Closed li-yiyang closed 3 months ago
I will be getting an Apple Silicon machine soon and then I will be able to test.
(This potentially looks like an SBCL bug too.)
I have tested it on an M1 Mac mini with macOS Sonoma 14.5 and SBCL 2.4.7 & 2.4.6, and neither of them has the loading issue. It may just be a bug in SBCL that has already been fixed.
The failed test test-matrix-stacking
is not related to this loading issue. It was introduced in commit 07edc0f7a17ac77ccae9cbd4506dfc51b58800d5 and can be fixed by:
diff --git a/tests/matrix-tests.lisp b/tests/matrix-tests.lisp
index 32a6f60..5ea3800 100644
--- a/tests/matrix-tests.lisp
+++ b/tests/matrix-tests.lisp
@@ -183,11 +183,11 @@
(is (magicl:= expected
(magicl:hstack
(loop :for j :below 3
- :collect (magicl:column expected j)))))
+ :collect (magicl:vector->column-matrix (magicl:column expected j))))))
(is (magicl:= expected
(magicl:vstack
(loop :for i :below 2
- :collect (magicl:row expected i)))))))
+ :collect (magicl:vector->row-matrix (magicl:row expected i))))))))
(deftest test-block-matrix-construction ()
Yes, thank you. After updating SBCL and it could be loaded successfully.
My machine: macbook air m1, 2020, Sonoma 14.3, sbcl 2.4.3 (homebrew), lapack 3.12.0 (homebrew).
When trying to load magicl, raise
The value 17314 is not of type (INTEGER -8192 16383) when binding SB-ASSEM:LABEL
error ((SB-ARM64-ASM::EMIT-TEST-BRANCH-IMM #<SB-ASSEM:SEGMENT {70098F1AF3}> 0 1 0 17314 14) [external]
).Could locate the error to
src/high-level/matrix-functions/mult-methods.lisp
at line 12:After commit the line, there's no error and could load magicl. But would fail one test for
((:METHOD NO-APPLICABLE-METHOD (T)) #<STANDARD-GENERIC-FUNCTION MAGICL:NROWS (1)> #S(MAGICL::VECTOR/INT32 :SIZE 2 :STORAGE #(1 4))) [fast-method]
(other test would pass though).Not know whether if for the apple's arm fault. (could load on amd64 machine)
Load code:
Test code: