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

[muh_kernel_map] fix syntax error + add bytes_in_flight from CCCL babelstream benchmark

Browse Source 
Two missing commas in BI_V100 dict caused Python SyntaxError.

Added bytes_in_flight=64KB from bench_bi100.py real-hardware data:
  bif=8 (64KB) beat bif=0 (32KB) at all problem sizes
  Top 30 transform/float16 results ALL have bif=8
  Physical: BW/SM × HBM_latency = 56 GB/s × 1100ns ≈ 62KB
  CCCL cc_to_min_bytes_in_flight: B200=64KB (54 GB/s/SM)
  BI-V100 matches B200 tier, not A100 (16KB) or V100 (12KB)

Source: cccl_upstream/cub/benchmarks/bench/transform/babelstream.cu
and cccl_upstream/cub/cub/device/dispatch/tuning/tuning_transform.cuh
cc_to_min_bytes_in_flight() function.
This commit is contained in:
muh-bot
2026-08-05 03:34:08 +00:00
parent f8153d492a
commit 162a45d4ea
1 changed files with 6 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
muh_kernel_map.py
View File

@@ -49,9 +49,13 @@ BI_V100 = {
"max_registers_per_thread": 255,
"l2_cache_size_bytes": 6 * 1024 * 1024, # 6 MiB
"memory_bandwidth_gbps": 900,
"sm_count": 16 # CONFIRMED 2026-08-01,
"sm_count": 16, # CONFIRMED 2026-08-01
# Derived
"bandwidth_per_sm_gbps": 900 / 16 # 56.25 GB/s per SM, # 18 GB/s ≈ A100 level
"bandwidth_per_sm_gbps": 900 / 16, # 56.25 GB/s per SM ≈ B200 level
# bytes_in_flight: BW/SM × HBM_latency = 56 GB/s × 1100ns ≈ 62KB → 64KB
# Confirmed by bench_bi100.py transform/float16: bif=8 (64KB) wins at all sizes
# CCCL ref: B200=64KB, H100=48KB, A100=16KB, V100=12KB
"bytes_in_flight": 64 * 1024,
}
SM100 = {