[unified-memory] Hierarchical cache for every unified pool shape (#37507)

This commit is contained in:
Cheng Wan
2026-09-21 16:50:37 -07:00
committed by GitHub
parent 22587fb15c
commit 506698761d
26 changed files with 1194 additions and 116 deletions
+3
View File
@@ -11,6 +11,7 @@
"registered/disaggregation/test_disaggregation_unified_memory.py",
"registered/e2e/disaggregation/test_disaggregation_unified_memory_swa.py",
"registered/e2e/disaggregation/test_disaggregation_unified_memory_tri.py",
"registered/e2e/hicache/test_hicache_unified_memory.py",
"registered/e2e/models/test_inkling_unified.py",
"registered/e2e/models/test_kimi_linear_models.py",
"registered/e2e/models/test_kimi_linear_unified_memory.py",
@@ -37,6 +38,8 @@
"registered/unit/mem_cache/test_unified_capacity_memo.py",
"registered/unit/mem_cache/test_unified_free_no_host_sync.py",
"registered/unit/mem_cache/test_unified_handout_zeroing.py",
"registered/unit/mem_cache/test_unified_hicache_regressions.py",
"registered/unit/mem_cache/test_unified_hicache_strided_state.py",
"registered/unit/mem_cache/test_unified_mamba_views.py",
"registered/unit/mem_cache/test_unified_mha_views.py",
"registered/unit/mem_cache/test_unified_mla_gpu_parity.py",