# 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 5.4.0.ad84daa81c
f16o 10240 2560 10240 256 -1086328105   # 2.41232 ms -> 0.331076 ms
f16o 10240 2560 10240 512 -1086328089   # 4.85504 ms -> 0.64038 ms
f16o 10240 2560 10240 768 -1086328073   # 7.40307 ms -> 0.956203 ms
f16o 10240 2560 10240 1024 -1086328073   # 9.91176 ms -> 1.28472 ms
f16o 10240 2560 10240 1536 default   # 1.87765 ms
f16o 10240 2560 10240 2048 default   # 2.45467 ms
f16o 10240 2560 10240 4096 default   # 5.14293 ms
f16o 10240 2560 10240 8192 -1086328066   # 11.2006 ms -> 10.3581 ms
f16o 12288 2560 12288 256 -1086328104   # 2.83052 ms -> 0.471702 ms
f16o 12288 2560 12288 512 -1086328089   # 0.978456 ms -> 0.782912 ms
f16o 12288 2560 12288 768 -1086328089   # 1.49446 ms -> 1.22893 ms
f16o 12288 2560 12288 1024 -1086328073   # 1.86293 ms -> 1.65612 ms
f16o 12288 2560 12288 1536 -1086328073   # 2.84602 ms -> 2.45139 ms
f16o 12288 2560 12288 2048 -1086328066   # 3.58288 ms -> 3.22614 ms
f16o 12288 2560 12288 4096 -1086328066   # 6.57714 ms -> 6.15869 ms
f16o 12288 2560 12288 8192 -1086328066   # 13.2365 ms -> 12.5598 ms
f16o 6144 2560 6144 256 -1086328303   # 0.978043 ms -> 0.26645 ms
f16o 6144 2560 6144 512 -1086328089   # 1.82723 ms -> 0.478555 ms
f16o 6144 2560 6144 768 default   # 0.634327 ms
f16o 6144 2560 6144 1024 default   # 0.788085 ms
f16o 6144 2560 6144 1536 default   # 1.2302 ms
f16o 6144 2560 6144 2048 -1086328073   # 2.30385 ms -> 1.61118 ms
f16o 6144 2560 6144 4096 default   # 3.05293 ms
f16o 6144 2560 6144 8192 default   # 5.96723 ms
f16o 2560 6144 2560 256 -1086328136   # 1.23379 ms -> 0.278677 ms
f16o 2560 6144 2560 512 -1086328105   # 1.27985 ms -> 0.446675 ms
f16o 2560 6144 2560 768 default   # 0.748205 ms
f16o 2560 6144 2560 1024 default   # 0.93767 ms
f16o 2560 6144 2560 1536 -1086328089   # 1.52944 ms -> 1.32132 ms
f16o 2560 6144 2560 2048 default   # 1.73281 ms
f16o 2560 6144 2560 4096 default   # 3.50691 ms
f16o 2560 6144 2560 8192 default   # 7.29727 ms
f16o 2560 640 2560 256 -1086328292   # 0.0415197 ms -> 0.0360263 ms
f16o 2560 640 2560 512 -1086328082   # 0.070039 ms -> 0.0551727 ms
f16o 2560 640 2560 768 -1086328080   # 0.0913723 ms -> 0.074799 ms
f16o 2560 640 2560 1024 -1086328099   # 0.102719 ms -> 0.0898123 ms
f16o 2560 640 2560 1536 -1086328099   # 0.148758 ms -> 0.130398 ms
f16o 2560 640 2560 2048 default   # 0.175625 ms
f16o 2560 640 2560 4096 default   # 0.33413 ms
f16o 2560 640 2560 8192 default   # 0.660913 ms
f16o 2560 2560 2560 256 default   # 0.121172 ms
f16o 2560 2560 2560 512 -1086328105   # 0.214691 ms -> 0.191611 ms
f16o 2560 2560 2560 768 default   # 0.28757 ms
f16o 2560 2560 2560 1024 -1086328073   # 0.393769 ms -> 0.332903 ms
f16o 2560 2560 2560 1536 default   # 0.496915 ms
f16o 2560 2560 2560 2048 default   # 0.660913 ms
f16o 2560 2560 2560 4096 -1086328089   # 1.5423 ms -> 1.27008 ms
f16o 2560 2560 2560 8192 -1086328073   # 2.7543 ms -> 2.50901 ms
f16o 640 2560 640 256 -1086328110   # 0.151425 ms -> 0.0534527 ms
f16o 640 2560 640 512 -1086328110   # 0.189491 ms -> 0.0783857 ms
f16o 640 2560 640 768 -1086328108   # 0.152652 ms -> 0.107425 ms
f16o 640 2560 640 1024 -1086328150   # 0.193811 ms -> 0.119865 ms
f16o 640 2560 640 1536 -1086328303   # 0.225251 ms -> 0.190064 ms
f16o 640 2560 640 2048 -1086328105   # 0.209491 ms -> 0.193105 ms
f16o 640 2560 640 4096 default   # 0.329503 ms
f16o 640 2560 640 8192 default   # 0.643126 ms
f16o 512 2560 512 256 -1086328160   # 0.145438 ms -> 0.0458663 ms
f16o 512 2560 512 512 -1086328108   # 0.189304 ms -> 0.0692127 ms
f16o 512 2560 512 768 -1086328160   # 0.137879 ms -> 0.0925323 ms
f16o 512 2560 512 1024 -1086328108   # 0.144239 ms -> 0.110732 ms
f16o 512 2560 512 1536 -1086328154   # 0.213091 ms -> 0.156625 ms
f16o 512 2560 512 2048 -1086328073   # 0.203198 ms -> 0.191864 ms
f16o 512 2560 512 4096 default   # 0.327583 ms
f16o 512 2560 512 8192 default   # 0.632553 ms
f16o 128 2560 128 256 -1086328166   # 0.0572793 ms -> 0.0304797 ms
f16o 128 2560 128 512 -1086328155   # 0.184798 ms -> 0.0352397 ms
f16o 128 2560 128 768 -1086328155   # 0.104545 ms -> 0.0426263 ms
f16o 128 2560 128 1024 -1086328166   # 0.109719 ms -> 0.0470927 ms
f16o 128 2560 128 1536 -1086328155   # 0.145785 ms -> 0.0635193 ms
f16o 128 2560 128 2048 -1086328108   # 0.119719 ms -> 0.0723993 ms
f16o 128 2560 128 4096 -1086328088   # 0.144412 ms -> 0.119399 ms
f16o 128 2560 128 8192 default   # 0.194691 ms
f16o 48 2560 48 256 -1086328226   # 0.0430263 ms -> 0.01968 ms
f16o 48 2560 48 512 -1086328167   # 0.186145 ms -> 0.0363597 ms
f16o 48 2560 48 768 -1086328165   # 0.0980123 ms -> 0.037853 ms
f16o 48 2560 48 1024 -1086328165   # 0.102039 ms -> 0.038493 ms
f16o 48 2560 48 1536 -1086328165   # 0.107106 ms -> 0.046373 ms
f16o 48 2560 48 2048 -1086328165   # 0.115452 ms -> 0.0523327 ms
f16o 48 2560 48 4096 -1086328311   # 0.178131 ms -> 0.0971057 ms
f16o 48 2560 48 8192 -1086328134   # 0.191385 ms -> 0.142972 ms
f16o 320 10240 320 256 -1086328165   # 0.37109 ms -> 0.123225 ms
f16o 320 10240 320 512 -1086328160   # 0.539701 ms -> 0.188625 ms
f16o 320 10240 320 768 -1086328108   # 0.865764 ms -> 0.219211 ms
f16o 320 10240 320 1024 -1086328166   # 1.17451 ms -> 0.344369 ms
f16o 320 10240 320 1536 -1086328104   # 1.56313 ms -> 0.442622 ms
f16o 320 10240 320 2048 -1086328109   # 2.14796 ms -> 0.57314 ms
f16o 320 10240 320 4096 -1086328103   # 9.17455 ms -> 1.04079 ms
f16o 320 10240 320 8192 default   # 2.14887 ms
f16o 10240 320 10240 256 -1086328321   # 0.0553727 ms -> 0.0521463 ms
f16o 10240 320 10240 512 default   # 0.092839 ms
f16o 10240 320 10240 768 default   # 0.133639 ms
f16o 10240 320 10240 1024 default   # 0.172398 ms
f16o 10240 320 10240 1536 default   # 0.251037 ms
f16o 10240 320 10240 2048 default   # 0.329716 ms
f16o 10240 320 10240 4096 default   # 0.64222 ms
f16o 10240 320 10240 8192 default   # 1.28463 ms
f16o 4 10240 4 256 -1086328238   # 0.090026 ms -> 0.047693 ms
f16o 4 10240 4 512 -1086328226   # 0.243144 ms -> 0.0523193 ms
f16o 4 10240 4 768 -1086328236   # 0.243077 ms -> 0.0584793 ms
f16o 4 10240 4 1024 -1086328236   # 0.243731 ms -> 0.0767593 ms
f16o 4 10240 4 1536 -1086328236   # 0.252277 ms -> 0.107026 ms
f16o 4 10240 4 2048 -1086328226   # 0.342943 ms -> 0.131998 ms
f16o 4 10240 4 4096 -1086328310   # 0.485461 ms -> 0.242717 ms
f16o 4 10240 4 8192 -1086328220   # 1.03616 ms -> 0.469662 ms
f16o 1 2560 1 256 -9   # 0.0410663 ms -> 0.00781333 ms
f16o 1 2560 1 512 -9   # 0.185958 ms -> 0.00988 ms
f16o 1 2560 1 768 -9   # 0.0816927 ms -> 0.0117597 ms
f16o 1 2560 1 1024 -9   # 0.0863723 ms -> 0.015733 ms
f16o 1 2560 1 1536 -9   # 0.0866123 ms -> 0.0187463 ms
f16o 1 2560 1 2048 -9   # 0.114119 ms -> 0.023613 ms
f16o 1 2560 1 4096 -9   # 0.0687993 ms -> 0.0440127 ms
f16o 1 2560 1 8192 -9   # 0.188958 ms -> 0.0770793 ms
