cmd/compile: insert complicated x86 addressing modes as a separate pass

Use a separate compiler pass to introduce complicated x86 addressing
modes.  Loads in the normal architecture rules (for x86 and all other
platforms) can have constant offsets (AuxInt values) and symbols (Aux
values), but no more.

The complex addressing modes (x+y, x+2*y, etc.) are introduced in a
separate pass that combines loads with LEAQx ops.

Organizing rewrites this way simplifies the number of rewrites
required, as there are lots of different rule orderings that have to
be specified to ensure these complex addressing modes are always found
if they are possible.

Update #36468

Change-Id: I5b4bf7b03a1e731d6dfeb9ef19b376175f3b4b44
Reviewed-on: https://go-review.googlesource.com/c/go/+/217097
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>

1 files changed, 88 insertions, 0 deletions

diff --git a/test/codegen/memops.go b/test/codegen/memops.go
index dcf5863666..9d18153a29 100644
--- a/test/codegen/memops.go
+++ b/test/codegen/memops.go
@@ -93,3 +93,91 @@ func compMem3(x, y *int) (int, bool) {
 	// 386:`CMPL\t\(`
 	return r, r < *y
 }
+
+// The following functions test that indexed load/store operations get generated.
+
+func idxInt8(x, y []int8, i int) {
+	var t int8
+	// amd64: `MOVBL[SZ]X\t1\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*1\), [A-Z]+[0-9]*`
+	t = x[i+1]
+	// amd64: `MOVB\t[A-Z]+[0-9]*, 1\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*1\)`
+	y[i+1] = t
+	// amd64: `MOVB\t[$]77, 1\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*1\)`
+	x[i+1] = 77
+}
+
+func idxInt16(x, y []int16, i int) {
+	var t int16
+	// amd64: `MOVWL[SZ]X\t2\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*2\), [A-Z]+[0-9]*`
+	t = x[i+1]
+	// amd64: `MOVW\t[A-Z]+[0-9]*, 2\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*2\)`
+	y[i+1] = t
+	// amd64: `MOVWL[SZ]X\t2\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*[12]\), [A-Z]+[0-9]*`
+	t = x[16*i+1]
+	// amd64: `MOVW\t[A-Z]+[0-9]*, 2\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*[12]\)`
+	y[16*i+1] = t
+	// amd64: `MOVW\t[$]77, 2\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*2\)`
+	x[i+1] = 77
+	// amd64: `MOVW\t[$]77, 2\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*[12]\)`
+	x[16*i+1] = 77
+}
+
+func idxInt32(x, y []int32, i int) {
+	var t int32
+	// amd64: `MOVL\t4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\), [A-Z]+[0-9]*`
+	t = x[i+1]
+	// amd64: `MOVL\t[A-Z]+[0-9]*, 4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\)`
+	y[i+1] = t
+	// amd64: `MOVL\t4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\), [A-Z]+[0-9]*`
+	t = x[2*i+1]
+	// amd64: `MOVL\t[A-Z]+[0-9]*, 4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)`
+	y[2*i+1] = t
+	// amd64: `MOVL\t4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*[14]\), [A-Z]+[0-9]*`
+	t = x[16*i+1]
+	// amd64: `MOVL\t[A-Z]+[0-9]*, 4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*[14]\)`
+	y[16*i+1] = t
+	// amd64: `MOVL\t[$]77, 4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\)`
+	x[i+1] = 77
+	// amd64: `MOVL\t[$]77, 4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*[14]\)`
+	x[16*i+1] = 77
+}
+
+func idxInt64(x, y []int64, i int) {
+	var t int64
+	// amd64: `MOVQ\t8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\), [A-Z]+[0-9]*`
+	t = x[i+1]
+	// amd64: `MOVQ\t[A-Z]+[0-9]*, 8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)`
+	y[i+1] = t
+	// amd64: `MOVQ\t8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*[18]\), [A-Z]+[0-9]*`
+	t = x[16*i+1]
+	// amd64: `MOVQ\t[A-Z]+[0-9]*, 8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*[18]\)`
+	y[16*i+1] = t
+	// amd64: `MOVQ\t[$]77, 8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)`
+	x[i+1] = 77
+	// amd64: `MOVQ\t[$]77, 8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*[18]\)`
+	x[16*i+1] = 77
+}
+
+func idxFloat32(x, y []float32, i int) {
+	var t float32
+	// amd64: `MOVSS\t4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\), X[0-9]+`
+	t = x[i+1]
+	// amd64: `MOVSS\tX[0-9]+, 4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*4\)`
+	y[i+1] = t
+	// amd64: `MOVSS\t4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*[14]\), X[0-9]+`
+	t = x[16*i+1]
+	// amd64: `MOVSS\tX[0-9]+, 4\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*[14]\)`
+	y[16*i+1] = t
+}
+
+func idxFloat64(x, y []float64, i int) {
+	var t float64
+	// amd64: `MOVSD\t8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\), X[0-9]+`
+	t = x[i+1]
+	// amd64: `MOVSD\tX[0-9]+, 8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\)`
+	y[i+1] = t
+	// amd64: `MOVSD\t8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*[18]\), X[0-9]+`
+	t = x[16*i+1]
+	// amd64: `MOVSD\tX[0-9]+, 8\([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*[18]\)`
+	y[16*i+1] = t
+}