AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add support for AQL dimensions (#262)

Browse Source 
* Add support for AQL dimension changes

Adds support for returning dimensions from AQLProfile through rocprofiler
to tools. Includes a much larger expanded test suite that covers nearly
all files in counter collection.

Specific changes below:

samples/counter_collection/print_functional_counters: Modified to check
the validity of dimensions returned in comparison to the actual underlying
data obtained from a kernel execution.

rocprofiler-sdk/aql/helpers: adds function calls to support fetching
dimension information from AQLProfile.

rocprofiler-sdk/aql/packet_construct: modified to allow for events
to be exported to aid evaluate_ast in decoding the output buffer.

lib/rocprofiler-sdk/counters: Instance count now derived from dimension
sizes. rocprofiler_query_counter_dimensions now moved to a callback format
to improve usability.

rocprofiler-sdk/counters/core: Code migrations and exports of functions
for testing.

rocprofiler-sdk/counters/dimensions: Generates a dimension cache to be
used when querying dimension information for a counter id.

rocprofiler-sdk/counters/evaluate_ast: Modified to pass back correct
dimension information and to check/determine output dimensions for derived
counters.

rocprofiler-sdk/counters/id_decode: Modified to have a map between
dimension name -> dimension along with a conversion from the aql profile
id for a dimension (string) -> integer based id (happens only once during
init).

rocprofiler-sdk/hsa/queue: Modified to allow for making testing easier.
Specifically to allow Queue to now be mocked in unit tests for counter
collection.

* Merge with changes for serialization

* Added suggestions

* source formatting (clang-format v11) (#457)

Co-authored-by: bwelton <bwelton@users.noreply.github.com>

* Minor fix

* Test change

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: bwelton <bwelton@users.noreply.github.com>
This commit is contained in:
Benjamin Welton
2024-02-07 20:03:21 -08:00
committed by GitHub
parent 8a25b239bc
commit 3eb6a27bc6
29 changed files with 1670 additions and 293 deletions
Download Patch File Download Diff File Expand all files Collapse all files
source/lib/rocprofiler-sdk/counters/dimensions.cpp
+68 -19
View File
@@ -29,10 +29,12 @@
#include <fmt/core.h>
#include "lib/common/static_object.hpp"
#include "lib/common/synchronized.hpp"
#include "lib/common/utility.hpp"
#include "lib/rocprofiler-sdk/aql/helpers.hpp"
#include "lib/rocprofiler-sdk/aql/packet_construct.hpp"
#include "lib/rocprofiler-sdk/counters/evaluate_ast.hpp"
#include "lib/rocprofiler-sdk/hsa/queue_controller.hpp"
namespace rocprofiler
@@ -45,35 +47,82 @@ getBlockDimensions(std::string_view agent, const Metric& metric)
if(!metric.special().empty())
{
// Special non-hardware counters without dimension data
return std::vector<MetricDimension>{
{dimension_map().at(ROCPROFILER_DIMENSION_NONE), 1, ROCPROFILER_DIMENSION_NONE}};
return std::vector<MetricDimension>{{dimension_map().at(ROCPROFILER_DIMENSION_INSTANCE),
1,
ROCPROFILER_DIMENSION_INSTANCE}};
}
std::unordered_map<rocprofiler_profile_counter_instance_types, uint64_t> count;
std::vector<MetricDimension> ret;
for(const auto& [_, maybe_agent] : hsa::get_queue_controller().get_supported_agents())
{
if(maybe_agent.name() == agent)
{
// To be returned when instance counting is functional with AQL profiler
// return std::vector<MetricDimension>{
// {dimension_map().at(ROCPROFILER_DIMENSION_SHADER_ENGINE),
// maybe_agent.get_rocp_agent()->num_shader_banks,
// ROCPROFILER_DIMENSION_SHADER_ENGINE},
// {dimension_map().at(ROCPROFILER_DIMENSION_XCC),
// maybe_agent.get_rocp_agent()->num_xcc,
// ROCPROFILER_DIMENSION_XCC},
// {dimension_map().at(ROCPROFILER_DIMENSION_CU),
// maybe_agent.get_rocp_agent()->cu_count,
// ROCPROFILER_DIMENSION_CU},
// {dimension_map().at(ROCPROFILER_DIMENSION_AGENT),
// maybe_agent.get_rocp_agent()->id.handle,
// ROCPROFILER_DIMENSION_AGENT}};
aql::AQLPacketConstruct pkt_gen(maybe_agent, {metric});
return std::vector<MetricDimension>{
{metric.block(), pkt_gen.get_all_events().size(), ROCPROFILER_DIMENSION_NONE}};
const auto& events = pkt_gen.get_counter_events(metric);
for(const auto& event : events)
{
std::map<int, uint64_t> dims;
auto status = aql::get_dim_info(maybe_agent.get_hsa_agent(), event, 0, dims);
CHECK_EQ(status, ROCPROFILER_STATUS_SUCCESS)
<< rocprofiler_get_status_string(status);
for(const auto& [id, extent] : dims)
{
if(const auto* inst_type =
rocprofiler::common::get_val(aqlprofile_id_to_rocprof_instance(), id))
{
count.emplace(*inst_type, 0).first->second = extent;
}
else
{
LOG(ERROR) << "Unknown AQL Profiler Dimension " << id << " " << extent;
}
}
}
}
}
return {};
ret.reserve(count.size());
for(const auto& [dim, size] : count)
{
ret.emplace_back(dimension_map().at(dim), size, dim);
}
return ret;
}
const std::unordered_map<uint64_t, std::vector<MetricDimension>>&
get_dimension_cache()
{
static auto*& cache =
common::static_object<std::unordered_map<uint64_t, std::vector<MetricDimension>>>::
construct([]() -> std::unordered_map<uint64_t, std::vector<MetricDimension>> {
std::unordered_map<uint64_t, std::vector<MetricDimension>> dims;
/**
* Fails if HSA is not loaded by retruning nothing. This should not remain after
* AQL is transistioned away from HSA.
*/
if(rocprofiler::hsa::get_queue_controller().get_supported_agents().empty())
{
return {};
}
const auto& asts = counters::get_ast_map();
for(const auto& [gfx, metrics] : asts)
{
for(const auto& [_, ast] : metrics)
{
auto ast_copy = ast;
dims.emplace(ast.out_id().handle, ast_copy.set_dimensions());
}
}
return dims;
}());
return *cache;
}
} // namespace counters