Skip to content

mtl5_bench capability card -- ryzen-9-8945hs

The platform benchmark’s capability cards for the ryzen-9-8945hs machine described on the benchmark systems page: every operation family, measured against this machine’s FMA peak and memory bandwidth (#721). This page shows the latest published session; every session’s raw results stay in the data directory.

Repeatability: 580 of 666 entries within +-5% of the previous session, measured after a cold boot on the same commit (median ratio 0.998; benchmarks/compare_sessions.py). The background_activity records show a Defender security-intelligence update at 03:22Z during this session’s native-t8 run, the run with the most entries outside the band; no Defender scan ran in either session. Relates to #721 and #737.

session20261112T032806Z (UTC)
commitcd6762a6e989
classmedium
runsblas at 1 thread, blas at 8 threads, native-fast at 1 thread, native-fast at 8 threads, native at 1 thread, native at 8 threads
failed checks0
CPUs0,2,4,6,8,10,12,14 (pinned)
clock policypower scheme copy of Balanced with processor state 100-100% and EPP 0, boost aggressive; AC (governor mtl5_performance_-_temporary_copy_of_Balanced)
raw databenchmarks/data/ryzen-9-8945hs/mtl5_bench/20261112T032806Z/
earlier sessions20261108T231039Z, 20261112T030915Z

Each column is one run of this session: the build variant and its thread count (1T is one thread). The runs follow below, each with its full entry table.

areanative 1Tnative 8Tnative-fast 1Tnative-fast 8Tblas 1Tblas 8T
blas120%84%103%110%18%81%
blas244%61%54%62%76%100%
blas312%2%14%5%71%44%
factor3%1%3%1%13%2%
eigen3%0%4%0%17%2%
elementwise7%5%7%6%6%6%
sparse22%32%21%36%22%21%
direct0%0%0%0%0%0%
orderingn/an/an/an/an/an/a
krylovn/an/an/an/an/an/a
pcn/an/an/an/an/an/a
smoother31%28%30%29%31%30%
mg29%29%29%29%29%29%
sparse_eign/an/an/an/an/an/a
mixed11%11%59%58%11%11%
integern/an/an/an/an/an/a

The median time of each of the 111 entries in each run. The fastest run of each entry is in bold; — is an entry the run did not measure.

areaoptypeshapenative 1Tnative 8Tnative-fast 1Tnative-fast 8Tblas 1Tblas 8T
blas1dotf64n=838860812.291 ms2.315 ms2.500 ms2.292 ms14.419 ms2.351 ms
blas1dotf32n=838860812.257 ms1.565 ms1.255 ms1.009 ms12.263 ms1.562 ms
blas1axpyf64n=838860814.961 ms3.541 ms3.595 ms3.519 ms14.961 ms3.521 ms
blas1axpyf32n=838860814.906 ms1.955 ms1.749 ms1.671 ms12.802 ms1.798 ms
blas1scalef64n=83886084.296 ms2.115 ms2.000 ms2.111 ms2.645 ms2.134 ms
blas1two_normf64n=838860811.731 ms1.501 ms1.181 ms1.061 ms15.421 ms1.698 ms
blas1one_normf64n=83886086.391 ms6.389 ms6.395 ms6.417 ms7.357 ms6.392 ms
blas1infinity_normf64n=83886086.405 ms6.406 ms6.400 ms6.401 ms6.411 ms6.405 ms
blas1sumf64n=83886086.391 ms6.393 ms6.390 ms6.394 ms6.396 ms6.394 ms
blas1dot_k3f64n=8388608109.635 ms109.711 ms105.029 ms105.179 ms111.774 ms111.710 ms
blas2gemvf64n=40963.382 ms2.385 ms3.498 ms2.335 ms3.276 ms2.570 ms
blas2gemvf32n=40962.967 ms1.181 ms1.840 ms1.071 ms1.651 ms1.148 ms
blas2gemv_colmajorf64n=40967.779 ms2.362 ms22.253 ms10.106 ms3.593 ms2.678 ms
blas2gemv_colmajorf32n=40962.538 ms1.182 ms8.261 ms5.084 ms1.556 ms1.244 ms
blas2symvf64n=409610.435 ms10.366 ms10.434 ms10.181 ms2.032 ms1.273 ms
blas2trmvf64n=40962.044 ms2.114 ms1.956 ms1.977 ms1.707 ms1.237 ms
blas2trmvf32n=40961.839 ms1.823 ms1.243 ms1.167 ms0.968 ms0.540 ms
blas2trsvf64n=40962.034 ms2.076 ms2.337 ms1.960 ms1.693 ms1.820 ms
blas2trsvf32n=40961.861 ms1.869 ms1.194 ms1.024 ms0.921 ms0.932 ms
blas2gerf64n=409616.267 ms16.210 ms20.381 ms16.364 ms16.270 ms16.217 ms
blas3gemmf64512x512x51220.011 ms2.533 ms21.614 ms2.732 ms4.563 ms0.682 ms
blas3gemmf32512x512x5129.957 ms1.266 ms10.825 ms1.553 ms2.374 ms0.379 ms
blas3gemmtf64512x512x51213.757 ms13.912 ms13.861 ms5.625 ms3.425 ms1.672 ms
blas3symmf64512x512x51221.110 ms19.756 ms22.713 ms3.886 ms4.749 ms0.684 ms
blas3syrkf64512x512x51216.862 ms16.886 ms21.422 ms17.038 ms3.710 ms2.007 ms
blas3syr2kf64512x512x51229.066 ms28.757 ms34.823 ms30.512 ms5.998 ms2.725 ms
blas3trmmf64512x512x51230.766 ms31.264 ms17.339 ms6.934 ms2.361 ms0.451 ms
blas3trsmf64512x512x51235.235 ms37.072 ms14.950 ms7.357 ms3.504 ms0.613 ms
blas3trsmf32512x512x51232.124 ms33.483 ms11.683 ms5.496 ms1.637 ms0.301 ms
blas3kronf6448x48 (x) 48x4810.902 ms10.884 ms14.767 ms11.229 ms11.129 ms11.133 ms
factorluf64512x51240.693 ms27.621 ms62.529 ms27.520 ms6.393 ms11.145 ms
factorluf32512x51238.427 ms30.019 ms60.922 ms26.059 ms3.063 ms5.082 ms
factorcholeskyf64512x51214.327 ms14.367 ms14.793 ms14.292 ms5.484 ms6.993 ms
factorcholeskyf32512x51213.712 ms13.720 ms13.714 ms13.720 ms2.114 ms2.442 ms
factorldltf64512x51214.412 ms14.416 ms14.410 ms14.397 ms14.811 ms14.818 ms
factorldlt_bkf64512x512203.839 ms205.959 ms205.022 ms211.274 ms211.072 ms211.419 ms
factorqrf64512x512810.843 ms500.573 ms815.861 ms498.148 ms12.960 ms14.125 ms
factorlqf64512x51283.463 ms54.157 ms81.457 ms52.735 ms81.665 ms54.165 ms
factorhessenbergf64512x512933.562 ms576.677 ms932.770 ms576.798 ms935.107 ms573.414 ms
factorhouseholderf64512x5124.961 ms1.267 ms4.911 ms1.262 ms4.952 ms1.269 ms
factorinvf64512x512110.021 ms98.837 ms111.077 ms98.454 ms76.588 ms82.207 ms
factorlu_iterative_refinef32>f64512x51241.893 ms29.059 ms41.410 ms29.304 ms6.977 ms8.818 ms
eigeneigenvalue_symmetricf64256x25649.700 ms49.481 ms48.094 ms47.850 ms2.121 ms4.261 ms
eigeneigen_symmetricf64256x256164.127 ms163.409 ms166.196 ms159.964 ms160.066 ms164.039 ms
eigeneigen_symmetric_jacobif64256x256638.466 ms641.051 ms649.868 ms634.102 ms650.132 ms619.276 ms
eigeneigenvaluef64256x25675.241 ms75.625 ms75.800 ms74.892 ms20.908 ms25.142 ms
eigenschurf64256x256144.383 ms156.657 ms141.079 ms143.368 ms27.263 ms31.143 ms
eigensvdf64256x256149.555 ms176.201 ms149.845 ms149.733 ms10.946 ms13.775 ms
eigenexpmf64256x25637.397 ms23.813 ms32.603 ms23.283 ms20.395 ms21.439 ms
eigensylvesterf64256x256329.581 ms324.020 ms332.324 ms326.885 ms91.010 ms101.372 ms
elementwiseexpf64n=838860839.838 ms50.395 ms33.426 ms33.551 ms33.461 ms33.422 ms
elementwiselogf64n=838860852.323 ms61.334 ms51.729 ms51.693 ms53.847 ms53.304 ms
elementwisesinf64n=838860848.041 ms48.002 ms48.967 ms48.948 ms47.973 ms47.965 ms
elementwisesqrtf64n=838860826.584 ms26.497 ms26.439 ms26.453 ms26.544 ms26.549 ms
elementwiseerff64n=838860838.837 ms38.790 ms38.976 ms39.016 ms42.476 ms42.414 ms
sparsespmvf64512^2 grid, nnz 13086720.932 ms0.277 ms0.919 ms0.213 ms0.914 ms0.202 ms
sparsespmv_transf64512^2 grid, nnz 13086722.156 ms1.618 ms1.456 ms1.441 ms1.433 ms2.480 ms
sparsecrs_densef64512^2 grid, nnz 1308672 x 167.913 ms2.355 ms7.954 ms2.293 ms7.879 ms2.351 ms
sparsespgemmf64256^2 grid, nnz 32665622.429 ms30.168 ms21.535 ms24.368 ms22.341 ms24.035 ms
sparselower_solve_levelf64192^2 grid, nnz(L) 10035641.960 ms1.946 ms1.932 ms1.950 ms1.955 ms1.878 ms
directsparse_choleskyf6496^2 grid, nnz 45696138.306 ms135.235 ms127.373 ms126.374 ms129.966 ms124.491 ms
directsparse_ldltf6496^2 grid, nnz 4569639.556 ms40.139 ms37.139 ms35.577 ms37.994 ms38.809 ms
directsupernodal_ldltf6496^2 grid, nnz 4569622.580 ms23.098 ms20.192 ms18.848 ms22.885 ms22.820 ms
directsparse_luf6496^2 grid, nnz 4569667.183 ms67.811 ms78.922 ms66.367 ms67.095 ms66.970 ms
directsupernodal_luf6496^2 grid, nnz 4569691.196 ms92.133 ms91.607 ms89.705 ms91.993 ms92.117 ms
directnative_kluf6496^2 grid, nnz 4569630.980 ms33.135 ms32.577 ms31.939 ms31.939 ms31.735 ms
directsparse_qrf6448^2 grid, nnz 1132895.396 ms96.896 ms104.595 ms104.711 ms101.803 ms100.413 ms
orderingamdf6496^2 grid, nnz 45696, scrambled4.575 ms4.619 ms4.589 ms4.627 ms4.551 ms4.632 ms
orderingcolamdf6496^2 grid, nnz 45696, columns scrambled6.975 ms6.994 ms6.967 ms6.960 ms6.957 ms7.013 ms
orderingrcmf6496^2 grid, nnz 45696, scrambled0.502 ms0.500 ms0.500 ms0.506 ms0.504 ms0.509 ms
orderingblock_triangular_formf6432 planted blocks, n 819217.880 ms17.897 ms17.939 ms18.409 ms17.597 ms17.462 ms
krylovcg+ic0f64128^2 grid, n 1638449.762 ms49.792 ms41.333 ms41.242 ms50.950 ms51.032 ms
krylovminres+diagf64128^2 grid, n 1638454.686 ms55.816 ms39.078 ms39.075 ms57.576 ms57.630 ms
krylovbicg+ilu0f64128^2 grid, n 1638438.497 ms38.615 ms34.910 ms34.667 ms39.396 ms39.467 ms
krylovbicgstab+ilu0f64128^2 grid, n 1638421.068 ms23.293 ms17.053 ms17.010 ms21.799 ms21.792 ms
krylovbicgstab_ell+ilu0f64128^2 grid, n 1638427.404 ms31.942 ms22.667 ms22.618 ms28.238 ms28.238 ms
krylovcgs+ilu0f64128^2 grid, n 1638422.309 ms26.083 ms19.877 ms19.880 ms22.672 ms22.678 ms
krylovgmres30+ilu0f64128^2 grid, n 1638464.560 ms75.592 ms39.245 ms39.030 ms68.911 ms68.981 ms
krylovqmr+ilu0f64128^2 grid, n 1638442.778 ms49.956 ms36.837 ms36.698 ms43.678 ms43.744 ms
krylovtfqmr+ilu0f64128^2 grid, n 1638425.174 ms29.295 ms21.978 ms21.926 ms25.558 ms25.561 ms
krylovidr_s1+ilu0f64128^2 grid, n 1638429.857 ms34.724 ms24.450 ms24.490 ms30.784 ms30.797 ms
krylovidr_s2+ilu0f64128^2 grid, n 1638428.791 ms33.654 ms23.283 ms23.294 ms29.718 ms29.839 ms
krylovidr_s4+ilu0f64128^2 grid, n 1638432.173 ms32.076 ms24.774 ms24.740 ms33.378 ms33.278 ms
krylovidr_s8+ilu0f64128^2 grid, n 1638439.159 ms39.219 ms27.857 ms28.077 ms41.427 ms41.389 ms
pcdiagonalf64128^2 grid, rows scaled0.046 ms0.045 ms0.048 ms0.048 ms0.046 ms0.048 ms
pcilu_0f64128^2 grid2.935 ms2.980 ms2.923 ms2.974 ms2.938 ms2.994 ms
pcilutf64128^2 grid28.281 ms28.990 ms28.356 ms29.144 ms27.106 ms28.485 ms
pcic_0f64128^2 grid2.260 ms2.243 ms2.284 ms2.275 ms2.287 ms2.259 ms
pcildlf64128^2 grid2.162 ms2.194 ms2.183 ms2.189 ms2.503 ms2.160 ms
pcssorf64128^2 grid0.028 ms0.026 ms0.027 ms0.027 ms0.031 ms0.027 ms
pcblock_diagonal4f64128^2 grid, rows scaled0.807 ms0.811 ms0.804 ms0.813 ms1.765 ms1.638 ms
smootherjacobif64512^2 grid, nnz 13086721.776 ms1.793 ms1.769 ms1.828 ms1.756 ms1.751 ms
smoothergauss_seidelf64512^2 grid, nnz 13086721.829 ms1.942 ms1.846 ms1.895 ms1.838 ms1.833 ms
smoothersor1.5f64512^2 grid, nnz 13086722.221 ms2.270 ms2.915 ms2.219 ms2.219 ms2.219 ms
mgpoisson2d_vcyclef64255^2 grid, 6 levels, 10 V(1,1)19.376 ms18.979 ms19.383 ms18.781 ms19.322 ms18.703 ms
sparse_eiglanczosf6464x65 Laplacian, k=4, m=8040.561 ms40.609 ms40.571 ms40.667 ms40.615 ms40.472 ms
sparse_eigarnoldif6464x65 convection-diffusion, k=4, m=8058.794 ms58.685 ms58.743 ms58.633 ms41.327 ms41.255 ms
sparse_eigshift_invertf6464x65 Laplacian, k=4 nearest 0, m=4066.987 ms66.880 ms72.118 ms71.690 ms69.312 ms70.565 ms
sparse_eigpower_iterationf6464x65 Laplacian, 300 steps9.700 ms9.719 ms8.099 ms8.076 ms9.928 ms9.892 ms
mixeddotf32>f64n=838860811.733 ms11.742 ms2.174 ms2.173 ms14.628 ms14.380 ms
mixedgemmf32>f64512x512x51236.662 ms4.817 ms19.251 ms3.154 ms37.348 ms4.877 ms
mixedblock_scaled_doti8+f32n=8388608, block 321.725 ms1.721 ms0.577 ms0.576 ms1.836 ms1.836 ms
integerdoti32n=8388608, |v| <= 154.276 ms5.016 ms1.445 ms1.453 ms4.277 ms4.322 ms
integerdoti16>i32n=8388608, |v| <= 153.213 ms3.219 ms0.497 ms0.496 ms3.235 ms3.236 ms
integerdoti8>i32n=8388608, |v| <= 151.727 ms1.727 ms0.430 ms0.429 ms1.726 ms1.726 ms
integerdotu8i8>i32n=8388608, |v| <= 151.727 ms1.728 ms0.386 ms0.384 ms1.726 ms1.726 ms
integergemmi32512^3, |v| <= 20479.867 ms1.278 ms18.877 ms2.880 ms10.238 ms1.304 ms
integergemmi16>i32512^3, |v| <= 204738.326 ms5.022 ms21.941 ms3.039 ms38.570 ms5.034 ms
integergemmi8>i32512^3, |v| <= 12738.661 ms5.043 ms22.451 ms3.063 ms38.705 ms5.041 ms
integergemm_quadi8>i32512^3, |v| <= 12738.664 ms5.053 ms4.603 ms0.787 ms38.695 ms5.046 ms
integergemm_quadu8i8>i32512^3, |v| <= 12734.464 ms5.173 ms4.559 ms0.783 ms34.459 ms4.533 ms

ryzen-9-8945hs — native, 1 thread, class medium

Section titled “ryzen-9-8945hs — native, 1 thread, class medium”
CPUAMD Ryzen 9 8945HS w/ Radeon 780M Graphics (x86_64, 16 logical cores)
OSWindows 10.0.26300 (26H2)
compilerMSVC 19.52.37029
build/DWIN32 /D_WINDOWS /EHsc /O2 /Ob2 /DNDEBUG, SIMD scalar
commitcd6762a6e989
FMA peak (ceiling)f64 78.4 GFLOP/s, f32 126.4 GFLOP/s
… hardware, AVX-512Ff64 78.4, f32 126.4 GFLOP/s
… this build, scalarf64 1.9, f32 1.9 GFLOP/s
bandwidthread 36.8, update 51.7, triad 37.8 GB/s (ceiling 51.7)
last-level cache16 MiB
areaentriesmedian %roofbest %rooffailed checks
blas11020%60%0
blas21044%77%0
blas31012%21%0
factor123%4%0
eigen83%9%0
elementwise57%10%0
sparse522%56%0
direct70%1%0
ordering4n/an/a0
krylov13n/an/a0
pc7n/an/a0
smoother331%32%0
mg129%29%0
sparse_eig4n/an/a0
mixed311%20%0
integer9n/an/a0
Every entry of this run
areaoptypeshapemedianGFLOP/sGB/s%roofcheck
blas1dotf64n=838860812.291 ms1.3710.9221%ok
blas1dotf32n=838860812.257 ms1.375.4711%ok
blas1axpyf64n=838860814.961 ms1.1213.4626%ok
blas1axpyf32n=838860814.906 ms1.136.7513%ok
blas1scalef64n=83886084.296 ms1.9531.2460%ok
blas1two_normf64n=838860811.731 ms1.435.7211%ok
blas1one_normf64n=83886086.391 ms1.3110.5020%ok
blas1infinity_normf64n=83886086.405 ms1.3110.4820%ok
blas1sumf64n=83886086.391 ms1.3110.5020%ok
blas1dot_k3f64n=8388608109.635 ms1.681.222%ok
blas2gemvf64n=40963.382 ms9.9239.7177%ok
blas2gemvf32n=40962.967 ms11.3122.6344%ok
blas2gemv_colmajorf64n=40967.779 ms4.3117.2633%ok
blas2gemv_colmajorf32n=40962.538 ms13.2226.4551%ok
blas2symvf64n=409610.435 ms3.226.4412%ok
blas2trmvf64n=40962.044 ms8.2132.8764%ok
blas2trmvf32n=40961.839 ms9.1218.2735%ok
blas2trsvf64n=40962.034 ms8.2533.0264%ok
blas2trsvf32n=40961.861 ms9.0218.0535%ok
blas2gerf64n=409616.267 ms2.0616.5132%ok
blas3gemmf64512x512x51220.011 ms13.410.3117%ok
blas3gemmf32512x512x5129.957 ms26.960.3221%ok
blas3gemmtf64512x512x51213.757 ms9.760.4612%ok
blas3symmf64512x512x51221.110 ms12.720.3016%ok
blas3syrkf64512x512x51216.862 ms7.960.2510%ok
blas3syr2kf64512x512x51229.066 ms9.240.2212%ok
blas3trmmf64512x512x51230.766 ms4.360.176%ok
blas3trsmf64512x512x51235.235 ms3.810.155%ok
blas3trsmf32512x512x51232.124 ms4.180.083%ok
blas3kronf6448x48 (x) 48x4810.902 ms0.493.908%ok
factorluf64512x51240.693 ms2.200.103%ok
factorluf32512x51238.427 ms2.330.052%ok
factorcholeskyf64512x51214.327 ms3.120.294%ok
factorcholeskyf32512x51213.712 ms3.260.153%ok
factorldltf64512x51214.412 ms3.100.294%ok
factorldlt_bkf64512x512203.839 ms0.220.020%ok
factorqrf64512x512810.843 ms0.220.010%ok
factorlqf64512x51283.463 ms2.140.053%ok
factorhessenbergf64512x512933.562 ms0.480.001%ok
factorhouseholderf64512x5124.961 ms0.210.852%*ok
factorinvf64512x512110.021 ms2.440.043%ok
factorlu_iterative_refinef32>f64512x51241.893 ms2.140.103%ok
eigeneigenvalue_symmetricf64256x25649.700 ms0.450.021%ok
eigeneigen_symmetricf64256x256164.127 ms0.920.011%ok
eigeneigen_symmetric_jacobif64256x256638.466 ms0.240.000%ok
eigeneigenvaluef64256x25675.241 ms2.230.013%ok
eigenschurf64256x256144.383 ms2.900.014%ok
eigensvdf64256x256149.555 ms2.470.013%ok
eigenexpmf64256x25637.397 ms6.730.039%ok
eigensylvesterf64256x256329.581 ms3.050.004%ok
elementwiseexpf64n=838860839.838 ms0.213.377%ok
elementwiselogf64n=838860852.323 ms0.162.575%ok
elementwisesinf64n=838860848.041 ms0.172.795%ok
elementwisesqrtf64n=838860826.584 ms0.325.0510%ok
elementwiseerff64n=838860838.837 ms0.223.467%ok
sparsespmvf64512^2 grid, nnz 13086720.932 ms2.8129.2256%ok
sparsespmv_transf64512^2 grid, nnz 13086722.156 ms1.2112.6324%ok
sparsecrs_densef64512^2 grid, nnz 1308672 x 167.913 ms5.2911.1322%ok
sparsespgemmf64256^2 grid, nnz 32665622.429 ms0.150.511%*ok
sparselower_solve_levelf64192^2 grid, nnz(L) 10035641.960 ms1.028.4916%*ok
directsparse_choleskyf6496^2 grid, nnz 45696138.306 ms0.080.010%ok
directsparse_ldltf6496^2 grid, nnz 4569639.556 ms0.280.020%ok
directsupernodal_ldltf6496^2 grid, nnz 4569622.580 ms0.500.031%ok
directsparse_luf6496^2 grid, nnz 4569667.183 ms0.330.010%ok
directsupernodal_luf6496^2 grid, nnz 4569691.196 ms0.250.010%ok
directnative_kluf6496^2 grid, nnz 4569630.980 ms0.720.021%ok
directsparse_qrf6448^2 grid, nnz 1132895.396 ms0.020.000%ok
orderingamdf6496^2 grid, nnz 45696, scrambled4.575 ms0.000.00n/aok
orderingcolamdf6496^2 grid, nnz 45696, columns scrambled6.975 ms0.000.00n/aok
orderingrcmf6496^2 grid, nnz 45696, scrambled0.502 ms0.000.00n/aok
orderingblock_triangular_formf6432 planted blocks, n 819217.880 ms0.000.00n/aok
krylovcg+ic0f64128^2 grid, n 1638449.762 ms0.000.00n/aok
krylovminres+diagf64128^2 grid, n 1638454.686 ms0.000.00n/aok
krylovbicg+ilu0f64128^2 grid, n 1638438.497 ms0.000.00n/aok
krylovbicgstab+ilu0f64128^2 grid, n 1638421.068 ms0.000.00n/aok
krylovbicgstab_ell+ilu0f64128^2 grid, n 1638427.404 ms0.000.00n/aok
krylovcgs+ilu0f64128^2 grid, n 1638422.309 ms0.000.00n/aok
krylovgmres30+ilu0f64128^2 grid, n 1638464.560 ms0.000.00n/aok
krylovqmr+ilu0f64128^2 grid, n 1638442.778 ms0.000.00n/aok
krylovtfqmr+ilu0f64128^2 grid, n 1638425.174 ms0.000.00n/aok
krylovidr_s1+ilu0f64128^2 grid, n 1638429.857 ms0.000.00n/aok
krylovidr_s2+ilu0f64128^2 grid, n 1638428.791 ms0.000.00n/aok
krylovidr_s4+ilu0f64128^2 grid, n 1638432.173 ms0.000.00n/aok
krylovidr_s8+ilu0f64128^2 grid, n 1638439.159 ms0.000.00n/aok
pcdiagonalf64128^2 grid, rows scaled0.046 ms0.000.00n/aok
pcilu_0f64128^2 grid2.935 ms0.000.00n/aok
pcilutf64128^2 grid28.281 ms0.000.00n/aok
pcic_0f64128^2 grid2.260 ms0.000.00n/aok
pcildlf64128^2 grid2.162 ms0.000.00n/aok
pcssorf64128^2 grid0.028 ms0.000.00n/aok
pcblock_diagonal4f64128^2 grid, rows scaled0.807 ms0.000.00n/aok
smootherjacobif64512^2 grid, nnz 13086721.776 ms1.4716.5132%ok
smoothergauss_seidelf64512^2 grid, nnz 13086721.829 ms1.4316.0331%ok
smoothersor1.5f64512^2 grid, nnz 13086722.221 ms1.1813.2126%ok
mgpoisson2d_vcyclef64255^2 grid, 6 levels, 10 V(1,1)19.376 ms1.9815.0029%ok
sparse_eiglanczosf6464x65 Laplacian, k=4, m=8040.561 ms0.000.00n/aok
sparse_eigarnoldif6464x65 convection-diffusion, k=4, m=8058.794 ms0.000.00n/aok
sparse_eigshift_invertf6464x65 Laplacian, k=4 nearest 0, m=4066.987 ms0.000.00n/aok
sparse_eigpower_iterationf6464x65 Laplacian, 300 steps9.700 ms0.000.00n/aok
mixeddotf32>f64n=838860811.733 ms1.435.7211%ok
mixedgemmf32>f64512x512x51236.662 ms7.320.119%ok
mixedblock_scaled_doti8+f32n=8388608, block 321.725 ms9.7310.3320%ok
integerdoti32n=8388608, |v| <= 154.276 ms3.9215.69n/aok
integerdoti16>i32n=8388608, |v| <= 153.213 ms5.2210.44n/aok
integerdoti8>i32n=8388608, |v| <= 151.727 ms9.719.71n/aok
integerdotu8i8>i32n=8388608, |v| <= 151.727 ms9.719.71n/aok
integergemmi32512^3, |v| <= 20479.867 ms27.210.32n/aok
integergemmi16>i32512^3, |v| <= 204738.326 ms7.000.05n/aok
integergemmi8>i32512^3, |v| <= 12738.661 ms6.940.04n/aok
integergemm_quadi8>i32512^3, |v| <= 12738.664 ms6.940.04n/aok
integergemm_quadu8i8>i32512^3, |v| <= 12734.464 ms7.790.05n/aok

* working set fits in the last-level cache: a cache rate, shown as a % of the DRAM roofline.

ryzen-9-8945hs — native, 8 threads, class medium

Section titled “ryzen-9-8945hs — native, 8 threads, class medium”
CPUAMD Ryzen 9 8945HS w/ Radeon 780M Graphics (x86_64, 16 logical cores)
OSWindows 10.0.26300 (26H2)
compilerMSVC 19.52.37029
build/DWIN32 /D_WINDOWS /EHsc /O2 /Ob2 /DNDEBUG, SIMD scalar
commitcd6762a6e989
FMA peak (ceiling)f64 544.4 GFLOP/s, f32 1011.9 GFLOP/s
… hardware, AVX-512Ff64 544.4, f32 1011.9 GFLOP/s
… this build, scalarf64 14.5, f32 14.4 GFLOP/s
bandwidthread 52.3, update 53.2, triad 40.9 GB/s (ceiling 53.2)
last-level cache16 MiB
areaentriesmedian %roofbest %rooffailed checksmedian parallel efficiency
blas11084%119%053%
blas21061%107%013%
blas3102%21%013%
factor121%6%018%
eigen80%2%013%
elementwise55%10%013%
sparse532%185%017%
direct70%0%012%
ordering4n/an/a012%
krylov13n/an/a011%
pc7n/an/a012%
smoother328%31%012%
mg129%29%013%
sparse_eig4n/an/a013%
mixed311%19%013%
integer9n/an/a083%
Every entry of this run
areaoptypeshapemedianGFLOP/sGB/s%roofcheckspeedup (efficiency)
blas1dotf64n=83886082.315 ms7.2557.99109%ok5.3x (66%)
blas1dotf32n=83886081.565 ms10.7242.8781%ok7.8x (98%)
blas1axpyf64n=83886083.541 ms4.7456.86107%ok4.2x (53%)
blas1axpyf32n=83886081.955 ms8.5851.5097%ok7.6x (95%)
blas1scalef64n=83886082.115 ms3.9763.47119%ok2.0x (25%)
blas1two_normf64n=83886081.501 ms11.1844.7284%ok7.8x (98%)
blas1one_normf64n=83886086.389 ms1.3110.5020%ok1.0x (13%)
blas1infinity_normf64n=83886086.406 ms1.3110.4820%ok1.0x (12%)
blas1sumf64n=83886086.393 ms1.3110.5020%ok1.0x (12%)
blas1dot_k3f64n=8388608109.711 ms1.681.222%ok1.0x (12%)
blas2gemvf64n=40962.385 ms14.0756.31106%ok1.4x (18%)
blas2gemvf32n=40961.181 ms28.4056.83107%ok2.5x (31%)
blas2gemv_colmajorf64n=40962.362 ms14.2156.85107%ok3.3x (41%)
blas2gemv_colmajorf32n=40961.182 ms28.3956.80107%ok2.1x (27%)
blas2symvf64n=409610.366 ms3.246.4812%ok1.0x (13%)
blas2trmvf64n=40962.114 ms7.9431.7860%ok1.0x (12%)
blas2trmvf32n=40961.823 ms9.2018.4235%ok1.0x (13%)
blas2trsvf64n=40962.076 ms8.0832.3561%ok1.0x (12%)
blas2trsvf32n=40961.869 ms8.9817.9734%ok1.0x (12%)
blas2gerf64n=409616.210 ms2.0716.5631%ok1.0x (13%)
blas3gemmf64512x512x5122.533 ms105.992.4819%ok7.9x (99%)
blas3gemmf32512x512x5121.266 ms212.002.4821%ok7.9x (98%)
blas3gemmtf64512x512x51213.912 ms9.650.452%ok1.0x (12%)
blas3symmf64512x512x51219.756 ms13.590.322%ok1.1x (13%)
blas3syrkf64512x512x51216.886 ms7.950.251%ok1.0x (12%)
blas3syr2kf64512x512x51228.757 ms9.330.222%ok1.0x (13%)
blas3trmmf64512x512x51231.264 ms4.290.171%ok1.0x (12%)
blas3trsmf64512x512x51237.072 ms3.620.141%ok1.0x (12%)
blas3trsmf32512x512x51233.483 ms4.010.080%ok1.0x (12%)
blas3kronf6448x48 (x) 48x4810.884 ms0.493.907%ok1.0x (13%)
factorluf64512x51227.621 ms3.240.151%ok1.5x (18%)
factorluf32512x51230.019 ms2.980.070%ok1.3x (16%)
factorcholeskyf64512x51214.367 ms3.110.291%ok1.0x (12%)
factorcholeskyf32512x51213.720 ms3.260.150%ok1.0x (12%)
factorldltf64512x51214.416 ms3.100.291%ok1.0x (12%)
factorldlt_bkf64512x512205.959 ms0.220.020%ok1.0x (12%)
factorqrf64512x512500.573 ms0.360.010%ok1.6x (20%)
factorlqf64512x51254.157 ms3.300.081%ok1.5x (19%)
factorhessenbergf64512x512576.677 ms0.780.010%ok1.6x (20%)
factorhouseholderf64512x5121.267 ms0.833.316%*ok3.9x (49%)
factorinvf64512x51298.837 ms2.720.040%ok1.1x (14%)
factorlu_iterative_refinef32>f64512x51229.059 ms3.080.141%ok1.4x (18%)
eigeneigenvalue_symmetricf64256x25649.481 ms0.450.020%ok1.0x (13%)
eigeneigen_symmetricf64256x256163.409 ms0.920.010%ok1.0x (13%)
eigeneigen_symmetric_jacobif64256x256641.051 ms0.240.000%ok1.0x (12%)
eigeneigenvaluef64256x25675.625 ms2.220.010%ok1.0x (12%)
eigenschurf64256x256156.657 ms2.680.010%ok0.9x (12%)
eigensvdf64256x256176.201 ms2.090.010%ok0.8x (11%)
eigenexpmf64256x25623.813 ms10.570.042%ok1.6x (20%)
eigensylvesterf64256x256324.020 ms3.110.001%ok1.0x (13%)
elementwiseexpf64n=838860850.395 ms0.172.665%ok0.8x (10%)
elementwiselogf64n=838860861.334 ms0.142.194%ok0.9x (11%)
elementwisesinf64n=838860848.002 ms0.172.805%ok1.0x (13%)
elementwisesqrtf64n=838860826.497 ms0.325.0710%ok1.0x (13%)
elementwiseerff64n=838860838.790 ms0.223.467%ok1.0x (13%)
sparsespmvf64512^2 grid, nnz 13086720.277 ms9.4598.27185% !ok3.4x (42%)
sparsespmv_transf64512^2 grid, nnz 13086721.618 ms1.6216.8332%ok1.3x (17%)
sparsecrs_densef64512^2 grid, nnz 1308672 x 162.355 ms17.7837.3870%ok3.4x (42%)
sparsespgemmf64256^2 grid, nnz 32665630.168 ms0.110.381%*ok0.7x (9%)
sparselower_solve_levelf64192^2 grid, nnz(L) 10035641.946 ms1.038.5516%*ok1.0x (13%)
directsparse_choleskyf6496^2 grid, nnz 45696135.235 ms0.080.010%ok1.0x (13%)
directsparse_ldltf6496^2 grid, nnz 4569640.139 ms0.280.020%ok1.0x (12%)
directsupernodal_ldltf6496^2 grid, nnz 4569623.098 ms0.480.030%ok1.0x (12%)
directsparse_luf6496^2 grid, nnz 4569667.811 ms0.330.010%ok1.0x (12%)
directsupernodal_luf6496^2 grid, nnz 4569692.133 ms0.240.010%ok1.0x (12%)
directnative_kluf6496^2 grid, nnz 4569633.135 ms0.680.020%ok0.9x (12%)
directsparse_qrf6448^2 grid, nnz 1132896.896 ms0.020.000%ok1.0x (12%)
orderingamdf6496^2 grid, nnz 45696, scrambled4.619 ms0.000.00n/aok1.0x (12%)
orderingcolamdf6496^2 grid, nnz 45696, columns scrambled6.994 ms0.000.00n/aok1.0x (12%)
orderingrcmf6496^2 grid, nnz 45696, scrambled0.500 ms0.000.00n/aok1.0x (13%)
orderingblock_triangular_formf6432 planted blocks, n 819217.897 ms0.000.00n/aok1.0x (12%)
krylovcg+ic0f64128^2 grid, n 1638449.792 ms0.000.00n/aok1.0x (12%)
krylovminres+diagf64128^2 grid, n 1638455.816 ms0.000.00n/aok1.0x (12%)
krylovbicg+ilu0f64128^2 grid, n 1638438.615 ms0.000.00n/aok1.0x (12%)
krylovbicgstab+ilu0f64128^2 grid, n 1638423.293 ms0.000.00n/aok0.9x (11%)
krylovbicgstab_ell+ilu0f64128^2 grid, n 1638431.942 ms0.000.00n/aok0.9x (11%)
krylovcgs+ilu0f64128^2 grid, n 1638426.083 ms0.000.00n/aok0.9x (11%)
krylovgmres30+ilu0f64128^2 grid, n 1638475.592 ms0.000.00n/aok0.9x (11%)
krylovqmr+ilu0f64128^2 grid, n 1638449.956 ms0.000.00n/aok0.9x (11%)
krylovtfqmr+ilu0f64128^2 grid, n 1638429.295 ms0.000.00n/aok0.9x (11%)
krylovidr_s1+ilu0f64128^2 grid, n 1638434.724 ms0.000.00n/aok0.9x (11%)
krylovidr_s2+ilu0f64128^2 grid, n 1638433.654 ms0.000.00n/aok0.9x (11%)
krylovidr_s4+ilu0f64128^2 grid, n 1638432.076 ms0.000.00n/aok1.0x (13%)
krylovidr_s8+ilu0f64128^2 grid, n 1638439.219 ms0.000.00n/aok1.0x (12%)
pcdiagonalf64128^2 grid, rows scaled0.045 ms0.000.00n/aok1.0x (13%)
pcilu_0f64128^2 grid2.980 ms0.000.00n/aok1.0x (12%)
pcilutf64128^2 grid28.990 ms0.000.00n/aok1.0x (12%)
pcic_0f64128^2 grid2.243 ms0.000.00n/aok1.0x (13%)
pcildlf64128^2 grid2.194 ms0.000.00n/aok1.0x (12%)
pcssorf64128^2 grid0.026 ms0.000.00n/aok1.0x (13%)
pcblock_diagonal4f64128^2 grid, rows scaled0.811 ms0.000.00n/aok1.0x (12%)
smootherjacobif64512^2 grid, nnz 13086721.793 ms1.4616.3631%ok1.0x (12%)
smoothergauss_seidelf64512^2 grid, nnz 13086721.942 ms1.3515.1028%ok0.9x (12%)
smoothersor1.5f64512^2 grid, nnz 13086722.270 ms1.1512.9224%ok1.0x (12%)
mgpoisson2d_vcyclef64255^2 grid, 6 levels, 10 V(1,1)18.979 ms2.0315.3129%ok1.0x (13%)
sparse_eiglanczosf6464x65 Laplacian, k=4, m=8040.609 ms0.000.00n/aok1.0x (12%)
sparse_eigarnoldif6464x65 convection-diffusion, k=4, m=8058.685 ms0.000.00n/aok1.0x (13%)
sparse_eigshift_invertf6464x65 Laplacian, k=4 nearest 0, m=4066.880 ms0.000.00n/aok1.0x (13%)
sparse_eigpower_iterationf6464x65 Laplacian, 300 steps9.719 ms0.000.00n/aok1.0x (12%)
mixeddotf32>f64n=838860811.742 ms1.435.7211%ok1.0x (12%)
mixedgemmf32>f64512x512x5124.817 ms55.720.8710%ok7.6x (95%)
mixedblock_scaled_doti8+f32n=8388608, block 321.721 ms9.7510.3619%ok1.0x (13%)
integerdoti32n=8388608, |v| <= 155.016 ms3.3413.38n/aok0.9x (11%)
integerdoti16>i32n=8388608, |v| <= 153.219 ms5.2110.42n/aok1.0x (12%)
integerdoti8>i32n=8388608, |v| <= 151.727 ms9.719.71n/aok1.0x (12%)
integerdotu8i8>i32n=8388608, |v| <= 151.728 ms9.719.71n/aok1.0x (12%)
integergemmi32512^3, |v| <= 20471.278 ms210.032.46n/aok7.7x (97%)
integergemmi16>i32512^3, |v| <= 20475.022 ms53.460.42n/aok7.6x (95%)
integergemmi8>i32512^3, |v| <= 1275.043 ms53.230.31n/aok7.7x (96%)
integergemm_quadi8>i32512^3, |v| <= 1275.053 ms53.120.31n/aok7.7x (96%)
integergemm_quadu8i8>i32512^3, |v| <= 1275.173 ms51.890.30n/aok6.7x (83%)

* working set fits in the last-level cache: a cache rate, shown as a % of the DRAM roofline.

! more than 150% of the DRAM roofline for a bandwidth-bound result: the entry’s byte model is probably wrong.

ryzen-9-8945hs — native-fast, 1 thread, class medium

Section titled “ryzen-9-8945hs — native-fast, 1 thread, class medium”
CPUAMD Ryzen 9 8945HS w/ Radeon 780M Graphics (x86_64, 16 logical cores)
OSWindows 10.0.26300 (26H2)
compilerMSVC 19.52.37029
build/DWIN32 /D_WINDOWS /EHsc /O2 /Ob2 /DNDEBUG, SIMD SSE2
commitcd6762a6e989
FMA peak (ceiling)f64 63.2 GFLOP/s, f32 126.3 GFLOP/s
… hardware, AVX-512Ff64 63.2, f32 126.3 GFLOP/s
… this build, SSE2f64 40.7, f32 81.6 GFLOP/s
bandwidthread 45.7, update 52.3, triad 37.9 GB/s (ceiling 52.3)
last-level cache16 MiB
areaentriesmedian %roofbest %rooffailed checks
blas110103%128%0
blas21054%73%0
blas31014%20%0
factor123%5%0
eigen84%12%0
elementwise57%10%0
sparse521%57%0
direct70%1%0
ordering4n/an/a0
krylov13n/an/a0
pc7n/an/a0
smoother330%32%0
mg129%29%0
sparse_eig4n/an/a0
mixed359%59%0
integer9n/an/a0
Every entry of this run
areaoptypeshapemedianGFLOP/sGB/s%roofcheck
blas1dotf64n=83886082.500 ms6.7153.68103%ok
blas1dotf32n=83886081.255 ms13.3753.48102%ok
blas1axpyf64n=83886083.595 ms4.6756.00107%ok
blas1axpyf32n=83886081.749 ms9.5957.54110%ok
blas1scalef64n=83886082.000 ms4.1967.12128%ok
blas1two_normf64n=83886081.181 ms14.2056.82109%ok
blas1one_normf64n=83886086.395 ms1.3110.4920%ok
blas1infinity_normf64n=83886086.400 ms1.3110.4920%ok
blas1sumf64n=83886086.390 ms1.3110.5020%ok
blas1dot_k3f64n=8388608105.029 ms1.761.283%ok
blas2gemvf64n=40963.498 ms9.5938.3873%ok
blas2gemvf32n=40961.840 ms18.2436.4970%ok
blas2gemv_colmajorf64n=409622.253 ms1.516.0312%ok
blas2gemv_colmajorf32n=40968.261 ms4.068.1316%ok
blas2symvf64n=409610.434 ms3.226.4412%ok
blas2trmvf64n=40961.956 ms8.5834.3466%ok
blas2trmvf32n=40961.243 ms13.5027.0252%ok
blas2trsvf64n=40962.337 ms7.1828.7555%ok
blas2trsvf32n=40961.194 ms14.0528.1254%ok
blas2gerf64n=409620.381 ms1.6513.1725%ok
blas3gemmf64512x512x51221.614 ms12.420.2920%ok
blas3gemmf32512x512x51210.825 ms24.800.2920%ok
blas3gemmtf64512x512x51213.861 ms9.680.4515%ok
blas3symmf64512x512x51222.713 ms11.820.2819%ok
blas3syrkf64512x512x51221.422 ms6.270.2010%ok
blas3syr2kf64512x512x51234.823 ms7.710.1812%ok
blas3trmmf64512x512x51217.339 ms7.740.3012%ok
blas3trsmf64512x512x51214.950 ms8.980.3514%ok
blas3trsmf32512x512x51211.683 ms11.490.229%ok
blas3kronf6448x48 (x) 48x4814.767 ms0.362.885%ok
factorluf64512x51262.529 ms1.430.072%ok
factorluf32512x51260.922 ms1.470.031%ok
factorcholeskyf64512x51214.793 ms3.020.285%ok
factorcholeskyf32512x51213.714 ms3.260.153%ok
factorldltf64512x51214.410 ms3.100.295%ok
factorldlt_bkf64512x512205.022 ms0.220.020%ok
factorqrf64512x512815.861 ms0.220.010%ok
factorlqf64512x51281.457 ms2.200.053%ok
factorhessenbergf64512x512932.770 ms0.480.001%ok
factorhouseholderf64512x5124.911 ms0.210.852%*ok
factorinvf64512x512111.077 ms2.420.044%ok
factorlu_iterative_refinef32>f64512x51241.410 ms2.160.103%ok
eigeneigenvalue_symmetricf64256x25648.094 ms0.470.021%ok
eigeneigen_symmetricf64256x256166.196 ms0.910.011%ok
eigeneigen_symmetric_jacobif64256x256649.868 ms0.230.000%ok
eigeneigenvaluef64256x25675.800 ms2.210.013%ok
eigenschurf64256x256141.079 ms2.970.015%ok
eigensvdf64256x256149.845 ms2.460.014%ok
eigenexpmf64256x25632.603 ms7.720.0312%ok
eigensylvesterf64256x256332.324 ms3.030.005%ok
elementwiseexpf64n=838860833.426 ms0.254.028%ok
elementwiselogf64n=838860851.729 ms0.162.595%ok
elementwisesinf64n=838860848.967 ms0.172.745%ok
elementwisesqrtf64n=838860826.439 ms0.325.0810%ok
elementwiseerff64n=838860838.976 ms0.223.447%ok
sparsespmvf64512^2 grid, nnz 13086720.919 ms2.8529.6257%ok
sparsespmv_transf64512^2 grid, nnz 13086721.456 ms1.8018.7036%ok
sparsecrs_densef64512^2 grid, nnz 1308672 x 167.954 ms5.2611.0721%ok
sparsespgemmf64256^2 grid, nnz 32665621.535 ms0.150.531%*ok
sparselower_solve_levelf64192^2 grid, nnz(L) 10035641.932 ms1.048.6216%*ok
directsparse_choleskyf6496^2 grid, nnz 45696127.373 ms0.090.010%ok
directsparse_ldltf6496^2 grid, nnz 4569637.139 ms0.300.020%ok
directsupernodal_ldltf6496^2 grid, nnz 4569620.192 ms0.550.041%ok
directsparse_luf6496^2 grid, nnz 4569678.922 ms0.280.010%ok
directsupernodal_luf6496^2 grid, nnz 4569691.607 ms0.240.010%ok
directnative_kluf6496^2 grid, nnz 4569632.577 ms0.690.021%ok
directsparse_qrf6448^2 grid, nnz 11328104.595 ms0.020.000%ok
orderingamdf6496^2 grid, nnz 45696, scrambled4.589 ms0.000.00n/aok
orderingcolamdf6496^2 grid, nnz 45696, columns scrambled6.967 ms0.000.00n/aok
orderingrcmf6496^2 grid, nnz 45696, scrambled0.500 ms0.000.00n/aok
orderingblock_triangular_formf6432 planted blocks, n 819217.939 ms0.000.00n/aok
krylovcg+ic0f64128^2 grid, n 1638441.333 ms0.000.00n/aok
krylovminres+diagf64128^2 grid, n 1638439.078 ms0.000.00n/aok
krylovbicg+ilu0f64128^2 grid, n 1638434.910 ms0.000.00n/aok
krylovbicgstab+ilu0f64128^2 grid, n 1638417.053 ms0.000.00n/aok
krylovbicgstab_ell+ilu0f64128^2 grid, n 1638422.667 ms0.000.00n/aok
krylovcgs+ilu0f64128^2 grid, n 1638419.877 ms0.000.00n/aok
krylovgmres30+ilu0f64128^2 grid, n 1638439.245 ms0.000.00n/aok
krylovqmr+ilu0f64128^2 grid, n 1638436.837 ms0.000.00n/aok
krylovtfqmr+ilu0f64128^2 grid, n 1638421.978 ms0.000.00n/aok
krylovidr_s1+ilu0f64128^2 grid, n 1638424.450 ms0.000.00n/aok
krylovidr_s2+ilu0f64128^2 grid, n 1638423.283 ms0.000.00n/aok
krylovidr_s4+ilu0f64128^2 grid, n 1638424.774 ms0.000.00n/aok
krylovidr_s8+ilu0f64128^2 grid, n 1638427.857 ms0.000.00n/aok
pcdiagonalf64128^2 grid, rows scaled0.048 ms0.000.00n/aok
pcilu_0f64128^2 grid2.923 ms0.000.00n/aok
pcilutf64128^2 grid28.356 ms0.000.00n/aok
pcic_0f64128^2 grid2.284 ms0.000.00n/aok
pcildlf64128^2 grid2.183 ms0.000.00n/aok
pcssorf64128^2 grid0.027 ms0.000.00n/aok
pcblock_diagonal4f64128^2 grid, rows scaled0.804 ms0.000.00n/aok
smootherjacobif64512^2 grid, nnz 13086721.769 ms1.4816.5832%ok
smoothergauss_seidelf64512^2 grid, nnz 13086721.846 ms1.4215.8930%ok
smoothersor1.5f64512^2 grid, nnz 13086722.915 ms0.9010.0619%ok
mgpoisson2d_vcyclef64255^2 grid, 6 levels, 10 V(1,1)19.383 ms1.9815.0029%ok
sparse_eiglanczosf6464x65 Laplacian, k=4, m=8040.571 ms0.000.00n/aok
sparse_eigarnoldif6464x65 convection-diffusion, k=4, m=8058.743 ms0.000.00n/aok
sparse_eigshift_invertf6464x65 Laplacian, k=4 nearest 0, m=4072.118 ms0.000.00n/aok
sparse_eigpower_iterationf6464x65 Laplacian, 300 steps8.099 ms0.000.00n/aok
mixeddotf32>f64n=83886082.174 ms7.7230.8759%ok
mixedgemmf32>f64512x512x51219.251 ms13.940.2222%ok
mixedblock_scaled_doti8+f32n=8388608, block 320.577 ms29.0630.8859%ok
integerdoti32n=8388608, |v| <= 151.445 ms11.6146.46n/aok
integerdoti16>i32n=8388608, |v| <= 150.497 ms33.7367.46n/aok
integerdoti8>i32n=8388608, |v| <= 150.430 ms38.9938.99n/aok
integerdotu8i8>i32n=8388608, |v| <= 150.386 ms43.5043.50n/aok
integergemmi32512^3, |v| <= 204718.877 ms14.220.17n/aok
integergemmi16>i32512^3, |v| <= 204721.941 ms12.230.10n/aok
integergemmi8>i32512^3, |v| <= 12722.451 ms11.960.07n/aok
integergemm_quadi8>i32512^3, |v| <= 1274.603 ms58.320.34n/aok
integergemm_quadu8i8>i32512^3, |v| <= 1274.559 ms58.880.35n/aok

* working set fits in the last-level cache: a cache rate, shown as a % of the DRAM roofline.

ryzen-9-8945hs — native-fast, 8 threads, class medium

Section titled “ryzen-9-8945hs — native-fast, 8 threads, class medium”
CPUAMD Ryzen 9 8945HS w/ Radeon 780M Graphics (x86_64, 16 logical cores)
OSWindows 10.0.26300 (26H2)
compilerMSVC 19.52.37029
build/DWIN32 /D_WINDOWS /EHsc /O2 /Ob2 /DNDEBUG, SIMD SSE2
commitcd6762a6e989
FMA peak (ceiling)f64 505.9 GFLOP/s, f32 1011.7 GFLOP/s
… hardware, AVX-512Ff64 505.9, f32 1011.7 GFLOP/s
… this build, SSE2f64 317.4, f32 633.8 GFLOP/s
bandwidthread 52.7, update 53.0, triad 40.9 GB/s (ceiling 53.0)
last-level cache16 MiB
areaentriesmedian %roofbest %rooffailed checksmedian parallel efficiency
blas110110%125%013%
blas21062%118%016%
blas3105%19%031%
factor121%6%019%
eigen80%2%013%
elementwise56%10%012%
sparse536%241%013%
direct70%0%013%
ordering4n/an/a012%
krylov13n/an/a013%
pc7n/an/a012%
smoother329%30%012%
mg129%29%013%
sparse_eig4n/an/a013%
mixed358%58%013%
integer9n/an/a073%
Every entry of this run
areaoptypeshapemedianGFLOP/sGB/s%roofcheckspeedup (efficiency)
blas1dotf64n=83886082.292 ms7.3258.56110%ok1.1x (14%)
blas1dotf32n=83886081.009 ms16.6366.51125%ok1.2x (16%)
blas1axpyf64n=83886083.519 ms4.7757.21108%ok1.0x (13%)
blas1axpyf32n=83886081.671 ms10.0460.25114%ok1.0x (13%)
blas1scalef64n=83886082.111 ms3.9763.60120%ok0.9x (12%)
blas1two_normf64n=83886081.061 ms15.8163.24119%ok1.1x (14%)
blas1one_normf64n=83886086.417 ms1.3110.4620%ok1.0x (12%)
blas1infinity_normf64n=83886086.401 ms1.3110.4820%ok1.0x (12%)
blas1sumf64n=83886086.394 ms1.3110.5020%ok1.0x (12%)
blas1dot_k3f64n=8388608105.179 ms1.751.282%ok1.0x (12%)
blas2gemvf64n=40962.335 ms14.3757.50108%ok1.5x (19%)
blas2gemvf32n=40961.071 ms31.3362.70118%ok1.7x (21%)
blas2gemv_colmajorf64n=409610.106 ms3.3213.2925%ok2.2x (28%)
blas2gemv_colmajorf32n=40965.084 ms6.6013.2125%ok1.6x (20%)
blas2symvf64n=409610.181 ms3.306.6012%ok1.0x (13%)
blas2trmvf64n=40961.977 ms8.4833.9764%ok1.0x (12%)
blas2trmvf32n=40961.167 ms14.3728.7854%ok1.1x (13%)
blas2trsvf64n=40961.960 ms8.5634.2665%ok1.2x (15%)
blas2trsvf32n=40961.024 ms16.3932.8162%ok1.2x (15%)
blas2gerf64n=409616.364 ms2.0516.4131%ok1.2x (16%)
blas3gemmf64512x512x5122.732 ms98.272.3019%ok7.9x (99%)
blas3gemmf32512x512x5121.553 ms172.862.0317%ok7.0x (87%)
blas3gemmtf64512x512x5125.625 ms23.861.125%ok2.5x (31%)
blas3symmf64512x512x5123.886 ms69.071.6214%ok5.8x (73%)
blas3syrkf64512x512x51217.038 ms7.880.252%ok1.3x (16%)
blas3syr2kf64512x512x51230.512 ms8.800.212%ok1.1x (14%)
blas3trmmf64512x512x5126.934 ms19.360.764%ok2.5x (31%)
blas3trsmf64512x512x5127.357 ms18.240.714%ok2.0x (25%)
blas3trsmf32512x512x5125.496 ms24.420.482%ok2.1x (27%)
blas3kronf6448x48 (x) 48x4811.229 ms0.473.787%ok1.3x (16%)
factorluf64512x51227.520 ms3.250.151%ok2.3x (28%)
factorluf32512x51226.059 ms3.430.080%ok2.3x (29%)
factorcholeskyf64512x51214.292 ms3.130.291%ok1.0x (13%)
factorcholeskyf32512x51213.720 ms3.260.150%ok1.0x (12%)
factorldltf64512x51214.397 ms3.110.291%ok1.0x (13%)
factorldlt_bkf64512x512211.274 ms0.210.020%ok1.0x (12%)
factorqrf64512x512498.148 ms0.360.010%ok1.6x (20%)
factorlqf64512x51252.735 ms3.390.081%ok1.5x (19%)
factorhessenbergf64512x512576.798 ms0.780.010%ok1.6x (20%)
factorhouseholderf64512x5121.262 ms0.833.326%*ok3.9x (49%)
factorinvf64512x51298.454 ms2.730.041%ok1.1x (14%)
factorlu_iterative_refinef32>f64512x51229.304 ms3.050.141%ok1.4x (18%)
eigeneigenvalue_symmetricf64256x25647.850 ms0.470.020%ok1.0x (13%)
eigeneigen_symmetricf64256x256159.964 ms0.940.010%ok1.0x (13%)
eigeneigen_symmetric_jacobif64256x256634.102 ms0.240.000%ok1.0x (13%)
eigeneigenvaluef64256x25674.892 ms2.240.010%ok1.0x (13%)
eigenschurf64256x256143.368 ms2.930.011%ok1.0x (12%)
eigensvdf64256x256149.733 ms2.470.010%ok1.0x (13%)
eigenexpmf64256x25623.283 ms10.810.052%ok1.4x (18%)
eigensylvesterf64256x256326.885 ms3.080.001%ok1.0x (13%)
elementwiseexpf64n=838860833.551 ms0.254.008%ok1.0x (12%)
elementwiselogf64n=838860851.693 ms0.162.605%ok1.0x (13%)
elementwisesinf64n=838860848.948 ms0.172.745%ok1.0x (13%)
elementwisesqrtf64n=838860826.453 ms0.325.0710%ok1.0x (12%)
elementwiseerff64n=838860839.016 ms0.223.446%ok1.0x (12%)
sparsespmvf64512^2 grid, nnz 13086720.213 ms12.29127.84241% !ok4.3x (54%)
sparsespmv_transf64512^2 grid, nnz 13086721.441 ms1.8218.9036%ok1.0x (13%)
sparsecrs_densef64512^2 grid, nnz 1308672 x 162.293 ms18.2738.4172%ok3.5x (43%)
sparsespgemmf64256^2 grid, nnz 32665624.368 ms0.130.471%*ok0.9x (11%)
sparselower_solve_levelf64192^2 grid, nnz(L) 10035641.950 ms1.038.5416%*ok1.0x (12%)
directsparse_choleskyf6496^2 grid, nnz 45696126.374 ms0.090.010%ok1.0x (13%)
directsparse_ldltf6496^2 grid, nnz 4569635.577 ms0.310.020%ok1.0x (13%)
directsupernodal_ldltf6496^2 grid, nnz 4569618.848 ms0.590.040%ok1.1x (13%)
directsparse_luf6496^2 grid, nnz 4569666.367 ms0.340.010%ok1.2x (15%)
directsupernodal_luf6496^2 grid, nnz 4569689.705 ms0.250.010%ok1.0x (13%)
directnative_kluf6496^2 grid, nnz 4569631.939 ms0.700.020%ok1.0x (13%)
directsparse_qrf6448^2 grid, nnz 11328104.711 ms0.020.000%ok1.0x (12%)
orderingamdf6496^2 grid, nnz 45696, scrambled4.627 ms0.000.00n/aok1.0x (12%)
orderingcolamdf6496^2 grid, nnz 45696, columns scrambled6.960 ms0.000.00n/aok1.0x (13%)
orderingrcmf6496^2 grid, nnz 45696, scrambled0.506 ms0.000.00n/aok1.0x (12%)
orderingblock_triangular_formf6432 planted blocks, n 819218.409 ms0.000.00n/aok1.0x (12%)
krylovcg+ic0f64128^2 grid, n 1638441.242 ms0.000.00n/aok1.0x (13%)
krylovminres+diagf64128^2 grid, n 1638439.075 ms0.000.00n/aok1.0x (13%)
krylovbicg+ilu0f64128^2 grid, n 1638434.667 ms0.000.00n/aok1.0x (13%)
krylovbicgstab+ilu0f64128^2 grid, n 1638417.010 ms0.000.00n/aok1.0x (13%)
krylovbicgstab_ell+ilu0f64128^2 grid, n 1638422.618 ms0.000.00n/aok1.0x (13%)
krylovcgs+ilu0f64128^2 grid, n 1638419.880 ms0.000.00n/aok1.0x (12%)
krylovgmres30+ilu0f64128^2 grid, n 1638439.030 ms0.000.00n/aok1.0x (13%)
krylovqmr+ilu0f64128^2 grid, n 1638436.698 ms0.000.00n/aok1.0x (13%)
krylovtfqmr+ilu0f64128^2 grid, n 1638421.926 ms0.000.00n/aok1.0x (13%)
krylovidr_s1+ilu0f64128^2 grid, n 1638424.490 ms0.000.00n/aok1.0x (12%)
krylovidr_s2+ilu0f64128^2 grid, n 1638423.294 ms0.000.00n/aok1.0x (12%)
krylovidr_s4+ilu0f64128^2 grid, n 1638424.740 ms0.000.00n/aok1.0x (13%)
krylovidr_s8+ilu0f64128^2 grid, n 1638428.077 ms0.000.00n/aok1.0x (12%)
pcdiagonalf64128^2 grid, rows scaled0.048 ms0.000.00n/aok1.0x (13%)
pcilu_0f64128^2 grid2.974 ms0.000.00n/aok1.0x (12%)
pcilutf64128^2 grid29.144 ms0.000.00n/aok1.0x (12%)
pcic_0f64128^2 grid2.275 ms0.000.00n/aok1.0x (13%)
pcildlf64128^2 grid2.189 ms0.000.00n/aok1.0x (12%)
pcssorf64128^2 grid0.027 ms0.000.00n/aok1.0x (13%)
pcblock_diagonal4f64128^2 grid, rows scaled0.813 ms0.000.00n/aok1.0x (12%)
smootherjacobif64512^2 grid, nnz 13086721.828 ms1.4316.0430%ok1.0x (12%)
smoothergauss_seidelf64512^2 grid, nnz 13086721.895 ms1.3815.4829%ok1.0x (12%)
smoothersor1.5f64512^2 grid, nnz 13086722.219 ms1.1813.2125%ok1.3x (16%)
mgpoisson2d_vcyclef64255^2 grid, 6 levels, 10 V(1,1)18.781 ms2.0515.4829%ok1.0x (13%)
sparse_eiglanczosf6464x65 Laplacian, k=4, m=8040.667 ms0.000.00n/aok1.0x (12%)
sparse_eigarnoldif6464x65 convection-diffusion, k=4, m=8058.633 ms0.000.00n/aok1.0x (13%)
sparse_eigshift_invertf6464x65 Laplacian, k=4 nearest 0, m=4071.690 ms0.000.00n/aok1.0x (13%)
sparse_eigpower_iterationf6464x65 Laplacian, 300 steps8.076 ms0.000.00n/aok1.0x (13%)
mixeddotf32>f64n=83886082.173 ms7.7230.8958%ok1.0x (13%)
mixedgemmf32>f64512x512x5123.154 ms85.101.3317%ok6.1x (76%)
mixedblock_scaled_doti8+f32n=8388608, block 320.576 ms29.1430.9658%ok1.0x (13%)
integerdoti32n=8388608, |v| <= 151.453 ms11.5546.19n/aok1.0x (12%)
integerdoti16>i32n=8388608, |v| <= 150.496 ms33.8467.68n/aok1.0x (13%)
integerdoti8>i32n=8388608, |v| <= 150.429 ms39.1439.14n/aok1.0x (13%)
integerdotu8i8>i32n=8388608, |v| <= 150.384 ms43.6943.69n/aok1.0x (13%)
integergemmi32512^3, |v| <= 20472.880 ms93.201.09n/aok6.6x (82%)
integergemmi16>i32512^3, |v| <= 20473.039 ms88.330.69n/aok7.2x (90%)
integergemmi8>i32512^3, |v| <= 1273.063 ms87.630.51n/aok7.3x (92%)
integergemm_quadi8>i32512^3, |v| <= 1270.787 ms340.872.00n/aok5.8x (73%)
integergemm_quadu8i8>i32512^3, |v| <= 1270.783 ms342.652.01n/aok5.8x (73%)

* working set fits in the last-level cache: a cache rate, shown as a % of the DRAM roofline.

! more than 150% of the DRAM roofline for a bandwidth-bound result: the entry’s byte model is probably wrong.

ryzen-9-8945hs — blas, 1 thread, class medium

Section titled “ryzen-9-8945hs — blas, 1 thread, class medium”
CPUAMD Ryzen 9 8945HS w/ Radeon 780M Graphics (x86_64, 16 logical cores)
OSWindows 10.0.26300 (26H2)
compilerMSVC 19.52.37029
build/DWIN32 /D_WINDOWS /EHsc /O2 /Ob2 /DNDEBUG, SIMD scalar
commitcd6762a6e989
FMA peak (ceiling)f64 63.3 GFLOP/s, f32 126.6 GFLOP/s
… hardware, AVX-512Ff64 63.3, f32 126.6 GFLOP/s
… this build, scalarf64 1.6, f32 1.9 GFLOP/s
bandwidthread 37.7, update 51.7, triad 37.5 GB/s (ceiling 51.7)
last-level cache16 MiB
areaentriesmedian %roofbest %rooffailed checks
blas11018%98%0
blas21076%83%0
blas31071%93%0
factor1213%23%0
eigen817%53%0
elementwise56%10%0
sparse522%58%0
direct70%1%0
ordering4n/an/a0
krylov13n/an/a0
pc7n/an/a0
smoother331%32%0
mg129%29%0
sparse_eig4n/an/a0
mixed311%19%0
integer9n/an/a0
Every entry of this run
areaoptypeshapemedianGFLOP/sGB/s%roofcheck
blas1dotf64n=838860814.419 ms1.169.3118%ok
blas1dotf32n=838860812.263 ms1.375.4711%ok
blas1axpyf64n=838860814.961 ms1.1213.4626%ok
blas1axpyf32n=838860812.802 ms1.317.8615%ok
blas1scalef64n=83886082.645 ms3.1750.7598%ok
blas1two_normf64n=838860815.421 ms1.094.358%ok
blas1one_normf64n=83886087.357 ms1.149.1218%ok
blas1infinity_normf64n=83886086.411 ms1.3110.4720%ok
blas1sumf64n=83886086.396 ms1.3110.4920%ok
blas1dot_k3f64n=8388608111.774 ms1.651.203%ok
blas2gemvf64n=40963.276 ms10.2440.9979%ok
blas2gemvf32n=40961.651 ms20.3240.6679%ok
blas2gemv_colmajorf64n=40963.593 ms9.3437.3872%ok
blas2gemv_colmajorf32n=40961.556 ms21.5743.1683%ok
blas2symvf64n=40962.032 ms16.5133.0664%ok
blas2trmvf64n=40961.707 ms9.8339.3676%ok
blas2trmvf32n=40960.968 ms17.3334.6967%ok
blas2trsvf64n=40961.693 ms9.9139.6877%ok
blas2trsvf32n=40960.921 ms18.2136.4570%ok
blas2gerf64n=409616.270 ms2.0616.5032%ok
blas3gemmf64512x512x5124.563 ms58.821.3893%ok
blas3gemmf32512x512x5122.374 ms113.071.3389%ok
blas3gemmtf64512x512x5123.425 ms39.181.8462%ok
blas3symmf64512x512x5124.749 ms56.531.3289%ok
blas3syrkf64512x512x5123.710 ms36.171.1357%ok
blas3syr2kf64512x512x5125.998 ms44.751.0571%ok
blas3trmmf64512x512x5122.361 ms56.852.2290%ok
blas3trsmf64512x512x5123.504 ms38.311.5060%ok
blas3trsmf32512x512x5121.637 ms82.011.6065%ok
blas3kronf6448x48 (x) 48x4811.129 ms0.483.827%ok
factorluf64512x5126.393 ms14.000.6622%ok
factorluf32512x5123.063 ms29.210.6823%ok
factorcholeskyf64512x5125.484 ms8.160.7613%ok
factorcholeskyf32512x5122.114 ms21.160.9917%ok
factorldltf64512x51214.811 ms3.020.285%ok
factorldlt_bkf64512x512211.072 ms0.210.020%ok
factorqrf64512x51212.960 ms13.810.3222%ok
factorlqf64512x51281.665 ms2.190.053%ok
factorhessenbergf64512x512935.107 ms0.480.001%ok
factorhouseholderf64512x5124.952 ms0.210.852%*ok
factorinvf64512x51276.588 ms3.500.056%ok
factorlu_iterative_refinef32>f64512x5126.977 ms12.820.6020%ok
eigeneigenvalue_symmetricf64256x2562.121 ms10.550.4917%ok
eigeneigen_symmetricf64256x256160.066 ms0.940.011%ok
eigeneigen_symmetric_jacobif64256x256650.132 ms0.230.000%ok
eigeneigenvaluef64256x25620.908 ms8.020.0513%ok
eigenschurf64256x25627.263 ms15.380.0424%ok
eigensvdf64256x25610.946 ms33.720.1053%ok
eigenexpmf64256x25620.395 ms12.340.0519%ok
eigensylvesterf64256x25691.010 ms11.060.0217%ok
elementwiseexpf64n=838860833.461 ms0.254.018%ok
elementwiselogf64n=838860853.847 ms0.162.495%ok
elementwisesinf64n=838860847.973 ms0.172.805%ok
elementwisesqrtf64n=838860826.544 ms0.325.0610%ok
elementwiseerff64n=838860842.476 ms0.203.166%ok
sparsespmvf64512^2 grid, nnz 13086720.914 ms2.8629.8058%ok
sparsespmv_transf64512^2 grid, nnz 13086721.433 ms1.8319.0137%ok
sparsecrs_densef64512^2 grid, nnz 1308672 x 167.879 ms5.3211.1822%ok
sparsespgemmf64256^2 grid, nnz 32665622.341 ms0.150.511%*ok
sparselower_solve_levelf64192^2 grid, nnz(L) 10035641.955 ms1.038.5116%*ok
directsparse_choleskyf6496^2 grid, nnz 45696129.966 ms0.090.010%ok
directsparse_ldltf6496^2 grid, nnz 4569637.994 ms0.290.020%ok
directsupernodal_ldltf6496^2 grid, nnz 4569622.885 ms0.490.031%ok
directsparse_luf6496^2 grid, nnz 4569667.095 ms0.330.011%ok
directsupernodal_luf6496^2 grid, nnz 4569691.993 ms0.240.010%ok
directnative_kluf6496^2 grid, nnz 4569631.939 ms0.700.021%ok
directsparse_qrf6448^2 grid, nnz 11328101.803 ms0.020.000%ok
orderingamdf6496^2 grid, nnz 45696, scrambled4.551 ms0.000.00n/aok
orderingcolamdf6496^2 grid, nnz 45696, columns scrambled6.957 ms0.000.00n/aok
orderingrcmf6496^2 grid, nnz 45696, scrambled0.504 ms0.000.00n/aok
orderingblock_triangular_formf6432 planted blocks, n 819217.597 ms0.000.00n/aok
krylovcg+ic0f64128^2 grid, n 1638450.950 ms0.000.00n/aok
krylovminres+diagf64128^2 grid, n 1638457.576 ms0.000.00n/aok
krylovbicg+ilu0f64128^2 grid, n 1638439.396 ms0.000.00n/aok
krylovbicgstab+ilu0f64128^2 grid, n 1638421.799 ms0.000.00n/aok
krylovbicgstab_ell+ilu0f64128^2 grid, n 1638428.238 ms0.000.00n/aok
krylovcgs+ilu0f64128^2 grid, n 1638422.672 ms0.000.00n/aok
krylovgmres30+ilu0f64128^2 grid, n 1638468.911 ms0.000.00n/aok
krylovqmr+ilu0f64128^2 grid, n 1638443.678 ms0.000.00n/aok
krylovtfqmr+ilu0f64128^2 grid, n 1638425.558 ms0.000.00n/aok
krylovidr_s1+ilu0f64128^2 grid, n 1638430.784 ms0.000.00n/aok
krylovidr_s2+ilu0f64128^2 grid, n 1638429.718 ms0.000.00n/aok
krylovidr_s4+ilu0f64128^2 grid, n 1638433.378 ms0.000.00n/aok
krylovidr_s8+ilu0f64128^2 grid, n 1638441.427 ms0.000.00n/aok
pcdiagonalf64128^2 grid, rows scaled0.046 ms0.000.00n/aok
pcilu_0f64128^2 grid2.938 ms0.000.00n/aok
pcilutf64128^2 grid27.106 ms0.000.00n/aok
pcic_0f64128^2 grid2.287 ms0.000.00n/aok
pcildlf64128^2 grid2.503 ms0.000.00n/aok
pcssorf64128^2 grid0.031 ms0.000.00n/aok
pcblock_diagonal4f64128^2 grid, rows scaled1.765 ms0.000.00n/aok
smootherjacobif64512^2 grid, nnz 13086721.756 ms1.4916.7032%ok
smoothergauss_seidelf64512^2 grid, nnz 13086721.838 ms1.4215.9631%ok
smoothersor1.5f64512^2 grid, nnz 13086722.219 ms1.1813.2226%ok
mgpoisson2d_vcyclef64255^2 grid, 6 levels, 10 V(1,1)19.322 ms1.9915.0429%ok
sparse_eiglanczosf6464x65 Laplacian, k=4, m=8040.615 ms0.000.00n/aok
sparse_eigarnoldif6464x65 convection-diffusion, k=4, m=8041.327 ms0.000.00n/aok
sparse_eigshift_invertf6464x65 Laplacian, k=4 nearest 0, m=4069.312 ms0.000.00n/aok
sparse_eigpower_iterationf6464x65 Laplacian, 300 steps9.928 ms0.000.00n/aok
mixeddotf32>f64n=838860814.628 ms1.154.599%ok
mixedgemmf32>f64512x512x51237.348 ms7.190.1111%ok
mixedblock_scaled_doti8+f32n=8388608, block 321.836 ms9.149.7119%ok
integerdoti32n=8388608, |v| <= 154.277 ms3.9215.69n/aok
integerdoti16>i32n=8388608, |v| <= 153.235 ms5.1910.37n/aok
integerdoti8>i32n=8388608, |v| <= 151.726 ms9.729.72n/aok
integerdotu8i8>i32n=8388608, |v| <= 151.726 ms9.729.72n/aok
integergemmi32512^3, |v| <= 204710.238 ms26.220.31n/aok
integergemmi16>i32512^3, |v| <= 204738.570 ms6.960.05n/aok
integergemmi8>i32512^3, |v| <= 12738.705 ms6.940.04n/aok
integergemm_quadi8>i32512^3, |v| <= 12738.695 ms6.940.04n/aok
integergemm_quadu8i8>i32512^3, |v| <= 12734.459 ms7.790.05n/aok

* working set fits in the last-level cache: a cache rate, shown as a % of the DRAM roofline.

ryzen-9-8945hs — blas, 8 threads, class medium

Section titled “ryzen-9-8945hs — blas, 8 threads, class medium”
CPUAMD Ryzen 9 8945HS w/ Radeon 780M Graphics (x86_64, 16 logical cores)
OSWindows 10.0.26300 (26H2)
compilerMSVC 19.52.37029
build/DWIN32 /D_WINDOWS /EHsc /O2 /Ob2 /DNDEBUG, SIMD scalar
commitcd6762a6e989
FMA peak (ceiling)f64 506.3 GFLOP/s, f32 1011.9 GFLOP/s
… hardware, AVX-512Ff64 506.3, f32 1011.9 GFLOP/s
… this build, scalarf64 12.3, f32 14.5 GFLOP/s
bandwidthread 52.7, update 52.9, triad 41.1 GB/s (ceiling 52.9)
last-level cache16 MiB
areaentriesmedian %roofbest %rooffailed checksmedian parallel efficiency
blas11081%119%053%
blas210100%117%017%
blas31044%78%068%
factor122%6%012%
eigen82%5%011%
elementwise56%10%013%
sparse521%254%013%
direct70%0%013%
ordering4n/an/a012%
krylov13n/an/a012%
pc7n/an/a013%
smoother330%32%013%
mg129%29%013%
sparse_eig4n/an/a013%
mixed311%18%013%
integer9n/an/a095%
Every entry of this run
areaoptypeshapemedianGFLOP/sGB/s%roofcheckspeedup (efficiency)
blas1dotf64n=83886082.351 ms7.1457.09108%ok6.1x (77%)
blas1dotf32n=83886081.562 ms10.7442.9681%ok7.9x (98%)
blas1axpyf64n=83886083.521 ms4.7657.18108%ok4.2x (53%)
blas1axpyf32n=83886081.798 ms9.3355.97106%ok7.1x (89%)
blas1scalef64n=83886082.134 ms3.9362.88119%ok1.2x (15%)
blas1two_normf64n=83886081.698 ms9.8839.5375%ok9.1x (114%)
blas1one_normf64n=83886086.392 ms1.3110.5020%ok1.2x (14%)
blas1infinity_normf64n=83886086.405 ms1.3110.4820%ok1.0x (13%)
blas1sumf64n=83886086.394 ms1.3110.5020%ok1.0x (13%)
blas1dot_k3f64n=8388608111.710 ms1.651.202%ok1.0x (13%)
blas2gemvf64n=40962.570 ms13.0652.2599%ok1.3x (16%)
blas2gemvf32n=40961.148 ms29.2358.50111%ok1.4x (18%)
blas2gemv_colmajorf64n=40962.678 ms12.5350.1595%ok1.3x (17%)
blas2gemv_colmajorf32n=40961.244 ms26.9753.96102%ok1.3x (16%)
blas2symvf64n=40961.273 ms26.3652.77100%ok1.6x (20%)
blas2trmvf64n=40961.237 ms13.5654.29103%ok1.4x (17%)
blas2trmvf32n=40960.540 ms31.0462.14117%ok1.8x (22%)
blas2trsvf64n=40961.820 ms9.2236.9170%ok0.9x (12%)
blas2trsvf32n=40960.932 ms18.0036.0368%ok1.0x (12%)
blas2gerf64n=409616.217 ms2.0716.5631%ok1.0x (13%)
blas3gemmf64512x512x5120.682 ms393.839.2378%ok6.7x (84%)
blas3gemmf32512x512x5120.379 ms709.028.3170%ok6.3x (78%)
blas3gemmtf64512x512x5121.672 ms80.253.7616%ok2.0x (26%)
blas3symmf64512x512x5120.684 ms392.629.2078%ok6.9x (87%)
blas3syrkf64512x512x5122.007 ms66.882.0913%ok1.8x (23%)
blas3syr2kf64512x512x5122.725 ms98.512.3119%ok2.2x (28%)
blas3trmmf64512x512x5120.451 ms297.4711.6259%ok5.2x (65%)
blas3trsmf64512x512x5120.613 ms218.818.5543%ok5.7x (71%)
blas3trsmf32512x512x5120.301 ms445.918.7144%ok5.4x (68%)
blas3kronf6448x48 (x) 48x4811.133 ms0.483.817%ok1.0x (12%)
factorluf64512x51211.145 ms8.030.382%ok0.6x (7%)
factorluf32512x5125.082 ms17.610.412%ok0.6x (8%)
factorcholeskyf64512x5126.993 ms6.400.601%ok0.8x (10%)
factorcholeskyf32512x5122.442 ms18.320.862%ok0.9x (11%)
factorldltf64512x51214.818 ms3.020.281%ok1.0x (12%)
factorldlt_bkf64512x512211.419 ms0.210.020%ok1.0x (12%)
factorqrf64512x51214.125 ms12.670.303%ok0.9x (11%)
factorlqf64512x51254.165 ms3.300.081%ok1.5x (19%)
factorhessenbergf64512x512573.414 ms0.780.010%ok1.6x (20%)
factorhouseholderf64512x5121.269 ms0.833.306%*ok3.9x (49%)
factorinvf64512x51282.207 ms3.270.051%ok0.9x (12%)
factorlu_iterative_refinef32>f64512x5128.818 ms10.150.482%ok0.8x (10%)
eigeneigenvalue_symmetricf64256x2564.261 ms5.250.251%ok0.5x (6%)
eigeneigen_symmetricf64256x256164.039 ms0.920.010%ok1.0x (12%)
eigeneigen_symmetric_jacobif64256x256619.276 ms0.240.000%ok1.0x (13%)
eigeneigenvaluef64256x25625.142 ms6.670.041%ok0.8x (10%)
eigenschurf64256x25631.143 ms13.470.033%ok0.9x (11%)
eigensvdf64256x25613.775 ms26.790.085%ok0.8x (10%)
eigenexpmf64256x25621.439 ms11.740.052%ok1.0x (12%)
eigensylvesterf64256x256101.372 ms9.930.022%ok0.9x (11%)
elementwiseexpf64n=838860833.422 ms0.254.028%ok1.0x (13%)
elementwiselogf64n=838860853.304 ms0.162.525%ok1.0x (13%)
elementwisesinf64n=838860847.965 ms0.172.805%ok1.0x (13%)
elementwisesqrtf64n=838860826.549 ms0.325.0610%ok1.0x (12%)
elementwiseerff64n=838860842.414 ms0.203.166%ok1.0x (13%)
sparsespmvf64512^2 grid, nnz 13086720.202 ms12.94134.60254% !ok4.5x (56%)
sparsespmv_transf64512^2 grid, nnz 13086722.480 ms1.0610.9821%ok0.6x (7%)
sparsecrs_densef64512^2 grid, nnz 1308672 x 162.351 ms17.8137.4571%ok3.4x (42%)
sparsespgemmf64256^2 grid, nnz 32665624.035 ms0.140.481%*ok0.9x (12%)
sparselower_solve_levelf64192^2 grid, nnz(L) 10035641.878 ms1.078.8617%*ok1.0x (13%)
directsparse_choleskyf6496^2 grid, nnz 45696124.491 ms0.090.010%ok1.0x (13%)
directsparse_ldltf6496^2 grid, nnz 4569638.809 ms0.290.020%ok1.0x (12%)
directsupernodal_ldltf6496^2 grid, nnz 4569622.820 ms0.490.030%ok1.0x (13%)
directsparse_luf6496^2 grid, nnz 4569666.970 ms0.330.010%ok1.0x (13%)
directsupernodal_luf6496^2 grid, nnz 4569692.117 ms0.240.010%ok1.0x (12%)
directnative_kluf6496^2 grid, nnz 4569631.735 ms0.700.020%ok1.0x (13%)
directsparse_qrf6448^2 grid, nnz 11328100.413 ms0.020.000%ok1.0x (13%)
orderingamdf6496^2 grid, nnz 45696, scrambled4.632 ms0.000.00n/aok1.0x (12%)
orderingcolamdf6496^2 grid, nnz 45696, columns scrambled7.013 ms0.000.00n/aok1.0x (12%)
orderingrcmf6496^2 grid, nnz 45696, scrambled0.509 ms0.000.00n/aok1.0x (12%)
orderingblock_triangular_formf6432 planted blocks, n 819217.462 ms0.000.00n/aok1.0x (13%)
krylovcg+ic0f64128^2 grid, n 1638451.032 ms0.000.00n/aok1.0x (12%)
krylovminres+diagf64128^2 grid, n 1638457.630 ms0.000.00n/aok1.0x (12%)
krylovbicg+ilu0f64128^2 grid, n 1638439.467 ms0.000.00n/aok1.0x (12%)
krylovbicgstab+ilu0f64128^2 grid, n 1638421.792 ms0.000.00n/aok1.0x (13%)
krylovbicgstab_ell+ilu0f64128^2 grid, n 1638428.238 ms0.000.00n/aok1.0x (12%)
krylovcgs+ilu0f64128^2 grid, n 1638422.678 ms0.000.00n/aok1.0x (12%)
krylovgmres30+ilu0f64128^2 grid, n 1638468.981 ms0.000.00n/aok1.0x (12%)
krylovqmr+ilu0f64128^2 grid, n 1638443.744 ms0.000.00n/aok1.0x (12%)
krylovtfqmr+ilu0f64128^2 grid, n 1638425.561 ms0.000.00n/aok1.0x (12%)
krylovidr_s1+ilu0f64128^2 grid, n 1638430.797 ms0.000.00n/aok1.0x (12%)
krylovidr_s2+ilu0f64128^2 grid, n 1638429.839 ms0.000.00n/aok1.0x (12%)
krylovidr_s4+ilu0f64128^2 grid, n 1638433.278 ms0.000.00n/aok1.0x (13%)
krylovidr_s8+ilu0f64128^2 grid, n 1638441.389 ms0.000.00n/aok1.0x (13%)
pcdiagonalf64128^2 grid, rows scaled0.048 ms0.000.00n/aok0.9x (12%)
pcilu_0f64128^2 grid2.994 ms0.000.00n/aok1.0x (12%)
pcilutf64128^2 grid28.485 ms0.000.00n/aok1.0x (12%)
pcic_0f64128^2 grid2.259 ms0.000.00n/aok1.0x (13%)
pcildlf64128^2 grid2.160 ms0.000.00n/aok1.2x (14%)
pcssorf64128^2 grid0.027 ms0.000.00n/aok1.2x (15%)
pcblock_diagonal4f64128^2 grid, rows scaled1.638 ms0.000.00n/aok1.1x (13%)
smootherjacobif64512^2 grid, nnz 13086721.751 ms1.4916.7532%ok1.0x (13%)
smoothergauss_seidelf64512^2 grid, nnz 13086721.833 ms1.4316.0030%ok1.0x (13%)
smoothersor1.5f64512^2 grid, nnz 13086722.219 ms1.1813.2125%ok1.0x (12%)
mgpoisson2d_vcyclef64255^2 grid, 6 levels, 10 V(1,1)18.703 ms2.0615.5429%ok1.0x (13%)
sparse_eiglanczosf6464x65 Laplacian, k=4, m=8040.472 ms0.000.00n/aok1.0x (13%)
sparse_eigarnoldif6464x65 convection-diffusion, k=4, m=8041.255 ms0.000.00n/aok1.0x (13%)
sparse_eigshift_invertf6464x65 Laplacian, k=4 nearest 0, m=4070.565 ms0.000.00n/aok1.0x (12%)
sparse_eigpower_iterationf6464x65 Laplacian, 300 steps9.892 ms0.000.00n/aok1.0x (13%)
mixeddotf32>f64n=838860814.380 ms1.174.679%ok1.0x (13%)
mixedgemmf32>f64512x512x5124.877 ms55.040.8611%ok7.7x (96%)
mixedblock_scaled_doti8+f32n=8388608, block 321.836 ms9.149.7118%ok1.0x (13%)
integerdoti32n=8388608, |v| <= 154.322 ms3.8815.53n/aok1.0x (12%)
integerdoti16>i32n=8388608, |v| <= 153.236 ms5.1910.37n/aok1.0x (12%)
integerdoti8>i32n=8388608, |v| <= 151.726 ms9.729.72n/aok1.0x (13%)
integerdotu8i8>i32n=8388608, |v| <= 151.726 ms9.729.72n/aok1.0x (13%)
integergemmi32512^3, |v| <= 20471.304 ms205.922.41n/aok7.9x (98%)
integergemmi16>i32512^3, |v| <= 20475.034 ms53.330.42n/aok7.7x (96%)
integergemmi8>i32512^3, |v| <= 1275.041 ms53.250.31n/aok7.7x (96%)
integergemm_quadi8>i32512^3, |v| <= 1275.046 ms53.200.31n/aok7.7x (96%)
integergemm_quadu8i8>i32512^3, |v| <= 1274.533 ms59.220.35n/aok7.6x (95%)

* working set fits in the last-level cache: a cache rate, shown as a % of the DRAM roofline.

! more than 150% of the DRAM roofline for a bandwidth-bound result: the entry’s byte model is probably wrong.