Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
PaddlePaddle
Paddle
提交
3a77d027
P
Paddle
项目概览
PaddlePaddle
/
Paddle
1 年多 前同步成功
通知
2310
Star
20933
Fork
5423
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1423
列表
看板
标记
里程碑
合并请求
543
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
Paddle
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
1,423
Issue
1,423
列表
看板
标记
里程碑
合并请求
543
合并请求
543
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
未验证
提交
3a77d027
编写于
3月 08, 2022
作者:
C
Chen Weihang
提交者：
GitHub
3月 08, 2022
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
[Phi] Remove gpudnn suffix & polish cmake (#40239)
* remove gpudnn suffix & polish cmake * fix typo
上级
9aa6bfc7
变更
6
隐藏空白更改
内联
并排
Showing
6 changed file
with
59 addition
and
63 deletion
+59
-63
cmake/phi.cmake
cmake/phi.cmake
+59
-63
paddle/phi/kernels/gpudnn/conv_grad_grad_kernel.cu
paddle/phi/kernels/gpudnn/conv_grad_grad_kernel.cu
+0
-0
paddle/phi/kernels/gpudnn/conv_grad_kernel.cu
paddle/phi/kernels/gpudnn/conv_grad_kernel.cu
+0
-0
paddle/phi/kernels/gpudnn/conv_kernel.cu
paddle/phi/kernels/gpudnn/conv_kernel.cu
+0
-0
paddle/phi/kernels/gpudnn/softmax_grad_kernel.cu
paddle/phi/kernels/gpudnn/softmax_grad_kernel.cu
+0
-0
paddle/phi/kernels/gpudnn/softmax_kernel.cu
paddle/phi/kernels/gpudnn/softmax_kernel.cu
+0
-0
未找到文件。
cmake/phi.cmake
浏览文件 @
3a77d027
...
...
@@ -134,8 +134,8 @@ function(kernel_library TARGET)
if
(
EXISTS
${
CMAKE_CURRENT_SOURCE_DIR
}
/kps/
${
TARGET
}
.cu
)
list
(
APPEND gpu_srcs
${
CMAKE_CURRENT_SOURCE_DIR
}
/kps/
${
TARGET
}
.cu
)
endif
()
if
(
EXISTS
${
CMAKE_CURRENT_SOURCE_DIR
}
/gpudnn/
${
TARGET
}
_gpudnn
.cu
)
list
(
APPEND gpudnn_srcs
${
CMAKE_CURRENT_SOURCE_DIR
}
/gpudnn/
${
TARGET
}
_gpudnn
.cu
)
if
(
EXISTS
${
CMAKE_CURRENT_SOURCE_DIR
}
/gpudnn/
${
TARGET
}
.cu
)
list
(
APPEND gpudnn_srcs
${
CMAKE_CURRENT_SOURCE_DIR
}
/gpudnn/
${
TARGET
}
.cu
)
endif
()
endif
()
if
(
WITH_XPU
)
...
...
@@ -197,92 +197,88 @@ function(kernel_library TARGET)
# kernel source file level
# level 1: base device kernel
# - cpu_srcs / gpu_srcs / xpu_srcs / kps_srcs
# - cpu_srcs / gpu_srcs / xpu_srcs /
gpudnn_srcs /
kps_srcs
# level 2: device-independent kernel
# - common_srcs
# level 3: Kernel implemented by reusing device-independent kernel
# - selected_rows_srcs
set
(
base_device_kernels
)
set
(
device_independent_kernel
)
set
(
high_level_kernels
)
# Build Target according different src organization
if
((
${
cpu_srcs_len
}
GREATER 0 OR
${
gpu_srcs_len
}
GREATER 0 OR
${
xpu_srcs_len
}
GREATER 0 OR
${
gpudnn_srcs_len
}
GREATER 0 OR
${
kps_srcs_len
}
GREATER 0
)
AND
(
${
common_srcs_len
}
GREATER 0 OR
${
selected_rows_srcs_len
}
GREATER 0
))
# If the common_srcs/selected_rows_srcs depends on specific device srcs, build target using this rule.
# 1. Base device kernel compile
if
(
${
cpu_srcs_len
}
GREATER 0
)
cc_library
(
${
TARGET
}
_cpu SRCS
${
cpu_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
list
(
APPEND base_device_kernels
${
TARGET
}
_cpu
)
endif
()
if
(
${
gpu_srcs_len
}
GREATER 0
)
if
(
WITH_GPU
)
if
(
${
cpu_srcs_len
}
GREATER 0 OR
${
gpu_srcs_len
}
GREATER 0 OR
${
gpudnn_srcs_len
}
GREATER 0
)
nv_library
(
${
TARGET
}
_part SRCS
${
cpu_srcs
}
${
gpu_srcs
}
${
gpudnn_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
nv_library
(
${
TARGET
}
SRCS
${
common_srcs
}
${
selected_rows_srcs
}
DEPS
${
TARGET
}
_part
)
endif
()
nv_library
(
${
TARGET
}
_gpu SRCS
${
gpu_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
elseif
(
WITH_ROCM
)
if
(
${
cpu_srcs_len
}
GREATER 0 OR
${
gpu_srcs_len
}
GREATER 0 OR
${
gpudnn_srcs_len
}
GREATER 0
)
hip_library
(
${
TARGET
}
_part SRCS
${
cpu_srcs
}
${
gpu_srcs
}
${
gpudnn_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
hip_library
(
${
TARGET
}
SRCS
${
common_srcs
}
${
selected_rows_srcs
}
DEPS
${
TARGET
}
_part
)
endif
()
elseif
(
WITH_XPU_KP
)
if
(
${
cpu_srcs_len
}
GREATER 0 OR
${
xpu_srcs_len
}
GREATER 0 OR
${
kps_srcs_len
}
GREATER 0
)
xpu_library
(
${
TARGET
}
_part SRCS
${
cpu_srcs
}
${
xpu_srcs
}
${
kps_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
xpu_library
(
${
TARGET
}
SRCS
${
common_srcs
}
${
selected_rows_srcs
}
DEPS
${
TARGET
}
_part
)
endif
()
else
()
if
(
${
cpu_srcs_len
}
GREATER 0 OR
${
xpu_srcs_len
}
GREATER 0
)
cc_library
(
${
TARGET
}
_part SRCS
${
cpu_srcs
}
${
xpu_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
cc_library
(
${
TARGET
}
SRCS
${
common_srcs
}
${
selected_rows_srcs
}
DEPS
${
TARGET
}
_part
)
endif
()
hip_library
(
${
TARGET
}
_gpu SRCS
${
gpu_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
endif
()
# If there are only specific device srcs, build target using this rule.
elseif
(
${
cpu_srcs_len
}
GREATER 0 OR
${
gpu_srcs_len
}
GREATER 0 OR
${
xpu_srcs_len
}
GREATER 0 OR
${
gpudnn_srcs_len
}
GREATER 0 OR
${
kps_srcs_len
}
GREATER 0
)
list
(
APPEND base_device_kernels
${
TARGET
}
_gpu
)
endif
()
if
(
${
xpu_srcs_len
}
GREATER 0
)
cc_library
(
${
TARGET
}
_xpu SRCS
${
xpu_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
list
(
APPEND base_device_kernels
${
TARGET
}
_xpu
)
endif
()
if
(
${
gpudnn_srcs_len
}
GREATER 0
)
if
(
WITH_GPU
)
if
(
${
cpu_srcs_len
}
GREATER 0 OR
${
gpu_srcs_len
}
GREATER 0 OR
${
gpudnn_srcs_len
}
GREATER 0
)
nv_library
(
${
TARGET
}
SRCS
${
cpu_srcs
}
${
gpu_srcs
}
${
gpudnn_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
endif
()
nv_library
(
${
TARGET
}
_gpudnn SRCS
${
gpudnn_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
elseif
(
WITH_ROCM
)
if
(
${
cpu_srcs_len
}
GREATER 0 OR
${
gpu_srcs_len
}
GREATER 0 OR
${
gpudnn_srcs_len
}
GREATER 0
)
hip_library
(
${
TARGET
}
SRCS
${
cpu_srcs
}
${
gpu_srcs
}
${
gpudnn_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
endif
()
elseif
(
WITH_XPU_KP
)
if
(
${
cpu_srcs_len
}
GREATER 0 OR
${
xpu_srcs_len
}
GREATER 0 OR
${
kps_srcs_len
}
GREATER 0
)
xpu_library
(
${
TARGET
}
SRCS
${
cpu_srcs
}
${
xpu_srcs
}
${
kps_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
endif
()
else
()
if
(
${
cpu_srcs_len
}
GREATER 0 OR
${
xpu_srcs_len
}
GREATER 0
)
cc_library
(
${
TARGET
}
SRCS
${
cpu_srcs
}
${
xpu_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
endif
()
hip_library
(
${
TARGET
}
_gpudnn SRCS
${
gpudnn_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
endif
()
# If the selected_rows_srcs depends on common_srcs, build target using this rule.
elseif
(
${
common_srcs_len
}
GREATER 0 AND
${
selected_rows_srcs_len
}
GREATER 0
)
list
(
APPEND base_device_kernels
${
TARGET
}
_gpudnn
)
endif
()
if
(
${
kps_srcs_len
}
GREATER 0
)
# only when WITH_XPU_KP, the kps_srcs_len can be > 0
xpu_library
(
${
TARGET
}
_kps SRCS
${
kps_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
list
(
APPEND base_device_kernels
${
TARGET
}
_kps
)
endif
()
# 2. Device-independent kernel compile
if
(
${
common_srcs_len
}
GREATER 0
)
if
(
WITH_GPU
)
nv_library
(
${
TARGET
}
_part SRCS
${
common_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
nv_library
(
${
TARGET
}
SRCS
${
selected_rows_srcs
}
DEPS
${
TARGET
}
_part
)
nv_library
(
${
TARGET
}
_common SRCS
${
common_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
${
base_device_kernels
}
)
elseif
(
WITH_ROCM
)
hip_library
(
${
TARGET
}
_part SRCS
${
common_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
hip_library
(
${
TARGET
}
SRCS
${
selected_rows_srcs
}
DEPS
${
TARGET
}
_part
)
hip_library
(
${
TARGET
}
_common SRCS
${
common_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
${
base_device_kernels
}
)
elseif
(
WITH_XPU_KP
)
xpu_library
(
${
TARGET
}
_part SRCS
${
common_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
xpu_library
(
${
TARGET
}
SRCS
${
selected_rows_srcs
}
DEPS
${
TARGET
}
_part
)
xpu_library
(
${
TARGET
}
_common SRCS
${
common_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
${
base_device_kernels
}
)
else
()
cc_library
(
${
TARGET
}
_part SRCS
${
common_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
cc_library
(
${
TARGET
}
SRCS
${
selected_rows_srcs
}
DEPS
${
TARGET
}
_part
)
cc_library
(
${
TARGET
}
_common SRCS
${
common_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
${
base_device_kernels
}
)
endif
()
# If there are only common_srcs or selected_rows_srcs, build target using below rules.
elseif
(
${
common_srcs_len
}
GREATER 0
)
list
(
APPEND device_independent_kernel
${
TARGET
}
_common
)
endif
()
# 3. Reusing kernel compile
if
(
${
selected_rows_srcs_len
}
GREATER 0
)
if
(
WITH_GPU
)
nv_library
(
${
TARGET
}
SRCS
${
common_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
nv_library
(
${
TARGET
}
_sr SRCS
${
selected_rows_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
${
base_device_kernels
}
${
device_independent_kernel
}
)
elseif
(
WITH_ROCM
)
hip_library
(
${
TARGET
}
SRCS
${
common_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
hip_library
(
${
TARGET
}
_sr SRCS
${
selected_rows_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
${
base_device_kernels
}
${
device_independent_kernel
}
)
elseif
(
WITH_XPU_KP
)
xpu_library
(
${
TARGET
}
SRCS
${
common_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
xpu_library
(
${
TARGET
}
_sr SRCS
${
selected_rows_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
${
base_device_kernels
}
${
device_independent_kernel
}
)
else
()
cc_library
(
${
TARGET
}
SRCS
${
common_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
)
cc_library
(
${
TARGET
}
_sr SRCS
${
selected_rows_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
${
base_device_kernels
}
${
device_independent_kernel
}
)
endif
()
elseif
(
${
selected_rows_srcs_len
}
GREATER 0
)
list
(
APPEND high_level_kernels
${
TARGET
}
_sr
)
endif
()
# 4. Unify target compile
list
(
LENGTH base_device_kernels base_device_kernels_len
)
list
(
LENGTH device_independent_kernel device_independent_kernel_len
)
list
(
LENGTH high_level_kernels high_level_kernels_len
)
if
(
${
base_device_kernels_len
}
GREATER 0 OR
${
device_independent_kernel_len
}
GREATER 0 OR
${
high_level_kernels_len
}
GREATER 0
)
if
(
WITH_GPU
)
nv_library
(
${
TARGET
}
SRCS
${
selected_rows_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_dep
s
}
)
nv_library
(
${
TARGET
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
${
base_device_kernels
}
${
device_independent_kernel
}
${
high_level_kernel
s
}
)
elseif
(
WITH_ROCM
)
hip_library
(
${
TARGET
}
SRCS
${
selected_rows_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_dep
s
}
)
hip_library
(
${
TARGET
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
${
base_device_kernels
}
${
device_independent_kernel
}
${
high_level_kernel
s
}
)
elseif
(
WITH_XPU_KP
)
xpu_library
(
${
TARGET
}
SRCS
${
selected_rows_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_dep
s
}
)
xpu_library
(
${
TARGET
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
${
base_device_kernels
}
${
device_independent_kernel
}
${
high_level_kernel
s
}
)
else
()
cc_library
(
${
TARGET
}
SRCS
${
selected_rows_srcs
}
DEPS
${
kernel_library_DEPS
}
${
kernel_dep
s
}
)
cc_library
(
${
TARGET
}
DEPS
${
kernel_library_DEPS
}
${
kernel_deps
}
${
base_device_kernels
}
${
device_independent_kernel
}
${
high_level_kernel
s
}
)
endif
()
else
()
set
(
target_build_flag 0
)
...
...
paddle/phi/kernels/gpudnn/conv_grad_grad_kernel
_gpudnn
.cu
→
paddle/phi/kernels/gpudnn/conv_grad_grad_kernel.cu
浏览文件 @
3a77d027
文件已移动
paddle/phi/kernels/gpudnn/conv_grad_kernel
_gpudnn
.cu
→
paddle/phi/kernels/gpudnn/conv_grad_kernel.cu
浏览文件 @
3a77d027
文件已移动
paddle/phi/kernels/gpudnn/conv_kernel
_gpudnn
.cu
→
paddle/phi/kernels/gpudnn/conv_kernel.cu
浏览文件 @
3a77d027
文件已移动
paddle/phi/kernels/gpudnn/softmax_grad_kernel
_gpudnn
.cu
→
paddle/phi/kernels/gpudnn/softmax_grad_kernel.cu
浏览文件 @
3a77d027
文件已移动
paddle/phi/kernels/gpudnn/softmax_kernel
_gpudnn
.cu
→
paddle/phi/kernels/gpudnn/softmax_kernel.cu
浏览文件 @
3a77d027
文件已移动
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑！
取消
想要评论请
注册
或
登录
