summaryrefslogtreecommitdiff
path: root/src/gallium/drivers/nv50/nv50_program.c
diff options
context:
space:
mode:
Diffstat (limited to 'src/gallium/drivers/nv50/nv50_program.c')
-rw-r--r--src/gallium/drivers/nv50/nv50_program.c70
1 files changed, 38 insertions, 32 deletions
diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c
index 2372cbbef6..b8b6b12120 100644
--- a/src/gallium/drivers/nv50/nv50_program.c
+++ b/src/gallium/drivers/nv50/nv50_program.c
@@ -30,6 +30,7 @@
#include "tgsi/tgsi_util.h"
#include "nv50_context.h"
+#include "nv50_transfer.h"
#define NV50_SU_MAX_TEMP 127
#define NV50_SU_MAX_ADDR 4
@@ -599,8 +600,8 @@ static void
emit_interp(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *iv,
unsigned mode)
{
- assert(dst->rhw != -1);
struct nv50_program_exec *e = exec(pc);
+ assert(dst->rhw != -1);
e->inst[0] |= 0x80000000;
set_dst(pc, dst, e);
@@ -1614,7 +1615,7 @@ emit_lit(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask,
struct nv50_reg *zero = alloc_immd(pc, 0.0);
struct nv50_reg *neg128 = alloc_immd(pc, -127.999999);
struct nv50_reg *pos128 = alloc_immd(pc, 127.999999);
- struct nv50_reg *tmp[4];
+ struct nv50_reg *tmp[4] = { 0 };
boolean allow32 = pc->allow32;
pc->allow32 = FALSE;
@@ -2552,7 +2553,7 @@ nv50_program_tx_insn(struct nv50_pc *pc,
const struct tgsi_full_instruction *inst)
{
struct nv50_reg *rdst[4], *dst[4], *brdc, *src[3][4], *temp;
- unsigned mask, sat, unit;
+ unsigned mask, sat, unit = 0;
int i, c;
mask = inst->Dst[0].Register.WriteMask;
@@ -3530,7 +3531,7 @@ nv50_program_tx_prep(struct nv50_pc *pc)
struct tgsi_parse_context tp;
struct nv50_program *p = pc->p;
boolean ret = FALSE;
- unsigned i, c, instance_id, vertex_id, flat_nr = 0;
+ unsigned i, c, instance_id = 0, vertex_id = 0, flat_nr = 0;
tgsi_parse_init(&tp, pc->p->pipe.tokens);
while (!tgsi_parse_end_of_tokens(&tp)) {
@@ -3728,7 +3729,7 @@ nv50_program_tx_prep(struct nv50_pc *pc)
copy_semantic_info(p);
} else
if (p->type == PIPE_SHADER_FRAGMENT) {
- int rid, aid;
+ int rid = 0, aid;
unsigned n = 0, m = pc->attr_nr - flat_nr;
pc->allow32 = TRUE;
@@ -3762,7 +3763,7 @@ nv50_program_tx_prep(struct nv50_pc *pc)
p->cfg.in[n].hw = rid = aid;
i = p->cfg.in[n].id;
- if (p->info.input_semantic_name[n] ==
+ if (p->info.input_semantic_name[i] ==
TGSI_SEMANTIC_FACE) {
load_frontfacing(pc, &pc->attr[i * 4]);
continue;
@@ -3821,7 +3822,7 @@ nv50_program_tx_prep(struct nv50_pc *pc)
for (rid = 0; i < pc->result_nr * 4; i++)
pc->result[i].rhw = rid++;
if (p->info.writes_z)
- pc->result[2].rhw = rid;
+ pc->result[2].rhw = rid++;
p->cfg.high_result = rid;
@@ -4157,7 +4158,8 @@ nv50_program_upload_data(struct nv50_context *nv50, uint32_t *map,
static void
nv50_program_validate_data(struct nv50_context *nv50, struct nv50_program *p)
{
- struct pipe_screen *pscreen = nv50->pipe.screen;
+ struct pipe_context *pipe = &nv50->pipe;
+ struct pipe_transfer *transfer;
if (!p->data[0] && p->immd_nr) {
struct nouveau_resource *heap = nv50->screen->immd_heap[0];
@@ -4182,9 +4184,10 @@ nv50_program_validate_data(struct nv50_context *nv50, struct nv50_program *p)
if (p->param_nr) {
unsigned cb;
- uint32_t *map = pipe_buffer_map(pscreen,
+ uint32_t *map = pipe_buffer_map(pipe,
nv50->constbuf[p->type],
- PIPE_BUFFER_USAGE_CPU_READ);
+ PIPE_TRANSFER_READ,
+ &transfer);
switch (p->type) {
case PIPE_SHADER_GEOMETRY: cb = NV50_CB_PGP; break;
case PIPE_SHADER_FRAGMENT: cb = NV50_CB_PFP; break;
@@ -4195,7 +4198,8 @@ nv50_program_validate_data(struct nv50_context *nv50, struct nv50_program *p)
}
nv50_program_upload_data(nv50, map, 0, p->param_nr, cb);
- pipe_buffer_unmap(pscreen, nv50->constbuf[p->type]);
+ pipe_buffer_unmap(pipe, nv50->constbuf[p->type],
+ transfer);
}
}
@@ -4270,7 +4274,7 @@ nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p)
FREE(up);
}
-void
+struct nouveau_stateobj *
nv50_vertprog_validate(struct nv50_context *nv50)
{
struct nouveau_grobj *tesla = nv50->screen->tesla;
@@ -4286,6 +4290,9 @@ nv50_vertprog_validate(struct nv50_context *nv50)
nv50_program_validate_data(nv50, p);
nv50_program_validate_code(nv50, p);
+ if (!(nv50->dirty & NV50_NEW_VERTPROG))
+ return NULL;
+
so = so_new(5, 7, 2);
so_method(so, tesla, NV50TCL_VP_ADDRESS_HIGH, 2);
so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
@@ -4301,11 +4308,10 @@ nv50_vertprog_validate(struct nv50_context *nv50)
so_data (so, p->cfg.high_temp);
so_method(so, tesla, NV50TCL_VP_START_ID, 1);
so_data (so, 0); /* program start offset */
- so_ref(so, &nv50->state.vertprog);
- so_ref(NULL, &so);
+ return so;
}
-void
+struct nouveau_stateobj *
nv50_fragprog_validate(struct nv50_context *nv50)
{
struct nouveau_grobj *tesla = nv50->screen->tesla;
@@ -4321,6 +4327,9 @@ nv50_fragprog_validate(struct nv50_context *nv50)
nv50_program_validate_data(nv50, p);
nv50_program_validate_code(nv50, p);
+ if (!(nv50->dirty & NV50_NEW_FRAGPROG))
+ return NULL;
+
so = so_new(6, 7, 2);
so_method(so, tesla, NV50TCL_FP_ADDRESS_HIGH, 2);
so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
@@ -4337,11 +4346,10 @@ nv50_fragprog_validate(struct nv50_context *nv50)
so_data (so, p->cfg.regs[3]);
so_method(so, tesla, NV50TCL_FP_START_ID, 1);
so_data (so, 0); /* program start offset */
- so_ref(so, &nv50->state.fragprog);
- so_ref(NULL, &so);
+ return so;
}
-void
+struct nouveau_stateobj *
nv50_geomprog_validate(struct nv50_context *nv50)
{
struct nouveau_grobj *tesla = nv50->screen->tesla;
@@ -4357,6 +4365,9 @@ nv50_geomprog_validate(struct nv50_context *nv50)
nv50_program_validate_data(nv50, p);
nv50_program_validate_code(nv50, p);
+ if (!(nv50->dirty & NV50_NEW_GEOMPROG))
+ return NULL;
+
so = so_new(6, 7, 2);
so_method(so, tesla, NV50TCL_GP_ADDRESS_HIGH, 2);
so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
@@ -4373,8 +4384,7 @@ nv50_geomprog_validate(struct nv50_context *nv50)
so_data (so, p->cfg.vert_count);
so_method(so, tesla, NV50TCL_GP_START_ID, 1);
so_data (so, 0);
- so_ref(so, &nv50->state.geomprog);
- so_ref(NULL, &so);
+ return so;
}
static uint32_t
@@ -4454,7 +4464,7 @@ nv50_vec4_map(uint32_t *map32, int mid, uint8_t zval, uint32_t lin[4],
return mid;
}
-void
+struct nouveau_stateobj *
nv50_fp_linkage_validate(struct nv50_context *nv50)
{
struct nouveau_grobj *tesla = nv50->screen->tesla;
@@ -4540,7 +4550,7 @@ nv50_fp_linkage_validate(struct nv50_context *nv50)
so = so_new(10, 54, 0);
n = (m + 3) / 4;
- assert(m <= 32);
+ assert(m <= 64);
if (vp->type == PIPE_SHADER_GEOMETRY) {
so_method(so, tesla, NV50TCL_GP_RESULT_MAP_SIZE, 1);
so_data (so, m);
@@ -4580,8 +4590,7 @@ nv50_fp_linkage_validate(struct nv50_context *nv50)
so_method(so, tesla, NV50TCL_GP_ENABLE, 1);
so_data (so, (vp->type == PIPE_SHADER_GEOMETRY) ? 1 : 0);
- so_ref(so, &nv50->state.fp_linkage);
- so_ref(NULL, &so);
+ return so;
}
static int
@@ -4592,7 +4601,7 @@ construct_vp_gp_mapping(uint32_t *map32, int m,
int i, j, c;
for (i = 0; i < gp->cfg.in_nr; ++i) {
- uint8_t oid, mv = 0, mg = gp->cfg.in[i].mask;
+ uint8_t oid = 0, mv = 0, mg = gp->cfg.in[i].mask;
for (j = 0; j < vp->cfg.out_nr; ++j) {
if (vp->cfg.out[j].sn == gp->cfg.in[i].sn &&
@@ -4615,7 +4624,7 @@ construct_vp_gp_mapping(uint32_t *map32, int m,
return m;
}
-void
+struct nouveau_stateobj *
nv50_gp_linkage_validate(struct nv50_context *nv50)
{
struct nouveau_grobj *tesla = nv50->screen->tesla;
@@ -4625,10 +4634,8 @@ nv50_gp_linkage_validate(struct nv50_context *nv50)
uint32_t map[16];
int m = 0;
- if (!gp) {
- so_ref(NULL, &nv50->state.gp_linkage);
- return;
- }
+ if (!gp)
+ return NULL;
memset(map, 0, sizeof(map));
m = construct_vp_gp_mapping(map, m, vp, gp);
@@ -4646,8 +4653,7 @@ nv50_gp_linkage_validate(struct nv50_context *nv50)
so_method(so, tesla, NV50TCL_VP_RESULT_MAP(0), m);
so_datap (so, map, m);
- so_ref(so, &nv50->state.gp_linkage);
- so_ref(NULL, &so);
+ return so;
}
void
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-14 05:14:04 +0000