JIT: some cleaning of the matrix reduction profile

This commit is contained in:
Philippe Tillet
2016-07-01 20:23:33 -07:00
parent add123da11
commit 35260a49d4
2 changed files with 31 additions and 47 deletions

View File

@@ -80,9 +80,8 @@ std::string reduce_2d::generate_impl(std::string const & suffix, expression_tree
name[0] += suffix; name[0] += suffix;
name[1] += suffix; name[1] += suffix;
int col_simd_width = (reduction_type_ == REDUCE_COLUMNS) ? 1 : p_.simd_width; unsigned int ldls = p_.local_size_0;
unsigned int local_size_0_ld = p_.local_size_0; std::string ls0ldstr = to_string(ldls);
std::string ls0ldstr = to_string(local_size_0_ld);
auto unroll_tmp = [&]() auto unroll_tmp = [&]()
{ {
@@ -122,44 +121,47 @@ std::string reduce_2d::generate_impl(std::string const & suffix, expression_tree
stream.inc_tab(); stream.inc_tab();
//Unroll //Unroll
unroll_tmp(); unroll_tmp();
//Declare Buffers
for (symbolic::reduce_2d* rd : reductions)
stream << rd->process("$LOCAL " + append_width("#scalartype", col_simd_width) + " #name_buf[" + to_string(p_.local_size_1*local_size_0_ld) + "];") << std::endl;
//Loop r
stream << "for($SIZE_T r = $GLOBAL_IDX_1*" << col_simd_width << "; r < (M +" << p_.local_size_1 - 1 << ")/" << p_.local_size_1 << "*" << p_.local_size_1*col_simd_width << "; r += $GLOBAL_SIZE_1*" << col_simd_width << ")" << std::endl;
stream << "{" << std::endl;
stream.inc_tab();
stream << "$SIZE_T lidx = $LOCAL_IDX_0;" << std::endl; stream << "$SIZE_T lidx = $LOCAL_IDX_0;" << std::endl;
stream << "$SIZE_T lidy = $LOCAL_IDX_1;" << std::endl; stream << "$SIZE_T lidy = $LOCAL_IDX_1;" << std::endl;
//Loop r
std::ostringstream upper;
upper << "(M +" << p_.local_size_1 - 1 << ")/" << p_.local_size_1 << "*" << p_.local_size_1;
element_wise_loop_1D(stream, p_.fetch_policy, (reduction_type_==REDUCE_ROWS)?p_.simd_width:1, "r", upper.str(), "$GLOBAL_IDX_1", "$GLOBAL_SIZE_1", device, [&](unsigned int cwidth)
{
//Declare Buffers
for (symbolic::reduce_2d* rd : reductions)
stream << rd->process("$LOCAL " + append_width("#scalartype", cwidth) + " #name_buf[" + to_string(p_.local_size_1*ldls) + "];") << std::endl;
//Accumulators //Accumulators
for (symbolic::reduce_2d* rd : reductions){ for (symbolic::reduce_2d* rd : reductions){
std::string data_type = append_width("#scalartype",col_simd_width); std::string data_type = append_width("#scalartype",cwidth);
stream << rd->process(data_type + " #name_acc = " + InitPrefix(backend, data_type).get() + "(" + neutral_element((rd)->op(), backend, "#scalartype") + ");") << std::endl; stream << rd->process(data_type + " #name_acc = " + InitPrefix(backend, data_type).get() + "(" + neutral_element((rd)->op(), backend, "#scalartype") + ");") << std::endl;
} }
//Loop c //Loop c
stream << "if (r < M)" << std::endl; stream << "if (r < M)" << std::endl;
stream << "{" << std::endl; stream << "{" << std::endl;
stream.inc_tab(); stream.inc_tab();
element_wise_loop_1D(stream, p_.fetch_policy, (reduction_type_==REDUCE_COLUMNS)?p_.simd_width:1, "c", "N", "$GLOBAL_IDX_0", "$GLOBAL_SIZE_0", device, [&](unsigned int row_simd_width) element_wise_loop_1D(stream, p_.fetch_policy, (reduction_type_==REDUCE_COLUMNS)?p_.simd_width:1, "c", "N", "$GLOBAL_IDX_0", "$GLOBAL_SIZE_0", device, [&](unsigned int rwidth)
{ {
std::string rdtype = append_width("#scalartype", row_simd_width); std::string rdtype = append_width("#scalartype", rwidth);
std::string cdtype = append_width("#scalartype", col_simd_width); std::string cdtype = append_width("#scalartype", cwidth);
//Fetch //Fetch
std::set<std::string> fetched; std::set<std::string> fetched;
for (symbolic::reduce_2d* rd : reductions) for (symbolic::reduce_2d* rd : reductions)
for(symbolic::leaf* sym: symbolic::extract<symbolic::leaf>(tree, symbols, rd->root(), false)) for(symbolic::leaf* sym: symbolic::extract<symbolic::leaf>(tree, symbols, rd->root(), false))
if(fetched.insert(sym->process("#name")).second){ if(fetched.insert(sym->process("#name")).second){
if(reduction_type_==REDUCE_COLUMNS) if(reduction_type_==REDUCE_COLUMNS)
stream << sym->process(rdtype + " #name = " + append_width("loadv",row_simd_width) + "(c,r);") << std::endl; stream << sym->process(rdtype + " #name = " + append_width("loadv",rwidth) + "(c,r);") << std::endl;
else else
stream << sym->process(cdtype + " #name = " + append_width("loadv",col_simd_width) + "(r,c);") << std::endl; stream << sym->process(cdtype + " #name = " + append_width("loadv",cwidth) + "(r,c);") << std::endl;
} }
//Compute //Compute
for (symbolic::reduce_2d* rd : reductions) for (symbolic::reduce_2d* rd : reductions)
for (unsigned int s = 0; s < row_simd_width; ++s){ for (unsigned int s = 0; s < rwidth; ++s){
std::string value = rd->lhs()->evaluate({{"leaf", access_vector_type("#name", s, row_simd_width)}}); std::string value = rd->lhs()->evaluate({{"leaf", access_vector_type("#name", s, rwidth)}});
if (is_indexing(rd->op().type)) if (is_indexing(rd->op().type))
compute_index_reduce_1d(stream, rd->process("#name_acc"), "c*"+to_string(row_simd_width) + to_string(s), rd->process("#name_acc_value"), value, rd->op()); compute_index_reduce_1d(stream, rd->process("#name_acc"), "c*"+to_string(rwidth) + to_string(s), rd->process("#name_acc_value"), value, rd->op());
else else
compute_reduce_1d(stream, rd->process("#name_acc"), value,rd->op()); compute_reduce_1d(stream, rd->process("#name_acc"), value,rd->op());
} }
@@ -190,42 +192,24 @@ std::string reduce_2d::generate_impl(std::string const & suffix, expression_tree
stream.dec_tab(); stream.dec_tab();
stream << "}" << std::endl; stream << "}" << std::endl;
//Write result/temporary //Write result/temporary
stream << "if (lidx == 0 && r < M)" << std::endl; stream << "if (r < M && lidx == 0)" << std::endl;
stream << "{" << std::endl; stream << "{" << std::endl;
stream.inc_tab(); stream.inc_tab();
if(p_.num_groups_0==1) if(p_.num_groups_0==1)
for(size_t idx: assignments) for(size_t idx: assignments)
for(int s = 0 ; s < col_simd_width ; ++s) for(size_t s = 0 ; s < cwidth ; ++s)
stream << symbols.at(idx)->evaluate({{"leaf", "at(r+" + to_string(s) + ")"}, stream << symbols.at(idx)->evaluate({{"leaf", "at(r+" + to_string(s) + ")"},
{"reduce_2d", access_vector_type("#name_buf[lidy*" + ls0ldstr + "]", s, col_simd_width)}}) << ";" << std::endl; {"reduce_2d", access_vector_type("#name_buf[lidy*" + ls0ldstr + "]", s, cwidth)}}) << ";" << std::endl;
else else
{
for (symbolic::reduction const * rd : reductions) for (symbolic::reduction const * rd : reductions)
{ for(size_t s = 0 ; s < cwidth ; ++s){
if(col_simd_width > 1)
stream << "if(M - r > " << col_simd_width << "){" << std::endl;
if (is_indexing(rd->op().type)) if (is_indexing(rd->op().type))
stream << rd->process(vstore(col_simd_width,"uint", "#name_buf_value[lidy*" + ls0ldstr + "]", "0", "#name_temp_value + r + M*$GROUP_IDX_0", "1", backend, false)) << ";" << std::endl; stream << "if(r + " << s << "< M) " << rd->process("#name_temp_value[r + " + to_string(s) + " + M*$GROUP_IDX_0] = " + access_vector_type("#name_buf_value[lidy*" + ls0ldstr + "]", s, cwidth)) << ";" << std::endl;
stream << rd->process(vstore(col_simd_width,"#scalartype", "#name_buf[lidy*" + ls0ldstr + "]", "0", "#name_temp + r + M*$GROUP_IDX_0", "1", backend, false)) << ";" << std::endl; stream << "if(r + " << s << "< M) " << rd->process("#name_temp[r + " + to_string(s) + " + M*$GROUP_IDX_0] = " + access_vector_type("#name_buf[lidy*" + ls0ldstr + "]", s, cwidth)) << ";" << std::endl;
if(col_simd_width > 1)
{
stream << "}" << std::endl;
stream << "else{" << std::endl;
stream.inc_tab();
for(int s = 0 ; s < col_simd_width ; ++s){
if (is_indexing(rd->op().type))
stream << "if(r + " << s << "< M) " << rd->process("#name_temp_value[r + " + to_string(s) + " + M*$GROUP_IDX_0] = " + access_vector_type("#name_buf_value[lidy*" + ls0ldstr + "]", s)) << ";" << std::endl;
stream << "if(r + " << s << "< M) " << rd->process("#name_temp[r + " + to_string(s) + " + M*$GROUP_IDX_0] = " + access_vector_type("#name_buf[lidy*" + ls0ldstr + "]", s)) << ";" << std::endl;
} }
stream.dec_tab(); stream.dec_tab();
stream << "}" << std::endl; stream << "}" << std::endl;
} });
}
}
stream.dec_tab();
stream << "}" << std::endl;
stream.dec_tab();
stream << "}" << std::endl;
stream.dec_tab(); stream.dec_tab();
stream << "}" << std::endl; stream << "}" << std::endl;
@@ -242,7 +226,7 @@ std::string reduce_2d::generate_impl(std::string const & suffix, expression_tree
stream.inc_tab(); stream.inc_tab();
unroll_tmp(); unroll_tmp();
for (symbolic::reduce_2d* rd : reductions) for (symbolic::reduce_2d* rd : reductions)
stream << rd->process("$LOCAL #scalartype #name_buf[" + to_string(p_.local_size_1*local_size_0_ld) + "];") << std::endl; stream << rd->process("$LOCAL #scalartype #name_buf[" + to_string(p_.local_size_1*ldls) + "];") << std::endl;
stream << "for($SIZE_T r = $GLOBAL_IDX_1; r < (M +" << p_.local_size_1 - 1 << ")/" << p_.local_size_1 << "*" << p_.local_size_1 << "; r += " << GlobalSize1(backend) << "){" << std::endl; stream << "for($SIZE_T r = $GLOBAL_IDX_1; r < (M +" << p_.local_size_1 - 1 << ")/" << p_.local_size_1 << "*" << p_.local_size_1 << "; r += " << GlobalSize1(backend) << "){" << std::endl;
stream.inc_tab(); stream.inc_tab();
stream << "$SIZE_T lidx = $LOCAL_IDX_0;" << std::endl; stream << "$SIZE_T lidx = $LOCAL_IDX_0;" << std::endl;

View File

@@ -55,7 +55,7 @@ void test(sc::driver::Context const & ctx, int& nfail, int& npass)
int STARTM = std::min(5,M-1), STARTN = std::min(7, N-1), STRIDEM = std::min(3,M); int STARTM = std::min(5,M-1), STARTN = std::min(7, N-1), STRIDEM = std::min(3,M);
INIT_VECTOR(M, SUBM, STARTM, STRIDEM, cy, y, ctx); INIT_VECTOR(M, SUBM, STARTM, STRIDEM, cy, y, ctx);
INIT_VECTOR(N, SUBN, STARTN, STRIDEN, cx, x, ctx); INIT_VECTOR(N, SUBN, STARTN, STRIDEN, cx, x, ctx);
INIT_MATRIX(M, SUBM, STARTM, STRIDEM, N, SUBN, STARTN, STRIDEN, cA, A, ctx); INIT_MATRIX(M, SUBM, STARTM, 1, N, SUBN, STARTN, 1, cA, A, ctx);
std::string suffix = "MN"; std::string suffix = "MN";
if(M==1) suffix[0] = '1'; if(M==1) suffix[0] = '1';
if(N==1) suffix[1] = '1'; if(N==1) suffix[1] = '1';