[GH-ISSUE #27] why your guess? "st.global.L1::no_allocate" that doesn't mean L2 won't be allocated cache. #27

New issue

Closed

opened 2026-10-26 09:52:43 +00:00 by bsakaguchi · 2 comments

bsakaguchi commented

2026-10-26 09:52:43 +00:00

Owner

Originally created by @zzJingHong on GitHub (Feb 26, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepEP/issues/27

// st.global.L1::no_allocate will be translated into ST.E.NA.[width] in SASS,
// which does not have cache allocation (obviously in L1, I guess not in L2 too)
#ifndef DISABLE_AGGRESSIVE_PTX_INSTRS
#define ST_NA_FUNC "st.global.L1::no_allocate"
#else
#define ST_NA_FUNC "st.global"
#endif

Originally created by @zzJingHong on GitHub (Feb 26, 2025). Original GitHub issue: https://github.com/deepseek-ai/DeepEP/issues/27 // `st.global.L1::no_allocate` will be translated into `ST.E.NA.[width]` in SASS, // which does not have cache allocation (obviously in L1, I guess not in L2 too) #ifndef DISABLE_AGGRESSIVE_PTX_INSTRS #define ST_NA_FUNC "st.global.L1::no_allocate" #else #define ST_NA_FUNC "st.global" #endif

bsakaguchi closed this issue

2026-10-26 09:52:44 +00:00

bsakaguchi commented

2026-10-26 09:52:45 +00:00

Author

Owner

@LyricZhao commented on GitHub (Feb 27, 2025):

Sorry, the comment is somehow misleading: here, I'm guessing the "NA" in ST.E.NA not the PTX. Will update related docs later.

@LyricZhao commented on GitHub (Feb 27, 2025): Sorry, the comment is somehow misleading: here, I'm guessing the "NA" in `ST.E.NA` not the PTX. Will update related docs later.

bsakaguchi commented

2026-10-26 09:52:46 +00:00

Author

Owner

@zsnoob commented on GitHub (Mar 29, 2025):

Yeah, though a bit misleading here, it's actually great to see these kinds of speculations and discussions. Even if unverified, they're valuable—after all, the PTX's undefined behavior is exactly the kind of thing we are speculating about in the first place.

@zsnoob commented on GitHub (Mar 29, 2025): Yeah, though a bit misleading here, it's actually great to see these kinds of speculations and discussions. Even if unverified, they're valuable—after all, the PTX's undefined behavior is exactly the kind of thing we are speculating about in the first place.

bsakaguchi referenced this issue

2026-10-26 09:53:39 +00:00

[GH-ISSUE #74] test_internode.py hangs at internode.cu:dispatch(...) #68

bsakaguchi referenced this issue

2026-10-26 09:56:29 +00:00

[GH-ISSUE #270] A timeout error occurred when running test_internode.py on four devices. #200

bsakaguchi referenced this issue

2026-10-26 09:56:40 +00:00

[GH-ISSUE #290] Running the test case on a 2-node H20 3E machine failed with internode.cu:1029 'unspecified launch failure' #208