#include "nir/nir_vla.h"
#include "nir/nir_control_flow.h"
#include "nir/nir_constant_expressions.h"
+#include "spirv_info.h"
+
+void
+_vtn_warn(const char *file, int line, const char *msg, ...)
+{
+ char *formatted;
+ va_list args;
+
+ va_start(args, msg);
+ formatted = ralloc_vasprintf(NULL, msg, args);
+ va_end(args);
+
+ fprintf(stderr, "%s:%d WARNING: %s\n", file, line, formatted);
+
+ ralloc_free(formatted);
+}
static struct vtn_ssa_value *
vtn_undef_ssa_value(struct vtn_builder *b, const struct glsl_type *type)
nir_load_const_instr *load =
nir_load_const_instr_create(b->shader, num_components, 32);
- for (unsigned i = 0; i < num_components; i++)
- load->value.u32[i] = constant->value.u[i];
+ load->value = constant->values[0];
nir_instr_insert_before_cf_list(&b->impl->body, &load->instr);
val->def = &load->def;
nir_load_const_instr *load =
nir_load_const_instr_create(b->shader, rows, 32);
- for (unsigned j = 0; j < rows; j++)
- load->value.u32[j] = constant->value.u[rows * i + j];
+ load->value = constant->values[i];
nir_instr_insert_before_cf_list(&b->impl->body, &load->instr);
col_val->def = &load->def;
case SpvDecorationUniform:
break; /* FIXME: Do nothing with this for now. */
case SpvDecorationNoPerspective:
- ctx->fields[member].interpolation = INTERP_QUALIFIER_NOPERSPECTIVE;
+ ctx->fields[member].interpolation = INTERP_MODE_NOPERSPECTIVE;
break;
case SpvDecorationFlat:
- ctx->fields[member].interpolation = INTERP_QUALIFIER_FLAT;
+ ctx->fields[member].interpolation = INTERP_MODE_FLAT;
break;
case SpvDecorationCentroid:
ctx->fields[member].centroid = true;
break;
case SpvDecorationPatch:
- unreachable("Tessellation not yet supported");
+ vtn_warn("Tessellation not yet supported");
+ break;
case SpvDecorationSpecId:
case SpvDecorationBlock:
case SpvDecorationIndex:
case SpvDecorationBinding:
case SpvDecorationDescriptorSet:
+ case SpvDecorationLinkageAttributes:
case SpvDecorationNoContraction:
case SpvDecorationInputAttachmentIndex:
- unreachable("Decoration not allowed on struct members");
+ vtn_warn("Decoration not allowed on struct members: %s",
+ spirv_decoration_to_string(dec->decoration));
+ break;
case SpvDecorationXfbBuffer:
case SpvDecorationXfbStride:
- unreachable("Vulkan does not have transform feedback");
+ vtn_warn("Vulkan does not have transform feedback");
+ break;
case SpvDecorationCPacked:
case SpvDecorationSaturatedConversion:
case SpvDecorationFPRoundingMode:
case SpvDecorationFPFastMathMode:
case SpvDecorationAlignment:
- unreachable("Decoraiton only allowed for CL-style kernels");
-
- default:
- unreachable("Unhandled member decoration");
+ vtn_warn("Decoraiton only allowed for CL-style kernels: %s",
+ spirv_decoration_to_string(dec->decoration));
+ break;
}
}
case SpvDecorationOffset:
case SpvDecorationXfbBuffer:
case SpvDecorationXfbStride:
- unreachable("Decoraiton only allowed for struct members");
+ vtn_warn("Decoraiton only allowed for struct members: %s",
+ spirv_decoration_to_string(dec->decoration));
+ break;
case SpvDecorationRelaxedPrecision:
case SpvDecorationSpecId:
case SpvDecorationLinkageAttributes:
case SpvDecorationNoContraction:
case SpvDecorationInputAttachmentIndex:
- unreachable("Decoraiton not allowed on types");
+ vtn_warn("Decoraiton not allowed on types: %s",
+ spirv_decoration_to_string(dec->decoration));
+ break;
case SpvDecorationCPacked:
case SpvDecorationSaturatedConversion:
case SpvDecorationFPRoundingMode:
case SpvDecorationFPFastMathMode:
case SpvDecorationAlignment:
- unreachable("Decoraiton only allowed for CL-style kernels");
+ vtn_warn("Decoraiton only allowed for CL-style kernels: %s",
+ spirv_decoration_to_string(dec->decoration));
+ break;
}
}
val->type->type = (signedness ? glsl_int_type() : glsl_uint_type());
break;
}
- case SpvOpTypeFloat:
- val->type->type = glsl_float_type();
+ case SpvOpTypeFloat: {
+ int bit_size = w[2];
+ val->type->type = bit_size == 64 ? glsl_double_type() : glsl_float_type();
break;
+ }
case SpvOpTypeVector: {
struct vtn_type *base = vtn_value(b, w[2], vtn_value_type_type)->type;
length = 0;
} else {
length =
- vtn_value(b, w[3], vtn_value_type_constant)->constant->value.u[0];
+ vtn_value(b, w[3], vtn_value_type_constant)->constant->values[0].u32[0];
}
val->type->type = glsl_array_type(array_element->type, length);
case SpvDimCube: dim = GLSL_SAMPLER_DIM_CUBE; break;
case SpvDimRect: dim = GLSL_SAMPLER_DIM_RECT; break;
case SpvDimBuffer: dim = GLSL_SAMPLER_DIM_BUF; break;
+ case SpvDimSubpassData: dim = GLSL_SAMPLER_DIM_SUBPASS; break;
default:
unreachable("Invalid SPIR-V Sampler dimension");
}
val->type->type = glsl_sampler_type(dim, is_shadow, is_array,
glsl_get_base_type(sampled_type));
} else if (sampled == 2) {
- assert(format);
+ assert((dim == GLSL_SAMPLER_DIM_SUBPASS) || format);
assert(!is_shadow);
val->type->type = glsl_image_type(dim, is_array,
glsl_get_base_type(sampled_type));
}
static void
-spec_constant_deocoration_cb(struct vtn_builder *b, struct vtn_value *v,
+spec_constant_decoration_cb(struct vtn_builder *b, struct vtn_value *v,
int member, const struct vtn_decoration *dec,
void *data)
{
get_specialization(struct vtn_builder *b, struct vtn_value *val,
uint32_t const_value)
{
- vtn_foreach_decoration(b, val, spec_constant_deocoration_cb, &const_value);
+ vtn_foreach_decoration(b, val, spec_constant_decoration_cb, &const_value);
return const_value;
}
assert(val->const_type == glsl_vector_type(GLSL_TYPE_UINT, 3));
- b->shader->info.cs.local_size[0] = val->constant->value.u[0];
- b->shader->info.cs.local_size[1] = val->constant->value.u[1];
- b->shader->info.cs.local_size[2] = val->constant->value.u[2];
+ b->shader->info->cs.local_size[0] = val->constant->values[0].u32[0];
+ b->shader->info->cs.local_size[1] = val->constant->values[0].u32[1];
+ b->shader->info->cs.local_size[2] = val->constant->values[0].u32[2];
}
static void
switch (opcode) {
case SpvOpConstantTrue:
assert(val->const_type == glsl_bool_type());
- val->constant->value.u[0] = NIR_TRUE;
+ val->constant->values[0].u32[0] = NIR_TRUE;
break;
case SpvOpConstantFalse:
assert(val->const_type == glsl_bool_type());
- val->constant->value.u[0] = NIR_FALSE;
+ val->constant->values[0].u32[0] = NIR_FALSE;
break;
case SpvOpSpecConstantTrue:
assert(val->const_type == glsl_bool_type());
uint32_t int_val =
get_specialization(b, val, (opcode == SpvOpSpecConstantTrue));
- val->constant->value.u[0] = int_val ? NIR_TRUE : NIR_FALSE;
+ val->constant->values[0].u32[0] = int_val ? NIR_TRUE : NIR_FALSE;
break;
}
case SpvOpConstant:
assert(glsl_type_is_scalar(val->const_type));
- val->constant->value.u[0] = w[3];
+ val->constant->values[0].u32[0] = w[3];
break;
case SpvOpSpecConstant:
assert(glsl_type_is_scalar(val->const_type));
- val->constant->value.u[0] = get_specialization(b, val, w[3]);
+ val->constant->values[0].u32[0] = get_specialization(b, val, w[3]);
break;
case SpvOpSpecConstantComposite:
case SpvOpConstantComposite: {
case GLSL_TYPE_FLOAT:
case GLSL_TYPE_BOOL:
if (glsl_type_is_matrix(val->const_type)) {
- unsigned rows = glsl_get_vector_elements(val->const_type);
assert(glsl_get_matrix_columns(val->const_type) == elem_count);
for (unsigned i = 0; i < elem_count; i++)
- for (unsigned j = 0; j < rows; j++)
- val->constant->value.u[rows * i + j] = elems[i]->value.u[j];
+ val->constant->values[i] = elems[i]->values[0];
} else {
assert(glsl_type_is_vector(val->const_type));
assert(glsl_get_vector_elements(val->const_type) == elem_count);
for (unsigned i = 0; i < elem_count; i++)
- val->constant->value.u[i] = elems[i]->value.u[0];
+ val->constant->values[0].u32[i] = elems[i]->values[0].u32[0];
}
ralloc_free(elems);
break;
uint32_t u[8];
for (unsigned i = 0; i < len0; i++)
- u[i] = v0->constant->value.u[i];
+ u[i] = v0->constant->values[0].u32[i];
for (unsigned i = 0; i < len1; i++)
- u[len0 + i] = v1->constant->value.u[i];
+ u[len0 + i] = v1->constant->values[0].u32[i];
for (unsigned i = 0; i < count - 6; i++) {
uint32_t comp = w[i + 6];
if (comp == (uint32_t)-1) {
- val->constant->value.u[i] = 0xdeadbeef;
+ val->constant->values[0].u32[i] = 0xdeadbeef;
} else {
- val->constant->value.u[i] = u[comp];
+ val->constant->values[0].u32[i] = u[comp];
}
}
break;
}
int elem = -1;
+ int col = 0;
const struct glsl_type *type = comp->const_type;
for (unsigned i = deref_start; i < count; i++) {
switch (glsl_get_base_type(type)) {
case GLSL_TYPE_FLOAT:
case GLSL_TYPE_BOOL:
/* If we hit this granularity, we're picking off an element */
- if (elem < 0)
- elem = 0;
-
if (glsl_type_is_matrix(type)) {
- elem += w[i] * glsl_get_vector_elements(type);
+ assert(col == 0 && elem == -1);
+ col = w[i];
+ elem = 0;
type = glsl_get_column_type(type);
} else {
- assert(glsl_type_is_vector(type));
- elem += w[i];
+ assert(elem <= 0 && glsl_type_is_vector(type));
+ elem = w[i];
type = glsl_scalar_type(glsl_get_base_type(type));
}
continue;
} else {
unsigned num_components = glsl_get_vector_elements(type);
for (unsigned i = 0; i < num_components; i++)
- val->constant->value.u[i] = (*c)->value.u[elem + i];
+ val->constant->values[0].u32[i] = (*c)->values[col].u32[elem + i];
}
} else {
struct vtn_value *insert =
} else {
unsigned num_components = glsl_get_vector_elements(type);
for (unsigned i = 0; i < num_components; i++)
- (*c)->value.u[elem + i] = insert->constant->value.u[i];
+ (*c)->values[col].u32[elem + i] = insert->constant->values[0].u32[i];
}
}
break;
unsigned j = swap ? 1 - i : i;
assert(bit_size == 32);
- for (unsigned k = 0; k < num_components; k++)
- src[j].u32[k] = c->value.u[k];
+ src[j] = c->values[0];
}
- nir_const_value res = nir_eval_const_opcode(op, num_components,
- bit_size, src);
-
- for (unsigned k = 0; k < num_components; k++)
- val->constant->value.u[k] = res.u32[k];
-
+ val->constant->values[0] =
+ nir_eval_const_opcode(op, num_components, bit_size, src);
break;
} /* default */
}
struct vtn_value *arg = vtn_untyped_value(b, arg_id);
if (arg->value_type == vtn_value_type_access_chain) {
nir_deref_var *d = vtn_access_chain_to_deref(b, arg->access_chain);
- call->params[i] = nir_deref_as_var(nir_copy_deref(call, &d->deref));
+ call->params[i] = nir_deref_var_clone(d, call);
} else {
struct vtn_ssa_value *arg_ssa = vtn_ssa_value(b, arg_id);
} else {
image_type = sampled.sampler->var->var->interface_type;
}
-
- nir_tex_src srcs[8]; /* 8 should be enough */
- nir_tex_src *p = srcs;
-
- unsigned idx = 4;
-
- bool has_coord = false;
- switch (opcode) {
- case SpvOpImageSampleImplicitLod:
- case SpvOpImageSampleExplicitLod:
- case SpvOpImageSampleDrefImplicitLod:
- case SpvOpImageSampleDrefExplicitLod:
- case SpvOpImageSampleProjImplicitLod:
- case SpvOpImageSampleProjExplicitLod:
- case SpvOpImageSampleProjDrefImplicitLod:
- case SpvOpImageSampleProjDrefExplicitLod:
- case SpvOpImageFetch:
- case SpvOpImageGather:
- case SpvOpImageDrefGather:
- case SpvOpImageQueryLod: {
- /* All these types have the coordinate as their first real argument */
- struct vtn_ssa_value *coord = vtn_ssa_value(b, w[idx++]);
- has_coord = true;
- p->src = nir_src_for_ssa(coord->def);
- p->src_type = nir_tex_src_coord;
- p++;
- break;
- }
-
- default:
- break;
- }
-
- /* These all have an explicit depth value as their next source */
- switch (opcode) {
- case SpvOpImageSampleDrefImplicitLod:
- case SpvOpImageSampleDrefExplicitLod:
- case SpvOpImageSampleProjDrefImplicitLod:
- case SpvOpImageSampleProjDrefExplicitLod:
- (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_comparitor);
- break;
- default:
- break;
- }
-
- /* For OpImageQuerySizeLod, we always have an LOD */
- if (opcode == SpvOpImageQuerySizeLod)
- (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_lod);
+ const enum glsl_sampler_dim sampler_dim = glsl_get_sampler_dim(image_type);
+ const bool is_array = glsl_sampler_type_is_array(image_type);
+ const bool is_shadow = glsl_sampler_type_is_shadow(image_type);
/* Figure out the base texture operation */
nir_texop texop;
break;
case SpvOpImageQuerySamples:
+ texop = nir_texop_texture_samples;
+ break;
+
default:
unreachable("Unhandled opcode");
}
+ nir_tex_src srcs[8]; /* 8 should be enough */
+ nir_tex_src *p = srcs;
+
+ unsigned idx = 4;
+
+ struct nir_ssa_def *coord;
+ unsigned coord_components;
+ switch (opcode) {
+ case SpvOpImageSampleImplicitLod:
+ case SpvOpImageSampleExplicitLod:
+ case SpvOpImageSampleDrefImplicitLod:
+ case SpvOpImageSampleDrefExplicitLod:
+ case SpvOpImageSampleProjImplicitLod:
+ case SpvOpImageSampleProjExplicitLod:
+ case SpvOpImageSampleProjDrefImplicitLod:
+ case SpvOpImageSampleProjDrefExplicitLod:
+ case SpvOpImageFetch:
+ case SpvOpImageGather:
+ case SpvOpImageDrefGather:
+ case SpvOpImageQueryLod: {
+ /* All these types have the coordinate as their first real argument */
+ switch (sampler_dim) {
+ case GLSL_SAMPLER_DIM_1D:
+ case GLSL_SAMPLER_DIM_BUF:
+ coord_components = 1;
+ break;
+ case GLSL_SAMPLER_DIM_2D:
+ case GLSL_SAMPLER_DIM_RECT:
+ case GLSL_SAMPLER_DIM_MS:
+ coord_components = 2;
+ break;
+ case GLSL_SAMPLER_DIM_3D:
+ case GLSL_SAMPLER_DIM_CUBE:
+ coord_components = 3;
+ break;
+ default:
+ unreachable("Invalid sampler type");
+ }
+
+ if (is_array && texop != nir_texop_lod)
+ coord_components++;
+
+ coord = vtn_ssa_value(b, w[idx++])->def;
+ p->src = nir_src_for_ssa(coord);
+ p->src_type = nir_tex_src_coord;
+ p++;
+ break;
+ }
+
+ default:
+ coord = NULL;
+ coord_components = 0;
+ break;
+ }
+
+ switch (opcode) {
+ case SpvOpImageSampleProjImplicitLod:
+ case SpvOpImageSampleProjExplicitLod:
+ case SpvOpImageSampleProjDrefImplicitLod:
+ case SpvOpImageSampleProjDrefExplicitLod:
+ /* These have the projector as the last coordinate component */
+ p->src = nir_src_for_ssa(nir_channel(&b->nb, coord, coord_components));
+ p->src_type = nir_tex_src_projector;
+ p++;
+ break;
+
+ default:
+ break;
+ }
+
+ unsigned gather_component = 0;
+ switch (opcode) {
+ case SpvOpImageSampleDrefImplicitLod:
+ case SpvOpImageSampleDrefExplicitLod:
+ case SpvOpImageSampleProjDrefImplicitLod:
+ case SpvOpImageSampleProjDrefExplicitLod:
+ case SpvOpImageDrefGather:
+ /* These all have an explicit depth value as their next source */
+ (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_comparator);
+ break;
+
+ case SpvOpImageGather:
+ /* This has a component as its next source */
+ gather_component =
+ vtn_value(b, w[idx++], vtn_value_type_constant)->constant->values[0].u32[0];
+ break;
+
+ default:
+ break;
+ }
+
+ /* For OpImageQuerySizeLod, we always have an LOD */
+ if (opcode == SpvOpImageQuerySizeLod)
+ (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_lod);
+
/* Now we need to handle some number of optional arguments */
+ const struct vtn_ssa_value *gather_offsets = NULL;
if (idx < count) {
uint32_t operands = w[idx++];
if (operands & SpvImageOperandsLodMask) {
assert(texop == nir_texop_txl || texop == nir_texop_txf ||
- texop == nir_texop_txf_ms || texop == nir_texop_txs);
+ texop == nir_texop_txs);
(*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_lod);
}
if (operands & SpvImageOperandsGradMask) {
- assert(texop == nir_texop_tex);
+ assert(texop == nir_texop_txl);
texop = nir_texop_txd;
(*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_ddx);
(*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_ddy);
operands & SpvImageOperandsConstOffsetMask)
(*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_offset);
- if (operands & SpvImageOperandsConstOffsetsMask)
- assert(!"Constant offsets to texture gather not yet implemented");
+ if (operands & SpvImageOperandsConstOffsetsMask) {
+ gather_offsets = vtn_ssa_value(b, w[idx++]);
+ (*p++) = (nir_tex_src){};
+ }
if (operands & SpvImageOperandsSampleMask) {
assert(texop == nir_texop_txf_ms);
memcpy(instr->src, srcs, instr->num_srcs * sizeof(*instr->src));
- instr->sampler_dim = glsl_get_sampler_dim(image_type);
- instr->is_array = glsl_sampler_type_is_array(image_type);
- instr->is_shadow = glsl_sampler_type_is_shadow(image_type);
- instr->is_new_style_shadow = instr->is_shadow;
-
- if (has_coord) {
- switch (instr->sampler_dim) {
- case GLSL_SAMPLER_DIM_1D:
- case GLSL_SAMPLER_DIM_BUF:
- instr->coord_components = 1;
- break;
- case GLSL_SAMPLER_DIM_2D:
- case GLSL_SAMPLER_DIM_RECT:
- case GLSL_SAMPLER_DIM_MS:
- instr->coord_components = 2;
- break;
- case GLSL_SAMPLER_DIM_3D:
- case GLSL_SAMPLER_DIM_CUBE:
- instr->coord_components = 3;
- break;
- default:
- assert("Invalid sampler type");
- }
-
- if (instr->is_array)
- instr->coord_components++;
- } else {
- instr->coord_components = 0;
- }
+ instr->coord_components = coord_components;
+ instr->sampler_dim = sampler_dim;
+ instr->is_array = is_array;
+ instr->is_shadow = is_shadow;
+ instr->is_new_style_shadow =
+ is_shadow && glsl_get_components(ret_type->type) == 1;
+ instr->component = gather_component;
switch (glsl_get_sampler_result_type(image_type)) {
case GLSL_TYPE_FLOAT: instr->dest_type = nir_type_float; break;
}
nir_deref_var *sampler = vtn_access_chain_to_deref(b, sampled.sampler);
+ nir_deref_var *texture;
if (sampled.image) {
nir_deref_var *image = vtn_access_chain_to_deref(b, sampled.image);
- instr->texture = nir_deref_as_var(nir_copy_deref(instr, &image->deref));
+ texture = image;
} else {
- instr->texture = nir_deref_as_var(nir_copy_deref(instr, &sampler->deref));
+ texture = sampler;
}
+ instr->texture = nir_deref_var_clone(texture, instr);
+
switch (instr->op) {
case nir_texop_tex:
case nir_texop_txb:
case nir_texop_txl:
case nir_texop_txd:
/* These operations require a sampler */
- instr->sampler = nir_deref_as_var(nir_copy_deref(instr, &sampler->deref));
+ instr->sampler = nir_deref_var_clone(sampler, instr);
break;
case nir_texop_txf:
case nir_texop_txf_ms:
assert(glsl_get_vector_elements(ret_type->type) ==
nir_tex_instr_dest_size(instr));
+ nir_ssa_def *def;
+ nir_instr *instruction;
+ if (gather_offsets) {
+ assert(glsl_get_base_type(gather_offsets->type) == GLSL_TYPE_ARRAY);
+ assert(glsl_get_length(gather_offsets->type) == 4);
+ nir_tex_instr *instrs[4] = {instr, NULL, NULL, NULL};
+
+ /* Copy the current instruction 4x */
+ for (uint32_t i = 1; i < 4; i++) {
+ instrs[i] = nir_tex_instr_create(b->shader, instr->num_srcs);
+ instrs[i]->op = instr->op;
+ instrs[i]->coord_components = instr->coord_components;
+ instrs[i]->sampler_dim = instr->sampler_dim;
+ instrs[i]->is_array = instr->is_array;
+ instrs[i]->is_shadow = instr->is_shadow;
+ instrs[i]->is_new_style_shadow = instr->is_new_style_shadow;
+ instrs[i]->component = instr->component;
+ instrs[i]->dest_type = instr->dest_type;
+ instrs[i]->texture = nir_deref_var_clone(texture, instrs[i]);
+ instrs[i]->sampler = NULL;
+
+ memcpy(instrs[i]->src, srcs, instr->num_srcs * sizeof(*instr->src));
+
+ nir_ssa_dest_init(&instrs[i]->instr, &instrs[i]->dest,
+ nir_tex_instr_dest_size(instr), 32, NULL);
+ }
+
+ /* Fill in the last argument with the offset from the passed in offsets
+ * and insert the instruction into the stream.
+ */
+ for (uint32_t i = 0; i < 4; i++) {
+ nir_tex_src src;
+ src.src = nir_src_for_ssa(gather_offsets->elems[i]->def);
+ src.src_type = nir_tex_src_offset;
+ instrs[i]->src[instrs[i]->num_srcs - 1] = src;
+ nir_builder_instr_insert(&b->nb, &instrs[i]->instr);
+ }
+
+ /* Combine the results of the 4 instructions by taking their .w
+ * components
+ */
+ nir_alu_instr *vec4 = nir_alu_instr_create(b->shader, nir_op_vec4);
+ nir_ssa_dest_init(&vec4->instr, &vec4->dest.dest, 4, 32, NULL);
+ vec4->dest.write_mask = 0xf;
+ for (uint32_t i = 0; i < 4; i++) {
+ vec4->src[i].src = nir_src_for_ssa(&instrs[i]->dest.ssa);
+ vec4->src[i].swizzle[0] = 3;
+ }
+ def = &vec4->dest.dest.ssa;
+ instruction = &vec4->instr;
+ } else {
+ def = &instr->dest.ssa;
+ instruction = &instr->instr;
+ }
+
val->ssa = vtn_create_ssa_value(b, ret_type->type);
- val->ssa->def = &instr->dest.ssa;
+ val->ssa->def = def;
+
+ nir_builder_instr_insert(&b->nb, instruction);
+}
+
+static void
+fill_common_atomic_sources(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, nir_src *src)
+{
+ switch (opcode) {
+ case SpvOpAtomicIIncrement:
+ src[0] = nir_src_for_ssa(nir_imm_int(&b->nb, 1));
+ break;
+
+ case SpvOpAtomicIDecrement:
+ src[0] = nir_src_for_ssa(nir_imm_int(&b->nb, -1));
+ break;
+
+ case SpvOpAtomicISub:
+ src[0] =
+ nir_src_for_ssa(nir_ineg(&b->nb, vtn_ssa_value(b, w[6])->def));
+ break;
- nir_builder_instr_insert(&b->nb, &instr->instr);
+ case SpvOpAtomicCompareExchange:
+ src[0] = nir_src_for_ssa(vtn_ssa_value(b, w[8])->def);
+ src[1] = nir_src_for_ssa(vtn_ssa_value(b, w[7])->def);
+ break;
+
+ case SpvOpAtomicExchange:
+ case SpvOpAtomicIAdd:
+ case SpvOpAtomicSMin:
+ case SpvOpAtomicUMin:
+ case SpvOpAtomicSMax:
+ case SpvOpAtomicUMax:
+ case SpvOpAtomicAnd:
+ case SpvOpAtomicOr:
+ case SpvOpAtomicXor:
+ src[0] = nir_src_for_ssa(vtn_ssa_value(b, w[6])->def);
+ break;
+
+ default:
+ unreachable("Invalid SPIR-V atomic");
+ }
}
static nir_ssa_def *
case SpvOpAtomicIDecrement:
case SpvOpAtomicIAdd:
case SpvOpAtomicISub:
+ case SpvOpAtomicLoad:
case SpvOpAtomicSMin:
case SpvOpAtomicUMin:
case SpvOpAtomicSMax:
image = *vtn_value(b, w[3], vtn_value_type_image_pointer)->image;
break;
+ case SpvOpAtomicStore:
+ image = *vtn_value(b, w[1], vtn_value_type_image_pointer)->image;
+ break;
+
case SpvOpImageQuerySize:
image.image =
vtn_value(b, w[3], vtn_value_type_access_chain)->access_chain;
OP(ImageQuerySize, size)
OP(ImageRead, load)
OP(ImageWrite, store)
+ OP(AtomicLoad, load)
+ OP(AtomicStore, store)
OP(AtomicExchange, atomic_exchange)
OP(AtomicCompareExchange, atomic_comp_swap)
OP(AtomicIIncrement, atomic_add)
nir_intrinsic_instr *intrin = nir_intrinsic_instr_create(b->shader, op);
nir_deref_var *image_deref = vtn_access_chain_to_deref(b, image.image);
- intrin->variables[0] =
- nir_deref_as_var(nir_copy_deref(&intrin->instr, &image_deref->deref));
+ intrin->variables[0] = nir_deref_var_clone(image_deref, intrin);
/* ImageQuerySize doesn't take any extra parameters */
if (opcode != SpvOpImageQuerySize) {
}
switch (opcode) {
+ case SpvOpAtomicLoad:
case SpvOpImageQuerySize:
case SpvOpImageRead:
break;
+ case SpvOpAtomicStore:
+ intrin->src[2] = nir_src_for_ssa(vtn_ssa_value(b, w[4])->def);
+ break;
case SpvOpImageWrite:
intrin->src[2] = nir_src_for_ssa(vtn_ssa_value(b, w[3])->def);
break;
+
case SpvOpAtomicIIncrement:
- intrin->src[2] = nir_src_for_ssa(nir_imm_int(&b->nb, 1));
- break;
case SpvOpAtomicIDecrement:
- intrin->src[2] = nir_src_for_ssa(nir_imm_int(&b->nb, -1));
- break;
-
case SpvOpAtomicExchange:
case SpvOpAtomicIAdd:
case SpvOpAtomicSMin:
case SpvOpAtomicAnd:
case SpvOpAtomicOr:
case SpvOpAtomicXor:
- intrin->src[2] = nir_src_for_ssa(vtn_ssa_value(b, w[6])->def);
- break;
-
- case SpvOpAtomicCompareExchange:
- intrin->src[2] = nir_src_for_ssa(vtn_ssa_value(b, w[7])->def);
- intrin->src[3] = nir_src_for_ssa(vtn_ssa_value(b, w[6])->def);
- break;
-
- case SpvOpAtomicISub:
- intrin->src[2] = nir_src_for_ssa(nir_ineg(&b->nb, vtn_ssa_value(b, w[6])->def));
+ fill_common_atomic_sources(b, opcode, w, &intrin->src[2]);
break;
default:
get_ssbo_nir_atomic_op(SpvOp opcode)
{
switch (opcode) {
+ case SpvOpAtomicLoad: return nir_intrinsic_load_ssbo;
+ case SpvOpAtomicStore: return nir_intrinsic_store_ssbo;
#define OP(S, N) case SpvOp##S: return nir_intrinsic_ssbo_##N;
OP(AtomicExchange, atomic_exchange)
OP(AtomicCompareExchange, atomic_comp_swap)
get_shared_nir_atomic_op(SpvOp opcode)
{
switch (opcode) {
+ case SpvOpAtomicLoad: return nir_intrinsic_load_var;
+ case SpvOpAtomicStore: return nir_intrinsic_store_var;
#define OP(S, N) case SpvOp##S: return nir_intrinsic_var_##N;
OP(AtomicExchange, atomic_exchange)
OP(AtomicCompareExchange, atomic_comp_swap)
}
static void
-fill_common_atomic_sources(struct vtn_builder *b, SpvOp opcode,
- const uint32_t *w, nir_src *src)
+vtn_handle_ssbo_or_shared_atomic(struct vtn_builder *b, SpvOp opcode,
+ const uint32_t *w, unsigned count)
{
+ struct vtn_access_chain *chain;
+ nir_intrinsic_instr *atomic;
+
switch (opcode) {
+ case SpvOpAtomicLoad:
+ case SpvOpAtomicExchange:
+ case SpvOpAtomicCompareExchange:
+ case SpvOpAtomicCompareExchangeWeak:
case SpvOpAtomicIIncrement:
- src[0] = nir_src_for_ssa(nir_imm_int(&b->nb, 1));
- break;
-
case SpvOpAtomicIDecrement:
- src[0] = nir_src_for_ssa(nir_imm_int(&b->nb, -1));
- break;
-
- case SpvOpAtomicISub:
- src[0] =
- nir_src_for_ssa(nir_ineg(&b->nb, vtn_ssa_value(b, w[6])->def));
- break;
-
- case SpvOpAtomicCompareExchange:
- src[0] = nir_src_for_ssa(vtn_ssa_value(b, w[7])->def);
- src[1] = nir_src_for_ssa(vtn_ssa_value(b, w[8])->def);
- break;
- /* Fall through */
-
- case SpvOpAtomicExchange:
case SpvOpAtomicIAdd:
+ case SpvOpAtomicISub:
case SpvOpAtomicSMin:
case SpvOpAtomicUMin:
case SpvOpAtomicSMax:
case SpvOpAtomicAnd:
case SpvOpAtomicOr:
case SpvOpAtomicXor:
- src[0] = nir_src_for_ssa(vtn_ssa_value(b, w[6])->def);
+ chain =
+ vtn_value(b, w[3], vtn_value_type_access_chain)->access_chain;
+ break;
+
+ case SpvOpAtomicStore:
+ chain =
+ vtn_value(b, w[1], vtn_value_type_access_chain)->access_chain;
break;
default:
unreachable("Invalid SPIR-V atomic");
}
-}
-
-static void
-vtn_handle_ssbo_or_shared_atomic(struct vtn_builder *b, SpvOp opcode,
- const uint32_t *w, unsigned count)
-{
- struct vtn_access_chain *chain =
- vtn_value(b, w[3], vtn_value_type_access_chain)->access_chain;
- nir_intrinsic_instr *atomic;
/*
SpvScope scope = w[4];
*/
if (chain->var->mode == vtn_variable_mode_workgroup) {
- nir_deref *deref = &vtn_access_chain_to_deref(b, chain)->deref;
+ struct vtn_type *type = chain->var->type;
+ nir_deref_var *deref = vtn_access_chain_to_deref(b, chain);
nir_intrinsic_op op = get_shared_nir_atomic_op(opcode);
atomic = nir_intrinsic_instr_create(b->nb.shader, op);
- atomic->variables[0] = nir_deref_as_var(nir_copy_deref(atomic, deref));
- fill_common_atomic_sources(b, opcode, w, &atomic->src[0]);
+ atomic->variables[0] = nir_deref_var_clone(deref, atomic);
+
+ switch (opcode) {
+ case SpvOpAtomicLoad:
+ atomic->num_components = glsl_get_vector_elements(type->type);
+ break;
+
+ case SpvOpAtomicStore:
+ atomic->num_components = glsl_get_vector_elements(type->type);
+ nir_intrinsic_set_write_mask(atomic, (1 << atomic->num_components) - 1);
+ atomic->src[0] = nir_src_for_ssa(vtn_ssa_value(b, w[4])->def);
+ break;
+
+ case SpvOpAtomicExchange:
+ case SpvOpAtomicCompareExchange:
+ case SpvOpAtomicCompareExchangeWeak:
+ case SpvOpAtomicIIncrement:
+ case SpvOpAtomicIDecrement:
+ case SpvOpAtomicIAdd:
+ case SpvOpAtomicISub:
+ case SpvOpAtomicSMin:
+ case SpvOpAtomicUMin:
+ case SpvOpAtomicSMax:
+ case SpvOpAtomicUMax:
+ case SpvOpAtomicAnd:
+ case SpvOpAtomicOr:
+ case SpvOpAtomicXor:
+ fill_common_atomic_sources(b, opcode, w, &atomic->src[0]);
+ break;
+
+ default:
+ unreachable("Invalid SPIR-V atomic");
+
+ }
} else {
assert(chain->var->mode == vtn_variable_mode_ssbo);
struct vtn_type *type;
nir_intrinsic_op op = get_ssbo_nir_atomic_op(opcode);
atomic = nir_intrinsic_instr_create(b->nb.shader, op);
- atomic->src[0] = nir_src_for_ssa(index);
- atomic->src[1] = nir_src_for_ssa(offset);
- fill_common_atomic_sources(b, opcode, w, &atomic->src[2]);
+
+ switch (opcode) {
+ case SpvOpAtomicLoad:
+ atomic->num_components = glsl_get_vector_elements(type->type);
+ atomic->src[0] = nir_src_for_ssa(index);
+ atomic->src[1] = nir_src_for_ssa(offset);
+ break;
+
+ case SpvOpAtomicStore:
+ atomic->num_components = glsl_get_vector_elements(type->type);
+ nir_intrinsic_set_write_mask(atomic, (1 << atomic->num_components) - 1);
+ atomic->src[0] = nir_src_for_ssa(vtn_ssa_value(b, w[4])->def);
+ atomic->src[1] = nir_src_for_ssa(index);
+ atomic->src[2] = nir_src_for_ssa(offset);
+ break;
+
+ case SpvOpAtomicExchange:
+ case SpvOpAtomicCompareExchange:
+ case SpvOpAtomicCompareExchangeWeak:
+ case SpvOpAtomicIIncrement:
+ case SpvOpAtomicIDecrement:
+ case SpvOpAtomicIAdd:
+ case SpvOpAtomicISub:
+ case SpvOpAtomicSMin:
+ case SpvOpAtomicUMin:
+ case SpvOpAtomicSMax:
+ case SpvOpAtomicUMax:
+ case SpvOpAtomicAnd:
+ case SpvOpAtomicOr:
+ case SpvOpAtomicXor:
+ atomic->src[0] = nir_src_for_ssa(index);
+ atomic->src[1] = nir_src_for_ssa(offset);
+ fill_common_atomic_sources(b, opcode, w, &atomic->src[2]);
+ break;
+
+ default:
+ unreachable("Invalid SPIR-V atomic");
+ }
}
- nir_ssa_dest_init(&atomic->instr, &atomic->dest, 1, 32, NULL);
+ if (opcode != SpvOpAtomicStore) {
+ struct vtn_type *type = vtn_value(b, w[1], vtn_value_type_type)->type;
- struct vtn_type *type = vtn_value(b, w[1], vtn_value_type_type)->type;
- struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
- val->ssa = rzalloc(b, struct vtn_ssa_value);
- val->ssa->def = &atomic->dest.ssa;
- val->ssa->type = type->type;
+ nir_ssa_dest_init(&atomic->instr, &atomic->dest,
+ glsl_get_vector_elements(type->type),
+ glsl_get_bit_size(type->type), NULL);
+
+ struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
+ val->ssa = rzalloc(b, struct vtn_ssa_value);
+ val->ssa->def = &atomic->dest.ssa;
+ val->ssa->type = type->type;
+ }
nir_builder_instr_insert(&b->nb, &atomic->instr);
}
}
}
+#define spv_check_supported(name, cap) do { \
+ if (!(b->ext && b->ext->name)) \
+ vtn_warn("Unsupported SPIR-V capability: %s", \
+ spirv_capability_to_string(cap)); \
+ } while(0)
+
static bool
vtn_handle_preamble_instruction(struct vtn_builder *b, SpvOp opcode,
const uint32_t *w, unsigned count)
case SpvCapabilityMatrix:
case SpvCapabilityShader:
case SpvCapabilityGeometry:
- case SpvCapabilityTessellationPointSize:
case SpvCapabilityGeometryPointSize:
case SpvCapabilityUniformBufferArrayDynamicIndexing:
case SpvCapabilitySampledImageArrayDynamicIndexing:
case SpvCapabilitySampledBuffer:
case SpvCapabilityImageBuffer:
case SpvCapabilityImageQuery:
- break;
+ case SpvCapabilityDerivativeControl:
+ case SpvCapabilityInterpolationFunction:
+ case SpvCapabilityMultiViewport:
+ case SpvCapabilitySampleRateShading:
case SpvCapabilityClipDistance:
case SpvCapabilityCullDistance:
+ case SpvCapabilityInputAttachment:
+ case SpvCapabilityImageGatherExtended:
+ case SpvCapabilityStorageImageExtendedFormats:
+ break;
+
case SpvCapabilityGeometryStreams:
- fprintf(stderr, "WARNING: Unsupported SPIR-V Capability\n");
+ case SpvCapabilityTessellation:
+ case SpvCapabilityTessellationPointSize:
+ case SpvCapabilityLinkage:
+ case SpvCapabilityVector16:
+ case SpvCapabilityFloat16Buffer:
+ case SpvCapabilityFloat16:
+ case SpvCapabilityFloat64:
+ case SpvCapabilityInt64:
+ case SpvCapabilityInt64Atomics:
+ case SpvCapabilityAtomicStorage:
+ case SpvCapabilityInt16:
+ case SpvCapabilityStorageImageMultisample:
+ case SpvCapabilityImageCubeArray:
+ case SpvCapabilityInt8:
+ case SpvCapabilitySparseResidency:
+ case SpvCapabilityMinLod:
+ case SpvCapabilityTransformFeedback:
+ case SpvCapabilityStorageImageReadWithoutFormat:
+ case SpvCapabilityStorageImageWriteWithoutFormat:
+ vtn_warn("Unsupported SPIR-V capability: %s",
+ spirv_capability_to_string(cap));
+ break;
+
+ case SpvCapabilityAddresses:
+ case SpvCapabilityKernel:
+ case SpvCapabilityImageBasic:
+ case SpvCapabilityImageReadWrite:
+ case SpvCapabilityImageMipmap:
+ case SpvCapabilityPipes:
+ case SpvCapabilityGroups:
+ case SpvCapabilityDeviceEnqueue:
+ case SpvCapabilityLiteralSampler:
+ case SpvCapabilityGenericPointer:
+ vtn_warn("Unsupported OpenCL-style SPIR-V capability: %s",
+ spirv_capability_to_string(cap));
+ break;
+
+ case SpvCapabilityImageMSArray:
+ spv_check_supported(image_ms_array, cap);
break;
- default:
- assert(!"Unsupported capability");
}
break;
}
case SpvExecutionModeEarlyFragmentTests:
assert(b->shader->stage == MESA_SHADER_FRAGMENT);
- b->shader->info.fs.early_fragment_tests = true;
+ b->shader->info->fs.early_fragment_tests = true;
break;
case SpvExecutionModeInvocations:
assert(b->shader->stage == MESA_SHADER_GEOMETRY);
- b->shader->info.gs.invocations = MAX2(1, mode->literals[0]);
+ b->shader->info->gs.invocations = MAX2(1, mode->literals[0]);
break;
case SpvExecutionModeDepthReplacing:
assert(b->shader->stage == MESA_SHADER_FRAGMENT);
- b->shader->info.fs.depth_layout = FRAG_DEPTH_LAYOUT_ANY;
+ b->shader->info->fs.depth_layout = FRAG_DEPTH_LAYOUT_ANY;
break;
case SpvExecutionModeDepthGreater:
assert(b->shader->stage == MESA_SHADER_FRAGMENT);
- b->shader->info.fs.depth_layout = FRAG_DEPTH_LAYOUT_GREATER;
+ b->shader->info->fs.depth_layout = FRAG_DEPTH_LAYOUT_GREATER;
break;
case SpvExecutionModeDepthLess:
assert(b->shader->stage == MESA_SHADER_FRAGMENT);
- b->shader->info.fs.depth_layout = FRAG_DEPTH_LAYOUT_LESS;
+ b->shader->info->fs.depth_layout = FRAG_DEPTH_LAYOUT_LESS;
break;
case SpvExecutionModeDepthUnchanged:
assert(b->shader->stage == MESA_SHADER_FRAGMENT);
- b->shader->info.fs.depth_layout = FRAG_DEPTH_LAYOUT_UNCHANGED;
+ b->shader->info->fs.depth_layout = FRAG_DEPTH_LAYOUT_UNCHANGED;
break;
case SpvExecutionModeLocalSize:
assert(b->shader->stage == MESA_SHADER_COMPUTE);
- b->shader->info.cs.local_size[0] = mode->literals[0];
- b->shader->info.cs.local_size[1] = mode->literals[1];
- b->shader->info.cs.local_size[2] = mode->literals[2];
+ b->shader->info->cs.local_size[0] = mode->literals[0];
+ b->shader->info->cs.local_size[1] = mode->literals[1];
+ b->shader->info->cs.local_size[2] = mode->literals[2];
break;
case SpvExecutionModeLocalSizeHint:
- break; /* Nothing do do with this */
+ break; /* Nothing to do with this */
case SpvExecutionModeOutputVertices:
assert(b->shader->stage == MESA_SHADER_GEOMETRY);
- b->shader->info.gs.vertices_out = mode->literals[0];
+ b->shader->info->gs.vertices_out = mode->literals[0];
break;
case SpvExecutionModeInputPoints:
case SpvExecutionModeQuads:
case SpvExecutionModeIsolines:
if (b->shader->stage == MESA_SHADER_GEOMETRY) {
- b->shader->info.gs.vertices_in =
+ b->shader->info->gs.vertices_in =
vertices_in_from_spv_execution_mode(mode->exec_mode);
} else {
assert(!"Tesselation shaders not yet supported");
case SpvExecutionModeOutputLineStrip:
case SpvExecutionModeOutputTriangleStrip:
assert(b->shader->stage == MESA_SHADER_GEOMETRY);
- b->shader->info.gs.output_primitive =
+ b->shader->info->gs.output_primitive =
gl_primitive_from_spv_execution_mode(mode->exec_mode);
break;
break;
}
+ case SpvOpAtomicLoad:
case SpvOpAtomicExchange:
case SpvOpAtomicCompareExchange:
case SpvOpAtomicCompareExchangeWeak:
break;
}
+ case SpvOpAtomicStore: {
+ struct vtn_value *pointer = vtn_untyped_value(b, w[1]);
+ if (pointer->value_type == vtn_value_type_image_pointer) {
+ vtn_handle_image(b, opcode, w, count);
+ } else {
+ assert(pointer->value_type == vtn_value_type_access_chain);
+ vtn_handle_ssbo_or_shared_atomic(b, opcode, w, count);
+ }
+ break;
+ }
+
case SpvOpSNegate:
case SpvOpFNegate:
case SpvOpNot:
spirv_to_nir(const uint32_t *words, size_t word_count,
struct nir_spirv_specialization *spec, unsigned num_spec,
gl_shader_stage stage, const char *entry_point_name,
+ const struct nir_spirv_supported_extensions *ext,
const nir_shader_compiler_options *options)
{
const uint32_t *word_end = words + word_count;
exec_list_make_empty(&b->functions);
b->entry_point_stage = stage;
b->entry_point_name = entry_point_name;
+ b->ext = ext;
/* Handle all the preamble instructions */
words = vtn_foreach_instruction(b, words, word_end,
return NULL;
}
- b->shader = nir_shader_create(NULL, stage, options);
+ b->shader = nir_shader_create(NULL, stage, options, NULL);
/* Set shader info defaults */
- b->shader->info.gs.invocations = 1;
+ b->shader->info->gs.invocations = 1;
/* Parse execution modes */
vtn_foreach_execution_mode(b, b->entry_point,