Add op remainder for all platform #4912

FisherWY · 2023-08-03T15:05:02Z

tencent-adm · 2023-08-03T15:05:20Z

Thank you for your submission, we really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ nihui
❌ FisherWY
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

nihui · 2023-08-03T15:31:17Z

remainder 应该实现在 binaryop 里的...

FisherWY · 2023-08-04T02:10:04Z

remainder 应该实现在 binaryop 里的...

是的，昨天参考了Paddle的文档，提PR后才发现Paddle和Torch的Remainder不一样😂，下一个commit会修正的
Paddle文档：链接
Torch文档：链接

codecov-commenter · 2023-08-05T00:47:09Z

Codecov Report

Merging #4912 (1fd5705) into master (c45c01c) will decrease coverage by 0.05%.
Report is 32 commits behind head on master.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master    #4912      +/-   ##
==========================================
- Coverage   89.81%   89.76%   -0.05%     
==========================================
  Files         306      306              
  Lines       86875    86997     +122     
==========================================
+ Hits        78024    78091      +67     
- Misses       8851     8906      +55

Files Changed	Coverage Δ
src/layer/binaryop.cpp	`97.19% <0.00%> (-2.20%)`	⬇️
src/layer/x86/avx512_mathfun.h	`99.00% <0.00%> (-1.00%)`	⬇️
src/layer/x86/avx_mathfun.h	`98.79% <0.00%> (-1.21%)`	⬇️
src/layer/x86/binaryop_x86.cpp	`98.11% <0.00%> (-1.69%)`	⬇️
src/layer/x86/sse_mathfun.h	`98.78% <0.00%> (-1.22%)`	⬇️

... and 6 files with indirect coverage changes

nihui · 2023-09-21T02:32:37Z

ci 很多编译失败，需要修复

FisherWY · 2023-09-21T14:01:49Z

ci 很多编译失败，需要修复

目前在x86上根据Torch提供的计算公式进行实现，但貌似结果没法对齐（test_binaryop挂）：torch.remainder(a, b) == a - a.div(b, rounding_mode="floor") * b，链接，🤔

nihui · 2023-09-26T07:04:45Z

ci 很多编译失败，需要修复

目前在x86上根据Torch提供的计算公式进行实现，但貌似结果没法对齐（test_binaryop挂）：torch.remainder(a, b) == a - a.div(b, rounding_mode="floor") * b，链接，🤔

        float div_result = x / y;
        float round_result = roundf(div_result);
        float res = x - y * round_result;
        return res;

是这里的 roundf( x / y ) 和 div floor 不一样吧

FisherWY · 2023-10-16T08:09:10Z

ci 很多编译失败，需要修复

目前在x86上根据Torch提供的计算公式进行实现，但貌似结果没法对齐（test_binaryop挂）：torch.remainder(a, b) == a - a.div(b, rounding_mode="floor") * b，链接，🤔
        float div_result = x / y;
        float round_result = roundf(div_result);
        float res = x - y * round_result;
        return res;
是这里的 roundf( x / y ) 和 div floor 不一样吧

遇到了一个奇怪的问题，复现步骤如下：

在src/layer/binaryop.cpp中写一个实现，返回值为0：

struct binary_op_remainder
{
    float operator()(const float& x, const float& y) const
    {
        return 0.0f;
    }
};

在src/layer/x86/binaryop_x86.cpp中实现x86平台，返回值同样为0：

struct binary_op_remainder
{
    float func(const float& x, const float& y) const
    {

        return 0.0f;
    }
#if __SSE2__
    __m128 func_pack4(const __m128& x, const __m128& y) const
    {
        __m128 res = _mm_setzero_ps();
        return res;
    }
#if __AVX__
    __m256 func_pack8(const __m256& x, const __m256& y) const
    {
        __m256 res = _mm256_setzero_ps();
        return res;
    }
#if __AVX512F__
    __m512 func_pack16(const __m512& x, const __m512& y) const
    {
        __m512 res = _mm512_setzero_ps();
        return res;
    }
#endif // __AVX512F__
#endif // __AVX__
#endif // __SSE2__

编译并运行单测，却会得到不同的结果：
请问这是什么原因造成的呢？（我的理解是单测是用src/layer/binaryop.cpp的计算结果跟对应平台的实现进行比对，请问是这理解有误吗？）

nihui · 2023-10-16T08:24:03Z

ci 很多编译失败，需要修复

目前在x86上根据Torch提供的计算公式进行实现，但貌似结果没法对齐（test_binaryop挂）：torch.remainder(a, b) == a - a.div(b, rounding_mode="floor") * b，链接，🤔
        float div_result = x / y;
        float round_result = roundf(div_result);
        float res = x - y * round_result;
        return res;
是这里的 roundf( x / y ) 和 div floor 不一样吧

遇到了一个奇怪的问题，复现步骤如下：

1. 在`src/layer/binaryop.cpp`中写一个实现，返回值为0：

struct binary_op_remainder
{
    float operator()(const float& x, const float& y) const
    {
        return 0.0f;
    }
};

2. 在`src/layer/x86/binaryop_x86.cpp`中实现x86平台，返回值同样为0：

struct binary_op_remainder
{
    float func(const float& x, const float& y) const
    {

        return 0.0f;
    }
#if __SSE2__
    __m128 func_pack4(const __m128& x, const __m128& y) const
    {
        __m128 res = _mm_setzero_ps();
        return res;
    }
#if __AVX__
    __m256 func_pack8(const __m256& x, const __m256& y) const
    {
        __m256 res = _mm256_setzero_ps();
        return res;
    }
#if __AVX512F__
    __m512 func_pack16(const __m512& x, const __m512& y) const
    {
        __m512 res = _mm512_setzero_ps();
        return res;
    }
#endif // __AVX512F__
#endif // __AVX__
#endif // __SSE2__

3. 编译并运行单测，却会得到不同的结果：
   ![image](https://user-images.githubusercontent.com/32707008/275434958-8e9949fa-2e45-420b-949d-b218cbb2a881.png)

4. 请问这是什么原因造成的呢？（我的理解是单测是用`src/layer/binaryop.cpp`的计算结果跟对应平台的实现进行比对，请问是这理解有误吗？）

test layer gpu failed 表明 vulkan 的实现没有和 binaryop.cpp 对齐

FisherWY · 2023-10-16T10:07:28Z

ci 很多编译失败，需要修复

目前在x86上根据Torch提供的计算公式进行实现，但貌似结果没法对齐（test_binaryop挂）：torch.remainder(a, b) == a - a.div(b, rounding_mode="floor") * b，链接，🤔
        float div_result = x / y;
        float round_result = roundf(div_result);
        float res = x - y * round_result;
        return res;
是这里的 roundf( x / y ) 和 div floor 不一样吧

遇到了一个奇怪的问题，复现步骤如下：

1. 在`src/layer/binaryop.cpp`中写一个实现，返回值为0：

struct binary_op_remainder
{
    float operator()(const float& x, const float& y) const
    {
        return 0.0f;
    }
};

2. 在`src/layer/x86/binaryop_x86.cpp`中实现x86平台，返回值同样为0：

struct binary_op_remainder
{
    float func(const float& x, const float& y) const
    {

        return 0.0f;
    }
#if __SSE2__
    __m128 func_pack4(const __m128& x, const __m128& y) const
    {
        __m128 res = _mm_setzero_ps();
        return res;
    }
#if __AVX__
    __m256 func_pack8(const __m256& x, const __m256& y) const
    {
        __m256 res = _mm256_setzero_ps();
        return res;
    }
#if __AVX512F__
    __m512 func_pack16(const __m512& x, const __m512& y) const
    {
        __m512 res = _mm512_setzero_ps();
        return res;
    }
#endif // __AVX512F__
#endif // __AVX__
#endif // __SSE2__

3. 编译并运行单测，却会得到不同的结果：
   ![image](https://user-images.githubusercontent.com/32707008/275434958-8e9949fa-2e45-420b-949d-b218cbb2a881.png)

4. 请问这是什么原因造成的呢？（我的理解是单测是用`src/layer/binaryop.cpp`的计算结果跟对应平台的实现进行比对，请问是这理解有误吗？）

test layer gpu failed 表明 vulkan 的实现没有和 binaryop.cpp 对齐

原来如此，非常感谢！

nihui · 2023-10-20T02:48:14Z

ci 很多测试失败了 qaq

Add draft code for op remainder on x86

242dd3c

FisherWY added 3 commits August 4, 2023 23:10

Remove old remainder_x86

3d77066

Refactor remainder_x86 to binary op

173d199

Remove headers

460b4d4

FisherWY added 10 commits August 8, 2023 22:16

Use sse4

5ab6be7

Try support remainder in vulkan

a8d4d97

Try support remainder in riscv

184e4ef

Try support remainder in mips

a10197b

Try support remainder in loongarch

42950e2

Try support remainder in arm

1fd5705

Change tests/test_binaryop OP_TYPE_MAX from 12 to 13

0f3c34f

Fix build error on riscv

66887d9

Add pnnx convertor

5df5c8d

Add remainder python unittest

a3e022f

FisherWY changed the title ~~[WIP] Add op remainder for all platform~~ Add op remainder for all platform Sep 13, 2023

Try fix result error on x86

21e6ca0

Fix args in binaryop.cpp

cea0026

nihui closed this Oct 11, 2023

nihui reopened this Oct 11, 2023

github-actions bot added core riscv tool pnnx labels Oct 11, 2023

github-actions bot added vulkan test layer arm loongarch mips x86 labels Oct 11, 2023

FisherWY added 5 commits October 17, 2023 09:14

Fix compute error for remainder on x86

22ce01f

Fix compute error for remainder on vulkan

6097a17

Fix compute error for remainder on loongarch

77f908d

Fix compute error for remainder on mips

8f7770d

Preprocess divisor for remainder on unittest

a370f41

github-actions bot removed the core label Oct 17, 2023

Fix build error

b4b198e

nihui and others added 2 commits December 22, 2023 11:28

Merge branch 'master' into op_remainder

b5c371a

apply code-format changes

d917d97

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add op remainder for all platform #4912

Add op remainder for all platform #4912

FisherWY commented Aug 3, 2023 •

edited

Loading

tencent-adm commented Aug 3, 2023 •

edited

Loading

nihui commented Aug 3, 2023

FisherWY commented Aug 4, 2023

codecov-commenter commented Aug 5, 2023 •

edited

Loading

nihui commented Sep 21, 2023

FisherWY commented Sep 21, 2023

nihui commented Sep 26, 2023

FisherWY commented Oct 16, 2023

nihui commented Oct 16, 2023

FisherWY commented Oct 16, 2023

nihui commented Oct 20, 2023

Add op remainder for all platform #4912

Are you sure you want to change the base?

Add op remainder for all platform #4912

Conversation

FisherWY commented Aug 3, 2023 • edited Loading

tencent-adm commented Aug 3, 2023 • edited Loading

nihui commented Aug 3, 2023

FisherWY commented Aug 4, 2023

codecov-commenter commented Aug 5, 2023 • edited Loading

Codecov Report

nihui commented Sep 21, 2023

FisherWY commented Sep 21, 2023

nihui commented Sep 26, 2023

FisherWY commented Oct 16, 2023

nihui commented Oct 16, 2023

FisherWY commented Oct 16, 2023

nihui commented Oct 20, 2023

FisherWY commented Aug 3, 2023 •

edited

Loading

tencent-adm commented Aug 3, 2023 •

edited

Loading

codecov-commenter commented Aug 5, 2023 •

edited

Loading