cmd/compile: simplify specific addition operations using the ADDV16 instruction

On loong64, the addi.d instruction can only directly handle 12-bit
immediate numbers. If a larger immediate number needs to be processed,
it must first be placed in a register, and then the add.d instruction
is used to complete the processing of the larger immediate number.
If a larger immediate number c satisfies is32Bit(c) && c&0xffff == 0,
then the ADDV16 instruction can be used to complete the addition operation.

Removes 164 instructions from the go binary on loong64.

Change-Id: I404de93cc4eaaa12fe424f5a0d61b03231215d1a
Reviewed-on: https://go-review.googlesource.com/c/go/+/700536
Reviewed-by: Meidan Li <[email protected]>
Reviewed-by: Keith Randall <[email protected]>
Reviewed-by: Keith Randall <[email protected]>
Auto-Submit: Michael Pratt <[email protected]>
Reviewed-by: abner chenc <[email protected]>
LUCI-TryBot-Result: Go LUCI <[email protected]>
Reviewed-by: Michael Pratt <[email protected]>

Author: sophie-zhao

src/cmd/compile/internal/loong64/ssa.go

@@ -276,6 +276,7 @@ func ssaGenValue(s *ssagen.State, v *ssa.Value) {
 		p.To.Type = obj.TYPE_REG
 		p.To.Reg = v.Reg()
 	case ssa.OpLOONG64ADDVconst,
+		ssa.OpLOONG64ADDV16const,
 		ssa.OpLOONG64SUBVconst,
 		ssa.OpLOONG64ANDconst,
 		ssa.OpLOONG64ORconst,

src/cmd/compile/internal/ssa/_gen/LOONG64.rules

@@ -794,6 +794,7 @@
 (SUBVconst [c] (SUBVconst [d] x)) && is32Bit(-c-d) => (ADDVconst [-c-d] x)
 (SUBVconst [c] (ADDVconst [d] x)) && is32Bit(-c+d) => (ADDVconst [-c+d] x)
 (SUBV (MOVVconst [c]) (NEGV (SUBVconst [d] x))) => (ADDVconst [c-d] x)
+(ADDVconst [c] x) && is32Bit(c) && c&0xffff == 0 && c != 0 => (ADDV16const [c] x)
 (SLLVconst [c] (MOVVconst [d]))  => (MOVVconst [d<<uint64(c)])
 (SRLVconst [c] (MOVVconst [d]))  => (MOVVconst [int64(uint64(d)>>uint64(c))])
 (SRAVconst [c] (MOVVconst [d]))  => (MOVVconst [d>>uint64(c)])

src/cmd/compile/internal/ssa/_gen/LOONG64Ops.go

@@ -189,10 +189,11 @@ func init() {
 		{name: "VPCNT16", argLength: 1, reg: fp11, asm: "VPCNTH"}, // count set bits for each 16-bit unit and store the result in each 16-bit unit
 
 		// binary ops
-		{name: "ADDV", argLength: 2, reg: gp21, asm: "ADDVU", commutative: true},   // arg0 + arg1
-		{name: "ADDVconst", argLength: 1, reg: gp11sp, asm: "ADDVU", aux: "Int64"}, // arg0 + auxInt. auxInt is 32-bit, also in other *const ops.
-		{name: "SUBV", argLength: 2, reg: gp21, asm: "SUBVU"},                      // arg0 - arg1
-		{name: "SUBVconst", argLength: 1, reg: gp11, asm: "SUBVU", aux: "Int64"},   // arg0 - auxInt
+		{name: "ADDV", argLength: 2, reg: gp21, asm: "ADDVU", commutative: true},      // arg0 + arg1
+		{name: "ADDVconst", argLength: 1, reg: gp11sp, asm: "ADDVU", aux: "Int64"},    // arg0 + auxInt. auxInt is 32-bit, also in other *const ops.
+		{name: "ADDV16const", argLength: 1, reg: gp11sp, asm: "ADDV16", aux: "Int64"}, // arg0 + auxInt. auxInt is signed 32-bit and is a multiple of 65536, also in other *const ops.
+		{name: "SUBV", argLength: 2, reg: gp21, asm: "SUBVU"},                         // arg0 - arg1
+		{name: "SUBVconst", argLength: 1, reg: gp11, asm: "SUBVU", aux: "Int64"},      // arg0 - auxInt
 
 		{name: "MULV", argLength: 2, reg: gp21, asm: "MULV", commutative: true, typ: "Int64"},      // arg0 * arg1
 		{name: "MULHV", argLength: 2, reg: gp21, asm: "MULHV", commutative: true, typ: "Int64"},    // (arg0 * arg1) >> 64, signed

src/cmd/compile/internal/ssa/opGen.go

@@ -51,6 +51,11 @@ func AddLargeConst(a uint64, out []uint64) {
 	out[9] = a - 32769
 }
 
+func AddLargeConst2(a int, out []int) {
+	// loong64: -"ADDVU","ADDV16"
+	out[0] = a + 0x10000
+}
+
 // ----------------- //
 //    Subtraction    //
 // ----------------- //