[llvm] [RISCV] Update SpacemiT-X60 vector floating-point instructions latencies (PR #150618)

[Min-Yih Hsu via llvm-commits]

================
@@ -541,83 +546,169 @@ foreach mx = SchedMxListF in {
   foreach sew = SchedSEWSet<mx, isF=1>.val in {
     defvar IsWorstCase = SMX60IsWorstCaseMXSEW<mx, sew, SchedMxListF, isF=1>.c;
 
-    defm "" : LMULSEWWriteResMXSEW<"WriteVFALUV", [SMX60_VFP], mx, sew, IsWorstCase>;
-    defm "" : LMULSEWWriteResMXSEW<"WriteVFALUF", [SMX60_VFP], mx, sew, IsWorstCase>;
-    defm "" : LMULSEWWriteResMXSEW<"WriteVFMulV", [SMX60_VFP], mx, sew, IsWorstCase>;
-    defm "" : LMULSEWWriteResMXSEW<"WriteVFMulF", [SMX60_VFP], mx, sew, IsWorstCase>;
-    defm "" : LMULSEWWriteResMXSEW<"WriteVFMulAddV", [SMX60_VFP], mx, sew, IsWorstCase>;
-    defm "" : LMULSEWWriteResMXSEW<"WriteVFMulAddF", [SMX60_VFP], mx, sew, IsWorstCase>;
-  }
-}
+    defvar VFALULat = Get4458Latency<mx>.c;
+    defvar VFALUOcc = ConstOneUntilM1ThenDouble<mx>.c;
+    let Latency = VFALULat, ReleaseAtCycles = [VFALUOcc] in {
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFALUV", [SMX60_VFP], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFALUF", [SMX60_VFP], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMinMaxV", [SMX60_VFP], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMinMaxF", [SMX60_VFP], mx, sew, IsWorstCase>;
+    }
 
-foreach mx = SchedMxListF in {
-  foreach sew = SchedSEWSet<mx, isF=1>.val in {
-    defvar IsWorstCase = SMX60IsWorstCaseMXSEW<mx, sew, SchedMxListF, isF=1>.c;
+    // Slightly increased latency for sew == 64
+    defvar VFMulVLatAndOcc = !if(!eq(sew, 64), ConstValueUntilLMULThenDoubleBase<"M8", 5, 8, mx>.c,
+                                               Get4458Latency<mx>.c);
+    let Latency = VFMulVLatAndOcc, ReleaseAtCycles = [VFMulVLatAndOcc] in {
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMulV", [SMX60_VFP], mx, sew, IsWorstCase>;
+    }
+    // VFMulF has the same latency as VFMulV, but slighlty lower ReleaseAtCycles
+    let Latency = VFMulVLatAndOcc, ReleaseAtCycles = [ConstOneUntilM1ThenDouble<mx>.c] in {
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMulF", [SMX60_VFP], mx, sew, IsWorstCase>;
+    }
 
-    defm "" : LMULSEWWriteResMXSEW<"WriteVFRecpV", [SMX60_VFP], mx, sew, IsWorstCase>;
-    defm "" : LMULSEWWriteResMXSEW<"WriteVFSgnjV", [SMX60_VFP], mx, sew, IsWorstCase>;
-    defm "" : LMULSEWWriteResMXSEW<"WriteVFSgnjF", [SMX60_VFP], mx, sew, IsWorstCase>;
-    defm "" : LMULSEWWriteResMXSEW<"WriteVFMinMaxV", [SMX60_VFP], mx, sew, IsWorstCase>;
-    defm "" : LMULSEWWriteResMXSEW<"WriteVFMinMaxF", [SMX60_VFP], mx, sew, IsWorstCase>;
+    defvar VFSgnjLat = ConstValueUntilLMULThenDouble<"M2", 4, mx>.c;
+    defvar VFSgnjOcc = ConstOneUntilMF2ThenDouble<mx>.c;
+    let Latency = VFSgnjLat, ReleaseAtCycles = [VFSgnjOcc] in {
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFRecpV", [SMX60_VFP], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFSgnjV", [SMX60_VFP], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFSgnjF", [SMX60_VFP], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFCvtIToFV", [SMX60_VFP], mx, sew, IsWorstCase>;
+    }
 
-    defm "" : LMULSEWWriteResMXSEW<"WriteVFCvtIToFV", [SMX60_VFP], mx, sew, IsWorstCase>;
+    // The following covers vfmacc, vfmsac, and their vfn* variants in the same group, but the
+    // ReleaseAtCycles takes one extra cycle for the vfn* variants.
+    // TODO: Should we split them?
+    // TODO: for some reason, the following cond is not working, and always use ConstValueUntilLMULThenDoubleBase<"M4", 5, 8, mx>.c
----------------
mshockwave wrote:

is this TODO still valid?

https://github.com/llvm/llvm-project/pull/150618