diff --git a/src/fpu/postproc/cvtshiftcalc.sv b/src/fpu/postproc/cvtshiftcalc.sv
index ff3d29b90..77b8543fd 100644
--- a/src/fpu/postproc/cvtshiftcalc.sv
+++ b/src/fpu/postproc/cvtshiftcalc.sv
@@ -98,6 +98,6 @@ module cvtshiftcalc import cvw::*;  #(parameter cvw_t P) (
   // determine if the result underflows ??? -> fp
   //      - if the first 1 is shifted out of the result then the result underflows
   //      - can't underflow an integer to fp conversions
-  assign CvtResUf = ($signed(CvtCe) < $signed({{P.NE-$clog2(P.NF){1'b1}}, ResNegNF}))&~XZero&~IntToFp;
+  assign CvtResUf = ($signed(CvtCe) < $signed({{P.NE-$clog2(P.NF){1'b1}}, ResNegNF}))&~XZero; // dh &~IntToFp not necessary because integer to float conversion never underflows
   
 endmodule
diff --git a/src/fpu/postproc/fmashiftcalc.sv b/src/fpu/postproc/fmashiftcalc.sv
index 0a8ac3035..b13b6d3da 100644
--- a/src/fpu/postproc/fmashiftcalc.sv
+++ b/src/fpu/postproc/fmashiftcalc.sv
@@ -101,9 +101,9 @@ module fmashiftcalc import cvw::*;  #(parameter cvw_t P) (
     assign Sum2GEFL = $signed(PreNormSumExp) >= $signed((P.NE+2)'(-P.NF2-2+P.BIAS-P.BIAS2)) | ~|PreNormSumExp;
     always_comb begin
       case (Fmt)
-        P.FMT: FmaPreResultSubnorm   = Sum0LEZ & Sum0GEFL & ~FmaSZero;
-        P.FMT1: FmaPreResultSubnorm  = Sum1LEZ & Sum1GEFL & ~FmaSZero;
-        P.FMT2: FmaPreResultSubnorm  = Sum2LEZ & Sum2GEFL & ~FmaSZero;
+        P.FMT: FmaPreResultSubnorm   = Sum0LEZ & Sum0GEFL; // & ~FmaSZero; // checking sum is not zero is harmless but turns out to be unnecessary
+        P.FMT1: FmaPreResultSubnorm  = Sum1LEZ & Sum1GEFL; // & ~FmaSZero;
+        P.FMT2: FmaPreResultSubnorm  = Sum2LEZ & Sum2GEFL; // & ~FmaSZero; 
         default: FmaPreResultSubnorm = 1'bx;
       endcase
     end
diff --git a/tests/coverage/fpu.S b/tests/coverage/fpu.S
index cfb3fb4b0..3bf4c5422 100644
--- a/tests/coverage/fpu.S
+++ b/tests/coverage/fpu.S
@@ -164,6 +164,28 @@ main:
     fcvt.s.w ft2, zero
     fdiv.s ft3, ft1, ft2
 
+/*  These didn't make a difference 
+
+    // Test subtraction of identical denormalized numbers with to exercise FmaPreResultSubnorm logic with FmaSZero
+    li t0, 0xFFFFFFFF00000001
+    fmv.w.x ft1, t0        // smallest denorm
+    fsub.s ft1, ft1, ft1    // difference = 0
+
+    li t0, 0x0000000000000001
+    fmv.d.x ft1, t0        // smallest denorm
+    fsub.d ft1, ft1, ft1    // difference = 0
+
+ // Test subtraction of identical denormalized numbers with to exercise FmaPreResultSubnorm logic with FmaSZero
+    li t0, 0xFFFFFFFF00100001
+    fmv.w.x ft1, t0        // almost largest denorm
+    fsub.s ft1, ft1, ft1    // difference = 0
+
+    li t0, 0x0001000000000001
+    fmv.d.x ft1, t0        // smallest denorm
+    fsub.d ft1, ft1, ft1    // difference = 0
+
+*/
+
     # Test floating point convert to integer and using result
     fcvt.w.s t0, f0
     add t1, t0, t0