Purpose Monte Carlo simulation is indispensable in particle physics experiments, but explicitly sampling for very large stochastic particle populations could become a major computational bottleneck. Gas Electron Multiplier (GEM)-based detectors are a representative example: Cascade avalanche multiplication across multiple GEM foils can generate <InlineEquation ID="IEq1"> <EquationSource Format="TEX">$\mathcal {O}(10^7)$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi mathvariant="script">O</mi> <mo stretchy="false">(</mo> <msup> <mn>10</mn> <mn>7</mn> </msup> <mo stretchy="false">)</mo> </mrow> </math></EquationSource> </InlineEquation> electrons per track, and the stochastic transport and signal induction sampling or computation required for each electron make large-scale Monte Carlo production prohibitively expensive. Methods We present two acceleration methods, developed and validated for the digitization of the Cylindrical GEM (CGEM) detector at Beijing Spectrometer (BESIII). The first method, random number pool, pre-generates random numbers in contiguous memory and retrieves them through pointer arithmetic. Combined with batch processing, this approach improves cache locality. The second method, statistical scaling approximation, reduces the number of simulated electrons by a downscaling factor while amplifying back the magnitude in signal induction, thereby preserving signal fidelity. Results For the BESIII CGEM digitization software, the random number pool optimization together with batch processing accelerates the simulation by over <InlineEquation ID="IEq2"> <EquationSource Format="TEX">$4\times $</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>4</mn> <mo>×</mo> </mrow> </math></EquationSource> </InlineEquation> for <InlineEquation ID="IEq3"> <EquationSource Format="TEX">$\pi ^+$</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>π</mi> <mo>+</mo> </msup> </math></EquationSource> </InlineEquation> samples and <InlineEquation ID="IEq4"> <EquationSource Format="TEX">$5\times $</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>5</mn> <mo>×</mo> </mrow> </math></EquationSource> </InlineEquation> for proton samples. The statistical scaling approximation provides a reduction of the simulated electron population while preserving the detector response distributions. With both acceleration methods enabled, the overall speedup reaches <InlineEquation ID="IEq5"> <EquationSource Format="TEX">$22\times $</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>22</mn> <mo>×</mo> </mrow> </math></EquationSource> </InlineEquation> for <InlineEquation ID="IEq6"> <EquationSource Format="TEX">$\pi ^+$</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>π</mi> <mo>+</mo> </msup> </math></EquationSource> </InlineEquation> samples and <InlineEquation ID="IEq7"> <EquationSource Format="TEX">$56\times $</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>56</mn> <mo>×</mo> </mrow> </math></EquationSource> </InlineEquation> for proton samples without visible signal distortion. Conclusion These results demonstrate that substantial speedups can be achieved without compromising accuracy, and that the proposed methods could be readily extended to other detectors involving massive avalanche multiplication after necessary validations.

Acceleration of GEM detector digitization via random number pool and statistical scaling approximation

期刊论文
发表日期：

2026年7月9日

查看全文

Xinnan Wang,
Liangliang Wang,
Huaimin Liu,
Linghui Wu,
Guang Zhao,
Shengsen Sun

摘要

Purpose

Monte Carlo simulation is indispensable in particle physics experiments, but explicitly sampling for very large stochastic particle populations could become a major computational bottleneck. Gas Electron Multiplier (GEM)-based detectors are a representative example: Cascade avalanche multiplication across multiple GEM foils can generate $\mathcal {O}(10^7)$ $O (10^{7})$ electrons per track, and the stochastic transport and signal induction sampling or computation required for each electron make large-scale Monte Carlo production prohibitively expensive.

Methods

We present two acceleration methods, developed and validated for the digitization of the Cylindrical GEM (CGEM) detector at Beijing Spectrometer (BESIII). The first method, random number pool, pre-generates random numbers in contiguous memory and retrieves them through pointer arithmetic. Combined with batch processing, this approach improves cache locality. The second method, statistical scaling approximation, reduces the number of simulated electrons by a downscaling factor while amplifying back the magnitude in signal induction, thereby preserving signal fidelity.

Results

For the BESIII CGEM digitization software, the random number pool optimization together with batch processing accelerates the simulation by over $4\times $ $4 \times$ for $\pi ^+$ $π^{+}$ samples and $5\times $ $5 \times$ for proton samples. The statistical scaling approximation provides a reduction of the simulated electron population while preserving the detector response distributions. With both acceleration methods enabled, the overall speedup reaches $22\times $ $22 \times$ for $\pi ^+$ $π^{+}$ samples and $56\times $ $56 \times$ for proton samples without visible signal distortion.

Conclusion

These results demonstrate that substantial speedups can be achieved without compromising accuracy, and that the proposed methods could be readily extended to other detectors involving massive avalanche multiplication after necessary validations.