#include "util/u_inlines.h"
#include "nv50_context.h"
-#include "nv50_transfer.h"
-static void
-nv50_transfer_constbuf(struct nv50_context *nv50,
- struct pipe_resource *buf, unsigned size, unsigned cbi)
+void
+nv50_constbufs_validate(struct nv50_context *nv50)
{
- struct pipe_context *pipe = &nv50->pipe;
- struct pipe_transfer *transfer;
struct nouveau_channel *chan = nv50->screen->base.channel;
- struct nouveau_grobj *tesla = nv50->screen->tesla;
- uint32_t *map;
- unsigned count, start;
+ unsigned s;
- map = pipe_buffer_map(pipe, buf, PIPE_TRANSFER_READ, &transfer);
- if (!map)
- return;
+ for (s = 0; s < 3; ++s) {
+ struct nv04_resource *res;
+ int i;
+ unsigned p, b;
- count = (buf->width0 + 3) / 4;
- start = 0;
+ if (s == PIPE_SHADER_FRAGMENT)
+ p = NV50_3D_SET_PROGRAM_CB_PROGRAM_FRAGMENT;
+ else
+ if (s == PIPE_SHADER_GEOMETRY)
+ p = NV50_3D_SET_PROGRAM_CB_PROGRAM_GEOMETRY;
+ else
+ p = NV50_3D_SET_PROGRAM_CB_PROGRAM_VERTEX;
+
+ while (nv50->constbuf_dirty[s]) {
+ struct nouveau_bo *bo;
+ unsigned start = 0;
+ unsigned words = 0;
+
+ i = ffs(nv50->constbuf_dirty[s]) - 1;
+ nv50->constbuf_dirty[s] &= ~(1 << i);
+
+ res = nv04_resource(nv50->constbuf[s][i]);
+ if (!res) {
+ if (i != 0) {
+ BEGIN_RING(chan, RING_3D(SET_PROGRAM_CB), 1);
+ OUT_RING (chan, (i << 8) | p | 0);
+ }
+ continue;
+ }
- while (count) {
- unsigned nr = count;
- nr = MIN2(nr, 2047);
+ if (i == 0) {
+ b = NV50_CB_PVP + s;
- /* FIXME: emit relocs for unsuiTed MM */
- BEGIN_RING(chan, tesla, NV50TCL_CB_ADDR, 1);
- OUT_RING (chan, (start << 8) | cbi);
- BEGIN_RING_NI(chan, tesla, NV50TCL_CB_DATA(0), nr);
- OUT_RINGp (chan, map, nr);
+ /* always upload GL uniforms through CB DATA */
+ bo = nv50->screen->uniforms;
+ words = res->base.width0 / 4;
+ } else {
+ b = s * 16 + i;
- count -= nr;
- start += nr;
- map += nr;
- }
+ assert(0);
- pipe_buffer_unmap(pipe, buf, transfer);
-}
+ if (!nouveau_resource_mapped_by_gpu(&res->base)) {
+ nouveau_buffer_migrate(&nv50->base, res, NOUVEAU_BO_VRAM);
-static void
-nv50_program_validate_data(struct nv50_context *nv50, struct nv50_program *p)
-{
- struct nouveau_channel *chan = nv50->screen->base.channel;
- struct nouveau_grobj *tesla = nv50->screen->tesla;
- unsigned cbi;
-
- if (p->immd_size) {
- uint32_t *data = p->immd;
- unsigned count = p->immd_size / 4;
- unsigned start = 0;
-
- while (count) {
- unsigned nr = count;
- nr = MIN2(nr, 2047);
-
- BEGIN_RING(chan, tesla, NV50TCL_CB_ADDR, 1);
- OUT_RING (chan, (start << 8) | NV50_CB_PMISC);
- BEGIN_RING_NI(chan, tesla, NV50TCL_CB_DATA(0), nr);
- OUT_RINGp (chan, data, nr);
-
- count -= nr;
- start += nr;
- data += nr;
- }
- }
+ BEGIN_RING(chan, RING_3D(CODE_CB_FLUSH), 1);
+ OUT_RING (chan, 0);
+ }
+ MARK_RING (chan, 6, 2);
+ BEGIN_RING(chan, RING_3D(CB_DEF_ADDRESS_HIGH), 3);
+ OUT_RESRCh(chan, res, 0, NOUVEAU_BO_RD);
+ OUT_RESRCl(chan, res, 0, NOUVEAU_BO_RD);
+ OUT_RING (chan, (b << 16) | (res->base.width0 & 0xffff));
+ BEGIN_RING(chan, RING_3D(SET_PROGRAM_CB), 1);
+ OUT_RING (chan, (b << 12) | (i << 8) | p | 1);
- /* If the state tracker doesn't change the constbuf, and it is first
- * validated with a program that doesn't use it, this check prevents
- * it from even being uploaded. */
- /*
- if (p->parm_size == 0)
- return;
- */
-
- switch (p->type) {
- case PIPE_SHADER_VERTEX:
- cbi = NV50_CB_PVP;
- break;
- case PIPE_SHADER_FRAGMENT:
- cbi = NV50_CB_PFP;
- break;
- case PIPE_SHADER_GEOMETRY:
- cbi = NV50_CB_PGP;
- break;
- default:
- assert(0);
- break;
- }
+ bo = res->bo;
- nv50_transfer_constbuf(nv50, nv50->constbuf[p->type], p->parm_size, cbi);
-}
+ nv50_bufctx_add_resident(nv50, NV50_BUFCTX_CONSTANT, res,
+ res->domain | NOUVEAU_BO_RD);
+ }
-static void
-nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p)
-{
- struct nouveau_channel *chan = nv50->screen->base.channel;
- struct nouveau_grobj *tesla = nv50->screen->tesla;
- struct nouveau_grobj *eng2d = nv50->screen->eng2d;
- int ret;
- unsigned offset;
- unsigned size = p->code_size;
- uint32_t *data = p->code;
+ if (words) {
+ MARK_RING(chan, 8, 1);
- assert(p->translated);
+ nouveau_bo_validate(chan, bo, res->domain | NOUVEAU_BO_WR);
+ }
- /* TODO: use a single bo (for each type) for shader code */
- if (p->bo)
- return;
- ret = nouveau_bo_new(chan->device, NOUVEAU_BO_VRAM, 0x100, size, &p->bo);
- assert(!ret);
-
- offset = p->code_start = 0;
-
- BEGIN_RING(chan, eng2d, NV50_2D_DST_FORMAT, 2);
- OUT_RING (chan, NV50_2D_DST_FORMAT_R8_UNORM);
- OUT_RING (chan, 1);
- BEGIN_RING(chan, eng2d, NV50_2D_DST_PITCH, 1);
- OUT_RING (chan, 0x40000);
- BEGIN_RING(chan, eng2d, NV50_2D_DST_WIDTH, 2);
- OUT_RING (chan, 0x10000);
- OUT_RING (chan, 1);
-
- while (size) {
- unsigned nr = size / 4;
-
- if (AVAIL_RING(chan) < 32)
- FIRE_RING(chan);
-
- nr = MIN2(nr, AVAIL_RING(chan) - 18);
- nr = MIN2(nr, 1792);
- if (nr < (size / 4))
- nr &= ~0x3f;
- assert(!(size & 3));
-
- BEGIN_RING(chan, eng2d, NV50_2D_DST_ADDRESS_HIGH, 2);
- OUT_RELOCh(chan, p->bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
- OUT_RELOCl(chan, p->bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
- BEGIN_RING(chan, eng2d, NV50_2D_SIFC_BITMAP_ENABLE, 2);
- OUT_RING (chan, 0);
- OUT_RING (chan, NV50_2D_SIFC_FORMAT_R8_UNORM);
- BEGIN_RING(chan, eng2d, NV50_2D_SIFC_WIDTH, 10);
- OUT_RING (chan, nr * 4);
- OUT_RING (chan, 1);
- OUT_RING (chan, 0);
- OUT_RING (chan, 1);
- OUT_RING (chan, 0);
- OUT_RING (chan, 1);
- OUT_RING (chan, 0);
- OUT_RING (chan, 0);
- OUT_RING (chan, 0);
- OUT_RING (chan, 0);
-
- BEGIN_RING_NI(chan, eng2d, NV50_2D_SIFC_DATA, nr);
- OUT_RINGp (chan, data, nr);
-
- data += nr;
- offset += nr * 4;
- size -= nr * 4;
- }
+ while (words) {
+ unsigned nr = AVAIL_RING(chan);
- BEGIN_RING(chan, tesla, NV50TCL_CODE_CB_FLUSH, 1);
- OUT_RING (chan, 0);
-}
+ if (nr < 16) {
+ FIRE_RING(chan);
+ nouveau_bo_validate(chan, bo, res->domain | NOUVEAU_BO_WR);
+ continue;
+ }
+ nr = MIN2(MIN2(nr - 3, words), NV04_PFIFO_MAX_PACKET_LEN);
-static void
-nv50_vp_update_stateobj(struct nv50_context *nv50, struct nv50_program *p)
-{
- struct nouveau_grobj *tesla = nv50->screen->tesla;
- struct nouveau_stateobj *so = so_new(5, 7, 2);
-
- nv50_program_validate_code(nv50, p);
-
- so_method(so, tesla, NV50TCL_VP_ADDRESS_HIGH, 2);
- so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
- NOUVEAU_BO_HIGH, 0, 0);
- so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
- NOUVEAU_BO_LOW, 0, 0);
- so_method(so, tesla, NV50TCL_VP_ATTR_EN_0, 2);
- so_data (so, p->vp.attrs[0]);
- so_data (so, p->vp.attrs[1]);
- so_method(so, tesla, NV50TCL_VP_REG_ALLOC_RESULT, 1);
- so_data (so, p->max_out);
- so_method(so, tesla, NV50TCL_VP_REG_ALLOC_TEMP, 1);
- so_data (so, p->max_gpr);
- so_method(so, tesla, NV50TCL_VP_START_ID, 1);
- so_data (so, p->code_start);
-
- so_ref(so, &p->so);
- so_ref(NULL, &so);
-}
+ BEGIN_RING(chan, RING_3D(CB_ADDR), 1);
+ OUT_RING (chan, (start << 8) | b);
+ BEGIN_RING_NI(chan, RING_3D(CB_DATA(0)), nr);
+ OUT_RINGp (chan, &res->data[start * 4], nr);
-static void
-nv50_fp_update_stateobj(struct nv50_context *nv50, struct nv50_program *p)
-{
- struct nouveau_grobj *tesla = nv50->screen->tesla;
- struct nouveau_stateobj *so = so_new(6, 7, 2);
-
- nv50_program_validate_code(nv50, p);
-
- so_method(so, tesla, NV50TCL_FP_ADDRESS_HIGH, 2);
- so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
- NOUVEAU_BO_HIGH, 0, 0);
- so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
- NOUVEAU_BO_LOW, 0, 0);
- so_method(so, tesla, NV50TCL_FP_REG_ALLOC_TEMP, 1);
- so_data (so, p->max_gpr);
- so_method(so, tesla, NV50TCL_FP_RESULT_COUNT, 1);
- so_data (so, p->max_out);
- so_method(so, tesla, NV50TCL_FP_CONTROL, 1);
- so_data (so, p->fp.flags[0]);
- so_method(so, tesla, NV50TCL_FP_CTRL_UNK196C, 1);
- so_data (so, p->fp.flags[1]);
- so_method(so, tesla, NV50TCL_FP_START_ID, 1);
- so_data (so, p->code_start);
-
- so_ref(so, &p->so);
- so_ref(NULL, &so);
-}
-
-static void
-nv50_gp_update_stateobj(struct nv50_context *nv50, struct nv50_program *p)
-{
- struct nouveau_grobj *tesla = nv50->screen->tesla;
- struct nouveau_stateobj *so = so_new(6, 7, 2);
-
- nv50_program_validate_code(nv50, p);
-
- so_method(so, tesla, NV50TCL_GP_ADDRESS_HIGH, 2);
- so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
- NOUVEAU_BO_HIGH, 0, 0);
- so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
- NOUVEAU_BO_LOW, 0, 0);
- so_method(so, tesla, NV50TCL_GP_REG_ALLOC_TEMP, 1);
- so_data (so, p->max_gpr);
- so_method(so, tesla, NV50TCL_GP_REG_ALLOC_RESULT, 1);
- so_data (so, p->max_out);
- so_method(so, tesla, NV50TCL_GP_OUTPUT_PRIMITIVE_TYPE, 1);
- so_data (so, p->gp.prim_type);
- so_method(so, tesla, NV50TCL_GP_VERTEX_OUTPUT_COUNT, 1);
- so_data (so, p->gp.vert_count);
- so_method(so, tesla, NV50TCL_GP_START_ID, 1);
- so_data (so, p->code_start);
-
- so_ref(so, &p->so);
- so_ref(NULL, &so);
+ start += nr;
+ words -= nr;
+ }
+ }
+ }
}
static boolean
-nv50_program_validate(struct nv50_program *p)
-{
- p->translated = nv50_program_tx(p);
- assert(p->translated);
- return p->translated;
-}
-
-struct nouveau_stateobj *
-nv50_vertprog_validate(struct nv50_context *nv50)
+nv50_program_validate(struct nv50_context *nv50, struct nv50_program *prog)
{
- struct nv50_program *p = nv50->vertprog;
- struct nouveau_stateobj *so = NULL;
-
- if (!p->translated) {
- if (nv50_program_validate(p))
- nv50_vp_update_stateobj(nv50, p);
- else
- return NULL;
- }
+ struct nouveau_resource *heap;
+ int ret;
+ unsigned size;
- if (nv50->dirty & NV50_NEW_VERTPROG_CB)
- nv50_program_validate_data(nv50, p);
+ if (prog->translated)
+ return TRUE;
- if (!(nv50->dirty & NV50_NEW_VERTPROG))
- return NULL;
+ prog->translated = nv50_program_translate(prog);
+ if (!prog->translated)
+ return FALSE;
- nv50_program_validate_code(nv50, p);
+ if (prog->type == PIPE_SHADER_FRAGMENT) heap = nv50->screen->fp_code_heap;
+ else
+ if (prog->type == PIPE_SHADER_GEOMETRY) heap = nv50->screen->gp_code_heap;
+ else
+ heap = nv50->screen->vp_code_heap;
- so_ref(p->so, &so);
- return so;
-}
+ size = align(prog->code_size, 0x100);
-struct nouveau_stateobj *
-nv50_fragprog_validate(struct nv50_context *nv50)
-{
- struct nv50_program *p = nv50->fragprog;
- struct nouveau_stateobj *so = NULL;
-
- if (!p->translated) {
- if (nv50_program_validate(p))
- nv50_fp_update_stateobj(nv50, p);
- else
- return NULL;
+ ret = nouveau_resource_alloc(heap, size, prog, &prog->res);
+ if (ret) {
+ NOUVEAU_ERR("out of code space for shader type %i\n", prog->type);
+ return FALSE;
}
+ prog->code_base = prog->res->start;
- if (nv50->dirty & NV50_NEW_FRAGPROG_CB)
- nv50_program_validate_data(nv50, p);
+ nv50_relocate_program(prog, prog->code_base, 0);
- if (!(nv50->dirty & NV50_NEW_FRAGPROG))
- return NULL;
+ nv50_sifc_linear_u8(&nv50->base, nv50->screen->code,
+ (prog->type << NV50_CODE_BO_SIZE_LOG2) + prog->code_base,
+ NOUVEAU_BO_VRAM, prog->code_size, prog->code);
- nv50_program_validate_code(nv50, p);
+ BEGIN_RING(nv50->screen->base.channel, RING_3D(CODE_CB_FLUSH), 1);
+ OUT_RING (nv50->screen->base.channel, 0);
- so_ref(p->so, &so);
- return so;
+ return TRUE;
}
-struct nouveau_stateobj *
-nv50_geomprog_validate(struct nv50_context *nv50)
+void
+nv50_vertprog_validate(struct nv50_context *nv50)
{
- struct nv50_program *p = nv50->geomprog;
- struct nouveau_stateobj *so = NULL;
-
- if (!p->translated) {
- if (nv50_program_validate(p))
- nv50_gp_update_stateobj(nv50, p);
- else
- return NULL;
- }
+ struct nouveau_channel *chan = nv50->screen->base.channel;
+ struct nv50_program *vp = nv50->vertprog;
- if (nv50->dirty & NV50_NEW_GEOMPROG_CB)
- nv50_program_validate_data(nv50, p);
+ if (!nv50_program_validate(nv50, vp))
+ return;
+
+ BEGIN_RING(chan, RING_3D(VP_ATTR_EN(0)), 2);
+ OUT_RING (chan, vp->vp.attrs[0]);
+ OUT_RING (chan, vp->vp.attrs[1]);
+ BEGIN_RING(chan, RING_3D(VP_REG_ALLOC_RESULT), 1);
+ OUT_RING (chan, vp->max_out);
+ BEGIN_RING(chan, RING_3D(VP_REG_ALLOC_TEMP), 1);
+ OUT_RING (chan, vp->max_gpr);
+ BEGIN_RING(chan, RING_3D(VP_START_ID), 1);
+ OUT_RING (chan, vp->code_base);
+}
- if (!(nv50->dirty & NV50_NEW_GEOMPROG))
- return NULL;
+void
+nv50_fragprog_validate(struct nv50_context *nv50)
+{
+ struct nouveau_channel *chan = nv50->screen->base.channel;
+ struct nv50_program *fp = nv50->fragprog;
- nv50_program_validate_code(nv50, p);
+ if (!nv50_program_validate(nv50, fp))
+ return;
+
+ BEGIN_RING(chan, RING_3D(FP_REG_ALLOC_TEMP), 1);
+ OUT_RING (chan, fp->max_gpr);
+ BEGIN_RING(chan, RING_3D(FP_RESULT_COUNT), 1);
+ OUT_RING (chan, fp->max_out);
+ BEGIN_RING(chan, RING_3D(FP_CONTROL), 1);
+ OUT_RING (chan, fp->fp.flags[0]);
+ BEGIN_RING(chan, RING_3D(FP_CTRL_UNK196C), 1);
+ OUT_RING (chan, fp->fp.flags[1]);
+ BEGIN_RING(chan, RING_3D(FP_START_ID), 1);
+ OUT_RING (chan, fp->code_base);
+}
- so_ref(p->so, &so);
- return so;
+void
+nv50_gmtyprog_validate(struct nv50_context *nv50)
+{
+ struct nouveau_channel *chan = nv50->screen->base.channel;
+ struct nv50_program *gp = nv50->vertprog;
+
+ if (!nv50_program_validate(nv50, gp))
+ return;
+
+ BEGIN_RING(chan, RING_3D(GP_REG_ALLOC_TEMP), 1);
+ OUT_RING (chan, gp->max_gpr);
+ BEGIN_RING(chan, RING_3D(GP_REG_ALLOC_RESULT), 1);
+ OUT_RING (chan, gp->max_out);
+ BEGIN_RING(chan, RING_3D(GP_OUTPUT_PRIMITIVE_TYPE), 1);
+ OUT_RING (chan, gp->gp.prim_type);
+ BEGIN_RING(chan, RING_3D(GP_VERTEX_OUTPUT_COUNT), 1);
+ OUT_RING (chan, gp->gp.vert_count);
+ BEGIN_RING(chan, RING_3D(GP_START_ID), 1);
+ OUT_RING (chan, gp->code_base);
}
-/* XXX: this might not work correctly in all cases yet: we assume that
- * an FP generic input that is not written in the VP is gl_PointCoord.
- */
-static uint32_t
-nv50_pntc_replace(struct nv50_context *nv50, uint32_t pntc[8], unsigned m)
+static void
+nv50_sprite_coords_validate(struct nv50_context *nv50)
{
- struct nv50_program *vp = nv50->vertprog;
+ struct nouveau_channel *chan = nv50->screen->base.channel;
+ uint32_t pntc[8], mode;
struct nv50_program *fp = nv50->fragprog;
unsigned i, c;
+ unsigned m = (nv50->state.interpolant_ctrl >> 8) & 0xff;
- memset(pntc, 0, 8 * sizeof(uint32_t));
+ if (!nv50->rast->pipe.point_quad_rasterization) {
+ if (nv50->state.point_sprite) {
+ BEGIN_RING(chan, RING_3D(POINT_COORD_REPLACE_MAP(0)), 8);
+ for (i = 0; i < 8; ++i)
+ OUT_RING(chan, 0);
+
+ nv50->state.point_sprite = FALSE;
+ }
+ return;
+ } else {
+ nv50->state.point_sprite = TRUE;
+ }
- if (nv50->geomprog)
- vp = nv50->geomprog;
+ memset(pntc, 0, sizeof(pntc));
for (i = 0; i < fp->in_nr; i++) {
- unsigned j, n = util_bitcount(fp->in[i].mask);
+ unsigned n = util_bitcount(fp->in[i].mask);
if (fp->in[i].sn != TGSI_SEMANTIC_GENERIC) {
m += n;
continue;
}
+ if (!(nv50->rast->pipe.sprite_coord_enable & (1 << fp->in[i].si))) {
+ m += n;
+ continue;
+ }
- for (j = 0; j < vp->out_nr; ++j)
- if (vp->out[j].sn == fp->in[i].sn && vp->out[j].si == fp->in[i].si)
- break;
-
- if (j < vp->out_nr) {
- uint32_t en = nv50->rasterizer->pipe.sprite_coord_enable;
-
- if (!(en & (1 << vp->out[j].si))) {
- m += n;
- continue;
+ for (c = 0; c < 4; ++c) {
+ if (fp->in[i].mask & (1 << c)) {
+ pntc[m / 8] |= (c + 1) << ((m % 8) * 4);
+ ++m;
}
}
+ }
- /* this is either PointCoord or replaced by sprite coords */
- for (c = 0; c < 4; c++) {
- if (!(fp->in[i].mask & (1 << c)))
- continue;
- pntc[m / 8] |= (c + 1) << ((m % 8) * 4);
- ++m;
- }
+ if (nv50->rast->pipe.sprite_coord_mode == PIPE_SPRITE_COORD_LOWER_LEFT)
+ mode = 0x00;
+ else
+ mode = 0x10;
+
+ BEGIN_RING(chan, RING_3D(POINT_SPRITE_CTRL), 1);
+ OUT_RING (chan, mode);
+
+ BEGIN_RING(chan, RING_3D(POINT_COORD_REPLACE_MAP(0)), 8);
+ OUT_RINGp (chan, pntc, 8);
+}
+
+/* Validate state derived from shaders and the rasterizer cso. */
+void
+nv50_validate_derived_rs(struct nv50_context *nv50)
+{
+ struct nouveau_channel *chan = nv50->screen->base.channel;
+ uint32_t color, psize;
+
+ nv50_sprite_coords_validate(nv50);
+
+ if (nv50->dirty & NV50_NEW_FRAGPROG)
+ return;
+ psize = nv50->state.semantic_psize & ~NV50_3D_MAP_SEMANTIC_3_PTSZ_EN__MASK;
+ color = nv50->state.semantic_color & ~NV50_3D_MAP_SEMANTIC_0_CLMP_EN;
+
+ if (nv50->rast->pipe.clamp_vertex_color)
+ color |= NV50_3D_MAP_SEMANTIC_0_CLMP_EN;
+
+ if (color != nv50->state.semantic_color) {
+ nv50->state.semantic_color = color;
+ BEGIN_RING(chan, RING_3D(MAP_SEMANTIC_0), 1);
+ OUT_RING (chan, color);
+ }
+
+ if (nv50->rast->pipe.point_size_per_vertex)
+ psize |= NV50_3D_MAP_SEMANTIC_3_PTSZ_EN__MASK;
+
+ if (psize != nv50->state.semantic_psize) {
+ nv50->state.semantic_psize = psize;
+ BEGIN_RING(chan, RING_3D(MAP_SEMANTIC_3), 1);
+ OUT_RING (chan, psize);
}
- if (nv50->rasterizer->pipe.sprite_coord_mode == PIPE_SPRITE_COORD_LOWER_LEFT)
- return 0;
- return (1 << 4);
}
static int
-nv50_vec4_map(uint32_t *map32, int mid, uint32_t lin[4],
+nv50_vec4_map(uint8_t *map, int mid, uint32_t lin[4],
struct nv50_varying *in, struct nv50_varying *out)
{
int c;
uint8_t mv = out->mask, mf = in->mask, oid = out->hw;
- uint8_t *map = (uint8_t *)map32;
for (c = 0; c < 4; ++c) {
if (mf & 1) {
return mid;
}
-struct nouveau_stateobj *
+void
nv50_fp_linkage_validate(struct nv50_context *nv50)
{
- struct nouveau_grobj *tesla = nv50->screen->tesla;
- struct nv50_program *vp;
+ struct nouveau_channel *chan = nv50->screen->base.channel;
+ struct nv50_program *vp = nv50->gmtyprog ? nv50->gmtyprog : nv50->vertprog;
struct nv50_program *fp = nv50->fragprog;
- struct nouveau_stateobj *so;
struct nv50_varying dummy;
int i, n, c, m;
-
- uint32_t map[16], lin[4], pntc[8];
-
+ uint32_t primid = 0;
+ uint32_t psiz = 0x000;
uint32_t interp = fp->fp.interp;
uint32_t colors = fp->fp.colors;
- uint32_t clip = 0x04;
- uint32_t psiz = 0x000;
- uint32_t primid = 0;
- uint32_t sysval = 0;
+ uint32_t lin[4];
+ uint8_t map[64];
- if (nv50->geomprog) {
- vp = nv50->geomprog;
- memset(map, 0x80, sizeof(map));
- } else {
- vp = nv50->vertprog;
- memset(map, 0x40, sizeof(map));
- }
- memset(lin, 0, sizeof(lin));
+ memset(lin, 0x00, sizeof(lin));
+
+ /* XXX: in buggy-endian mode, is the first element of map (u32)0x000000xx
+ * or is it the first byte ?
+ */
+ memset(map, nv50->gmtyprog ? 0x80 : 0x40, sizeof(map));
- dummy.linear = 0;
dummy.mask = 0xf; /* map all components of HPOS */
+ dummy.linear = 0;
m = nv50_vec4_map(map, 0, lin, &dummy, &vp->out[0]);
- if (vp->vp.clpd < 0x40) {
- for (c = 0; c < vp->vp.clpd_nr; ++c) {
- map[m / 4] |= (vp->vp.clpd + c) << ((m % 4) * 8);
- ++m;
- }
- clip |= vp->vp.clpd_nr << 8;
- }
+ for (c = 0; c < vp->vp.clpd_nr; ++c)
+ map[m++] |= vp->vp.clpd + c;
colors |= m << 8; /* adjust BFC0 id */
- /* if light_twoside is active, it seems FFC0_ID == BFC0_ID is bad */
- if (nv50->rasterizer->pipe.light_twoside) {
+ /* if light_twoside is active, FFC0_ID == BFC0_ID is invalid */
+ if (nv50->rast->pipe.light_twoside) {
for (i = 0; i < 2; ++i)
m = nv50_vec4_map(map, m, lin,
- &fp->in[fp->vp.bfc[i]],
- &vp->out[vp->vp.bfc[i]]);
+ &fp->in[fp->vp.bfc[i]], &vp->out[vp->vp.bfc[i]]);
}
-
colors += m - 4; /* adjust FFC0 id */
- interp |= m << 8; /* set mid where 'normal' FP inputs start */
+ interp |= m << 8; /* set map id where 'normal' FP inputs start */
dummy.mask = 0x0;
- for (i = 0; i < fp->in_nr; i++) {
+ for (i = 0; i < fp->in_nr; ++i) {
for (n = 0; n < vp->out_nr; ++n)
if (vp->out[n].sn == fp->in[i].sn &&
vp->out[n].si == fp->in[i].si)
break;
-
m = nv50_vec4_map(map, m, lin,
&fp->in[i], (n < vp->out_nr) ? &vp->out[n] : &dummy);
- }
+ }
/* PrimitiveID either is replaced by the system value, or
* written by the geometry shader into an output register
*/
if (fp->gp.primid < 0x40) {
- i = (m % 4) * 8;
- map[m / 4] = (map[m / 4] & ~(0xff << i)) | (vp->gp.primid << i);
- primid = m++;
+ primid = m;
+ map[m++] = vp->gp.primid;
}
- if (nv50->rasterizer->pipe.point_size_per_vertex) {
- i = (m % 4) * 8;
- map[m / 4] = (map[m / 4] & ~(0xff << i)) | (vp->vp.psiz << i);
- psiz = (m++ << 4) | 1;
+ if (nv50->rast->pipe.point_size_per_vertex) {
+ psiz = (m << 4) | 1;
+ map[m++] = vp->vp.psiz;
}
- /* now fill the stateobj (at most 28 so_data) */
- so = so_new(10, 54, 0);
+ if (nv50->rast->pipe.clamp_vertex_color)
+ colors |= NV50_3D_MAP_SEMANTIC_0_CLMP_EN;
n = (m + 3) / 4;
assert(m <= 64);
- if (vp->type == PIPE_SHADER_GEOMETRY) {
- so_method(so, tesla, NV50TCL_GP_RESULT_MAP_SIZE, 1);
- so_data (so, m);
- so_method(so, tesla, NV50TCL_GP_RESULT_MAP(0), n);
- so_datap (so, map, n);
+
+ if (unlikely(nv50->gmtyprog)) {
+ BEGIN_RING(chan, RING_3D(GP_RESULT_MAP_SIZE), 1);
+ OUT_RING (chan, m);
+ BEGIN_RING(chan, RING_3D(GP_RESULT_MAP(0)), n);
+ OUT_RINGp (chan, map, n);
} else {
- so_method(so, tesla, NV50TCL_VP_GP_BUILTIN_ATTR_EN, 1);
- so_data (so, vp->vp.attrs[2]);
+ BEGIN_RING(chan, RING_3D(VP_GP_BUILTIN_ATTR_EN), 1);
+ OUT_RING (chan, vp->vp.attrs[2]);
- so_method(so, tesla, NV50TCL_MAP_SEMANTIC_4, 1);
- so_data (so, primid);
+ BEGIN_RING(chan, RING_3D(MAP_SEMANTIC_4), 1);
+ OUT_RING (chan, primid);
- so_method(so, tesla, NV50TCL_VP_RESULT_MAP_SIZE, 1);
- so_data (so, m);
- so_method(so, tesla, NV50TCL_VP_RESULT_MAP(0), n);
- so_datap (so, map, n);
+ BEGIN_RING(chan, RING_3D(VP_RESULT_MAP_SIZE), 1);
+ OUT_RING (chan, m);
+ BEGIN_RING(chan, RING_3D(VP_RESULT_MAP(0)), n);
+ OUT_RINGp (chan, map, n);
}
- so_method(so, tesla, NV50TCL_MAP_SEMANTIC_0, 4);
- so_data (so, colors);
- so_data (so, clip);
- so_data (so, sysval);
- so_data (so, psiz);
-
- so_method(so, tesla, NV50TCL_FP_INTERPOLANT_CTRL, 1);
- so_data (so, interp);
+ BEGIN_RING(chan, RING_3D(MAP_SEMANTIC_0), 4);
+ OUT_RING (chan, colors);
+ OUT_RING (chan, (vp->vp.clpd_nr << 8) | 4);
+ OUT_RING (chan, 0);
+ OUT_RING (chan, psiz);
- so_method(so, tesla, NV50TCL_NOPERSPECTIVE_BITMAP(0), 4);
- so_datap (so, lin, 4);
+ BEGIN_RING(chan, RING_3D(FP_INTERPOLANT_CTRL), 1);
+ OUT_RING (chan, interp);
- if (nv50->rasterizer->pipe.point_quad_rasterization) {
- so_method(so, tesla, NV50TCL_POINT_SPRITE_CTRL, 1);
- so_data (so,
- nv50_pntc_replace(nv50, pntc, (interp >> 8) & 0xff));
+ nv50->state.interpolant_ctrl = interp;
- so_method(so, tesla, NV50TCL_POINT_COORD_REPLACE_MAP(0), 8);
- so_datap (so, pntc, 8);
- }
+ nv50->state.semantic_color = colors;
+ nv50->state.semantic_psize = psiz;
- so_method(so, tesla, NV50TCL_GP_ENABLE, 1);
- so_data (so, (vp->type == PIPE_SHADER_GEOMETRY) ? 1 : 0);
+ BEGIN_RING(chan, RING_3D(NOPERSPECTIVE_BITMAP(0)), 4);
+ OUT_RINGp (chan, lin, 4);
- return so;
+ BEGIN_RING(chan, RING_3D(GP_ENABLE), 1);
+ OUT_RING (chan, nv50->gmtyprog ? 1 : 0);
}
static int
-nv50_vp_gp_mapping(uint32_t *map32, int m,
+nv50_vp_gp_mapping(uint8_t *map, int m,
struct nv50_program *vp, struct nv50_program *gp)
{
- uint8_t *map = (uint8_t *)map32;
int i, j, c;
for (i = 0; i < gp->in_nr; ++i) {
return m;
}
-struct nouveau_stateobj *
+void
nv50_gp_linkage_validate(struct nv50_context *nv50)
{
- struct nouveau_grobj *tesla = nv50->screen->tesla;
- struct nouveau_stateobj *so;
+ struct nouveau_channel *chan = nv50->screen->base.channel;
struct nv50_program *vp = nv50->vertprog;
- struct nv50_program *gp = nv50->geomprog;
- uint32_t map[16];
+ struct nv50_program *gp = nv50->gmtyprog;
int m = 0;
+ int n;
+ uint8_t map[64];
if (!gp)
- return NULL;
+ return;
memset(map, 0, sizeof(map));
m = nv50_vp_gp_mapping(map, m, vp, gp);
- so = so_new(3, 24 - 3, 0);
-
- so_method(so, tesla, NV50TCL_VP_GP_BUILTIN_ATTR_EN, 1);
- so_data (so, vp->vp.attrs[2] | gp->vp.attrs[2]);
-
- assert(m <= 32);
- so_method(so, tesla, NV50TCL_VP_RESULT_MAP_SIZE, 1);
- so_data (so, m);
+ n = (m + 3) / 4;
- m = (m + 3) / 4;
- so_method(so, tesla, NV50TCL_VP_RESULT_MAP(0), m);
- so_datap (so, map, m);
+ BEGIN_RING(chan, RING_3D(VP_GP_BUILTIN_ATTR_EN), 1);
+ OUT_RING (chan, vp->vp.attrs[2] | gp->vp.attrs[2]);
- return so;
+ BEGIN_RING(chan, RING_3D(VP_RESULT_MAP_SIZE), 1);
+ OUT_RING (chan, m);
+ BEGIN_RING(chan, RING_3D(VP_RESULT_MAP(0)), n);
+ OUT_RINGp (chan, map, n);
}
projects / mesa.git / blobdiff