Gao Tianlin aaef246c74 remove log2 decomposition; add log2 lowering (#123112) ... Same reason as `log10`. `log2` is a core aten op, we should not decompose it. As https://github.com/pytorch/pytorch/pull/110882 suggested, it often maps to a hardware intrinsic; Furthermore, decomposing it will negatively impact the numerical precision of the output. Pull Request resolved: https://github.com/pytorch/pytorch/pull/123112 Approved by: https://github.com/peterbell10		2024-04-02 16:16:26 +00:00
..
aoti_runtime	[Inductor] Enable ABI-compatible mode for cpp-wrapper JIT (#121309)	2024-03-07 14:22:06 +00:00
cuda	Defer selection of triton template (#120275)	2024-03-20 01:40:33 +00:00
__init__.py
common.py	[inductor] Remove identity from ops.scan (#119727)	2024-04-01 22:47:26 +00:00
cpp_prefix.h	[inductor][cpp] unified the vectorized conversion with at::vec::convert for all data types (#119979)	2024-03-29 21:48:29 +00:00
cpp_wrapper_cpu.py	[AOTInductor] Support quantized linear on CPU with fbgemm (#123069)	2024-04-01 09:15:05 +00:00
cpp_wrapper_cuda.py	Add torch.cond support to AOT Inductor (#121120)	2024-03-07 22:39:57 +00:00
cpp.py	remove log2 decomposition; add log2 lowering (#123112)	2024-04-02 16:16:26 +00:00
cuda_combined_scheduling.py	[mypy] added type annotations to codegen_nodes methods (#119080)	2024-02-05 18:33:52 +00:00
memory_planning.py	Remove follow_imports = skip from sympy (#118469)	2024-01-28 13:38:38 +00:00
multi_kernel.py	[inductor] make multi-kernel work with cpp-wrapper (#117813)	2024-02-05 23:35:41 +00:00
triton_foreach.py	[inductor] Refactor common triton imports into one function (#121438)	2024-03-09 18:17:36 +00:00
triton_split_scan.py	[inductor] Remove identity from ops.scan (#119727)	2024-04-01 22:47:26 +00:00
triton_utils.py	Precompile triton templates (#121998)	2024-03-25 21:33:36 +00:00
triton.py	remove log2 decomposition; add log2 lowering (#123112)	2024-04-02 16:16:26 +00:00
wrapper.py	[AOTInductor] Support quantized linear on CPU with fbgemm (#123069)	2024-04-01 09:15:05 +00:00