diff --git a/.gitignore b/.gitignore
index 4b1d77a..c1f9720 100644
--- a/.gitignore
+++ b/.gitignore
@@ -2,3 +2,5 @@
 /bin
 /zig-cache
 /zig-out
+/docs
+**/*.log
\ No newline at end of file
diff --git a/src/cpu.zig b/src/cpu.zig
index d3bb169..02b7718 100644
--- a/src/cpu.zig
+++ b/src/cpu.zig
@@ -2,8 +2,10 @@ const std = @import("std");
 const util = @import("util.zig");
 const bitfield = @import("bitfield");
 
+const BarrelShifter = @import("cpu/barrel_shifter.zig");
 const Bus = @import("bus.zig").Bus;
 const Scheduler = @import("scheduler.zig").Scheduler;
+
 const Bitfield = bitfield.Bitfield;
 const Bit = bitfield.Bit;
 
@@ -160,7 +162,7 @@ fn populate() [0x1000]InstrFn {
     };
 }
 
-const CPSR = extern union {
+pub const CPSR = extern union {
     mode: Bitfield(u32, 0, 5),
     t: Bit(u32, 5),
     f: Bit(u32, 6),
diff --git a/src/cpu/barrel_shifter.zig b/src/cpu/barrel_shifter.zig
new file mode 100644
index 0000000..1e8c9fb
--- /dev/null
+++ b/src/cpu/barrel_shifter.zig
@@ -0,0 +1,83 @@
+const std = @import("std");
+
+const arm = @import("../cpu.zig");
+const Arm7tdmi = arm.Arm7tdmi;
+const CPSR = arm.CPSR;
+
+pub inline fn exec(cpu: *Arm7tdmi, opcode: u32) u32 {
+    var shift_amt: u8 = undefined;
+    if (opcode >> 4 & 1 == 1) {
+        shift_amt = @truncate(u8, cpu.r[opcode >> 8 & 0xF]);
+    } else {
+        shift_amt = @truncate(u8, opcode >> 7 & 0x1F);
+    }
+
+    const rm = cpu.r[opcode & 0xF];
+
+    return switch (@truncate(u2, opcode >> 5)) {
+        0b00 => logical_left(&cpu.cpsr, rm, shift_amt),
+        0b01 => logical_right(&cpu.cpsr, rm, shift_amt),
+        0b10 => arithmetic_right(&cpu.cpsr, rm, shift_amt),
+        0b11 => rotate_right(&cpu.cpsr, rm, shift_amt),
+    };
+}
+
+pub inline fn logical_left(cpsr: *CPSR, rm: u32, shift_byte: u8) u32 {
+    const shift_amt = @truncate(u5, shift_byte);
+    const bit_count: u8 = @typeInfo(u32).Int.bits;
+
+    var result: u32 = 0x0000_0000;
+
+    if (shift_byte < bit_count) {
+        // We can perform a well-defined shift here
+
+        // FIXME: We assume cpu.r[rs] == 0 and imm_shift == 0 are equivalent
+        if (shift_amt != 0) {
+            const carry_bit = @truncate(u5, bit_count - shift_amt);
+            cpsr.c.write(rm >> carry_bit & 1 == 1);
+        }
+
+        result = rm << shift_amt;
+    } else if (shift_byte == bit_count) {
+        // Shifted all bits out, carry bit is bit 0 of rm
+        cpsr.c.write(rm & 1 == 1);
+    } else {
+        // Shifted all bits out, carry bit has also been shifted out
+        cpsr.c.write(false);
+    }
+
+    return result;
+}
+
+pub inline fn logical_right(cpsr: *CPSR, rm: u32, shift_byte: u8) u32 {
+    const shift_amt = @truncate(u5, shift_byte);
+    const bit_count: u8 = @typeInfo(u32).Int.bits;
+
+    var result: u32 = 0x0000_0000;
+
+    if (shift_byte == 0 or shift_byte == bit_count) {
+        // Actualy LSR #32
+        cpsr.c.write(rm >> 31 & 1 == 1);
+    } else if (shift_byte < bit_count) {
+        // We can perform a well-defined shift
+        const carry_bit = shift_amt - 1;
+        cpsr.c.write(rm >> carry_bit & 1 == 1);
+
+        result = rm >> shift_amt;
+    } else {
+        // All bits have been shifted out, including carry bit
+        cpsr.c.write(false);
+    }
+
+    return result;
+}
+
+pub fn arithmetic_right(_: *CPSR, _: u32, _: u8) u32 {
+    // @bitCast(u32, @bitCast(i32, r_val) >> @truncate(u5, amount))
+    std.debug.panic("[BarrelShifter] implement arithmetic shift right", .{});
+}
+
+pub fn rotate_right(_: *CPSR, _: u32, _: u8) u32 {
+    // std.math.rotr(u32, r_val, amount)
+    std.debug.panic("[BarrelShifter] implement rotate right", .{});
+}
diff --git a/src/cpu/data_processing.zig b/src/cpu/data_processing.zig
index 4bfcb9d..1f73716 100644
--- a/src/cpu/data_processing.zig
+++ b/src/cpu/data_processing.zig
@@ -1,9 +1,10 @@
 const std = @import("std");
-const processor = @import("../cpu.zig");
+const arm = @import("../cpu.zig");
 
+const BarrelShifter = @import("barrel_shifter.zig");
 const Bus = @import("../bus.zig").Bus;
-const Arm7tdmi = processor.Arm7tdmi;
-const InstrFn = processor.InstrFn;
+const Arm7tdmi = arm.Arm7tdmi;
+const InstrFn = arm.InstrFn;
 
 pub fn comptimeDataProcessing(comptime I: bool, comptime S: bool, comptime instrKind: u4) InstrFn {
     return struct {
@@ -15,7 +16,7 @@ pub fn comptimeDataProcessing(comptime I: bool, comptime S: bool, comptime instr
             if (I) {
                 op2 = std.math.rotr(u32, opcode & 0xFF, (opcode >> 8 & 0xF) << 1);
             } else {
-                op2 = registerOp2(cpu, opcode);
+                op2 = BarrelShifter.exec(cpu, opcode);
             }
 
             switch (instrKind) {
@@ -45,7 +46,7 @@ pub fn comptimeDataProcessing(comptime I: bool, comptime S: bool, comptime instr
                     cpu.cpsr.n.write(result >> 31 & 1 == 1);
                     cpu.cpsr.z.write(result == 0);
                     cpu.cpsr.c.write(op2 <= op1_val);
-                    cpu.cpsr.v.write(v_ctx and (result >> 31 & 1== 1));
+                    cpu.cpsr.v.write(v_ctx and (result >> 31 & 1 == 1));
                 },
                 else => std.debug.panic("[CPU] TODO: implement data processing type {}", .{instrKind}),
             }
@@ -53,22 +54,22 @@ pub fn comptimeDataProcessing(comptime I: bool, comptime S: bool, comptime instr
     }.dataProcessing;
 }
 
-fn registerOp2(cpu: *const Arm7tdmi, opcode: u32) u32 {
-    var amount: u32 = undefined;
-    if (opcode >> 4 & 0x01 == 0x01) {
-        amount = cpu.r[opcode >> 8 & 0xF] & 0xFF;
-    } else {
-        amount = opcode >> 7 & 0x1F;
-    }
+// fn registerOp2(cpu: *const Arm7tdmi, opcode: u32) u32 {
+//     var amount: u32 = undefined;
+//     if (opcode >> 4 & 0x01 == 0x01) {
+//         amount = cpu.r[opcode >> 8 & 0xF] & 0xFF;
+//     } else {
+//         amount = opcode >> 7 & 0x1F;
+//     }
 
-    const rm = opcode & 0xF;
-    const r_val = cpu.r[rm];
+//     const rm = opcode & 0xF;
+//     const r_val = cpu.r[rm];
 
-    return switch (opcode >> 5 & 0x03) {
-        0b00 => r_val << @truncate(u5, amount),
-        0b01 => r_val >> @truncate(u5, amount),
-        0b10 => @bitCast(u32, @bitCast(i32, r_val) >> @truncate(u5, amount)),
-        0b11 => std.math.rotr(u32, r_val, amount),
-        else => unreachable,
-    };
-}
+//     return switch (opcode >> 5 & 0x03) {
+//         0b00 => r_val << @truncate(u5, amount),
+//         0b01 => r_val >> @truncate(u5, amount),
+//         0b10 => @bitCast(u32, @bitCast(i32, r_val) >> @truncate(u5, amount)),
+//         0b11 => std.math.rotr(u32, r_val, amount),
+//         else => unreachable,
+//     };
+// }
diff --git a/src/cpu/half_signed_data_transfer.zig b/src/cpu/half_signed_data_transfer.zig
index 7adfb2a..23814fc 100644
--- a/src/cpu/half_signed_data_transfer.zig
+++ b/src/cpu/half_signed_data_transfer.zig
@@ -1,10 +1,10 @@
 const std = @import("std");
-const processor = @import("../cpu.zig");
+const arm = @import("../cpu.zig");
 const util = @import("../util.zig");
 
 const Bus = @import("../bus.zig").Bus;
-const Arm7tdmi = processor.Arm7tdmi;
-const InstrFn = processor.InstrFn;
+const Arm7tdmi = arm.Arm7tdmi;
+const InstrFn = arm.InstrFn;
 
 pub fn comptimeHalfSignedDataTransfer(comptime P: bool, comptime U: bool, comptime I: bool, comptime W: bool, comptime L: bool) InstrFn {
     return struct {
diff --git a/src/cpu/single_data_transfer.zig b/src/cpu/single_data_transfer.zig
index acef639..c30bb00 100644
--- a/src/cpu/single_data_transfer.zig
+++ b/src/cpu/single_data_transfer.zig
@@ -1,10 +1,11 @@
 const std = @import("std");
 const util = @import("../util.zig");
-const processor = @import("../cpu.zig");
+const arm = @import("../cpu.zig");
 
+const BarrelShifter = @import("barrel_shifter.zig");
 const Bus = @import("../bus.zig").Bus;
-const Arm7tdmi = processor.Arm7tdmi;
-const InstrFn = processor.InstrFn;
+const Arm7tdmi = arm.Arm7tdmi;
+const InstrFn = arm.InstrFn;
 
 pub fn comptimeSingleDataTransfer(comptime I: bool, comptime P: bool, comptime U: bool, comptime B: bool, comptime W: bool, comptime L: bool) InstrFn {
     return struct {
@@ -48,15 +49,14 @@ pub fn comptimeSingleDataTransfer(comptime I: bool, comptime P: bool, comptime U
 }
 
 fn registerOffset(cpu: *Arm7tdmi, opcode: u32) u32 {
-    const amount = opcode >> 7 & 0x1F;
-    const rm = opcode & 0xF;
-    const r_val = cpu.r[rm];
+    const shift_byte = @truncate(u8, opcode >> 7 & 0x1F);
 
-    return switch (opcode >> 5 & 0x03) {
-        0b00 => r_val << @truncate(u5, amount),
-        0b01 => r_val >> @truncate(u5, amount),
-        0b10 => @bitCast(u32, @bitCast(i32, r_val) >> @truncate(u5, amount)),
-        0b11 => std.math.rotr(u32, r_val, amount),
-        else => unreachable,
+    const rm = cpu.r[opcode & 0xF];
+
+    return switch (@truncate(u2, opcode >> 5)) {
+        0b00 => BarrelShifter.logical_left(&cpu.cpsr, rm, shift_byte),
+        0b01 => BarrelShifter.logical_right(&cpu.cpsr, rm, shift_byte),
+        0b10 => BarrelShifter.arithmetic_right(&cpu.cpsr, rm, shift_byte),
+        0b11 => BarrelShifter.rotate_right(&cpu.cpsr, rm, shift_byte),
     };
 }