[llvm] [X86] lowerFPToIntToFP - handle UI2FP on AVX512VL targets and i64 types on AVX512DQ targets (PR #162656)

[Kavin Gnanapandithan via llvm-commits]

================
@@ -143,11 +173,20 @@ define double @ucvtf64_i64(double %a0) {
 ; AVX2-NEXT:    vaddsd %xmm0, %xmm1, %xmm0
 ; AVX2-NEXT:    retq
 ;
-; AVX512-LABEL: ucvtf64_i64:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    vcvttsd2usi %xmm0, %rax
-; AVX512-NEXT:    vcvtusi2sd %rax, %xmm15, %xmm0
-; AVX512-NEXT:    retq
+; AVX512-VL-LABEL: ucvtf64_i64:
----------------
KavinTheG wrote:

My bad, I thought I was to focus on i64 handling for this patch. 

I can push a commit that removes the check for i64 like below to add optimization for `ucvtf64_i32`.
```
  if (Subtarget.hasVLX() && IntVT == MVT::i64) {
```
to 
```
  if (Subtarget.hasVLX()) {
```

https://github.com/llvm/llvm-project/pull/162656