Additional optimizations for C compilation

2 files changed, 67 insertions, 1 deletions

diff --git a/lib/sail.c b/lib/sail.c
index 5c83690d..c66c057c 100644
--- a/lib/sail.c
+++ b/lib/sail.c
@@ -680,6 +680,11 @@ void zero_extend(lbits *rop, const lbits op, const sail_int len)
   mpz_set(*rop->bits, *op.bits);
 }
 
+fbits fast_zero_extend(const sbits op, const uint64_t n)
+{
+  return op.bits;
+}
+
 void sign_extend(lbits *rop, const lbits op, const sail_int len)
 {
   assert(op.len <= mpz_get_ui(len));
@@ -694,6 +699,32 @@ void sign_extend(lbits *rop, const lbits op, const sail_int len)
   }
 }
 
+fbits fast_sign_extend(const fbits op, const uint64_t n, const uint64_t m)
+{
+  uint64_t rop = op;
+  if (op & (UINT64_C(1) << (n - 1))) {
+    for (uint64_t i = m - 1; i >= n; i--) {
+      rop = rop | (UINT64_C(1) << i);
+    }
+    return rop;
+  } else {
+    return rop;
+  }
+}
+
+fbits fast_sign_extend2(const sbits op, const uint64_t m)
+{
+  uint64_t rop = op.bits;
+  if (op.bits & (UINT64_C(1) << (op.len - 1))) {
+    for (uint64_t i = m - 1; i >= op.len; i--) {
+      rop = rop | (UINT64_C(1) << i);
+    }
+    return rop;
+  } else {
+    return rop;
+  }
+}
+
 void length_lbits(sail_int *rop, const lbits op)
 {
   mpz_set_ui(*rop, op.len);
@@ -783,12 +814,21 @@ void sail_signed(sail_int *rop, const lbits op)
   }
 }
 
-inline
 mach_int fast_unsigned(const fbits op)
 {
   return (mach_int) op;
 }
 
+mach_int fast_signed(const fbits op, const uint64_t n)
+{
+  if (op & (UINT64_C(1) << (n - 1))) {
+    uint64_t rop = op & ~(UINT64_C(1) << (n - 1));
+    return (mach_int) (rop - (UINT64_C(1) << (n - 1)));
+  } else {
+    return (mach_int) op;
+  }
+}
+
 void append(lbits *rop, const lbits op1, const lbits op2)
 {
   rop->len = op1.len + op2.len;
@@ -911,6 +951,23 @@ void vector_update_subrange_lbits(lbits *rop,
   }
 }
 
+fbits fast_update_subrange(const fbits op,
+			   const mach_int n,
+			   const mach_int m,
+			   const fbits slice)
+{
+  fbits rop = op;
+  for (mach_int i = 0; i < n - (m - UINT64_C(1)); i++) {
+    uint64_t bit = UINT64_C(1) << ((uint64_t) i);
+    if (slice & bit) {
+      rop |= (bit << m);
+    } else {
+      rop &= ~(bit << m);
+    }
+  }
+  return rop;
+}
+
 void slice(lbits *rop, const lbits op, const sail_int start_mpz, const sail_int len_mpz)
 {
   assert(mpz_get_ui(start_mpz) + mpz_get_ui(len_mpz) <= op.len);
diff --git a/lib/sail.h b/lib/sail.h
index 8f113339..d5597a64 100644
--- a/lib/sail.h
+++ b/lib/sail.h
@@ -246,7 +246,10 @@ void mult_vec(lbits *rop, const lbits op1, const lbits op2);
 void zeros(lbits *rop, const sail_int op);
 
 void zero_extend(lbits *rop, const lbits op, const sail_int len);
+fbits fast_zero_extend(const sbits op, const uint64_t n);
 void sign_extend(lbits *rop, const lbits op, const sail_int len);
+fbits fast_sign_extend(const fbits op, const uint64_t n, const uint64_t m);
+fbits fast_sign_extend2(const sbits op, const uint64_t m);
 
 void length_lbits(sail_int *rop, const lbits op);
 
@@ -267,6 +270,7 @@ fbits bitvector_access(const lbits op, const sail_int n_mpz);
 void sail_unsigned(sail_int *rop, const lbits op);
 void sail_signed(sail_int *rop, const lbits op);
 
+mach_int fast_signed(const fbits, const uint64_t);
 mach_int fast_unsigned(const fbits);
 
 void append(lbits *rop, const lbits op1, const lbits op2);
@@ -292,6 +296,11 @@ void vector_update_subrange_lbits(lbits *rop,
 				      const sail_int m_mpz,
 				      const lbits slice);
 
+fbits fast_update_subrange(const fbits op,
+			   const mach_int n,
+			   const mach_int m,
+			   const fbits slice);
+
 void slice(lbits *rop, const lbits op, const sail_int start_mpz, const sail_int len_mpz);
 
 sbits sslice(const fbits op, const mach_int start, const mach_int len);