Host trap PC sampling uses new record type (#1207)

* Host trap PC sampling uses new record type

* removing redundant field

* formatting

* simplifying templates in the parser - no need for HostTrap boolean

* reviving some parser tests

* hw_id decoding on GFX9

* HW id parser test

* parser CID test

* Parser multigpu test

* removing rocprofiler_pc_sampling_record_t and some fields from hw_id

* simplifying parser context

* keep bench test internally

* initializing gfx9_hw_id_t differently

* anonymous struct first

* avoiding inlining initialization of struct
Этот коммит содержится в:
Vladimir Indic
2024-11-20 21:02:47 +01:00
коммит произвёл GitHub
родитель 55aea3ef10
Коммит bc52c17e64
21 изменённых файлов: 1319 добавлений и 738 удалений
+335 -193
Просмотреть файл
@@ -100,14 +100,15 @@
ARBCHECK2(ISSUE_EXP); \
ARBCHECK2(ISSUE_MISC);
template <typename PcSamplingRecordT>
class WaveSnapTest
{
public:
WaveSnapTest()
{
buffer = std::make_shared<MockRuntimeBuffer>();
queue = std::make_shared<MockQueue>(16, buffer);
dispatch = std::make_shared<MockDispatch>(queue);
buffer = std::make_shared<MockRuntimeBuffer<PcSamplingRecordT>>();
queue = std::make_shared<MockQueue<PcSamplingRecordT>>(16, buffer);
dispatch = std::make_shared<MockDispatch<PcSamplingRecordT>>(queue);
}
void Test()
@@ -140,25 +141,27 @@ public:
dispatch->submit(packet_union_t{.snap = snap});
};
std::shared_ptr<MockRuntimeBuffer> buffer;
std::shared_ptr<MockQueue> queue;
std::shared_ptr<MockDispatch> dispatch;
std::shared_ptr<MockRuntimeBuffer<PcSamplingRecordT>> buffer;
std::shared_ptr<MockQueue<PcSamplingRecordT>> queue;
std::shared_ptr<MockDispatch<PcSamplingRecordT>> dispatch;
};
class WaveCntTest : public WaveSnapTest
template <typename PcSamplingRecordT>
class WaveCntTest : public WaveSnapTest<PcSamplingRecordT>
{
public:
void FillBuffers() override
{
// Loop over all possible wave_cnt
buffer->genUpcomingSamples(max_wave_number);
this->buffer->genUpcomingSamples(max_wave_number);
for(size_t i = 0; i < max_wave_number; i++)
genPCSample(i, GFX9::TYPE_LDS, GFX9::REASON_ALU, GFX9::ISSUE_VALU, GFX9::ISSUE_VALU);
this->genPCSample(
i, GFX9::TYPE_LDS, GFX9::REASON_ALU, GFX9::ISSUE_VALU, GFX9::ISSUE_VALU);
}
void CheckBuffers() override
{
auto parsed = buffer->get_parsed_buffer(9); // GFXIP==9
auto parsed = this->buffer->get_parsed_buffer(9); // GFXIP==9
assert(parsed.size() == 1);
assert(parsed[0].size() == max_wave_number);
@@ -166,204 +169,336 @@ public:
assert(parsed[0][i].wave_count == i);
}
const size_t max_wave_number = 64;
std::vector<rocprofiler_pc_sampling_snapshot_v1_t> snapshots;
const size_t max_wave_number = 64;
std::vector<PcSamplingRecordT> snapshots;
};
class InstTypeTest : public WaveSnapTest
// class InstTypeTest : public WaveSnapTest
// {
// public:
// void FillBuffers() override
// {
// // Loop over inst_type_issued
// UNROLL_TYPECHECK();
// buffer->genUpcomingSamples(GFX9::TYPE_LAST);
// for(int i = 0; i < GFX9::TYPE_LAST; i++)
// genPCSample(i, i, GFX9::REASON_ALU, GFX9::ISSUE_MATRIX, GFX9::ISSUE_MATRIX);
// }
// void CheckBuffers() override
// {
// auto parsed = buffer->get_parsed_buffer(9); // GFXIP==9
// assert(parsed.size() == 1);
// assert(parsed[0].size() == GFX9::TYPE_LAST);
// assert(snapshots.size() == GFX9::TYPE_LAST);
// for(size_t i = 0; i < GFX9::TYPE_LAST; i++)
// assert(snapshots[i].inst_type == parsed[0][i].snapshot.inst_type);
// }
// std::vector<rocprofiler_pc_sampling_snapshot_v1_t> snapshots;
// };
// class StallReasonTest : public WaveSnapTest
// {
// public:
// void FillBuffers() override
// {
// // Loop over reason_not_issued
// UNROLL_REASONCHECK();
// buffer->genUpcomingSamples(GFX9::REASON_LAST);
// for(int i = 0; i < GFX9::REASON_LAST; i++)
// genPCSample(i, GFX9::TYPE_MATRIX, i, GFX9::ISSUE_MATRIX, GFX9::ISSUE_MATRIX);
// }
// void CheckBuffers() override
// {
// auto parsed = buffer->get_parsed_buffer(9); // GFXIP==9
// assert(parsed.size() == 1);
// assert(parsed[0].size() == GFX9::REASON_LAST);
// assert(snapshots.size() == GFX9::REASON_LAST);
// for(size_t i = 0; i < GFX9::REASON_LAST; i++)
// assert(snapshots[i].reason_not_issued == parsed[0][i].snapshot.reason_not_issued);
// }
// std::vector<rocprofiler_pc_sampling_snapshot_v1_t> snapshots;
// };
// class ArbStateTest : public WaveSnapTest
// {
// public:
// void FillBuffers() override
// {
// // Loop over arb_state_issue
// UNROLL_ARBCHECK();
// buffer->genUpcomingSamples(GFX9::ISSUE_LAST * GFX9::ISSUE_LAST);
// for(int i = 0; i < GFX9::ISSUE_LAST; i++)
// for(int j = 0; j < GFX9::ISSUE_LAST; j++)
// genPCSample(i, GFX9::TYPE_MATRIX, GFX9::REASON_ALU, 1 << i, 1 << j);
// }
// void CheckBuffers() override
// {
// auto parsed = buffer->get_parsed_buffer(9); // GFXIP==9
// assert(parsed.size() == 1);
// assert(parsed[0].size() == GFX9::ISSUE_LAST * GFX9::ISSUE_LAST);
// assert(snapshots.size() == GFX9::ISSUE_LAST * GFX9::ISSUE_LAST);
// for(size_t i = 0; i < GFX9::ISSUE_LAST * GFX9::ISSUE_LAST; i++)
// {
// auto& snap = snapshots[i];
// assert(snap.arb_state_issue == parsed[0][i].snapshot.arb_state_issue);
// assert(snap.arb_state_stall == parsed[0][i].snapshot.arb_state_stall);
// }
// }
// std::vector<rocprofiler_pc_sampling_snapshot_v1_t> snapshots;
// };
// class WaveIssueAndErrorTest : public WaveSnapTest
// {
// void FillBuffers() override
// {
// buffer->genUpcomingSamples(16);
// for(int valid = 0; valid <= 1; valid++)
// for(int issued = 0; issued <= 1; issued++)
// for(int dual = 0; dual <= 1; dual++)
// for(int error = 0; error <= 1; error++)
// genPCSample(valid, issued, dual, error);
// }
// void CheckBuffers() override
// {
// const int num_combinations = 16;
// auto parsed = buffer->get_parsed_buffer(9); // GFXIP==9
// assert(parsed.size() == 1);
// assert(parsed[0].size() == num_combinations);
// assert(compare.size() == num_combinations);
// for(size_t i = 0; i < num_combinations; i++)
// {
// assert(compare[i].flags.valid == parsed[0][i].flags.valid);
// assert(compare[i].wave_issued == parsed[0][i].wave_issued);
// assert(compare[i].snapshot.dual_issue_valu == parsed[0][i].snapshot.dual_issue_valu);
// }
// }
// union trap_snapshot_v1
// {
// struct
// {
// uint32_t valid : 1;
// uint32_t issued : 1;
// uint32_t dual : 1;
// uint32_t reserved : 23;
// uint32_t error : 1;
// uint32_t reserved2 : 5;
// };
// uint32_t raw;
// };
// void genPCSample(bool valid, bool issued, bool dual, bool error)
// {
// rocprofiler_pc_sampling_record_t sample;
// ::memset(&sample, 0, sizeof(sample));
// // TODO: Since code objects are not mocked, use pc.code_object_offset
// // as the absolute physical address of the mocked PC.
// sample.pc.code_object_offset = dispatch->unique_id;
// sample.correlation_id.internal = dispatch->getMockId().raw;
// sample.flags.valid = valid && !error;
// sample.wave_issued = issued;
// sample.snapshot.dual_issue_valu = dual;
// assert(dispatch.get());
// compare.push_back(sample);
// trap_snapshot_v1 snap;
// snap.valid = valid;
// snap.issued = issued;
// snap.dual = dual;
// snap.error = error;
// perf_sample_snapshot_v1 pss;
// pss.perf_snapshot_data = snap.raw;
// pss.correlation_id = dispatch->getMockId().raw;
// dispatch->submit(std::move(pss));
// };
// std::vector<rocprofiler_pc_sampling_record_t> compare;
// };
template <typename PcSamplingRecordT>
class HwIdTest : public WaveSnapTest<PcSamplingRecordT>
{
public:
void FillBuffers() override
{
// Loop over inst_type_issued
UNROLL_TYPECHECK();
buffer->genUpcomingSamples(GFX9::TYPE_LAST);
for(int i = 0; i < GFX9::TYPE_LAST; i++)
genPCSample(i, i, GFX9::REASON_ALU, GFX9::ISSUE_MATRIX, GFX9::ISSUE_MATRIX);
}
void CheckBuffers() override
{
auto parsed = buffer->get_parsed_buffer(9); // GFXIP==9
assert(parsed.size() == 1);
assert(parsed[0].size() == GFX9::TYPE_LAST);
assert(snapshots.size() == GFX9::TYPE_LAST);
for(size_t i = 0; i < GFX9::TYPE_LAST; i++)
assert(snapshots[i].inst_type == parsed[0][i].snapshot.inst_type);
}
std::vector<rocprofiler_pc_sampling_snapshot_v1_t> snapshots;
};
class StallReasonTest : public WaveSnapTest
{
public:
void FillBuffers() override
{
// Loop over reason_not_issued
UNROLL_REASONCHECK();
buffer->genUpcomingSamples(GFX9::REASON_LAST);
for(int i = 0; i < GFX9::REASON_LAST; i++)
genPCSample(i, GFX9::TYPE_MATRIX, i, GFX9::ISSUE_MATRIX, GFX9::ISSUE_MATRIX);
}
void CheckBuffers() override
{
auto parsed = buffer->get_parsed_buffer(9); // GFXIP==9
assert(parsed.size() == 1);
assert(parsed[0].size() == GFX9::REASON_LAST);
assert(snapshots.size() == GFX9::REASON_LAST);
for(size_t i = 0; i < GFX9::REASON_LAST; i++)
assert(snapshots[i].reason_not_issued == parsed[0][i].snapshot.reason_not_issued);
}
std::vector<rocprofiler_pc_sampling_snapshot_v1_t> snapshots;
};
class ArbStateTest : public WaveSnapTest
{
public:
void FillBuffers() override
{
// Loop over arb_state_issue
UNROLL_ARBCHECK();
buffer->genUpcomingSamples(GFX9::ISSUE_LAST * GFX9::ISSUE_LAST);
for(int i = 0; i < GFX9::ISSUE_LAST; i++)
for(int j = 0; j < GFX9::ISSUE_LAST; j++)
genPCSample(i, GFX9::TYPE_MATRIX, GFX9::REASON_ALU, 1 << i, 1 << j);
}
void CheckBuffers() override
{
auto parsed = buffer->get_parsed_buffer(9); // GFXIP==9
assert(parsed.size() == 1);
assert(parsed[0].size() == GFX9::ISSUE_LAST * GFX9::ISSUE_LAST);
assert(snapshots.size() == GFX9::ISSUE_LAST * GFX9::ISSUE_LAST);
for(size_t i = 0; i < GFX9::ISSUE_LAST * GFX9::ISSUE_LAST; i++)
{
auto& snap = snapshots[i];
assert(snap.arb_state_issue == parsed[0][i].snapshot.arb_state_issue);
assert(snap.arb_state_stall == parsed[0][i].snapshot.arb_state_stall);
}
}
std::vector<rocprofiler_pc_sampling_snapshot_v1_t> snapshots;
};
class WaveIssueAndErrorTest : public WaveSnapTest
{
void FillBuffers() override
{
buffer->genUpcomingSamples(16);
for(int valid = 0; valid <= 1; valid++)
for(int issued = 0; issued <= 1; issued++)
for(int dual = 0; dual <= 1; dual++)
for(int error = 0; error <= 1; error++)
genPCSample(valid, issued, dual, error);
}
void CheckBuffers() override
{
const int num_combinations = 16;
auto parsed = buffer->get_parsed_buffer(9); // GFXIP==9
assert(parsed.size() == 1);
assert(parsed[0].size() == num_combinations);
assert(compare.size() == num_combinations);
for(size_t i = 0; i < num_combinations; i++)
{
assert(compare[i].flags.valid == parsed[0][i].flags.valid);
assert(compare[i].wave_issued == parsed[0][i].wave_issued);
assert(compare[i].snapshot.dual_issue_valu == parsed[0][i].snapshot.dual_issue_valu);
}
}
union trap_snapshot_v1
union gfx9_hw_id_t
{
uint32_t raw;
struct
{
uint32_t valid : 1;
uint32_t issued : 1;
uint32_t dual : 1;
uint32_t reserved : 23;
uint32_t error : 1;
uint32_t reserved2 : 5;
uint32_t wave_id : 4; ///< wave slot index
uint32_t simd_id : 2; ///< SIMD index
uint32_t pipe_id : 2; ///< pipe index
uint32_t cu_id : 4; ///< Index of compute unit on GFX9 or workgroup processer on other
///< architectures
uint32_t shader_array_id : 1; ///< Shared array index
uint32_t shader_engine_id : 3; ///< shared engine index
uint32_t
threadgroup_id : 4; ///< thread_group index on GFX9, and workgroup index on GFX10+
uint32_t vm_id : 4; ///< virtual memory ID
uint32_t queue_id : 3; ///< queue id
uint32_t gfx_context_state_id : 3; ///< GFX context (state) id (only on GFX9) - ignored
uint32_t microengine_id : 2; ///< ACE (microengine) index
};
uint32_t raw;
};
void genPCSample(bool valid, bool issued, bool dual, bool error)
{
rocprofiler_pc_sampling_record_t sample;
::memset(&sample, 0, sizeof(sample));
// TODO: Since code objects are not mocked, use pc.loaded_code_object_offset
// as the absolute physical address of the mocked PC.
sample.pc.loaded_code_object_offset = dispatch->unique_id;
sample.correlation_id.internal = dispatch->getMockId().raw;
sample.flags.valid = valid && !error;
sample.wave_issued = issued;
sample.snapshot.dual_issue_valu = dual;
assert(dispatch.get());
compare.push_back(sample);
trap_snapshot_v1 snap;
snap.valid = valid;
snap.issued = issued;
snap.dual = dual;
snap.error = error;
perf_sample_snapshot_v1 pss;
pss.perf_snapshot_data = snap.raw;
pss.correlation_id = dispatch->getMockId().raw;
dispatch->submit(std::move(pss));
};
std::vector<rocprofiler_pc_sampling_record_t> compare;
};
class WaveOtherFieldsTest : public WaveSnapTest
{
void FillBuffers() override
{
buffer->genUpcomingSamples(3);
genPCSample(1, 2, 3, 4, 5, 6, 7, 8); // Counting
genPCSample(3, 5, 7, 11, 13, 17, 19, 23); // Some prime numbers
genPCSample(23, 19, 17, 13, 11, 7, 5, 3); // Some reversed primes
gfx9_hw_id_t hw_id_val0;
hw_id_val0.wave_id = 0;
hw_id_val0.simd_id = 0;
hw_id_val0.pipe_id = 0;
hw_id_val0.cu_id = 0;
hw_id_val0.shader_array_id = 0;
hw_id_val0.shader_engine_id = 0;
hw_id_val0.threadgroup_id = 0;
hw_id_val0.vm_id = 0;
hw_id_val0.queue_id = 0;
hw_id_val0.gfx_context_state_id = 0;
hw_id_val0.microengine_id = 0;
gfx9_hw_id_t hw_id_val1;
hw_id_val0.wave_id = 15;
hw_id_val0.simd_id = 3;
hw_id_val0.pipe_id = 3;
hw_id_val0.cu_id = 15;
hw_id_val0.shader_array_id = 1;
hw_id_val0.shader_engine_id = 7;
hw_id_val0.threadgroup_id = 15;
hw_id_val0.vm_id = 15;
hw_id_val0.queue_id = 7;
hw_id_val0.gfx_context_state_id = 7;
hw_id_val0.microengine_id = 3;
gfx9_hw_id_t hw_id_val2;
hw_id_val2.wave_id = 7;
hw_id_val2.simd_id = 2;
hw_id_val2.pipe_id = 2;
hw_id_val2.cu_id = 6;
hw_id_val2.shader_array_id = 0;
hw_id_val2.shader_engine_id = 3;
hw_id_val2.threadgroup_id = 8;
hw_id_val2.vm_id = 9;
hw_id_val2.queue_id = 3;
hw_id_val2.gfx_context_state_id = 2;
hw_id_val2.microengine_id = 1;
this->buffer->genUpcomingSamples(3);
genPCSample(hw_id_val0);
genPCSample(hw_id_val1);
genPCSample(hw_id_val2);
}
void CheckBuffers() override
{
auto parsed = buffer->get_parsed_buffer(9); // GFXIP==9
auto parsed = this->buffer->get_parsed_buffer(9); // GFXIP==9
assert(parsed.size() == 1);
assert(parsed[0].size() == 3);
assert(compare.size() == 3);
for(size_t i = 0; i < 3; i++)
{
assert(parsed[0][i].flags.has_stall_reason == true);
assert(parsed[0][i].flags.has_wave_cnt == true);
assert(parsed[0][i].flags.reserved == false);
// Comparing individual fields
assert(compare[i].hw_id.wave_id == parsed[0][i].hw_id.wave_id);
assert(compare[i].hw_id.simd_id == parsed[0][i].hw_id.simd_id);
assert(compare[i].hw_id.pipe_id == parsed[0][i].hw_id.pipe_id);
assert(compare[i].hw_id.cu_or_wgp_id == parsed[0][i].hw_id.cu_or_wgp_id);
assert(compare[i].hw_id.shader_array_id == parsed[0][i].hw_id.shader_array_id);
assert(compare[i].hw_id.shader_engine_id == parsed[0][i].hw_id.shader_engine_id);
assert(compare[i].hw_id.workgroup_id == parsed[0][i].hw_id.workgroup_id);
assert(compare[i].hw_id.vm_id == parsed[0][i].hw_id.vm_id);
assert(compare[i].hw_id.queue_id == parsed[0][i].hw_id.queue_id);
assert(compare[i].hw_id.microengine_id == parsed[0][i].hw_id.microengine_id);
}
}
void genPCSample(gfx9_hw_id_t hw_id)
{
PcSamplingRecordT sample;
::memset(&sample, 0, sizeof(sample));
// Unpacking individual fields
// NOTE: chiplet is tested in a WaveOtherFieldsTest test, becuase it's not
// transferred via hw_id, but chiplet_and_wave_id field.
sample.hw_id.wave_id = hw_id.wave_id;
sample.hw_id.simd_id = hw_id.simd_id;
sample.hw_id.pipe_id = hw_id.pipe_id;
sample.hw_id.cu_or_wgp_id = hw_id.cu_id;
sample.hw_id.shader_array_id = hw_id.shader_array_id;
sample.hw_id.shader_engine_id = hw_id.shader_engine_id;
sample.hw_id.workgroup_id = hw_id.threadgroup_id;
sample.hw_id.vm_id = hw_id.vm_id;
sample.hw_id.queue_id = hw_id.queue_id;
sample.hw_id.microengine_id = hw_id.microengine_id;
compare.push_back(sample);
perf_sample_snapshot_v1 snap;
::memset(&snap, 0, sizeof(snap));
// raw register value
snap.hw_id = hw_id.raw;
snap.correlation_id = this->dispatch->getMockId().raw;
assert(this->dispatch.get());
this->dispatch->submit(snap);
};
std::vector<PcSamplingRecordT> compare;
};
template <typename PcSamplingRecordT>
class WaveOtherFieldsTest : public WaveSnapTest<PcSamplingRecordT>
{
void FillBuffers() override
{
this->buffer->genUpcomingSamples(3);
genPCSample(1, 2, 3, 4, 5, 6, 7); // Counting
genPCSample(3, 5, 7, 11, 13, 17, 19); // Some prime numbers
genPCSample(23, 19, 17, 13, 11, 7, 5); // Some reversed primes
}
void CheckBuffers() override
{
auto parsed = this->buffer->get_parsed_buffer(9); // GFXIP==9
assert(parsed.size() == 1);
assert(parsed[0].size() == 3);
assert(compare.size() == 3);
for(size_t i = 0; i < 3; i++)
{
// TODO: if we decide to test flags, make specialization for
// rocprofiler_pc_sampling_record_stochastic_v0_t
// assert(parsed[0][i].flags.has_stall_reason == true);
// assert(parsed[0][i].flags.has_wave_cnt == true);
// assert(parsed[0][i].flags.reserved == false);
assert(compare[i].exec_mask == parsed[0][i].exec_mask);
assert(compare[i].workgroup_id == parsed[0][i].workgroup_id);
assert(compare[i].chiplet == parsed[0][i].chiplet);
assert(compare[i].wave_id == parsed[0][i].wave_id);
assert(compare[i].hw_id == parsed[0][i].hw_id);
assert(compare[i].hw_id.chiplet == parsed[0][i].hw_id.chiplet);
assert(compare[i].wave_in_group == parsed[0][i].wave_in_group);
// TODO: handle HW_ID as well.
// assert(compare[i].hw_id == parsed[0][i].hw_id);
assert(compare[i].correlation_id.internal == parsed[0][i].correlation_id.internal);
}
}
void genPCSample(int pc, int exec, int blkx, int blky, int blkz, int chip, int wave, int hwid)
void genPCSample(int pc, int exec, int blkx, int blky, int blkz, int chip, int wave)
{
rocprofiler_pc_sampling_record_t sample;
PcSamplingRecordT sample;
::memset(&sample, 0, sizeof(sample));
sample.exec_mask = exec;
@@ -371,13 +506,15 @@ class WaveOtherFieldsTest : public WaveSnapTest
sample.workgroup_id.y = blky;
sample.workgroup_id.z = blkz;
sample.chiplet = chip;
sample.wave_id = wave;
sample.hw_id = hwid;
sample.correlation_id.internal = dispatch->unique_id;
sample.hw_id.chiplet = chip;
sample.wave_in_group = wave;
sample.correlation_id.internal = this->dispatch->unique_id;
compare.push_back(sample);
// We're testing fields commong for both perf_sample_host_trap_v1 and
// perf_sample_snapshot_v1, so either struct is suitable here. No need to make
// specialization,
perf_sample_snapshot_v1 snap;
::memset(&snap, 0, sizeof(snap));
snap.exec_mask = exec;
@@ -386,26 +523,31 @@ class WaveOtherFieldsTest : public WaveSnapTest
snap.workgroup_id_y = blky;
snap.workgroup_id_z = blkz;
snap.chiplet_and_wave_id = (chip << 8) | (wave & 0x3F);
snap.hw_id = hwid;
snap.correlation_id = dispatch->getMockId().raw;
snap.correlation_id = this->dispatch->getMockId().raw;
assert(dispatch.get());
dispatch->submit(snap);
assert(this->dispatch.get());
this->dispatch->submit(snap);
(void) pc;
};
std::vector<rocprofiler_pc_sampling_record_t> compare;
std::vector<PcSamplingRecordT> compare;
};
TEST(pcs_parser, gfx9_test)
{
WaveCntTest{}.Test();
InstTypeTest{}.Test();
StallReasonTest{}.Test();
ArbStateTest{}.Test();
WaveIssueAndErrorTest{}.Test();
WaveOtherFieldsTest{}.Test();
// Tests specific to stochastic sampling only
WaveCntTest<rocprofiler_pc_sampling_record_stochastic_v0_t>{}.Test();
// InstTypeTest{}.Test();
// StallReasonTest{}.Test();
// ArbStateTest{}.Test();
// WaveIssueAndErrorTest{}.Test();
// Tests commong for both host trap and stochastic sampling.
HwIdTest<rocprofiler_pc_sampling_record_host_trap_v0_t>{}.Test();
HwIdTest<rocprofiler_pc_sampling_record_stochastic_v0_t>{}.Test();
WaveOtherFieldsTest<rocprofiler_pc_sampling_record_host_trap_v0_t>{}.Test();
WaveOtherFieldsTest<rocprofiler_pc_sampling_record_stochastic_v0_t>{}.Test();
std::cout << "GFX9 Test Done." << std::endl;
}