Re: [PATCH v4 110/163] tcg/riscv: Drop support for add2/sub2
On 4/15/25 12:24, Richard Henderson wrote:
We now produce exactly the same code via generic expansion.
Signed-off-by: Richard Henderson
---
tcg/riscv/tcg-target-con-set.h | 1 -
tcg/riscv/tcg-target-has.h | 6 +--
tcg/riscv/tcg-target.c.inc | 86 +-
3 files changed, 3 insertions(+), 90 deletions(-)
diff --git a/tcg/riscv/tcg-target-con-set.h b/tcg/riscv/tcg-target-con-set.h
index 5ff2c2db60..0fc26d3f98 100644
--- a/tcg/riscv/tcg-target-con-set.h
+++ b/tcg/riscv/tcg-target-con-set.h
@@ -18,7 +18,6 @@ C_O1_I2(r, r, ri)
C_O1_I2(r, r, rI)
C_N1_I2(r, r, rM)
C_O1_I4(r, r, rI, rM, rM)
-C_O2_I4(r, r, rz, rz, rM, rM)
C_O0_I2(v, r)
C_O1_I1(v, r)
C_O1_I1(v, v)
diff --git a/tcg/riscv/tcg-target-has.h b/tcg/riscv/tcg-target-has.h
index b2814f8ef9..c95dc1921e 100644
--- a/tcg/riscv/tcg-target-has.h
+++ b/tcg/riscv/tcg-target-has.h
@@ -10,13 +10,11 @@
#include "host/cpuinfo.h"
/* optional instructions */
-#define TCG_TARGET_HAS_add2_i32 1
-#define TCG_TARGET_HAS_sub2_i32 1
#define TCG_TARGET_HAS_qemu_st8_i32 0
#define TCG_TARGET_HAS_extr_i64_i32 1
-#define TCG_TARGET_HAS_add2_i64 1
-#define TCG_TARGET_HAS_sub2_i64 1
+#define TCG_TARGET_HAS_add2_i64 0
+#define TCG_TARGET_HAS_sub2_i64 0
#define TCG_TARGET_HAS_qemu_ldst_i128 0
diff --git a/tcg/riscv/tcg-target.c.inc b/tcg/riscv/tcg-target.c.inc
index d74ac7587a..dce46dcba6 100644
--- a/tcg/riscv/tcg-target.c.inc
+++ b/tcg/riscv/tcg-target.c.inc
@@ -401,7 +401,7 @@ static bool tcg_target_const_match(int64_t val, int ct,
}
/*
* Sign extended from 12 bits, +/- matching: [-0x7ff, 0x7ff].
- * Used by addsub2 and movcond, which may need the negative value,
+ * Used by movcond, which may need the negative value,
* and requires the modified constant to be representable.
*/
if ((ct & TCG_CT_CONST_M12) && val >= -0x7ff && val <= 0x7ff) {
@@ -1073,67 +1073,6 @@ static bool tcg_out_sti(TCGContext *s, TCGType type,
TCGArg val,
return false;
}
-static void tcg_out_addsub2(TCGContext *s,
-TCGReg rl, TCGReg rh,
-TCGReg al, TCGReg ah,
-TCGArg bl, TCGArg bh,
-bool cbl, bool cbh, bool is_sub, bool is32bit)
-{
-const RISCVInsn opc_add = is32bit ? OPC_ADDW : OPC_ADD;
-const RISCVInsn opc_addi = is32bit ? OPC_ADDIW : OPC_ADDI;
-const RISCVInsn opc_sub = is32bit ? OPC_SUBW : OPC_SUB;
-TCGReg th = TCG_REG_TMP1;
-
-/* If we have a negative constant such that negating it would
- make the high part zero, we can (usually) eliminate one insn. */
-if (cbl && cbh && bh == -1 && bl != 0) {
-bl = -bl;
-bh = 0;
-is_sub = !is_sub;
-}
-
-/* By operating on the high part first, we get to use the final
- carry operation to move back from the temporary. */
-if (!cbh) {
-tcg_out_opc_reg(s, (is_sub ? opc_sub : opc_add), th, ah, bh);
-} else if (bh != 0 || ah == rl) {
-tcg_out_opc_imm(s, opc_addi, th, ah, (is_sub ? -bh : bh));
-} else {
-th = ah;
-}
-
-/* Note that tcg optimization should eliminate the bl == 0 case. */
-if (is_sub) {
-if (cbl) {
-tcg_out_opc_imm(s, OPC_SLTIU, TCG_REG_TMP0, al, bl);
-tcg_out_opc_imm(s, opc_addi, rl, al, -bl);
-} else {
-tcg_out_opc_reg(s, OPC_SLTU, TCG_REG_TMP0, al, bl);
-tcg_out_opc_reg(s, opc_sub, rl, al, bl);
-}
-tcg_out_opc_reg(s, opc_sub, rh, th, TCG_REG_TMP0);
-} else {
-if (cbl) {
-tcg_out_opc_imm(s, opc_addi, rl, al, bl);
-tcg_out_opc_imm(s, OPC_SLTIU, TCG_REG_TMP0, rl, bl);
-} else if (al == bl) {
-/*
- * If the input regs overlap, this is a simple doubling
- * and carry-out is the input msb. This special case is
- * required when the output reg overlaps the input,
- * but we might as well use it always.
- */
-tcg_out_opc_imm(s, OPC_SLTI, TCG_REG_TMP0, al, 0);
-tcg_out_opc_reg(s, opc_add, rl, al, al);
-} else {
-tcg_out_opc_reg(s, opc_add, rl, al, bl);
-tcg_out_opc_reg(s, OPC_SLTU, TCG_REG_TMP0,
-rl, (rl == bl ? al : bl));
-}
-tcg_out_opc_reg(s, opc_add, rh, th, TCG_REG_TMP0);
-}
-}
-
static bool tcg_out_dup_vec(TCGContext *s, TCGType type, unsigned vece,
TCGReg dst, TCGReg src)
{
@@ -2608,23 +2547,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
TCGType type,
tcg_out_ldst(s, OPC_SD, a0, a1, a2);
break;
-case INDEX_op_add2_i32:
-tcg_out_addsub2(s, a0, a1, a2, args[3], args[4], args[5],
-const_args[4], const_args[5], false, true);
-
[PATCH v4 110/163] tcg/riscv: Drop support for add2/sub2
We now produce exactly the same code via generic expansion.
Signed-off-by: Richard Henderson
---
tcg/riscv/tcg-target-con-set.h | 1 -
tcg/riscv/tcg-target-has.h | 6 +--
tcg/riscv/tcg-target.c.inc | 86 +-
3 files changed, 3 insertions(+), 90 deletions(-)
diff --git a/tcg/riscv/tcg-target-con-set.h b/tcg/riscv/tcg-target-con-set.h
index 5ff2c2db60..0fc26d3f98 100644
--- a/tcg/riscv/tcg-target-con-set.h
+++ b/tcg/riscv/tcg-target-con-set.h
@@ -18,7 +18,6 @@ C_O1_I2(r, r, ri)
C_O1_I2(r, r, rI)
C_N1_I2(r, r, rM)
C_O1_I4(r, r, rI, rM, rM)
-C_O2_I4(r, r, rz, rz, rM, rM)
C_O0_I2(v, r)
C_O1_I1(v, r)
C_O1_I1(v, v)
diff --git a/tcg/riscv/tcg-target-has.h b/tcg/riscv/tcg-target-has.h
index b2814f8ef9..c95dc1921e 100644
--- a/tcg/riscv/tcg-target-has.h
+++ b/tcg/riscv/tcg-target-has.h
@@ -10,13 +10,11 @@
#include "host/cpuinfo.h"
/* optional instructions */
-#define TCG_TARGET_HAS_add2_i32 1
-#define TCG_TARGET_HAS_sub2_i32 1
#define TCG_TARGET_HAS_qemu_st8_i32 0
#define TCG_TARGET_HAS_extr_i64_i32 1
-#define TCG_TARGET_HAS_add2_i64 1
-#define TCG_TARGET_HAS_sub2_i64 1
+#define TCG_TARGET_HAS_add2_i64 0
+#define TCG_TARGET_HAS_sub2_i64 0
#define TCG_TARGET_HAS_qemu_ldst_i128 0
diff --git a/tcg/riscv/tcg-target.c.inc b/tcg/riscv/tcg-target.c.inc
index d74ac7587a..dce46dcba6 100644
--- a/tcg/riscv/tcg-target.c.inc
+++ b/tcg/riscv/tcg-target.c.inc
@@ -401,7 +401,7 @@ static bool tcg_target_const_match(int64_t val, int ct,
}
/*
* Sign extended from 12 bits, +/- matching: [-0x7ff, 0x7ff].
- * Used by addsub2 and movcond, which may need the negative value,
+ * Used by movcond, which may need the negative value,
* and requires the modified constant to be representable.
*/
if ((ct & TCG_CT_CONST_M12) && val >= -0x7ff && val <= 0x7ff) {
@@ -1073,67 +1073,6 @@ static bool tcg_out_sti(TCGContext *s, TCGType type,
TCGArg val,
return false;
}
-static void tcg_out_addsub2(TCGContext *s,
-TCGReg rl, TCGReg rh,
-TCGReg al, TCGReg ah,
-TCGArg bl, TCGArg bh,
-bool cbl, bool cbh, bool is_sub, bool is32bit)
-{
-const RISCVInsn opc_add = is32bit ? OPC_ADDW : OPC_ADD;
-const RISCVInsn opc_addi = is32bit ? OPC_ADDIW : OPC_ADDI;
-const RISCVInsn opc_sub = is32bit ? OPC_SUBW : OPC_SUB;
-TCGReg th = TCG_REG_TMP1;
-
-/* If we have a negative constant such that negating it would
- make the high part zero, we can (usually) eliminate one insn. */
-if (cbl && cbh && bh == -1 && bl != 0) {
-bl = -bl;
-bh = 0;
-is_sub = !is_sub;
-}
-
-/* By operating on the high part first, we get to use the final
- carry operation to move back from the temporary. */
-if (!cbh) {
-tcg_out_opc_reg(s, (is_sub ? opc_sub : opc_add), th, ah, bh);
-} else if (bh != 0 || ah == rl) {
-tcg_out_opc_imm(s, opc_addi, th, ah, (is_sub ? -bh : bh));
-} else {
-th = ah;
-}
-
-/* Note that tcg optimization should eliminate the bl == 0 case. */
-if (is_sub) {
-if (cbl) {
-tcg_out_opc_imm(s, OPC_SLTIU, TCG_REG_TMP0, al, bl);
-tcg_out_opc_imm(s, opc_addi, rl, al, -bl);
-} else {
-tcg_out_opc_reg(s, OPC_SLTU, TCG_REG_TMP0, al, bl);
-tcg_out_opc_reg(s, opc_sub, rl, al, bl);
-}
-tcg_out_opc_reg(s, opc_sub, rh, th, TCG_REG_TMP0);
-} else {
-if (cbl) {
-tcg_out_opc_imm(s, opc_addi, rl, al, bl);
-tcg_out_opc_imm(s, OPC_SLTIU, TCG_REG_TMP0, rl, bl);
-} else if (al == bl) {
-/*
- * If the input regs overlap, this is a simple doubling
- * and carry-out is the input msb. This special case is
- * required when the output reg overlaps the input,
- * but we might as well use it always.
- */
-tcg_out_opc_imm(s, OPC_SLTI, TCG_REG_TMP0, al, 0);
-tcg_out_opc_reg(s, opc_add, rl, al, al);
-} else {
-tcg_out_opc_reg(s, opc_add, rl, al, bl);
-tcg_out_opc_reg(s, OPC_SLTU, TCG_REG_TMP0,
-rl, (rl == bl ? al : bl));
-}
-tcg_out_opc_reg(s, opc_add, rh, th, TCG_REG_TMP0);
-}
-}
-
static bool tcg_out_dup_vec(TCGContext *s, TCGType type, unsigned vece,
TCGReg dst, TCGReg src)
{
@@ -2608,23 +2547,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
TCGType type,
tcg_out_ldst(s, OPC_SD, a0, a1, a2);
break;
-case INDEX_op_add2_i32:
-tcg_out_addsub2(s, a0, a1, a2, args[3], args[4], args[5],
-const_args[4], const_args[5], false, true);
-break;
-case INDEX_op_add2_i64:
-tcg_out_addsub2(s, a0, a1, a2, args[3],
