feat(vpto): add AICore ld_dev and st_dev ops - #1125
Merged
Merged
Conversation
mouliangyu
force-pushed
the
codex/aicore-ld-dev-st-dev
branch
7 times, most recently
from
August 4, 2026 11:28
30a55f6 to
e205ea1
Compare
WenboCodes
approved these changes
Aug 4, 2026
mouliangyu
marked this pull request as ready for review
August 4, 2026 12:14
mouliangyu
force-pushed
the
codex/aicore-ld-dev-st-dev
branch
2 times, most recently
from
August 5, 2026 01:56
d14a316 to
70e413f
Compare
mouliangyu
force-pushed
the
codex/aicore-ld-dev-st-dev
branch
from
August 5, 2026 02:02
70e413f to
4b8c3c4
Compare
A5 板测成功
|
A3 板测失败
失败用例
|
A3 板测失败详情:PR #1125comm_p2p_binding_variants
comm_p2p
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
背景
普通 AICore 的 scalar GM data-cache bypass 与 SIMT
pto.ldg/pto.stg属于不同执行域和 intrinsic 家族。此前普通pto.aicore函数中的误用会通过 PTO verifier,随后使 Bisheng 在 device LLVM 编译阶段崩溃。Closes #1032
主要变更
pto.ldg/pto.stg的 placement verifier,只允许其位于pto.simt_entry或pto.section.simt。pto.ld_dev与pto.st_dev,不继承 SIMT op。!pto.ptr<T, gm>与T = i8/i16/i32/i64;offset 为元素步长,load result/store value 必须与指针元素类型一致。l1cache/l2cachepolicy 属性,也不隐含 atomic、同步、内存序、invalidate 或 writeback 语义。llvm.hivm.LD.DEV.u*.GM/llvm.hivm.ST.DEV.u*。arith、Group 17 为 SIMT;新增 Group 18Special Scalar Operations,收录 10 个 PTO scalar kernel query、pointer/address 与 scalar-memory op。docs/vpto-spec.md保留总体说明和索引,同一批 scalar op 的重复逐项定义改为链接到 Group 18,避免两处规范成为不同真源。测试
cmake --build build --target PTOASPythonPackage -j16test/vpto/scripts/run_host_vpto_validation.sh的micro-op/scalar-load-store/ld-st-devA5 simulator case 通过 fatobj/device-object emission、host build、runtime 与 golden compare。git diff --check,并检查 Group 18 恰好包含 10 个 PTO scalar op 的独立手册小节,docs/vpto-spec.md不再重复定义这些 op。Runtime 与 object compilation 当前覆盖 i32 非零 offset roundtrip;i8/i16/i32/i64 四种宽度均由 LLVM emission 用例覆盖。