static void
nv50_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
struct pipe_constant_buffer *cb)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct pipe_resource *res = cb ? cb->buffer : NULL;
const unsigned s = nv50_context_shader_stage(shader);
const unsigned i = index;
if (shader == PIPE_SHADER_COMPUTE)
return;
if (nv50->constbuf[s][i].user)
nv50->constbuf[s][i].u.buf = NULL;
else
if (nv50->constbuf[s][i].u.buf)
nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_CB(s, i));
pipe_resource_reference(&nv50->constbuf[s][i].u.buf, res);
nv50->constbuf[s][i].user = (cb && cb->user_buffer) ? TRUE : FALSE;
if (nv50->constbuf[s][i].user) {
nv50->constbuf[s][i].u.data = cb->user_buffer;
nv50->constbuf[s][i].size = cb->buffer_size;
nv50->constbuf_valid[s] |= 1 << i;
} else
if (res) {
nv50->constbuf[s][i].offset = cb->buffer_offset;
nv50->constbuf[s][i].size = align(cb->buffer_size, 0x100);
nv50->constbuf_valid[s] |= 1 << i;
} else {
nv50->constbuf_valid[s] &= ~(1 << i);
}
nv50->constbuf_dirty[s] |= 1 << i;
nv50->dirty |= NV50_NEW_CONSTBUF;
}
static void
nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_pushbuf *push = nv50->base.pushbuf;
struct nv50_query *q = nv50_query(pq);
switch (q->type) {
case PIPE_QUERY_OCCLUSION_COUNTER:
nv50_query_get(push, q, 0, 0x0100f002);
PUSH_SPACE(push, 2);
BEGIN_NV04(push, NV50_3D(SAMPLECNT_ENABLE), 1);
PUSH_DATA (push, 0);
break;
case PIPE_QUERY_PRIMITIVES_GENERATED:
nv50_query_get(push, q, 0, 0x06805002);
break;
case PIPE_QUERY_PRIMITIVES_EMITTED:
nv50_query_get(push, q, 0, 0x05805002);
break;
case PIPE_QUERY_SO_STATISTICS:
nv50_query_get(push, q, 0x00, 0x05805002);
nv50_query_get(push, q, 0x10, 0x06805002);
break;
case PIPE_QUERY_TIMESTAMP_DISJOINT:
case PIPE_QUERY_TIME_ELAPSED:
nv50_query_get(push, q, 0, 0x00005002);
break;
case PIPE_QUERY_GPU_FINISHED:
nv50_query_get(push, q, 0, 0x1000f010);
break;
default:
assert(0);
break;
}
q->flushed = FALSE;
}
static void
nv50_set_vertex_buffers(struct pipe_context *pipe,
unsigned start_slot, unsigned count,
const struct pipe_vertex_buffer *vb)
{
struct nv50_context *nv50 = nv50_context(pipe);
unsigned i;
util_set_vertex_buffers_count(nv50->vtxbuf, &nv50->num_vtxbufs, vb,
start_slot, count);
if (!vb) {
nv50->vbo_user &= ~(((1ull << count) - 1) << start_slot);
nv50->vbo_constant &= ~(((1ull << count) - 1) << start_slot);
return;
}
for (i = 0; i < count; ++i) {
unsigned dst_index = start_slot + i;
if (!vb[i].buffer && vb[i].user_buffer) {
nv50->vbo_user |= 1 << dst_index;
if (!vb[i].stride)
nv50->vbo_constant |= 1 << dst_index;
else
nv50->vbo_constant &= ~(1 << dst_index);
} else {
nv50->vbo_user &= ~(1 << dst_index);
nv50->vbo_constant &= ~(1 << dst_index);
}
}
nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_VERTEX);
nv50->dirty |= NV50_NEW_ARRAYS;
}
void
nv50_push_elements_instanced(struct pipe_context *pipe,
struct pipe_resource *idxbuf,
unsigned idxsize, int idxbias,
unsigned mode, unsigned start, unsigned count,
unsigned i_start, unsigned i_count)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_grobj *tesla = nv50->screen->tesla;
struct nouveau_channel *chan = tesla->channel;
struct push_context ctx;
const unsigned p_overhead = 4 + /* begin/end */
4; /* potential edgeflag enable/disable */
const unsigned v_overhead = 1 + /* VERTEX_DATA packet header */
2; /* potential edgeflag modification */
struct util_split_prim s;
unsigned vtx_size;
boolean nzi = FALSE;
int i;
ctx.nv50 = nv50;
ctx.attr_nr = 0;
ctx.idxbuf = NULL;
ctx.vtx_size = 0;
ctx.edgeflag = 0.5f;
ctx.edgeflag_attr = nv50->vertprog->vp.edgeflag;
/* map vertex buffers, determine vertex size */
for (i = 0; i < nv50->vtxelt->num_elements; i++) {
struct pipe_vertex_element *ve = &nv50->vtxelt->pipe[i];
struct pipe_vertex_buffer *vb = &nv50->vtxbuf[ve->vertex_buffer_index];
struct nouveau_bo *bo = nv50_resource(vb->buffer)->bo;
unsigned size, nr_components, n;
if (!(nv50->vbo_fifo & (1 << i)))
continue;
n = ctx.attr_nr++;
if (nouveau_bo_map(bo, NOUVEAU_BO_RD)) {
assert(bo->map);
return;
}
ctx.attr[n].map = (uint8_t *)bo->map + vb->buffer_offset + ve->src_offset;
nouveau_bo_unmap(bo);
ctx.attr[n].stride = vb->stride;
ctx.attr[n].divisor = ve->instance_divisor;
if (ctx.attr[n].divisor) {
ctx.attr[n].step = i_start % ve->instance_divisor;
ctx.attr[n].map = (uint8_t *)ctx.attr[n].map + i_start * vb->stride;
}
size = util_format_get_component_bits(ve->src_format,
UTIL_FORMAT_COLORSPACE_RGB, 0);
nr_components = util_format_get_nr_components(ve->src_format);
switch (size) {
case 8:
switch (nr_components) {
case 1: ctx.attr[n].push = emit_b08_1; break;
case 2: ctx.attr[n].push = emit_b16_1; break;
case 3: ctx.attr[n].push = emit_b08_3; break;
case 4: ctx.attr[n].push = emit_b32_1; break;
}
ctx.vtx_size++;
break;
case 16:
switch (nr_components) {
case 1: ctx.attr[n].push = emit_b16_1; break;
case 2: ctx.attr[n].push = emit_b32_1; break;
case 3: ctx.attr[n].push = emit_b16_3; break;
case 4: ctx.attr[n].push = emit_b32_2; break;
}
ctx.vtx_size += (nr_components + 1) >> 1;
break;
case 32:
switch (nr_components) {
case 1: ctx.attr[n].push = emit_b32_1; break;
case 2: ctx.attr[n].push = emit_b32_2; break;
case 3: ctx.attr[n].push = emit_b32_3; break;
case 4: ctx.attr[n].push = emit_b32_4; break;
}
ctx.vtx_size += nr_components;
break;
default:
assert(0);
return;
}
}
vtx_size = ctx.vtx_size + v_overhead;
/* map index buffer, if present */
if (idxbuf) {
struct nouveau_bo *bo = nv50_resource(idxbuf)->bo;
if (nouveau_bo_map(bo, NOUVEAU_BO_RD)) {
assert(bo->map);
return;
}
ctx.idxbuf = bo->map;
ctx.idxbias = idxbias;
ctx.idxsize = idxsize;
nouveau_bo_unmap(bo);
}
s.priv = &ctx;
s.edge = emit_edgeflag;
if (idxbuf) {
if (idxsize == 1)
s.emit = idxbias ? emit_elt08_biased : emit_elt08;
else
if (idxsize == 2)
s.emit = idxbias ? emit_elt16_biased : emit_elt16;
else
s.emit = idxbias ? emit_elt32_biased : emit_elt32;
} else
s.emit = emit_verts;
/* per-instance loop */
BEGIN_RING(chan, tesla, NV50TCL_CB_ADDR, 2);
OUT_RING (chan, NV50_CB_AUX | (24 << 8));
OUT_RING (chan, i_start);
while (i_count--) {
unsigned max_verts;
boolean done;
for (i = 0; i < ctx.attr_nr; i++) {
if (!ctx.attr[i].divisor ||
ctx.attr[i].divisor != ++ctx.attr[i].step)
continue;
ctx.attr[i].step = 0;
ctx.attr[i].map = (uint8_t *)ctx.attr[i].map + ctx.attr[i].stride;
}
util_split_prim_init(&s, mode, start, count);
do {
if (AVAIL_RING(chan) < p_overhead + (6 * vtx_size)) {
FIRE_RING(chan);
if (!nv50_state_validate(nv50, p_overhead + (6 * vtx_size))) {
assert(0);
return;
}
}
max_verts = AVAIL_RING(chan);
max_verts -= p_overhead;
max_verts /= vtx_size;
BEGIN_RING(chan, tesla, NV50TCL_VERTEX_BEGIN, 1);
OUT_RING (chan, nv50_prim(s.mode) | (nzi ? (1 << 28) : 0));
done = util_split_prim_next(&s, max_verts);
BEGIN_RING(chan, tesla, NV50TCL_VERTEX_END, 1);
OUT_RING (chan, 0);
} while (!done);
nzi = TRUE;
}
}
static void
nv50_resource_resolve(struct pipe_context *pipe,
const struct pipe_resolve_info *info)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nv50_screen *screen = nv50->screen;
struct nv50_blitctx *blit = screen->blitctx;
struct nouveau_channel *chan = screen->base.channel;
struct pipe_resource *src = info->src.res;
struct pipe_resource *dst = info->dst.res;
float x0, x1, y0, y1, z;
float x_range, y_range;
nv50_blitctx_get_color_mask_and_fp(blit, dst->format, info->mask);
blit->filter = util_format_is_depth_or_stencil(dst->format) ? 0 : 1;
nv50_blitctx_pre_blit(blit, nv50);
nv50_blit_set_dst(nv50, dst, info->dst.level, info->dst.layer);
nv50_blit_set_src(nv50, src, 0, info->src.layer);
nv50_blitctx_prepare_state(blit);
nv50_state_validate(nv50, 36);
x_range =
(float)(info->src.x1 - info->src.x0) /
(float)(info->dst.x1 - info->dst.x0);
y_range =
(float)(info->src.y1 - info->src.y0) /
(float)(info->dst.y1 - info->dst.y0);
x0 = (float)info->src.x0 - x_range * (float)info->dst.x0;
y0 = (float)info->src.y0 - y_range * (float)info->dst.y0;
x1 = x0 + 16384.0f * x_range;
y1 = y0 + 16384.0f * y_range;
x0 *= (float)(1 << nv50_miptree(src)->ms_x);
x1 *= (float)(1 << nv50_miptree(src)->ms_x);
y0 *= (float)(1 << nv50_miptree(src)->ms_y);
y1 *= (float)(1 << nv50_miptree(src)->ms_y);
z = (float)info->src.layer;
BEGIN_RING(chan, RING_3D(FP_START_ID), 1);
OUT_RING (chan,
blit->fp.code_base + blit->fp_offset);
BEGIN_RING(chan, RING_3D(VIEWPORT_TRANSFORM_EN), 1);
OUT_RING (chan, 0);
/* Draw a large triangle in screen coordinates covering the whole
* render target, with scissors defining the destination region.
* The vertex is supplied with non-normalized texture coordinates
* arranged in a way to yield the desired offset and scale.
*/
BEGIN_RING(chan, RING_3D(SCISSOR_HORIZ(0)), 2);
OUT_RING (chan, (info->dst.x1 << 16) | info->dst.x0);
OUT_RING (chan, (info->dst.y1 << 16) | info->dst.y0);
BEGIN_RING(chan, RING_3D(VERTEX_BEGIN_GL), 1);
OUT_RING (chan, NV50_3D_VERTEX_BEGIN_GL_PRIMITIVE_TRIANGLES);
BEGIN_RING(chan, RING_3D(VTX_ATTR_3F_X(1)), 3);
OUT_RINGf (chan, x0);
OUT_RINGf (chan, y0);
OUT_RINGf (chan, z);
BEGIN_RING(chan, RING_3D(VTX_ATTR_2F_X(0)), 2);
OUT_RINGf (chan, 0.0f);
OUT_RINGf (chan, 0.0f);
BEGIN_RING(chan, RING_3D(VTX_ATTR_3F_X(1)), 3);
OUT_RINGf (chan, x1);
OUT_RINGf (chan, y0);
OUT_RINGf (chan, z);
BEGIN_RING(chan, RING_3D(VTX_ATTR_2F_X(0)), 2);
OUT_RINGf (chan, 16384 << nv50_miptree(dst)->ms_x);
OUT_RINGf (chan, 0.0f);
BEGIN_RING(chan, RING_3D(VTX_ATTR_3F_X(1)), 3);
OUT_RINGf (chan, x0);
OUT_RINGf (chan, y1);
OUT_RINGf (chan, z);
BEGIN_RING(chan, RING_3D(VTX_ATTR_2F_X(0)), 2);
OUT_RINGf (chan, 0.0f);
OUT_RINGf (chan, 16384 << nv50_miptree(dst)->ms_y);
BEGIN_RING(chan, RING_3D(VERTEX_END_GL), 1);
OUT_RING (chan, 0);
/* re-enable normally constant state */
BEGIN_RING(chan, RING_3D(VIEWPORT_TRANSFORM_EN), 1);
OUT_RING (chan, 1);
nv50_blitctx_post_blit(nv50, blit);
}
static void
nv50_resource_copy_region(struct pipe_context *pipe,
struct pipe_resource *dst, unsigned dst_level,
unsigned dstx, unsigned dsty, unsigned dstz,
struct pipe_resource *src, unsigned src_level,
const struct pipe_box *src_box)
{
struct nv50_screen *screen = nv50_context(pipe)->screen;
int ret;
boolean m2mf;
unsigned dst_layer = dstz, src_layer = src_box->z;
/* Fallback for buffers. */
if (dst->target == PIPE_BUFFER && src->target == PIPE_BUFFER) {
util_resource_copy_region(pipe, dst, dst_level, dstx, dsty, dstz,
src, src_level, src_box);
return;
}
assert(src->nr_samples == dst->nr_samples);
m2mf = (src->format == dst->format) ||
(util_format_get_blocksizebits(src->format) ==
util_format_get_blocksizebits(dst->format));
nv04_resource(dst)->status |= NOUVEAU_BUFFER_STATUS_GPU_WRITING;
if (m2mf) {
struct nv50_m2mf_rect drect, srect;
unsigned i;
unsigned nx = util_format_get_nblocksx(src->format, src_box->width);
unsigned ny = util_format_get_nblocksy(src->format, src_box->height);
nv50_m2mf_rect_setup(&drect, dst, dst_level, dstx, dsty, dstz);
nv50_m2mf_rect_setup(&srect, src, src_level,
src_box->x, src_box->y, src_box->z);
for (i = 0; i < src_box->depth; ++i) {
nv50_m2mf_transfer_rect(&screen->base.base, &drect, &srect, nx, ny);
if (nv50_miptree(dst)->layout_3d)
drect.z++;
else
drect.base += nv50_miptree(dst)->layer_stride;
if (nv50_miptree(src)->layout_3d)
srect.z++;
else
srect.base += nv50_miptree(src)->layer_stride;
}
return;
}
assert((src->format == dst->format) ||
(nv50_2d_format_faithful(src->format) &&
nv50_2d_format_faithful(dst->format)));
for (; dst_layer < dstz + src_box->depth; ++dst_layer, ++src_layer) {
ret = nv50_2d_texture_do_copy(screen->base.channel,
nv50_miptree(dst), dst_level,
dstx, dsty, dst_layer,
nv50_miptree(src), src_level,
src_box->x, src_box->y, src_layer,
src_box->width, src_box->height);
if (ret)
return;
}
}
static void
nv50_gp_sampler_states_bind(struct pipe_context *pipe, unsigned nr, void **s)
{
nv50_stage_sampler_states_bind(nv50_context(pipe), 1, nr, s);
}
static void
nv50_query_destroy(struct pipe_context *pipe, struct pipe_query *pq)
{
nv50_query_allocate(nv50_context(pipe), nv50_query(pq), 0);
FREE(nv50_query(pq));
}
static boolean
nv50_query_result(struct pipe_context *pipe, struct pipe_query *pq,
boolean wait, union pipe_query_result *result)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nv50_query *q = nv50_query(pq);
uint64_t *res64 = (uint64_t *)result;
uint32_t *res32 = (uint32_t *)result;
boolean *res8 = (boolean *)result;
uint64_t *data64 = (uint64_t *)q->data;
int i;
if (!q->ready) /* update ? */
q->ready = nv50_query_ready(q);
if (!q->ready) {
if (!wait) {
/* for broken apps that spin on GL_QUERY_RESULT_AVAILABLE */
if (!q->flushed) {
q->flushed = TRUE;
PUSH_KICK(nv50->base.pushbuf);
}
return FALSE;
}
if (nouveau_bo_wait(q->bo, NOUVEAU_BO_RD, nv50->screen->base.client))
return FALSE;
}
q->ready = TRUE;
switch (q->type) {
case PIPE_QUERY_GPU_FINISHED:
res8[0] = TRUE;
break;
case PIPE_QUERY_OCCLUSION_COUNTER: /* u32 sequence, u32 count, u64 time */
res64[0] = q->data[1];
break;
case PIPE_QUERY_PRIMITIVES_GENERATED: /* u64 count, u64 time */
case PIPE_QUERY_PRIMITIVES_EMITTED: /* u64 count, u64 time */
res64[0] = data64[0] - data64[2];
break;
case PIPE_QUERY_SO_STATISTICS:
res64[0] = data64[0] - data64[4];
res64[1] = data64[2] - data64[6];
break;
case PIPE_QUERY_PIPELINE_STATISTICS:
for (i = 0; i < 8; ++i)
res64[i] = data64[i * 2] - data64[16 + i * 2];
break;
case PIPE_QUERY_TIMESTAMP:
res64[0] = data64[1];
break;
case PIPE_QUERY_TIMESTAMP_DISJOINT:
res64[0] = 1000000000;
res8[8] = FALSE;
break;
case PIPE_QUERY_TIME_ELAPSED:
res64[0] = data64[1] - data64[3];
break;
case NVA0_QUERY_STREAM_OUTPUT_BUFFER_OFFSET:
res32[0] = q->data[1];
break;
default:
return FALSE;
}
return TRUE;
}
static void *
nv50_rasterizer_state_create(struct pipe_context *pipe,
const struct pipe_rasterizer_state *cso)
{
struct nouveau_stateobj *so = so_new(15, 21, 0);
struct nouveau_grobj *tesla = nv50_context(pipe)->screen->tesla;
struct nv50_rasterizer_stateobj *rso =
CALLOC_STRUCT(nv50_rasterizer_stateobj);
/*XXX: ignored
* - light_twosize
* - point_smooth
* - multisample
* - point_sprite / sprite_coord_mode
*/
so_method(so, tesla, NV50TCL_SHADE_MODEL, 1);
so_data (so, cso->flatshade ? NV50TCL_SHADE_MODEL_FLAT :
NV50TCL_SHADE_MODEL_SMOOTH);
so_method(so, tesla, NV50TCL_PROVOKING_VERTEX_LAST, 1);
so_data (so, cso->flatshade_first ? 0 : 1);
so_method(so, tesla, NV50TCL_VERTEX_TWO_SIDE_ENABLE, 1);
so_data (so, cso->light_twoside);
so_method(so, tesla, NV50TCL_LINE_WIDTH, 1);
so_data (so, fui(cso->line_width));
so_method(so, tesla, NV50TCL_LINE_SMOOTH_ENABLE, 1);
so_data (so, cso->line_smooth ? 1 : 0);
if (cso->line_stipple_enable) {
so_method(so, tesla, NV50TCL_LINE_STIPPLE_ENABLE, 1);
so_data (so, 1);
so_method(so, tesla, NV50TCL_LINE_STIPPLE_PATTERN, 1);
so_data (so, (cso->line_stipple_pattern << 8) |
cso->line_stipple_factor);
} else {
so_method(so, tesla, NV50TCL_LINE_STIPPLE_ENABLE, 1);
so_data (so, 0);
}
so_method(so, tesla, NV50TCL_POINT_SIZE, 1);
so_data (so, fui(cso->point_size));
so_method(so, tesla, NV50TCL_POINT_SPRITE_ENABLE, 1);
so_data (so, cso->point_quad_rasterization ? 1 : 0);
so_method(so, tesla, NV50TCL_POLYGON_MODE_FRONT, 3);
if (cso->front_winding == PIPE_WINDING_CCW) {
so_data(so, nvgl_polygon_mode(cso->fill_ccw));
so_data(so, nvgl_polygon_mode(cso->fill_cw));
} else {
so_data(so, nvgl_polygon_mode(cso->fill_cw));
so_data(so, nvgl_polygon_mode(cso->fill_ccw));
}
so_data(so, cso->poly_smooth ? 1 : 0);
so_method(so, tesla, NV50TCL_CULL_FACE_ENABLE, 3);
so_data (so, cso->cull_mode != PIPE_WINDING_NONE);
if (cso->front_winding == PIPE_WINDING_CCW) {
so_data(so, NV50TCL_FRONT_FACE_CCW);
switch (cso->cull_mode) {
case PIPE_WINDING_CCW:
so_data(so, NV50TCL_CULL_FACE_FRONT);
break;
case PIPE_WINDING_CW:
so_data(so, NV50TCL_CULL_FACE_BACK);
break;
case PIPE_WINDING_BOTH:
so_data(so, NV50TCL_CULL_FACE_FRONT_AND_BACK);
break;
default:
so_data(so, NV50TCL_CULL_FACE_BACK);
break;
}
} else {
so_data(so, NV50TCL_FRONT_FACE_CW);
switch (cso->cull_mode) {
case PIPE_WINDING_CCW:
so_data(so, NV50TCL_CULL_FACE_BACK);
break;
case PIPE_WINDING_CW:
so_data(so, NV50TCL_CULL_FACE_FRONT);
break;
case PIPE_WINDING_BOTH:
so_data(so, NV50TCL_CULL_FACE_FRONT_AND_BACK);
break;
default:
so_data(so, NV50TCL_CULL_FACE_BACK);
break;
}
}
so_method(so, tesla, NV50TCL_POLYGON_STIPPLE_ENABLE, 1);
so_data (so, cso->poly_stipple_enable ? 1 : 0);
so_method(so, tesla, NV50TCL_POLYGON_OFFSET_POINT_ENABLE, 3);
if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_POINT) ||
(cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_POINT))
so_data(so, 1);
else
so_data(so, 0);
if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_LINE) ||
(cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_LINE))
so_data(so, 1);
else
so_data(so, 0);
if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_FILL) ||
(cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_FILL))
so_data(so, 1);
else
so_data(so, 0);
if (cso->offset_cw || cso->offset_ccw) {
so_method(so, tesla, NV50TCL_POLYGON_OFFSET_FACTOR, 1);
so_data (so, fui(cso->offset_scale));
so_method(so, tesla, NV50TCL_POLYGON_OFFSET_UNITS, 1);
so_data (so, fui(cso->offset_units * 2.0f));
}
rso->pipe = *cso;
so_ref(so, &rso->so);
so_ref(NULL, &so);
return (void *)rso;
}
static int
nv50_invalidate_resource_storage(struct nouveau_context *ctx,
struct pipe_resource *res,
int ref)
{
struct nv50_context *nv50 = nv50_context(&ctx->pipe);
unsigned bind = res->bind ? res->bind : PIPE_BIND_VERTEX_BUFFER;
unsigned s, i;
if (bind & PIPE_BIND_RENDER_TARGET) {
assert(nv50->framebuffer.nr_cbufs <= PIPE_MAX_COLOR_BUFS);
for (i = 0; i < nv50->framebuffer.nr_cbufs; ++i) {
if (nv50->framebuffer.cbufs[i] &&
nv50->framebuffer.cbufs[i]->texture == res) {
nv50->dirty_3d |= NV50_NEW_3D_FRAMEBUFFER;
nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_3D_FB);
if (!--ref)
return ref;
}
}
}
if (bind & PIPE_BIND_DEPTH_STENCIL) {
if (nv50->framebuffer.zsbuf &&
nv50->framebuffer.zsbuf->texture == res) {
nv50->dirty_3d |= NV50_NEW_3D_FRAMEBUFFER;
nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_3D_FB);
if (!--ref)
return ref;
}
}
if (bind & (PIPE_BIND_VERTEX_BUFFER |
PIPE_BIND_INDEX_BUFFER |
PIPE_BIND_CONSTANT_BUFFER |
PIPE_BIND_STREAM_OUTPUT |
PIPE_BIND_SAMPLER_VIEW)) {
assert(nv50->num_vtxbufs <= PIPE_MAX_ATTRIBS);
for (i = 0; i < nv50->num_vtxbufs; ++i) {
if (nv50->vtxbuf[i].buffer == res) {
nv50->dirty_3d |= NV50_NEW_3D_ARRAYS;
nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_3D_VERTEX);
if (!--ref)
return ref;
}
}
if (nv50->idxbuf.buffer == res) {
/* Just rebind to the bufctx as there is no separate dirty bit */
nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_3D_INDEX);
BCTX_REFN(nv50->bufctx_3d, 3D_INDEX, nv04_resource(res), RD);
if (!--ref)
return ref;
}
for (s = 0; s < 3; ++s) {
assert(nv50->num_textures[s] <= PIPE_MAX_SAMPLERS);
for (i = 0; i < nv50->num_textures[s]; ++i) {
if (nv50->textures[s][i] &&
nv50->textures[s][i]->texture == res) {
nv50->dirty_3d |= NV50_NEW_3D_TEXTURES;
nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_3D_TEXTURES);
if (!--ref)
return ref;
}
}
}
for (s = 0; s < 3; ++s) {
for (i = 0; i < NV50_MAX_PIPE_CONSTBUFS; ++i) {
if (!(nv50->constbuf_valid[s] & (1 << i)))
continue;
if (!nv50->constbuf[s][i].user &&
nv50->constbuf[s][i].u.buf == res) {
nv50->dirty_3d |= NV50_NEW_3D_CONSTBUF;
nv50->constbuf_dirty[s] |= 1 << i;
nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_3D_CB(s, i));
if (!--ref)
return ref;
}
}
}
}
return ref;
}
static void *
nv50_depth_stencil_alpha_state_create(struct pipe_context *pipe,
const struct pipe_depth_stencil_alpha_state *cso)
{
struct nouveau_grobj *tesla = nv50_context(pipe)->screen->tesla;
struct nv50_zsa_stateobj *zsa = CALLOC_STRUCT(nv50_zsa_stateobj);
struct nouveau_stateobj *so = so_new(64, 0);
so_method(so, tesla, NV50TCL_DEPTH_WRITE_ENABLE, 1);
so_data (so, cso->depth.writemask ? 1 : 0);
if (cso->depth.enabled) {
so_method(so, tesla, NV50TCL_DEPTH_TEST_ENABLE, 1);
so_data (so, 1);
so_method(so, tesla, NV50TCL_DEPTH_TEST_FUNC, 1);
so_data (so, nvgl_comparison_op(cso->depth.func));
} else {
so_method(so, tesla, NV50TCL_DEPTH_TEST_ENABLE, 1);
so_data (so, 0);
}
/* XXX: keep hex values until header is updated (names reversed) */
if (cso->stencil[0].enabled) {
so_method(so, tesla, 0x1380, 8);
so_data (so, 1);
so_data (so, nvgl_stencil_op(cso->stencil[0].fail_op));
so_data (so, nvgl_stencil_op(cso->stencil[0].zfail_op));
so_data (so, nvgl_stencil_op(cso->stencil[0].zpass_op));
so_data (so, nvgl_comparison_op(cso->stencil[0].func));
so_data (so, cso->stencil[0].ref_value);
so_data (so, cso->stencil[0].writemask);
so_data (so, cso->stencil[0].valuemask);
} else {
so_method(so, tesla, 0x1380, 1);
so_data (so, 0);
}
if (cso->stencil[1].enabled) {
so_method(so, tesla, 0x1594, 5);
so_data (so, 1);
so_data (so, nvgl_stencil_op(cso->stencil[1].fail_op));
so_data (so, nvgl_stencil_op(cso->stencil[1].zfail_op));
so_data (so, nvgl_stencil_op(cso->stencil[1].zpass_op));
so_data (so, nvgl_comparison_op(cso->stencil[1].func));
so_method(so, tesla, 0x0f54, 3);
so_data (so, cso->stencil[1].ref_value);
so_data (so, cso->stencil[1].writemask);
so_data (so, cso->stencil[1].valuemask);
} else {
so_method(so, tesla, 0x1594, 1);
so_data (so, 0);
}
if (cso->alpha.enabled) {
so_method(so, tesla, NV50TCL_ALPHA_TEST_ENABLE, 1);
so_data (so, 1);
so_method(so, tesla, NV50TCL_ALPHA_TEST_REF, 2);
so_data (so, fui(cso->alpha.ref_value));
so_data (so, nvgl_comparison_op(cso->alpha.func));
} else {
so_method(so, tesla, NV50TCL_ALPHA_TEST_ENABLE, 1);
so_data (so, 0);
}
zsa->pipe = *cso;
so_ref(so, &zsa->so);
so_ref(NULL, &so);
return (void *)zsa;
}
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);
}
static boolean
nv50_begin_query(struct pipe_context *pipe, struct pipe_query *pq)
{
struct nv50_query *q = nv50_query(pq);
return q->funcs->begin_query(nv50_context(pipe), q);
}
static void
nv50_destroy_query(struct pipe_context *pipe, struct pipe_query *pq)
{
struct nv50_query *q = nv50_query(pq);
q->funcs->destroy_query(nv50_context(pipe), q);
}
static int
nv50_invalidate_resource_storage(struct nouveau_context *ctx,
struct pipe_resource *res,
int ref)
{
struct nv50_context *nv50 = nv50_context(&ctx->pipe);
unsigned s, i;
if (res->bind & PIPE_BIND_RENDER_TARGET) {
for (i = 0; i < nv50->framebuffer.nr_cbufs; ++i) {
if (nv50->framebuffer.cbufs[i] &&
nv50->framebuffer.cbufs[i]->texture == res) {
nv50->dirty |= NV50_NEW_FRAMEBUFFER;
nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_FB);
if (!--ref)
return ref;
}
}
}
if (res->bind & PIPE_BIND_DEPTH_STENCIL) {
if (nv50->framebuffer.zsbuf &&
nv50->framebuffer.zsbuf->texture == res) {
nv50->dirty |= NV50_NEW_FRAMEBUFFER;
nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_FB);
if (!--ref)
return ref;
}
}
if (res->bind & PIPE_BIND_VERTEX_BUFFER) {
for (i = 0; i < nv50->num_vtxbufs; ++i) {
if (nv50->vtxbuf[i].buffer == res) {
nv50->dirty |= NV50_NEW_ARRAYS;
nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_VERTEX);
if (!--ref)
return ref;
}
}
}
if (res->bind & PIPE_BIND_INDEX_BUFFER) {
if (nv50->idxbuf.buffer == res)
if (!--ref)
return ref;
}
if (res->bind & PIPE_BIND_SAMPLER_VIEW) {
for (s = 0; s < 5; ++s) {
for (i = 0; i < nv50->num_textures[s]; ++i) {
if (nv50->textures[s][i] &&
nv50->textures[s][i]->texture == res) {
nv50->dirty |= NV50_NEW_TEXTURES;
nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_TEXTURES);
if (!--ref)
return ref;
}
}
}
}
if (res->bind & PIPE_BIND_CONSTANT_BUFFER) {
for (s = 0; s < 5; ++s) {
for (i = 0; i < nv50->num_vtxbufs; ++i) {
if (!nv50->constbuf[s][i].user &&
nv50->constbuf[s][i].u.buf == res) {
nv50->dirty |= NV50_NEW_CONSTBUF;
nv50->constbuf_dirty[s] |= 1 << i;
nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_CB(s, i));
if (!--ref)
return ref;
}
}
}
}
return ref;
}
static boolean
nv50_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_pushbuf *push = nv50->base.pushbuf;
struct nv50_query *q = nv50_query(pq);
/* For occlusion queries we have to change the storage, because a previous
* query might set the initial render conition to FALSE even *after* we re-
* initialized it to TRUE.
*/
if (q->type == PIPE_QUERY_OCCLUSION_COUNTER) {
q->offset += 32;
q->data += 32 / sizeof(*q->data);
if (q->offset - q->base == NV50_QUERY_ALLOC_SPACE)
nv50_query_allocate(nv50, q, NV50_QUERY_ALLOC_SPACE);
/* XXX: can we do this with the GPU, and sync with respect to a previous
* query ?
*/
q->data[0] = q->sequence; /* initialize sequence */
q->data[1] = 1; /* initial render condition = TRUE */
q->data[4] = q->sequence + 1; /* for comparison COND_MODE */
q->data[5] = 0;
}
if (!q->is64bit)
q->data[0] = q->sequence++; /* the previously used one */
switch (q->type) {
case PIPE_QUERY_OCCLUSION_COUNTER:
PUSH_SPACE(push, 4);
BEGIN_NV04(push, NV50_3D(COUNTER_RESET), 1);
PUSH_DATA (push, NV50_3D_COUNTER_RESET_SAMPLECNT);
BEGIN_NV04(push, NV50_3D(SAMPLECNT_ENABLE), 1);
PUSH_DATA (push, 1);
break;
case PIPE_QUERY_PRIMITIVES_GENERATED:
nv50_query_get(push, q, 0x10, 0x06805002);
break;
case PIPE_QUERY_PRIMITIVES_EMITTED:
nv50_query_get(push, q, 0x10, 0x05805002);
break;
case PIPE_QUERY_SO_STATISTICS:
nv50_query_get(push, q, 0x20, 0x05805002);
nv50_query_get(push, q, 0x30, 0x06805002);
break;
case PIPE_QUERY_PIPELINE_STATISTICS:
nv50_query_get(push, q, 0x80, 0x00801002); /* VFETCH, VERTICES */
nv50_query_get(push, q, 0x90, 0x01801002); /* VFETCH, PRIMS */
nv50_query_get(push, q, 0xa0, 0x02802002); /* VP, LAUNCHES */
nv50_query_get(push, q, 0xb0, 0x03806002); /* GP, LAUNCHES */
nv50_query_get(push, q, 0xc0, 0x04806002); /* GP, PRIMS_OUT */
nv50_query_get(push, q, 0xd0, 0x07804002); /* RAST, PRIMS_IN */
nv50_query_get(push, q, 0xe0, 0x08804002); /* RAST, PRIMS_OUT */
nv50_query_get(push, q, 0xf0, 0x0980a002); /* ROP, PIXELS */
break;
case PIPE_QUERY_TIME_ELAPSED:
nv50_query_get(push, q, 0x10, 0x00005002);
break;
default:
break;
}
q->ready = FALSE;
return true;
}
static void *
nv50_depth_stencil_alpha_state_create(struct pipe_context *pipe,
const struct pipe_depth_stencil_alpha_state *cso)
{
struct nouveau_grobj *tesla = nv50_context(pipe)->screen->tesla;
struct nv50_zsa_stateobj *zsa = CALLOC_STRUCT(nv50_zsa_stateobj);
struct nouveau_stateobj *so = so_new(9, 21, 0);
so_method(so, tesla, NV50TCL_DEPTH_WRITE_ENABLE, 1);
so_data (so, cso->depth.writemask ? 1 : 0);
if (cso->depth.enabled) {
so_method(so, tesla, NV50TCL_DEPTH_TEST_ENABLE, 1);
so_data (so, 1);
so_method(so, tesla, NV50TCL_DEPTH_TEST_FUNC, 1);
so_data (so, nvgl_comparison_op(cso->depth.func));
} else {
so_method(so, tesla, NV50TCL_DEPTH_TEST_ENABLE, 1);
so_data (so, 0);
}
if (cso->stencil[0].enabled) {
so_method(so, tesla, NV50TCL_STENCIL_FRONT_ENABLE, 5);
so_data (so, 1);
so_data (so, nvgl_stencil_op(cso->stencil[0].fail_op));
so_data (so, nvgl_stencil_op(cso->stencil[0].zfail_op));
so_data (so, nvgl_stencil_op(cso->stencil[0].zpass_op));
so_data (so, nvgl_comparison_op(cso->stencil[0].func));
so_method(so, tesla, NV50TCL_STENCIL_FRONT_MASK, 2);
so_data (so, cso->stencil[0].writemask);
so_data (so, cso->stencil[0].valuemask);
} else {
so_method(so, tesla, NV50TCL_STENCIL_FRONT_ENABLE, 1);
so_data (so, 0);
}
if (cso->stencil[1].enabled) {
so_method(so, tesla, NV50TCL_STENCIL_BACK_ENABLE, 5);
so_data (so, 1);
so_data (so, nvgl_stencil_op(cso->stencil[1].fail_op));
so_data (so, nvgl_stencil_op(cso->stencil[1].zfail_op));
so_data (so, nvgl_stencil_op(cso->stencil[1].zpass_op));
so_data (so, nvgl_comparison_op(cso->stencil[1].func));
so_method(so, tesla, NV50TCL_STENCIL_BACK_MASK, 2);
so_data (so, cso->stencil[1].writemask);
so_data (so, cso->stencil[1].valuemask);
} else {
so_method(so, tesla, NV50TCL_STENCIL_BACK_ENABLE, 1);
so_data (so, 0);
}
if (cso->alpha.enabled) {
so_method(so, tesla, NV50TCL_ALPHA_TEST_ENABLE, 1);
so_data (so, 1);
so_method(so, tesla, NV50TCL_ALPHA_TEST_REF, 2);
so_data (so, fui(cso->alpha.ref_value));
so_data (so, nvgl_comparison_op(cso->alpha.func));
} else {
so_method(so, tesla, NV50TCL_ALPHA_TEST_ENABLE, 1);
so_data (so, 0);
}
zsa->pipe = *cso;
so_ref(so, &zsa->so);
so_ref(NULL, &so);
return (void *)zsa;
}
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);
}
static void *
nv50_blend_state_create(struct pipe_context *pipe,
const struct pipe_blend_state *cso)
{
struct nouveau_stateobj *so = so_new(5, 24, 0);
struct nouveau_grobj *tesla = nv50_context(pipe)->screen->tesla;
struct nv50_blend_stateobj *bso = CALLOC_STRUCT(nv50_blend_stateobj);
unsigned i, blend_enabled = 0;
/*XXX ignored:
* - dither
*/
so_method(so, tesla, NV50TCL_BLEND_ENABLE(0), 8);
if (cso->independent_blend_enable) {
for (i = 0; i < 8; ++i) {
so_data(so, cso->rt[i].blend_enable);
if (cso->rt[i].blend_enable)
blend_enabled = 1;
}
} else if (cso->rt[0].blend_enable) {
blend_enabled = 1;
for (i = 0; i < 8; i++)
so_data(so, 1);
} else {
for (i = 0; i < 8; i++)
so_data(so, 0);
}
if (blend_enabled) {
so_method(so, tesla, NV50TCL_BLEND_EQUATION_RGB, 5);
so_data (so, nvgl_blend_eqn(cso->rt[0].rgb_func));
so_data (so, 0x4000 | nvgl_blend_func(cso->rt[0].rgb_src_factor));
so_data (so, 0x4000 | nvgl_blend_func(cso->rt[0].rgb_dst_factor));
so_data (so, nvgl_blend_eqn(cso->rt[0].alpha_func));
so_data (so, 0x4000 | nvgl_blend_func(cso->rt[0].alpha_src_factor));
so_method(so, tesla, NV50TCL_BLEND_FUNC_DST_ALPHA, 1);
so_data (so, 0x4000 | nvgl_blend_func(cso->rt[0].alpha_dst_factor));
}
if (cso->logicop_enable == 0 ) {
so_method(so, tesla, NV50TCL_LOGIC_OP_ENABLE, 1);
so_data (so, 0);
} else {
so_method(so, tesla, NV50TCL_LOGIC_OP_ENABLE, 2);
so_data (so, 1);
so_data (so, nvgl_logicop_func(cso->logicop_func));
}
so_method(so, tesla, NV50TCL_COLOR_MASK(0), 8);
if (cso->independent_blend_enable)
for (i = 0; i < 8; ++i)
so_data(so, nv50_colormask(cso->rt[i].colormask));
else {
uint32_t cmask = nv50_colormask(cso->rt[0].colormask);
for (i = 0; i < 8; i++)
so_data(so, cmask);
}
bso->pipe = *cso;
so_ref(so, &bso->so);
so_ref(NULL, &so);
return (void *)bso;
}
static void *
nv50_blend_state_create(struct pipe_context *pipe,
const struct pipe_blend_state *cso)
{
struct nv50_blend_stateobj *so = CALLOC_STRUCT(nv50_blend_stateobj);
int i;
boolean emit_common_func = cso->rt[0].blend_enable;
uint32_t ms;
if (nv50_context(pipe)->screen->tesla->oclass >= NVA3_3D_CLASS) {
SB_BEGIN_3D(so, BLEND_INDEPENDENT, 1);
SB_DATA (so, cso->independent_blend_enable);
}
so->pipe = *cso;
SB_BEGIN_3D(so, COLOR_MASK_COMMON, 1);
SB_DATA (so, !cso->independent_blend_enable);
SB_BEGIN_3D(so, BLEND_ENABLE_COMMON, 1);
SB_DATA (so, !cso->independent_blend_enable);
if (cso->independent_blend_enable) {
SB_BEGIN_3D(so, BLEND_ENABLE(0), 8);
for (i = 0; i < 8; ++i) {
SB_DATA(so, cso->rt[i].blend_enable);
if (cso->rt[i].blend_enable)
emit_common_func = TRUE;
}
if (nv50_context(pipe)->screen->tesla->oclass >= NVA3_3D_CLASS) {
emit_common_func = FALSE;
for (i = 0; i < 8; ++i) {
if (!cso->rt[i].blend_enable)
continue;
SB_BEGIN_3D_(so, NVA3_3D_IBLEND_EQUATION_RGB(i), 6);
SB_DATA (so, nvgl_blend_eqn(cso->rt[i].rgb_func));
SB_DATA (so, nv50_blend_fac(cso->rt[i].rgb_src_factor));
SB_DATA (so, nv50_blend_fac(cso->rt[i].rgb_dst_factor));
SB_DATA (so, nvgl_blend_eqn(cso->rt[i].alpha_func));
SB_DATA (so, nv50_blend_fac(cso->rt[i].alpha_src_factor));
SB_DATA (so, nv50_blend_fac(cso->rt[i].alpha_dst_factor));
}
}
} else {
SB_BEGIN_3D(so, BLEND_ENABLE(0), 1);
SB_DATA (so, cso->rt[0].blend_enable);
}
if (emit_common_func) {
SB_BEGIN_3D(so, BLEND_EQUATION_RGB, 5);
SB_DATA (so, nvgl_blend_eqn(cso->rt[0].rgb_func));
SB_DATA (so, nv50_blend_fac(cso->rt[0].rgb_src_factor));
SB_DATA (so, nv50_blend_fac(cso->rt[0].rgb_dst_factor));
SB_DATA (so, nvgl_blend_eqn(cso->rt[0].alpha_func));
SB_DATA (so, nv50_blend_fac(cso->rt[0].alpha_src_factor));
SB_BEGIN_3D(so, BLEND_FUNC_DST_ALPHA, 1);
SB_DATA (so, nv50_blend_fac(cso->rt[0].alpha_dst_factor));
}
if (cso->logicop_enable) {
SB_BEGIN_3D(so, LOGIC_OP_ENABLE, 2);
SB_DATA (so, 1);
SB_DATA (so, nvgl_logicop_func(cso->logicop_func));
} else {
SB_BEGIN_3D(so, LOGIC_OP_ENABLE, 1);
SB_DATA (so, 0);
}
if (cso->independent_blend_enable) {
SB_BEGIN_3D(so, COLOR_MASK(0), 8);
for (i = 0; i < 8; ++i)
SB_DATA(so, nv50_colormask(cso->rt[i].colormask));
} else {
SB_BEGIN_3D(so, COLOR_MASK(0), 1);
SB_DATA (so, nv50_colormask(cso->rt[0].colormask));
}
ms = 0;
if (cso->alpha_to_coverage)
ms |= NV50_3D_MULTISAMPLE_CTRL_ALPHA_TO_COVERAGE;
if (cso->alpha_to_one)
ms |= NV50_3D_MULTISAMPLE_CTRL_ALPHA_TO_ONE;
SB_BEGIN_3D(so, MULTISAMPLE_CTRL, 1);
SB_DATA (so, ms);
assert(so->size <= (sizeof(so->state) / sizeof(so->state[0])));
return so;
}
static void *
nv50_blend_state_create(struct pipe_context *pipe,
const struct pipe_blend_state *cso)
{
struct nv50_blend_stateobj *so = CALLOC_STRUCT(nv50_blend_stateobj);
int i;
bool emit_common_func = cso->rt[0].blend_enable;
if (nv50_context(pipe)->screen->tesla->oclass >= NVA3_3D_CLASS) {
SB_BEGIN_3D(so, BLEND_INDEPENDENT, 1);
SB_DATA (so, cso->independent_blend_enable);
}
so->pipe = *cso;
SB_BEGIN_3D(so, COLOR_MASK_COMMON, 1);
SB_DATA (so, !cso->independent_blend_enable);
SB_BEGIN_3D(so, BLEND_ENABLE_COMMON, 1);
SB_DATA (so, !cso->independent_blend_enable);
if (cso->independent_blend_enable) {
SB_BEGIN_3D(so, BLEND_ENABLE(0), 8);
for (i = 0; i < 8; ++i) {
SB_DATA(so, cso->rt[i].blend_enable);
if (cso->rt[i].blend_enable)
emit_common_func = true;
}
if (nv50_context(pipe)->screen->tesla->oclass >= NVA3_3D_CLASS) {
emit_common_func = false;
for (i = 0; i < 8; ++i) {
if (!cso->rt[i].blend_enable)
continue;
SB_BEGIN_3D_(so, NVA3_3D_IBLEND_EQUATION_RGB(i), 6);
SB_DATA (so, nvgl_blend_eqn(cso->rt[i].rgb_func));
SB_DATA (so, nv50_blend_fac(cso->rt[i].rgb_src_factor));
SB_DATA (so, nv50_blend_fac(cso->rt[i].rgb_dst_factor));
SB_DATA (so, nvgl_blend_eqn(cso->rt[i].alpha_func));
SB_DATA (so, nv50_blend_fac(cso->rt[i].alpha_src_factor));
SB_DATA (so, nv50_blend_fac(cso->rt[i].alpha_dst_factor));
}
}
} else {
SB_BEGIN_3D(so, BLEND_ENABLE(0), 1);
SB_DATA (so, cso->rt[0].blend_enable);
}
if (emit_common_func) {
SB_BEGIN_3D(so, BLEND_EQUATION_RGB, 5);
SB_DATA (so, nvgl_blend_eqn(cso->rt[0].rgb_func));
SB_DATA (so, nv50_blend_fac(cso->rt[0].rgb_src_factor));
SB_DATA (so, nv50_blend_fac(cso->rt[0].rgb_dst_factor));
SB_DATA (so, nvgl_blend_eqn(cso->rt[0].alpha_func));
SB_DATA (so, nv50_blend_fac(cso->rt[0].alpha_src_factor));
SB_BEGIN_3D(so, BLEND_FUNC_DST_ALPHA, 1);
SB_DATA (so, nv50_blend_fac(cso->rt[0].alpha_dst_factor));
}
if (cso->logicop_enable) {
SB_BEGIN_3D(so, LOGIC_OP_ENABLE, 2);
SB_DATA (so, 1);
SB_DATA (so, nvgl_logicop_func(cso->logicop_func));
} else {
SB_BEGIN_3D(so, LOGIC_OP_ENABLE, 1);
SB_DATA (so, 0);
}
if (cso->independent_blend_enable) {
SB_BEGIN_3D(so, COLOR_MASK(0), 8);
for (i = 0; i < 8; ++i)
SB_DATA(so, nv50_colormask(cso->rt[i].colormask));
} else {
SB_BEGIN_3D(so, COLOR_MASK(0), 1);
SB_DATA (so, nv50_colormask(cso->rt[0].colormask));
}
assert(so->size <= ARRAY_SIZE(so->state));
return so;
}
static void
nv50_resource_copy_region(struct pipe_context *pipe,
struct pipe_resource *dst, unsigned dst_level,
unsigned dstx, unsigned dsty, unsigned dstz,
struct pipe_resource *src, unsigned src_level,
const struct pipe_box *src_box)
{
struct nv50_context *nv50 = nv50_context(pipe);
int ret;
boolean m2mf;
unsigned dst_layer = dstz, src_layer = src_box->z;
/* Fallback for buffers. */
if (dst->target == PIPE_BUFFER && src->target == PIPE_BUFFER) {
util_resource_copy_region(pipe, dst, dst_level, dstx, dsty, dstz,
src, src_level, src_box);
return;
}
/* 0 and 1 are equal, only supporting 0/1, 2, 4 and 8 */
assert((src->nr_samples | 1) == (dst->nr_samples | 1));
m2mf = (src->format == dst->format) ||
(util_format_get_blocksizebits(src->format) ==
util_format_get_blocksizebits(dst->format));
nv04_resource(dst)->status |= NOUVEAU_BUFFER_STATUS_GPU_WRITING;
if (m2mf) {
struct nv50_m2mf_rect drect, srect;
unsigned i;
unsigned nx = util_format_get_nblocksx(src->format, src_box->width);
unsigned ny = util_format_get_nblocksy(src->format, src_box->height);
nv50_m2mf_rect_setup(&drect, dst, dst_level, dstx, dsty, dstz);
nv50_m2mf_rect_setup(&srect, src, src_level,
src_box->x, src_box->y, src_box->z);
for (i = 0; i < src_box->depth; ++i) {
nv50_m2mf_transfer_rect(nv50, &drect, &srect, nx, ny);
if (nv50_miptree(dst)->layout_3d)
drect.z++;
else
drect.base += nv50_miptree(dst)->layer_stride;
if (nv50_miptree(src)->layout_3d)
srect.z++;
else
srect.base += nv50_miptree(src)->layer_stride;
}
return;
}
assert((src->format == dst->format) ||
(nv50_2d_format_faithful(src->format) &&
nv50_2d_format_faithful(dst->format)));
BCTX_REFN(nv50->bufctx, 2D, nv04_resource(src), RD);
BCTX_REFN(nv50->bufctx, 2D, nv04_resource(dst), WR);
nouveau_pushbuf_bufctx(nv50->base.pushbuf, nv50->bufctx);
nouveau_pushbuf_validate(nv50->base.pushbuf);
for (; dst_layer < dstz + src_box->depth; ++dst_layer, ++src_layer) {
ret = nv50_2d_texture_do_copy(nv50->base.pushbuf,
nv50_miptree(dst), dst_level,
dstx, dsty, dst_layer,
nv50_miptree(src), src_level,
src_box->x, src_box->y, src_layer,
src_box->width, src_box->height);
if (ret)
break;
}
nouveau_bufctx_reset(nv50->bufctx, NV50_BIND_2D);
}
static void
nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq)
{
struct nv50_context *nv50 = nv50_context(pipe);
struct nouveau_pushbuf *push = nv50->base.pushbuf;
struct nv50_query *q = nv50_query(pq);
q->state = NV50_QUERY_STATE_ENDED;
switch (q->type) {
case PIPE_QUERY_OCCLUSION_COUNTER:
nv50_query_get(push, q, 0, 0x0100f002);
if (--nv50->screen->num_occlusion_queries_active == 0) {
PUSH_SPACE(push, 2);
BEGIN_NV04(push, NV50_3D(SAMPLECNT_ENABLE), 1);
PUSH_DATA (push, 0);
}
break;
case PIPE_QUERY_PRIMITIVES_GENERATED:
nv50_query_get(push, q, 0, 0x06805002);
break;
case PIPE_QUERY_PRIMITIVES_EMITTED:
nv50_query_get(push, q, 0, 0x05805002);
break;
case PIPE_QUERY_SO_STATISTICS:
nv50_query_get(push, q, 0x00, 0x05805002);
nv50_query_get(push, q, 0x10, 0x06805002);
break;
case PIPE_QUERY_PIPELINE_STATISTICS:
nv50_query_get(push, q, 0x00, 0x00801002); /* VFETCH, VERTICES */
nv50_query_get(push, q, 0x10, 0x01801002); /* VFETCH, PRIMS */
nv50_query_get(push, q, 0x20, 0x02802002); /* VP, LAUNCHES */
nv50_query_get(push, q, 0x30, 0x03806002); /* GP, LAUNCHES */
nv50_query_get(push, q, 0x40, 0x04806002); /* GP, PRIMS_OUT */
nv50_query_get(push, q, 0x50, 0x07804002); /* RAST, PRIMS_IN */
nv50_query_get(push, q, 0x60, 0x08804002); /* RAST, PRIMS_OUT */
nv50_query_get(push, q, 0x70, 0x0980a002); /* ROP, PIXELS */
break;
case PIPE_QUERY_TIMESTAMP:
q->sequence++;
/* fall through */
case PIPE_QUERY_TIME_ELAPSED:
nv50_query_get(push, q, 0, 0x00005002);
break;
case PIPE_QUERY_GPU_FINISHED:
q->sequence++;
nv50_query_get(push, q, 0, 0x1000f010);
break;
case NVA0_QUERY_STREAM_OUTPUT_BUFFER_OFFSET:
q->sequence++;
nv50_query_get(push, q, 0, 0x0d005002 | (q->index << 5));
break;
case PIPE_QUERY_TIMESTAMP_DISJOINT:
/* This query is not issued on GPU because disjoint is forced to false */
q->state = NV50_QUERY_STATE_READY;
break;
default:
assert(0);
break;
}
if (q->is64bit)
nouveau_fence_ref(nv50->screen->base.fence.current, &q->fence);
}
