diff --git a/vvp/opcodes.txt b/vvp/opcodes.txt
index 2c18aabbd..db870f20a 100644
--- a/vvp/opcodes.txt
+++ b/vvp/opcodes.txt
@@ -417,7 +417,7 @@ These instructions are like the %ix/get instructions, except that they
 read directly from a functor label instead of from thread bits. They
 set bit 4 just like %ix/get.
 
-* %ix/load <idx>, <low> <high>
+* %ix/load <idx>, <low>, <high>
 
 This instruction loads an immediate value into the addressed index
 register. The index register holds signed 64 bit numeric values, so
@@ -428,9 +428,9 @@ vector. The values are unsigned decimal values and are combined as
 <high> << 32 | <low> to produce the final value.
 
 
-* %ix/add <idx>, <low> <high>
-* %ix/sub <idx>, <low> <high>
-* %ix/mul <idx>, <low> <high>
+* %ix/add <idx>, <low>, <high>
+* %ix/sub <idx>, <low>, <high>
+* %ix/mul <idx>, <low>, <high>
 
 These instructions add, subtract, or multiply the selected index
 register by the immediate value. The 64 bit immediate value is built
@@ -758,6 +758,8 @@ bits) by the amount in index register 0. The <bit> is the address of
 the LSB of the vector, and <wid> the width of the vector. The shift is
 done in place. Zero values are shifted in.
 
+For a negative shift the value is padded with 'bx.
+
 * %shiftr/i0 <bit>, <wid>
 * %shiftr/s/i0 <bit>, <wid>
 
@@ -769,6 +771,8 @@ shift is done in place.
 %shiftr/i0 is an unsigned down shift, so zeros are shifted into the
 top bits. %shiftr/s/i0 is a signed shift, so the value is sign-extended.
 
+For a negative shift %shiftr/i0 will pad the value with 'bx.
+
 * %sub <bit-l>, <bit-r>, <wid>
 
 This instruction arithmetically subtracts the right vector out of the
diff --git a/vvp/vthread.cc b/vvp/vthread.cc
index 1b87cbc22..9a5fbe5b7 100644
--- a/vvp/vthread.cc
+++ b/vvp/vthread.cc
@@ -2642,7 +2642,7 @@ bool of_IX_GET_S(vthread_t thr, vvp_code_t cp)
       unsigned base  = cp->bit_idx[1];
       unsigned width = cp->number;
 
-      unsigned long v = 0;
+      uint64_t v = 0;
       bool unknown_flag = false;
 
       vvp_bit4_t vv = BIT4_0;
@@ -2654,7 +2654,7 @@ bool of_IX_GET_S(vthread_t thr, vvp_code_t cp)
 		  break;
 	    }
 
-	    v |= (unsigned long) vv << i;
+	    v |= (uint64_t) vv << i;
 
 	    if (base >= 4)
 		  base += 1;
@@ -2662,7 +2662,7 @@ bool of_IX_GET_S(vthread_t thr, vvp_code_t cp)
 
 	/* Sign-extend to fill the integer value. */
       if (!unknown_flag) {
-	    unsigned long pad = vv;
+	    uint64_t pad = vv;
 	    for (unsigned i = width ; i < 8*sizeof(v) ;  i += 1) {
 		  v |= pad << i;
 	    }
@@ -4206,7 +4206,7 @@ bool of_SHIFTL_I0(vthread_t thr, vvp_code_t cp)
 {
       unsigned base = cp->bit_idx[0];
       unsigned wid = cp->number;
-      unsigned long shift = thr->words[0].w_int;
+      long shift = thr->words[0].w_int;
 
       assert(base >= 4);
       thr_check_addr(thr, base+wid-1);
@@ -4215,7 +4215,8 @@ bool of_SHIFTL_I0(vthread_t thr, vvp_code_t cp)
 	    // The result is 'bx if the shift amount is undefined.
 	    vvp_vector4_t tmp (wid, BIT4_X);
 	    thr->bits4.set_vec(base, tmp);
-      } else if (shift >= wid) {
+
+      } else if (shift >= (long)wid) {
 	      // Shift is so far that all value is shifted out. Write
 	      // in a constant 0 result.
 	    vvp_vector4_t tmp (wid, BIT4_0);
@@ -4228,6 +4229,17 @@ bool of_SHIFTL_I0(vthread_t thr, vvp_code_t cp)
 	      // Fill zeros on the bottom
 	    vvp_vector4_t fil (shift, BIT4_0);
 	    thr->bits4.set_vec(base, fil);
+
+      } else if (shift < 0) {
+	      // For a negative shift we pad with 'bx.
+	    unsigned idx;
+	    for (idx = 0 ;  (idx-shift) < wid ;  idx += 1) {
+		  unsigned src = base + idx - shift;
+		  unsigned dst = base + idx;
+		  thr_put_bit(thr, dst, thr_get_bit(thr, src));
+	    }
+	    for ( ;  idx < wid ;  idx += 1)
+		  thr_put_bit(thr, base+idx, BIT4_X);
       }
       return true;
 }
@@ -4244,12 +4256,16 @@ bool of_SHIFTR_I0(vthread_t thr, vvp_code_t cp)
 {
       unsigned base = cp->bit_idx[0];
       unsigned wid = cp->number;
-      unsigned long shift = thr->words[0].w_int;
+      long shift = thr->words[0].w_int;
+
+      assert(base >= 4);
+      thr_check_addr(thr, base+wid-1);
 
       if (thr_get_bit(thr, 4) == BIT4_1) {
-	    // The result is 'bx if the shift amount is undefined.
+	      // The result is 'bx if the shift amount is undefined.
 	    vvp_vector4_t tmp (wid, BIT4_X);
 	    thr->bits4.set_vec(base, tmp);
+
       } else if (shift > 0) {
 	    unsigned idx;
 	    for (idx = 0 ;  (idx+shift) < wid ;  idx += 1) {
@@ -4259,6 +4275,21 @@ bool of_SHIFTR_I0(vthread_t thr, vvp_code_t cp)
 	    }
 	    for ( ;  idx < wid ;  idx += 1)
 		  thr_put_bit(thr, base+idx, BIT4_0);
+
+      } else if (shift < -(long)wid) {
+	      // Negative shift is so far that all the value is shifted out.
+	      // Write in a constant 'bx result.
+	    vvp_vector4_t tmp (wid, BIT4_X);
+	    thr->bits4.set_vec(base, tmp);
+
+      } else if (shift < 0) {
+
+	      // For a negative shift we pad with 'bx.
+	    vvp_vector4_t tmp (thr->bits4, base, wid+shift);
+	    thr->bits4.set_vec(base-shift, tmp);
+
+	    vvp_vector4_t fil (-shift, BIT4_X);
+	    thr->bits4.set_vec(base, fil);
       }
       return true;
 }
@@ -4271,7 +4302,7 @@ bool of_SHIFTR_S_I0(vthread_t thr, vvp_code_t cp)
       vvp_bit4_t sign = thr_get_bit(thr, base+wid-1);
 
       if (thr_get_bit(thr, 4) == BIT4_1) {
-	    // The result is 'bx if the shift amount is undefined.
+	      // The result is 'bx if the shift amount is undefined.
 	    vvp_vector4_t tmp (wid, BIT4_X);
 	    thr->bits4.set_vec(base, tmp);
       } else if (shift >= wid) {