lapack: add Dlagv2 and its tests #1893
Conversation
Codecov ReportAttention:
❗ Your organization needs to install the Codecov GitHub app to enable full functionality. Additional details and impacted files@@ Coverage Diff @@
## master #1893 +/- ##
==========================================
- Coverage 74.00% 73.89% -0.11%
==========================================
Files 526 534 +8
Lines 63210 63694 +484
==========================================
+ Hits 46778 47069 +291
- Misses 13854 14025 +171
- Partials 2578 2600 +22 ☔ View full report in Codecov by Sentry. |
|
@kortschak @vladimir-ch What else is needed for the merge of this PR? |
|
I need time to review but I'm now busy with other tasks. I'd also like to get to the CMatrix formatting PR that's been waiting for too long. |
lapack/gonum/dlagv2.go
Outdated
| case len(beta) < 2: | ||
| panic(badLenBeta) | ||
| case b[0] <= 0 || b[ldb+1] <= 0 || b[0] < b[ldb+1]: | ||
| panic(badBDiag) |
There was a problem hiding this comment.
I know the documentation says b11 >= b22 > 0 but the code below seems to be able to handle non-positive values, so is a panic justified here? If I remove the panic here and the max(ulp,b[0]) in the test, the test still passes.
There was a problem hiding this comment.
It does not pass on my machine. Might be because I'm no longer silently returning below.
| t.Fatalf("%s: wi1 != -wi2; wi1=%g,wi2=%g,\n%s\n%s", name, wi1, wi2, aStr, bStr) | ||
| return | ||
| } | ||
| if math.Abs(b.Data[b.Stride]) > tol { |
There was a problem hiding this comment.
Do you have a test case where b[stride] is actually non-zero?
There was a problem hiding this comment.
B as input has a NaN in this position which is modified to be 0. We ensure it is zero on output.
lapack/testlapack/dlagv2.go
Outdated
|
|
||
| res, err := residualDlag2(aCopy, bCopy, beta1, complex(wr1, wi1)) | ||
| if err != nil { | ||
| // t.Logf("%s: invalid input data: %v\n%s\n%s\n%s\n%s", name, err, aStr, bStr, aOrigStr, bOrigStr) |
There was a problem hiding this comment.
Please clean up. And why silently return? Is the error an indication of a bad input data for the test?
| t.Fatalf("%s: wi1 != -wi2; wi1=%g,wi2=%g,\n%s\n%s", name, wi1, wi2, aStr, bStr) | ||
| return | ||
| } | ||
| if math.Abs(b.Data[b.Stride]) > tol { |
There was a problem hiding this comment.
B as input has a NaN in this position which is modified to be 0. We ensure it is zero on output.
lapack/gonum/dlagv2.go
Outdated
| case len(beta) < 2: | ||
| panic(badLenBeta) | ||
| case b[0] <= 0 || b[ldb+1] <= 0 || b[0] < b[ldb+1]: | ||
| panic(badBDiag) |
There was a problem hiding this comment.
It does not pass on my machine. Might be because I'm no longer silently returning below.
Part of #1651.
Additional notes:
isSchurCanonical*functions to take in a tolerance and left existing call sites as taking 0 tolerance.badBDiag