testsuite results

[Daniel Fischer]

On Thursday 12 May 2011 17:49:16, Simon Peyton-Jones wrote:
> |    hpc_markup_multi_001(normal)
> |    hpc_markup_multi_002(normal)
> |    hpc_markup_multi_003(normal)
> | 
> | Unexpected passes:
> |    mc01(hpc,ghci)
> |    mc06(hpc,ghci)
> |    mc08(hpc,ghci)
> |    mc11(hpc)
> |    mc16(hpc)
> |    mc18(hpc)
> 
> I pushed patches for all of these today
> 

Great!
Now I've built HEAD (7b3a746294d3d034da0052644237e4d1ab1f08c8, minus the 
patch removing the import Config from AsmCodeGen, cf #5194) with profiling, 
leading to

====================
OVERALL SUMMARY for test run started at Do 12. Mai 20:47:35 CEST 2011
    2766 total tests, which gave rise to
   11839 test cases, of which
       0 caused framework failures
    2123 were skipped

    9428 expected passes
     280 expected failures
       1 unexpected passes
       7 unexpected failures
====================
Unexpected passes:
   conc016(threaded2)

Well, that one occasionally happens.
====================
Unexpected failures:
   T3016(profasm)

timeout-killed, too much other stuff running, rerunning T3016 on a less 
busy machine made it pass, but I'm not too happy with the stats:

<<ghc: 29262116704 bytes, 55739 GCs, 88426196/318733428 avg/max bytes 
residency (35 samples), 668M in use, 0.00 INIT (0.00 elapsed), 81.46 MUT 
(104.38 elapsed), 42.76 GC (103.41 elapsed) :ghc>>
====================
   T3064(normal)

This one failed due to ghc being too good:

bytes allocated 59305220 is less than minimum allowed 65000000
If this is because you have improved GHC, please
update the test so that GHC doesn't regress again
*** unexpected failure for T3064(normal)

On an almost idle machine, the figures are even lower:

 [("bytes allocated", "58378436")
 ,("num_GCs", "30")
 ,("average_bytes_used", "1594860")
 ,("max_bytes_used", "3064640")
 ,("num_byte_usage_samples", "2")
 ,("peak_megabytes_allocated", "8")
 ,("init_cpu_seconds", "0.00")
 ,("init_wall_seconds", "0.00")
 ,("mutator_cpu_seconds", "0.16")
 ,("mutator_wall_seconds", "0.17")
 ,("GC_cpu_seconds", "0.07")
 ,("GC_wall_seconds", "0.07")
 ]
====================

   T5084(normal)
   dph-diophantine-opt(normal,threaded1,threaded2)

The same as before, nothing to worry about.
====================
   dph-words-opt(normal)

This one was also timeout-killed on a busy machine, a later re-run let the 
compilation finish, but it took quite long:

[1 of 2] Compiling WordsVect        ( WordsVect.hs, WordsVect.o )
[2 of 2] Compiling Main             ( Main.hs, Main.o )
Linking dph-words-opt ...
<<ghc: 52843523420 bytes, 101270 GCs, 53522750/118551624 avg/max bytes 
residency (67 samples), 282M in use, 0.00 INIT (0.01 elapsed), 153.19 MUT 
(163.87 elapsed), 70.10 GC (70.05 elapsed) :ghc>>

The run of the programme again ended with an invalid slice.
====================

> | Summing up: Yay!
> 
> Indeed!

Still very much yay, I can't find the new unexpected failure of T3064 
disappointing in the least, only T3016(profasm) and dph-words-opt leave 
something to be desired.