Failure to build hmatrix-tests-0.4.1.0 with GHC 7.10.2 and stack #144

samvher · 2015-08-28T07:36:12Z

I updated my haskell-platform to try to build your current master (my base version was insufficient earlier), I have not yet managed to build the version in this repository but I now also get an error if I stack install hmatrix-tests after successfully installing hmatrix and hmatrix-gsl:

Run from outside a project, using implicit global config
Using resolver: lts-3.2 from global config file: /Users/sam/.stack/global/stack.yaml
NOTE: the install command is functionally equivalent to 'build --copy-bins'
Populated index cache.    
hmatrix-tests-0.4.1.0: configure
hmatrix-tests-0.4.1.0: build

--  While building package hmatrix-tests-0.4.1.0 using:
      /Users/sam/.stack/programs/x86_64-osx/ghc-7.10.2/bin/runhaskell -package=Cabal-1.22.4.0 -clear-package-db -global-package-db -package-db=/Users/sam/.stack/snapshots/x86_64-osx/lts-3.2/7.10.2/pkgdb/ /var/folders/19/5wrdz7h93tq2qx27bvhbkt9r0000gn/T/stack5697/Setup.hs --builddir=.stack-work/dist/x86_64-osx/Cabal-1.22.4.0/ build --ghc-options -hpcdir .stack-work/dist/x86_64-osx/Cabal-1.22.4.0/hpc/.hpc/ -ddump-hi -ddump-to-file
    Process exited with code: ExitFailure 1
    Logs have been written to: /Users/sam/.stack/global/.stack-work/logs/hmatrix-tests-0.4.1.0.log

    Configuring hmatrix-tests-0.4.1.0...
    Building hmatrix-tests-0.4.1.0...
    Preprocessing library hmatrix-tests-0.4.1.0...
    [1 of 4] Compiling Numeric.LinearAlgebra.Tests.Properties ( src/Numeric/LinearAlgebra/Tests/Properties.hs, .stack-work/dist/x86_64-osx/Cabal-1.22.4.0/build/Numeric/LinearAlgebra/Tests/Properties.o )

    src/Numeric/LinearAlgebra/Tests/Properties.hs:197:1:
        Illegal equational constraint ComplexOf t ~ Complex Double
        (Use GADTs or TypeFamilies to permit this)
        When checking that ‘eigProp’ has the inferred type
          eigProp :: forall t.
                     (Field t, ComplexOf t ~ Complex Double) =>
                     Matrix t -> Bool

The reason I haven't managed to install the master version yet is related to stack (I installed the hackage version first, then if I install the master version of hmatrix stack installs it locally in the project directory, which means I can't seem to make it available when I try to install gsl and tests, because the old version is still installed globally).

The text was updated successfully, but these errors were encountered:

samvher · 2015-08-28T08:47:20Z

Ok! After some more fiddling and nosing around, I found that hmatrix-tests does in fact build (the master version) on the latest GHC so that problem seems to be solved. Also, the mult tests that failed for me before are now successful. I managed to install the master version using stack (I overlooked something in the FAQ allowing me to install multiple packages locally).

However! The tests still don't completely pass on my machine :( Here is the output:

------ index
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ mult Double
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ sub-trans
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ ctrans
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ lu
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ inv (linearSolve)
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ luSolve
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ ldlSolve
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ cholSolve
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ luSolveLS
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests (5% trivial).
+++ OK, passed 100 tests (5% trivial).
------ pinv (linearSolveSVD)
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ det
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ svd
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ eig
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ nullSpace
+++ OK, passed 100 tests (51% trivial).
+++ OK, passed 100 tests (61% trivial).
------ qr
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ hess
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ schur
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
------ chol
+++ OK, passed 100 tests.
+++ OK, passed 100 tests.
*** Failed! Falsifiable (after 13 tests): 
PosDef (11><11)
 [  39.730968786211314,  0.1979928766936253,  12.465706523953791, -0.5112077392772991,  7.5960057283175235, -0.9670593597820314,  -5.935870547515698,   1.797363859395338, -1.0094703027385559,   5.577468310659227, -0.4972226349924339
 ,  0.1979928766936253,   75.14561462562708, -1.9974254725609648,  -1.770717997023017,  -3.503265657233948, 0.17261250443893594, -1.2879266102914397,   4.885056752728854,   6.156667250780151, -0.9378297974404508,   4.135399937757443
 ,  12.465706523953791, -1.9974254725609648,    55.4639573498602,  -4.012516656111143, -1.3331487266976585,  -3.722591446859042,   4.247657716781713, -2.0487171370893766,  -0.217294516677691,  1.6483887963949144,  0.2305992452092958
 , -0.5112077392772991,  -1.770717997023017,  -4.012516656111143,    57.5933371596545,   4.353451911593413, -2.4467335943194275, -3.2416994423985446,   5.227408327522042,  0.2937690710760158,   5.132618559884971, -2.6989748313524697
 ,  7.5960057283175235,  -3.503265657233948, -1.3331487266976585,   4.353451911593413,  48.142589092661886,  1.4195383322100628,   6.885535899430806,  1.1577267216715637,  10.790656256591262,  -6.367045247973574,   1.308532987164213
 , -0.9670593597820314, 0.17261250443893594,  -3.722591446859042, -2.4467335943194275,  1.4195383322100628,   77.33759994064724, -0.8067064745944341,  -3.306402338475162, -0.6385043765254745,  2.3508866716134627,  2.1978094219888655
 ,  -5.935870547515698, -1.2879266102914397,   4.247657716781713, -3.2416994423985446,   6.885535899430806, -0.8067064745944341,  42.692662060282096,  -4.873939713820163,   5.856905128308519,   -5.62027240991787,   5.284443287348966
 ,   1.797363859395338,   4.885056752728854, -2.0487171370893766,   5.227408327522042,  1.1577267216715637,  -3.306402338475162,  -4.873939713820163,   59.36434759773856,    5.10949999350114,  3.6303068127909004,  3.4362190675202573
 , -1.0094703027385559,   6.156667250780151,  -0.217294516677691,  0.2937690710760158,  10.790656256591262, -0.6385043765254745,   5.856905128308519,    5.10949999350114,  37.417326290979275,  1.4344468313481462,  -3.282705193366731
 ,   5.577468310659227, -0.9378297974404508,  1.6483887963949144,   5.132618559884971,  -6.367045247973574,  2.3508866716134627,   -5.62027240991787,  3.6303068127909004,  1.4344468313481462,  48.883781657971305,    1.27888571805805
 , -0.4972226349924339,   4.135399937757443,  0.2305992452092958, -2.6989748313524697,   1.308532987164213,  2.1978094219888655,   5.284443287348966,  3.4362190675202573,  -3.282705193366731,    1.27888571805805,    60.1246709433958 ]

albertoruiz · 2015-08-28T17:28:07Z

Interesting! Can you please run the following small program:

{-# LANGUAGE FlexibleContexts #-}

import Numeric.LinearAlgebra

a = (11><11)
[ 39.730968786211314, 0.1979928766936253, 12.465706523953791, -0.5112077392772991, 7.5960057283175235, -0.9670593597820314, -5.935870547515698, 1.797363859395338, -1.0094703027385559, 5.577468310659227, -0.4972226349924339
, 0.1979928766936253, 75.14561462562708, -1.9974254725609648, -1.770717997023017, -3.503265657233948, 0.17261250443893594, -1.2879266102914397, 4.885056752728854, 6.156667250780151, -0.9378297974404508, 4.135399937757443
, 12.465706523953791, -1.9974254725609648, 55.4639573498602, -4.012516656111143, -1.3331487266976585, -3.722591446859042, 4.247657716781713, -2.0487171370893766, -0.217294516677691, 1.6483887963949144, 0.2305992452092958
, -0.5112077392772991, -1.770717997023017, -4.012516656111143, 57.5933371596545, 4.353451911593413, -2.4467335943194275, -3.2416994423985446, 5.227408327522042, 0.2937690710760158, 5.132618559884971, -2.6989748313524697
, 7.5960057283175235, -3.503265657233948, -1.3331487266976585, 4.353451911593413, 48.142589092661886, 1.4195383322100628, 6.885535899430806, 1.1577267216715637, 10.790656256591262, -6.367045247973574, 1.308532987164213
, -0.9670593597820314, 0.17261250443893594, -3.722591446859042, -2.4467335943194275, 1.4195383322100628, 77.33759994064724, -0.8067064745944341, -3.306402338475162, -0.6385043765254745, 2.3508866716134627, 2.1978094219888655
, -5.935870547515698, -1.2879266102914397, 4.247657716781713, -3.2416994423985446, 6.885535899430806, -0.8067064745944341, 42.692662060282096, -4.873939713820163, 5.856905128308519, -5.62027240991787, 5.284443287348966
, 1.797363859395338, 4.885056752728854, -2.0487171370893766, 5.227408327522042, 1.1577267216715637, -3.306402338475162, -4.873939713820163, 59.36434759773856, 5.10949999350114, 3.6303068127909004, 3.4362190675202573
, -1.0094703027385559, 6.156667250780151, -0.217294516677691, 0.2937690710760158, 10.790656256591262, -0.6385043765254745, 5.856905128308519, 5.10949999350114, 37.417326290979275, 1.4344468313481462, -3.282705193366731
, 5.577468310659227, -0.9378297974404508, 1.6483887963949144, 5.132618559884971, -6.367045247973574, 2.3508866716134627, -5.62027240991787, 3.6303068127909004, 1.4344468313481462, 48.883781657971305, 1.27888571805805
, -0.4972226349924339, 4.135399937757443, 0.2305992452092958, -2.6989748313524697, 1.308532987164213, 2.1978094219888655, 5.284443287348966, 3.4362190675202573, -3.282705193366731, 1.27888571805805, 60.1246709433958 ] :: Matrix R

exactProp m = chol (trustSym m) == chol (trustSym (m+0))

main = do
print (exactProp a)
print $ chol (trustSym a) - chol (trustSym (a+0))

albertoruiz · 2015-08-28T17:34:08Z

Sorry for copying directly the code. The program is also here:

http://dis.um.es/profesores/alberto/material/hmatrix/bugchol.hs

samvher · 2015-08-28T17:40:51Z

This is the output:

True
(11><11)
 [ 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,                    0.0,                    0.0,                     0.0
 , 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,                    0.0,                    0.0,                     0.0
 , 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,                    0.0,                    0.0,                     0.0
 , 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,                    0.0,                    0.0,                     0.0
 , 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,                    0.0,                    0.0,                     0.0
 , 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,                    0.0,                    0.0,                     0.0
 , 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,                    0.0,                    0.0,                     0.0
 , 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,                    0.0,                    0.0,                     0.0
 , 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -8.881784197001252e-16, 1.1102230246251565e-16, -1.1102230246251565e-16
 , 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,                    0.0,                    0.0,                     0.0
 , 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,                    0.0,                    0.0,                     0.0 ]

samvher · 2015-08-28T17:45:46Z

(As an aside, out of personal interest, have you ever considered to/do you think it would be very difficult to put a GPU accelerated back-end behind hmatrix? E.g. cuBLAS + MAGMA/CULAtools? I think something like that behind a pure interface could be really powerful, for example for building Theano-like tools.)

albertoruiz · 2015-08-28T18:18:44Z

Again this is a problem with the tolerance of the test, the computation is correct. But in this case it is interesting to see that in some machines a+0 may not be exactly equal to a. This test was introduced to catch some old problem, but now it must be fixed or removed. You can comment out lines 666 and 667 in packages/tests/src/Numeric/LinearAlgebra/Tests.hs and recompile and run the tests to see if there are any other failures.

Thanks for testing!

A GPU back-end would be very useful but I have not yet tried it (some time ago only float elements were supported, if I am not mistaken) but now things are different with the explosion of deep learning. It should be easy to link with a GPU based blas/lapack library if the API is the same (although I don't know if memory transfers could be an issue). I will look into this.

I have just seen that there are some linear algebra libraries using cublas and magma:

https://github.com/bmsherman/LinAlg

samvher · 2015-08-28T19:57:19Z

That package looks interesting, I hadn't seen it before.

I commented out the faulty tests and everything works now.

Cheers

samvher closed this as completed Aug 28, 2015

albertoruiz mentioned this issue Sep 1, 2015

Test failure #145

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Failure to build hmatrix-tests-0.4.1.0 with GHC 7.10.2 and stack #144

Failure to build hmatrix-tests-0.4.1.0 with GHC 7.10.2 and stack #144

samvher commented Aug 28, 2015

samvher commented Aug 28, 2015

albertoruiz commented Aug 28, 2015

albertoruiz commented Aug 28, 2015

samvher commented Aug 28, 2015

samvher commented Aug 28, 2015

albertoruiz commented Aug 28, 2015

samvher commented Aug 28, 2015

Failure to build hmatrix-tests-0.4.1.0 with GHC 7.10.2 and stack #144

Failure to build hmatrix-tests-0.4.1.0 with GHC 7.10.2 and stack #144

Comments

samvher commented Aug 28, 2015

samvher commented Aug 28, 2015

albertoruiz commented Aug 28, 2015

albertoruiz commented Aug 28, 2015

samvher commented Aug 28, 2015

samvher commented Aug 28, 2015

albertoruiz commented Aug 28, 2015

samvher commented Aug 28, 2015