[Qemu-devel] [PULL 02/11] tcg-sparc: Use ADDXC in addsub2_i64

[Richard Henderson <rth@twiddle.net>]

On T4 and newer Sparc chips we have an add-with-carry insn
that takes its input from %xcc instead of %icc.

Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 disas/sparc.c          |  3 +++
 include/elf.h          | 37 +++++++++++++++++++++++++++++--------
 tcg/sparc/tcg-target.c | 28 +++++++++++++++++++++++-----
 tcg/sparc/tcg-target.h |  6 ++++++
 4 files changed, 61 insertions(+), 13 deletions(-)

diff --git a/disas/sparc.c b/disas/sparc.c
index 8eb22e6..092e1b6 100644
--- a/disas/sparc.c
+++ b/disas/sparc.c
@@ -2042,6 +2042,9 @@ IMPDEP ("impdep2", 0x37),
 
 #undef IMPDEP
 
+{ "addxc", F3F(2, 0x36, 0x011), F3F(~2, ~0x36, ~0x011), "1,2,d", 0, v9b },
+{ "addxccc", F3F(2, 0x36, 0x013), F3F(~2, ~0x36, ~0x013), "1,2,d", 0, v9b },
+
 };
 
 static const int sparc_num_opcodes = ((sizeof sparc_opcodes)/(sizeof sparc_opcodes[0]));
diff --git a/include/elf.h b/include/elf.h
index 70107f0..a516584 100644
--- a/include/elf.h
+++ b/include/elf.h
@@ -473,14 +473,35 @@ typedef struct {
 #define PPC_FEATURE_TRUE_LE             0x00000002
 #define PPC_FEATURE_PPC_LE              0x00000001
 
-/* Bits present in AT_HWCAP, primarily for Sparc32.  */
-
-#define HWCAP_SPARC_FLUSH       1    /* CPU supports flush instruction. */
-#define HWCAP_SPARC_STBAR       2
-#define HWCAP_SPARC_SWAP        4
-#define HWCAP_SPARC_MULDIV      8
-#define HWCAP_SPARC_V9		16
-#define HWCAP_SPARC_ULTRA3	32
+/* Bits present in AT_HWCAP for Sparc.  */
+
+#define HWCAP_SPARC_FLUSH               0x00000001
+#define HWCAP_SPARC_STBAR               0x00000002
+#define HWCAP_SPARC_SWAP                0x00000004
+#define HWCAP_SPARC_MULDIV              0x00000008
+#define HWCAP_SPARC_V9                  0x00000010
+#define HWCAP_SPARC_ULTRA3              0x00000020
+#define HWCAP_SPARC_BLKINIT             0x00000040
+#define HWCAP_SPARC_N2                  0x00000080
+#define HWCAP_SPARC_MUL32               0x00000100
+#define HWCAP_SPARC_DIV32               0x00000200
+#define HWCAP_SPARC_FSMULD              0x00000400
+#define HWCAP_SPARC_V8PLUS              0x00000800
+#define HWCAP_SPARC_POPC                0x00001000
+#define HWCAP_SPARC_VIS                 0x00002000
+#define HWCAP_SPARC_VIS2                0x00004000
+#define HWCAP_SPARC_ASI_BLK_INIT        0x00008000
+#define HWCAP_SPARC_FMAF                0x00010000
+#define HWCAP_SPARC_VIS3                0x00020000
+#define HWCAP_SPARC_HPC                 0x00040000
+#define HWCAP_SPARC_RANDOM              0x00080000
+#define HWCAP_SPARC_TRANS               0x00100000
+#define HWCAP_SPARC_FJFMAU              0x00200000
+#define HWCAP_SPARC_IMA                 0x00400000
+#define HWCAP_SPARC_ASI_CACHE_SPARING   0x00800000
+#define HWCAP_SPARC_PAUSE               0x01000000
+#define HWCAP_SPARC_CBCOND              0x02000000
+#define HWCAP_SPARC_CRYPTO              0x04000000
 
 /* Bits present in AT_HWCAP for s390.  */
 
diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c
index 21981d8..08ca482 100644
--- a/tcg/sparc/tcg-target.c
+++ b/tcg/sparc/tcg-target.c
@@ -209,6 +209,8 @@ static const int tcg_target_call_oarg_regs[] = {
 #define ARITH_MOVCC (INSN_OP(2) | INSN_OP3(0x2c))
 #define ARITH_MOVR (INSN_OP(2) | INSN_OP3(0x2f))
 
+#define ARITH_ADDXC (INSN_OP(2) | INSN_OP3(0x36) | INSN_OPF(0x11))
+
 #define SHIFT_SLL  (INSN_OP(2) | INSN_OP3(0x25))
 #define SHIFT_SRL  (INSN_OP(2) | INSN_OP3(0x26))
 #define SHIFT_SRA  (INSN_OP(2) | INSN_OP3(0x27))
@@ -262,6 +264,10 @@ static const int tcg_target_call_oarg_regs[] = {
 #define STW_LE     (STWA  | INSN_ASI(ASI_PRIMARY_LITTLE))
 #define STX_LE     (STXA  | INSN_ASI(ASI_PRIMARY_LITTLE))
 
+#ifndef use_vis3_instructions
+bool use_vis3_instructions;
+#endif
+
 static inline int check_fit_i64(int64_t val, unsigned int bits)
 {
     return val == sextract64(val, 0, bits);
@@ -748,11 +754,14 @@ static void tcg_out_addsub2_i64(TCGContext *s, TCGReg rl, TCGReg rh,
 
     tcg_out_arithc(s, tmp, al, bl, blconst, is_sub ? ARITH_SUBCC : ARITH_ADDCC);
 
-    /* Note that ADDX/SUBX take the carry-in from %icc, the 32-bit carry,
-       while we want %xcc, the 64-bit carry.  */
-    /* ??? There is a 2011 VIS3 ADDXC insn that does take a 64-bit carry.  */
-
-    if (bh == TCG_REG_G0) {
+    if (use_vis3_instructions && !is_sub) {
+        /* Note that ADDXC doesn't accept immediates.  */
+        if (bhconst && bh != 0) {
+           tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_T2, bh);
+           bh = TCG_REG_T2;
+        }
+        tcg_out_arith(s, rh, ah, bh, ARITH_ADDXC);
+    } else if (bh == TCG_REG_G0) {
 	/* If we have a zero, we can perform the operation in two insns,
            with the arithmetic first, and a conditional move into place.  */
 	if (rh == ah) {
@@ -1517,6 +1526,15 @@ static const TCGTargetOpDef sparc_op_defs[] = {
 
 static void tcg_target_init(TCGContext *s)
 {
+    /* Only probe for the platform and capabilities if we havn't already
+       determined maximum values at compile time.  */
+#ifndef use_vis3_instructions
+    {
+        unsigned long hwcap = qemu_getauxval(AT_HWCAP);
+        use_vis3_instructions = (hwcap & HWCAP_SPARC_VIS3) != 0;
+    }
+#endif
+
     tcg_regset_set32(tcg_target_available_regs[TCG_TYPE_I32], 0, 0xffffffff);
     tcg_regset_set32(tcg_target_available_regs[TCG_TYPE_I64], 0, ALL_64);
 
diff --git a/tcg/sparc/tcg-target.h b/tcg/sparc/tcg-target.h
index a44d34f..099b308 100644
--- a/tcg/sparc/tcg-target.h
+++ b/tcg/sparc/tcg-target.h
@@ -85,6 +85,12 @@ typedef enum {
 #define TCG_TARGET_EXTEND_ARGS 1
 #endif
 
+#if defined(__VIS__) && __VIS__ >= 0x300
+#define use_vis3_instructions  1
+#else
+extern bool use_vis3_instructions;
+#endif
+
 /* optional instructions */
 #define TCG_TARGET_HAS_div_i32		1
 #define TCG_TARGET_HAS_rem_i32		0
-- 
1.9.3