[CPU:Bugfix] supplement missing headfile, and fix a bug in MNNLineDep…

[Sherlockzhangjinge]

Description

Fix the bugs below:
ssize_t has not been declared:
source/backend/cpu/riscv/rvv/MNNFloat2Int8.cpp
source/backend/cpu/riscv/rvv/MNNInt8ScaleToFloat.cpp
source/backend/cpu/riscv/rvv/MNNAvgPoolInt8.cpp
不能在初始化时将 const int32_t* const 转换为 const float* :
source/backend/cpu/riscv/rvv/MNNLineDepthWiseInt8AddBiasScaleUnit.cpp

Module

CPU

Type

• Feature
• Bugfix
• Perf
• Refact
• Style
• Doc
• Test
• Chore

Checklist

• Commit message follows [Module:Type] Description format
• Code compiles without errors
• Tested on relevant platform(s)
• No unrelated format or style changes included

[wangzhaode]

Code Review

感谢提交！头文件补充的修复是正确的，但 MNNLineDepthWiseInt8AddBiasScaleUnit.cpp 的修复存在未解决的编译错误。

问题

在 MNNLineDepthWiseInt8AddBiasScaleUnit.cpp 第 24 行：

const float* bias_z = parameters->bias;

parameters->bias 的类型是 const int32_t*（见 QuanPostTreatParameters 结构体定义），C++ 中不允许从 const int32_t* 隐式转换为 const float*，此行会编译失败。

同时，修改后的第 59 行：

vint32m4_t vec_bias_int = __riscv_vle32_v_i32m4(bias_z, vl);

__riscv_vle32_v_i32m4 期望第一个参数为 const int32_t*，传入 const float* 类型的 bias_z 同样会导致类型不匹配错误。

建议修复

将第 24 行改为：

const int32_t* bias_z = parameters->bias;

同时建议删除注释掉的旧代码（第 58 行），git history 已经保留了记录：

-        // vfloat32m4_t vec_bias = __riscv_vle32_v_f32m4(bias_z, vl);

验证

已通过 clang++ 模拟验证两个错误：

1. error: cannot initialize a variable of type 'const float *' with an lvalue of type 'const int32_t *const'
2. error: no known conversion from 'const float *' to 'const int32_t *' for 1st argument

[wangzhaode]

Hi @Sherlockzhangjinge, the bugfix looks correct — especially the bias_z type change from float* to int32_t* with explicit vfcvt_f_x_v_f32m4 conversion, which is more accurate than the current cast in master. However, this PR now has merge conflicts because some of the header file fixes were already merged via #4359. Could you please rebase onto the latest master? The conflict should be straightforward — just keep both #include <sys/types.h> (from master) and your #include "Int8FunctionsOpt.h". The bias_z fix still has independent value. Thanks!