cmd/compile,cmd/internal/obj/riscv,runtime: use Duff's devices on riscv64

Implement runtime.duffzero and runtime.duffcopy for riscv64.
Use obj.ADUFFZERO/obj.ADUFFCOPY for medium size, word aligned
zeroing/moving.

Change-Id: I42ec622055630c94cb77e286d8d33dbe7c9f846c
Reviewed-on: https://go-review.googlesource.com/c/go/+/237797
Run-TryBot: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Joel Sing <joel@sing.id.au>
Reviewed-by: Cherry Zhang <cherryyz@google.com>

1 files changed, 33 insertions, 0 deletions

diff --git a/src/cmd/compile/internal/ssa/rewriteRISCV64.go b/src/cmd/compile/internal/ssa/rewriteRISCV64.go
index ac92945753..c337ffbfe7 100644
--- a/src/cmd/compile/internal/ssa/rewriteRISCV64.go
+++ b/src/cmd/compile/internal/ssa/rewriteRISCV64.go
@@ -2017,6 +2017,23 @@ func rewriteValueRISCV64_OpMove(v *Value) bool {
 		return true
 	}
 	// match: (Move [s] {t} dst src mem)
+	// cond: s%8 == 0 && s >= 16 && s <= 8*128 && t.Alignment()%8 == 0 && !config.noDuffDevice && logLargeCopy(v, s)
+	// result: (DUFFCOPY [16 * (128 - s/8)] dst src mem)
+	for {
+		s := auxIntToInt64(v.AuxInt)
+		t := auxToType(v.Aux)
+		dst := v_0
+		src := v_1
+		mem := v_2
+		if !(s%8 == 0 && s >= 16 && s <= 8*128 && t.Alignment()%8 == 0 && !config.noDuffDevice && logLargeCopy(v, s)) {
+			break
+		}
+		v.reset(OpRISCV64DUFFCOPY)
+		v.AuxInt = int64ToAuxInt(16 * (128 - s/8))
+		v.AddArg3(dst, src, mem)
+		return true
+	}
+	// match: (Move [s] {t} dst src mem)
 	// cond: (s <= 16 || logLargeCopy(v, s))
 	// result: (LoweredMove [t.Alignment()] dst src (ADDI <src.Type> [s-moveSize(t.Alignment(), config)] src) mem)
 	for {
@@ -5650,6 +5667,22 @@ func rewriteValueRISCV64_OpZero(v *Value) bool {
 		return true
 	}
 	// match: (Zero [s] {t} ptr mem)
+	// cond: s%8 == 0 && s >= 16 && s <= 8*128 && t.Alignment()%8 == 0 && !config.noDuffDevice
+	// result: (DUFFZERO [8 * (128 - s/8)] ptr mem)
+	for {
+		s := auxIntToInt64(v.AuxInt)
+		t := auxToType(v.Aux)
+		ptr := v_0
+		mem := v_1
+		if !(s%8 == 0 && s >= 16 && s <= 8*128 && t.Alignment()%8 == 0 && !config.noDuffDevice) {
+			break
+		}
+		v.reset(OpRISCV64DUFFZERO)
+		v.AuxInt = int64ToAuxInt(8 * (128 - s/8))
+		v.AddArg2(ptr, mem)
+		return true
+	}
+	// match: (Zero [s] {t} ptr mem)
 	// result: (LoweredZero [t.Alignment()] ptr (ADD <ptr.Type> ptr (MOVDconst [s-moveSize(t.Alignment(), config)])) mem)
 	for {
 		s := auxIntToInt64(v.AuxInt)