERIZER | NVC0_NEW_SCISSOR);
+ if (!ctx_to->zsa)
+ ctx_to->dirty &= ~NVC0_NEW_ZSA;
+
+ 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_sprite_coords, NVC0_NEW_RASTERIZER | NVC0_NEW_FRAGPROG },
+ { 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 },
+ { nve4_set_tex_handles, NVC0_NEW_TEXTURES | NVC0_NEW_SAMPLERS },
{ nvc0_vertex_arrays_validate, NVC0_NEW_VERTEX | NVC0_NEW_ARRAYS },
- { nvc0_tfb_validate, NVC0_NEW_TFB | NVC0_NEW_TFB_BUFFERS }
+ { nvc0_validate_surfaces, NVC0_NEW_SURFACES },
+ { nvc0_idxbuf_validate, NVC0_NEW_IDXBUF },
+ { 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;
+ int ret;
unsigned i;
-#if 0
- if (nvc0->screen->cur_ctx != nvc0) /* FIXME: not everything is valid */
- nvc0->dirty = 0xffffffff;
-#endif
- nvc0->screen->cur_ctx = nvc0;
- if (nvc0->dirty) {
+ if (nvc0->screen->cur_ctx != nvc0)
+ nvc0_switch_pipe_context(nvc0);
+
+ 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;
+
+ nvc0_bufctx_fence(nvc0, nvc0->bufctx_3d, FALSE);
}
- nvc0_bufctx_emit_relocs(nvc0);
+ nouveau_pushbuf_bufctx(nvc0->base.pushbuf, nvc0->bufctx_3d);
+ ret = nouveau_pushbuf_validate(nvc0->base.pushbuf);
+ if (unlikely(ret))
+ return FALSE;
+
+ if (unlikely(nvc0->state.flushed))
+ nvc0_bufctx_fence(nvc0, nvc0->bufctx_3d, TRUE);
return TRUE;
}