phase 3: GEMM kernels (naive, tiled, cuBLAS)

- Naive GEMM kernel: one thread per output element (F32 + BF16)
- Tiled GEMM kernel: 32x32 shared memory tiles (F32 + BF16)
- cuBLAS wrapper: cublasGemmEx with row-major trick
- GemmBackend enum for runtime backend selection
- CublasContext RAII handle
- Made error::check public for cross-crate use
- 17 GEMM tests: small/medium/rect sizes, all backends, F32+BF16
- Cross-backend consistency verified (naive vs tiled vs cuBLAS)
- All 44 tests pass across all crates

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

This commit is contained in:

Gahow Wang

2026-05-21 19:48:05 +08:00

parent a83971fa25

commit d77f921a12

9 changed files with 519 additions and 1 deletions

									
										1

Cargo.toml
									
												View File
												
				@@ -3,6 +3,7 @@ resolver = "2"

				members = [

				    "crates/xserv-cuda",

				    "crates/xserv-tensor",

				    "crates/xserv-kernels",

				]

				[workspace.package]

phase 3: GEMM kernels (naive, tiled, cuBLAS)

1 Cargo.toml Unescape Escape View File

1

Cargo.toml

View File