not fixed yet but switching machines

hercules_samples/call/build.rs

@@ -4,6 +4,8 @@ fn main() {
     JunoCompiler::new()
         .ir_in_src("call.hir")
         .unwrap()
+        .schedule_in_src(if cfg!(feature = "cuda") { "gpu.sch" } else { "cpu.sch" })
+        .unwrap()
         .build()
         .unwrap();
 }

hercules_samples/call/src/cpu.sch

+gvn(*);
+phi-elim(*);
+dce(*);
+
+auto-outline(*);
+
+ip-sroa(*);
+sroa(*);
+fork-split(*);
+unforkify(*);
+dce(*);
+float-collections(*);
+gvn(*);
+phi-elim(*);
+dce(*);
+
+infer-schedules(*);
+
+gcm(*);

hercules_samples/call/src/gpu.sch

+gvn(*);
+phi-elim(*);
+dce(*);
+
+let out = auto-outline(*);
+gpu(out.add);
+
+ip-sroa(*);
+sroa(*);
+dce(*);
+float-collections(*);
+gvn(*);
+phi-elim(*);
+dce(*);
+
+infer-schedules(*);
+
+gcm(*);

hercules_samples/ccp/build.rs

@@ -4,6 +4,8 @@ fn main() {
     JunoCompiler::new()
         .ir_in_src("ccp.hir")
         .unwrap()
+        .schedule_in_src(if cfg!(feature = "cuda") { "gpu.sch" } else { "cpu.sch" })
+        .unwrap()
         .build()
         .unwrap();
 }

hercules_samples/ccp/src/cpu.sch

+gvn(*);
+phi-elim(*);
+dce(*);
+
+auto-outline(*);
+
+ip-sroa(*);
+sroa(*);
+fork-split(*);
+unforkify(*);
+dce(*);
+float-collections(*);
+gvn(*);
+phi-elim(*);
+dce(*);
+
+infer-schedules(*);
+
+gcm(*);

hercules_samples/ccp/src/gpu.sch

+gvn(*);
+phi-elim(*);
+dce(*);
+
+let out = auto-outline(*);
+gpu(out.tricky);
+
+ip-sroa(*);
+sroa(*);
+dce(*);
+float-collections(*);
+gvn(*);
+phi-elim(*);
+dce(*);
+
+infer-schedules(*);
+
+gcm(*);

hercules_samples/dot/src/gpu.sch

@@ -2,9 +2,8 @@ gvn(*);
 phi-elim(*);
 dce(*);
 
-auto-outline(*);
-gpu(*);
-host(dot);
+let out = auto-outline(*);
+gpu(out.dot);
 
 ip-sroa(*);
 sroa(*);

hercules_samples/fac/build.rs

@@ -4,6 +4,8 @@ fn main() {
     JunoCompiler::new()
         .ir_in_src("fac.hir")
         .unwrap()
+        .schedule_in_src(if cfg!(feature = "cuda") { "gpu.sch" } else { "cpu.sch" })
+        .unwrap()
         .build()
         .unwrap();
 }

hercules_samples/fac/src/cpu.sch

+gvn(*);
+phi-elim(*);
+dce(*);
+
+auto-outline(*);
+
+ip-sroa(*);
+sroa(*);
+fork-split(*);
+unforkify(*);
+dce(*);
+float-collections(*);
+gvn(*);
+phi-elim(*);
+dce(*);
+
+infer-schedules(*);
+
+gcm(*);

hercules_samples/fac/src/gpu.sch

+gvn(*);
+phi-elim(*);
+dce(*);
+
+let out = auto-outline(*);
+gpu(out.fac);
+
+ip-sroa(*);
+sroa(*);
+dce(*);
+float-collections(*);
+gvn(*);
+phi-elim(*);
+dce(*);
+
+infer-schedules(*);
+
+gcm(*);

hercules_samples/matmul/src/gpu.sch

@@ -2,9 +2,8 @@ gvn(*);
 phi-elim(*);
 dce(*);
 
-auto-outline(*);
-gpu(*);
-host(matmul);
+let out = auto-outline(*);
+gpu(out.matmul);
 
 ip-sroa(*);
 sroa(*);

hercules_samples/matmul/src/main.rs

@@ -33,15 +33,13 @@ fn main() {
         }
         #[cfg(feature = "cuda")]
         {
-            let a_box = CUDABox::from_cpu_ref(HerculesCPURef::from_slice(&mut a));
-            let a = a_box.get_ref();
-            let b_box = CUDABox::from_cpu_ref(HerculesCPURef::from_slice(&mut b));
-            let b = b_box.get_ref();
+            let a = CUDABox::from_cpu_ref(HerculesCPURef::from_slice(&mut a));
+            let b = CUDABox::from_cpu_ref(HerculesCPURef::from_slice(&mut b));
             let mut r = runner!(matmul);
-            let c = r.run(I as u64, J as u64, K as u64, a, b).await;
+            let c = r.run(I as u64, J as u64, K as u64, a.get_ref(), b.get_ref()).await;
             let mut c_cpu: Box<[i32]> = vec![0; correct_c.len()].into_boxed_slice();
             c.to_cpu_ref(&mut c_cpu);
-            assert_eq!(c_cpu.as_ref(), &*correct_c);
+            assert_eq!(&*c_cpu, &*correct_c);
         }
     });
 }

juno_samples/antideps/build.rs

 use juno_build::JunoCompiler;
 
 fn main() {
+    #[cfg(feature = "cuda")]
+    JunoCompiler::new()
+        .file_in_src("antideps.jn")
+        .unwrap()
+        .schedule_in_src("gpu.sch")
+        .unwrap()
+        .build()
+        .unwrap();
+    #[cfg(not(feature = "cuda"))]
     JunoCompiler::new()
         .file_in_src("antideps.jn")
         .unwrap()

juno_samples/antideps/src/gpu.sch

+gvn(*);
+phi-elim(*);
+dce(*);
+
+let out = auto-outline(*);
+gpu(out.simple_antideps, out.loop_antideps, out.complex_antideps1, out.complex_antideps2, out.very_complex_antideps, out.read_chains, out.array_of_structs);
+
+ip-sroa(*);
+sroa(*);
+dce(*);
+float-collections(*);
+gvn(*);
+phi-elim(*);
+dce(*);
+
+infer-schedules(*);
+
+gcm(*);

juno_samples/casts_and_intrinsics/build.rs

 use juno_build::JunoCompiler;
 
 fn main() {
+    #[cfg(feature = "cuda")]
+    JunoCompiler::new()
+        .file_in_src("casts_and_intrinsics.jn")
+        .unwrap()
+        .schedule_in_src("gpu.sch")
+        .unwrap()
+        .build()
+        .unwrap();
+    #[cfg(not(feature = "cuda"))]
     JunoCompiler::new()
         .file_in_src("casts_and_intrinsics.jn")
         .unwrap()

juno_samples/casts_and_intrinsics/src/gpu.sch

+gvn(*);
+phi-elim(*);
+dce(*);
+
+let out = auto-outline(*);
+gpu(out.casts_and_intrinsics);
+
+ip-sroa(*);
+sroa(*);
+dce(*);
+float-collections(*);
+gvn(*);
+phi-elim(*);
+dce(*);
+
+infer-schedules(*);
+
+gcm(*);

juno_samples/cava/build.rs

@@ -2,6 +2,15 @@ extern crate juno_build;
 use juno_build::JunoCompiler;
 
 fn main() {
+    #[cfg(feature = "cuda")]
+    JunoCompiler::new()
+        .file_in_src("cava.jn")
+        .unwrap()
+        .schedule_in_src("gpu.sch")
+        .unwrap()
+        .build()
+        .unwrap();
+    #[cfg(not(feature = "cuda"))]
     JunoCompiler::new()
         .file_in_src("cava.jn")
         .unwrap()

juno_samples/cava/src/gpu.sch

+gvn(*);
+phi-elim(*);
+dce(*);
+
+gpu(scale, demosaic, medianMatrix, transform, gamut, tone_map, descale);
+
+ip-sroa(*);
+sroa(*);
+
+dce(*);
+float-collections(*);
+gvn(*);
+phi-elim(*);
+dce(*);
+
+infer-schedules(*);
+
+gcm(*);

juno_samples/concat/build.rs

@@ -4,6 +4,8 @@ fn main() {
     JunoCompiler::new()
         .file_in_src("concat.jn")
         .unwrap()
+        .schedule_in_src(if cfg!(feature = "cuda") { "gpu.sch" } else { "cpu.sch" })
+        .unwrap()
         .build()
         .unwrap();
 }

juno_samples/concat/src/concat.jn

@@ -18,15 +18,7 @@ fn sum<t : number, c : usize>(arr : t[c]) -> t {
 }
 
 #[entry]
-fn concat_entry(a : i32) -> i32 {
-  let arr1 : i32[3];
-  let arr2 : i32[6];
-  arr1[0] = a;
-  arr1[1] = a;
-  arr2[0] = a;
-  arr2[1] = a;
-  arr2[4] = a;
-  arr2[5] = a;
+fn concat_entry(arr1 : i32[3], arr2 : i32[6]) -> i32 {
   let arr3 = concat::<i32, 3, 6>(arr1, arr2);
   return sum::<i32, 9>(arr3);
 }