gpu-compute: Added register file cache support (#730)

The RFC is defaulted to a size of 0 which removes it completely. To use
the RFC set the --register-file-cache-size to a non-zero multiple of
two. In addition, rfc_pipe_length may be altered to increase or decrease
RFC latency benefit.

configs/example/apu_se.py

@@ -335,6 +335,12 @@ parser.add_argument(
     default="dynamic",
     help="register allocation policy (simple/dynamic)",
 )
+parser.add_argument(
+    "--register-file-cache-size",
+    type=int,
+    default=0,
+    help="number of registers in cache",
+)
 
 parser.add_argument(
     "--dgpu",
@@ -493,6 +499,7 @@ for i in range(n_cu):
     vrfs = []
     vrf_pool_mgrs = []
     srfs = []
+    rfcs = []
     srf_pool_mgrs = []
     for j in range(args.simds_per_cu):
         for k in range(shader.n_wf):
@@ -537,10 +544,16 @@ for i in range(n_cu):
                 simd_id=j, wf_size=args.wf_size, num_regs=args.sreg_file_size
             )
         )
+        rfcs.append(
+            RegisterFileCache(
+                simd_id=j, cache_size=args.register_file_cache_size
+            )
+        )
 
     compute_units[-1].wavefronts = wavefronts
     compute_units[-1].vector_register_file = vrfs
     compute_units[-1].scalar_register_file = srfs
+    compute_units[-1].register_file_cache = rfcs
     compute_units[-1].register_manager = RegisterManager(
         policy=args.registerManagerPolicy,
         vrf_pool_managers=vrf_pool_mgrs,

configs/example/gpufs/amd/AmdGPUOptions.py

@@ -247,3 +247,9 @@ def addAmdGPUOptions(parser):
         default="simple",
         help="register allocation policy (simple/dynamic)",
     )
+    parser.add_argument(
+        "--register-file-cache-size",
+        type=int,
+        default=0,
+        help="number of registers in cache",
+    )

configs/example/gpufs/system/amdgpu.py

@@ -84,6 +84,7 @@ def createGPU(system, args):
         vrfs = []
         vrf_pool_mgrs = []
         srfs = []
+        rfcs = []
         srf_pool_mgrs = []
         for j in range(args.simds_per_cu):
             for k in range(shader.n_wf):
@@ -133,10 +134,16 @@ def createGPU(system, args):
                     num_regs=args.sreg_file_size,
                 )
             )
+            rfcs.append(
+                RegisterFileCache(
+                    simd_id=j, cache_size=args.register_file_cache_size
+                )
+            )
 
         compute_units[-1].wavefronts = wavefronts
         compute_units[-1].vector_register_file = vrfs
         compute_units[-1].scalar_register_file = srfs
+        compute_units[-1].register_file_cache = rfcs
         compute_units[-1].register_manager = RegisterManager(
             policy=args.registerManagerPolicy,
             vrf_pool_managers=vrf_pool_mgrs,