diff --git a/scripts/eval.r b/scripts/eval.r
deleted file mode 100644
index 37106f9..0000000
--- a/scripts/eval.r
+++ /dev/null
@@ -1,78 +0,0 @@
-aggr_runs <- function(df, n, fun){
-    aggregate(x = df,
-              by = list(gl(ceiling(nrow(df)/n), n)[1:nrow(df)]),
-              FUN = fun)
-}
-
-evalfiles <- function(csvfiles) {
-    result <- list()
-    for(csvfile in csvfiles) {
-        ## import
-        r<-read.table(csvfile,
-                      sep=",",
-                      quote='"',
-                      header=T,
-                      skip=3)
-        ## only successful runs
-        r<-r[r[11]=="Success",]
-
-        ## collapse benchmark runs : static column body
-        tmp <- r[seq(1,nrow(r),5),1:9]
-        total <- tmp$nx * ifelse(tmp$ny>0, tmp$ny, 1) * ifelse(tmp$nz>0, tmp$nz, 1)
-        body_aggr <- cbind(tmp, total)
-
-        ## collapse benchmark runs : mean
-        tmp <- aggr_runs(r[,12:ncol(r)], nruns, mean)
-        ## union of benchmark data and aggregated runs
-        ## excluding run and success column
-        df.means<-cbind(body_aggr,tmp)
-
-        ## bandwidth in MiB/s
-        tmp <- df.means[,grepl( "Transfer", names(df.means) )] / 1048.576 /
-            df.means[,grepl( "Upload|Download", names(df.means) )]
-        bw.0 <- cbind(body_aggr,tmp)
-        ## further aggregated bandwidth
-        bw.1 <- aggregate(x=tmp,
-                          by=list(bw.0$inplace,bw.0$complex,bw.0$precision,bw.0$dim),
-                          FUN=mean)
-        ## even further aggregation, collapse dim column
-        bw.2 <- aggregate(x=tmp,
-                          by=list(bw.0$inplace,bw.0$complex,bw.0$precision),
-                          FUN=mean)
-        ## times
-        powerof2 <- df.means[df.means$kind=="powerof2",]
-        oddshape <- df.means[df.means$kind=="oddshape",]
-        radix357 <- df.means[df.means$kind=="radix357",]
-        
-        lib <- tolower(as.character(r[1,1]))
-        print(lib)
-        result[[lib]] <- list(bw.0=bw.0,
-                    bw.1=bw.1,
-                    bw.2=bw.2,
-                    powerof2=powerof2,
-                    oddshape=oddshape,
-                    radix357=radix357)
-    }
-    return(result)
-}
-
-colours <- c("red", "orange", "blue", "yellow", "green")
-nruns <- 5 # number of benchmark runs
-csvfiles <- c("../results/K80/cuda-7.5.18/clfft_gcc5.3.0_RHEL6.7.csv",
-              "../results/K80/cuda-7.5.18/cufft_gcc5.3.0_RHEL6.7.csv")
-k80 <- evalfiles(csvfiles)
-## join columns inplace, complex, precision into one column
-fftypes <- do.call(paste,c(k80[[1]]$bw.2[,1:3],sep=" "))
-## prepare data matrix for barchart
-bw<-as.matrix(cbind(k80[[1]]$bw.2[4],
-                    k80[[2]]$bw.2[4]))
-colnames(bw)<-names(k80)
-rownames(bw)<-fftypes
-
-barplot(t(bw),
-        main="Average Bandwidth on K80",
-        xlab="FFT type",
-        ylab="MiB/s",
-        col=colours[1:2],
-        legend = colnames(bw),
-        beside=TRUE)