Use MTLCompiler directly (v2) (#7920)

* Use MTLCompiler directly (v2)

* to_block_literal and REQUEST_TYPE_COMPILE

* Rewrite command encoding

* Revert to_block_literal

* Maybe that's more readable to some people?

* Typo and comment about stdlib caching

* Update ops_metal.py

* Update ops_metal.py

* Update ops_metal.py

---------

Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com>

test/external/external_metal_compile_fail.py

@@ -225,6 +225,6 @@ from tinygrad.runtime.ops_metal import MetalDevice, MetalCompiler, MetalProgram
 
 if __name__ == "__main__":
   dev = MetalDevice("METAL")
-  lib = MetalCompiler(dev).compile(src)
+  lib = MetalCompiler().compile(src)
   prg = MetalProgram(dev, "r_64_32_8_16_4_6_6_4", lib)
 

test/test_metal.py

@@ -10,14 +10,12 @@ class TestMetal(unittest.TestCase):
       device.allocator.alloc(10000000000000000000)
 
   def test_compile_error(self):
-    device = MetalDevice("metal")
-    compiler = MetalCompiler(device)
+    compiler = MetalCompiler()
     with self.assertRaises(CompileError):
       compiler.compile("this is not valid metal")
 
   def test_compile_success(self):
-    device = MetalDevice("metal")
-    compiler = MetalCompiler(device)
+    compiler = MetalCompiler()
     ret = compiler.compile("""
 #include <metal_stdlib>
   using namespace metal;
@@ -41,7 +39,7 @@ class TestMetal(unittest.TestCase):
 
   def test_failed_newLibraryWithData(self):
     device = MetalDevice("metal")
-    compiler = MetalCompiler(device)
+    compiler = MetalCompiler()
     compiled = compiler.compile("""
 #include <metal_stdlib>
 kernel void r_5(device int* data0, const device int* data1, uint3 gid [[threadgroup_position_in_grid]], uint3 lid [[thread_position_in_threadgroup]]){