static void
nv30_draw_elements(struct nv30_context *nv30, boolean shorten,
unsigned mode, unsigned start, unsigned count,
unsigned instance_count, int32_t index_bias)
{
const unsigned index_size = nv30->idxbuf.index_size;
struct nouveau_pushbuf *push = nv30->base.pushbuf;
struct nouveau_object *eng3d = nv30->screen->eng3d;
unsigned prim = nv30_prim_gl(mode);
#if 0 /*XXX*/
if (index_bias != nv30->state.index_bias) {
BEGIN_NV04(push, NV30_3D(VB_ELEMENT_BASE), 1);
PUSH_DATA (push, index_bias);
nv30->state.index_bias = index_bias;
}
#endif
if (eng3d->oclass == NV40_3D_CLASS && index_size > 1 &&
nv30->idxbuf.buffer) {
struct nv04_resource *res = nv04_resource(nv30->idxbuf.buffer);
unsigned offset = nv30->idxbuf.offset;
assert(nouveau_resource_mapped_by_gpu(&res->base));
BEGIN_NV04(push, NV30_3D(IDXBUF_OFFSET), 2);
PUSH_RESRC(push, NV30_3D(IDXBUF_OFFSET), BUFCTX_IDXBUF, res, offset,
NOUVEAU_BO_LOW | NOUVEAU_BO_RD, 0, 0);
PUSH_MTHD (push, NV30_3D(IDXBUF_FORMAT), BUFCTX_IDXBUF, res->bo,
(index_size == 2) ? 0x00000010 : 0x00000000,
res->domain | NOUVEAU_BO_RD,
0, NV30_3D_IDXBUF_FORMAT_DMA1);
BEGIN_NV04(push, NV30_3D(VERTEX_BEGIN_END), 1);
PUSH_DATA (push, prim);
while (count) {
const unsigned mpush = 2047 * 256;
unsigned npush = (count > mpush) ? mpush : count;
unsigned wpush = ((npush + 255) & ~255) >> 8;
count -= npush;
BEGIN_NI04(push, NV30_3D(VB_INDEX_BATCH), wpush);
while (npush >= 256) {
PUSH_DATA (push, 0xff000000 | start);
start += 256;
npush -= 256;
}
if (npush)
PUSH_DATA (push, ((npush - 1) << 24) | start);
}
BEGIN_NV04(push, NV30_3D(VERTEX_BEGIN_END), 1);
PUSH_DATA (push, NV30_3D_VERTEX_BEGIN_END_STOP);
PUSH_RESET(push, BUFCTX_IDXBUF);
} else {
static void
nv30_render_draw_elements(struct vbuf_render *render,
const ushort *indices, uint count)
{
struct nv30_render *r = nv30_render(render);
struct nv30_context *nv30 = r->nv30;
struct nouveau_pushbuf *push = nv30->screen->base.pushbuf;
unsigned i;
BEGIN_NV04(push, NV30_3D(VTXBUF(0)), r->vertex_info.num_attribs);
for (i = 0; i < r->vertex_info.num_attribs; i++) {
PUSH_RESRC(push, NV30_3D(VTXBUF(i)), BUFCTX_VTXTMP,
nv04_resource(r->buffer), r->offset + r->vtxptr[i],
NOUVEAU_BO_LOW | NOUVEAU_BO_RD, 0, NV30_3D_VTXBUF_DMA1);
}
if (!nv30_state_validate(nv30, FALSE))
return;
BEGIN_NV04(push, NV30_3D(VERTEX_BEGIN_END), 1);
PUSH_DATA (push, r->prim);
if (count & 1) {
BEGIN_NV04(push, NV30_3D(VB_ELEMENT_U32), 1);
PUSH_DATA (push, *indices++);
}
count >>= 1;
while (count) {
unsigned npush = MIN2(count, NV04_PFIFO_MAX_PACKET_LEN);
count -= npush;
BEGIN_NI04(push, NV30_3D(VB_ELEMENT_U16), npush);
while (npush--) {
PUSH_DATA(push, (indices[1] << 16) | indices[0]);
indices += 2;
}
}
BEGIN_NV04(push, NV30_3D(VERTEX_BEGIN_END), 1);
PUSH_DATA (push, NV30_3D_VERTEX_BEGIN_END_STOP);
PUSH_RESET(push, BUFCTX_VTXTMP);
}
static void
nv30_update_user_vbufs(struct nv30_context *nv30)
{
struct nouveau_pushbuf *push = nv30->base.pushbuf;
uint32_t base, offset, size;
int i;
uint32_t written = 0;
for (i = 0; i < nv30->vertex->num_elements; i++) {
struct pipe_vertex_element *ve = &nv30->vertex->pipe[i];
const int b = ve->vertex_buffer_index;
struct pipe_vertex_buffer *vb = &nv30->vtxbuf[b];
struct nv04_resource *buf = nv04_resource(vb->buffer);
if (!(nv30->vbo_user & (1 << b)))
continue;
if (!vb->stride) {
nv30_emit_vtxattr(nv30, vb, ve, i);
continue;
}
nv30_vbuf_range(nv30, b, &base, &size);
if (!(written & (1 << b))) {
written |= 1 << b;
nouveau_user_buffer_upload(&nv30->base, buf, base, size);
}
offset = vb->buffer_offset + ve->src_offset;
BEGIN_NV04(push, NV30_3D(VTXBUF(i)), 1);
PUSH_RESRC(push, NV30_3D(VTXBUF(i)), BUFCTX_VTXTMP, buf, offset,
NOUVEAU_BO_LOW | NOUVEAU_BO_RD,
0, NV30_3D_VTXBUF_DMA1);
}
nv30->base.vbo_dirty = TRUE;
}
static void
nv30_render_draw_arrays(struct vbuf_render *render, unsigned start, uint nr)
{
struct nv30_render *r = nv30_render(render);
struct nv30_context *nv30 = r->nv30;
struct nouveau_pushbuf *push = nv30->base.pushbuf;
unsigned fn = nr >> 8, pn = nr & 0xff;
unsigned ps = fn + (pn ? 1 : 0);
unsigned i;
BEGIN_NV04(push, NV30_3D(VTXBUF(0)), r->vertex_info.num_attribs);
for (i = 0; i < r->vertex_info.num_attribs; i++) {
PUSH_RESRC(push, NV30_3D(VTXBUF(i)), BUFCTX_VTXTMP,
nv04_resource(r->buffer), r->offset + r->vtxptr[i],
NOUVEAU_BO_LOW | NOUVEAU_BO_RD, 0, NV30_3D_VTXBUF_DMA1);
}
if (!nv30_state_validate(nv30, FALSE))
return;
BEGIN_NV04(push, NV30_3D(VERTEX_BEGIN_END), 1);
PUSH_DATA (push, r->prim);
BEGIN_NI04(push, NV30_3D(VB_VERTEX_BATCH), ps);
while (fn--) {
PUSH_DATA (push, 0xff000000 | start);
start += 256;
}
if (pn)
PUSH_DATA (push, ((pn - 1) << 24) | start);
BEGIN_NV04(push, NV30_3D(VERTEX_BEGIN_END), 1);
PUSH_DATA (push, NV30_3D_VERTEX_BEGIN_END_STOP);
PUSH_RESET(push, BUFCTX_VTXTMP);
}
void
nv30_vbo_validate(struct nv30_context *nv30)
{
struct nouveau_pushbuf *push = nv30->base.pushbuf;
struct nv30_vertex_stateobj *vertex = nv30->vertex;
struct pipe_vertex_element *ve;
struct pipe_vertex_buffer *vb;
unsigned i, redefine;
nouveau_bufctx_reset(nv30->bufctx, BUFCTX_VTXBUF);
if (!nv30->vertex || nv30->draw_flags)
return;
if (unlikely(vertex->need_conversion)) {
nv30->vbo_fifo = ~0;
nv30->vbo_user = 0;
} else {
nv30_prevalidate_vbufs(nv30);
}
if (!PUSH_SPACE(push, 128))
return;
redefine = MAX2(vertex->num_elements, nv30->state.num_vtxelts);
BEGIN_NV04(push, NV30_3D(VTXFMT(0)), redefine);
for (i = 0; i < vertex->num_elements; i++) {
ve = &vertex->pipe[i];
vb = &nv30->vtxbuf[ve->vertex_buffer_index];
if (likely(vb->stride) || nv30->vbo_fifo)
PUSH_DATA (push, (vb->stride << 8) | vertex->element[i].state);
else
PUSH_DATA (push, NV30_3D_VTXFMT_TYPE_V32_FLOAT);
}
for (; i < nv30->state.num_vtxelts; i++) {
PUSH_DATA (push, NV30_3D_VTXFMT_TYPE_V32_FLOAT);
}
for (i = 0; i < vertex->num_elements; i++) {
struct nv04_resource *res;
unsigned offset;
boolean user;
ve = &vertex->pipe[i];
vb = &nv30->vtxbuf[ve->vertex_buffer_index];
user = (nv30->vbo_user & (1 << ve->vertex_buffer_index));
res = nv04_resource(vb->buffer);
if (nv30->vbo_fifo || unlikely(vb->stride == 0)) {
if (!nv30->vbo_fifo)
nv30_emit_vtxattr(nv30, vb, ve, i);
continue;
}
offset = ve->src_offset + vb->buffer_offset;
BEGIN_NV04(push, NV30_3D(VTXBUF(i)), 1);
PUSH_RESRC(push, NV30_3D(VTXBUF(i)), user ? BUFCTX_VTXTMP : BUFCTX_VTXBUF,
res, offset, NOUVEAU_BO_LOW | NOUVEAU_BO_RD,
0, NV30_3D_VTXBUF_DMA1);
}
nv30->state.num_vtxelts = vertex->num_elements;
}
void
nv30_fragprog_validate(struct nv30_context *nv30)
{
struct nouveau_pushbuf *push = nv30->base.pushbuf;
struct nouveau_object *eng3d = nv30->screen->eng3d;
struct nv30_fragprog *fp = nv30->fragprog.program;
bool upload = false;
int i;
if (!fp->translated) {
_nvfx_fragprog_translate(eng3d->oclass, fp);
if (!fp->translated)
return;
upload = true;
}
/* update constants, also needs to be done on every fp switch as we
* have no idea whether the constbuf changed in the meantime
*/
if (nv30->fragprog.constbuf) {
struct pipe_resource *constbuf = nv30->fragprog.constbuf;
uint32_t *cbuf = (uint32_t *)nv04_resource(constbuf)->data;
for (i = 0; i < fp->nr_consts; i++) {
unsigned off = fp->consts[i].offset;
unsigned idx = fp->consts[i].index * 4;
if (!memcmp(&fp->insn[off], &cbuf[idx], 4 * 4))
continue;
memcpy(&fp->insn[off], &cbuf[idx], 4 * 4);
upload = true;
}
}
if (upload)
nv30_fragprog_upload(nv30);
/* FP_ACTIVE_PROGRAM needs to be done again even if only the consts
* were updated. TEX_CACHE_CTL magic is not enough to convince the
* GPU that it should re-read the fragprog from VRAM... sigh.
*/
if (nv30->state.fragprog != fp || upload) {
struct nv04_resource *r = nv04_resource(fp->buffer);
if (!PUSH_SPACE(push, 8))
return;
PUSH_RESET(push, BUFCTX_FRAGPROG);
BEGIN_NV04(push, NV30_3D(FP_ACTIVE_PROGRAM), 1);
PUSH_RESRC(push, NV30_3D(FP_ACTIVE_PROGRAM), BUFCTX_FRAGPROG, r, 0,
NOUVEAU_BO_LOW | NOUVEAU_BO_RD | NOUVEAU_BO_OR,
NV30_3D_FP_ACTIVE_PROGRAM_DMA0,
NV30_3D_FP_ACTIVE_PROGRAM_DMA1);
BEGIN_NV04(push, NV30_3D(FP_CONTROL), 1);
PUSH_DATA (push, fp->fp_control);
if (eng3d->oclass < NV40_3D_CLASS) {
BEGIN_NV04(push, NV30_3D(FP_REG_CONTROL), 1);
PUSH_DATA (push, 0x00010004);
BEGIN_NV04(push, NV30_3D(TEX_UNITS_ENABLE), 1);
PUSH_DATA (push, fp->texcoords);
} else {
BEGIN_NV04(push, SUBC_3D(0x0b40), 1);
PUSH_DATA (push, 0x00000000);
}
nv30->state.fragprog = fp;
}
}
