Closed philipturner closed 1 year ago
It doesn't yield meaningful improvement (show slightly regression):
./kmeans1d (before the change)
elapsed: 6895ms
elapsed: 6771ms
elapsed: 6762ms
elapsed: 6771ms
elapsed: 6777ms
elapsed: 6768ms
elapsed: 6787ms
elapsed: 6807ms
elapsed: 6765ms
elapsed: 6736ms
./kmeans1d (after the change)
elapsed: 7123ms
elapsed: 6942ms
elapsed: 6914ms
elapsed: 7132ms
elapsed: 6961ms
elapsed: 6934ms
elapsed: 6950ms
elapsed: 6850ms
elapsed: 6892ms
elapsed: 6863ms
Changed used for above test: https://github.com/liuliu/ccv/commit/2210ba756bdd7077e1527d2e324fa76b46667111
liu@sz77:~/workspace/ccv/lib$ clang --version
clang version 16.0.0
Target: x86_64-unknown-linux-gnu
Thread model: posix
InstalledDir: /usr/local/bin
Compile flags:
clang ccv_numeric.c -o ccv_numeric.o -c -O3 -ffast-math -Wall -I"." -I/usr/local/cuda/include -fopenmp -fblocks -msse2 -D HAVE_CBLAS -D HAVE_LIBPNG -D HAVE_LIBJPEG -D HAVE_FFTW3 -D HAVE_PTHREAD -D HAVE_LIBLINEAR -D HAVE_TESSERACT -D HAVE_AVCODEC -D HAVE_AVFORMAT -D HAVE_AVUTIL -D HAVE_SWSCALE -D USE_OPENMP -D USE_DISPATCH -D HAVE_SSE2 -D HAVE_GSL -D HAVE_CUDA -D HAVE_CUDNN -D HAVE_NCCL -D USE_SYSTEM_CUB -I/usr/local/include
Link flags:
clang -o kmeans1d kmeans1d.o -L"../lib" -lccv -L/usr/local/lib -lm -lcblas -latlas -lrt -lpng -ljpeg -lfftw3 -lfftw3f -lpthread -lpthread -llinear -ltesseract -lavcodec -lavformat -lavutil -lswscale -fopenmp -ldispatch -lBlocksRuntime -lgsl -lgslcblas -lcuda -lcudart -lcublas -lcurand -lstdc++ -L/usr/local/cuda/lib64 -lcudnn -lnccl
Machine info:
liu@sz77:~/workspace/ccv/lib$ uname -a
Linux sz77 5.19.0-45-generic #46~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Wed Jun 7 15:06:04 UTC 20 x86_64 x86_64 x86_64 GNU/Linux
liu@sz77:~/workspace/ccv/lib$ lscpu
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 43 bits physical, 48 bits virtual
Byte Order: Little Endian
CPU(s): 64
On-line CPU(s) list: 0-63
Vendor ID: AuthenticAMD
Model name: AMD Ryzen Threadripper 3970X 32-Core Processor
CPU family: 23
Model: 49
Thread(s) per core: 2
Core(s) per socket: 32
Socket(s): 1
Stepping: 0
Frequency boost: enabled
CPU max MHz: 3700.0000
CPU min MHz: 2200.0000
BogoMIPS: 7400.61
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp
lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes
xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_co
re perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed ad
x smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru w
bnoinvd amd_ppin arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload v
gif v_spec_ctrl umip rdpid overflow_recov succor smca sev sev_es
Virtualization features:
Virtualization: AMD-V
Caches (sum of all):
L1d: 1 MiB (32 instances)
L1i: 1 MiB (32 instances)
L2: 16 MiB (32 instances)
L3: 128 MiB (8 instances)
NUMA:
NUMA node(s): 1
NUMA node0 CPU(s): 0-63
Vulnerabilities:
Itlb multihit: Not affected
L1tf: Not affected
Mds: Not affected
Meltdown: Not affected
Mmio stale data: Not affected
Retbleed: Mitigation; untrained return thunk; SMT enabled with STIBP protection
Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl
Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Spectre v2: Mitigation; Retpolines, IBPB conditional, STIBP always-on, RSB filling, PBRSB-eIBRS Not affected
Srbds: Not affected
Tsx async abort: Not affected
This pull request speeds up the single-core CPU performance for un-vectorized K-means. The optimization is generally applicable, even to a vectorized implementation, because it simply elides computations in the critical path.
I am not sure how much speedup this actually provides, and have not tested its correctness. I have not compiled CCV and any respective Swift code calling into K-means. However, you could just test the improvement to
_kmeans1d_lookup
, check for any regressions, then test_kmeans1d_lookup_compare
afterward.