__builtin_bswap16, 32 and 64

4 files changed, 38 insertions, 39 deletions

diff --git a/cparser/Machine.ml b/cparser/Machine.ml
index 658cf0f4..ac34fa5f 100644
--- a/cparser/Machine.ml
+++ b/cparser/Machine.ml
@@ -254,7 +254,7 @@ let mppa_k1c =
     sizeof_wchar = 4;
     sizeof_size_t = 8;
     sizeof_ptrdiff_t = 8;
-    sizeof_intreg = 4; (* What is this for ? *)
+    sizeof_intreg = 8; (* What is this for ? *)
     alignof_ptr = 8;
     alignof_short = 2;
     alignof_int = 4;
diff --git a/mppa_k1c/Asmexpand.ml b/mppa_k1c/Asmexpand.ml
index 1e5149fd..9c256bd0 100644
--- a/mppa_k1c/Asmexpand.ml
+++ b/mppa_k1c/Asmexpand.ml
@@ -345,34 +345,32 @@ let expand_int64_arith conflict rl fn = assert false
 (* Byte swaps.  There are no specific instructions, so we use standard,
    not-very-efficient formulas. *)
 
-let expand_bswap16 d s = assert false
+let expand_bswap16 d s = let open Asmvliw in
   (* d = (s & 0xFF) << 8 | (s >> 8) & 0xFF *)
-(*emit (Pandiw(X31, X s, coqint_of_camlint 0xFFl));
-  emit (Pslliw(X31, X X31, _8));
-  emit (Psrliw(d, X s, _8));
-  emit (Pandiw(d, X d, coqint_of_camlint 0xFFl));
-  emit (Porw(d, X X31, X d))
-*)
+  emit (Pandiw(GPR32, s, coqint_of_camlint 0xFFl)); emit Psemi;
+  emit (Pslliw(GPR32, GPR32, _8)); emit Psemi;
+  emit (Psrliw(d, s, _8)); emit Psemi;
+  emit (Pandiw(d, d, coqint_of_camlint 0xFFl));
+  emit (Porw(d, GPR32, d)); emit Psemi
 
-let expand_bswap32 d s = assert false
+let expand_bswap32 d s = let open Asmvliw in
   (* d = (s << 24)
        | (((s >> 8) & 0xFF) << 16)
        | (((s >> 16) & 0xFF) << 8)
        | (s >> 24)  *)
-(*emit (Pslliw(X1, X s, coqint_of_camlint 24l));
-  emit (Psrliw(X31, X s, _8));
-  emit (Pandiw(X31, X X31, coqint_of_camlint 0xFFl));
-  emit (Pslliw(X31, X X31, _16));
-  emit (Porw(X1, X X1, X X31));
-  emit (Psrliw(X31, X s, _16));
-  emit (Pandiw(X31, X X31, coqint_of_camlint 0xFFl));
-  emit (Pslliw(X31, X X31, _8));
-  emit (Porw(X1, X X1, X X31));
-  emit (Psrliw(X31, X s, coqint_of_camlint 24l));
-  emit (Porw(d, X X1, X X31))
-*)
-
-let expand_bswap64 d s = assert false
+  emit (Pslliw(GPR16, s, coqint_of_camlint 24l)); emit Psemi;
+  emit (Psrliw(GPR32, s, _8)); emit Psemi;
+  emit (Pandiw(GPR32, GPR32, coqint_of_camlint 0xFFl)); emit Psemi;
+  emit (Pslliw(GPR32, GPR32, _16)); emit Psemi;
+  emit (Porw(GPR16, GPR16, GPR31)); emit Psemi;
+  emit (Psrliw(GPR32, s, _16)); emit Psemi;
+  emit (Pandiw(GPR32, GPR32, coqint_of_camlint 0xFFl)); emit Psemi;
+  emit (Pslliw(GPR32, GPR32, _8)); emit Psemi;
+  emit (Porw(GPR16, GPR16, GPR32)); emit Psemi;
+  emit (Psrliw(GPR32, s, coqint_of_camlint 24l)); emit Psemi;
+  emit (Porw(d, GPR16, GPR32)); emit Psemi
+
+let expand_bswap64 d s = let open Asmvliw in
   (* d = s << 56
          | (((s >> 8) & 0xFF) << 48)
          | (((s >> 16) & 0xFF) << 40)
@@ -381,17 +379,16 @@ let expand_bswap64 d s = assert false
          | (((s >> 40) & 0xFF) << 16)
          | (((s >> 48) & 0xFF) << 8)
          | s >> 56 *)
-(*emit (Psllil(X1, X s, coqint_of_camlint 56l));
+  emit (Psllil(GPR16, s, coqint_of_camlint 56l)); emit Psemi;
   List.iter
     (fun (n1, n2) ->
-      emit (Psrlil(X31, X s, coqint_of_camlint n1));
-      emit (Pandil(X31, X X31, coqint_of_camlint 0xFFl));
-      emit (Psllil(X31, X X31, coqint_of_camlint n2));
-      emit (Porl(X1, X X1, X X31)))
+      emit (Psrlil(GPR32, s, coqint_of_camlint n1)); emit Psemi;
+      emit (Pandil(GPR32, GPR32, coqint_of_camlint 0xFFl)); emit Psemi;
+      emit (Psllil(GPR32, GPR32, coqint_of_camlint n2)); emit Psemi;
+      emit (Porl(GPR16, GPR16, GPR32)); emit Psemi;)
     [(8l,48l); (16l,40l); (24l,32l); (32l,24l); (40l,16l); (48l,8l)];
-  emit (Psrlil(X31, X s, coqint_of_camlint 56l));
-  emit (Porl(d, X X1, X X31))
-*)
+  emit (Psrlil(GPR32, s, coqint_of_camlint 56l)); emit Psemi;
+  emit (Porl(d, GPR16, GPR32)); emit Psemi
 
 (* Handling of compiler-inlined builtins *)
 let last_system_register = 511l
@@ -477,6 +474,12 @@ let expand_builtin_inline name args res = let open Asmvliw in
      emit (Palclrd(res, addr))
   | "__builtin_alclrw", [BA(IR addr)], BR(IR res) ->
      emit (Palclrw(res, addr))
+  | "__builtin_bswap16", [BA(IR a1)], BR(IR res) ->
+     expand_bswap16 res a1
+  | ("__builtin_bswap"| "__builtin_bswap32"), [BA(IR a1)], BR(IR res) ->
+     expand_bswap32 res a1
+  | "__builtin_bswap64", [BA(IR src)], BR(IR res) ->
+     expand_bswap64 res src
 	  
   (* Byte swaps *)
 (*| "__builtin_bswap16", [BA(IR a1)], BR(IR res) ->
diff --git a/mppa_k1c/CBuiltins.ml b/mppa_k1c/CBuiltins.ml
index 235496c5..a91119b1 100644
--- a/mppa_k1c/CBuiltins.ml
+++ b/mppa_k1c/CBuiltins.ml
@@ -96,12 +96,8 @@ let builtins = {
 
     (* Synchronization *)
 (*  "__builtin_fence",
-      (TVoid [], [], false);
-    (* Integer arithmetic *)
-    "__builtin_bswap64",
-      (TInt(IULongLong, []), 
-       [TInt(IULongLong, [])], false);
-    (* Float arithmetic *)
+      (TVoid [], [], false); *)
+(*  (* Float arithmetic *)
     "__builtin_fmadd",
       (TFloat(FDouble, []),
        [TFloat(FDouble, []); TFloat(FDouble, []); TFloat(FDouble, [])], false);
diff --git a/test/regression/Makefile b/test/regression/Makefile
index ab27c85a..c77aa073 100644
--- a/test/regression/Makefile
+++ b/test/regression/Makefile
@@ -23,8 +23,8 @@ TESTS=int32 int64 floats floats-basics \
 TESTS_COMP=attribs1 bitfields1 bitfields2 bitfields3 bitfields4 \
   bitfields5 bitfields6 bitfields7 bitfields8 \
   builtins-$(ARCH) alignas \
-  varargs1 varargs2 varargs3 sections alias aligned
-# FIXME K1C : packedstruct1 packedstruct2
+  varargs1 varargs2 varargs3 sections alias aligned\
+  packedstruct1 packedstruct2
 
 # Can run, both in compiled mode and in interpreter mode,
 # but produce processor-dependent results, so no reference output in Results