886 kB

Ctrl+K

1 contributor

History: 7 commits

superexpai

Sync: clean cross-references in L2_BAND_CLIFF.md (MK, powered by Claude)

4bb4e7c verified 2 days ago

bench
v2: H100 (sm_90) retune — Hopper sub-16MB launch branch, H100 benchmarks/docs, version bump (MK, powered by Claude) 3 days ago
benchmarks
v3: fix pre-Hopper (32MiB, L2] launch cliff (independently reported); retract position-biased 4090 claims; buffer-fair dtype benchmarks (MK, powered by Claude) 3 days ago
build
v3: fix pre-Hopper (32MiB, L2] launch cliff (independently reported); retract position-biased 4090 claims; buffer-fair dtype benchmarks (MK, powered by Claude) 3 days ago
relu_cpu
Optimized relu: cpu/cuda/xpu, 1.2-1.85x faster on RTX 4090, benchmarked vs upstream and torch.relu about 1 month ago
relu_cuda
v3: fix pre-Hopper (32MiB, L2] launch cliff (independently reported); retract position-biased 4090 claims; buffer-fair dtype benchmarks (MK, powered by Claude) 3 days ago
relu_metal
v2: H100 (sm_90) retune — Hopper sub-16MB launch branch, H100 benchmarks/docs, version bump (MK, powered by Claude) 3 days ago
relu_xpu
Optimized relu: cpu/cuda/xpu, 1.2-1.85x faster on RTX 4090, benchmarked vs upstream and torch.relu about 1 month ago
tests
Optimized relu: cpu/cuda/xpu, 1.2-1.85x faster on RTX 4090, benchmarked vs upstream and torch.relu about 1 month ago
torch-ext
Optimized relu: cpu/cuda/xpu, 1.2-1.85x faster on RTX 4090, benchmarked vs upstream and torch.relu about 1 month ago
.gitattributes

129 Bytes
Optimized relu: cpu/cuda/xpu, 1.2-1.85x faster on RTX 4090, benchmarked vs upstream and torch.relu about 1 month ago
.gitignore

33 Bytes
v2: H100 (sm_90) retune — Hopper sub-16MB launch branch, H100 benchmarks/docs, version bump (MK, powered by Claude) 3 days ago
DTYPE_THROUGHPUT.md

3.36 kB
Optimized relu: cpu/cuda/xpu, 1.2-1.85x faster on RTX 4090, benchmarked vs upstream and torch.relu about 1 month ago
H100_BENCHMARK.md

7.11 kB
Align remaining docs with v3 corrections — errata for position-biased L2 ratios, H100 table fix (MK, powered by Claude) 2 days ago
INVESTIGATION_L2.md

3.49 kB
Optimized relu: cpu/cuda/xpu, 1.2-1.85x faster on RTX 4090, benchmarked vs upstream and torch.relu about 1 month ago
L2_BAND_CLIFF.md

3.69 kB
Sync: clean cross-references in L2_BAND_CLIFF.md (MK, powered by Claude) 2 days ago
OPTIMIZATION.md

4.31 kB
Align remaining docs with v3 corrections — errata for position-biased L2 ratios, H100 table fix (MK, powered by Claude) 2 days ago
PERFORMANCE.md

3.62 kB
Align remaining docs with v3 corrections — errata for position-biased L2 ratios, H100 table fix (MK, powered by Claude) 2 days ago
README.md

7.71 kB
v3: fix pre-Hopper (32MiB, L2] launch cliff (independently reported); retract position-biased 4090 claims; buffer-fair dtype benchmarks (MK, powered by Claude) 3 days ago
build.toml

521 Bytes
v3: fix pre-Hopper (32MiB, L2] launch cliff (independently reported); retract position-biased 4090 claims; buffer-fair dtype benchmarks (MK, powered by Claude) 3 days ago
flake.lock

3.03 kB
Optimized relu: cpu/cuda/xpu, 1.2-1.85x faster on RTX 4090, benchmarked vs upstream and torch.relu about 1 month ago
flake.nix

281 Bytes
Optimized relu: cpu/cuda/xpu, 1.2-1.85x faster on RTX 4090, benchmarked vs upstream and torch.relu about 1 month ago