Merge branch 'main' of github.com:davidharrishmc/riscv-wally into main

pipelined/config/shared/wally-shared.vh

@@ -26,7 +26,7 @@
 `include "wally-constants.vh"
 
 // macros to define supported modes
-// NOTE: No hardware support fo Q yet
+// NOTE: No hardware support for Q yet
 
 `define A_SUPPORTED ((`MISA >> 0) % 2 == 1)
 `define C_SUPPORTED ((`MISA >> 2) % 2 == 1)
@@ -111,15 +111,15 @@
 // division constants
 `define RADIX 32'h4
 `define DIVCOPIES 32'h3
-`define DIVLEN ((`NF < `XLEN) ? (`XLEN) : (`NF + 3))
+`define DIVLEN ((`NF < `XLEN) ? (`XLEN) : `NF+3)
 // `define DIVN (`NF < `XLEN ? `XLEN : `NF+1) // length of input
-`define DIVN (`NF < `XLEN ? `XLEN : `NF+3) // length of input
+`define DIVN (`NF<`XLEN ? `XLEN : (`NF + 3)) // length of input
 `define EXTRAFRACBITS ((`NF < (`XLEN)) ? (`XLEN - `NF) : 3)
-`define EXTRAINTBITS ((`NF<(`XLEN)) ? 0 : (`NF - `XLEN + 3))
+`define EXTRAINTBITS ((`NF < `XLEN) ? 0 : (`NF - `XLEN + 3))
-`define DIVRESLEN ((`NF>`XLEN) ? `NF+4 : `XLEN)
+`define DIVRESLEN ((`NF>`XLEN) ? (`NF + 4) : `XLEN)
 `define LOGR ((`RADIX==2) ? 32'h1 : 32'h2)
 // FPDUR = ceil(DIVRESLEN/(LOGR*DIVCOPIES))
-// one interation is required for the integer bit for minimally redundent radix-4
+// one iteration is required for the integer bit for minimally redundent radix-4
 `define FPDUR ((`DIVN+2+(`LOGR*`DIVCOPIES)-1)/(`LOGR*`DIVCOPIES)+(`RADIX/4))
 `define DURLEN ($clog2(`FPDUR+1))
 `define QLEN (`FPDUR*`LOGR*`DIVCOPIES)

pipelined/src/fpu/fdivsqrt/fdivsqrt.sv

@@ -67,7 +67,7 @@ module fdivsqrt(
 
   fdivsqrtpreproc fdivsqrtpreproc(
     .clk, .DivStartE, .Xm(XmE), .QeM, .Xe(XeE), .Fmt(FmtE), .Ye(YeE), 
-    .Sqrt(SqrtE), .Int(MDUE), .Ym(YmE), .XZero(XZeroE), .X, .Dpreproc, 
+    .Sqrt(SqrtE), .Ym(YmE), .XZero(XZeroE), .X, .Dpreproc, 
     .ForwardedSrcAE, .ForwardedSrcBE, .Funct3E, .Funct3M, .MDUE, .W64E);
   fdivsqrtfsm fdivsqrtfsm(
     .clk, .reset, .FmtE, .XsE, .SqrtE, 

pipelined/src/fpu/fdivsqrt/fdivsqrtiter.sv

@@ -50,7 +50,7 @@ module fdivsqrtiter(
 //QLEN = 1.(number of bits created for division)
 // N is NF+1 or XLEN
 // WC/WS is dependent on D so 4.N-1 ie N+3 bits or N+2:0 + one more bit in fraction for possible sqrt right shift
-// D is 1.N-1, but the msb is always 1 so 0.N-1 or N-1 bits or N-1:0
+// D is 1.N-1, but the msb is always 1 so 0.N-1 or N-1 bits or N-2:0
 // Dsel should match WC/WS so 4.N-1 ie N+3 bits or N+2:0
 // U/UM should be 1.b so b+1 bits or b:0
 // C needs to be the lenght of the final fraction 0.b so b or b-1:0

pipelined/src/fpu/fdivsqrt/fdivsqrtpreproc.sv

@@ -37,7 +37,6 @@ module fdivsqrtpreproc (
   input  logic [`NE-1:0] Xe, Ye,
   input  logic [`FMTBITS-1:0] Fmt,
   input  logic Sqrt,
-  input  logic Int,
   input  logic XZero,
   input  logic [`XLEN-1:0] ForwardedSrcAE, ForwardedSrcBE, // *** these are the src outputs before the mux choosing between them and PCE to put in srcA/B
 	input  logic [2:0] 	Funct3E, Funct3M,
@@ -46,15 +45,17 @@ module fdivsqrtpreproc (
   output logic [`DIVb+3:0] X,
   output logic [`DIVN-2:0] Dpreproc
 );
-  // logic  [`XLEN-1:0] PosA, PosB;
   // logic  [`DIVLEN-1:0] ExtraA, ExtraB, PreprocA, PreprocB, PreprocX, PreprocY;
   logic  [`NF-1:0] PreprocA, PreprocX;
   logic  [`NF-1:0] PreprocB, PreprocY;
-  // logic  [`DIVN-1:0] ZeroBufX, ZeroBufY; add after Cedar Commit
   logic  [`NF+1:0] SqrtX;
   logic  [`DIVb+3:0] DivX;
   logic  [$clog2(`NF+2)-1:0] XZeroCnt, YZeroCnt;
   logic  [`NE+1:0] Qe;
+  // Intdiv signals
+  // logic  [`DIVN-1:0] ZeroBufX, ZeroBufY; add after Cedar Commit
+  logic  [`XLEN-1:0] PosA, PosB;
+  logic  Signed, Aneg, Bneg;
 
   // ***can probably merge X LZC with conversion
   // cout the number of leading zeros
@@ -64,6 +65,12 @@ module fdivsqrtpreproc (
   lzc #(`NF+1) lzcX (Xm, XZeroCnt);
   lzc #(`NF+1) lzcY (Ym, YZeroCnt);
 
+  assign Signed = Funct3E[0];
+  assign Aneg = ForwardedSrcAE[`XLEN-1] & Signed;
+  assign Bneg = ForwardedSrcBE[`XLEN-1] & Signed;
+  assign PosA = Aneg ? -ForwardedSrcAE : ForwardedSrcAE;
+  assign PosB = Bneg ? -ForwardedSrcBE : ForwardedSrcBE;
+
   assign PreprocX = Xm[`NF-1:0]<<XZeroCnt;
   assign PreprocY = Ym[`NF-1:0]<<YZeroCnt;