CMake: Enabled using Accelerate on x86_64 / x64

[stemann]

Proposed changes

CMake: Enabled using Accelerate on x86_64 / x64.

Contributes to #1201 .

Cf. JuliaPackaging/Yggdrasil#9761

Checklist

Put an x in the boxes that apply.

• I have read the CONTRIBUTING document
• I have run pre-commit run --all-files to format my code / installed pre-commit prior to committing changes
• I have added tests that prove my fix is effective or that my feature works
• I have updated the necessary documentation (if needed)

[awni]

This change seems pretty reasonable.. but I think we can just remove MLX_BUILD_ARM since it's entirely unused at this point. Do you mind removing it?

[awni]

Also if we remove it we can also close #1626 which will be irrelevant.

[stemann]

Right, I will look into it.

[awni]

Thanks, will merge when tests clear!

[zcbenz]

@stemann @awni This change makes mac x64 build crash when running tests, you can simply reproduce by running python -m unittest discover python/tests:

Crashed Thread:        5

Exception Type:        EXC_BAD_ACCESS (SIGSEGV)
Exception Codes:       UNKNOWN_0xD at 0x0000000000000000
Exception Codes:       0x000000000000000d, 0x0000000000000000

Termination Reason:    Namespace SIGNAL, Code 11 Segmentation fault: 11
Terminating Process:   exc handler [21651]

VM Region Info: 0 is not in any region.  Bytes before following region: 4312813568
      REGION TYPE                    START - END         [ VSIZE] PRT/MAX SHRMOD  REGION DETAIL
      UNUSED SPACE AT START
--->  
      __TEXT                      101105000-101107000    [    8K] r-x/r-x SM=COW  /usr/local/Cellar/python@3.12/3.12.6/Frameworks/Python.framework/Versions/3.12/Resources/Python.app/Contents/MacOS/Python

Thread 0::  Dispatch queue: com.apple.main-thread
0   libsystem_kernel.dylib        	    0x7ff804e9cf7a __psynch_cvwait + 10
1   libsystem_pthread.dylib       	    0x7ff804eda6f3 _pthread_cond_wait + 1211
2   libc++.1.dylib                	    0x7ff804e13d12 std::__1::condition_variable::wait(std::__1::unique_lock<std::__1::mutex>&) + 18
3   libmlx.dylib                  	       0x103d3336b mlx::core::Event::wait() + 75
4   libmlx.dylib                  	       0x103139b42 mlx::core::array::wait() + 50
5   libmlx.dylib                  	       0x103207099 mlx::core::eval(std::__1::vector<mlx::core::array, std::__1::allocator<mlx::core::array>>) + 137
6   libmlx.dylib                  	       0x103139bf8 mlx::core::array::eval() + 152
7   core.cpython-312-darwin.so    	       0x101ad130e 0x101a8e000 + 275214
8   core.cpython-312-darwin.so    	       0x101ab17d6 0x101a8e000 + 145366
9   core.cpython-312-darwin.so    	       0x101a949f0 0x101a8e000 + 27120
10  Python                        	       0x101e53d1f _PyEval_EvalFrameDefault + 50484
11  Python                        	       0x101d5f290 method_vectorcall + 219
12  Python                        	       0x101e55c9b _PyEval_EvalFrameDefault + 58544
13  Python                        	       0x101d5bd5f _PyObject_FastCallDictTstate + 87
14  Python                        	       0x101d5d254 _PyObject_Call_Prepend + 116
15  Python                        	       0x101dceb71 slot_tp_call + 114
16  Python                        	       0x101d5bf3f _PyObject_MakeTpCall + 140
17  Python                        	       0x101e53ec1 _PyEval_EvalFrameDefault + 50902
18  Python                        	       0x101d5f290 method_vectorcall + 219
19  Python                        	       0x101e55c9b _PyEval_EvalFrameDefault + 58544
20  Python                        	       0x101d5bd5f _PyObject_FastCallDictTstate + 87
21  Python                        	       0x101d5d254 _PyObject_Call_Prepend + 116
22  Python                        	       0x101dceb71 slot_tp_call + 114
23  Python                        	       0x101d5bf3f _PyObject_MakeTpCall + 140
24  Python                        	       0x101e53ec1 _PyEval_EvalFrameDefault + 50902
25  Python                        	       0x101d5f290 method_vectorcall + 219
26  Python                        	       0x101e55c9b _PyEval_EvalFrameDefault + 58544
27  Python                        	       0x101d5bd5f _PyObject_FastCallDictTstate + 87
28  Python                        	       0x101d5d254 _PyObject_Call_Prepend + 116
29  Python                        	       0x101dceb71 slot_tp_call + 114
30  Python                        	       0x101d5bf3f _PyObject_MakeTpCall + 140
31  Python                        	       0x101e53ec1 _PyEval_EvalFrameDefault + 50902
32  Python                        	       0x101d5f290 method_vectorcall + 219
33  Python                        	       0x101e55c9b _PyEval_EvalFrameDefault + 58544
34  Python                        	       0x101d5bd5f _PyObject_FastCallDictTstate + 87
35  Python                        	       0x101d5d254 _PyObject_Call_Prepend + 116
36  Python                        	       0x101dceb71 slot_tp_call + 114
37  Python                        	       0x101d5bf3f _PyObject_MakeTpCall + 140
38  Python                        	       0x101e53ec1 _PyEval_EvalFrameDefault + 50902
39  Python                        	       0x101d5bdda _PyObject_FastCallDictTstate + 210
40  Python                        	       0x101dcf8b8 slot_tp_init + 209
41  Python                        	       0x101dc67ff type_call + 135
42  Python                        	       0x101d5bf3f _PyObject_MakeTpCall + 140
43  Python                        	       0x101e53ec1 _PyEval_EvalFrameDefault + 50902
44  Python                        	       0x101e4762f PyEval_EvalCode + 197
45  Python                        	       0x101e43da3 builtin_exec + 469
46  Python                        	       0x101dab4cb cfunction_vectorcall_FASTCALL_KEYWORDS + 94
47  Python                        	       0x101e53d1f _PyEval_EvalFrameDefault + 50484
48  Python                        	       0x101ecf89f pymain_run_module + 234
49  Python                        	       0x101ecef0e Py_RunMain + 553
50  Python                        	       0x101ecf615 pymain_main + 378
51  Python                        	       0x101ecf6c8 Py_BytesMain + 42
52  dyld                          	    0x7ff804b4e366 start + 1942

Thread 1:
0   libsystem_kernel.dylib        	    0x7ff804e9cf7a __psynch_cvwait + 10
1   libsystem_pthread.dylib       	    0x7ff804eda6f3 _pthread_cond_wait + 1211
2   libc++.1.dylib                	    0x7ff804e13d12 std::__1::condition_variable::wait(std::__1::unique_lock<std::__1::mutex>&) + 18
3   libmlx.dylib                  	       0x103cea31c ThreadPool::ThreadPool(unsigned long)::'lambda'()::operator()() const + 156
4   libmlx.dylib                  	       0x103cea232 void* std::__1::__thread_proxy[abi:ne180100]<std::__1::tuple<std::__1::unique_ptr<std::__1::__thread_struct, std::__1::default_delete<std::__1::__thread_struct>>, ThreadPool::ThreadPool(unsigned long)::'lambda'()>>(void*) + 50
5   libsystem_pthread.dylib       	    0x7ff804eda18b _pthread_start + 99
6   libsystem_pthread.dylib       	    0x7ff804ed5ae3 thread_start + 15

Thread 2:
0   libsystem_kernel.dylib        	    0x7ff804e9cf7a __psynch_cvwait + 10
1   libsystem_pthread.dylib       	    0x7ff804eda6f3 _pthread_cond_wait + 1211
2   libc++.1.dylib                	    0x7ff804e13d12 std::__1::condition_variable::wait(std::__1::unique_lock<std::__1::mutex>&) + 18
3   libmlx.dylib                  	       0x103cea31c ThreadPool::ThreadPool(unsigned long)::'lambda'()::operator()() const + 156
4   libmlx.dylib                  	       0x103cea232 void* std::__1::__thread_proxy[abi:ne180100]<std::__1::tuple<std::__1::unique_ptr<std::__1::__thread_struct, std::__1::default_delete<std::__1::__thread_struct>>, ThreadPool::ThreadPool(unsigned long)::'lambda'()>>(void*) + 50
5   libsystem_pthread.dylib       	    0x7ff804eda18b _pthread_start + 99
6   libsystem_pthread.dylib       	    0x7ff804ed5ae3 thread_start + 15

Thread 3:
0   libsystem_kernel.dylib        	    0x7ff804e9cf7a __psynch_cvwait + 10
1   libsystem_pthread.dylib       	    0x7ff804eda6f3 _pthread_cond_wait + 1211
2   libc++.1.dylib                	    0x7ff804e13d12 std::__1::condition_variable::wait(std::__1::unique_lock<std::__1::mutex>&) + 18
3   libmlx.dylib                  	       0x103cea31c ThreadPool::ThreadPool(unsigned long)::'lambda'()::operator()() const + 156
4   libmlx.dylib                  	       0x103cea232 void* std::__1::__thread_proxy[abi:ne180100]<std::__1::tuple<std::__1::unique_ptr<std::__1::__thread_struct, std::__1::default_delete<std::__1::__thread_struct>>, ThreadPool::ThreadPool(unsigned long)::'lambda'()>>(void*) + 50
5   libsystem_pthread.dylib       	    0x7ff804eda18b _pthread_start + 99
6   libsystem_pthread.dylib       	    0x7ff804ed5ae3 thread_start + 15

Thread 4:
0   libsystem_kernel.dylib        	    0x7ff804e9cf7a __psynch_cvwait + 10
1   libsystem_pthread.dylib       	    0x7ff804eda6f3 _pthread_cond_wait + 1211
2   libc++.1.dylib                	    0x7ff804e13d12 std::__1::condition_variable::wait(std::__1::unique_lock<std::__1::mutex>&) + 18
3   libmlx.dylib                  	       0x103cea31c ThreadPool::ThreadPool(unsigned long)::'lambda'()::operator()() const + 156
4   libmlx.dylib                  	       0x103cea232 void* std::__1::__thread_proxy[abi:ne180100]<std::__1::tuple<std::__1::unique_ptr<std::__1::__thread_struct, std::__1::default_delete<std::__1::__thread_struct>>, ThreadPool::ThreadPool(unsigned long)::'lambda'()>>(void*) + 50
5   libsystem_pthread.dylib       	    0x7ff804eda18b _pthread_start + 99
6   libsystem_pthread.dylib       	    0x7ff804ed5ae3 thread_start + 15

Thread 5 Crashed:
0   libmlx.dylib                  	       0x103d2d7d0 mlx::core::Reduce::eval_cpu(std::__1::vector<mlx::core::array, std::__1::allocator<mlx::core::array>> const&, mlx::core::array&) + 6464
1   libmlx.dylib                  	       0x1032118bc std::__1::__function::__func<mlx::core::eval_impl(std::__1::vector<mlx::core::array, std::__1::allocator<mlx::core::array>>, bool)::$_0, std::__1::allocator<mlx::core::eval_impl(std::__1::vector<mlx::core::array, std::__1::allocator<mlx::core::array>>, bool)::$_0>, void ()>::operator()() + 172
2   libmlx.dylib                  	       0x1032038e2 mlx::core::scheduler::StreamThread::thread_fn() + 482
3   libmlx.dylib                  	       0x103203a93 void* std::__1::__thread_proxy[abi:ne180100]<std::__1::tuple<std::__1::unique_ptr<std::__1::__thread_struct, std::__1::default_delete<std::__1::__thread_struct>>, void (mlx::core::scheduler::StreamThread::*)(), mlx::core::scheduler::StreamThread*>>(void*) + 67
4   libsystem_pthread.dylib       	    0x7ff804eda18b _pthread_start + 99
5   libsystem_pthread.dylib       	    0x7ff804ed5ae3 thread_start + 15

[awni]

I guess the only difference is it's using the accelerate back-end (which should work). We'll have to debug that..