Skip to content

Feature fp16 arm32#690

Merged
seanxcwang merged 46 commits into
masterfrom
feature-fp16-arm32
Jan 2, 2021
Merged

Feature fp16 arm32#690
seanxcwang merged 46 commits into
masterfrom
feature-fp16-arm32

Conversation

@seanxcwang
Copy link
Copy Markdown
Collaborator

@seanxcwang seanxcwang commented Jan 2, 2021

No description provided.

seanxcwang and others added 30 commits December 18, 2020 15:21
@seanxcwang
init for arm32-fp16 build, compile failed
eca6c5c
@powerpwang
[ARM][WIP] update add and pooling arm32 fp16
6edf568
@powerpwang
[ARM][WIP] update half8 for fp16 arm32
8c42b5d
@seanxcwang
[TMP] add conv_c3&conv_dw_s1 arm32 fp16 impl
44b0fd8
@seanxcwang
reopen ARM82 macro, add arm32 fp16 impl except conv&deconv
19c588c
@seanxcwang
add half2float cvt asm
3201f1f
@seanxcwang
push reg in subroutine
e782aac
@seanxcwang
fix conv c3 asm error, conv dw && c3 ut pass
09d03f5
@quinnrong94
[ARM] exclude exynos 9810 for arm82
b5afb49
@quinnrong94
[ARM] cpu info add copyright
c592a8d
@quinnrong94
[ARM] check fp16arith in arm32 using midr
0222db0
@quinnrong94
[ARM] fp16arith check A14
903ba5f
@quinnrong94
[ARM] fp16arith check for apple m1
3872e1a
@quinnrong94
[ARM] support apple m1 compile
afd4616
@powerpwang
[ARM] add arm32 fp16 tile gemm
ebab479
@quinnrong94
[ARM] blob converter support fp16 reformat
0670cbc
@quinnrong94
[ARM] support convert RESERVED_FP16_TEST to float blob
295555c
@powerpwang
[ARM] add mls in TNNVector
acc42ae
@powerpwang
[ARM] add a32 fp16 conv 3x3
f270594
@seanxcwang
delete duplicate ARM82 macro
888b2f7
@powerpwang
[ARM] skip fp16 neon if turn off arm82 in armv7
fbc4f0f
@seanxcwang
[TMP] prepare for arm82 compile singly
c6cc789
@seanxcwang @quinnrong94
mov fp16 impl to single folder
f4ef0ee
@quinnrong94
[ARM] move fp16 computation to single file
7ca642c
@seanxcwang
arm32&arm64 fp16 compiler ok, ut ok
eb320ea
@powerpwang
[ARM] update arm32 fp16 deconv
bb69b96
@quinnrong94
[ARM] refactor to separate fp16 compute
19546d6
@powerpwang
[ARM] add conv and deconv fp16 simu impl
035e43a
@powerpwang
[ARM] remove unused files
ed907ac
@powerpwang
[ARM] remove unused header
a013bb3
@codecov-io
Copy link
Copy Markdown

codecov-io commented Jan 2, 2021 •
edited
Loading

Codecov Report

Merging #690 (1453e79) into master (20af475) will decrease coverage by 0.04%.
The diff coverage is 100.00%.

Impacted file tree graph

@@            Coverage Diff             @@
##           master     #690      +/-   ##
==========================================
- Coverage   32.52%   32.48%   -0.05%     
==========================================
  Files         298      298              
  Lines        8679     8679              
==========================================
- Hits         2823     2819       -4     
- Misses       5856     5860       +4
Impacted Files Coverage Δ
...ource/tnn/optimizer/net_optimizer_fuse_conv_add.cc 31.14% <100.00%> (ø)
source/tnn/utils/cpu_utils.cc 13.33% <100.00%> (ø)
include/tnn/utils/half.hpp 0.00% <0.00%> (-40.00%) ⬇️
source/tnn/utils/naive_compute.cc 59.61% <0.00%> (-0.49%) ⬇️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 20af475...1453e79. Read the comment docs.

@seanxcwang seanxcwang merged commit 86298db into master Jan 2, 2021
darrenyao87 added a commit that referenced this pull request Jan 4, 2021
@darrenyao87
Merge branch 'master' into feature_dynamic_dims
9b30550 
* master:
  [CONVERTER2TNN][UPD] 1. default do not clean build directory; (#694)
  Feature fp16 arm32 (#690)
  pose demo (#665)

# Conflicts:
#	examples/base/tnn_sdk_sample.cc
#	source/tnn/layer/upsample_layer.cc
#	tools/convert2tnn/utils/align_model.py
darrenyao87 added a commit that referenced this pull request Jan 4, 2021
@darrenyao87
Merge branch 'master' into feature_dynamic_dims
f231dfe 
* master:
  [CONVERTER2TNN][UPD] 1. default do not clean build directory; (#694)
  Feature fp16 arm32 (#690)
  pose demo (#665)

# Conflicts:
#	examples/base/tnn_sdk_sample.cc
#	source/tnn/layer/upsample_layer.cc
#	tools/convert2tnn/utils/align_model.py
@seanxcwang seanxcwang deleted the feature-fp16-arm32 branch February 1, 2021 10:37
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@quinnrong94 quinnrong94 quinnrong94 approved these changes

@darrenyao87 darrenyao87 Awaiting requested review from darrenyao87

+1 more reviewer

@powerpwang powerpwang powerpwang approved these changes

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@seanxcwang @codecov-io @quinnrong94 @powerpwang