static Bool
NV40EXAPictSolid(NVPtr pNv, PicturePtr pPict, int unit)
{
struct nouveau_pushbuf *push = pNv->pushbuf;
PUSH_DATAu(push, pNv->scratch, SOLID(unit), 2);
PUSH_DATA (push, pPict->pSourcePict->solidFill.color);
PUSH_DATA (push, 0);
BEGIN_NV04(push, NV30_3D(TEX_OFFSET(unit)), 8);
PUSH_MTHDl(push, NV30_3D(TEX_OFFSET(unit)), pNv->scratch, SOLID(unit),
NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
PUSH_DATA (push, NV40_3D_TEX_FORMAT_FORMAT_A8R8G8B8 | 0x8000 |
NV40_3D_TEX_FORMAT_LINEAR |
NV30_3D_TEX_FORMAT_DIMS_2D |
NV30_3D_TEX_FORMAT_NO_BORDER |
(1 << NV40_3D_TEX_FORMAT_MIPMAP_COUNT__SHIFT) |
NV30_3D_TEX_FORMAT_DMA0);
PUSH_DATA (push, NV30_3D_TEX_WRAP_S_REPEAT |
NV30_3D_TEX_WRAP_T_REPEAT |
NV30_3D_TEX_WRAP_R_REPEAT);
PUSH_DATA (push, NV40_3D_TEX_ENABLE_ENABLE);
PUSH_DATA (push, 0x0000aae4);
PUSH_DATA (push, NV30_3D_TEX_FILTER_MIN_NEAREST |
NV30_3D_TEX_FILTER_MAG_NEAREST | 0x3fd6);
PUSH_DATA (push, 0x00010001);
PUSH_DATA (push, 0x00000000);
BEGIN_NV04(push, NV40_3D(TEX_SIZE1(unit)), 1);
PUSH_DATA (push, 0x00100040);
BEGIN_NV04(push, NV30_3D(VP_UPLOAD_CONST_ID), 17);
PUSH_DATA (push, unit * 4);
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
return TRUE;
}
void
nv40_verttex_validate(struct nv30_context *nv30)
{
struct nouveau_pushbuf *push = nv30->base.pushbuf;
unsigned dirty = nv30->vertprog.dirty_samplers;
while (dirty) {
unsigned unit = ffs(dirty) - 1;
struct nv30_sampler_view *sv = (void *)nv30->fragprog.textures[unit];
struct nv30_sampler_state *ss = nv30->fragprog.samplers[unit];
if (ss && sv) {
} else {
BEGIN_NV04(push, NV40_3D(VTXTEX_ENABLE(unit)), 1);
PUSH_DATA (push, 0);
}
}
nv30->vertprog.dirty_samplers = 0;
}
static void
nv30_clear_depth_stencil(struct pipe_context *pipe, struct pipe_surface *ps,
unsigned buffers, double depth, unsigned stencil,
unsigned x, unsigned y, unsigned w, unsigned h)
{
struct nv30_context *nv30 = nv30_context(pipe);
struct nv30_surface *sf = nv30_surface(ps);
struct nv30_miptree *mt = nv30_miptree(ps->texture);
struct nouveau_pushbuf *push = nv30->base.pushbuf;
struct nouveau_object *eng3d = nv30->screen->eng3d;
struct nouveau_pushbuf_refn refn;
uint32_t rt_format, mode = 0;
rt_format = nv30_format(pipe->screen, ps->format)->hw;
if (util_format_get_blocksize(ps->format) == 4)
rt_format |= NV30_3D_RT_FORMAT_COLOR_A8R8G8B8;
else
rt_format |= NV30_3D_RT_FORMAT_COLOR_R5G6B5;
if (nv30_miptree(ps->texture)->swizzled) {
rt_format |= NV30_3D_RT_FORMAT_TYPE_SWIZZLED;
rt_format |= util_logbase2(sf->width) << 16;
rt_format |= util_logbase2(sf->height) << 24;
} else {
rt_format |= NV30_3D_RT_FORMAT_TYPE_LINEAR;
}
if (buffers & PIPE_CLEAR_DEPTH)
mode |= NV30_3D_CLEAR_BUFFERS_DEPTH;
if (buffers & PIPE_CLEAR_STENCIL)
mode |= NV30_3D_CLEAR_BUFFERS_STENCIL;
refn.bo = mt->base.bo;
refn.flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_WR;
if (nouveau_pushbuf_space(push, 32, 1, 0) ||
nouveau_pushbuf_refn (push, &refn, 1))
return;
BEGIN_NV04(push, NV30_3D(RT_ENABLE), 1);
PUSH_DATA (push, 0);
BEGIN_NV04(push, NV30_3D(RT_HORIZ), 3);
PUSH_DATA (push, sf->width << 16);
PUSH_DATA (push, sf->height << 16);
PUSH_DATA (push, rt_format);
if (eng3d->oclass < NV40_3D_CLASS) {
BEGIN_NV04(push, NV30_3D(COLOR0_PITCH), 1);
PUSH_DATA (push, (sf->pitch << 16) | sf->pitch);
} else {
BEGIN_NV04(push, NV40_3D(ZETA_PITCH), 1);
PUSH_DATA (push, sf->pitch);
}
BEGIN_NV04(push, NV30_3D(ZETA_OFFSET), 1);
PUSH_RELOC(push, mt->base.bo, sf->offset, NOUVEAU_BO_LOW, 0, 0);
BEGIN_NV04(push, NV30_3D(SCISSOR_HORIZ), 2);
PUSH_DATA (push, (w << 16) | x);
PUSH_DATA (push, (h << 16) | y);
BEGIN_NV04(push, NV30_3D(CLEAR_DEPTH_VALUE), 1);
PUSH_DATA (push, pack_zeta(ps->format, depth, stencil));
BEGIN_NV04(push, NV30_3D(CLEAR_BUFFERS), 1);
PUSH_DATA (push, mode);
nv30->dirty |= NV30_NEW_FRAMEBUFFER | NV30_NEW_SCISSOR;
}
Bool
NV40EXAPrepareComposite(int op, PicturePtr psPict,
PicturePtr pmPict,
PicturePtr pdPict,
PixmapPtr psPix,
PixmapPtr pmPix,
PixmapPtr pdPix)
{
ScrnInfoPtr pScrn = xf86ScreenToScrn(pdPix->drawable.pScreen);
NVPtr pNv = NVPTR(pScrn);
nv_pict_op_t *blend = NV40_GetPictOpRec(op);
struct nouveau_pushbuf *push = pNv->pushbuf;
uint32_t fragprog;
if (!PUSH_SPACE(push, 128))
NOUVEAU_FALLBACK("space\n");
PUSH_RESET(push);
NV40_SetupBlend(pScrn, blend, pdPict->format,
(pmPict && pmPict->componentAlpha &&
PICT_FORMAT_RGB(pmPict->format)));
if (!NV40_SetupSurface(pScrn, pdPix, pdPict->format) ||
!NV40EXAPicture(pNv, psPix, psPict, 0))
return FALSE;
if (pmPict) {
if (!NV40EXAPicture(pNv, pmPix, pmPict, 1))
return FALSE;
if (pdPict->format == PICT_a8) {
fragprog = PFP_C_A8;
} else
if (pmPict->componentAlpha && PICT_FORMAT_RGB(pmPict->format)) {
if (blend->src_alpha)
fragprog = PFP_CCASA;
else
fragprog = PFP_CCA;
} else {
fragprog = PFP_C;
}
} else {
if (pdPict->format == PICT_a8)
fragprog = PFP_S_A8;
else
fragprog = PFP_S;
}
BEGIN_NV04(push, NV30_3D(FP_ACTIVE_PROGRAM), 1);
PUSH_MTHD (push, NV30_3D(FP_ACTIVE_PROGRAM), pNv->scratch, fragprog,
NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW |
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, 0x02000000);
/* Appears to be some kind of cache flush, needed here at least
* sometimes.. funky text rendering otherwise :)
*/
BEGIN_NV04(push, NV40_3D(TEX_CACHE_CTL), 1);
PUSH_DATA (push, 2);
BEGIN_NV04(push, NV40_3D(TEX_CACHE_CTL), 1);
PUSH_DATA (push, 1);
nouveau_pushbuf_bufctx(push, pNv->bufctx);
if (nouveau_pushbuf_validate(push)) {
nouveau_pushbuf_bufctx(push, NULL);
return FALSE;
}
return TRUE;
}
static Bool
NV40EXAPictTexture(NVPtr pNv, PixmapPtr pPix, PicturePtr pPict, int unit)
{
unsigned reloc = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_WR;
struct nouveau_pushbuf *push = pNv->pushbuf;
struct nouveau_bo *bo = nouveau_pixmap_bo(pPix);
nv_pict_texture_format_t *fmt;
fmt = NV40_GetPictTextureFormat(pPict->format);
if (!fmt)
return FALSE;
BEGIN_NV04(push, NV30_3D(TEX_OFFSET(unit)), 8);
PUSH_MTHDl(push, NV30_3D(TEX_OFFSET(unit)), bo, 0, reloc);
PUSH_MTHDs(push, NV30_3D(TEX_FORMAT(unit)), bo, fmt->card_fmt |
NV40_3D_TEX_FORMAT_LINEAR |
NV30_3D_TEX_FORMAT_DIMS_2D | 0x8000 |
NV30_3D_TEX_FORMAT_NO_BORDER |
(1 << NV40_3D_TEX_FORMAT_MIPMAP_COUNT__SHIFT),
reloc | NOUVEAU_BO_OR,
NV30_3D_TEX_FORMAT_DMA0, NV30_3D_TEX_FORMAT_DMA1);
if (pPict->repeat) {
switch(pPict->repeatType) {
case RepeatPad:
PUSH_DATA (push, NV30_3D_TEX_WRAP_S_CLAMP |
NV30_3D_TEX_WRAP_T_CLAMP |
NV30_3D_TEX_WRAP_R_CLAMP);
break;
case RepeatReflect:
PUSH_DATA (push, NV30_3D_TEX_WRAP_S_MIRRORED_REPEAT |
NV30_3D_TEX_WRAP_T_MIRRORED_REPEAT |
NV30_3D_TEX_WRAP_R_MIRRORED_REPEAT);
break;
case RepeatNormal:
default:
PUSH_DATA (push, NV30_3D_TEX_WRAP_S_REPEAT |
NV30_3D_TEX_WRAP_T_REPEAT |
NV30_3D_TEX_WRAP_R_REPEAT);
break;
}
} else {
PUSH_DATA (push, NV30_3D_TEX_WRAP_S_CLAMP_TO_BORDER |
NV30_3D_TEX_WRAP_T_CLAMP_TO_BORDER |
NV30_3D_TEX_WRAP_R_CLAMP_TO_BORDER);
}
PUSH_DATA (push, NV40_3D_TEX_ENABLE_ENABLE);
PUSH_DATA (push, fmt->card_swz);
if (pPict->filter == PictFilterBilinear) {
PUSH_DATA (push, NV30_3D_TEX_FILTER_MIN_LINEAR |
NV30_3D_TEX_FILTER_MAG_LINEAR | 0x3fd6);
} else {
PUSH_DATA (push, NV30_3D_TEX_FILTER_MIN_NEAREST |
NV30_3D_TEX_FILTER_MAG_NEAREST | 0x3fd6);
}
PUSH_DATA (push, (pPix->drawable.width << 16) | pPix->drawable.height);
PUSH_DATA (push, 0); /* border ARGB */
BEGIN_NV04(push, NV40_3D(TEX_SIZE1(unit)), 1);
PUSH_DATA (push, (1 << NV40_3D_TEX_SIZE1_DEPTH__SHIFT) |
(uint32_t)exaGetPixmapPitch(pPix));
BEGIN_NV04(push, NV30_3D(VP_UPLOAD_CONST_ID), 17);
PUSH_DATA (push, unit * 4);
if (pPict->transform) {
PUSH_DATAf(push, xFixedToFloat(pPict->transform->matrix[0][0]));
PUSH_DATAf(push, xFixedToFloat(pPict->transform->matrix[0][1]));
PUSH_DATAf(push, xFixedToFloat(pPict->transform->matrix[0][2]));
PUSH_DATAf(push, 0);
PUSH_DATAf(push, xFixedToFloat(pPict->transform->matrix[1][0]));
PUSH_DATAf(push, xFixedToFloat(pPict->transform->matrix[1][1]));
PUSH_DATAf(push, xFixedToFloat(pPict->transform->matrix[1][2]));
PUSH_DATAf(push, 0);
PUSH_DATAf(push, xFixedToFloat(pPict->transform->matrix[2][0]));
PUSH_DATAf(push, xFixedToFloat(pPict->transform->matrix[2][1]));
PUSH_DATAf(push, xFixedToFloat(pPict->transform->matrix[2][2]));
PUSH_DATAf(push, 0);
} else {
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 0.0);
}
PUSH_DATAf(push, 1.0 / pPix->drawable.width);
PUSH_DATAf(push, 1.0 / pPix->drawable.height);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 1.0);
return TRUE;
}
void
nv30_push_vbo(struct nv30_context *nv30, const struct pipe_draw_info *info)
{
struct push_context ctx;
unsigned i, index_size;
boolean apply_bias = info->indexed && info->index_bias;
ctx.push = nv30->base.pushbuf;
ctx.translate = nv30->vertex->translate;
ctx.packet_vertex_limit = nv30->vertex->vtx_per_packet_max;
ctx.vertex_words = nv30->vertex->vtx_size;
for (i = 0; i < nv30->num_vtxbufs; ++i) {
uint8_t *data;
struct pipe_vertex_buffer *vb = &nv30->vtxbuf[i];
struct nv04_resource *res = nv04_resource(vb->buffer);
if (!vb->buffer && !vb->user_buffer) {
continue;
}
data = nouveau_resource_map_offset(&nv30->base, res,
vb->buffer_offset, NOUVEAU_BO_RD);
if (apply_bias)
data += info->index_bias * vb->stride;
ctx.translate->set_buffer(ctx.translate, i, data, vb->stride, ~0);
}
if (info->indexed) {
if (nv30->idxbuf.buffer)
ctx.idxbuf = nouveau_resource_map_offset(&nv30->base,
nv04_resource(nv30->idxbuf.buffer), nv30->idxbuf.offset,
NOUVEAU_BO_RD);
else
ctx.idxbuf = nv30->idxbuf.user_buffer;
if (!ctx.idxbuf) {
nv30_state_release(nv30);
return;
}
index_size = nv30->idxbuf.index_size;
ctx.primitive_restart = info->primitive_restart;
ctx.restart_index = info->restart_index;
} else {
ctx.idxbuf = NULL;
index_size = 0;
ctx.primitive_restart = FALSE;
ctx.restart_index = 0;
}
if (nv30->screen->eng3d->oclass >= NV40_3D_CLASS) {
BEGIN_NV04(ctx.push, NV40_3D(PRIM_RESTART_ENABLE), 2);
PUSH_DATA (ctx.push, info->primitive_restart);
PUSH_DATA (ctx.push, info->restart_index);
nv30->state.prim_restart = info->primitive_restart;
}
ctx.prim = nv30_prim_gl(info->mode);
PUSH_RESET(ctx.push, BUFCTX_IDXBUF);
BEGIN_NV04(ctx.push, NV30_3D(VERTEX_BEGIN_END), 1);
PUSH_DATA (ctx.push, ctx.prim);
switch (index_size) {
case 0:
emit_vertices_seq(&ctx, info->start, info->count);
break;
case 1:
emit_vertices_i08(&ctx, info->start, info->count);
break;
case 2:
emit_vertices_i16(&ctx, info->start, info->count);
break;
case 4:
emit_vertices_i32(&ctx, info->start, info->count);
break;
default:
assert(0);
break;
}
BEGIN_NV04(ctx.push, NV30_3D(VERTEX_BEGIN_END), 1);
PUSH_DATA (ctx.push, NV30_3D_VERTEX_BEGIN_END_STOP);
if (info->indexed)
nouveau_resource_unmap(nv04_resource(nv30->idxbuf.buffer));
for (i = 0; i < nv30->num_vtxbufs; ++i) {
if (nv30->vtxbuf[i].buffer) {
nouveau_resource_unmap(nv04_resource(nv30->vtxbuf[i].buffer));
}
}
nv30_state_release(nv30);
}
static boolean
nv30_render_validate(struct nv30_context *nv30)
{
struct nv30_render *r = nv30_render(nv30->draw->render);
struct nv30_rasterizer_stateobj *rast = nv30->rast;
struct pipe_screen *pscreen = &nv30->screen->base.base;
struct nouveau_pushbuf *push = nv30->screen->base.pushbuf;
struct nouveau_object *eng3d = nv30->screen->eng3d;
struct nv30_vertprog *vp = nv30->vertprog.program;
struct vertex_info *vinfo = &r->vertex_info;
unsigned vp_attribs = 0;
unsigned vp_results = 0;
unsigned attrib = 0;
unsigned pntc;
int i;
if (!r->vertprog) {
struct nouveau_heap *heap = nv30_screen(pscreen)->vp_exec_heap;
if (nouveau_heap_alloc(heap, 16, &r->vertprog, &r->vertprog)) {
while (heap->next && heap->size < 16) {
struct nouveau_heap **evict = heap->next->priv;
nouveau_heap_free(evict);
}
if (nouveau_heap_alloc(heap, 16, &r->vertprog, &r->vertprog))
return FALSE;
}
}
vinfo->num_attribs = 0;
vinfo->size = 0;
/* setup routing for all necessary vp outputs */
for (i = 0; i < vp->info.num_outputs && attrib < 16; i++) {
uint semantic = vp->info.output_semantic_name[i];
uint index = vp->info.output_semantic_index[i];
if (vroute_add(r, attrib, semantic, &index)) {
vp_attribs |= (1 << attrib++);
vp_results |= index;
}
}
/* setup routing for replaced point coords not written by vp */
if (rast && rast->pipe.point_quad_rasterization)
pntc = rast->pipe.sprite_coord_enable & 0x000002ff;
else
pntc = 0;
while (pntc && attrib < 16) {
uint index = ffs(pntc) - 1; pntc &= ~(1 << index);
if (vroute_add(r, attrib, TGSI_SEMANTIC_TEXCOORD, &index)) {
vp_attribs |= (1 << attrib++);
vp_results |= index;
}
}
/* modify vertex format for correct stride, and stub out unused ones */
BEGIN_NV04(push, NV30_3D(VP_UPLOAD_FROM_ID), 1);
PUSH_DATA (push, r->vertprog->start);
r->vtxprog[attrib - 1][3] |= 1;
for (i = 0; i < attrib; i++) {
BEGIN_NV04(push, NV30_3D(VP_UPLOAD_INST(0)), 4);
PUSH_DATAp(push, r->vtxprog[i], 4);
r->vtxfmt[i] |= vinfo->size << 8;
}
for (; i < 16; i++)
r->vtxfmt[i] = NV30_3D_VTXFMT_TYPE_V32_FLOAT;
BEGIN_NV04(push, NV30_3D(VIEWPORT_TRANSLATE_X), 8);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 1.0);
BEGIN_NV04(push, NV30_3D(DEPTH_RANGE_NEAR), 2);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 1.0);
BEGIN_NV04(push, NV30_3D(VTXFMT(0)), 16);
PUSH_DATAp(push, r->vtxfmt, 16);
BEGIN_NV04(push, NV30_3D(VP_START_FROM_ID), 1);
PUSH_DATA (push, r->vertprog->start);
BEGIN_NV04(push, NV30_3D(ENGINE), 1);
PUSH_DATA (push, 0x00000103);
if (eng3d->oclass >= NV40_3D_CLASS) {
BEGIN_NV04(push, NV40_3D(VP_ATTRIB_EN), 2);
PUSH_DATA (push, vp_attribs);
PUSH_DATA (push, vp_results);
}
vinfo->size /= 4;
return TRUE;
}
void
nv30_fragtex_validate(struct nv30_context *nv30)
{
struct pipe_screen *pscreen = &nv30->screen->base.base;
struct nouveau_object *eng3d = nv30->screen->eng3d;
struct nouveau_pushbuf *push = nv30->base.pushbuf;
unsigned dirty = nv30->fragprog.dirty_samplers;
while (dirty) {
unsigned unit = ffs(dirty) - 1;
struct nv30_sampler_view *sv = (void *)nv30->fragprog.textures[unit];
struct nv30_sampler_state *ss = nv30->fragprog.samplers[unit];
PUSH_RESET(push, BUFCTX_FRAGTEX(unit));
if (ss && sv) {
const struct nv30_texfmt *fmt = nv30_texfmt(pscreen, sv->pipe.format);
struct pipe_resource *pt = sv->pipe.texture;
struct nv30_miptree *mt = nv30_miptree(pt);
unsigned min_lod, max_lod;
u32 filter = sv->filt | (ss->filt & sv->filt_mask);
u32 format = sv->fmt | ss->fmt;
u32 enable = ss->en;
/* handle base_level when not using a mip filter, min/max level
* is unfortunately ignored by the hardware otherwise
*/
if (ss->pipe.min_mip_filter == PIPE_TEX_MIPFILTER_NONE) {
if (sv->base_lod)
filter += 0x00020000; /* N/L -> NMN/LMN */
max_lod = sv->base_lod;
min_lod = sv->base_lod;
} else {
max_lod = MIN2(ss->max_lod + sv->base_lod, sv->high_lod);
min_lod = MIN2(ss->min_lod + sv->base_lod, max_lod);
}
if (eng3d->oclass >= NV40_3D_CLASS) {
/* this is a tad stupid of the hardware, but there's no non-rcomp
* z16/z24 texture formats to be had, we have to suffer and lose
* some precision to handle this case.
*/
if (ss->pipe.compare_mode != PIPE_TEX_COMPARE_R_TO_TEXTURE) {
if (fmt->nv40 == NV40_3D_TEX_FORMAT_FORMAT_Z16)
format |= NV40_3D_TEX_FORMAT_FORMAT_A8L8;
else
if (fmt->nv40 == NV40_3D_TEX_FORMAT_FORMAT_Z24)
format |= NV40_3D_TEX_FORMAT_FORMAT_A16L16;
else
format |= fmt->nv40;
} else {
format |= fmt->nv40;
}
enable |= (min_lod << 19) | (max_lod << 7);
enable |= NV40_3D_TEX_ENABLE_ENABLE;
BEGIN_NV04(push, NV40_3D(TEX_SIZE1(unit)), 1);
PUSH_DATA (push, sv->npot_size1);
} else {
/* this is a tad stupid of the hardware, but there's no non-rcomp
* z16/z24 texture formats to be had, we have to suffer and lose
* some precision to handle this case.
*/
if (ss->pipe.compare_mode != PIPE_TEX_COMPARE_R_TO_TEXTURE) {
if (fmt->nv30 == NV30_3D_TEX_FORMAT_FORMAT_Z16) {
if (ss->pipe.normalized_coords)
format |= NV30_3D_TEX_FORMAT_FORMAT_A8L8;
else
format |= NV30_3D_TEX_FORMAT_FORMAT_A8L8_RECT;
} else
if (fmt->nv30 == NV30_3D_TEX_FORMAT_FORMAT_Z24) {
if (ss->pipe.normalized_coords)
format |= NV30_3D_TEX_FORMAT_FORMAT_HILO16;
else
format |= NV30_3D_TEX_FORMAT_FORMAT_HILO16_RECT;
} else {
if (ss->pipe.normalized_coords)
format |= fmt->nv30;
else
format |= fmt->nv30_rect;
}
} else {
if (ss->pipe.normalized_coords)
format |= fmt->nv30;
else
format |= fmt->nv30_rect;
}
enable |= NV30_3D_TEX_ENABLE_ENABLE;
enable |= (min_lod << 18) | (max_lod << 6);
}
BEGIN_NV04(push, NV30_3D(TEX_OFFSET(unit)), 8);
PUSH_MTHDl(push, NV30_3D(TEX_OFFSET(unit)), BUFCTX_FRAGTEX(unit),
mt->base.bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
PUSH_MTHDs(push, NV30_3D(TEX_FORMAT(unit)), BUFCTX_FRAGTEX(unit),
mt->base.bo, format, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD,
NV30_3D_TEX_FORMAT_DMA0,
NV30_3D_TEX_FORMAT_DMA1);
PUSH_DATA (push, sv->wrap | (ss->wrap & sv->wrap_mask));
PUSH_DATA (push, enable);
PUSH_DATA (push, sv->swz);
PUSH_DATA (push, filter);
PUSH_DATA (push, sv->npot_size0);
PUSH_DATA (push, ss->bcol);
BEGIN_NV04(push, NV30_3D(TEX_FILTER_OPTIMIZATION(unit)), 1);
PUSH_DATA (push, nv30->config.filter);
} else {
BEGIN_NV04(push, NV30_3D(TEX_ENABLE(unit)), 1);
PUSH_DATA (push, 0);
}
dirty &= ~(1 << unit);
}
nv30->fragprog.dirty_samplers = 0;
}
static void
nv30_transfer_rect_blit(XFER_ARGS)
{
struct nv04_resource *fp = nv30_transfer_rect_fragprog(nv30);
struct nouveau_heap *vp = nv30_transfer_rect_vertprog(nv30);
struct nouveau_pushbuf *push = nv30->base.pushbuf;
struct nouveau_pushbuf_refn refs[] = {
{ fp->bo, fp->domain | NOUVEAU_BO_RD },
{ src->bo, src->domain | NOUVEAU_BO_RD },
{ dst->bo, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR },
};
u32 texfmt, texswz;
u32 format, stride;
if (nouveau_pushbuf_space(push, 512, 8, 0) ||
nouveau_pushbuf_refn (push, refs, sizeof(refs) / sizeof(refs[0])))
return;
/* various switches depending on cpp of the transfer */
switch (dst->cpp) {
case 4:
format = NV30_3D_RT_FORMAT_COLOR_A8R8G8B8 |
NV30_3D_RT_FORMAT_ZETA_Z24S8;
texfmt = NV40_3D_TEX_FORMAT_FORMAT_A8R8G8B8;
texswz = 0x0000aae4;
break;
case 2:
format = NV30_3D_RT_FORMAT_COLOR_R5G6B5 |
NV30_3D_RT_FORMAT_ZETA_Z16;
texfmt = NV40_3D_TEX_FORMAT_FORMAT_R5G6B5;
texswz = 0x0000a9e4;
break;
case 1:
format = NV30_3D_RT_FORMAT_COLOR_B8 |
NV30_3D_RT_FORMAT_ZETA_Z16;
texfmt = NV40_3D_TEX_FORMAT_FORMAT_L8;
texswz = 0x0000aaff;
break;
default:
assert(0);
return;
}
/* render target */
if (!dst->pitch) {
format |= NV30_3D_RT_FORMAT_TYPE_SWIZZLED;
format |= util_logbase2(dst->w) << 16;
format |= util_logbase2(dst->h) << 24;
stride = 64;
} else {
format |= NV30_3D_RT_FORMAT_TYPE_LINEAR;
stride = dst->pitch;
}
BEGIN_NV04(push, NV30_3D(VIEWPORT_HORIZ), 2);
PUSH_DATA (push, dst->w << 16);
PUSH_DATA (push, dst->h << 16);
BEGIN_NV04(push, NV30_3D(RT_HORIZ), 5);
PUSH_DATA (push, dst->w << 16);
PUSH_DATA (push, dst->h << 16);
PUSH_DATA (push, format);
PUSH_DATA (push, stride);
PUSH_RELOC(push, dst->bo, dst->offset, NOUVEAU_BO_LOW, 0, 0);
BEGIN_NV04(push, NV30_3D(RT_ENABLE), 1);
PUSH_DATA (push, NV30_3D_RT_ENABLE_COLOR0);
nv30->dirty |= NV30_NEW_FRAMEBUFFER;
/* viewport state */
BEGIN_NV04(push, NV30_3D(VIEWPORT_TRANSLATE_X), 8);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 1.0);
PUSH_DATAf(push, 1.0);
BEGIN_NV04(push, NV30_3D(DEPTH_RANGE_NEAR), 2);
PUSH_DATAf(push, 0.0);
PUSH_DATAf(push, 1.0);
nv30->dirty |= NV30_NEW_VIEWPORT;
/* blend state */
BEGIN_NV04(push, NV30_3D(COLOR_LOGIC_OP_ENABLE), 1);
PUSH_DATA (push, 0);
BEGIN_NV04(push, NV30_3D(DITHER_ENABLE), 1);
PUSH_DATA (push, 0);
BEGIN_NV04(push, NV30_3D(BLEND_FUNC_ENABLE), 1);
PUSH_DATA (push, 0);
BEGIN_NV04(push, NV30_3D(COLOR_MASK), 1);
PUSH_DATA (push, 0x01010101);
nv30->dirty |= NV30_NEW_BLEND;
/* depth-stencil-alpha state */
BEGIN_NV04(push, NV30_3D(DEPTH_WRITE_ENABLE), 2);
PUSH_DATA (push, 0);
PUSH_DATA (push, 0);
BEGIN_NV04(push, NV30_3D(STENCIL_ENABLE(0)), 1);
PUSH_DATA (push, 0);
BEGIN_NV04(push, NV30_3D(STENCIL_ENABLE(1)), 1);
PUSH_DATA (push, 0);
BEGIN_NV04(push, NV30_3D(ALPHA_FUNC_ENABLE), 1);
PUSH_DATA (push, 0);
nv30->dirty |= NV30_NEW_ZSA;
/* rasterizer state */
BEGIN_NV04(push, NV30_3D(SHADE_MODEL), 1);
PUSH_DATA (push, NV30_3D_SHADE_MODEL_FLAT);
BEGIN_NV04(push, NV30_3D(CULL_FACE_ENABLE), 1);
PUSH_DATA (push, 0);
BEGIN_NV04(push, NV30_3D(POLYGON_MODE_FRONT), 2);
PUSH_DATA (push, NV30_3D_POLYGON_MODE_FRONT_FILL);
PUSH_DATA (push, NV30_3D_POLYGON_MODE_BACK_FILL);
BEGIN_NV04(push, NV30_3D(POLYGON_OFFSET_FILL_ENABLE), 1);
PUSH_DATA (push, 0);
BEGIN_NV04(push, NV30_3D(POLYGON_STIPPLE_ENABLE), 1);
PUSH_DATA (push, 0);
nv30->state.scissor_off = 0;
nv30->dirty |= NV30_NEW_RASTERIZER;
/* vertex program */
BEGIN_NV04(push, NV30_3D(VP_START_FROM_ID), 1);
PUSH_DATA (push, vp->start);
BEGIN_NV04(push, NV40_3D(VP_ATTRIB_EN), 2);
PUSH_DATA (push, 0x00000101); /* attrib: 0, 8 */
PUSH_DATA (push, 0x00004000); /* result: hpos, tex0 */
BEGIN_NV04(push, NV30_3D(ENGINE), 1);
PUSH_DATA (push, 0x00000103);
BEGIN_NV04(push, NV30_3D(VP_CLIP_PLANES_ENABLE), 1);
PUSH_DATA (push, 0x00000000);
nv30->dirty |= NV30_NEW_VERTPROG;
nv30->dirty |= NV30_NEW_CLIP;
/* fragment program */
BEGIN_NV04(push, NV30_3D(FP_ACTIVE_PROGRAM), 1);
PUSH_RELOC(push, fp->bo, fp->offset, fp->domain |
NOUVEAU_BO_LOW | 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, 0x02000000);
nv30->state.fragprog = NULL;
nv30->dirty |= NV30_NEW_FRAGPROG;
/* texture */
texfmt |= 1 << NV40_3D_TEX_FORMAT_MIPMAP_COUNT__SHIFT;
texfmt |= NV30_3D_TEX_FORMAT_NO_BORDER;
texfmt |= NV40_3D_TEX_FORMAT_RECT;
texfmt |= 0x00008000;
if (src->d < 2)
texfmt |= NV30_3D_TEX_FORMAT_DIMS_2D;
else
texfmt |= NV30_3D_TEX_FORMAT_DIMS_3D;
if (src->pitch)
texfmt |= NV40_3D_TEX_FORMAT_LINEAR;
BEGIN_NV04(push, NV30_3D(TEX_OFFSET(0)), 8);
PUSH_RELOC(push, src->bo, src->offset, NOUVEAU_BO_LOW, 0, 0);
PUSH_RELOC(push, src->bo, texfmt, NOUVEAU_BO_OR,
NV30_3D_TEX_FORMAT_DMA0, NV30_3D_TEX_FORMAT_DMA1);
PUSH_DATA (push, NV30_3D_TEX_WRAP_S_CLAMP_TO_EDGE |
NV30_3D_TEX_WRAP_T_CLAMP_TO_EDGE |
NV30_3D_TEX_WRAP_R_CLAMP_TO_EDGE);
PUSH_DATA (push, NV40_3D_TEX_ENABLE_ENABLE);
PUSH_DATA (push, texswz);
switch (filter) {
case BILINEAR:
PUSH_DATA (push, NV30_3D_TEX_FILTER_MIN_LINEAR |
NV30_3D_TEX_FILTER_MAG_LINEAR | 0x00002000);
break;
default:
PUSH_DATA (push, NV30_3D_TEX_FILTER_MIN_NEAREST |
NV30_3D_TEX_FILTER_MAG_NEAREST | 0x00002000);
break;
}
PUSH_DATA (push, (src->w << 16) | src->h);
PUSH_DATA (push, 0x00000000);
BEGIN_NV04(push, NV40_3D(TEX_SIZE1(0)), 1);
PUSH_DATA (push, 0x00100000 | src->pitch);
BEGIN_NV04(push, SUBC_3D(0x0b40), 1);
PUSH_DATA (push, src->d < 2 ? 0x00000001 : 0x00000000);
BEGIN_NV04(push, NV40_3D(TEX_CACHE_CTL), 1);
PUSH_DATA (push, 1);
nv30->fragprog.dirty_samplers |= 1;
nv30->dirty |= NV30_NEW_FRAGTEX;
/* blit! */
BEGIN_NV04(push, NV30_3D(SCISSOR_HORIZ), 2);
PUSH_DATA (push, (dst->x1 - dst->x0) << 16 | dst->x0);
PUSH_DATA (push, (dst->y1 - dst->y0) << 16 | dst->y0);
BEGIN_NV04(push, NV30_3D(VERTEX_BEGIN_END), 1);
PUSH_DATA (push, NV30_3D_VERTEX_BEGIN_END_QUADS);
BEGIN_NV04(push, NV30_3D(VTX_ATTR_3F(8)), 3);
PUSH_DATAf(push, src->x0);
PUSH_DATAf(push, src->y0);
PUSH_DATAf(push, src->z);
BEGIN_NV04(push, NV30_3D(VTX_ATTR_2I(0)), 1);
PUSH_DATA (push, (dst->y0 << 16) | dst->x0);
BEGIN_NV04(push, NV30_3D(VTX_ATTR_3F(8)), 3);
PUSH_DATAf(push, src->x1);
PUSH_DATAf(push, src->y0);
PUSH_DATAf(push, src->z);
BEGIN_NV04(push, NV30_3D(VTX_ATTR_2I(0)), 1);
PUSH_DATA (push, (dst->y0 << 16) | dst->x1);
BEGIN_NV04(push, NV30_3D(VTX_ATTR_3F(8)), 3);
PUSH_DATAf(push, src->x1);
PUSH_DATAf(push, src->y1);
PUSH_DATAf(push, src->z);
BEGIN_NV04(push, NV30_3D(VTX_ATTR_2I(0)), 1);
PUSH_DATA (push, (dst->y1 << 16) | dst->x1);
BEGIN_NV04(push, NV30_3D(VTX_ATTR_3F(8)), 3);
PUSH_DATAf(push, src->x0);
PUSH_DATAf(push, src->y1);
PUSH_DATAf(push, src->z);
BEGIN_NV04(push, NV30_3D(VTX_ATTR_2I(0)), 1);
PUSH_DATA (push, (dst->y1 << 16) | dst->x0);
BEGIN_NV04(push, NV30_3D(VERTEX_BEGIN_END), 1);
PUSH_DATA (push, NV30_3D_VERTEX_BEGIN_END_STOP);
}
void
nv30_vertprog_validate(struct nv30_context *nv30)
{
struct nouveau_pushbuf *push = nv30->base.pushbuf;
struct nouveau_object *eng3d = nv30->screen->eng3d;
struct nv30_vertprog *vp = nv30->vertprog.program;
struct nv30_fragprog *fp = nv30->fragprog.program;
boolean upload_code = FALSE;
boolean upload_data = FALSE;
unsigned i;
if (nv30->dirty & NV30_NEW_FRAGPROG) {
if (memcmp(vp->texcoord, fp->texcoord, sizeof(vp->texcoord))) {
if (vp->translated)
nv30_vertprog_destroy(vp);
memcpy(vp->texcoord, fp->texcoord, sizeof(vp->texcoord));
}
}
if (nv30->rast && nv30->rast->pipe.clip_plane_enable != vp->enabled_ucps) {
vp->enabled_ucps = nv30->rast->pipe.clip_plane_enable;
if (vp->translated)
nv30_vertprog_destroy(vp);
}
if (!vp->translated) {
vp->translated = _nvfx_vertprog_translate(eng3d->oclass, vp);
if (!vp->translated) {
nv30->draw_flags |= NV30_NEW_VERTPROG;
return;
}
nv30->dirty |= NV30_NEW_VERTPROG;
}
if (!vp->exec) {
struct nouveau_heap *heap = nv30->screen->vp_exec_heap;
struct nv30_shader_reloc *reloc = vp->branch_relocs.data;
unsigned nr_reloc = vp->branch_relocs.size / sizeof(*reloc);
uint32_t *inst, target;
if (nouveau_heap_alloc(heap, vp->nr_insns, &vp->exec, &vp->exec)) {
while (heap->next && heap->size < vp->nr_insns) {
struct nouveau_heap **evict = heap->next->priv;
nouveau_heap_free(evict);
}
if (nouveau_heap_alloc(heap, vp->nr_insns, &vp->exec, &vp->exec)) {
nv30->draw_flags |= NV30_NEW_VERTPROG;
return;
}
}
if (eng3d->oclass < NV40_3D_CLASS) {
while (nr_reloc--) {
inst = vp->insns[reloc->location].data;
target = vp->exec->start + reloc->target;
inst[2] &= ~0x000007fc;
inst[2] |= target << 2;
reloc++;
}
} else {
while (nr_reloc--) {
inst = vp->insns[reloc->location].data;
target = vp->exec->start + reloc->target;
inst[2] &= ~0x0000003f;
inst[2] |= target >> 3;
inst[3] &= ~0xe0000000;
inst[3] |= target << 29;
reloc++;
}
}
upload_code = TRUE;
}
if (vp->nr_consts && !vp->data) {
struct nouveau_heap *heap = nv30->screen->vp_data_heap;
struct nv30_shader_reloc *reloc = vp->const_relocs.data;
unsigned nr_reloc = vp->const_relocs.size / sizeof(*reloc);
uint32_t *inst, target;
if (nouveau_heap_alloc(heap, vp->nr_consts, vp, &vp->data)) {
while (heap->next && heap->size < vp->nr_consts) {
struct nv30_vertprog *evp = heap->next->priv;
nouveau_heap_free(&evp->data);
}
if (nouveau_heap_alloc(heap, vp->nr_consts, vp, &vp->data)) {
nv30->draw_flags |= NV30_NEW_VERTPROG;
return;
}
}
if (eng3d->oclass < NV40_3D_CLASS) {
while (nr_reloc--) {
inst = vp->insns[reloc->location].data;
target = vp->data->start + reloc->target;
inst[1] &= ~0x0007fc000;
inst[1] |= (target & 0x1ff) << 14;
reloc++;
}
} else {
while (nr_reloc--) {
inst = vp->insns[reloc->location].data;
target = vp->data->start + reloc->target;
inst[1] &= ~0x0001ff000;
inst[1] |= (target & 0x1ff) << 12;
reloc++;
}
}
upload_code = TRUE;
upload_data = TRUE;
}
if (vp->nr_consts) {
struct nv04_resource *res = nv04_resource(nv30->vertprog.constbuf);
for (i = 0; i < vp->nr_consts; i++) {
struct nv30_vertprog_data *data = &vp->consts[i];
if (data->index < 0) {
if (!upload_data)
continue;
} else {
float *constbuf = (float *)res->data;
if (!upload_data &&
!memcmp(data->value, &constbuf[data->index * 4], 16))
continue;
memcpy(data->value, &constbuf[data->index * 4], 16);
}
BEGIN_NV04(push, NV30_3D(VP_UPLOAD_CONST_ID), 5);
PUSH_DATA (push, vp->data->start + i);
PUSH_DATAp(push, data->value, 4);
}
}
if (upload_code) {
BEGIN_NV04(push, NV30_3D(VP_UPLOAD_FROM_ID), 1);
PUSH_DATA (push, vp->exec->start);
for (i = 0; i < vp->nr_insns; i++) {
BEGIN_NV04(push, NV30_3D(VP_UPLOAD_INST(0)), 4);
PUSH_DATAp(push, vp->insns[i].data, 4);
}
}
if (nv30->dirty & (NV30_NEW_VERTPROG | NV30_NEW_FRAGPROG)) {
BEGIN_NV04(push, NV30_3D(VP_START_FROM_ID), 1);
PUSH_DATA (push, vp->exec->start);
if (eng3d->oclass < NV40_3D_CLASS) {
BEGIN_NV04(push, NV30_3D(ENGINE), 1);
PUSH_DATA (push, 0x00000013); /* vp instead of ff, somehow */
} else {
BEGIN_NV04(push, NV40_3D(VP_ATTRIB_EN), 2);
PUSH_DATA (push, vp->ir);
PUSH_DATA (push, vp->or | fp->vp_or);
BEGIN_NV04(push, NV30_3D(ENGINE), 1);
PUSH_DATA (push, 0x00000011);
}
}
}
