{
struct r600_shader_io *input = &ps->input[id];
- /* position/face doesn't get/need a semantic index */
- if (input->name == TGSI_SEMANTIC_POSITION || input->name == TGSI_SEMANTIC_FACE)
- return 0;
-
for (int i = 0; i < vs->noutput; i++) {
if (input->name == vs->output[i].name &&
input->sid == vs->output[i].sid) {
/* build state */
switch (rshader->processor_type) {
case TGSI_PROCESSOR_VERTEX:
- if (rshader->family >= CHIP_CEDAR) {
+ if (rctx->chip_class >= EVERGREEN) {
evergreen_pipe_shader_vs(ctx, shader);
} else {
r600_pipe_shader_vs(ctx, shader);
}
break;
case TGSI_PROCESSOR_FRAGMENT:
- if (rshader->family >= CHIP_CEDAR) {
+ if (rctx->chip_class >= EVERGREEN) {
evergreen_pipe_shader_ps(ctx, shader);
} else {
r600_pipe_shader_ps(ctx, shader);
return 0;
}
-static int r600_shader_from_tgsi(const struct tgsi_token *tokens, struct r600_shader *shader);
+static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pipe_shader *pipeshader);
-int r600_pipe_shader_create(struct pipe_context *ctx, struct r600_pipe_shader *shader, const struct tgsi_token *tokens)
+int r600_pipe_shader_create(struct pipe_context *ctx, struct r600_pipe_shader *shader)
{
static int dump_shaders = -1;
struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx;
if (dump_shaders) {
fprintf(stderr, "--------------------------------------------------------------\n");
- tgsi_dump(tokens, 0);
+ tgsi_dump(shader->tokens, 0);
}
- shader->shader.family = r600_get_family(rctx->radeon);
- r = r600_shader_from_tgsi(tokens, &shader->shader);
+ r = r600_shader_from_tgsi(rctx, shader);
if (r) {
R600_ERR("translation from TGSI failed !\n");
return r;
r600_bo_reference(rctx->radeon, &shader->bo, NULL);
r600_bc_clear(&shader->shader.bc);
+
+ memset(&shader->shader,0,sizeof(struct r600_shader));
}
/*
struct r600_shader_tgsi_instruction *inst_info;
struct r600_bc *bc;
struct r600_shader *shader;
- struct r600_shader_src src[3];
+ struct r600_shader_src src[4];
u32 *literals;
u32 nliterals;
u32 max_driver_temp_used;
ctx->shader->input[i].interpolate = d->Declaration.Interpolate;
ctx->shader->input[i].centroid = d->Declaration.Centroid;
ctx->shader->input[i].gpr = ctx->file_offset[TGSI_FILE_INPUT] + i;
- if (ctx->type == TGSI_PROCESSOR_FRAGMENT && ctx->bc->chiprev >= CHIPREV_EVERGREEN) {
+ if (ctx->type == TGSI_PROCESSOR_FRAGMENT && ctx->bc->chip_class >= EVERGREEN) {
/* turn input into interpolate on EG */
if (ctx->shader->input[i].name != TGSI_SEMANTIC_POSITION) {
if (ctx->shader->input[i].interpolate > 0) {
return 0;
}
-static int r600_shader_from_tgsi(const struct tgsi_token *tokens, struct r600_shader *shader)
+static int r600_shader_from_tgsi(struct r600_pipe_context * rctx, struct r600_pipe_shader *pipeshader)
{
+ struct r600_shader *shader = &pipeshader->shader;
+ struct tgsi_token *tokens = pipeshader->tokens;
struct tgsi_full_immediate *immediate;
struct tgsi_full_property *property;
struct r600_shader_ctx ctx;
struct r600_bc_output output[32];
unsigned output_done, noutput;
unsigned opcode;
- int i, r = 0, pos0;
+ int i, j, r = 0, pos0;
ctx.bc = &shader->bc;
ctx.shader = shader;
- r = r600_bc_init(ctx.bc, shader->family);
- if (r)
- return r;
+ r600_bc_init(ctx.bc, rctx->chip_class);
ctx.tokens = tokens;
tgsi_scan_shader(tokens, &ctx.info);
tgsi_parse_init(&ctx.parse, tokens);
shader->processor_type = ctx.type;
ctx.bc->type = shader->processor_type;
+ shader->clamp_color = (((ctx.type == TGSI_PROCESSOR_FRAGMENT) && rctx->clamp_fragment_color) ||
+ ((ctx.type == TGSI_PROCESSOR_VERTEX) && rctx->clamp_vertex_color));
+
+ shader->nr_cbufs = rctx->nr_cbufs;
+
/* register allocations */
/* Values [0,127] correspond to GPR[0..127].
* Values [128,159] correspond to constant buffer bank 0
}
if (ctx.type == TGSI_PROCESSOR_VERTEX) {
ctx.file_offset[TGSI_FILE_INPUT] = 1;
- if (ctx.bc->chiprev >= CHIPREV_EVERGREEN) {
+ if (ctx.bc->chip_class >= EVERGREEN) {
r600_bc_add_cfinst(ctx.bc, EG_V_SQ_CF_WORD1_SQ_CF_INST_CALL_FS);
} else {
r600_bc_add_cfinst(ctx.bc, V_SQ_CF_WORD1_SQ_CF_INST_CALL_FS);
}
}
- if (ctx.type == TGSI_PROCESSOR_FRAGMENT && ctx.bc->chiprev >= CHIPREV_EVERGREEN) {
+ if (ctx.type == TGSI_PROCESSOR_FRAGMENT && ctx.bc->chip_class >= EVERGREEN) {
ctx.file_offset[TGSI_FILE_INPUT] = evergreen_gpr_count(&ctx);
}
ctx.file_offset[TGSI_FILE_OUTPUT] = ctx.file_offset[TGSI_FILE_INPUT] +
goto out_err;
if ((r = tgsi_split_literal_constant(&ctx)))
goto out_err;
- if (ctx.bc->chiprev == CHIPREV_CAYMAN)
+ if (ctx.bc->chip_class == CAYMAN)
ctx.inst_info = &cm_shader_tgsi_instruction[opcode];
- else if (ctx.bc->chiprev >= CHIPREV_EVERGREEN)
+ else if (ctx.bc->chip_class >= EVERGREEN)
ctx.inst_info = &eg_shader_tgsi_instruction[opcode];
else
ctx.inst_info = &r600_shader_tgsi_instruction[opcode];
goto out_err;
}
}
- /* export output */
+
noutput = shader->noutput;
+
+ /* clamp color outputs */
+ if (shader->clamp_color) {
+ for (i = 0; i < noutput; i++) {
+ if (shader->output[i].name == TGSI_SEMANTIC_COLOR ||
+ shader->output[i].name == TGSI_SEMANTIC_BCOLOR) {
+
+ int j;
+ for (j = 0; j < 4; j++) {
+ struct r600_bc_alu alu;
+ memset(&alu, 0, sizeof(struct r600_bc_alu));
+
+ /* MOV_SAT R, R */
+ alu.inst = BC_INST(ctx.bc, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV);
+ alu.dst.sel = shader->output[i].gpr;
+ alu.dst.chan = j;
+ alu.dst.write = 1;
+ alu.dst.clamp = 1;
+ alu.src[0].sel = alu.dst.sel;
+ alu.src[0].chan = j;
+
+ if (j == 3) {
+ alu.last = 1;
+ }
+ r = r600_bc_add_alu(ctx.bc, &alu);
+ if (r)
+ return r;
+ }
+ }
+ }
+ }
+
+ /* export output */
+ j = 0;
for (i = 0, pos0 = 0; i < noutput; i++) {
memset(&output[i], 0, sizeof(struct r600_bc_output));
- output[i].gpr = shader->output[i].gpr;
- output[i].elem_size = 3;
- output[i].swizzle_x = 0;
- output[i].swizzle_y = 1;
- output[i].swizzle_z = 2;
- output[i].swizzle_w = 3;
- output[i].burst_count = 1;
- output[i].barrier = 1;
- output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM;
- output[i].array_base = i - pos0;
- output[i].inst = BC_INST(ctx.bc, V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT);
+ output[i + j].gpr = shader->output[i].gpr;
+ output[i + j].elem_size = 3;
+ output[i + j].swizzle_x = 0;
+ output[i + j].swizzle_y = 1;
+ output[i + j].swizzle_z = 2;
+ output[i + j].swizzle_w = 3;
+ output[i + j].burst_count = 1;
+ output[i + j].barrier = 1;
+ output[i + j].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM;
+ output[i + j].array_base = i - pos0;
+ output[i + j].inst = BC_INST(ctx.bc, V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT);
switch (ctx.type) {
case TGSI_PROCESSOR_VERTEX:
if (shader->output[i].name == TGSI_SEMANTIC_POSITION) {
- output[i].array_base = 60;
- output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_POS;
+ output[i + j].array_base = 60;
+ output[i + j].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_POS;
/* position doesn't count in array_base */
pos0++;
}
if (shader->output[i].name == TGSI_SEMANTIC_PSIZE) {
- output[i].array_base = 61;
- output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_POS;
+ output[i + j].array_base = 61;
+ output[i + j].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_POS;
/* position doesn't count in array_base */
pos0++;
}
break;
case TGSI_PROCESSOR_FRAGMENT:
if (shader->output[i].name == TGSI_SEMANTIC_COLOR) {
- output[i].array_base = shader->output[i].sid;
- output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL;
+ output[i + j].array_base = shader->output[i].sid;
+ output[i + j].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL;
+ if (shader->fs_write_all && (rctx->chip_class >= EVERGREEN)) {
+ for (j = 1; j < shader->nr_cbufs; j++) {
+ memset(&output[i + j], 0, sizeof(struct r600_bc_output));
+ output[i + j].gpr = shader->output[i].gpr;
+ output[i + j].elem_size = 3;
+ output[i + j].swizzle_x = 0;
+ output[i + j].swizzle_y = 1;
+ output[i + j].swizzle_z = 2;
+ output[i + j].swizzle_w = 3;
+ output[i + j].burst_count = 1;
+ output[i + j].barrier = 1;
+ output[i + j].array_base = shader->output[i].sid + j;
+ output[i + j].inst = BC_INST(ctx.bc, V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT);
+ output[i + j].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL;
+ }
+ j--;
+ }
} else if (shader->output[i].name == TGSI_SEMANTIC_POSITION) {
- output[i].array_base = 61;
- output[i].swizzle_x = 2;
- output[i].swizzle_y = 7;
- output[i].swizzle_z = output[i].swizzle_w = 7;
- output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL;
+ output[i + j].array_base = 61;
+ output[i + j].swizzle_x = 2;
+ output[i + j].swizzle_y = 7;
+ output[i + j].swizzle_z = output[i + j].swizzle_w = 7;
+ output[i + j].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL;
} else if (shader->output[i].name == TGSI_SEMANTIC_STENCIL) {
- output[i].array_base = 61;
- output[i].swizzle_x = 7;
- output[i].swizzle_y = 1;
- output[i].swizzle_z = output[i].swizzle_w = 7;
- output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL;
+ output[i + j].array_base = 61;
+ output[i + j].swizzle_x = 7;
+ output[i + j].swizzle_y = 1;
+ output[i + j].swizzle_z = output[i + j].swizzle_w = 7;
+ output[i + j].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL;
} else {
R600_ERR("unsupported fragment output name %d\n", shader->output[i].name);
r = -EINVAL;
goto out_err;
}
}
+ noutput += j;
/* add fake param output for vertex shader if no param is exported */
if (ctx.type == TGSI_PROCESSOR_VERTEX) {
for (i = 0, pos0 = 0; i < noutput; i++) {
}
/* set export done on last export of each type */
for (i = noutput - 1, output_done = 0; i >= 0; i--) {
- if (ctx.bc->chiprev < CHIPREV_CAYMAN) {
+ if (ctx.bc->chip_class < CAYMAN) {
if (i == (noutput - 1)) {
output[i].end_of_program = 1;
}
goto out_err;
}
/* add program end */
- if (ctx.bc->chiprev == CHIPREV_CAYMAN)
+ if (ctx.bc->chip_class == CAYMAN)
cm_bc_add_cf_end(ctx.bc);
free(ctx.literals);
bc_src->value = shader_src->value[bc_src->chan];
}
+static void r600_bc_src_set_abs(struct r600_bc_alu_src *bc_src)
+{
+ bc_src->abs = 1;
+ bc_src->neg = 0;
+}
+
+static void r600_bc_src_toggle_neg(struct r600_bc_alu_src *bc_src)
+{
+ bc_src->neg = !bc_src->neg;
+}
+
static void tgsi_dst(struct r600_shader_ctx *ctx,
const struct tgsi_full_dst_register *tgsi_dst,
unsigned swizzle,
/* handle some special cases */
switch (ctx->inst_info->tgsi_opcode) {
case TGSI_OPCODE_SUB:
- alu.src[1].neg = 1;
+ r600_bc_src_toggle_neg(&alu.src[1]);
break;
case TGSI_OPCODE_ABS:
- alu.src[0].abs = 1;
- if (alu.src[0].neg)
- alu.src[0].neg = 0;
+ r600_bc_src_set_abs(&alu.src[0]);
break;
default:
break;
alu.src[2].sel = V_SQ_ALU_SRC_LITERAL;
alu.src[2].chan = 0;
- if (ctx->bc->chiprev == CHIPREV_R600) {
+ if (ctx->bc->chip_class == R600) {
alu.src[1].value = *(uint32_t *)&double_pi;
alu.src[2].value = *(uint32_t *)&neg_pi;
} else {
/* dst.x = COS */
if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_X) {
- if (ctx->bc->chiprev == CHIPREV_CAYMAN) {
+ if (ctx->bc->chip_class == CAYMAN) {
for (i = 0 ; i < 3; i++) {
memset(&alu, 0, sizeof(struct r600_bc_alu));
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS);
/* dst.y = SIN */
if (inst->Dst[0].Register.WriteMask & TGSI_WRITEMASK_Y) {
- if (ctx->bc->chiprev == CHIPREV_CAYMAN) {
+ if (ctx->bc->chip_class == CAYMAN) {
for (i = 0 ; i < 3; i++) {
memset(&alu, 0, sizeof(struct r600_bc_alu));
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN);
struct r600_bc_alu alu;
int r;
- /* dst.x, <- 1.0 */
- memset(&alu, 0, sizeof(struct r600_bc_alu));
- alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV);
- alu.src[0].sel = V_SQ_ALU_SRC_1; /*1.0*/
- alu.src[0].chan = 0;
- tgsi_dst(ctx, &inst->Dst[0], 0, &alu.dst);
- alu.dst.write = (inst->Dst[0].Register.WriteMask >> 0) & 1;
- r = r600_bc_add_alu(ctx->bc, &alu);
- if (r)
- return r;
-
- /* dst.y = max(src.x, 0.0) */
+ /* tmp.x = max(src.y, 0.0) */
memset(&alu, 0, sizeof(struct r600_bc_alu));
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX);
- r600_bc_src(&alu.src[0], &ctx->src[0], 0);
+ r600_bc_src(&alu.src[0], &ctx->src[0], 1);
alu.src[1].sel = V_SQ_ALU_SRC_0; /*0.0*/
- alu.src[1].chan = 0;
- tgsi_dst(ctx, &inst->Dst[0], 1, &alu.dst);
- alu.dst.write = (inst->Dst[0].Register.WriteMask >> 1) & 1;
- r = r600_bc_add_alu(ctx->bc, &alu);
- if (r)
- return r;
+ alu.src[1].chan = 1;
+
+ alu.dst.sel = ctx->temp_reg;
+ alu.dst.chan = 0;
+ alu.dst.write = 1;
- /* dst.w, <- 1.0 */
- memset(&alu, 0, sizeof(struct r600_bc_alu));
- alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV);
- alu.src[0].sel = V_SQ_ALU_SRC_1;
- alu.src[0].chan = 0;
- tgsi_dst(ctx, &inst->Dst[0], 3, &alu.dst);
- alu.dst.write = (inst->Dst[0].Register.WriteMask >> 3) & 1;
alu.last = 1;
r = r600_bc_add_alu(ctx->bc, &alu);
if (r)
int sel;
int i;
- if (ctx->bc->chiprev == CHIPREV_CAYMAN) {
+ if (ctx->bc->chip_class == CAYMAN) {
for (i = 0; i < 3; i++) {
- /* dst.z = log(src.y) */
+ /* tmp.z = log(tmp.x) */
memset(&alu, 0, sizeof(struct r600_bc_alu));
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_CLAMPED);
- r600_bc_src(&alu.src[0], &ctx->src[0], 1);
- tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
+ alu.src[0].sel = ctx->temp_reg;
+ alu.src[0].chan = 0;
+ alu.dst.sel = ctx->temp_reg;
+ alu.dst.chan = i;
if (i == 2) {
alu.dst.write = 1;
alu.last = 1;
return r;
}
} else {
- /* dst.z = log(src.y) */
+ /* tmp.z = log(tmp.x) */
memset(&alu, 0, sizeof(struct r600_bc_alu));
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_CLAMPED);
- r600_bc_src(&alu.src[0], &ctx->src[0], 1);
- tgsi_dst(ctx, &inst->Dst[0], 2, &alu.dst);
+ alu.src[0].sel = ctx->temp_reg;
+ alu.src[0].chan = 0;
+ alu.dst.sel = ctx->temp_reg;
+ alu.dst.chan = 2;
+ alu.dst.write = 1;
alu.last = 1;
r = r600_bc_add_alu(ctx->bc, &alu);
if (r)
chan = alu.dst.chan;
sel = alu.dst.sel;
- /* tmp.x = amd MUL_LIT(src.w, dst.z, src.x ) */
+ /* tmp.x = amd MUL_LIT(tmp.z, src.w, src.x ) */
memset(&alu, 0, sizeof(struct r600_bc_alu));
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT);
- r600_bc_src(&alu.src[0], &ctx->src[0], 3);
- alu.src[1].sel = sel;
- alu.src[1].chan = chan;
-
+ alu.src[0].sel = sel;
+ alu.src[0].chan = chan;
+ r600_bc_src(&alu.src[1], &ctx->src[0], 3);
r600_bc_src(&alu.src[2], &ctx->src[0], 0);
alu.dst.sel = ctx->temp_reg;
alu.dst.chan = 0;
if (r)
return r;
- if (ctx->bc->chiprev == CHIPREV_CAYMAN) {
+ if (ctx->bc->chip_class == CAYMAN) {
for (i = 0; i < 3; i++) {
/* dst.z = exp(tmp.x) */
memset(&alu, 0, sizeof(struct r600_bc_alu));
return r;
}
}
+
+ /* dst.x, <- 1.0 */
+ memset(&alu, 0, sizeof(struct r600_bc_alu));
+ alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV);
+ alu.src[0].sel = V_SQ_ALU_SRC_1; /*1.0*/
+ alu.src[0].chan = 0;
+ tgsi_dst(ctx, &inst->Dst[0], 0, &alu.dst);
+ alu.dst.write = (inst->Dst[0].Register.WriteMask >> 0) & 1;
+ r = r600_bc_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+
+ /* dst.y = max(src.x, 0.0) */
+ memset(&alu, 0, sizeof(struct r600_bc_alu));
+ alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX);
+ r600_bc_src(&alu.src[0], &ctx->src[0], 0);
+ alu.src[1].sel = V_SQ_ALU_SRC_0; /*0.0*/
+ alu.src[1].chan = 0;
+ tgsi_dst(ctx, &inst->Dst[0], 1, &alu.dst);
+ alu.dst.write = (inst->Dst[0].Register.WriteMask >> 1) & 1;
+ r = r600_bc_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+
+ /* dst.w, <- 1.0 */
+ memset(&alu, 0, sizeof(struct r600_bc_alu));
+ alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV);
+ alu.src[0].sel = V_SQ_ALU_SRC_1;
+ alu.src[0].chan = 0;
+ tgsi_dst(ctx, &inst->Dst[0], 3, &alu.dst);
+ alu.dst.write = (inst->Dst[0].Register.WriteMask >> 3) & 1;
+ alu.last = 1;
+ r = r600_bc_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+
return 0;
}
for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
r600_bc_src(&alu.src[i], &ctx->src[i], 0);
- alu.src[i].abs = 1;
+ r600_bc_src_set_abs(&alu.src[i]);
}
alu.dst.sel = ctx->temp_reg;
alu.dst.write = 1;
return 0;
}
+static inline boolean tgsi_tex_src_requires_loading(struct r600_shader_ctx *ctx,
+ unsigned index)
+{
+ struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+ return (inst->Src[index].Register.File != TGSI_FILE_TEMPORARY &&
+ inst->Src[index].Register.File != TGSI_FILE_INPUT) ||
+ ctx->src[index].neg || ctx->src[index].abs;
+}
+
+static inline unsigned tgsi_tex_get_src_gpr(struct r600_shader_ctx *ctx,
+ unsigned index)
+{
+ struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+ return ctx->file_offset[inst->Src[index].Register.File] + inst->Src[index].Register.Index;
+}
+
static int tgsi_tex(struct r600_shader_ctx *ctx)
{
static float one_point_five = 1.5f;
struct r600_bc_tex tex;
struct r600_bc_alu alu;
unsigned src_gpr;
- int r, i;
+ int r, i, j;
int opcode;
/* Texture fetch instructions can only use gprs as source.
* Also they cannot negate the source or take the absolute value */
- const boolean src_requires_loading =
- (inst->Src[0].Register.File != TGSI_FILE_TEMPORARY &&
- inst->Src[0].Register.File != TGSI_FILE_INPUT) ||
- ctx->src[0].neg || ctx->src[0].abs;
+ const boolean src_requires_loading = tgsi_tex_src_requires_loading(ctx, 0);
boolean src_loaded = FALSE;
+ unsigned sampler_src_reg = 1;
+
+ src_gpr = tgsi_tex_get_src_gpr(ctx, 0);
+
+ if (inst->Instruction.Opcode == TGSI_OPCODE_TXD) {
+ /* TGSI moves the sampler to src reg 3 for TXD */
+ sampler_src_reg = 3;
+
+ for (i = 1; i < 3; i++) {
+ /* set gradients h/v */
+ memset(&tex, 0, sizeof(struct r600_bc_tex));
+ tex.inst = (i == 1) ? SQ_TEX_INST_SET_GRADIENTS_H :
+ SQ_TEX_INST_SET_GRADIENTS_V;
+ tex.sampler_id = tgsi_tex_get_src_gpr(ctx, sampler_src_reg);
+ tex.resource_id = tex.sampler_id + R600_MAX_CONST_BUFFERS;
+
+ if (tgsi_tex_src_requires_loading(ctx, i)) {
+ tex.src_gpr = r600_get_temp(ctx);
+ tex.src_sel_x = 0;
+ tex.src_sel_y = 1;
+ tex.src_sel_z = 2;
+ tex.src_sel_w = 3;
+
+ for (j = 0; j < 4; j++) {
+ memset(&alu, 0, sizeof(struct r600_bc_alu));
+ alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV);
+ r600_bc_src(&alu.src[0], &ctx->src[i], j);
+ alu.dst.sel = tex.src_gpr;
+ alu.dst.chan = j;
+ if (j == 3)
+ alu.last = 1;
+ alu.dst.write = 1;
+ r = r600_bc_add_alu(ctx->bc, &alu);
+ if (r)
+ return r;
+ }
- src_gpr = ctx->file_offset[inst->Src[0].Register.File] + inst->Src[0].Register.Index;
-
- if (inst->Instruction.Opcode == TGSI_OPCODE_TXP) {
+ } else {
+ tex.src_gpr = tgsi_tex_get_src_gpr(ctx, i);
+ tex.src_sel_x = ctx->src[i].swizzle[0];
+ tex.src_sel_y = ctx->src[i].swizzle[1];
+ tex.src_sel_z = ctx->src[i].swizzle[2];
+ tex.src_sel_w = ctx->src[i].swizzle[3];
+ tex.src_rel = ctx->src[i].rel;
+ }
+ tex.dst_gpr = ctx->temp_reg; /* just to avoid confusing the asm scheduler */
+ tex.dst_sel_x = tex.dst_sel_y = tex.dst_sel_z = tex.dst_sel_w = 7;
+ if (inst->Texture.Texture != TGSI_TEXTURE_RECT) {
+ tex.coord_type_x = 1;
+ tex.coord_type_y = 1;
+ tex.coord_type_z = 1;
+ tex.coord_type_w = 1;
+ }
+ r = r600_bc_add_tex(ctx->bc, &tex);
+ if (r)
+ return r;
+ }
+ } else if (inst->Instruction.Opcode == TGSI_OPCODE_TXP) {
int out_chan;
/* Add perspective divide */
- if (ctx->bc->chiprev == CHIPREV_CAYMAN) {
+ if (ctx->bc->chip_class == CAYMAN) {
out_chan = 2;
for (i = 0; i < 3; i++) {
memset(&alu, 0, sizeof(struct r600_bc_alu));
}
/* tmp1.z = RCP_e(|tmp1.z|) */
- if (ctx->bc->chiprev == CHIPREV_CAYMAN) {
+ if (ctx->bc->chip_class == CAYMAN) {
for (i = 0; i < 3; i++) {
memset(&alu, 0, sizeof(struct r600_bc_alu));
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE);
}
opcode = ctx->inst_info->r600_opcode;
- if (opcode == SQ_TEX_INST_SAMPLE &&
- (inst->Texture.Texture == TGSI_TEXTURE_SHADOW1D || inst->Texture.Texture == TGSI_TEXTURE_SHADOW2D))
- opcode = SQ_TEX_INST_SAMPLE_C;
+ if (inst->Texture.Texture == TGSI_TEXTURE_SHADOW1D || inst->Texture.Texture == TGSI_TEXTURE_SHADOW2D) {
+ switch (opcode) {
+ case SQ_TEX_INST_SAMPLE:
+ opcode = SQ_TEX_INST_SAMPLE_C;
+ break;
+ case SQ_TEX_INST_SAMPLE_L:
+ opcode = SQ_TEX_INST_SAMPLE_C_L;
+ break;
+ case SQ_TEX_INST_SAMPLE_G:
+ opcode = SQ_TEX_INST_SAMPLE_C_G;
+ break;
+ }
+ }
memset(&tex, 0, sizeof(struct r600_bc_tex));
tex.inst = opcode;
- tex.sampler_id = ctx->file_offset[inst->Src[1].Register.File] + inst->Src[1].Register.Index;
+
+ tex.sampler_id = tgsi_tex_get_src_gpr(ctx, sampler_src_reg);
tex.resource_id = tex.sampler_id + R600_MAX_CONST_BUFFERS;
tex.src_gpr = src_gpr;
tex.dst_gpr = ctx->file_offset[inst->Dst[0].Register.File] + inst->Dst[0].Register.Index;
alu.src[0].sel = V_SQ_ALU_SRC_1;
alu.src[0].chan = 0;
r600_bc_src(&alu.src[1], &ctx->src[0], i);
- alu.src[1].neg = 1;
+ r600_bc_src_toggle_neg(&alu.src[1]);
alu.dst.sel = ctx->temp_reg;
alu.dst.chan = i;
if (i == lasti) {
if (r)
return r;
- if (ctx->bc->chiprev == CHIPREV_CAYMAN) {
+ if (ctx->bc->chip_class == CAYMAN) {
for (i = 0; i < 3; i++) {
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE);
alu.src[0].sel = ctx->temp_reg;
/* result.z = RoughApprox2ToX(tmp);*/
if ((inst->Dst[0].Register.WriteMask >> 2) & 0x1) {
- if (ctx->bc->chiprev == CHIPREV_CAYMAN) {
+ if (ctx->bc->chip_class == CAYMAN) {
for (i = 0; i < 3; i++) {
memset(&alu, 0, sizeof(struct r600_bc_alu));
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE);
int r;
int i;
- /* result.x = floor(log2(src)); */
+ /* result.x = floor(log2(|src|)); */
if (inst->Dst[0].Register.WriteMask & 1) {
- if (ctx->bc->chiprev == CHIPREV_CAYMAN) {
+ if (ctx->bc->chip_class == CAYMAN) {
for (i = 0; i < 3; i++) {
memset(&alu, 0, sizeof(struct r600_bc_alu));
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE);
r600_bc_src(&alu.src[0], &ctx->src[0], 0);
+ r600_bc_src_set_abs(&alu.src[0]);
alu.dst.sel = ctx->temp_reg;
alu.dst.chan = i;
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE);
r600_bc_src(&alu.src[0], &ctx->src[0], 0);
+ r600_bc_src_set_abs(&alu.src[0]);
alu.dst.sel = ctx->temp_reg;
alu.dst.chan = 0;
return r;
}
- /* result.y = src.x / (2 ^ floor(log2(src.x))); */
+ /* result.y = |src.x| / (2 ^ floor(log2(|src.x|))); */
if ((inst->Dst[0].Register.WriteMask >> 1) & 1) {
- if (ctx->bc->chiprev == CHIPREV_CAYMAN) {
+ if (ctx->bc->chip_class == CAYMAN) {
for (i = 0; i < 3; i++) {
memset(&alu, 0, sizeof(struct r600_bc_alu));
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE);
r600_bc_src(&alu.src[0], &ctx->src[0], 0);
+ r600_bc_src_set_abs(&alu.src[0]);
alu.dst.sel = ctx->temp_reg;
alu.dst.chan = i;
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE);
r600_bc_src(&alu.src[0], &ctx->src[0], 0);
+ r600_bc_src_set_abs(&alu.src[0]);
alu.dst.sel = ctx->temp_reg;
alu.dst.chan = 1;
if (r)
return r;
- if (ctx->bc->chiprev == CHIPREV_CAYMAN) {
+ if (ctx->bc->chip_class == CAYMAN) {
for (i = 0; i < 3; i++) {
memset(&alu, 0, sizeof(struct r600_bc_alu));
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE);
return r;
}
- if (ctx->bc->chiprev == CHIPREV_CAYMAN) {
+ if (ctx->bc->chip_class == CAYMAN) {
for (i = 0; i < 3; i++) {
memset(&alu, 0, sizeof(struct r600_bc_alu));
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE);
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL);
r600_bc_src(&alu.src[0], &ctx->src[0], 0);
+ r600_bc_src_set_abs(&alu.src[0]);
alu.src[1].sel = ctx->temp_reg;
alu.src[1].chan = 1;
return r;
}
- /* result.z = log2(src);*/
+ /* result.z = log2(|src|);*/
if ((inst->Dst[0].Register.WriteMask >> 2) & 1) {
- if (ctx->bc->chiprev == CHIPREV_CAYMAN) {
+ if (ctx->bc->chip_class == CAYMAN) {
for (i = 0; i < 3; i++) {
memset(&alu, 0, sizeof(struct r600_bc_alu));
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE);
r600_bc_src(&alu.src[0], &ctx->src[0], 0);
+ r600_bc_src_set_abs(&alu.src[0]);
alu.dst.sel = ctx->temp_reg;
if (i == 2)
alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE);
r600_bc_src(&alu.src[0], &ctx->src[0], 0);
+ r600_bc_src_set_abs(&alu.src[0]);
alu.dst.sel = ctx->temp_reg;
alu.dst.write = 1;
{TGSI_OPCODE_SNE, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETNE, tgsi_op2},
{TGSI_OPCODE_STR, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
{TGSI_OPCODE_TEX, 0, SQ_TEX_INST_SAMPLE, tgsi_tex},
- {TGSI_OPCODE_TXD, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+ {TGSI_OPCODE_TXD, 0, SQ_TEX_INST_SAMPLE_G, tgsi_tex},
{TGSI_OPCODE_TXP, 0, SQ_TEX_INST_SAMPLE, tgsi_tex},
{TGSI_OPCODE_UP2H, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
{TGSI_OPCODE_UP2US, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
{TGSI_OPCODE_MOV, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV, tgsi_op2},
{TGSI_OPCODE_LIT, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_lit},
{TGSI_OPCODE_RCP, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE, tgsi_trans_srcx_replicate},
- {TGSI_OPCODE_RSQ, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_IEEE, tgsi_trans_srcx_replicate},
+ {TGSI_OPCODE_RSQ, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_IEEE, tgsi_rsq},
{TGSI_OPCODE_EXP, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_exp},
{TGSI_OPCODE_LOG, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_log},
{TGSI_OPCODE_MUL, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL, tgsi_op2},
{TGSI_OPCODE_SNE, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETNE, tgsi_op2},
{TGSI_OPCODE_STR, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
{TGSI_OPCODE_TEX, 0, SQ_TEX_INST_SAMPLE, tgsi_tex},
- {TGSI_OPCODE_TXD, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+ {TGSI_OPCODE_TXD, 0, SQ_TEX_INST_SAMPLE_G, tgsi_tex},
{TGSI_OPCODE_TXP, 0, SQ_TEX_INST_SAMPLE, tgsi_tex},
{TGSI_OPCODE_UP2H, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
{TGSI_OPCODE_UP2US, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
{TGSI_OPCODE_SNE, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETNE, tgsi_op2},
{TGSI_OPCODE_STR, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
{TGSI_OPCODE_TEX, 0, SQ_TEX_INST_SAMPLE, tgsi_tex},
- {TGSI_OPCODE_TXD, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+ {TGSI_OPCODE_TXD, 0, SQ_TEX_INST_SAMPLE_G, tgsi_tex},
{TGSI_OPCODE_TXP, 0, SQ_TEX_INST_SAMPLE, tgsi_tex},
{TGSI_OPCODE_UP2H, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
{TGSI_OPCODE_UP2US, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},