[large tensor] Use auto in intermediate var

[zrr1999]

PR Category

Operator Mechanism

PR Types

Bug fixes

Description

将通过四则运算计算得到的且原本类型为int的变量统一使用auto，可以大幅度减少后续将其他变量修改为int64时的工作量。

应用下列rule

id: const-int-arith-combo
language: cpp
files:
  - paddle/phi/kernels/**
ignores:
  - paddle/phi/kernels/legacy/**
  - paddle/phi/kernels/xpu/**
  - paddle/phi/kernels/custom/**
  - paddle/phi/kernels/sparse/**
severity: warning
message: const int/int32_t initialized from a (possibly complex) arithmetic expression
rule:
  any:
    # 赋值初始化：const T var = <complex arithmetic expr>;
    - pattern: const $T $VAR = $EXPR
    # 直接初始化：const T var(<complex arithmetic expr>);
    - pattern: const $T $VAR($EXPR)
constraints:
  T:
    regex: ^(int|int32_t)$
  EXPR:
    all:
      - any:
        - has:
            pattern: $A + $B
        - has:
            pattern: $A - $B
        - has:
            pattern: $A * $B
        - has:
            pattern: $A / $B
      - not:
          regex: \b(threadIdx|blockDim|blockIdx)\b
fix: |
  const auto $VAR($EXPR);


id: nonconst-int-arith-combo
language: cpp
files:
  - paddle/phi/kernels/**
ignores:
  - paddle/phi/kernels/legacy/**
  - paddle/phi/kernels/xpu/**
  - paddle/phi/kernels/custom/**
  - paddle/phi/kernels/sparse/**
severity: warning
message: int/int32_t initialized from a (possibly complex) arithmetic expression
rule:
  any:
    # 赋值初始化：T var = <complex arithmetic expr>;
    - pattern: $T $VAR = $EXPR
    # 直接初始化：T var(<complex arithmetic expr>);
    - pattern: $T $VAR($EXPR)
constraints:
  T:
    regex: ^(int|int32_t)$
  EXPR:
    all:
      - any:
        - has:
            pattern: $A + $B
        - has:
            pattern: $A - $B
        - has:
            pattern: $A * $B
        - has:
            pattern: $A / $B
      - not:
          regex: \b(threadIdx|blockDim|blockIdx)\b
fix: |
  auto $VAR = $EXPR;

pcard-93269

[paddle-bot]

你的PR提交成功，感谢你对开源项目的贡献!
请关注后续CI自动化测试结果，详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[Copilot]

Pull Request Overview

This pull request systematically converts explicit integer type declarations (primarily int, int32_t, int64_t) to use auto type deduction across a large number of files in the PaddlePaddle codebase. The changes affect kernel implementations, utility functions, and CUTLASS extensions.

Key Changes:

• Mass conversion of explicit integer types to auto for variables initialized with arithmetic expressions
• Introduction of direct initialization syntax with parentheses (e.g., const auto var(expr)) instead of copy initialization
• Changes span GPU kernels, CPU kernels, fusion kernels, and utility functions

Reviewed Changes

Copilot reviewed 180 out of 180 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
paddle/phi/kernels/strings/gpu/copy_utils.h	Changed int64_t numel to auto numel - potential overflow risk
paddle/phi/kernels/stride/reduce_grad_stride_kernel.cu	Changed int tmp_dim to auto tmp_dim
paddle/phi/kernels/primitive/datamover_primitives_xpu2.h	Multiple int to auto conversions for buffer calculations
paddle/phi/kernels/primitive/compute_primitives.h	Index calculations changed from int to auto
paddle/phi/kernels/onednn/*.cc	Multiple conversions with direct initialization syntax
paddle/phi/kernels/impl/*.h	Extensive int to auto conversions in implementation headers
paddle/phi/kernels/gpu/*.cu	GPU kernel index calculations converted to auto
paddle/phi/kernels/fusion/gpu/*.cu	Fusion kernel optimizations with auto type deduction
paddle/phi/kernels/funcs/*.{cu,cc,h}	Utility function conversions to auto

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.