Use sNaN value for flw, like other single-precision ops
This commit is contained in:
parent
cc389bea90
commit
f505aba1ac
@ -536,7 +536,7 @@ class FPU(cfg: FPUParams)(implicit p: Parameters) extends FPUModule()(p) {
|
|||||||
case 32 => rec_s
|
case 32 => rec_s
|
||||||
case 64 =>
|
case 64 =>
|
||||||
val rec_d = hardfloat.recFNFromFN(dExpWidth, dSigWidth, load_wb_data)
|
val rec_d = hardfloat.recFNFromFN(dExpWidth, dSigWidth, load_wb_data)
|
||||||
Mux(load_wb_single, Cat(UInt((BigInt(1) << (fLen - 32)) - 1), rec_s), rec_d)
|
Mux(load_wb_single, rec_s | CanonicalNaN.signaling(maxExpWidth, maxSigWidth), rec_d)
|
||||||
}
|
}
|
||||||
|
|
||||||
// regfile
|
// regfile
|
||||||
|
Loading…
Reference in New Issue
Block a user