Faster general unary op (#2472)

* faster general unary op

* faster general ops + reorg

* fix + comment

* binary two

* copy general
This commit is contained in:
Awni Hannun
2025-08-15 15:04:12 -07:00
committed by GitHub
parent dfb5022eab
commit 6441c21a94
62 changed files with 1215 additions and 203 deletions

View File

@@ -8,7 +8,6 @@ target_sources(
PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/allocator.cpp
${CMAKE_CURRENT_SOURCE_DIR}/arange.cu
${CMAKE_CURRENT_SOURCE_DIR}/arg_reduce.cu
${CMAKE_CURRENT_SOURCE_DIR}/binary.cu
${CMAKE_CURRENT_SOURCE_DIR}/binary_two.cu
${CMAKE_CURRENT_SOURCE_DIR}/compiled.cpp
${CMAKE_CURRENT_SOURCE_DIR}/copy.cu
@@ -45,12 +44,14 @@ target_sources(
${CMAKE_CURRENT_SOURCE_DIR}/softmax.cu
${CMAKE_CURRENT_SOURCE_DIR}/sort.cu
${CMAKE_CURRENT_SOURCE_DIR}/ternary.cu
${CMAKE_CURRENT_SOURCE_DIR}/unary.cu
${CMAKE_CURRENT_SOURCE_DIR}/utils.cpp
${CMAKE_CURRENT_SOURCE_DIR}/quantized/affine_quantize.cu
${CMAKE_CURRENT_SOURCE_DIR}/quantized/quantized.cpp
${CMAKE_CURRENT_SOURCE_DIR}/worker.cpp)
add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/binary)
add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/unary)
if(CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 12.9.0)
target_sources(
mlx PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/gemms/cublas_gemm_batched_12_9.cu)