Skip to content

Clarify Step 9 operand-sharing exercise #214

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
tlopex merged 1 commit into from
Aug 26, 2026
+2 −2
Merged

Clarify Step 9 operand-sharing exercise #214

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion chapter_gemm_advanced/index.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -906,4 +906,4 @@ In this B200 run, combining TMA, software pipelining, persistent scheduling, war

1. What happens if you set the initial `phase` to `0` for both the TMA and MMA `PipelineState` in Step 7? Draw the deadlock scenario.
2. With `cta_group=2` in Step 8, the TMA arrive byte count is `CTA_GROUP * (BLK_M*BLK_K + BLK_N*BLK_K) * F16_SIZE`. Why multiply by `CTA_GROUP` when each CTA loads its own data?
3. In Step 9, each consumer handles different M rows but the same B tile. Why is sharing B (not A) the right choice?
3. In Step 9, the two consumers extend the output tile along M: they handle different M rows but cover the same N columns. Why can they reuse B in this mapping? If they instead extended along N, which operand could be reused?
2 changes: 1 addition & 1 deletion zh/chapter_gemm_advanced/index.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -906,4 +906,4 @@ def hgemm_v9(M, N, K):

1. 第 7 步中，如果 TMA 和 MMA 的 `PipelineState` 都将初始 `phase` 设为 `0`，会发生什么？画出 deadlock 过程。
2. 第 8 步使用 `cta_group=2` 时，TMA arrival byte count 为 `CTA_GROUP * (BLK_M*BLK_K + BLK_N*BLK_K) * F16_SIZE`。既然每个 CTA 分别加载自己的数据，为什么还要乘以 `CTA_GROUP`？
3. 第 9 步中，每个 consumer 处理不同的 M rows，但使用相同的 B tile。为什么应该共享 B，而不是 A？
3. 第 9 步让两个 consumers 沿 M 维扩展输出：它们处理不同的 M rows，但覆盖相同的 N columns。为什么这种映射可以复用 B？如果改为沿 N 维扩展，哪个操作数可以被复用？
Loading