Gaussian Error Linear Unit
⏳ PENDING — not yet measured| Operation | GELU (bpd_gelu) |
|---|---|
| Formula | f(x) = 0.5x · (1 + erf(x/√2)) |
| Category | Transcendental (erf) |
This is the backend-neutral op_expr fact that specifies the operation. The same term is lowered to LLVM IR, CUDA C, MLIR, and every other backend — one specification, many targets.
op_expr(bpd_gelu,
mul(mul(var, const(0.5)),
add(const(1.0),
call(erf, mul(var, const(0.7071067811865476)))))).
| Backend | LLVM CPU (llvm_cpu) |
|---|---|
| Target | cpu |
| Description | LLVM IR targeting x86-64 CPU. Compiled with llc + clang. |
This (operation, backend) pair has not yet been measured. When the backend emitter and verification infrastructure are ready, this cell will be tested against the PyTorch reference.