Xavi/cvw
1
0
Fork 0
forked from Github_Repos/cvw
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fixed Issue #65 fmv sign selection. Sign needs to come from most significant bit of raw X source without doing NaN Box fixes first.

Browse Source
This commit is contained in:
David Harris 2023-02-21 09:57:57 -08:00
parent 5ce476241b
commit 8df7768d32
1 changed files with 14 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
src/fpu/fpu.sv
View File

@ -160,6 +160,7 @@ module fpu (
logic [`FLEN-1:0] BoxedOneE; // One value for Z for multiplication, with NaN boxing if needed logic [`FLEN-1:0] BoxedOneE; // One value for Z for multiplication, with NaN boxing if needed
logic StallUnpackedM; // Stall unpacker outputs during multicycle fdivsqrt logic StallUnpackedM; // Stall unpacker outputs during multicycle fdivsqrt
logic [`FLEN-1:0] SgnExtXE; // Sign-extended X input for move to integer logic [`FLEN-1:0] SgnExtXE; // Sign-extended X input for move to integer
logic mvsgn; // sign bit for extending move
////////////////////////////////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////////////////////////////////
// Decode Stage: fctrl decoder, read register file // Decode Stage: fctrl decoder, read register file
@ -278,21 +279,25 @@ module fpu (
mux3 #(`FLEN) FResMux(SgnResE, AlignedSrcAE, CmpFpResE, {OpCtrlE[2], &OpCtrlE[1:0]}, PreFpResE); mux3 #(`FLEN) FResMux(SgnResE, AlignedSrcAE, CmpFpResE, {OpCtrlE[2], &OpCtrlE[1:0]}, PreFpResE);
assign PreNVE = CmpNVE&(OpCtrlE[2]|FWriteIntE); assign PreNVE = CmpNVE&(OpCtrlE[2]|FWriteIntE);
// select the result that may be written to the integer register - to IEU // select the result that may be written to the integer register with fmv - to IEU
if(`FPSIZES == 1) if(`FPSIZES == 1) begin
assign mvsgn = XE[`FLEN-1];
assign SgnExtXE = XE; assign SgnExtXE = XE;
else if(`FPSIZES == 2) end else if(`FPSIZES == 2) begin
mux2 #(`FLEN) sgnextmux ({{`FLEN-`LEN1{XsE}}, XE[`LEN1-1:0]}, XE, FmtE, SgnExtXE); mux2 #(1) sgnmux (XE[`LEN1-1], XE[`FLEN-1],FmtE, mvsgn);
else if(`FPSIZES == 3 | `FPSIZES == 4) mux2 #(`FLEN) sgnextmux ({{`FLEN-`LEN1{mvsgn}}, XE[`LEN1-1:0]}, XE, FmtE, SgnExtXE);
mux4 #(`FLEN) fmulzeromux ({{`FLEN-`H_LEN{XsE}}, XE[`H_LEN-1:0]}, end else if(`FPSIZES == 3 | `FPSIZES == 4) begin
{{`FLEN-`S_LEN{XsE}}, XE[`S_LEN-1:0]}, mux4 #(1) sgnmux (XE[`H_LEN-1], XE[`S_LEN-1], XE[`D_LEN-1], XE[`LLEN-1], FmtE, mvsgn);
{{`FLEN-`D_LEN{XsE}}, XE[`D_LEN-1:0]}, mux4 #(`FLEN) fmulzeromux ({{`FLEN-`H_LEN{mvsgn}}, XE[`H_LEN-1:0]},
{{`FLEN-`S_LEN{mvsgn}}, XE[`S_LEN-1:0]},
{{`FLEN-`D_LEN{mvsgn}}, XE[`D_LEN-1:0]},
XE, FmtE, SgnExtXE); XE, FmtE, SgnExtXE);
end
if (`FLEN>`XLEN) if (`FLEN>`XLEN)
assign IntSrcXE = SgnExtXE[`XLEN-1:0]; assign IntSrcXE = SgnExtXE[`XLEN-1:0];
else else
assign IntSrcXE = {{`XLEN-`FLEN{XsE}}, SgnExtXE}; assign IntSrcXE = {{`XLEN-`FLEN{mvsgn}}, SgnExtXE};
mux3 #(`XLEN) IntResMux (ClassResE, IntSrcXE, CmpIntResE, {~FResSelE[1], FResSelE[0]}, FIntResE); mux3 #(`XLEN) IntResMux (ClassResE, IntSrcXE, CmpIntResE, {~FResSelE[1], FResSelE[0]}, FIntResE);
// E/M pipe registers // E/M pipe registers