cmd/link, cmd/internal/obj/loong64: support the PCALIGN directive

Allow writing `PCALIGN $imm` where imm is a power-of-2 between 8 and
2048 (inclusive), for ensuring that the following instruction is
placed at an imm-byte boundary relative to the beginning of the
function. If the PC is not sufficiently aligned, NOOPs will be
inserted to make it so, otherwise the directive will do nothing.

This could be useful for both asm performance hand-tuning, and future
scenarios where a certain bigger alignment might be required.

Change-Id: Iad6244669a3d5adea88eceb0dc7be1af4f0d4fc9
Reviewed-on: https://go-review.googlesource.com/c/go/+/479815
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
Run-TryBot: WANG Xuerui <git@xen0n.name>
Auto-Submit: Ian Lance Taylor <iant@golang.org>
Reviewed-by: abner chenc <chenguoqi@loongson.cn>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>

1 files changed, 41 insertions, 4 deletions

diff --git a/src/cmd/internal/obj/loong64/asm.go b/src/cmd/internal/obj/loong64/asm.go
index 32b13d91c0..e23165a874 100644
--- a/src/cmd/internal/obj/loong64/asm.go
+++ b/src/cmd/internal/obj/loong64/asm.go
@@ -345,6 +345,7 @@ var optab = []Optab{
 	{ARDTIMED, C_NONE, C_NONE, C_NONE, C_REG, C_REG, 62, 4, 0, 0},
 
 	{obj.AUNDEF, C_NONE, C_NONE, C_NONE, C_NONE, C_NONE, 49, 4, 0, 0},
+	{obj.APCALIGN, C_SCON, C_NONE, C_NONE, C_NONE, C_NONE, 0, 0, 0, 0},
 	{obj.APCDATA, C_LCON, C_NONE, C_NONE, C_LCON, C_NONE, 0, 0, 0, 0},
 	{obj.APCDATA, C_DCON, C_NONE, C_NONE, C_DCON, C_NONE, 0, 0, 0, 0},
 	{obj.AFUNCDATA, C_SCON, C_NONE, C_NONE, C_ADDR, C_NONE, 0, 0, 0, 0},
@@ -359,6 +360,15 @@ var optab = []Optab{
 	{obj.AXXX, C_NONE, C_NONE, C_NONE, C_NONE, C_NONE, 0, 4, 0, 0},
 }
 
+// pcAlignPadLength returns the number of bytes required to align pc to alignedValue,
+// reporting an error if alignedValue is not a power of two or is out of range.
+func pcAlignPadLength(ctxt *obj.Link, pc int64, alignedValue int64) int {
+	if !((alignedValue&(alignedValue-1) == 0) && 8 <= alignedValue && alignedValue <= 2048) {
+		ctxt.Diag("alignment value of an instruction must be a power of two and in the range [8, 2048], got %d\n", alignedValue)
+	}
+	return int(-pc & (alignedValue - 1))
+}
+
 var oprange [ALAST & obj.AMask][]Optab
 
 var xcmp [C_NCLASS][C_NCLASS]bool
@@ -390,10 +400,20 @@ func span0(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 		o = c.oplook(p)
 		m = int(o.size)
 		if m == 0 {
-			if p.As != obj.ANOP && p.As != obj.AFUNCDATA && p.As != obj.APCDATA {
+			switch p.As {
+			case obj.APCALIGN:
+				alignedValue := p.From.Offset
+				m = pcAlignPadLength(ctxt, pc, alignedValue)
+				// Update the current text symbol alignment value.
+				if int32(alignedValue) > cursym.Func().Align {
+					cursym.Func().Align = int32(alignedValue)
+				}
+				break
+			case obj.ANOP, obj.AFUNCDATA, obj.APCDATA:
+				continue
+			default:
 				c.ctxt.Diag("zero-width instruction\n%v", p)
 			}
-			continue
 		}
 
 		pc += int64(m)
@@ -443,10 +463,16 @@ func span0(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 
 			m = int(o.size)
 			if m == 0 {
-				if p.As != obj.ANOP && p.As != obj.AFUNCDATA && p.As != obj.APCDATA {
+				switch p.As {
+				case obj.APCALIGN:
+					alignedValue := p.From.Offset
+					m = pcAlignPadLength(ctxt, pc, alignedValue)
+					break
+				case obj.ANOP, obj.AFUNCDATA, obj.APCDATA:
+					continue
+				default:
 					c.ctxt.Diag("zero-width instruction\n%v", p)
 				}
-				continue
 			}
 
 			pc += int64(m)
@@ -470,6 +496,16 @@ func span0(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 		if int(o.size) > 4*len(out) {
 			log.Fatalf("out array in span0 is too small, need at least %d for %v", o.size/4, p)
 		}
+		if p.As == obj.APCALIGN {
+			alignedValue := p.From.Offset
+			v := pcAlignPadLength(c.ctxt, p.Pc, alignedValue)
+			for i = 0; i < int32(v/4); i++ {
+				// emit ANOOP instruction by the padding size
+				c.ctxt.Arch.ByteOrder.PutUint32(bp, c.oprrr(ANOOP))
+				bp = bp[4:]
+			}
+			continue
+		}
 		c.asmout(p, o, out[:])
 		for i = 0; i < int32(o.size/4); i++ {
 			c.ctxt.Arch.ByteOrder.PutUint32(bp, out[i])
@@ -1062,6 +1098,7 @@ func buildop(ctxt *obj.Link) {
 			obj.ATEXT,
 			obj.AUNDEF,
 			obj.AFUNCDATA,
+			obj.APCALIGN,
 			obj.APCDATA,
 			obj.ADUFFZERO,
 			obj.ADUFFCOPY: