[PR4] Ray-benchmark tool to test throughput of batch query API added in #195

[Waqar-ukaea]

This PR adds a new tool/miniapp that I have been using to benchmark the batch query API pure ray throughput on various GPUs.

This PR adds tools/ray_benchmark/ which includes:

• ray_benchmark.cpp - The main application
• ray_benchmark.h - Header file containing some helper functions including definition of the callback method which "mocks" an external application filling XDG's internal ray buffers
• ray_benchmark_deviceCode.slang - Slang compute shader for filling XDG's internal ray buffer on device
• ray_benchmark_shared.h - Shared code between host and device
• ray_benchmark_driver.py - A python script used to drive multiple runs of the ray_benchmark miniapp for more consistent results measurement.

Copied from #178:

Benchmark parameters

model	Volume	No. of Elements	No. of Rays	Location	No. of Runs
simple_tokamak	2	280K	50M	(180,250,-27)	100

A render of the simple_tokamak model [1] used in these preliminary benchmarks along with a depiction (significantly smaller number of rays plotted) of the rays launched is shown in the image below, along with the volume queried against highlghted in blue:

Ray tracer performance (trace-only)

Baseline = Embree (CPU), 2× Intel® Xeon® Platinum 8480+ (Sapphire Rapids) × 112 threads

Times and throughput averaged over the 100 runs.

Ray tracer performance (trace-only).

Baseline = Embree (CPU), 2x Intel Xeon Platinum 8480+ (Sapphire Rapids), 112 threads.

Ray Tracer backend	Hardware (Threads / Device)	Trace Time (s)	Throughput (ray/s)	Speedup vs 2x8480+ (112-thread)	Peak FP32/FP64 (TFLOPS) + RT cores	Node %
Embree	13th Gen Intel Core i7-13850HX x 28 threads	0.749128	1.06791e+08	~0.25x	N/A	Not HPC
Embree	1x Intel Xeon Platinum 8480+ (Sapphire Rapids) x 56 threads	0.367881	2.17462e+08	~0.51x	N/A	50%
Embree	2x Intel Xeon Platinum 8480+ (Sapphire Rapids) x 112 threads	0.189140	4.22967e+08	1x (baseline)	N/A	100%
GPRT (mixed FP64)	NVIDIA RTX 2000 Ada	1.112610	7.19033e+07	~0.17x	FP32: 12.0 FP64: 0.19 RT cores: 22	Not HPC
GPRT (FP32 + RT cores)	NVIDIA RTX 2000 Ada	0.0302956	2.64065e+09	~6.24x	FP32: 12.0 FP64: 0.19 RT cores: 22	Not HPC
cuBQL (FP64)	NVIDIA RTX 2000 Ada	~1.188013*	6.73393e+07	~0.16x	FP32: 12.0 FP64: 0.19 RT cores: 22 (unused)	Not HPC
GPRT (mixed FP64)	NVIDIA L40	0.185930	4.31900e+08	~1.02x	FP32: 90.5 FP64: 1.41 RT cores: 142	25%
GPRT (FP32 + RT cores)	NVIDIA L40	0.008051	9.95400e+09	~23.5x	FP32: 90.5 FP64: 1.41 RT cores: 142	25%
GPRT (mixed FP64)	NVIDIA A100	0.183721	4.50600e+08	~1.07x	FP32: 19.5 FP64: 9.7 RT cores: N/A	25%
cuBQL (FP64)	AMD MI300X	~0.0978379*	8.17679e+08	~1.93x	FP32: 163.4 FP64: 81.7 RT cores: N/A	12.5%
GPRT (mixed FP64)	AMD MI300X - no Vulkan ray-tracing support	N/A	N/A	N/A	FP32: 163.4 FP64: 81.7 RT cores: N/A	N/A
GPRT (mixed FP64)	Intel Data Center GPU Max 1100 (Ponte Vecchio) - Vulkan does not recognise PVCs as physical devices	N/A	N/A	N/A	N/A	N/A

* The cuBQL runs are 50m rays, so I derive the equivalent 80m throughput from the trace time and throughput at 50m

References

[1] Valentine, A., Berry, T., Bradnam, S., Hagues, J., & Hodson, J. (2022). Benchmarking of emergent radiation transport codes for fusion neutronics applications. Fusion Engineering and Design, 180, 113197. https://doi.org/10.1016/j.fusengdes.2022.113197

[Waqar-ukaea]

Closed in favour of #238