Memory: Fix bug in which swp_oldval does not get latched (oops!). Make align_rddata...
[firearm.git] / Execute.v
index 06912a2..c1049ba 100644 (file)
--- a/Execute.v
+++ b/Execute.v
@@ -19,11 +19,15 @@ module Execute(
        output reg outbubble = 1,
        output reg [31:0] outcpsr = 0,
        output reg [31:0] outspsr = 0,
+       output reg outcpsrup = 0,
        output reg write_reg = 1'bx,
        output reg [3:0] write_num = 4'bxxxx,
        output reg [31:0] write_data = 32'hxxxxxxxx,
+       output reg [31:0] jmppc,
+       output reg jmp,
        output reg [31:0] outpc,
-       output reg outflush
+       output reg [31:0] outinsn,
+       output reg [31:0] outop0, outop1, outop2
        );
        
        reg mult_start;
@@ -39,10 +43,12 @@ module Execute(
        
        reg next_outbubble;
        reg [31:0] next_outcpsr, next_outspsr;
+       reg next_outcpsrup;
        reg next_write_reg;
        reg [3:0] next_write_num;
+
        reg [31:0] next_write_data;
-       
+
        Multiplier multiplier(
                .clk(clk), .Nrst(Nrst),
                .start(mult_start), .acc0(mult_acc0), .in0(mult_in0),
@@ -53,7 +59,7 @@ module Execute(
                .in0(alu_in0), .in1(alu_in1), .cpsr(cpsr), .op(alu_op),
                .setflags(alu_setflags), .shifter_carry(carry),
                .result(alu_result), .cpsr_out(alu_outcpsr), .setres(alu_setres));
-       
+
        always @(posedge clk)
        begin
                if (!stall)
@@ -61,55 +67,67 @@ module Execute(
                        outbubble <= next_outbubble;
                        outcpsr <= next_outcpsr;
                        outspsr <= next_outspsr;
+                       outcpsrup <= next_outcpsrup;
                        write_reg <= next_write_reg;
                        write_num <= next_write_num;
                        write_data <= next_write_data;
+                       outpc <= pc;
+                       outinsn <= insn;
+                       outop0 <= op0;
+                       outop1 <= op1;
+                       outop2 <= op2;
                end
        end
+       
+       reg delayedflush = 0;
+       always @(posedge clk)
+               if (flush && outstall /* halp! I can't do it now, maybe later? */)
+                       delayedflush <= 1;
+               else if (!outstall /* anything has been handled this time around */)
+                       delayedflush <= 0;
 
        reg prevstall = 0;
        always @(posedge clk)
                prevstall <= outstall;
-
+       
        always @(*)
        begin
                outstall = stall;
-               next_outbubble = inbubble;
+               
+               casez (insn)
+               `DECODE_ALU_MULT:       /* Multiply -- must come before ALU, because it pattern matches a specific case of ALU */
+                       outstall = outstall | ((!prevstall | !mult_done) && !inbubble);
+               endcase
+       end
+       
+       /* ALU inputs */
+       always @(*)
+       begin
+               alu_in0 = op0;
+               alu_in1 = op1;
+               alu_op = insn[24:21];
+               alu_setflags = insn[20] /* S */;
+       end
+       
+       /* Register outputs */
+       always @(*)
+       begin
                next_outcpsr = cpsr;
                next_outspsr = spsr;
+               next_outcpsrup = 0;
                next_write_reg = 0;
                next_write_num = 4'hx;
                next_write_data = 32'hxxxxxxxx;
-       
-               mult_start = 0;
-               mult_acc0 = 32'hxxxxxxxx;
-               mult_in0 = 32'hxxxxxxxx;
-               mult_in1 = 32'hxxxxxxxx;
-       
-               alu_in0 = 32'hxxxxxxxx;
-               alu_in1 = 32'hxxxxxxxx;
-               alu_op = 4'hx;  /* hax! */
-               alu_setflags = 1'bx;
                
-               casez (insn)
+               casez(insn)
                `DECODE_ALU_MULT:       /* Multiply -- must come before ALU, because it pattern matches a specific case of ALU */
                begin
-                       if (!prevstall && !inbubble)
-                       begin
-                               mult_start = 1;
-                               mult_acc0 = insn[21] /* A */ ? op0 /* Rn */ : 32'h0;
-                               mult_in0 = op1 /* Rm */;
-                               mult_in1 = op2 /* Rs */;
-                               $display("New MUL instruction");
-                       end
-                       outstall = stall | ((!prevstall | !mult_done) && !inbubble);
-                       next_outbubble = inbubble | !mult_done | !prevstall;
                        next_outcpsr = insn[20] /* S */ ? {mult_result[31] /* N */, mult_result == 0 /* Z */, 1'b0 /* C */, cpsr[28] /* V */, cpsr[27:0]} : cpsr;
+                       next_outcpsrup = insn[20] /* S */;
                        next_write_reg = 1;
                        next_write_num = insn[19:16] /* Rd -- why the fuck isn't this the same place as ALU */;
                        next_write_data = mult_result;
                end
-//             `DECODE_ALU_MUL_LONG,   /* Multiply long */
                `DECODE_ALU_MRS:        /* MRS (Transfer PSR to register) */
                begin
                        next_write_reg = 1;
@@ -121,6 +139,7 @@ module Execute(
                end
                `DECODE_ALU_MSR,        /* MSR (Transfer register to PSR) */
                `DECODE_ALU_MSR_FLAGS:  /* MSR (Transfer register or immediate to PSR, flag bits only) */
+               begin
                        if ((cpsr[4:0] == `MODE_USR) || (insn[16] /* that random bit */ == 1'b0))       /* flags only */
                        begin
                                if (insn[22] /* Ps */)
@@ -133,6 +152,8 @@ module Execute(
                                else
                                        next_outcpsr = op0;
                        end
+                       next_outcpsrup = 1;
+               end
                `DECODE_ALU_SWP,        /* Atomic swap */
                `DECODE_ALU_BX,         /* Branch */
                `DECODE_ALU_HDATA_REG,  /* Halfword transfer - register offset */
@@ -140,31 +161,84 @@ module Execute(
                begin end
                `DECODE_ALU:            /* ALU */
                begin
-                       alu_in0 = op0;
-                       alu_in1 = op1;
-                       alu_op = insn[24:21];
-                       alu_setflags = insn[20] /* S */;
-                       
                        if (alu_setres) begin
                                next_write_reg = 1;
                                next_write_num = insn[15:12] /* Rd */;
                                next_write_data = alu_result;
                        end
                        
-                       next_outcpsr = ((insn[15:12] == 4'b1111) && insn[20]) ? spsr : alu_outcpsr;
+                       if (insn[20] /* S */) begin
+                               next_outcpsrup = 1;
+                               next_outcpsr = ((insn[15:12] == 4'b1111) && insn[20]) ? spsr : alu_outcpsr;
+                       end
                end
                `DECODE_LDRSTR_UNDEFINED,       /* Undefined. I hate ARM */
                `DECODE_LDRSTR,         /* Single data transfer */
                `DECODE_LDMSTM:         /* Block data transfer */
                begin end
-               `DECODE_BRANCH:
+               `DECODE_BRANCH:         /* Branch */
                begin
-                       outpc = pc + op0;
-                       if(insn[24]) begin
+                       if(insn[24] /* L */) begin
                                next_write_reg = 1;
                                next_write_num = 4'hE; /* link register */
                                next_write_data = pc + 32'h4;
                        end
+               end
+               endcase
+       end
+       
+       /* Multiplier inputs */
+       always @(*)
+       begin
+               mult_start = 0;
+               mult_acc0 = 32'hxxxxxxxx;
+               mult_in0 = 32'hxxxxxxxx;
+               mult_in1 = 32'hxxxxxxxx;
+               
+               casez(insn)
+               `DECODE_ALU_MULT:
+               begin
+                       if (!prevstall /* i.e., this is a new one */ && !inbubble /* i.e., this is a real one */)
+                       begin
+                               mult_start = 1;
+                               mult_acc0 = insn[21] /* A */ ? op0 /* Rn */ : 32'h0;
+                               mult_in0 = op1 /* Rm */;
+                               mult_in1 = op2 /* Rs */;
+                               $display("New MUL instruction");
+                       end
+               end
+               endcase
+       end
+
+       /* Miscellaneous cleanup. */
+       always @(*)
+       begin
+               next_outbubble = inbubble | flush | delayedflush;
+
+               jmp = 1'b0;
+               jmppc = 32'h00000000;
+
+               casez (insn)
+               `DECODE_ALU_MULT:       /* Multiply -- must come before ALU, because it pattern matches a specific case of ALU */
+                       next_outbubble = next_outbubble | !mult_done | !prevstall;
+               `DECODE_ALU_MRS,        /* MRS (Transfer PSR to register) */
+               `DECODE_ALU_MSR,        /* MSR (Transfer register to PSR) */
+               `DECODE_ALU_MSR_FLAGS,  /* MSR (Transfer register or immediate to PSR, flag bits only) */
+               `DECODE_ALU_SWP,        /* Atomic swap */
+               `DECODE_ALU_BX,         /* Branch */
+               `DECODE_ALU_HDATA_REG,  /* Halfword transfer - register offset */
+               `DECODE_ALU_HDATA_IMM,  /* Halfword transfer - immediate offset */
+               `DECODE_ALU,            /* ALU */
+               `DECODE_LDRSTR_UNDEFINED,       /* Undefined. I hate ARM */
+               `DECODE_LDRSTR,         /* Single data transfer */
+               `DECODE_LDMSTM:         /* Block data transfer */
+               begin end
+               `DECODE_BRANCH:
+               begin
+                       if(!inbubble && !flush && !delayedflush && !outstall /* Let someone else take precedence. */) begin
+                               jmppc = pc + op0 + 32'h8;
+                               jmp = 1'b1;
+                       end
                end                     /* Branch */
                `DECODE_LDCSTC,         /* Coprocessor data transfer */
                `DECODE_CDP,            /* Coprocessor data op */
@@ -229,14 +303,14 @@ module ALU(
        output reg [31:0] cpsr_out,
        output reg setres
 );
-       wire [31:0] res;
-       wire flag_n, flag_z, flag_c, flag_v, setres;
+       reg [31:0] res;
+       reg flag_n, flag_z, flag_c, flag_v;
        wire [32:0] sum, diff, rdiff;
        wire sum_v, diff_v, rdiff_v;
 
        assign sum = {1'b0, in0} + {1'b0, in1};
        assign diff = {1'b0, in0} - {1'b0, in1};
-       assign rdiff = {1'b0, in1} + {1'b0, in0};
+       assign rdiff = {1'b0, in1} - {1'b0, in0};
        assign sum_v = (in0[31] ^~ in1[31]) & (sum[31] ^ in0[31]);
        assign diff_v = (in0[31] ^ in1[31]) & (diff[31] ^ in0[31]);
        assign rdiff_v = (in0[31] ^ in1[31]) & (rdiff[31] ^ in1[31]);
@@ -259,11 +333,13 @@ module ALU(
                end
                `ALU_SUB: begin
                        {flag_c, result} = diff;
+                       flag_c = !flag_c;
                        flag_v = diff_v;
                        setres = 1'b1;
                end
                `ALU_RSB: begin
                        {flag_c, result} = rdiff;
+                       flag_c = !flag_c;
                        flag_v = rdiff_v;
                        setres = 1'b1;
                end
@@ -279,11 +355,13 @@ module ALU(
                end
                `ALU_SBC: begin
                        {flag_c, result} = diff - {32'b0, (~cpsr[`CPSR_C])};
+                       flag_c = !flag_c;
                        flag_v = diff_v | (diff[31] & ~result[31]);
                        setres = 1'b1;
                end
                `ALU_RSC: begin
                        {flag_c, result} = rdiff - {32'b0, (~cpsr[`CPSR_C])};
+                       flag_c = !flag_c;
                        flag_v = rdiff_v | (rdiff[31] & ~result[31]);
                        setres = 1'b1;
                end
@@ -299,6 +377,7 @@ module ALU(
                end
                `ALU_CMP: begin
                        {flag_c, result} = diff;
+                       flag_c = !flag_c;
                        flag_v = diff_v;
                        setres = 1'b0;
                end
This page took 0.026118 seconds and 4 git commands to generate.