diff --git a/fft-org.c b/fft-org.c
index e33de5b..a74dfad 100644
--- a/fft-org.c
+++ b/fft-org.c
@@ -126,6 +126,7 @@ int fix_fft_org(fixed fr[], fixed fi[], int m, int inverse)
                 wr >>= 1;
                 wi >>= 1;
             }
+            
             for(i=m; i<n; i+=istep)
             {
 
@@ -148,6 +149,7 @@ int fix_fft_org(fixed fr[], fixed fi[], int m, int inverse)
                 fi[i] = qi + ti;
             }
         }
+        
         --k;
         l = istep;
     }
diff --git a/fft.c b/fft.c
index b69eda7..e6fdbf0 100644
--- a/fft.c
+++ b/fft.c
@@ -112,26 +112,41 @@ int fix_fft(fixed fr[], fixed fi[], int m, int inverse)
         istep = l << 1;		//step width of current butterfly
         
         FFT_reg reg;
+        FFT_reg_simd simd_r, simd_i;
         fixed *reg_s = ((fixed*) &reg);
         
-        for(m=0; m<n; m+=istep)
+        if(l == 1)
         {
-        	for(i=m; i<m+l; ++i)
-            {
-                j = i + l;
-                
-                reg_s[3] = fr[i];
-                reg_s[2] = fr[j];
-                reg_s[1] = fi[i];
-                reg_s[0] = fi[j];
-                
-                FFT_calc(reg, i << k, (xtbool) shift, inverse);
-                
-                fr[i] = reg_s[3];
-                fr[j] = reg_s[2];
-                fi[i] = reg_s[1];
-                fi[j] = reg_s[0];
-            }
+        	for(i=0; i<n; i+=8)
+        	{
+        		simd_r = FFT_simd_load(fr, i);
+        		simd_i = FFT_simd_load(fi, i);
+    			FFT_simd_first(simd_r, simd_i, (xtbool) shift);
+        		FFT_simd_store(fr, i, simd_r);
+        		FFT_simd_store(fi, i, simd_i);
+        	}
+        }
+        else
+        {
+	        for(m=0; m<n; m+=istep)
+	        {
+	        	for(i=m; i<m+l; ++i)
+	            {
+	                j = i + l;
+	                
+	                reg_s[3] = fr[i];
+	                reg_s[2] = fr[j];
+	                reg_s[1] = fi[i];
+	                reg_s[0] = fi[j];
+	                
+	                FFT_calc(reg, i << k, (xtbool) shift, inverse);
+	                
+	                fr[i] = reg_s[3];
+	                fr[j] = reg_s[2];
+	                fi[i] = reg_s[1];
+	                fi[j] = reg_s[0];
+	            }
+        	}
         }
         --k;
         l = istep;
diff --git a/fft_inst.tie b/fft_inst.tie
index d61ba99..dc9825c 100644
--- a/fft_inst.tie
+++ b/fft_inst.tie
@@ -132,6 +132,62 @@ table SIN_wave 16 1024 {
 
 regfile FFT_reg 64 2 fftv
 
+regfile FFT_reg_simd 128 2 fftsv
+
+function [31:0] FFT_var_shift([31:0] data, [3:0] sh)
+{
+	assign FFT_var_shift = TIEmux(sh,
+		data[31:0],
+		{data[30:0], 1'b0},
+		{data[29:0], 2'b0},
+		{data[28:0], 3'b0},
+		{data[27:0], 4'b0},
+		{data[26:0], 5'b0},
+		{data[25:0], 6'b0},
+		{data[24:0], 7'b0},
+		{data[23:0], 8'b0},
+		{data[22:0], 9'b0},
+		{data[21:0], 10'b0},
+		{data[20:0], 11'b0},
+		{data[19:0], 12'b0},
+		{data[18:0], 13'b0},
+		{data[17:0], 14'b0},
+		{data[16:0], 15'b0}
+	);
+}
+
+operation FFT_simd_load {in AR *base, in AR offset, out FFT_reg_simd data} {out VAddr, in MemDataIn128}
+{
+	assign VAddr = TIEadd(base, {offset[30:0], 1'b0}, 1'b0);
+	
+	wire [15:0] o1 = MemDataIn128[15:0];
+	wire [15:0] o2 = MemDataIn128[31:16];
+	wire [15:0] o3 = MemDataIn128[47:32];
+	wire [15:0] o4 = MemDataIn128[63:48];
+	wire [15:0] o5 = MemDataIn128[79:64];
+	wire [15:0] o6 = MemDataIn128[95:80];
+	wire [15:0] o7 = MemDataIn128[111:96];
+	wire [15:0] o8 = MemDataIn128[127:112];
+	
+	assign data = {o1, o2, o3, o4, o5, o6, o7, o8 };
+}
+
+operation FFT_simd_store {in AR *base, in AR offset, in FFT_reg_simd data} {out VAddr, out MemDataOut128}
+{
+	assign VAddr = TIEadd(base, {offset[30:0], 1'b0}, 1'b0);
+	
+	wire [15:0] o1 = data[15:0];
+	wire [15:0] o2 = data[31:16];
+	wire [15:0] o3 = data[47:32];
+	wire [15:0] o4 = data[63:48];
+	wire [15:0] o5 = data[79:64];
+	wire [15:0] o6 = data[95:80];
+	wire [15:0] o7 = data[111:96];
+	wire [15:0] o8 = data[127:112];
+	
+	assign MemDataOut128 = {o1, o2, o3, o4, o5, o6, o7, o8 };
+}
+
 operation FFT_shift_check {in AR *base, in AR offset, out AR needs_shift} {out VAddr, in MemDataIn128}
 {
 	assign VAddr = TIEadd(base, offset[31:1], 1'b0);
@@ -243,6 +299,22 @@ operation FFT_calc {inout FFT_reg data, in AR i, in BR shift, in BR inverse} {}
 	assign data = FFT_butterfly(data, wr, wi, shift);
 }
 
+// 4 butterflies at once
+operation FFT_simd_first {inout FFT_reg_simd fr, inout FFT_reg_simd fi, in BR shift} {}
+{
+	
+	wire [15:0] wr = TIEmux(shift, 16'h7fff, 16'h3fff);
+	wire [15:0] wi = 16'b0;
+	
+	wire [63:0] res1 = FFT_butterfly({fr[127:96], fi[127:96]}, wr, wi, shift);
+	wire [63:0] res2 = FFT_butterfly({fr[95:64],  fi[95:64]},  wr, wi, shift);
+	wire [63:0] res3 = FFT_butterfly({fr[63:32],  fi[63:32]},  wr, wi, shift);
+	wire [63:0] res4 = FFT_butterfly({fr[31:0],   fi[31:0]},   wr, wi, shift);
+	
+	assign fr = { res1[63:32], res2[63:32], res3[63:32], res4[63:32] };
+	assign fi = { res1[31:0],  res2[31:0],  res3[31:0],  res4[31:0] };
+}
+
 ////////////////////////////////////////////////////////////////////////////
 //
 // Generated by XPRES v4.0.4
@@ -321,11 +393,11 @@ slot_opcodes Inst {
 	 FFT_bit_reverse, FFT_calc, ld.FFT_reg,
 	 mv.FFT_reg, st.FFT_reg }
 slot_opcodes flix64_0_slot0 {
-	 ADD, BEQZ, BGEU, BLTU, BNEZ, L16SI, MOV.N, NOP, S16I, SEXT, SUB, FFT_shift_check }
+	 ADD, BEQZ, BGEU, BLTU, BNEZ, L16SI, MOV.N, NOP, S16I, SEXT, SUB, FFT_shift_check, FFT_simd_load, FFT_simd_store }
 slot_opcodes flix64_0_slot1 {
 	 ADD, ADDX2, MOV.N, MOVNEZ, NEG, NOP }
 slot_opcodes flix64_0_slot2 {
-	 ADD, L16SI, NOP, S16I, SUB, FFT_shift_check }
+	 ADD, L16SI, NOP, S16I, SUB, FFT_shift_check, FFT_simd_load, FFT_simd_store }
 
 
 //--------------------------------------------------------------------------
@@ -335,3 +407,4 @@ slot_opcodes flix64_0_slot2 {
 //--------------------------------------------------------------------------
 
 ctype FFT_reg 64 64 FFT_reg default
+ctype FFT_reg_simd 128 128 FFT_reg_simd default