dylanyunlong/project_6
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

[muh] scan: add bi100_lookback_1B_o8 — close SM100 parity gap (7/7 lookback branches)

Browse Source 
CCCL SM100 scan lookback has 7 type-specialized branches:
  offset_size=4: 1B, 2B, 4B, 8B
  offset_size=8: 1B, 4B, 8B

muh BI-V100 previously had 6 (missing o8_1B).
This commit adds the o8_1B branch derived from SM100 ref:
  ipt_14.tpb_384.ns_228.dcid_7.l2w_775 → 1.107x
  BI-V100 delay: halved ns (L2 6MB vs 50MB): backon(114, 465)
  nominal_tile = 384*14*4 = 21504 ≤ 49152 ✓

Now: 7/7 lookback + 6/6 lookahead = 13/13 SM100 parity.
This commit is contained in:
dylanyunlon
2026-07-31 18:35:01 +08:00
parent c5a0d61851
commit 35ef79c5f8
1 changed files with 18 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
muh/include/muh/tuning/tuning_scan.cuh
View File

@@ -138,6 +138,20 @@ struct bi100_lookback_8B_o8 {
static constexpr CacheLoadModifier load_mod = LOAD_DEFAULT;
};
struct bi100_lookback_1B_o8 {
// CCCL SM100 ref: ipt_14.tpb_384.ns_228.dcid_7.l2w_775 → 1.107x
// BI-V100 derived: delay halved (L2 6MB vs 50MB), LOAD_CA matches SM100
// nominal_tile = 384*14*4 = 21504 ≤ 49152 ✓
static constexpr int threads = 384;
static constexpr int items = 14;
static constexpr LookbackDelayPolicy delay = {
LookbackDelayAlgorithm::exponential_backon, 114, 465};
static constexpr BlockLoadAlgorithm load_algo = BLOCK_LOAD_WARP_TRANSPOSE;
static constexpr BlockStoreAlgorithm store_algo = BLOCK_STORE_WARP_TRANSPOSE;
static constexpr CacheLoadModifier load_mod = LOAD_CA;
};
// --- Lookahead tunings for BI-V100 ---
struct bi100_lookahead_1B {
@@ -233,6 +247,10 @@ struct policy_selector {
}
} else if (offset_size == 8) {
switch (input_value_size) {
case 1: return {bi100_lookback_1B_o8::threads, bi100_lookback_1B_o8::items,
bi100_lookback_1B_o8::load_algo, bi100_lookback_1B_o8::load_mod,
bi100_lookback_1B_o8::store_algo, BLOCK_SCAN_WARP_SCANS,
bi100_lookback_1B_o8::delay};
case 4: return {bi100_lookback_4B_o8::threads, bi100_lookback_4B_o8::items,
bi100_lookback_4B_o8::load_algo, bi100_lookback_4B_o8::load_mod,
bi100_lookback_4B_o8::store_algo, BLOCK_SCAN_WARP_SCANS,