cmd/compile, runtime: intrinsify atomic And8 and Or8 on s390x

Intrinsify these functions to match other platforms. Update the
sequence of instructions used in the assembly implementations to
match the intrinsics.

Also, add a micro benchmark so we can more easily measure the
performance of these two functions:

name            old time/op  new time/op  delta
And8-8          5.33ns ± 7%  2.55ns ± 8%  -52.12%  (p=0.000 n=20+20)
And8Parallel-8  7.39ns ± 5%  3.74ns ± 4%  -49.34%  (p=0.000 n=20+20)
Or8-8           4.84ns ±15%  2.64ns ±11%  -45.50%  (p=0.000 n=20+20)
Or8Parallel-8   7.27ns ± 3%  3.84ns ± 4%  -47.10%  (p=0.000 n=19+20)

By using a 'rotate then xor selected bits' instruction combined with
either a 'load and and' or a 'load and or' instruction we can
implement And8 and Or8 with far fewer instructions. Replacing
'compare and swap' with atomic instructions may also improve
performance when there is contention.

Change-Id: I28bb8032052b73ae8ccdf6e4c612d2877085fa01
Reviewed-on: https://go-review.googlesource.com/c/go/+/204277
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

1 files changed, 34 insertions, 0 deletions

diff --git a/src/cmd/compile/internal/s390x/ssa.go b/src/cmd/compile/internal/s390x/ssa.go
index 885c14b33a..f1725bdda4 100644
--- a/src/cmd/compile/internal/s390x/ssa.go
+++ b/src/cmd/compile/internal/s390x/ssa.go
@@ -173,6 +173,21 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 		if r != r1 {
 			p.Reg = r1
 		}
+	case ssa.OpS390XRXSBG:
+		r1 := v.Reg()
+		if r1 != v.Args[0].Reg() {
+			v.Fatalf("input[0] and output not in same register %s", v.LongString())
+		}
+		r2 := v.Args[1].Reg()
+		i := v.Aux.(s390x.RotateParams)
+		p := s.Prog(v.Op.Asm())
+		p.From = obj.Addr{Type: obj.TYPE_CONST, Offset: int64(i.Start)}
+		p.RestArgs = []obj.Addr{
+			{Type: obj.TYPE_CONST, Offset: int64(i.End)},
+			{Type: obj.TYPE_CONST, Offset: int64(i.Amount)},
+			{Type: obj.TYPE_REG, Reg: r2},
+		}
+		p.To = obj.Addr{Type: obj.TYPE_REG, Reg: r1}
 	case ssa.OpS390XADD, ssa.OpS390XADDW,
 		ssa.OpS390XSUB, ssa.OpS390XSUBW,
 		ssa.OpS390XAND, ssa.OpS390XANDW,
@@ -736,6 +751,25 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 		p.To.Type = obj.TYPE_MEM
 		p.To.Reg = v.Args[0].Reg()
 		gc.AddAux(&p.To, v)
+	case ssa.OpS390XLANfloor, ssa.OpS390XLAOfloor:
+		r := v.Args[0].Reg() // clobbered, assumed R1 in comments
+
+		// Round ptr down to nearest multiple of 4.
+		// ANDW $~3, R1
+		ptr := s.Prog(s390x.AANDW)
+		ptr.From.Type = obj.TYPE_CONST
+		ptr.From.Offset = 0xfffffffc
+		ptr.To.Type = obj.TYPE_REG
+		ptr.To.Reg = r
+
+		// Redirect output of LA(N|O) into R1 since it is clobbered anyway.
+		// LA(N|O) Rx, R1, 0(R1)
+		op := s.Prog(v.Op.Asm())
+		op.From.Type = obj.TYPE_REG
+		op.From.Reg = v.Args[1].Reg()
+		op.Reg = r
+		op.To.Type = obj.TYPE_MEM
+		op.To.Reg = r
 	case ssa.OpS390XLAA, ssa.OpS390XLAAG:
 		p := s.Prog(v.Op.Asm())
 		p.Reg = v.Reg0()