void
vec4_tes_visitor::nir_emit_intrinsic(nir_intrinsic_instr *instr)
{
const struct brw_tes_prog_data *tes_prog_data =
(const struct brw_tes_prog_data *) prog_data;
switch (instr->intrinsic) {
case nir_intrinsic_load_tess_coord:
/* gl_TessCoord is part of the payload in g1 channels 0-2 and 4-6. */
emit(MOV(get_nir_dest(instr->dest, BRW_REGISTER_TYPE_F),
src_reg(brw_vec8_grf(1, 0))));
break;
case nir_intrinsic_load_tess_level_outer:
if (tes_prog_data->domain == BRW_TESS_DOMAIN_ISOLINE) {
emit(MOV(get_nir_dest(instr->dest, BRW_REGISTER_TYPE_F),
swizzle(src_reg(ATTR, 1, glsl_type::vec4_type),
BRW_SWIZZLE_ZWZW)));
} else {
emit(MOV(get_nir_dest(instr->dest, BRW_REGISTER_TYPE_F),
swizzle(src_reg(ATTR, 1, glsl_type::vec4_type),
BRW_SWIZZLE_WZYX)));
}
break;
case nir_intrinsic_load_tess_level_inner:
if (tes_prog_data->domain == BRW_TESS_DOMAIN_QUAD) {
emit(MOV(get_nir_dest(instr->dest, BRW_REGISTER_TYPE_F),
swizzle(src_reg(ATTR, 0, glsl_type::vec4_type),
BRW_SWIZZLE_WZYX)));
} else {
emit(MOV(get_nir_dest(instr->dest, BRW_REGISTER_TYPE_F),
src_reg(ATTR, 1, glsl_type::float_type)));
}
break;
case nir_intrinsic_load_primitive_id:
emit(TES_OPCODE_GET_PRIMITIVE_ID,
get_nir_dest(instr->dest, BRW_REGISTER_TYPE_UD));
break;
case nir_intrinsic_load_input:
case nir_intrinsic_load_per_vertex_input: {
src_reg indirect_offset = get_indirect_offset(instr);
unsigned imm_offset = instr->const_index[0];
src_reg header = input_read_header;
bool is_64bit = nir_dest_bit_size(instr->dest) == 64;
unsigned first_component = nir_intrinsic_component(instr);
if (is_64bit)
first_component /= 2;
if (indirect_offset.file != BAD_FILE) {
header = src_reg(this, glsl_type::uvec4_type);
emit(TES_OPCODE_ADD_INDIRECT_URB_OFFSET, dst_reg(header),
input_read_header, indirect_offset);
} else {
/* Arbitrarily only push up to 24 vec4 slots worth of data,
* which is 12 registers (since each holds 2 vec4 slots).
*/
const unsigned max_push_slots = 24;
if (imm_offset < max_push_slots) {
const glsl_type *src_glsl_type =
is_64bit ? glsl_type::dvec4_type : glsl_type::ivec4_type;
src_reg src = src_reg(ATTR, imm_offset, src_glsl_type);
src.swizzle = BRW_SWZ_COMP_INPUT(first_component);
const brw_reg_type dst_reg_type =
is_64bit ? BRW_REGISTER_TYPE_DF : BRW_REGISTER_TYPE_D;
emit(MOV(get_nir_dest(instr->dest, dst_reg_type), src));
prog_data->urb_read_length =
MAX2(prog_data->urb_read_length,
DIV_ROUND_UP(imm_offset + (is_64bit ? 2 : 1), 2));
break;
}
}
if (!is_64bit) {
dst_reg temp(this, glsl_type::ivec4_type);
vec4_instruction *read =
emit(VEC4_OPCODE_URB_READ, temp, src_reg(header));
read->offset = imm_offset;
read->urb_write_flags = BRW_URB_WRITE_PER_SLOT_OFFSET;
src_reg src = src_reg(temp);
src.swizzle = BRW_SWZ_COMP_INPUT(first_component);
/* Copy to target. We might end up with some funky writemasks landing
* in here, but we really don't want them in the above pseudo-ops.
*/
dst_reg dst = get_nir_dest(instr->dest, BRW_REGISTER_TYPE_D);
dst.writemask = brw_writemask_for_size(instr->num_components);
emit(MOV(dst, src));
} else {
/* For 64-bit we need to load twice as many 32-bit components, and for
* dvec3/4 we need to emit 2 URB Read messages
*/
dst_reg temp(this, glsl_type::dvec4_type);
dst_reg temp_d = retype(temp, BRW_REGISTER_TYPE_D);
vec4_instruction *read =
emit(VEC4_OPCODE_URB_READ, temp_d, src_reg(header));
read->offset = imm_offset;
read->urb_write_flags = BRW_URB_WRITE_PER_SLOT_OFFSET;
if (instr->num_components > 2) {
read = emit(VEC4_OPCODE_URB_READ, byte_offset(temp_d, REG_SIZE),
src_reg(header));
read->offset = imm_offset + 1;
read->urb_write_flags = BRW_URB_WRITE_PER_SLOT_OFFSET;
}
src_reg temp_as_src = src_reg(temp);
temp_as_src.swizzle = BRW_SWZ_COMP_INPUT(first_component);
dst_reg shuffled(this, glsl_type::dvec4_type);
shuffle_64bit_data(shuffled, temp_as_src, false);
dst_reg dst = get_nir_dest(instr->dest, BRW_REGISTER_TYPE_DF);
dst.writemask = brw_writemask_for_size(instr->num_components);
emit(MOV(dst, src_reg(shuffled)));
}
break;
}
default:
vec4_visitor::nir_emit_intrinsic(instr);
}
}
arma::Mat<T> shuffled_matrix(const arma::Mat<T>& matrix, std::mt19937* gen) {
arma::Mat<T> shuffled(matrix.n_rows, matrix.n_cols);
std::vector<std::size_t> indices(matrix.n_rows);
std::iota(indices.begin(), indices.end(), 0);
for (std::size_t j=0; j < matrix.n_cols; ++j) {
std::shuffle(indices.begin(), indices.end(), *gen);
for (std::size_t i=0; i < matrix.n_rows; ++i) {
shuffled(i, j) = matrix(indices[i], j);
}
}
return shuffled;
}
void pri_queue_test_ordered_iterators(void)
{
for (int i = 0; i != test_size; ++i) {
test_data data = make_test_data(i);
test_data shuffled (data);
std::random_shuffle(shuffled.begin(), shuffled.end());
pri_queue q;
BOOST_REQUIRE(q.ordered_begin() == q.ordered_end());
fill_q(q, shuffled);
test_data data_from_queue(q.ordered_begin(), q.ordered_end());
std::reverse(data_from_queue.begin(), data_from_queue.end());
BOOST_REQUIRE(data == data_from_queue);
for (unsigned long i = 0; i != data.size(); ++i)
BOOST_REQUIRE(std::find(q.ordered_begin(), q.ordered_end(), data[i]) != q.ordered_end());
for (unsigned long i = 0; i != data.size(); ++i)
BOOST_REQUIRE(std::find(q.ordered_begin(), q.ordered_end(), data[i] + data.size()) == q.ordered_end());
for (unsigned long i = 0; i != data.size(); ++i) {
BOOST_REQUIRE_EQUAL((long)std::distance(q.begin(), q.end()), (long)(data.size() - i));
q.pop();
}
}
}
void pri_queue_test_update_shuffled(void)
{
pri_queue q;
test_data data = make_test_data(test_size);
PUSH_WITH_HANDLES(handles, q, data);
test_data shuffled (data);
std::random_shuffle(shuffled.begin(), shuffled.end());
for (int i = 0; i != test_size; ++i)
q.update(handles[i], shuffled[i]);
check_q(q, data);
}
void pri_queue_test_random_push(void)
{
for (int i = 0; i != test_size; ++i)
{
pri_queue q;
test_data data = make_test_data(i);
test_data shuffled (data);
std::random_shuffle(shuffled.begin(), shuffled.end());
fill_q(q, shuffled);
check_q(q, data);
}
}
inline void shuffle_stream (
const char * messages_in,
const char * shuffled_out,
long seed,
double target_mem_bytes)
{
typedef std::vector<char> Message;
typedef uint32_t pos_t;
LOOM_ASSERT(
std::string(messages_in) != std::string(shuffled_out),
"cannot shuffle file in-place: " << messages_in);
const auto stats = protobuf::InFile::stream_stats(messages_in);
LOOM_ASSERT(stats.is_file, "shuffle input is not a file: " << messages_in);
const uint64_t max_message_count = std::numeric_limits<pos_t>::max();
LOOM_ASSERT(stats.message_count, max_message_count);
const size_t message_count = stats.message_count;
double index_bytes = sizeof(pos_t) * message_count;
double target_chunk_size = std::max(1.0, std::min(double(message_count),
(target_mem_bytes - index_bytes) / stats.max_message_size));
size_t chunk_size = static_cast<size_t>(std::round(target_chunk_size));
std::vector<pos_t> index(message_count);
for (size_t i = 0; i < message_count; ++i) {
index[i] = i;
}
std::shuffle(index.begin(), index.end(), loom::rng_t(seed));
Message message;
std::vector<Message> chunk;
protobuf::OutFile shuffled(shuffled_out);
for (size_t begin = 0; begin < message_count; begin += chunk_size) {
size_t end = std::min(begin + chunk_size, message_count);
chunk.resize(end - begin);
protobuf::InFile messages(messages_in);
for (size_t i : index) {
messages.try_read_stream(message);
if (begin <= i and i < end) {
std::swap(message, chunk[i - begin]);
}
}
for (const auto & message : chunk) {
shuffled.write_stream(message);
}
}
}
void pri_queue_test_swap(void)
{
for (int i = 0; i != test_size; ++i)
{
pri_queue q;
test_data data = make_test_data(i);
test_data shuffled (data);
std::random_shuffle(shuffled.begin(), shuffled.end());
fill_q(q, shuffled);
pri_queue r;
q.swap(r);
check_q(r, data);
BOOST_REQUIRE(q.empty());
}
}