GPU backend

Adds GPU backend and "CUDA" feature to all the tests. As of MR creation, there were no forks in tested IRs so only vanilla non-parallel (single block, single thread) codegen has been tested.

hercules_cg/src/rt.rs

@@ -324,7 +324,7 @@ impl<'a> RTContext<'a> {
             Node::Constant { id: cons_id } => {
                 let block = &mut blocks.get_mut(&self.bbs.0[id.idx()]).unwrap();
                 write!(block, "                {} = ", self.get_value(id))?;
-                let mut size = None;
+                let mut size_and_device = None;
                 match self.module.constants[cons_id.idx()] {
                     Constant::Boolean(val) => write!(block, "{}bool", val)?,
                     Constant::Integer8(val) => write!(block, "{}i8", val)?,
@@ -351,14 +351,15 @@ impl<'a> RTContext<'a> {
                         write!(block, "backing_{}.byte_add(", device.name())?;
                         self.codegen_dynamic_constant(offset, block)?;
                         write!(block, " as usize)")?;
-                        size = Some(self.codegen_type_size(ty));
+                        size_and_device = Some((self.codegen_type_size(ty), device));
                     }
                 }
                 write!(block, ";\n")?;
-                if let Some(size) = size {
+                if let Some((size, device)) = size_and_device {
                     write!(
                         block,
-                        "                ::core::ptr::write_bytes({}, 0, {} as usize);\n",
+                        "                ::hercules_rt::__{}_zero_mem({}, {} as usize);\n",
+                        device.name(),
                         self.get_value(id),
                         size
                     )?;