vfcvt: add fpCanonicalNAN for scalar fvt inst

OpenXiangShan · Apr 7, 2024 · 960dca4 · 960dca4
1 parent 9c4d97a
commit 960dca4
Show file tree

Hide file tree

Showing 3 changed files with 36 additions and 11 deletions.
diff --git a/src/main/scala/yunsuan/vector/VectorConvert/CVT64.scala b/src/main/scala/yunsuan/vector/VectorConvert/CVT64.scala
@@ -16,8 +16,8 @@ class CVT64(width: Int = 64) extends CVT(width){
   val widthExpAdder = 13 // 13bits is enough
 
   // input
-  val (fire, src, sew, opType, rmNext, input1H, output1H) =
-      (io.fire, io.src, io.sew, io.opType, io.rm, io.input1H, io.output1H)
+  val (fire, src, sew, opType, rmNext, input1H, output1H, isFpToVecInst) =
+      (io.fire, io.src, io.sew, io.opType, io.rm, io.input1H, io.output1H, io.isFpToVecInst)
   val fireReg = GatedValidRegNext(fire)
 
   // control for cycle 0
@@ -26,6 +26,18 @@ class CVT64(width: Int = 64) extends CVT(width){
   val inIsFpNext = opType.head(1).asBool
   val outIsFpNext = opType.tail(1).head(1).asBool
   val hasSignIntNext = opType(0).asBool
+  val s0_fpCanonicalNAN = isFpToVecInst & inIsFpNext & (input1H(1) & !src.head(48).andR | input1H(2) & !src.head(32).andR)
+  val s0_outIsF64 =  outIsFpNext && output1H(3)
+  val s0_outIsU32 = !outIsFpNext && output1H(2) && !hasSignIntNext
+  val s0_outIsS32 = !outIsFpNext && output1H(2) && hasSignIntNext
+  val s0_outIsU64 = !outIsFpNext && output1H(3) && !hasSignIntNext
+  val s0_outIsS64 = !outIsFpNext && output1H(3) && hasSignIntNext
+  val s2_fpCanonicalNAN = RegEnable(RegEnable(s0_fpCanonicalNAN, fire), fireReg)
+  val s2_outIsF64 = RegEnable(RegEnable(s0_outIsF64, fire), fireReg)
+  val s2_outIsU32 = RegEnable(RegEnable(s0_outIsU32, fire), fireReg)
+  val s2_outIsS32 = RegEnable(RegEnable(s0_outIsS32, fire), fireReg)
+  val s2_outIsU64 = RegEnable(RegEnable(s0_outIsU64, fire), fireReg)
+  val s2_outIsS64 = RegEnable(RegEnable(s0_outIsS64, fire), fireReg)
 
   val int1HSrcNext = input1H
   val float1HSrcNext = input1H.head(3)//exclude f8
@@ -609,7 +621,16 @@ class CVT64(width: Int = 64) extends CVT(width){
 
   fflagsNext := Cat(nv, dz, of, uf, nx)
 
-  io.result := result
-  io.fflags := fflags
+  io.result := Mux(s2_fpCanonicalNAN,
+      Mux1H(
+        Seq(s2_outIsF64, s2_outIsU32, s2_outIsS32, s2_outIsU64, s2_outIsS64),
+        Seq(~0.U((f64.expWidth+1).W) ## 0.U((f64.fracWidth-1).W),
+          ~0.U(32.W),
+          ~0.U(31.W),
+          ~0.U(64.W),
+          ~0.U(63.W))
+      ),
+    result)
+  io.fflags := Mux(s2_fpCanonicalNAN && !s2_outIsF64, "b10000".U ,fflags)
 }
 
diff --git a/src/main/scala/yunsuan/vector/VectorConvert/Convert.scala b/src/main/scala/yunsuan/vector/VectorConvert/Convert.scala
@@ -11,6 +11,7 @@ class VectorCvtIO(width: Int) extends Bundle {
   val opType = Input(UInt(8.W))
   val sew = Input(UInt(2.W))
   val rm = Input(UInt(3.W))
+  val isFpToVecInst = Input(Bool())
 
   val result = Output(UInt(width.W))
   val fflags = Output(UInt(20.W))
@@ -19,7 +20,7 @@ class VectorCvtIO(width: Int) extends Bundle {
 class VectorCvt(xlen :Int) extends Module{
 
   val io = IO(new VectorCvtIO(xlen))
-  val (fire, src, opType, sew, rm) = (io.fire, io.src, io.opType, io.sew, io.rm)
+  val (fire, src, opType, sew, rm, isFpToVecInst) = (io.fire, io.src, io.opType, io.sew, io.rm, io.isFpToVecInst)
   val widen = opType(4, 3) // 0->single 1->widen 2->norrow => width of result
 
   // input width 8， 16， 32， 64
@@ -82,16 +83,16 @@ class VectorCvt(xlen :Int) extends Module{
   element32 := src.asTypeOf(element32)
   element64 := src.asTypeOf(element64)
 
-  val in0 = Mux1H(inputWidth1H, Seq(element8(0), element16(0), element32(0), element64(0)))
+  val in0 = element64(0)
   val in1 = Mux1H(inputWidth1H, Seq(element8(1), element16(1), element32(1), 0.U))// input 0=> result 0 while norrow eg. 64b->32b
   val in2 = Mux1H(inputWidth1H, Seq(element8(2), element16(2), 0.U, 0.U))
   val in3 = Mux1H(inputWidth1H, Seq(element8(3), element16(3), 0.U, 0.U))
 
 
-  val (result0, fflags0) = VCVT(64)(fire, in0, opType, sew, rm, input1H, output1H)
-  val (result1, fflags1) = VCVT(32)(fire, in1, opType, sew, rm, input1H, output1H)
-  val (result2, fflags2) = VCVT(16)(fire, in2, opType, sew, rm, input1H, output1H)
-  val (result3, fflags3) = VCVT(16)(fire, in3, opType, sew, rm, input1H, output1H)
+  val (result0, fflags0) = VCVT(64)(fire, in0, opType, sew, rm, input1H, output1H, isFpToVecInst)
+  val (result1, fflags1) = VCVT(32)(fire, in1, opType, sew, rm, input1H, output1H, isFpToVecInst)
+  val (result2, fflags2) = VCVT(16)(fire, in2, opType, sew, rm, input1H, output1H, isFpToVecInst)
+  val (result3, fflags3) = VCVT(16)(fire, in3, opType, sew, rm, input1H, output1H, isFpToVecInst)
 
   io.result := Mux1H(outputWidth1H, Seq(
     result3(7,0) ## result2(7,0) ## result1(7,0) ## result0(7,0),

diff --git a/src/main/scala/yunsuan/vector/VectorConvert/VCVT.scala b/src/main/scala/yunsuan/vector/VectorConvert/VCVT.scala
@@ -11,6 +11,7 @@ class CVTIO(width: Int) extends Bundle {
   val rm = Input(UInt(3.W))
   val input1H = Input(UInt(4.W))
   val output1H = Input(UInt(4.W))
+  val isFpToVecInst = Input(Bool())
   val result = Output(UInt(width.W))
   val fflags = Output(UInt(5.W))
 }
@@ -37,7 +38,8 @@ object VCVT {
              sew:     UInt,
              rm:      UInt,
              input1H:      UInt,
-             output1H:      UInt
+             output1H:      UInt,
+             isFpToVecInst: Bool
            ): (UInt, UInt) = {
     val vcvtWraper = Module(new VCVT(width))
     vcvtWraper.io.fire := fire
@@ -47,6 +49,7 @@ object VCVT {
     vcvtWraper.io.rm := rm
     vcvtWraper.io.input1H := input1H
     vcvtWraper.io.output1H := output1H
+    vcvtWraper.io.isFpToVecInst := isFpToVecInst
     (vcvtWraper.io.result, vcvtWraper.io.fflags)
   }
 }