cmd/compile: added some intrinsics to SSA back end

One intrinsic was needed to help get the very best
performance out of a future GC; as long as that one was
being added, I also added Bswap since that is sometimes
a handy thing to have.  I had intended to fill out the
bit-scan intrinsic family, but the mismatch between the
"scan forward" instruction and "count leading zeroes"
was large enough to cause me to leave it out -- it poses
a dilemma that I'd rather dodge right now.

These intrinsics are not exposed for general use.
That's a separate issue requiring an API proposal change
( https://github.com/golang/proposal )

All intrinsics are tested, both that they are substituted
on the appropriate architecture, and that they produce the
expected result.

Change-Id: I5848037cfd97de4f75bdc33bdd89bba00af4a8ee
Reviewed-on: https://go-review.googlesource.com/20564
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

1 files changed, 109 insertions, 0 deletions

diff --git a/test/intrinsic.dir/main.go b/test/intrinsic.dir/main.go
new file mode 100644
index 0000000000..46e6cb3283
--- /dev/null
+++ b/test/intrinsic.dir/main.go
@@ -0,0 +1,109 @@
+// Copyright 2016 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package main
+
+import (
+	"fmt"
+	T "runtime/internal/sys"
+)
+
+var A = []uint64{0x0102030405060708, 0x1122334455667788}
+var B = []uint64{0x0807060504030201, 0x8877665544332211}
+
+var errors int
+
+func logf(f string, args ...interface{}) {
+	errors++
+	fmt.Printf(f, args...)
+	if errors > 100 { // 100 is enough spewage
+		panic("100 errors is plenty is enough")
+	}
+}
+
+func test(i, x uint64) {
+	t := T.Ctz64(x) // ERROR "intrinsic substitution for Ctz64"
+	if i != t {
+		logf("Ctz64(0x%x) expected %d but got %d\n", x, i, t)
+	}
+	x = -x
+	t = T.Ctz64(x) // ERROR "intrinsic substitution for Ctz64"
+	if i != t {
+		logf("Ctz64(0x%x) expected %d but got %d\n", x, i, t)
+	}
+
+	if i <= 32 {
+		x32 := uint32(x)
+		t32 := T.Ctz32(x32) // ERROR "intrinsic substitution for Ctz32"
+		if uint32(i) != t32 {
+			logf("Ctz32(0x%x) expected %d but got %d\n", x32, i, t32)
+		}
+		x32 = -x32
+		t32 = T.Ctz32(x32) // ERROR "intrinsic substitution for Ctz32"
+		if uint32(i) != t32 {
+			logf("Ctz32(0x%x) expected %d but got %d\n", x32, i, t32)
+		}
+	}
+	if i <= 16 {
+		x16 := uint16(x)
+		t16 := T.Ctz16(x16) // ERROR "intrinsic substitution for Ctz16"
+		if uint16(i) != t16 {
+			logf("Ctz16(0x%x) expected %d but got %d\n", x16, i, t16)
+		}
+		x16 = -x16
+		t16 = T.Ctz16(x16) // ERROR "intrinsic substitution for Ctz16"
+		if uint16(i) != t16 {
+			logf("Ctz16(0x%x) expected %d but got %d\n", x16, i, t16)
+		}
+	}
+}
+
+func main() {
+	// Test Bswap first because the other test relies on it
+	// working correctly (to implement bit reversal).
+	for i := range A {
+		x := A[i]
+		y := B[i]
+		X := T.Bswap64(x) // ERROR "intrinsic substitution for Bswap64"
+		Y := T.Bswap64(y) // ERROR "intrinsic substitution for Bswap64"
+		if y != X {
+			logf("Bswap64(0x%08x) expected 0x%08x but got 0x%08x\n", x, y, X)
+		}
+		if x != Y {
+			logf("Bswap64(0x%08x) expected 0x%08x but got 0x%08x\n", y, x, Y)
+		}
+
+		x32 := uint32(X)
+		y32 := uint32(Y >> 32)
+
+		X32 := T.Bswap32(x32) // ERROR "intrinsic substitution for Bswap32"
+		Y32 := T.Bswap32(y32) // ERROR "intrinsic substitution for Bswap32"
+		if y32 != X32 {
+			logf("Bswap32(0x%08x) expected 0x%08x but got 0x%08x\n", x32, y32, X32)
+		}
+		if x32 != Y32 {
+			logf("Bswap32(0x%08x) expected 0x%08x but got 0x%08x\n", y32, x32, Y32)
+		}
+	}
+
+	// Zero is a special case, be sure it is done right.
+	if T.Ctz16(0) != 16 { // ERROR "intrinsic substitution for Ctz16"
+		logf("ctz16(0) != 16")
+	}
+	if T.Ctz32(0) != 32 { // ERROR "intrinsic substitution for Ctz32"
+		logf("ctz32(0) != 32")
+	}
+	if T.Ctz64(0) != 64 { // ERROR "intrinsic substitution for Ctz64"
+		logf("ctz64(0) != 64")
+	}
+
+	for i := uint64(0); i <= 64; i++ {
+		for j := uint64(1); j <= 255; j += 2 {
+			for k := uint64(1); k <= 65537; k += 128 {
+				x := (j * k) << i
+				test(i, x)
+			}
+		}
+	}
+}