[arm]add gemm + relu6/leakyrelu fusion #2674
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
….cc to distinguish between conv3x3s1_depthwise_fp32.cc
it is coped from __gemm_sdot_meta_.h
update code
fix build error in kernels/x86/conv_compute.h
Merge branch 'develop' of git://github.com/PaddlePaddle/Paddle-Lite into PaddlePaddle-develop
Merge branch 'conv_pad' of https://github.com/chenjiaoAngel/Paddle-Lite into conv_pad
delete con2d_transpose test, this test can found in test/math/
yiicy
reviewed
Jan 14, 2020
| @@ -11,8 +11,6 @@ if((NOT LITE_WITH_OPENCL AND NOT LITE_WITH_FPGA) AND (LITE_WITH_X86 OR LITE_WITH | |||
| lite_cc_test(test_kernel_activation_compute SRCS activation_compute_test.cc DEPS arena_framework ${npu_kernels} ${xpu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels}) | |||
| lite_cc_test(test_kernel_argmax_compute SRCS argmax_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels}) | |||
| lite_cc_test(test_kernel_axpy_compute SRCS axpy_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels}) | |||
| lite_cc_test(test_kernel_conv_compute SRCS conv_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels}) | |||
| lite_cc_test(test_kernel_conv2d_transpose_compute SRCS conv2d_transpose_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels}) | |||
There was a problem hiding this comment.
不跑了，test/math下也有conv_compute和conv_transpose的单测。避免每次修改，要修改两个单测
There was a problem hiding this comment.
- conv的单测是我最近加的，用来跑npu kernel的，如果是arm kernel会跳过，不会多跑
- lite/tests/kernels 目录下的单测最好保留。1、这里面的单测更多是用来测试功能性的东西，比如padding_algorithm，不同参数组合下的结果是否正确。2、这里面的单测会验证operator部分是否正确，而不只是kernel本身是否正确。3、不同平台的单测都可以比较容易的加在这里，只需要改下place就可以，用LITE_WITH_XXX分隔。
- lite/tests/math 目录下的单测，我理解是测试同一个OP kernel的多种实现是否正确，应该不要和lite/tests/kernels下的单测合并
There was a problem hiding this comment.
嗯，好。我下次提交把这个恢复过来。删除原因是，我现在在加relu6融合，这个需要修改单测，给conv_param.activation_param设值，不然跑单测会挂。不想重复修改单测，所以删除了。
MyPandaShaoxiang
requested changes
Jan 14, 2020
| "vld1.32 {d10-d11}, [%[din_ptr]]! @ vld1q_f32(din_ptr) \n" \ | ||
| "vld1.32 {d12-d13}, [%[din_ptr]]! @ vld1q_f32(din_ptr) \n" \ | ||
| "vadd.f32 q3, q3, %q[vbias] @ add \n" \ | ||
| "vadd.f32 q4, q5, %q[vbias] @ add \n" \ |
There was a problem hiding this comment.
这应该是q4,q4?还是就是q4,q5
| "vbif q3, q8, q7 @ choose \n" \ | ||
| "vbif q4, q10, q9 @ choose \n" \ | ||
| "vbif q5, q12, q11 @ choose \n" \ | ||
| "vbif q6, q13, q13 @ choose \n" |
| int remain = channel_size % 16; | ||
| float32x4_t vzero = vdupq_n_f32(0.f); | ||
| for (int j = 0; j < channel; j++) { | ||
| } |
MyPandaShaoxiang
approved these changes
Jan 14, 2020
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.