[llvm] [AArch64] Neoverse V1 scheduling info (PR #126707)

[Julien Villette via llvm-commits]

================
@@ -2973,10 +2973,10 @@ drps
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -     0.25   0.25   0.25   0.25    -      -      -      -     clz	x26, x4
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -     0.25   0.25   0.25   0.25    -      -      -      -     cls	w3, w5
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -     0.25   0.25   0.25   0.25    -      -      -      -     cls	x20, x5
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -     5.00    -      -      -      -      -      -      -     udiv	w0, w7, w10
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -     5.00    -      -      -      -      -      -      -     udiv	x9, x22, x4
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -     5.00    -      -      -      -      -      -      -     sdiv	w12, w21, w0
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -     5.00    -      -      -      -      -      -      -     sdiv	x13, x2, x1
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -     12.00   -      -      -      -      -      -      -     udiv	w0, w7, w10
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -      -     20.00   -      -      -      -      -      -      -     udiv	x9, x22, x4
----------------
jvillette38 wrote:

Yes. I took the worst-case... I am agree with you. I am going to fix it and probably doing some experiment with small benchmark (llvm-exegesis). We have seen that `FDIV` tends to be faster in reality than in SOG.
For a quick fix, do you prefer best-case, or 1/3 between best and worst cases for example (`udiv w: 6, udiv x: 10`)?

https://github.com/llvm/llvm-project/pull/126707