# rocBLAS solution indices for the prompt path's FP16-out GEMMs; valid for this architecture and rocBLAS build only
# rows: type N K ldy T solution   (type f16o: FP16 in, FP16 out, FP32 compute; Gemm::f16_inplace)
# solution: rocBLAS's solution index, or `default` where its own choice measured best at that bucket
# measured on an RX 6900 XT (80 CUs); another gfx1030 card can make its own with tools/hip/tune_rocblas
STRATA_ROCBLAS_TUNING_V1 gfx1030 5.6.0.8d1ae90e
f16o 10240 2560 10240 256 -1086327703   # 2.35815 ms -> 0.347774 ms
f16o 10240 2560 10240 512 -1086327703   # 4.93049 ms -> 0.674375 ms
f16o 10240 2560 10240 768 -1086327719   # 7.56866 ms -> 1.01344 ms
f16o 10240 2560 10240 1024 -1086327687   # 10.1879 ms -> 1.40927 ms
f16o 10240 2560 10240 1536 default   # 1.93035 ms
f16o 10240 2560 10240 2048 default   # 2.56287 ms
f16o 10240 2560 10240 4096 default   # 5.22701 ms
f16o 10240 2560 10240 8192 default   # 12.256 ms
f16o 12288 2560 12288 256 -1086327718   # 3.00699 ms -> 0.504361 ms
f16o 12288 2560 12288 512 -1086327703   # 1.01009 ms -> 0.831414 ms
f16o 12288 2560 12288 768 -1086327703   # 1.54412 ms -> 1.30476 ms
f16o 12288 2560 12288 1024 -1086327687   # 1.84607 ms -> 1.62627 ms
f16o 12288 2560 12288 1536 -1086327687   # 2.60974 ms -> 2.41411 ms
f16o 12288 2560 12288 2048 -1086327687   # 3.39763 ms -> 3.2127 ms
f16o 12288 2560 12288 4096 default   # 6.8017 ms
f16o 12288 2560 12288 8192 default   # 14.3542 ms
f16o 6144 2560 6144 256 -1086327717   # 0.980093 ms -> 0.281347 ms
f16o 6144 2560 6144 512 -1086327703   # 1.83313 ms -> 0.509667 ms
f16o 6144 2560 6144 768 default   # 0.666733 ms
f16o 6144 2560 6144 1024 default   # 0.829613 ms
f16o 6144 2560 6144 1536 default   # 1.29944 ms
f16o 6144 2560 6144 2048 -1086327687   # 2.4454 ms -> 1.63103 ms
f16o 6144 2560 6144 4096 default   # 3.20715 ms
f16o 6144 2560 6144 8192 default   # 6.33023 ms
f16o 2560 6144 2560 256 -1086327764   # 1.30173 ms -> 0.292693 ms
f16o 2560 6144 2560 512 -1086327719   # 1.22932 ms -> 0.475053 ms
f16o 2560 6144 2560 768 default   # 0.795013 ms
f16o 2560 6144 2560 1024 -1086327719   # 1.01636 ms -> 0.886773 ms
f16o 2560 6144 2560 1536 -1086327703   # 1.59868 ms -> 1.43847 ms
f16o 2560 6144 2560 2048 default   # 1.83883 ms
f16o 2560 6144 2560 4096 default   # 3.45259 ms
f16o 2560 6144 2560 8192 default   # 7.14512 ms
f16o 2560 640 2560 256 -1086327695   # 0.04352 ms -> 0.0384667 ms
f16o 2560 640 2560 512 -1086327711   # 0.0728533 ms -> 0.0594 ms
f16o 2560 640 2560 768 -1086327694   # 0.0964533 ms -> 0.07984 ms
f16o 2560 640 2560 1024 -1086327709   # 0.108387 ms -> 0.0945467 ms
f16o 2560 640 2560 1536 -1086327713   # 0.156893 ms -> 0.13828 ms
f16o 2560 640 2560 2048 default   # 0.183253 ms
f16o 2560 640 2560 4096 default   # 0.348147 ms
f16o 2560 640 2560 8192 default   # 0.685773 ms
f16o 2560 2560 2560 256 default   # 0.1292 ms
f16o 2560 2560 2560 512 -1086327719   # 0.215 ms -> 0.204587 ms
f16o 2560 2560 2560 768 default   # 0.303533 ms
f16o 2560 2560 2560 1024 -1086327719   # 0.416427 ms -> 0.356733 ms
f16o 2560 2560 2560 1536 default   # 0.52424 ms
f16o 2560 2560 2560 2048 default   # 0.687973 ms
f16o 2560 2560 2560 4096 -1086327703   # 1.55271 ms -> 1.32725 ms
f16o 2560 2560 2560 8192 -1086327703   # 2.81991 ms -> 2.61531 ms
f16o 640 2560 640 256 -1086327725   # 0.152587 ms -> 0.0562933 ms
f16o 640 2560 640 512 -1086327724   # 0.200773 ms -> 0.0829867 ms
f16o 640 2560 640 768 -1086327723   # 0.165947 ms -> 0.11384 ms
f16o 640 2560 640 1024 -1086327764   # 0.206733 ms -> 0.125493 ms
f16o 640 2560 640 1536 -1086327719   # 0.284106 ms -> 0.201267 ms
f16o 640 2560 640 2048 -1086327703   # 0.217293 ms -> 0.206173 ms
f16o 640 2560 640 4096 default   # 0.346747 ms
f16o 640 2560 640 8192 default   # 0.689187 ms
f16o 512 2560 512 256 -1086327769   # 0.156933 ms -> 0.0474 ms
f16o 512 2560 512 512 -1086327722   # 0.202507 ms -> 0.0734 ms
f16o 512 2560 512 768 -1086327774   # 0.147413 ms -> 0.0978133 ms
f16o 512 2560 512 1024 -1086327722   # 0.15524 ms -> 0.11936 ms
f16o 512 2560 512 1536 -1086327768   # 0.2222 ms -> 0.17032 ms
f16o 512 2560 512 2048 default   # 0.2122 ms
f16o 512 2560 512 4096 default   # 0.344253 ms
f16o 512 2560 512 8192 default   # 0.664307 ms
f16o 128 2560 128 256 -1086327780   # 0.0608 ms -> 0.03712 ms
f16o 128 2560 128 512 -1086327780   # 0.1992 ms -> 0.0377333 ms
f16o 128 2560 128 768 -1086327780   # 0.113667 ms -> 0.0451733 ms
f16o 128 2560 128 1024 -1086327774   # 0.118013 ms -> 0.0492533 ms
f16o 128 2560 128 1536 -1086327772   # 0.151973 ms -> 0.0664 ms
f16o 128 2560 128 2048 -1086327768   # 0.12424 ms -> 0.07528 ms
f16o 128 2560 128 4096 -1086327764   # 0.151187 ms -> 0.124813 ms
f16o 128 2560 128 8192 default   # 0.206573 ms
f16o 48 2560 48 256 -1086327840   # 0.0455867 ms -> 0.0207733 ms
f16o 48 2560 48 512 -1086327781   # 0.202253 ms -> 0.0386667 ms
f16o 48 2560 48 768 -1086327779   # 0.102947 ms -> 0.04176 ms
f16o 48 2560 48 1024 -1086327779   # 0.10784 ms -> 0.04392 ms
f16o 48 2560 48 1536 -1086327776   # 0.11252 ms -> 0.0512133 ms
f16o 48 2560 48 2048 -1086327779   # 0.121 ms -> 0.0565467 ms
f16o 48 2560 48 4096 -1086327736   # 0.186493 ms -> 0.0971733 ms
f16o 48 2560 48 8192 -1086327762   # 0.204667 ms -> 0.148867 ms
f16o 320 10240 320 256 -1086327781   # 0.42828 ms -> 0.13252 ms
f16o 320 10240 320 512 -1086327769   # 0.497053 ms -> 0.20012 ms
f16o 320 10240 320 768 -1086327722   # 0.826453 ms -> 0.241627 ms
f16o 320 10240 320 1024 -1086327773   # 1.16967 ms -> 0.36232 ms
f16o 320 10240 320 1536 -1086327738   # 1.60872 ms -> 0.466893 ms
f16o 320 10240 320 2048 -1086327724   # 2.17916 ms -> 0.633547 ms
f16o 320 10240 320 4096 -1086327717   # 9.61694 ms -> 1.18415 ms
f16o 320 10240 320 8192 default   # 2.26508 ms
f16o 10240 320 10240 256 -1086327694   # 0.0583733 ms -> 0.05556 ms
f16o 10240 320 10240 512 default   # 0.0980133 ms
f16o 10240 320 10240 768 default   # 0.1398 ms
f16o 10240 320 10240 1024 default   # 0.181493 ms
f16o 10240 320 10240 1536 default   # 0.262587 ms
f16o 10240 320 10240 2048 default   # 0.343693 ms
f16o 10240 320 10240 4096 default   # 0.669293 ms
f16o 10240 320 10240 8192 default   # 1.35307 ms
f16o 4 10240 4 256 -1086327846   # 0.105 ms -> 0.05028 ms
f16o 4 10240 4 512 -1086327840   # 0.247013 ms -> 0.0543863 ms
f16o 4 10240 4 768 -1086327840   # 0.253867 ms -> 0.061 ms
f16o 4 10240 4 1024 -1086327850   # 0.255293 ms -> 0.0801467 ms
f16o 4 10240 4 1536 -1086327840   # 0.2556 ms -> 0.1116 ms
f16o 4 10240 4 2048 -1086327815   # 0.305107 ms -> 0.142907 ms
f16o 4 10240 4 4096 -1086327922   # 0.489133 ms -> 0.253813 ms
f16o 4 10240 4 8192 -1086327922   # 1.16808 ms -> 0.48448 ms
f16o 1 2560 1 256 -9   # 0.0432533 ms -> 0.00812 ms
f16o 1 2560 1 512 -9   # 0.2 ms -> 0.0107333 ms
f16o 1 2560 1 768 -9   # 0.0864667 ms -> 0.0125333 ms
f16o 1 2560 1 1024 -9   # 0.0914133 ms -> 0.0166533 ms
f16o 1 2560 1 1536 -9   # 0.09308 ms -> 0.0197333 ms
f16o 1 2560 1 2048 -9   # 0.120027 ms -> 0.030653 ms
f16o 1 2560 1 4096 -9   # 0.0734267 ms -> 0.04152 ms
f16o 1 2560 1 8192 -9   # 0.200733 ms -> 0.07284 ms
