Skip to content

feat: add compact mode to tile config#400

Open
FangRui0 wants to merge 1 commit intohw-native-sys:mainfrom
FangRui0:add_compact
Open

feat: add compact mode to tile config#400
FangRui0 wants to merge 1 commit intohw-native-sys:mainfrom
FangRui0:add_compact

Conversation

@FangRui0
Copy link
Copy Markdown
Contributor

@FangRui0 FangRui0 commented Mar 30, 2026

No description provided.

@FangRui0
Copy link
Copy Markdown
Contributor Author

FangRui0 commented Mar 31, 2026

/run a3

@reedhecre
Copy link
Copy Markdown

reedhecre commented Mar 31, 2026

A3 板测失败

失败用例

  • rowexpandadd (run, exit=1)
  • reshape (run, exit=1)
  • abs (run, exit=1)

@reedhecre
Copy link
Copy Markdown

reedhecre commented Mar 31, 2026

A3 板测失败详情：PR #400

rowexpandadd

stage=run info=exit=1

[ERROR] aclrtSynchronizeStream(stream) failed: 507035 (/tmp/ptoas-board-monitor/runs/20260331_111604_manual_pr400/npu_validation/Rowexpandadd/rowexpandadd/main.cpp:99)
[ERROR] RecentErrMsg: EZ9999: Inner Error!
EZ9999[PID: 920321] 2026-03-31-11:26:14.987.234 (EZ9999):  The error from device(chipId:1, dieId:0), serial number is 1081, there is an exception of aivec error, core id is 26, error code = 0x4000000000000000, dump info: pc start: 0x124400000000, current: 0x1244000000e0, vec error info: 0x2c, mte error info: 0x710300a846, ifu error info: 0x200003db50b00, ccu error info: 0xcc201000d80000f, cube error info: 0, biu error info: 0, aic error mask: 0x6500020bd00028c, para base: 0x12c100000000.[FUNC:PrintCoreInfo][FILE:device_error_core_proc.cc][LINE:389]
        TraceBack (most recent call last):
       The extend info: errcode:(0x4000000000000000, 0, 0) errorStr: VEC instruction error: the ub address out of bounds. fixp_error0 info: 0x300a846, fixp_error1 info: 0x71, fsmId:0, tslot:2, thread:0, ctxid:0, blk:0, sublk:0, subErrType:4.[FUNC:PrintCoreInfo][FILE:device_error_core_proc.cc][LINE:402]
       Kernel task happen error, retCode=0x31, [vector core exception].[FUNC:PreCheckTaskErr][FILE:davinci_kernel_task.cc][LINE:1497]
       AIV Kernel happen error, retCode=0x31.[FUNC:GetError][FILE:stream.cc][LINE:1346]
       [AIC_INFO] after execute:args print end[FUNC:GetError][FILE:stream.cc][LINE:1346]
       [DFX_INFO]Aicore kernel execute failed, device_id=2, stream_id=1471, report_stream_id=1471, task_id=0, flip_num=0, fault kernel_name=_Z23trowexpandadd_kernel_2dPfS_S_, fault kernel info ext=_Z23trowexpandadd_kernel_2dPfS_S_, program id=0, hash=9985848676763991650.[FUNC:GetError][FILE:stream.cc][LINE:1346]
       rtStreamSynchronize execution failed, reason=vector core exception[FUNC:FuncErrorReason][FILE:error_message_manage.cc][LINE:65]
       synchronize stream failed, runtime result = 507035[FUNC:ReportCallError][FILE:log_inner.cpp][LINE:148]
[2026-03-31 11:26:17] ERROR: testcase failed (exit 1): rowexpandadd
reshape

stage=run info=exit=1

[ERROR] aclrtSetDevice(deviceId) failed: 507033 (/tmp/ptoas-board-monitor/runs/20260331_111604_manual_pr400/npu_validation/Reshape/reshape/main.cpp:75)
[ERROR] RecentErrMsg: [PID: 1211395] 2026-03-31-11:26:28.107.102 Memory_Allocation_Failure(EL0004): Failed to allocate memory.
        Possible Cause: Available memory is insufficient.
        Solution: Close applications not in use.
        TraceBack (most recent call last):
        allocate device memory failed.[FUNC:Init][FILE:memory_pool_manager.cc][LINE:50]
        Fail to init MemoryPoolManager, retCode=0x711000e.[FUNC:Init][FILE:raw_device.cc][LINE:670]
        Check param failed, dev can not be NULL![FUNC:DeviceRetain][FILE:runtime.cc][LINE:3547]
        Check param failed, dev can not be NULL![FUNC:PrimaryContextRetain][FILE:runtime.cc][LINE:3171]
        Check param failed, ctx can not be NULL![FUNC:PrimaryContextRetain][FILE:runtime.cc][LINE:3202]
        Check param failed, context can not be null.[FUNC:SetDevice][FILE:api_impl.cc][LINE:3278]
        rtSetDevice execution failed, reason=device retain error[FUNC:FuncErrorReason][FILE:error_message_manage.cc][LINE:65]
        open device 2 failed, runtime result = 507033.[FUNC:ReportCallError][FILE:log_inner.cpp][LINE:148]
        ctx is NULL![FUNC:GetDevErrMsg][FILE:api_impl.cc][LINE:6029]
        The argument is invalid.Reason: rtGetDevMsg execution failed, the context is a null pointer.
[2026-03-31 11:26:29] ERROR: testcase failed (exit 1): reshape
abs

stage=run info=exit=1

[ERROR] aclrtSynchronizeStream(stream) failed: 507035 (/tmp/ptoas-board-monitor/runs/20260331_111604_manual_pr400/npu_validation/Abs/abs/main.cpp:91)
[ERROR] RecentErrMsg: EZ9999: Inner Error!
EZ9999[PID: 1745279] 2026-03-31-11:35:41.707.746 (EZ9999):  The error from device(chipId:1, dieId:0), serial number is 1082, there is an exception of aivec error, core id is 3, error code = 0x4000000000000000, dump info: pc start: 0x124400000000, current: 0x12440000008c, vec error info: 0xf01b, mte error info: 0x7a03000020, ifu error info: 0x212c2ff56f400, ccu error info: 0xcc2000029800075, cube error info: 0, biu error info: 0, aic error mask: 0x6500020bd00028c, para base: 0x12c100000000.[FUNC:PrintCoreInfo][FILE:device_error_core_proc.cc][LINE:389]
        TraceBack (most recent call last):
       The extend info: errcode:(0x4000000000000000, 0, 0) errorStr: VEC instruction error: the ub address out of bounds. fixp_error0 info: 0x3000020, fixp_error1 info: 0x7a, fsmId:1, tslot:4, thread:0, ctxid:0, blk:0, sublk:0, subErrType:4.[FUNC:PrintCoreInfo][FILE:device_error_core_proc.cc][LINE:402]
       Kernel task happen error, retCode=0x31, [vector core exception].[FUNC:PreCheckTaskErr][FILE:davinci_kernel_task.cc][LINE:1497]
       AIV Kernel happen error, retCode=0x31.[FUNC:GetError][FILE:stream.cc][LINE:1346]
       [AIC_INFO] after execute:args print end[FUNC:GetError][FILE:stream.cc][LINE:1346]
       [DFX_INFO]Aicore kernel execute failed, device_id=2, stream_id=1470, report_stream_id=1470, task_id=0, flip_num=0, fault kernel_name=_Z13abs_kernel_2dPfS_, fault kernel info ext=_Z13abs_kernel_2dPfS_, program id=0, hash=8649095210733992711.[FUNC:GetError][FILE:stream.cc][LINE:1346]
       rtStreamSynchronize execution failed, reason=vector core exception[FUNC:FuncErrorReason][FILE:error_message_manage.cc][LINE:65]
       synchronize stream failed, runtime result = 507035[FUNC:ReportCallError][FILE:log_inner.cpp][LINE:148]
[2026-03-31 11:35:43] ERROR: testcase failed (exit 1): abs
[2026-03-31 11:35:43] === SUMMARY ===
[2026-03-31 11:35:43] OK=160 FAIL=3 SKIP=0
[2026-03-31 11:35:43] RESULTS_TSV=/tmp/ptoas-board-monitor/runs/20260331_111604_manual_pr400/remote_npu_validation_results.tsv

@reedhecre
Copy link
Copy Markdown

reedhecre commented Mar 31, 2026 •
edited
Loading

Codex Review

该评论由 review 机器人自动更新。

  • PR: feat: add compact mode to tile config #400 feat: add compact mode to tile config
  • Author: FangRui0
  • Base/Head: main / add_compact
  • Head SHA: f7801ff330ae
  • Trigger: PR 有新提交
  • Generated At: 2026-03-31T13:28:36Z
  • Previous Head SHA: 29b105044d6a
  • Status: completed

Summary

PR #400 存在 2 个问题：新增的 CompactMode 没有接入现有的 C++ 编译级回归覆盖，且文档声明的文本语法与实际 parser 不一致。

Findings

  1. P2 `compile_cpp` 冒烟用例仍在编译旧版 `Tile<>` 形态，当前变更没有任何端到端编译覆盖 test/compile_cpp/abs_vec_core.cpp:35

本 PR 在 lib/PTO/Transforms/PTOToEmitC.cpp 中把所有生成的 tile 类型都改成了 Tile<..., PadValue::X, CompactMode::Y>，但仓库里唯一的编译级冒烟样例 test/compile_cpp/abs_vec_core.cpp 仍然是旧的 10 参数 Tile<..., PadValue::Null> 形式。test/compile_cpp/README.md:1 还明确说明这个文件是由 ptoas 生成的，因此它现在已经不再代表当前编译器输出。结果是：即使新的 EmitC 输出和 pto/common/pto_instr.hpp/外部工具链不兼容，仓库内的编译脚本也捕不到这个回归。对这类纯代码生成接口变更来说，这是一个重要的 CI/兼容性缺口。

  1. P3 新增的 `compact` 文本语法只接受数字，和 PR 文档宣称的枚举/示例语法不一致 lib/PTO/IR/PTOTypeDefs.cpp:223

TileBufType::parse 这里对 compact 直接调用了 parseInteger，因此 !pto.tile_buf<..., compact=normal> 这种与 blayout/slayout 一致的枚举写法会直接解析失败。与此同时，文档在 docs/PTO_IR_manual.md:204-206 声称 compact 支持 mnemonic，并给出了 #pto.tile_buf_config<row_major, none_box, 16, zero, null> 这样的示例；但 lib/PTO/IR/PTOAttrs.cpp 的 TileBufConfigAttr::parse 仍然只接受 keyed 形式和完整 attribute/int attr，因此该示例同样不可解析。也就是说，用户按 PR 新文档使用 compact mode 时会立即遇到 parser 错误。

@FangRui0
Copy link
Copy Markdown
Contributor Author

FangRui0 commented Mar 31, 2026

/run a3

@reedhecre
Copy link
Copy Markdown

reedhecre commented Mar 31, 2026

A3 板测失败

失败用例

  • mul (run, exit=1)

@reedhecre
Copy link
Copy Markdown

reedhecre commented Mar 31, 2026

A3 板测失败详情：PR #400

mul

stage=run info=exit=1

[ERROR] aclrtSynchronizeStream(stream) failed: 507035 (/tmp/ptoas-board-monitor/runs/20260331_164104_manual_pr400/npu_validation/Mul/mul/main.cpp:99)
[ERROR] RecentErrMsg: EZ9999: Inner Error!
EZ9999[PID: 4127433] 2026-03-31-16:55:12.811.284 (EZ9999):  The error from device(chipId:1, dieId:0), serial number is 1089, there is an exception of aivec error, core id is 41, error code = 0x4000000000000000, dump info: pc start: 0x124400000000, current: 0x1244000000b8, vec error info: 0xbf22, mte error info: 0x7a03000920, ifu error info: 0x212c0812000c0, ccu error info: 0x1cc600180d80000f, cube error info: 0, biu error info: 0, aic error mask: 0x6500020bd00028c, para base: 0x12c100000000.[FUNC:PrintCoreInfo][FILE:device_error_core_proc.cc][LINE:389]
        TraceBack (most recent call last):
       The extend info: errcode:(0x4000000000000000, 0, 0) errorStr: VEC instruction error: the ub address out of bounds. fixp_error0 info: 0x3000920, fixp_error1 info: 0x7a, fsmId:1, tslot:5, thread:0, ctxid:0, blk:0, sublk:0, subErrType:4.[FUNC:PrintCoreInfo][FILE:device_error_core_proc.cc][LINE:402]
       Kernel task happen error, retCode=0x31, [vector core exception].[FUNC:PreCheckTaskErr][FILE:davinci_kernel_task.cc][LINE:1497]
       AIV Kernel happen error, retCode=0x31.[FUNC:GetError][FILE:stream.cc][LINE:1346]
       [AIC_INFO] after execute:args print end[FUNC:GetError][FILE:stream.cc][LINE:1346]
       [DFX_INFO]Aicore kernel execute failed, device_id=2, stream_id=1469, report_stream_id=1469, task_id=0, flip_num=0, fault kernel_name=_Z13mul_kernel_2dPfS_S_, fault kernel info ext=_Z13mul_kernel_2dPfS_S_, program id=0, hash=16858064220790693416.[FUNC:GetError][FILE:stream.cc][LINE:1346]
       rtStreamSynchronize execution failed, reason=vector core exception[FUNC:FuncErrorReason][FILE:error_message_manage.cc][LINE:65]
       synchronize stream failed, runtime result = 507035[FUNC:ReportCallError][FILE:log_inner.cpp][LINE:148]
[2026-03-31 16:55:14] ERROR: testcase failed (exit 1): mul

@FangRui0
Copy link
Copy Markdown
Contributor Author

FangRui0 commented Mar 31, 2026

/run a3

@reedhecre
Copy link
Copy Markdown

reedhecre commented Mar 31, 2026

A3 板测成功

  • 触发方式：manual
  • 源码提交：12f55ce4e7e1
  • 结果汇总：OK 163 / FAIL 0 / SKIP 0
  • 日志：/tmp/ptoas-board-monitor/logs/20260331_184205_manual_pr400.log
  • 结果 TSV：/tmp/ptoas-board-monitor/logs/20260331_184205_manual_pr400.tsv
  • 手动指令：/run a3
  • 触发人：FangRui0
  • 触发评论：https://github.com/zhangstevenunity/PTOAS/pull/400#issuecomment-4161563812

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

pto project
Status: Todo

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@FangRui0 @reedhecre @jiashu