luminal cuda fixes

Author: jafioti

crates/luminal_cuda/src/block/interpreter.cu

@@ -1,4 +1,4 @@
-#include <device_atomic_functions.h>
+#include <device_atomic_functions.h> // need this for nanosleep
 enum OpCode {
   //%extra_op_codes%
 };

crates/luminal_cuda/src/runtime.rs

@@ -2,11 +2,13 @@ use crate::{
     block::{BlockOp, IntoBlockOp},
     kernel::KernelOp,
 };
-use cudarc::driver::{
-    CudaContext, CudaFunction, CudaSlice, CudaStream, DevicePtr, DevicePtrMut, DeviceRepr,
-    LaunchConfig, PushKernelArg, ValidAsZeroBits,
+use cudarc::{
+    driver::{
+        CudaContext, CudaFunction, CudaSlice, CudaStream, DevicePtr, DevicePtrMut, DeviceRepr,
+        LaunchConfig, PushKernelArg, ValidAsZeroBits,
+    },
+    nvrtc::{compile_ptx_with_opts, CompileOptions},
 };
-use cudarc::nvrtc::{compile_ptx_with_opts, CompileOptions};
 use fixedbitset::FixedBitSet;
 use itertools::Itertools;
 use luminal::op::Output;

src/egglog.egg

@@ -124,11 +124,11 @@
         (= ?n_elems (n_elements ?other))
     )
     (
-        (union ?e (ECons (MMul (MIter) ?n_elems) (RowMajor ?other)))
+        (union ?e (ECons ?n_elems (RowMajor ?other)))
     )
     :ruleset expr
 )
-(rewrite (RowMajor (ECons ?dim (ENil))) (ECons (MIter) (ENil)) :ruleset expr)
+(rewrite (RowMajor (ECons ?dim (ENil))) (ECons (MNum 1) (ENil)) :ruleset expr)
 
 (rewrite (MReplaceList (ECons ?expr ?list) ?from ?to) (ECons (MReplace ?expr ?from ?to) (MReplaceList ?list ?from ?to)) :ruleset expr)
 (rule