Craig-
I get the following test failures for commit id
8244725170a561f481c70933d057cc8aeb290284 ::
t/matrixops.t ............... 1/28 Can't call method "slice" on an undefined
value at t/matrixops.t line 45.
t/matmult.t ................. 1/5 Use of uninitialized value $PDL::Primitive::d
in concatenation (.) or string at /c/chm/pdl/git/pdl/blib/lib/PDL/Primitive.pm
line 248.
Use of uninitialized value $PDL::Primitive::d in concatenation (.) or string at
/c/chm/pdl/git/pdl/blib/lib/PDL/Primitive.pm line 248.
Test Summary Report
-------------------
t/matrixops.t (Wstat: 65280 Tests: 4 Failed: 0)
Non-zero exit status: 255
Parse errors: Bad plan. You planned 28 tests but ran 4.
t/subclass3.t (Wstat: 0 Tests: 7 Failed: 1)
Failed test: 3
Files=110, Tests=1184, 91 wallclock secs ( 0.20 usr 0.25 sys + 56.04 cusr
28.18 csys = 84.68 CPU)
Result: FAIL
Failed 2/110 test programs. 1/1184 subtests failed.
make: *** [test_dynamic] Error 255
An the following failures for the 64x64 modification in
commit id 7760030bd6520c606d8d9d5b2cee2c9ac554123c ::
Test Summary Report
-------------------
t/limits_range.t (Wstat: 256 Tests: 6 Failed: 1)
Failed test: 5
Non-zero exit status: 1
t/limits_ulimits.t (Wstat: 1280 Tests: 26 Failed: 5)
Failed tests: 19-23
Non-zero exit status: 5
t/matrixops.t (Wstat: 65280 Tests: 4 Failed: 0)
Non-zero exit status: 255
Parse errors: Bad plan. You planned 28 tests but ran 4.
t/poly.t (Wstat: 0 Tests: 1 Failed: 1)
Failed test: 1
t/subclass3.t (Wstat: 0 Tests: 7 Failed: 1)
Failed test: 3
Files=110, Tests=1184, 88 wallclock secs ( 0.11 usr 0.20 sys + 56.04 cusr
27.96 csys = 84.32 CPU)
Result: FAIL
Failed 5/110 test programs. 8/1184 subtests failed.
make: *** [test_dynamic] Error 255
Did these changes pass tests on your platform? They
don't on cygwin/XP.
--Chris
On 5/8/2010 9:37 AM, Craig DeForest wrote:
> Okay, I was stupid. I just checked in a new version that does the same
> operation in 30 seconds on my mac. Tile size is larger (64x64), and I
> avoid calculating offsets with PP macros inside the hottest loop - it
> walks through using dimincs in the summation direction. So the $a x $b
> case below now runs in 30 seconds, a factor-of-2.7 improvement (on my
> platform) over yesterday morning.
>
> Cheers,
> Craig
>
>
>
> On May 8, 2010, at 6:30 AM, Chris Marshall wrote:
>
>> On 5/8/2010 3:44 AM, Craig DeForest wrote:
>>>
>>> Er, sorry, I was noodling around and may have jumped the gun. I just
>>> checked in a small speed improvement for matmult. It just evaluates
>>> the terms in the matrix product in tiled order (multiplying 32x32
>>> tiles) rather than in direct threading order; that fits each tile into
>>> 16k in the double-precision case, which is small enough to fit in L1
>>> cache of most performance CPUs. Unsurprisingly, it helps.
>>> Surprisingly, not so very much. On my PowerBook:
>>>
>>> perldl> $a = random(2000,2000);
>>> perldl> $b = random(2000,2000);
>>> perldl> {$t0=time; $c = $a->dummy(1)->inner($b->xchg(0,1)->dummy(2));
>>> ..{> $t1=time; print $t1-$t0,"\n";}
>>> 82
>>>
>>> perldl> {$t0=time; $d = $a x $b;
>>> ..{> $t1=time; print $t1-$t0,"\n";}
>>> 70
>>>
>>> perldl> print all($d==$c)
>>> 1
>>>
>>> I am a bit puzzled how these other packages manage to go so much
>>> faster...
>>
>> The tiled calculation optimizes the memory accesses
>> but, if I understand correctly, the tile code still
>> uses the existing inner product algorithm.
>>
>> If so the total memory ops is still N**3 rather than
>> the optimal N**2, they've just been moved to a different
>> level of the memory hierarchy.
>>
>> Another possibility is that a 32x32 tile is not
>> big enough to hide the memory access time for the
>> data behind the floating point calculations.
>>
>> As to other packages performance, I'm sure an
>> optimized C matrix multiply routine that did the
>> entire optimization would be very fast. I don't
>> know how much threading would/could be supported.
>>
>> --Chris
>>> Cheers,
>>> Craig
_______________________________________________
Perldl mailing list
[email protected]
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
