% auto-generated by code/gen_numbers.py -- do not edit
\newcommand{\nActiveParams}{39.7}
\newcommand{\nAmpFactor}{1.54}
\newcommand{\nAmplification}{10.7}
\newcommand{\nBitsFreq}{1.51}
\newcommand{\nBitsNoLdlq}{1.51}
\newcommand{\nBitsNoRht}{1.51}
\newcommand{\nBitsOurs}{1.51}
\newcommand{\nBitsOursOne}{1.01}
\newcommand{\nBitsOursTwo}{2.01}
\newcommand{\nBitsRtnThree}{3.25}
\newcommand{\nBitsRtnTwo}{2.25}
\newcommand{\nCacheExperts}{2,543}
\newcommand{\nCacheExpertsFP}{238}
\newcommand{\nCacheFrac}{12.42}
\newcommand{\nCacheFracFP}{1.16}
\newcommand{\nDecodeFpEquiv}{2.4}
\newcommand{\nDecodeGap}{25}
\newcommand{\nDecodeGw}{1.20}
\newcommand{\nDecodePacked}{0.22}
\newcommand{\nDistinctEight}{38.2}
\newcommand{\nDistinctOne}{8.0}
\newcommand{\nDistinctThirtyTwo}{58.3}
\newcommand{\nExpertMB}{9.44}
\newcommand{\nExpertMBfp}{100.7}
\newcommand{\nExpertSlots}{20,480}
\newcommand{\nFootprintFP}{2,092}
\newcommand{\nFootprintOne}{131}
\newcommand{\nFootprintOnePFive}{196}
\newcommand{\nFootprintTwo}{261}
\newcommand{\nFusedBw}{116}
\newcommand{\nFusedRatio}{1.32}
\newcommand{\nGPUName}{NVIDIA RTX A500 Laptop GPU}
\newcommand{\nHitRate}{35.5}
\newcommand{\nHitRateFP}{5.8}
\newcommand{\nHybridFifteen}{24.9}
\newcommand{\nHybridFive}{10.6}
\newcommand{\nHybridForty}{67.3}
\newcommand{\nHybridTen}{18.4}
\newcommand{\nHybridTwelve}{21.9}
\newcommand{\nHybridTwentyFive}{38.2}
\newcommand{\nHybridTwo}{4.9}
\newcommand{\nIOatExpert}{5.46}
\newcommand{\nLruFifteen}{37.0}
\newcommand{\nLruFive}{0.0}
\newcommand{\nLruForty}{66.0}
\newcommand{\nLruTen}{0.0}
\newcommand{\nLruTwelve}{25.3}
\newcommand{\nLruTwentyFive}{48.4}
\newcommand{\nLruTwo}{0.0}
\newcommand{\nMBperTok}{3,021}
\newcommand{\nMassTopQ}{44.9}
\newcommand{\nNumRuns}{8}
\newcommand{\nPCIe}{6.11}
\newcommand{\nPPLFreq}{22.02}
\newcommand{\nPPLNoLdlq}{7,701.98}
\newcommand{\nPPLNoRht}{352.10}
\newcommand{\nPPLOurs}{25.54}
\newcommand{\nPPLOursOne}{156.07}
\newcommand{\nPPLOursTwo}{12.17}
\newcommand{\nPPLRtnThree}{10.49}
\newcommand{\nPPLRtnTwo}{22,793.90}
\newcommand{\nPPLfp}{8.11}
\newcommand{\nPeakTflops}{6.94}
\newcommand{\nRamCopy}{12.9}
\newcommand{\nRandBest}{6.01}
\newcommand{\nRandBestBlock}{1}
\newcommand{\nRandBestThreads}{4}
\newcommand{\nRandOneThread}{2.81}
\newcommand{\nRandSmall}{0.49}
\newcommand{\nReqDecode}{5.46}
\newcommand{\nResidentGB}{2.42}
\newcommand{\nReuseMean}{37.0}
\newcommand{\nSeqRead}{4.79}
\newcommand{\nSlots}{1,024}
\newcommand{\nSlotsFP}{238}
\newcommand{\nSpeedupOverFP}{16.4}
\newcommand{\nStaticFifteen}{31.0}
\newcommand{\nStaticFive}{13.5}
\newcommand{\nStaticForty}{62.4}
\newcommand{\nStaticMAE}{0.00}
\newcommand{\nStaticTen}{22.9}
\newcommand{\nStaticTwelve}{27.1}
\newcommand{\nStaticTwentyFive}{45.1}
\newcommand{\nStaticTwo}{6.2}
\newcommand{\nTokSecBatch}{3.08}
\newcommand{\nTokSecEight}{2.17}
\newcommand{\nTokSecFP}{0.11}
\newcommand{\nTokSecOne}{1.81}
\newcommand{\nTokSecOneBit}{3.21}
\newcommand{\nTokSecOneBitBatch}{5.48}
\newcommand{\nTokSecTwoBit}{1.22}
\newcommand{\nTokenWS}{128}
\newcommand{\nTokenWSFrac}{12.5}
\newcommand{\nTotalParams}{1,045.8}
\newcommand{\nTraceTokens}{49,152}
\newcommand{\nWorkSetKilo}{63.2}
\newcommand{\nWorkSetOne}{8.0}
\newcommand{\nWorkSetSixteen}{42.1}
\newcommand{\nWorkSetSixtyFour}{56.6}
\newcommand{\nZipfBias}{6.84}
\newcommand{\nZipfMAE}{6.84}
\newcommand{\nZipfS}{0.65}
\newcommand{\nZipfSmeas}{0.65}