Elliot/measurements

2026-05-21 16:21:07 +02:00
parent 3f26921177
commit ba19c3876f
30 changed files with 1115 additions and 33 deletions
@@ -0,0 +1,42 @@
+rm -f main_* sizes run_tests *.o *.s variants/*.s variants/*.o headers/*.s headers/*.o
+/usr/bin/g++ main.cpp ./variants/variant_avx512_1.cpp -o main_avx512_1 -O3 -march=native -std=c++23 -Wall -Wextra -flto -lsleef -DFLOPCOUNT=0 -DRANDOM_SEED=0 -DROOFLINE=0 -DBENCHMARK=1
+=== N_P=50 N_ITER=1000 ===
+W(n) per iteration: 184500.00 flops
+Q(n) per iteration: 6512 bytes
+I(n): 28.33 flops/byte
+W(n): 184500000.00 flops
+T(n): 63060640.00 cycles
+P(n): 2.93 flops/cycle
+
+=== N_P=100 N_ITER=1000 ===
+W(n) per iteration: 729000.00 flops
+Q(n) per iteration: 12912 bytes
+I(n): 56.46 flops/byte
+W(n): 729000000.00 flops
+T(n): 215114100.00 cycles
+P(n): 3.39 flops/cycle
+
+=== N_P=200 N_ITER=1000 ===
+W(n) per iteration: 2898000.00 flops
+Q(n) per iteration: 25712 bytes
+I(n): 112.71 flops/byte
+W(n): 2898000000.00 flops
+T(n): 744500640.00 cycles
+P(n): 3.89 flops/cycle
+
+=== N_P=500 N_ITER=1000 ===
+W(n) per iteration: 18045000.00 flops
+Q(n) per iteration: 64112 bytes
+I(n): 281.46 flops/byte
+W(n): 18045000000.00 flops
+T(n): 4480783600.00 cycles
+P(n): 4.03 flops/cycle
+
+=== N_P=1000 N_ITER=1000 ===
+W(n) per iteration: 72090000.00 flops
+Q(n) per iteration: 128112 bytes
+I(n): 562.71 flops/byte
+W(n): 72090000000.00 flops
+T(n): 16918374740.00 cycles
+P(n): 4.26 flops/cycle
+
@@ -0,0 +1,42 @@
+rm -f main_* sizes run_tests *.o *.s variants/*.s variants/*.o headers/*.s headers/*.o
+/usr/bin/g++ main.cpp ./variants/variant_avx512_1.cpp -o main_avx512_1 -O3 -march=native -std=c++23 -Wall -Wextra -flto -lsleef -DFLOPCOUNT=0 -DRANDOM_SEED=0 -DROOFLINE=0 -DBENCHMARK=1
+=== N_P=50 N_ITER=1000 ===
+W(n) per iteration: 184500.00 flops
+Q(n) per iteration: 6512 bytes
+I(n): 28.33 flops/byte
+W(n): 184500000.00 flops
+T(n): 62465750.00 cycles
+P(n): 2.95 flops/cycle
+
+=== N_P=100 N_ITER=1000 ===
+W(n) per iteration: 729000.00 flops
+Q(n) per iteration: 12912 bytes
+I(n): 56.46 flops/byte
+W(n): 729000000.00 flops
+T(n): 213111220.00 cycles
+P(n): 3.42 flops/cycle
+
+=== N_P=200 N_ITER=1000 ===
+W(n) per iteration: 2898000.00 flops
+Q(n) per iteration: 25712 bytes
+I(n): 112.71 flops/byte
+W(n): 2898000000.00 flops
+T(n): 749193840.00 cycles
+P(n): 3.87 flops/cycle
+
+=== N_P=500 N_ITER=1000 ===
+W(n) per iteration: 18045000.00 flops
+Q(n) per iteration: 64112 bytes
+I(n): 281.46 flops/byte
+W(n): 18045000000.00 flops
+T(n): 4481202620.00 cycles
+P(n): 4.03 flops/cycle
+
+=== N_P=1000 N_ITER=1000 ===
+W(n) per iteration: 72090000.00 flops
+Q(n) per iteration: 128112 bytes
+I(n): 562.71 flops/byte
+W(n): 72090000000.00 flops
+T(n): 17063409500.00 cycles
+P(n): 4.22 flops/cycle
+
@@ -0,0 +1,42 @@
+rm -f main_* sizes run_tests *.o *.s variants/*.s variants/*.o headers/*.s headers/*.o
+/usr/bin/g++ main.cpp ./variants/variant_avx512_1.cpp -o main_avx512_1 -O3 -march=native -std=c++23 -Wall -Wextra -flto -lsleef -DFLOPCOUNT=0 -DRANDOM_SEED=0 -DROOFLINE=0 -DBENCHMARK=1
+=== N_P=50 N_ITER=1000 ===
+W(n) per iteration: 184500.00 flops
+Q(n) per iteration: 6512 bytes
+I(n): 28.33 flops/byte
+W(n): 184500000.00 flops
+T(n): 62186140.00 cycles
+P(n): 2.97 flops/cycle
+
+=== N_P=100 N_ITER=1000 ===
+W(n) per iteration: 729000.00 flops
+Q(n) per iteration: 12912 bytes
+I(n): 56.46 flops/byte
+W(n): 729000000.00 flops
+T(n): 214174540.00 cycles
+P(n): 3.40 flops/cycle
+
+=== N_P=200 N_ITER=1000 ===
+W(n) per iteration: 2898000.00 flops
+Q(n) per iteration: 25712 bytes
+I(n): 112.71 flops/byte
+W(n): 2898000000.00 flops
+T(n): 739779420.00 cycles
+P(n): 3.92 flops/cycle
+
+=== N_P=500 N_ITER=1000 ===
+W(n) per iteration: 18045000.00 flops
+Q(n) per iteration: 64112 bytes
+I(n): 281.46 flops/byte
+W(n): 18045000000.00 flops
+T(n): 4486436580.00 cycles
+P(n): 4.02 flops/cycle
+
+=== N_P=1000 N_ITER=1000 ===
+W(n) per iteration: 72090000.00 flops
+Q(n) per iteration: 128112 bytes
+I(n): 562.71 flops/byte
+W(n): 72090000000.00 flops
+T(n): 16950058060.00 cycles
+P(n): 4.25 flops/cycle
+