Add (bnot (bxor x y)) lowerings for s390x/aarch64 (#5763)

* Add (bnot (bxor x y)) lowerings for s390x/aarch64

I originally thought that s390x's original lowering in #5709, but as was
rightfully pointed out `(bnot (bxor x y))` is equivalent to
`(bxor x (bnot y))` so the special lowering for one should apply as a
special lowering for the other. For the s390x and aarch64 backend that
have already have a fused lowering of the bxor/bnot add a lowering
additionally for the bnot/bxor combination.

* Add bnot(bxor(..)) tests for s390x 128-bit sizes

cranelift/codegen/src/isa/aarch64/lower.isle

@@ -1052,6 +1052,11 @@
 (rule -2 (lower (has_type (ty_vec128 ty) (bnot x)))
       (not x (vector_size ty)))
 
+;; Special-cases for fusing a bnot with bxor
+(rule 2 (lower (has_type (fits_in_64 ty) (bnot (bxor x y))))
+      (alu_rs_imm_logic (ALUOp.EorNot) ty x y))
+(rule 3 (lower (has_type $I128 (bnot (bxor x y)))) (i128_alu_bitop (ALUOp.EorNot) $I64 x y))
+
 ;;;; Rules for `band` ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 
 (rule -1 (lower (has_type (fits_in_64 ty) (band x y)))

cranelift/codegen/src/isa/s390x/lower.isle

@@ -956,6 +956,15 @@
 (rule (lower (has_type (vr128_ty ty) (bnot x)))
       (vec_not ty x))
 
+;; With z15 (bnot (bxor ...)) can be a single instruction, similar to the
+;; (bxor _ (bnot _)) lowering.
+(rule 3 (lower (has_type (and (mie2_enabled) (fits_in_64 ty)) (bnot (bxor x y))))
+      (not_xor_reg ty x y))
+
+;; Combine a not/xor operation of vector types into one.
+(rule 4 (lower (has_type (vr128_ty ty) (bnot (bxor x y))))
+      (vec_not_xor ty x y))
+
 
 ;;;; Rules for `band` ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 

cranelift/filetests/filetests/isa/aarch64/bitops.clif

@@ -924,3 +924,26 @@ block0(v0: i128, v1: i128):
 ;   csel x1, x15, x7, ne
 ;   ret
 
+function %bnot_of_bxor(i32, i32) -> i32 {
+block0(v0: i32, v1: i32):
+  v2 = bxor v0, v1
+  v3 = bnot v2
+  return v3
+}
+
+; block0:
+;   eon w0, w0, w1
+;   ret
+
+function %bnot_of_bxor(i128, i128) -> i128 {
+block0(v0: i128, v1: i128):
+  v2 = bxor v0, v1
+  v3 = bnot v2
+  return v3
+}
+
+; block0:
+;   eon x0, x0, x2
+;   eon x1, x1, x3
+;   ret
+

cranelift/filetests/filetests/isa/s390x/bitops-optimized.clif

@@ -64,3 +64,14 @@ block0(v0: i32, v1: i32):
 ; block0:
 ;   nxrk %r2, %r3, %r2
 ;   br %r14
+
+function %bnot_of_bxor(i32, i32) -> i32 {
+block0(v0: i32, v1: i32):
+  v2 = bxor v0, v1
+  v3 = bnot v2
+  return v3
+}
+
+; block0:
+;   nxrk %r2, %r2, %r3
+;   br %r14

cranelift/filetests/filetests/isa/s390x/bitwise.clif

@@ -632,3 +632,40 @@ block0(v0: i8, v1: i8, v2: i8):
 ;   or %r2, %r3
 ;   br %r14
 
+function %bnot_of_bxor(i32, i32) -> i32 {
+block0(v0: i32, v1: i32):
+  v2 = bxor v0, v1
+  v3 = bnot v2
+  return v3
+}
+
+; block0:
+;   xr %r2, %r3
+;   xilf %r2, 4294967295
+;   br %r14
+
+function %bnot_of_bxor(i128, i128) -> i128 {
+block0(v0: i128, v1: i128):
+  v2 = bxor v0, v1
+  v3 = bnot v2
+  return v3
+}
+
+; block0:
+;   vl %v1, 0(%r3)
+;   vl %v3, 0(%r4)
+;   vnx %v6, %v1, %v3
+;   vst %v6, 0(%r2)
+;   br %r14
+
+function %bnot_of_bxor(i32x4, i32x4) -> i32x4 {
+block0(v0: i32x4, v1: i32x4):
+  v2 = bxor v0, v1
+  v3 = bnot v2
+  return v3
+}
+
+; block0:
+;   vnx %v24, %v24, %v25
+;   br %r14
+