LeetGPU

Solutions to LeetGPU challenges, written in CUDA.

• problems/ — challenge statements fetched from leetgpu.com
• src/ — solutions (CUDA .cu; #41 Simple Inference is PyTorch-only)
• ROCm/ — HIP ports of the solutions (generated with hipify-perl)
• tests/ — local test harnesses built from each spec's examples
• tools/ — fetch / compile-check / remote-run scripts

Local verification (CUDA or ROCm)

Every problem has a harness in tests/ that feeds the spec's example inputs to solve() on your own GPU and checks the output — see tests/README.md for details.

# AMD GPU (ROCm / hipcc), tests link ROCm/*.hip
tests/run_tests.sh                    # all problems
GPU_ARCH=gfx1100 tests/run_tests.sh   # non-default arch (default gfx1201)

# NVIDIA GPU (CUDA / nvcc), tests link src/*.cu
tests/run_tests_cuda.sh
NVCC_ARCH="-arch=sm_86" tests/run_tests_cuda.sh

# single problem, either runtime
tests/run_tests.sh softmax
tests/run_tests_cuda.sh softmax

# 41 Simple Inference (PyTorch)
python3 tests/test_simple_inference.py

Status: all 88 GPU tests + the PyTorch test pass on ROCm (RX 9070 XT, gfx1201). These are the small spec examples only — LeetGPU's official judge additionally runs large hidden tests and performance measurement.

Problems

#	Title	Difficulty	Problem	Solution
1	Vector Addition	easy	spec	src
2	Matrix Multiplication	easy	spec	src
3	Matrix Transpose	easy	spec	src
4	Reduction	medium	spec	src
5	Softmax	medium	spec	src
6	Softmax Attention	medium	spec	src
7	Color Inversion	easy	spec	src
8	Matrix Addition	easy	spec	src
9	1D Convolution	easy	spec	src
10	2D Convolution	medium	spec	src
11	3D Convolution	medium	spec	src
12	Multi-Head Attention	hard	spec	src
13	Histogramming	medium	spec	src
14	Multi-Agent Simulation	hard	spec	src
15	Sorting	hard	spec	src
16	Prefix Sum	medium	spec	src
17	Dot Product	medium	spec	src
18	Sparse Matrix-Vector Multiplication	medium	spec	src
19	Reverse Array	easy	spec	src
20	K-Means Clustering	hard	spec	src
21	ReLU	easy	spec	src
22	General Matrix Multiplication (GEMM)	medium	spec	src
23	Leaky ReLU	easy	spec	src
24	Rainbow Table	easy	spec	src
25	Categorical Cross Entropy Loss	medium	spec	src
27	Mean Squared Error	medium	spec	src
28	Gaussian Blur	medium	spec	src
29	Top K Selection	medium	spec	src
30	Batched Matrix Multiplication	medium	spec	src
31	Matrix Copy	easy	spec	src
32	INT8 Quantized MatMul	medium	spec	src
33	Ordinary Least Squares	medium	spec	src
34	Logistic Regression	medium	spec	src
35	Monte Carlo Integration	medium	spec	src
36	Radix Sort	hard	spec	src
37	Matrix Power	medium	spec	src
38	Nearest Neighbor	medium	spec	src
39	Fast Fourier Transform	hard	spec	src
40	Batch Normalization	medium	spec	src
41	Simple Inference	easy	spec	src
42	2D Max Pooling	medium	spec	src
43	Count Array Element	medium	spec	src
44	Count 2D Array Element	medium	spec	src
45	Count 3D Array Element	medium	spec	src
46	BFS Shortest Path	hard	spec	src
47	Subarray Sum	medium	spec	src
48	2D Subarray Sum	medium	spec	src
49	3D Subarray Sum	medium	spec	src
50	RMS Normalization	medium	spec	src
51	Max Subarray Sum	medium	spec	src
52	Sigmoid Linear Unit	easy	spec	src
53	Causal Self-Attention	hard	spec	src
54	Swish-Gated Linear Unit	easy	spec	src
55	Attention with Linear Biases	medium	spec	src
56	Linear Self-Attention	hard	spec	src
57	FP16 Batched Matrix Multiplication	medium	spec	src
58	FP16 Dot Product	medium	spec	src
59	Sliding Window Self-Attention	hard	spec	src
60	Top-p Sampling	medium	spec	src
61	Rotary Positional Embedding	medium	spec	src
62	Value Clipping	easy	spec	src
63	Interleave Arrays	easy	spec	src
64	Weight Dequantization	medium	spec	src
65	Gaussian Error Gated Linear Unit	easy	spec	src
66	RGB to Grayscale	easy	spec	src
67	MoE Top-K Gating	medium	spec	src
68	Sigmoid Activation	easy	spec	src
69	2D Jacobi Stencil	medium	spec	src
70	Segmented Exclusive Prefix Sum	medium	spec	src
71	Parallel Merge	medium	spec	src
72	Stream Compaction	medium	spec	src
73	All-Pairs Shortest Paths	hard	spec	src
74	GPT-2 Transformer Block	hard	spec	src
75	Sparse Matrix-Dense Matrix Multiplication	medium	spec	src
76	Adder Transformer Inference	medium	spec	src
78	2D FFT	medium	spec	src
80	Grouped Query Attention	medium	spec	src
81	INT4 Weight-Only Quantized MatMul	medium	spec	src
82	Linear Recurrence	medium	spec	src
84	SwiGLU MLP Block	medium	spec	src
85	LoRA Linear	medium	spec	src
87	Speculative Decoding Verification	medium	spec	src
90	Causal Depthwise Conv1d	medium	spec	src
92	Decaying Causal Attention	medium	spec	src
93	Llama Transformer Block	hard	spec	src
94	SSM Selective Scan	medium	spec	src
96	INT8 KV-Cache Attention	medium	spec	src
105	Group Normalization	medium	spec	src
106	Token Embedding Layer	medium	spec	src