tcg: Add opcode for ctpop

The number of actual invocations of ctpop itself does not warrent
an opcode, but it is very helpful for POWER7 to use in generating
an expansion for ctz.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>

15 files changed, 69 insertions, 0 deletions

diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h
index 9d6b00fbba..1a5ea23844 100644
--- a/tcg/aarch64/tcg-target.h
+++ b/tcg/aarch64/tcg-target.h
@@ -64,6 +64,7 @@ typedef enum {
 #define TCG_TARGET_HAS_nor_i32          0
 #define TCG_TARGET_HAS_clz_i32          1
 #define TCG_TARGET_HAS_ctz_i32          1
+#define TCG_TARGET_HAS_ctpop_i32        0
 #define TCG_TARGET_HAS_deposit_i32      1
 #define TCG_TARGET_HAS_extract_i32      1
 #define TCG_TARGET_HAS_sextract_i32     1
@@ -98,6 +99,7 @@ typedef enum {
 #define TCG_TARGET_HAS_nor_i64          0
 #define TCG_TARGET_HAS_clz_i64          1
 #define TCG_TARGET_HAS_ctz_i64          1
+#define TCG_TARGET_HAS_ctpop_i64        0
 #define TCG_TARGET_HAS_deposit_i64      1
 #define TCG_TARGET_HAS_extract_i64      1
 #define TCG_TARGET_HAS_sextract_i64     1
diff --git a/tcg/arm/tcg-target.h b/tcg/arm/tcg-target.h
index 4cb94dc103..09a19c6f35 100644
--- a/tcg/arm/tcg-target.h
+++ b/tcg/arm/tcg-target.h
@@ -112,6 +112,7 @@ extern bool use_idiv_instructions;
 #define TCG_TARGET_HAS_nor_i32          0
 #define TCG_TARGET_HAS_clz_i32          use_armv5t_instructions
 #define TCG_TARGET_HAS_ctz_i32          use_armv7_instructions
+#define TCG_TARGET_HAS_ctpop_i32        0
 #define TCG_TARGET_HAS_deposit_i32      use_armv7_instructions
 #define TCG_TARGET_HAS_extract_i32      use_armv7_instructions
 #define TCG_TARGET_HAS_sextract_i32     use_armv7_instructions
diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h
index 8fff287caa..b8f73f5382 100644
--- a/tcg/i386/tcg-target.h
+++ b/tcg/i386/tcg-target.h
@@ -95,6 +95,7 @@ extern bool have_bmi1;
 #define TCG_TARGET_HAS_nor_i32          0
 #define TCG_TARGET_HAS_clz_i32          1
 #define TCG_TARGET_HAS_ctz_i32          1
+#define TCG_TARGET_HAS_ctpop_i32        0
 #define TCG_TARGET_HAS_deposit_i32      1
 #define TCG_TARGET_HAS_extract_i32      1
 #define TCG_TARGET_HAS_sextract_i32     1
@@ -129,6 +130,7 @@ extern bool have_bmi1;
 #define TCG_TARGET_HAS_nor_i64          0
 #define TCG_TARGET_HAS_clz_i64          1
 #define TCG_TARGET_HAS_ctz_i64          1
+#define TCG_TARGET_HAS_ctpop_i64        0
 #define TCG_TARGET_HAS_deposit_i64      1
 #define TCG_TARGET_HAS_extract_i64      1
 #define TCG_TARGET_HAS_sextract_i64     0
diff --git a/tcg/ia64/tcg-target.h b/tcg/ia64/tcg-target.h
index 9a829ae751..42aea03a8b 100644
--- a/tcg/ia64/tcg-target.h
+++ b/tcg/ia64/tcg-target.h
@@ -144,6 +144,8 @@ typedef enum {
 #define TCG_TARGET_HAS_clz_i64          0
 #define TCG_TARGET_HAS_ctz_i32          0
 #define TCG_TARGET_HAS_ctz_i64          0
+#define TCG_TARGET_HAS_ctpop_i32        0
+#define TCG_TARGET_HAS_ctpop_i64        0
 #define TCG_TARGET_HAS_nor_i64          1
 #define TCG_TARGET_HAS_orc_i32          1
 #define TCG_TARGET_HAS_orc_i64          1
diff --git a/tcg/mips/tcg-target.h b/tcg/mips/tcg-target.h
index a680f16e47..f46d64a3a7 100644
--- a/tcg/mips/tcg-target.h
+++ b/tcg/mips/tcg-target.h
@@ -165,6 +165,7 @@ extern bool use_mips32r2_instructions;
 #define TCG_TARGET_HAS_rot_i32          use_mips32r2_instructions
 #define TCG_TARGET_HAS_clz_i32          use_mips32r2_instructions
 #define TCG_TARGET_HAS_ctz_i32          0
+#define TCG_TARGET_HAS_ctpop_i32        0
 
 #if TCG_TARGET_REG_BITS == 64
 #define TCG_TARGET_HAS_movcond_i64      use_movnz_instructions
@@ -179,6 +180,7 @@ extern bool use_mips32r2_instructions;
 #define TCG_TARGET_HAS_rot_i64          use_mips32r2_instructions
 #define TCG_TARGET_HAS_clz_i64          use_mips32r2_instructions
 #define TCG_TARGET_HAS_ctz_i64          0
+#define TCG_TARGET_HAS_ctpop_i64        0
 #endif
 
 /* optional instructions automatically implemented */
diff --git a/tcg/optimize.c b/tcg/optimize.c
index e7ecce478e..adfc56ce62 100644
--- a/tcg/optimize.c
+++ b/tcg/optimize.c
@@ -308,6 +308,12 @@ static TCGArg do_constant_folding_2(TCGOpcode op, TCGArg x, TCGArg y)
     case INDEX_op_ctz_i64:
         return x ? ctz64(x) : y;
 
+    case INDEX_op_ctpop_i32:
+        return ctpop32(x);
+
+    case INDEX_op_ctpop_i64:
+        return ctpop64(x);
+
     CASE_OP_32_64(ext8s):
         return (int8_t)x;
 
@@ -918,6 +924,13 @@ void tcg_optimize(TCGContext *s)
             mask = temps[args[2]].mask | 63;
             break;
 
+        case INDEX_op_ctpop_i32:
+            mask = 32 | 31;
+            break;
+        case INDEX_op_ctpop_i64:
+            mask = 64 | 63;
+            break;
+
         CASE_OP_32_64(setcond):
         case INDEX_op_setcond2_i32:
             mask = 1;
@@ -1031,6 +1044,7 @@ void tcg_optimize(TCGContext *s)
         CASE_OP_32_64(ext8u):
         CASE_OP_32_64(ext16s):
         CASE_OP_32_64(ext16u):
+        CASE_OP_32_64(ctpop):
         case INDEX_op_ext32s_i64:
         case INDEX_op_ext32u_i64:
         case INDEX_op_ext_i32_i64:
diff --git a/tcg/ppc/tcg-target.h b/tcg/ppc/tcg-target.h
index c798c9c85d..57e66cf3ed 100644
--- a/tcg/ppc/tcg-target.h
+++ b/tcg/ppc/tcg-target.h
@@ -72,6 +72,7 @@ extern bool have_isa_3_00;
 #define TCG_TARGET_HAS_nor_i32          1
 #define TCG_TARGET_HAS_clz_i32          1
 #define TCG_TARGET_HAS_ctz_i32          have_isa_3_00
+#define TCG_TARGET_HAS_ctpop_i32        0
 #define TCG_TARGET_HAS_deposit_i32      1
 #define TCG_TARGET_HAS_extract_i32      1
 #define TCG_TARGET_HAS_sextract_i32     0
@@ -107,6 +108,7 @@ extern bool have_isa_3_00;
 #define TCG_TARGET_HAS_nor_i64          1
 #define TCG_TARGET_HAS_clz_i64          1
 #define TCG_TARGET_HAS_ctz_i64          have_isa_3_00
+#define TCG_TARGET_HAS_ctpop_i64        0
 #define TCG_TARGET_HAS_deposit_i64      1
 #define TCG_TARGET_HAS_extract_i64      1
 #define TCG_TARGET_HAS_sextract_i64     0
diff --git a/tcg/s390/tcg-target.h b/tcg/s390/tcg-target.h
index 22500ba858..cbdd2a6275 100644
--- a/tcg/s390/tcg-target.h
+++ b/tcg/s390/tcg-target.h
@@ -79,6 +79,7 @@ extern uint64_t s390_facilities;
 #define TCG_TARGET_HAS_nor_i32        0
 #define TCG_TARGET_HAS_clz_i32        0
 #define TCG_TARGET_HAS_ctz_i32        0
+#define TCG_TARGET_HAS_ctpop_i32      0
 #define TCG_TARGET_HAS_deposit_i32    (s390_facilities & FACILITY_GEN_INST_EXT)
 #define TCG_TARGET_HAS_extract_i32    (s390_facilities & FACILITY_GEN_INST_EXT)
 #define TCG_TARGET_HAS_sextract_i32   0
@@ -112,6 +113,7 @@ extern uint64_t s390_facilities;
 #define TCG_TARGET_HAS_nor_i64        0
 #define TCG_TARGET_HAS_clz_i64        (s390_facilities & FACILITY_EXT_IMM)
 #define TCG_TARGET_HAS_ctz_i64        0
+#define TCG_TARGET_HAS_ctpop_i64      0
 #define TCG_TARGET_HAS_deposit_i64    (s390_facilities & FACILITY_GEN_INST_EXT)
 #define TCG_TARGET_HAS_extract_i64    (s390_facilities & FACILITY_GEN_INST_EXT)
 #define TCG_TARGET_HAS_sextract_i64   0
diff --git a/tcg/sparc/tcg-target.h b/tcg/sparc/tcg-target.h
index 340837ae9a..b8b74f96ff 100644
--- a/tcg/sparc/tcg-target.h
+++ b/tcg/sparc/tcg-target.h
@@ -112,6 +112,7 @@ extern bool use_vis3_instructions;
 #define TCG_TARGET_HAS_nor_i32          0
 #define TCG_TARGET_HAS_clz_i32          0
 #define TCG_TARGET_HAS_ctz_i32          0
+#define TCG_TARGET_HAS_ctpop_i32        0
 #define TCG_TARGET_HAS_deposit_i32      0
 #define TCG_TARGET_HAS_extract_i32      0
 #define TCG_TARGET_HAS_sextract_i32     0
@@ -146,6 +147,7 @@ extern bool use_vis3_instructions;
 #define TCG_TARGET_HAS_nor_i64          0
 #define TCG_TARGET_HAS_clz_i64          0
 #define TCG_TARGET_HAS_ctz_i64          0
+#define TCG_TARGET_HAS_ctpop_i64        0
 #define TCG_TARGET_HAS_deposit_i64      0
 #define TCG_TARGET_HAS_extract_i64      0
 #define TCG_TARGET_HAS_sextract_i64     0
diff --git a/tcg/tcg-op.c b/tcg/tcg-op.c
index 620e26897f..6f4b1b62cb 100644
--- a/tcg/tcg-op.c
+++ b/tcg/tcg-op.c
@@ -550,6 +550,21 @@ void tcg_gen_clrsb_i32(TCGv_i32 ret, TCGv_i32 arg)
     }
 }
 
+void tcg_gen_ctpop_i32(TCGv_i32 ret, TCGv_i32 arg1)
+{
+    if (TCG_TARGET_HAS_ctpop_i32) {
+        tcg_gen_op2_i32(INDEX_op_ctpop_i32, ret, arg1);
+    } else if (TCG_TARGET_HAS_ctpop_i64) {
+        TCGv_i64 t = tcg_temp_new_i64();
+        tcg_gen_extu_i32_i64(t, arg1);
+        tcg_gen_ctpop_i64(t, t);
+        tcg_gen_extrl_i64_i32(ret, t);
+        tcg_temp_free_i64(t);
+    } else {
+        gen_helper_ctpop_i32(ret, arg1);
+    }
+}
+
 void tcg_gen_rotl_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2)
 {
     if (TCG_TARGET_HAS_rot_i32) {
@@ -1874,6 +1889,20 @@ void tcg_gen_clrsb_i64(TCGv_i64 ret, TCGv_i64 arg)
     }
 }
 
+void tcg_gen_ctpop_i64(TCGv_i64 ret, TCGv_i64 arg1)
+{
+    if (TCG_TARGET_HAS_ctpop_i64) {
+        tcg_gen_op2_i64(INDEX_op_ctpop_i64, ret, arg1);
+    } else if (TCG_TARGET_REG_BITS == 32 && TCG_TARGET_HAS_ctpop_i32) {
+        tcg_gen_ctpop_i32(TCGV_HIGH(ret), TCGV_HIGH(arg1));
+        tcg_gen_ctpop_i32(TCGV_LOW(ret), TCGV_LOW(arg1));
+        tcg_gen_add_i32(TCGV_LOW(ret), TCGV_LOW(ret), TCGV_HIGH(ret));
+        tcg_gen_movi_i32(TCGV_HIGH(ret), 0);
+    } else {
+        gen_helper_ctpop_i64(ret, arg1);
+    }
+}
+
 void tcg_gen_rotl_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2)
 {
     if (TCG_TARGET_HAS_rot_i64) {
diff --git a/tcg/tcg-op.h b/tcg/tcg-op.h
index c2f3db9288..c68e300a68 100644
--- a/tcg/tcg-op.h
+++ b/tcg/tcg-op.h
@@ -291,6 +291,7 @@ void tcg_gen_ctz_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2);
 void tcg_gen_clzi_i32(TCGv_i32 ret, TCGv_i32 arg1, uint32_t arg2);
 void tcg_gen_ctzi_i32(TCGv_i32 ret, TCGv_i32 arg1, uint32_t arg2);
 void tcg_gen_clrsb_i32(TCGv_i32 ret, TCGv_i32 arg);
+void tcg_gen_ctpop_i32(TCGv_i32 a1, TCGv_i32 a2);
 void tcg_gen_rotl_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2);
 void tcg_gen_rotli_i32(TCGv_i32 ret, TCGv_i32 arg1, unsigned arg2);
 void tcg_gen_rotr_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2);
@@ -479,6 +480,7 @@ void tcg_gen_ctz_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2);
 void tcg_gen_clzi_i64(TCGv_i64 ret, TCGv_i64 arg1, uint64_t arg2);
 void tcg_gen_ctzi_i64(TCGv_i64 ret, TCGv_i64 arg1, uint64_t arg2);
 void tcg_gen_clrsb_i64(TCGv_i64 ret, TCGv_i64 arg);
+void tcg_gen_ctpop_i64(TCGv_i64 a1, TCGv_i64 a2);
 void tcg_gen_rotl_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2);
 void tcg_gen_rotli_i64(TCGv_i64 ret, TCGv_i64 arg1, unsigned arg2);
 void tcg_gen_rotr_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2);
@@ -973,6 +975,7 @@ void tcg_gen_atomic_xor_fetch_i64(TCGv_i64, TCGv, TCGv_i64, TCGArg, TCGMemOp);
 #define tcg_gen_clzi_tl tcg_gen_clzi_i64
 #define tcg_gen_ctzi_tl tcg_gen_ctzi_i64
 #define tcg_gen_clrsb_tl tcg_gen_clrsb_i64
+#define tcg_gen_ctpop_tl tcg_gen_ctpop_i64
 #define tcg_gen_rotl_tl tcg_gen_rotl_i64
 #define tcg_gen_rotli_tl tcg_gen_rotli_i64
 #define tcg_gen_rotr_tl tcg_gen_rotr_i64
@@ -1069,6 +1072,7 @@ void tcg_gen_atomic_xor_fetch_i64(TCGv_i64, TCGv, TCGv_i64, TCGArg, TCGMemOp);
 #define tcg_gen_clzi_tl tcg_gen_clzi_i32
 #define tcg_gen_ctzi_tl tcg_gen_ctzi_i32
 #define tcg_gen_clrsb_tl tcg_gen_clrsb_i32
+#define tcg_gen_ctpop_tl tcg_gen_ctpop_i32
 #define tcg_gen_rotl_tl tcg_gen_rotl_i32
 #define tcg_gen_rotli_tl tcg_gen_rotli_i32
 #define tcg_gen_rotr_tl tcg_gen_rotr_i32
diff --git a/tcg/tcg-opc.h b/tcg/tcg-opc.h
index d00db4f47c..f06f89405e 100644
--- a/tcg/tcg-opc.h
+++ b/tcg/tcg-opc.h
@@ -106,6 +106,7 @@ DEF(nand_i32, 1, 2, 0, IMPL(TCG_TARGET_HAS_nand_i32))
 DEF(nor_i32, 1, 2, 0, IMPL(TCG_TARGET_HAS_nor_i32))
 DEF(clz_i32, 1, 2, 0, IMPL(TCG_TARGET_HAS_clz_i32))
 DEF(ctz_i32, 1, 2, 0, IMPL(TCG_TARGET_HAS_ctz_i32))
+DEF(ctpop_i32, 1, 1, 0, IMPL(TCG_TARGET_HAS_ctpop_i32))
 
 DEF(mov_i64, 1, 1, 0, TCG_OPF_64BIT | TCG_OPF_NOT_PRESENT)
 DEF(movi_i64, 1, 0, 1, TCG_OPF_64BIT | TCG_OPF_NOT_PRESENT)
@@ -175,6 +176,7 @@ DEF(nand_i64, 1, 2, 0, IMPL64 | IMPL(TCG_TARGET_HAS_nand_i64))
 DEF(nor_i64, 1, 2, 0, IMPL64 | IMPL(TCG_TARGET_HAS_nor_i64))
 DEF(clz_i64, 1, 2, 0, IMPL64 | IMPL(TCG_TARGET_HAS_clz_i64))
 DEF(ctz_i64, 1, 2, 0, IMPL64 | IMPL(TCG_TARGET_HAS_ctz_i64))
+DEF(ctpop_i64, 1, 1, 0, IMPL64 | IMPL(TCG_TARGET_HAS_ctpop_i64))
 
 DEF(add2_i64, 2, 4, 0, IMPL64 | IMPL(TCG_TARGET_HAS_add2_i64))
 DEF(sub2_i64, 2, 4, 0, IMPL64 | IMPL(TCG_TARGET_HAS_sub2_i64))
diff --git a/tcg/tcg-runtime.h b/tcg/tcg-runtime.h
index 0d30f1a90c..114ea6fecf 100644
--- a/tcg/tcg-runtime.h
+++ b/tcg/tcg-runtime.h
@@ -21,6 +21,8 @@ DEF_HELPER_FLAGS_2(clz_i64, TCG_CALL_NO_RWG_SE, i64, i64, i64)
 DEF_HELPER_FLAGS_2(ctz_i64, TCG_CALL_NO_RWG_SE, i64, i64, i64)
 DEF_HELPER_FLAGS_1(clrsb_i32, TCG_CALL_NO_RWG_SE, i32, i32)
 DEF_HELPER_FLAGS_1(clrsb_i64, TCG_CALL_NO_RWG_SE, i64, i64)
+DEF_HELPER_FLAGS_1(ctpop_i32, TCG_CALL_NO_RWG_SE, i32, i32)
+DEF_HELPER_FLAGS_1(ctpop_i64, TCG_CALL_NO_RWG_SE, i64, i64)
 
 DEF_HELPER_FLAGS_1(exit_atomic, TCG_CALL_NO_WG, noreturn, env)
 
diff --git a/tcg/tcg.h b/tcg/tcg.h
index e0262822cb..631c6f69b1 100644
--- a/tcg/tcg.h
+++ b/tcg/tcg.h
@@ -113,6 +113,7 @@ typedef uint64_t TCGRegSet;
 #define TCG_TARGET_HAS_nor_i64          0
 #define TCG_TARGET_HAS_clz_i64          0
 #define TCG_TARGET_HAS_ctz_i64          0
+#define TCG_TARGET_HAS_ctpop_i64        0
 #define TCG_TARGET_HAS_deposit_i64      0
 #define TCG_TARGET_HAS_extract_i64      0
 #define TCG_TARGET_HAS_sextract_i64     0
diff --git a/tcg/tci/tcg-target.h b/tcg/tci/tcg-target.h
index 0646444b87..838bf3a858 100644
--- a/tcg/tci/tcg-target.h
+++ b/tcg/tci/tcg-target.h
@@ -76,6 +76,7 @@
 #define TCG_TARGET_HAS_nor_i32          0
 #define TCG_TARGET_HAS_clz_i32          0
 #define TCG_TARGET_HAS_ctz_i32          0
+#define TCG_TARGET_HAS_ctpop_i32        0
 #define TCG_TARGET_HAS_neg_i32          1
 #define TCG_TARGET_HAS_not_i32          1
 #define TCG_TARGET_HAS_orc_i32          0
@@ -108,6 +109,7 @@
 #define TCG_TARGET_HAS_nor_i64          0
 #define TCG_TARGET_HAS_clz_i64          0
 #define TCG_TARGET_HAS_ctz_i64          0
+#define TCG_TARGET_HAS_ctpop_i64        0
 #define TCG_TARGET_HAS_neg_i64          1
 #define TCG_TARGET_HAS_not_i64          1
 #define TCG_TARGET_HAS_orc_i64          0