diff --git a/HSBencher/App.hs b/HSBencher/App.hs
--- a/HSBencher/App.hs
+++ b/HSBencher/App.hs
@@ -282,28 +282,30 @@
   ------------------------------------------
   -- (3) Produce output to the right places:
   ------------------------------------------
+  let pads n s = take (max 1 (n - length s)) $ repeat ' '
+      padl n x = pads n x ++ x 
+      padr n x = x ++ pads n x
   (t1,t2,t3,p1,p2,p3) <-
-    if not (all didComplete nruns) then do
-      log $ "\n >>> MIN/MEDIAN/MAX (TIME,PROD) -- got ERRORS: " ++show nruns
+    if all isError nruns then do
+      log $ "\n >>> MIN/MEDIAN/MAX (TIME,PROD) -- got only ERRORS: " ++show nruns
+      logOn [ResultsFile]$ 
+        printf "# %s %s %s %s %s" (padr 35 testRoot) (padr 20$ intercalate "_" args)
+                                  (padr 8$ sched) (padr 3$ show numthreads) (" ALL_ERRORS"::String)
       return ("","","","","","")
-    else do 
+    else do
+      let goodruns = filter (not . isError) nruns
       -- Extract the min, median, and max:
-      let sorted = sortBy (\ a b -> compare (realtime a) (realtime b)) nruns
+          sorted = sortBy (\ a b -> compare (gettime a) (gettime b)) goodruns
           minR = head sorted
           maxR = last sorted
           medianR = sorted !! (length sorted `quot` 2)
 
       let ts@[t1,t2,t3]    = map (\x -> showFFloat Nothing x "")
-                             [realtime minR, realtime medianR, realtime maxR]
-          prods@[p1,p2,p3] = map mshow [productivity minR, productivity medianR, productivity maxR]
+                             [gettime minR, gettime medianR, gettime maxR]
+          prods@[p1,p2,p3] = map mshow [getprod minR, getprod medianR, getprod maxR]
           mshow Nothing  = ""
           mshow (Just x) = showFFloat (Just 2) x "" 
 
-      let 
-          pads n s = take (max 1 (n - length s)) $ repeat ' '
-          padl n x = pads n x ++ x 
-          padr n x = x ++ pads n x
-
           -- These are really (time,prod) tuples, but a flat list of
           -- scalars is simpler and readable by gnuplot:
           formatted = (padl 15$ unwords $ ts)
@@ -321,13 +323,13 @@
             , _VARIANT  = sched
             , _ARGS     = args
             , _THREADS  = numthreads
-            , _MINTIME    =  realtime minR
-            , _MEDIANTIME =  realtime medianR
-            , _MAXTIME    =  realtime maxR
-            , _MINTIME_PRODUCTIVITY    = productivity minR
-            , _MEDIANTIME_PRODUCTIVITY = productivity medianR
-            , _MAXTIME_PRODUCTIVITY    = productivity maxR
-            , _ALLTIMES      =  unwords$ map (show . realtime) nruns
+            , _MINTIME    =  gettime minR
+            , _MEDIANTIME =  gettime medianR
+            , _MAXTIME    =  gettime maxR
+            , _MINTIME_PRODUCTIVITY    = getprod minR
+            , _MEDIANTIME_PRODUCTIVITY = getprod medianR
+            , _MAXTIME_PRODUCTIVITY    = getprod maxR
+            , _ALLTIMES      =  unwords$ map (show . gettime) goodruns
             , _TRIALS        =  trials
             }
       result' <- liftIO$ augmentResultWithConfig conf result
@@ -403,7 +405,7 @@
 ----------------------------------------------------------------------------------------------------
 
 
--- | TODO: Eventually this will be parameterized.
+-- | TODO: Eventually this will make sense when all config can be read from the environment, args, files.
 defaultMain :: IO ()
 defaultMain = do
   --      benchF = get "BENCHLIST" "benchlist.txt"
@@ -429,7 +431,7 @@
   writeIORef main_threadid id
 
   cli_args <- getArgs
-  let (options,args,errs) = getOpt Permute (concat$ map snd all_cli_options) cli_args
+  let (options,plainargs,errs) = getOpt Permute (concat$ map snd all_cli_options) cli_args
   let recomp  = NoRecomp `notElem` options
   
   when (ShowVersion `elem` options) $ do
@@ -438,18 +440,30 @@
       (unwords$ versionTags version)
     exitSuccess 
       
-  when (not (null errs && null args) || ShowHelp `elem` options) $ do
+  when (not (null errs) || ShowHelp `elem` options) $ do
     unless (ShowHelp `elem` options) $
       putStrLn$ "Errors parsing command line options:"
     mapM_ (putStr . ("   "++)) errs       
     putStrLn$ "\nUSAGE: [set ENV VARS] "++my_name++" [CMDLN OPTIONS]"
+    putStrLn$ "USAGE: command line options include patterns that select the benchmarks to run"
     mapM putStr (map (uncurry usageInfo) all_cli_options)
     putStrLn$ usageStr
     if (ShowHelp `elem` options) then exitSuccess else exitFailure
 
   conf0 <- getConfig options []
-  let conf1@Config{envs,benchlist,stdOut} = modConfig conf0
 
+  -- The list of benchmarks can optionally be narrowed to match any of the given patterns.
+  let conf1   = modConfig conf0
+      cutlist = case plainargs of
+                 [] -> benchlist conf1
+                 patterns -> filter (\ Benchmark{target,cmdargs} ->
+                                      any (\pat ->
+                                            isInfixOf pat target ||
+                                            any (isInfixOf pat) cmdargs)
+                                          patterns)
+                                    (benchlist conf1)
+  let conf2@Config{envs,benchlist,stdOut} = conf1{benchlist=cutlist}
+
   hasMakefile <- doesFileExist "Makefile"
   cabalFile   <- runLines "ls *.cabal"
   let hasCabalFile = (cabalFile /= []) &&
@@ -457,6 +471,9 @@
   rootDir <- getCurrentDirectory  
   runReaderT 
     (do
+        unless (null plainargs) $
+          logT$"There were "++show(length cutlist)++" benchmarks matching patterns: "++show plainargs
+        
         logT$"Beginning benchmarking, root directory: "++rootDir
         let globalBinDir = rootDir </> "bin"
         when recomp $ do
@@ -613,7 +630,7 @@
         log$ "--------------------------------------------------------------------------------"
 	liftIO$ exitSuccess
     )
-    conf1
+    conf2
 
 
 -- Several different options for how to display output in parallel:
@@ -655,6 +672,20 @@
 
 didComplete RunCompleted{} = True
 didComplete _              = False
+
+isError ExitError{} = True
+isError _           = False
+
+getprod RunCompleted{productivity} = productivity
+getprod RunTimeOut{}               = Nothing
+getprod x                          = error$"Cannot get productivity from: "++show x
+
+gettime RunCompleted{realtime} = realtime
+gettime RunTimeOut{}           = posInf
+gettime x                      = error$"Cannot get realtime from: "++show x
+
+posInf :: Double
+posInf = 1/0
 
 -- Shorthand for tagged version:
 logT str = log$hsbencher_tag++str
diff --git a/hsbencher.cabal b/hsbencher.cabal
--- a/hsbencher.cabal
+++ b/hsbencher.cabal
@@ -1,6 +1,6 @@
 
 name:                hsbencher
-version:             1.3.1
+version:             1.3.4
 -- CHANGELOG:
 -- 1.0   : Initial release, new flexible benchmark format.
 -- 1.1   : Change interface to RunInPlace
@@ -9,6 +9,8 @@
 -- 1.2.1 : Hack to shorten tuples, annoying URL length problem.
 -- 1.3   : breaking change to Config/RunResult types
 -- 1.3.1 : bugfix
+-- 1.3.2 : change policy for timeouts and errors
+-- 1.3.4 : Add ability to prune/select benchmarks with commmand line args
 
 synopsis:  Flexible benchmark runner for Haskell and non-Haskell benchmarks.