Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

TPG Threshold By Plane #177

Merged
merged 5 commits into from
May 14, 2024
Merged

TPG Threshold By Plane #177

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
808497d
Developing TPG threshold by plane
adam-abed-abud Apr 11, 2024
b3957e1
Added plane threshold config setting.
aeoranday Apr 25, 2024
1f10ee9
Added NAIVE thresholding per plane (channel).
aeoranday Apr 25, 2024
a411346
Fixed the plane numbering.
aeoranday May 6, 2024
119fc48
Updated TPG threshold word chioce for planes 0,1,2.
aeoranday May 9, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 12 additions & 4 deletions include/fdreadoutlibs/wibeth/WIBEthFrameProcessor.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,7 +58,7 @@ class WIBEthFrameHandler {

void reset();

void initialize(uint16_t threshold_value, uint16_t memory_factor, uint16_t scale_factor, int16_t frug_streaming_acclimt);
void initialize(uint16_t memory_factor, uint16_t scale_factor, int16_t frug_streaming_acclimt);

uint16_t* get_hits_dest();

Expand Down Expand Up @@ -146,7 +146,7 @@ class WIBEthFrameProcessor : public readoutlibs::TaskRawDataProcessorModel<types
private:
bool m_tpg_enabled;

bool m_enable_simple_threshold_on_collection = false;
bool m_enable_simple_threshold_on_plane2 = false;
// Selected TPG algorithm properties from configuration
std::string m_tpg_algorithm;
uint16_t m_tpg_rs_memory_factor;
Expand All @@ -157,7 +157,10 @@ class WIBEthFrameProcessor : public readoutlibs::TaskRawDataProcessorModel<types
uint32_t m_tp_max_width;
std::vector<int> m_channel_mask_vec;
std::set<uint> m_channel_mask_set;
uint16_t m_tpg_threshold;

uint16_t m_tpg_threshold_plane0;
uint16_t m_tpg_threshold_plane1;
uint16_t m_tpg_threshold_plane2;

// Algorithm used to form a trigger primitive
dunedaq::trgdataformats::TriggerPrimitive::Algorithm m_tp_algo = trgdataformats::TriggerPrimitive::Algorithm::kUnknown;
Expand All @@ -181,10 +184,15 @@ class WIBEthFrameProcessor : public readoutlibs::TaskRawDataProcessorModel<types


// Create an array to store the values of the memory factor
// AAA: silver bullet to be able to use SimpleThreshold on collection and RS on induction planes
// AAA: silver bullet to be able to use SimpleThreshold on plane 2 and RS on planes 0 and 1
// By default set all the values to the selected memory factor
std::array<uint16_t, swtpg_wibeth::NUM_REGISTERS_PER_FRAME * swtpg_wibeth::SAMPLES_PER_REGISTER> m_register_memory_factor = {0};

// Create an array to store the values of the TPG threshold
// This is to be used for setting a different value by plane
// By default set it to 150
std::array<uint16_t, swtpg_wibeth::NUM_REGISTERS_PER_FRAME * swtpg_wibeth::SAMPLES_PER_REGISTER> m_tpg_threshold = {150};


std::function<void(swtpg_wibeth::ProcessingInfo<swtpg_wibeth::NUM_REGISTERS_PER_FRAME>& info)> m_assigned_tpg_algorithm_function;

Expand Down
7 changes: 4 additions & 3 deletions include/fdreadoutlibs/wibeth/tpg/ProcessAVX2.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@ process_window_avx2(ProcessingInfo<NREGISTERS>& info)
// Pointer to keep track of where we'll write the next output hit
__m256i* output_loc = (__m256i*)(info.output); // NOLINT(readability/casting)

const __m256i iota = _mm256_set_epi16(15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0);
const __m256i iota = _mm256_lddqu_si256((__m256i*) IOTA);

int nhits = 0;

Expand All @@ -42,6 +42,8 @@ process_window_avx2(ProcessingInfo<NREGISTERS>& info)
// from the previous go-around.

ChanState<NREGISTERS>& state = info.chanState;
__m256i threshold = _mm256_lddqu_si256(reinterpret_cast<__m256i*>(state.threshold) + ireg); // NOLINT

__m256i median = _mm256_lddqu_si256(reinterpret_cast<__m256i*>(state.pedestals) + ireg); // NOLINT
// The accumulator that we increase/decrease when the current
// sample is greater/less than the median
Expand Down Expand Up @@ -93,8 +95,7 @@ process_window_avx2(ProcessingInfo<NREGISTERS>& info)
// --------------------------------------------------------------
// Mask for channels that are over the threshold in this step

// FIXED THRESHOLD
__m256i threshold = _mm256_set1_epi16(info.threshold);
// Define a register for elements above the threhsold
__m256i is_over = _mm256_cmpgt_epi16(s, threshold);

// Mask for channels that left "over threshold" state this step
Expand Down
7 changes: 4 additions & 3 deletions include/fdreadoutlibs/wibeth/tpg/ProcessAbsRSAVX2.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,7 +39,7 @@ process_window_rs_avx2(ProcessingInfo<NREGISTERS>& info)
// Pointer to keep track of where we'll write the next output hit
__m256i* output_loc = (__m256i*)(info.output); // NOLINT(readability/casting)

const __m256i iota = _mm256_set_epi16(15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0);
const __m256i iota = _mm256_lddqu_si256((__m256i*) IOTA);

int nhits = 0;

Expand All @@ -55,6 +55,8 @@ process_window_rs_avx2(ProcessingInfo<NREGISTERS>& info)
// from the previous go-around.

ChanState<NREGISTERS>& state = info.chanState;
__m256i threshold = _mm256_lddqu_si256(reinterpret_cast<__m256i*>(state.threshold) + ireg); // NOLINT

__m256i median = _mm256_lddqu_si256(reinterpret_cast<__m256i*>(state.pedestals) + ireg); // NOLINT
//__m256i quantile25 = _mm256_lddqu_si256(reinterpret_cast<__m256i*>(state.quantile25) + ireg); // NOLINT
//__m256i quantile75 = _mm256_lddqu_si256(reinterpret_cast<__m256i*>(state.quantile75) + ireg); // NOLINT
Expand Down Expand Up @@ -183,8 +185,7 @@ process_window_rs_avx2(ProcessingInfo<NREGISTERS>& info)
// IQR-based THRESHOLD
//__m256i is_over = _mm256_cmpgt_epi16(RS, sigma * info.threshold);

// FIXED THRESHOLD
__m256i threshold = _mm256_set1_epi16(info.threshold);
// Define a register for elements above the threhsold
__m256i is_over = _mm256_cmpgt_epi16(RS, threshold);


Expand Down
3 changes: 2 additions & 1 deletion include/fdreadoutlibs/wibeth/tpg/ProcessNaive.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,6 +63,7 @@ process_window_naive(ProcessingInfo<NREGISTERS>& info)
int16_t& accum = state.accum[ichan];

// Variables for hit finding
int16_t& threshold = state.threshold[ichan]; // Threshold for this channel.
uint16_t& prev_was_over = state.prev_was_over[ichan]; // was the previous sample over threshold?
uint16_t& hit_charge = state.hit_charge[ichan];
uint16_t& hit_tover = state.hit_tover[ichan]; // time over threshold
Expand Down Expand Up @@ -90,7 +91,7 @@ process_window_naive(ProcessingInfo<NREGISTERS>& info)
// --------------------------------------------------------------
// Hit finding
// --------------------------------------------------------------
bool is_over = sample > info.threshold;
bool is_over = sample > threshold;
//printf("% 5d % 5d % 5d % 5d\n", (uint16_t)ichan, (uint16_t)itime, sample, info.threshold); // NOLINT
if (is_over) {
// Simulate saturated add
Expand Down
3 changes: 2 additions & 1 deletion include/fdreadoutlibs/wibeth/tpg/ProcessNaiveRS.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -57,6 +57,7 @@ process_window_naive_RS(ProcessingInfo<NREGISTERS>& info)
//int16_t& quantile75 = state.quantile75[ichan];

// Variables for hit finding
int16_t& threshold = state.threshold[ichan]; // Threshold for this channel.
uint16_t& prev_was_over = state.prev_was_over[ichan]; // was the previous sample over threshold?
uint16_t& hit_charge = state.hit_charge[ichan];
uint16_t& hit_tover = state.hit_tover[ichan]; // time over threshold
Expand Down Expand Up @@ -116,7 +117,7 @@ process_window_naive_RS(ProcessingInfo<NREGISTERS>& info)
// --------------------------------------------------------------
// Hit finding
// --------------------------------------------------------------
bool is_over = RS > info.threshold;
bool is_over = RS > threshold;
if (is_over) {
// Simulate saturated add
int32_t tmp_charge = hit_charge;
Expand Down
9 changes: 5 additions & 4 deletions include/fdreadoutlibs/wibeth/tpg/ProcessStandardRSAVX2.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,15 +31,15 @@ process_window_standard_rs_avx2(ProcessingInfo<NREGISTERS>& info)

// Scaling factor to stop the ADCs from overflowing
// (may not needs this, depends on magnitude of FIR output)
const __m256i scale_factor = _mm256_set1_epi16(info.rs_scale_factor);
//const __m256i scale_factor = _mm256_set1_epi16(info.rs_scale_factor);

// The maximum value that sigma can have before the threshold overflows a 16-bit signed integer
//const __m256i sigmaMax = _mm256_set1_epi16((1 << 15) / (info.multiplier * info.threshold));

// Pointer to keep track of where we'll write the next output hit
__m256i* output_loc = (__m256i*)(info.output); // NOLINT(readability/casting)

const __m256i iota = _mm256_set_epi16(15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0);
const __m256i iota = _mm256_lddqu_si256((__m256i*) IOTA);

int nhits = 0;

Expand All @@ -55,6 +55,8 @@ process_window_standard_rs_avx2(ProcessingInfo<NREGISTERS>& info)
// from the previous go-around.

ChanState<NREGISTERS>& state = info.chanState;
__m256i threshold = _mm256_lddqu_si256(reinterpret_cast<__m256i*>(state.threshold) + ireg); // NOLINT

__m256i median = _mm256_lddqu_si256(reinterpret_cast<__m256i*>(state.pedestals) + ireg); // NOLINT
//__m256i quantile25 = _mm256_lddqu_si256(reinterpret_cast<__m256i*>(state.quantile25) + ireg); // NOLINT
//__m256i quantile75 = _mm256_lddqu_si256(reinterpret_cast<__m256i*>(state.quantile75) + ireg); // NOLINT
Expand Down Expand Up @@ -178,8 +180,7 @@ process_window_standard_rs_avx2(ProcessingInfo<NREGISTERS>& info)
// IQR-based THRESHOLD
//__m256i is_over = _mm256_cmpgt_epi16(RS, sigma * info.threshold);

// FIXED THRESHOLD
__m256i threshold = _mm256_set1_epi16(info.threshold);
// Define a register for elements above the threhsold
__m256i is_over = _mm256_cmpgt_epi16(RS, threshold);


Expand Down
17 changes: 14 additions & 3 deletions include/fdreadoutlibs/wibeth/tpg/ProcessingInfo.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,7 @@ struct ChanState
ChanState()
{
for (size_t i = 0; i < NREGISTERS * SAMPLES_PER_REGISTER; ++i) {
threshold[i] = 0;
pedestals[i] = 0;
accum[i] = 0;
RS[i] = 0;
Expand All @@ -39,6 +40,7 @@ struct ChanState
}
}

alignas(32) int16_t __restrict__ threshold[NREGISTERS * SAMPLES_PER_REGISTER];
alignas(32) int16_t __restrict__ pedestals[NREGISTERS * SAMPLES_PER_REGISTER];
alignas(32) int16_t __restrict__ accum[NREGISTERS * SAMPLES_PER_REGISTER];

Expand Down Expand Up @@ -74,7 +76,6 @@ struct ProcessingInfo
uint8_t last_register_, // NOLINT
uint16_t* __restrict__ output_, // NOLINT
const uint8_t exponent_, // NOLINT
uint16_t threshold_, // NOLINT
uint16_t rs_memory_factor_, // NOLINT
uint16_t rs_scale_factor_, // NOLINT
int16_t frugal_streaming_accumulator_limit_, // NOLINT
Expand All @@ -86,7 +87,6 @@ struct ProcessingInfo
, last_register(last_register_)
, output(output_)
, exponent(exponent_)
, threshold(threshold_)
, rs_memory_factor(rs_memory_factor_)
, rs_scale_factor(rs_scale_factor_)
, frugal_streaming_accumulator_limit(frugal_streaming_accumulator_limit_)
Expand All @@ -95,6 +95,18 @@ struct ProcessingInfo
, nhits(nhits_)
{}

// Set TPG threshold by plane
void setThresholdState(std::array<uint16_t, swtpg_wibeth::NUM_REGISTERS_PER_FRAME * swtpg_wibeth::SAMPLES_PER_REGISTER>& register_threshold
)
{
// Set the threshold values
for (size_t j = 0; j < NREGISTERS * SAMPLES_PER_REGISTER; ++j) {
const size_t i = IOTA[j % SAMPLES_PER_REGISTER];
const size_t reg_idx = j / SAMPLES_PER_REGISTER;
chanState.threshold[j] = register_threshold[i + reg_idx * SAMPLES_PER_REGISTER];
}

}

// Set the initial state from the window starting at first_msg_p
template<size_t N>
Expand Down Expand Up @@ -153,7 +165,6 @@ struct ProcessingInfo
uint8_t last_register; // NOLINT
uint16_t* __restrict__ output; // NOLINT
uint8_t exponent; // NOLINT
uint16_t threshold; // NOLINT
uint16_t rs_memory_factor; // NOLINT
uint16_t rs_scale_factor; // NOLINT
int16_t frugal_streaming_accumulator_limit; // NOLINT
Expand Down
3 changes: 3 additions & 0 deletions include/fdreadoutlibs/wibeth/tpg/TPGConstants_wibeth.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,9 @@ const constexpr std::size_t BYTES_PER_REGISTER = 32;
// How many samples are in a register
const constexpr std::size_t SAMPLES_PER_REGISTER = 16;

// Order of the channels after frame expansion.
const constexpr std::uint16_t IOTA[SAMPLES_PER_REGISTER] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15};

// One netio message's worth of channel ADCs after
// expansion: 64 frames per message times 4 registers per frame times
// 32 bytes (256 bits) per register
Expand Down
2 changes: 1 addition & 1 deletion src/wib2/WIB2FrameProcessor.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -209,7 +209,7 @@ WIB2FrameProcessor::conf(const nlohmann::json& cfg)
// AAA: The set provides faster look up than a std::vector
m_channel_mask_set.insert(m_channel_mask_vec.begin(), m_channel_mask_vec.end());

m_tpg_threshold_selected = config.tpg_threshold;
m_tpg_threshold_selected = config.tpg_threshold_default;

m_crate_no = config.crate_id;
m_slot_no = config.slot_id;
Expand Down
35 changes: 19 additions & 16 deletions src/wibeth/WIBEthFrameProcessor.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -72,7 +72,7 @@ WIBEthFrameHandler::reset()
}

void
WIBEthFrameHandler::initialize(uint16_t threshold_value, uint16_t memory_factor, uint16_t scale_factor, int16_t frug_streaming_acclimt)
WIBEthFrameHandler::initialize(uint16_t memory_factor, uint16_t scale_factor, int16_t frug_streaming_acclimt)
{

if(m_hits_dest == nullptr) {m_hits_dest = new uint16_t[100000];}
Expand All @@ -83,7 +83,6 @@ WIBEthFrameHandler::initialize(uint16_t threshold_value, uint16_t memory_factor,
swtpg_wibeth::NUM_REGISTERS_PER_FRAME,
m_hits_dest,
m_tpg_exponent,
threshold_value,
memory_factor,
scale_factor,
frug_streaming_acclimt,
Expand Down Expand Up @@ -116,8 +115,7 @@ WIBEthFrameProcessor::start(const nlohmann::json& args)
m_tps_suppressed_too_long = 0;
m_tps_send_failed = 0;

m_wibeth_frame_handler->initialize(m_tpg_threshold,
m_tpg_rs_memory_factor,
m_wibeth_frame_handler->initialize(m_tpg_rs_memory_factor,
m_tpg_rs_scale_factor,
m_tpg_frugal_streaming_accumulator_limit
);
Expand Down Expand Up @@ -185,13 +183,13 @@ WIBEthFrameProcessor::conf(const nlohmann::json& cfg)
} else if (m_tpg_algorithm == "AbsRS" ) {
m_tp_algo = trgdataformats::TriggerPrimitive::Algorithm::kAbsRunningSum;
m_assigned_tpg_algorithm_function = &swtpg_wibeth::process_window_rs_avx2<swtpg_wibeth::NUM_REGISTERS_PER_FRAME>;
// Enable simple threshold on collection is used only if you are using a Running Sum algorithm
m_enable_simple_threshold_on_collection = config.enable_simple_threshold_on_collection;
// Enable simple threshold on plane 2 is used only if you are using a Running Sum algorithm
m_enable_simple_threshold_on_plane2 = config.enable_simple_threshold_on_plane2;
} else if (m_tpg_algorithm == "StandardRS" ) {
m_tp_algo = trgdataformats::TriggerPrimitive::Algorithm::kRunningSum;
m_assigned_tpg_algorithm_function = &swtpg_wibeth::process_window_standard_rs_avx2<swtpg_wibeth::NUM_REGISTERS_PER_FRAME>;
// Enable simple threshold on collection is used only if you are using a Running Sum algorithm
m_enable_simple_threshold_on_collection = config.enable_simple_threshold_on_collection;
// Enable simple threshold on plane 2 is used only if you are using a Running Sum algorithm
m_enable_simple_threshold_on_plane2 = config.enable_simple_threshold_on_plane2;
} else {
throw TPGAlgorithmInexistent(ERS_HERE, m_tpg_algorithm);
}
Expand All @@ -217,7 +215,10 @@ WIBEthFrameProcessor::conf(const nlohmann::json& cfg)
// AAA: The set provides faster look up than a std::vector
m_channel_mask_set.insert(m_channel_mask_vec.begin(), m_channel_mask_vec.end());

m_tpg_threshold = config.tpg_threshold;
// Use config.<plane>_threshold if it is non-zero, else default to config.tpg_threshold.
m_tpg_threshold_plane2 = config.tpg_threshold_plane2 ? config.tpg_threshold_plane2 : config.tpg_threshold_default;
m_tpg_threshold_plane1 = config.tpg_threshold_plane1 ? config.tpg_threshold_plane1 : config.tpg_threshold_default;
m_tpg_threshold_plane0 = config.tpg_threshold_plane0 ? config.tpg_threshold_plane0 : config.tpg_threshold_default;

m_crate_no = config.crate_id;
m_slot_no = config.slot_id;
Expand Down Expand Up @@ -438,15 +439,16 @@ WIBEthFrameProcessor::find_hits(constframeptr fp, WIBEthFrameHandler* frame_hand
auto chan_value = frame_handler->register_channel_map.channel[i];
m_register_channels[i] = chan_value;

if (m_enable_simple_threshold_on_collection) {
// If the given channel is a collection then set R (memory factor) to zero
if (m_channel_map->get_plane_from_offline_channel(chan_value) == 0 ) {
m_register_memory_factor[i] = 0;
} else {
m_register_memory_factor[i] = m_tpg_rs_memory_factor;
}
if (m_channel_map->get_plane_from_offline_channel(chan_value) == 2 ) {
// If SimpleThreshold on plane 2, then set the memory factor to 0, else use the common memory factor.
m_register_memory_factor[i] = m_enable_simple_threshold_on_plane2 ? 0 : m_tpg_rs_memory_factor;
m_tpg_threshold[i] = m_tpg_threshold_plane2;
} else if (m_channel_map->get_plane_from_offline_channel(chan_value) == 1) {
m_register_memory_factor[i] = m_tpg_rs_memory_factor;
m_tpg_threshold[i] = m_tpg_threshold_plane1;
} else {
m_register_memory_factor[i] = m_tpg_rs_memory_factor;
m_tpg_threshold[i] = m_tpg_threshold_plane0;
}

//TLOG () << "Index number " << i << " offline channel " << frame_handler->register_channel_map.channel[i];
Expand All @@ -455,6 +457,7 @@ WIBEthFrameProcessor::find_hits(constframeptr fp, WIBEthFrameHandler* frame_hand
m_tp_channel_rate_map[frame_handler->register_channel_map.channel[i]] = 0;
}

frame_handler->m_tpg_processing_info->setThresholdState(m_tpg_threshold);
frame_handler->m_tpg_processing_info->setState(registers_array, m_register_memory_factor);


Expand Down