Implement the extswsli instruction

This mainly required the addition of an entry to the opcode 31 decode
table and a 32-bit sign-extender in the rotator.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>

decode1.vhdl

@@ -185,7 +185,8 @@ architecture behaviour of decode1 is
 		2#1110111010#  =>       (ALU,    OP_EXTS,      NONE,       NONE,        RS,   RA,   '0', '0', '0', '0', ZERO, '0', is1B, '0', '0', '0', '0', '0', '0', RC,   '0', '0'), -- extsb
 		2#1110011010#  =>       (ALU,    OP_EXTS,      NONE,       NONE,        RS,   RA,   '0', '0', '0', '0', ZERO, '0', is2B, '0', '0', '0', '0', '0', '0', RC,   '0', '0'), -- extsh
 		2#1111011010#  =>       (ALU,    OP_EXTS,      NONE,       NONE,        RS,   RA,   '0', '0', '0', '0', ZERO, '0', is4B, '0', '0', '0', '0', '0', '0', RC,   '0', '0'), -- extsw
-		-- 2#110111101-# extswsli
+                2#1101111010#  =>       (ALU,    OP_EXTSWSLI,  NONE,       CONST_SH,    RS,   RA,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC,   '0', '0'), -- extswsli
+                2#1101111011#  =>       (ALU,    OP_EXTSWSLI,  NONE,       CONST_SH,    RS,   RA,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC,   '0', '0'), -- extswsli
 		2#1111010110#  =>       (ALU,    OP_ICBI,      NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '1'), -- icbi
 		2#0000010110#  =>       (ALU,    OP_NOP,       NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '1'), -- icbt
 		2#0000001111#  =>       (ALU,    OP_ISEL,      RA_OR_ZERO, RB,          NONE, RT,   '1', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '1'), -- isel

execute1.vhdl

@@ -63,6 +63,7 @@ architecture behaviour of execute1 is
     signal ctrl: ctrl_t := (irq_state => WRITE_SRR0, others => (others => '0'));
     signal ctrl_tmp: ctrl_t := (irq_state => WRITE_SRR0, others => (others => '0'));
     signal right_shift, rot_clear_left, rot_clear_right: std_ulogic;
+    signal rot_sign_ext: std_ulogic;
     signal rotator_result: std_ulogic_vector(63 downto 0);
     signal rotator_carry: std_ulogic;
     signal logical_result: std_ulogic_vector(63 downto 0);
@@ -174,6 +175,7 @@ begin
 	    arith => e_in.is_signed,
 	    clear_left => rot_clear_left,
 	    clear_right => rot_clear_right,
+            sign_ext_rs => rot_sign_ext,
 	    result => rotator_result,
 	    carry_out => rotator_carry
 	    );
@@ -429,6 +431,7 @@ begin
 	right_shift <= '1' when e_in.insn_type = OP_SHR else '0';
 	rot_clear_left <= '1' when e_in.insn_type = OP_RLC or e_in.insn_type = OP_RLCL else '0';
 	rot_clear_right <= '1' when e_in.insn_type = OP_RLC or e_in.insn_type = OP_RLCR else '0';
+        rot_sign_ext <= '1' when e_in.insn_type = OP_EXTSWSLI else '0';
 
 	ctrl_tmp.irq_state <= WRITE_SRR0;
 	exception := '0';
@@ -828,7 +831,7 @@ begin
 	    when OP_PRTY =>
 		result := parity_result;
 		result_en := '1';
-	    when OP_RLC | OP_RLCL | OP_RLCR | OP_SHL | OP_SHR =>
+	    when OP_RLC | OP_RLCL | OP_RLCR | OP_SHL | OP_SHR | OP_EXTSWSLI =>
 		result := rotator_result;
 		if e_in.output_carry = '1' then
 		    set_carry(v.e, rotator_carry, rotator_carry);

rotator.vhdl

@@ -15,6 +15,7 @@ entity rotator is
 	  arith: in std_ulogic;
 	  clear_left: in std_ulogic;
 	  clear_right: in std_ulogic;
+          sign_ext_rs: in std_ulogic;
 	  result: out std_ulogic_vector(63 downto 0);
 	  carry_out: out std_ulogic
       );
@@ -57,13 +58,18 @@ architecture behaviour of rotator is
 
 begin
     rotator_0: process(all)
+        variable hi32: std_ulogic_vector(31 downto 0);
     begin
 	-- First replicate bottom 32 bits to both halves if 32-bit
 	if is_32bit = '1' then
-	    repl32 <= rs(31 downto 0) & rs(31 downto 0);
+            hi32 := rs(31 downto 0);
-	else
+	elsif sign_ext_rs = '1' then
-	    repl32 <= rs;
+            -- sign extend bottom 32 bits
+            hi32 := (others => rs(31));
+        else
+	    hi32 := rs(63 downto 32);
 	end if;
+        repl32 <= hi32 & rs(31 downto 0);
 
 	-- Negate shift count for right shifts
 	if right_shift = '1' then

rotator_tb.vhdl

@@ -19,6 +19,7 @@ architecture behave of rotator_tb is
     signal is_32bit, right_shift, arith, clear_left, clear_right: std_ulogic := '0';
     signal result: std_ulogic_vector(63 downto 0);
     signal carry_out: std_ulogic;
+    signal extsw: std_ulogic;
 
 begin
     rotator_0: entity work.rotator
@@ -32,6 +33,7 @@ begin
 	    arith => arith,
 	    clear_left => clear_left,
 	    clear_right => clear_right,
+            sign_ext_rs => extsw,
 	    result => result,
 	    carry_out => carry_out
 	);
@@ -48,6 +50,7 @@ begin
 	arith <= '0';
 	clear_left <= '1';
 	clear_right <= '1';
+        extsw <= '0';
         rlwnm_loop : for i in 0 to 1000 loop
 	    rs <= pseudorand(64);
 	    shift <= pseudorand(7);
@@ -263,6 +266,31 @@ begin
 		report "bad srad expected " & to_hstring(behave_ca_ra) & " got " & to_hstring(carry_out & result);
         end loop;
 
+        -- extswsli
+        report "test extswsli";
+        ra <= (others => '0');
+	is_32bit <= '0';
+	right_shift <= '0';
+	arith <= '0';
+	clear_left <= '0';
+	clear_right <= '0';
+        extsw <= '1';
+        extswsli_loop : for i in 0 to 1000 loop
+	    rs <= pseudorand(64);
+	    shift <= '0' & pseudorand(6);
+	    wait for clk_period;
+	    behave_ra := rs;
+            behave_ra(63 downto 32) := (others => rs(31));
+            behave_ra := std_ulogic_vector(shift_left(unsigned(behave_ra),
+                                                      to_integer(unsigned(shift))));
+	    --report "rs = " & to_hstring(rs);
+	    --report "ra = " & to_hstring(ra);
+	    --report "shift = " & to_hstring(shift);
+	    --report "result = " & to_hstring(carry_out & result);
+	    assert behave_ra = result
+		report "bad extswsli expected " & to_hstring(behave_ra) & " got " & to_hstring(result);
+        end loop;
+
         assert false report "end of test" severity failure;
         wait;
     end process;