[Qemu-devel] [PATCH v4 00/25] tcg-aarch64 improvments

[Qemu-devel] [PATCH v4 00/25] tcg-aarch64 improvments

[Richard Henderson]

V4?  The last patch set had a bizzare mix of v2/v3 markers...

Anyway, changes since last:
  * Patch 1: Fix error in sigsegv write detection,
  * Patch 21: Fix error in REV64; patch largely rewritten,
  * Patch 23: Squashed 3 patches cleaning up loads; having these separate
    didn't seem to help understanding and seemed like needless churn.

These are, in fact, the only patches left without the R-B tag.
Thanks, Claudio.


r~


Richard Henderson (25):
  tcg-aarch64: Properly detect SIGSEGV writes
  tcg-aarch64: Use intptr_t apropriately
  tcg-aarch64: Use TCGType and TCGMemOp constants
  tcg-aarch64: Use MOVN in tcg_out_movi
  tcg-aarch64: Use ORRI in tcg_out_movi
  tcg-aarch64: Special case small constants in tcg_out_movi
  tcg-aarch64: Use adrp in tcg_out_movi
  tcg-aarch64: Use symbolic names for branches
  tcg-aarch64: Create tcg_out_brcond
  tcg-aarch64: Use CBZ and CBNZ
  tcg-aarch64: Reuse LR in translated code
  tcg-aarch64: Introduce tcg_out_insn_3314
  tcg-aarch64: Implement tcg_register_jit
  tcg-aarch64: Avoid add with zero in tlb load
  tcg-aarch64: Use tcg_out_call for qemu_ld/st
  tcg-aarch64: Use ADR to pass the return address to the ld/st helpers
  tcg-aarch64: Use TCGMemOp in qemu_ld/st
  tcg-aarch64: Pass qemu_ld/st arguments directly
  tcg-aarch64: Implement TCG_TARGET_HAS_new_ldst
  tcg-aarch64: Support stores of zero
  tcg-aarch64: Introduce tcg_out_insn_3507
  tcg-aarch64: Merge aarch64_ldst_get_data/type into tcg_out_op
  tcg-aarch64: Introduce tcg_out_insn_3312, _3310, _3313
  tcg-aarch64: Prefer unsigned offsets before signed offsets for ldst
  tcg-aarch64: Use tcg_out_mov in preference to tcg_out_movr

 tcg/aarch64/tcg-target.c | 1127 ++++++++++++++++++++++++----------------------
 tcg/aarch64/tcg-target.h |   34 +-
 user-exec.c              |   29 +-
 3 files changed, 632 insertions(+), 558 deletions(-)

-- 
1.9.0

[Qemu-devel] [PATCH v4 01/25] tcg-aarch64: Properly detect SIGSEGV writes

[Richard Henderson]

Since the kernel doesn't pass any info on the reason for the fault,
disassemble the instruction to detect a store.

Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 user-exec.c | 29 +++++++++++++++++++++++------
 1 file changed, 23 insertions(+), 6 deletions(-)

diff --git a/user-exec.c b/user-exec.c
index bc58056..c7106a6 100644
--- a/user-exec.c
+++ b/user-exec.c
@@ -465,16 +465,33 @@ int cpu_signal_handler(int host_signum, void *pinfo,
 
 #elif defined(__aarch64__)
 
-int cpu_signal_handler(int host_signum, void *pinfo,
-                       void *puc)
+int cpu_signal_handler(int host_signum, void *pinfo, void *puc)
 {
     siginfo_t *info = pinfo;
     struct ucontext *uc = puc;
-    uint64_t pc;
-    int is_write = 0; /* XXX how to determine? */
+    uintptr_t pc = uc->uc_mcontext.pc;
+    uint32_t insn = *(uint32_t *)pc;
+    bool is_write;
 
-    pc = uc->uc_mcontext.pc;
-    return handle_cpu_signal(pc, (uint64_t)info->si_addr,
+    /* XXX: need kernel patch to get write flag faster.  */
+    /* XXX: several of these could be combined.  */
+    is_write = (   (insn & 0xbfff0000) == 0x0c000000   /* C3.3.1 */
+                || (insn & 0xbfe00000) == 0x0c800000   /* C3.3.2 */
+                || (insn & 0xbfdf0000) == 0x0d000000   /* C3.3.3 */
+                || (insn & 0xbfc00000) == 0x0d800000   /* C3.3.4 */
+                || (insn & 0x3f400000) == 0x08000000   /* C3.3.6 */
+                || (insn & 0x3bc00000) == 0x28000000   /* C3.3.7 */
+                || (insn & 0x3be00c00) == 0x38000400   /* C3.3.8 */
+                || (insn & 0x3be00c00) == 0x38000c00   /* C3.3.9 */
+                || (insn & 0x3be00c00) == 0x38200800   /* C3.3.10 */
+                || (insn & 0x3be00c00) == 0x38000800   /* C3.3.11 */
+                || (insn & 0x3be00c00) == 0x38000000   /* C3.3.12 */
+                || (insn & 0x3bc00000) == 0x39000000   /* C3.3.13 */
+                || (insn & 0x3bc00000) == 0x29000000   /* C3.3.14 */
+                || (insn & 0x3bc00000) == 0x28800000   /* C3.3.15 */
+                || (insn & 0x3bc00000) == 0x29800000); /* C3.3.16 */
+
+    return handle_cpu_signal(pc, (uintptr_t)info->si_addr,
                              is_write, &uc->uc_sigmask, puc);
 }
 
-- 
1.9.0

[Qemu-devel] [PATCH v4 02/25] tcg-aarch64: Use intptr_t apropriately

[Richard Henderson]

As opposed to tcg_target_long.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 661a5af..6938248 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -497,7 +497,7 @@ static void tcg_out_insn_3509(TCGContext *s, AArch64Insn insn, TCGType ext,
 static inline void tcg_out_ldst_9(TCGContext *s,
                                   enum aarch64_ldst_op_data op_data,
                                   enum aarch64_ldst_op_type op_type,
-                                  TCGReg rd, TCGReg rn, tcg_target_long offset)
+                                  TCGReg rd, TCGReg rn, intptr_t offset)
 {
     /* use LDUR with BASE register with 9bit signed unscaled offset */
     tcg_out32(s, op_data << 24 | op_type << 20
@@ -566,7 +566,7 @@ static inline void tcg_out_ldst_r(TCGContext *s,
 /* solve the whole ldst problem */
 static inline void tcg_out_ldst(TCGContext *s, enum aarch64_ldst_op_data data,
                                 enum aarch64_ldst_op_type type,
-                                TCGReg rd, TCGReg rn, tcg_target_long offset)
+                                TCGReg rd, TCGReg rn, intptr_t offset)
 {
     if (offset >= -256 && offset < 256) {
         tcg_out_ldst_9(s, data, type, rd, rn, offset);
@@ -954,9 +954,9 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
     tcg_out_movr(s, 1, TCG_REG_X0, TCG_AREG0);
     tcg_out_movr(s, (TARGET_LONG_BITS == 64), TCG_REG_X1, lb->addrlo_reg);
     tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_X2, lb->mem_index);
-    tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_X3, (tcg_target_long)lb->raddr);
+    tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_X3, (intptr_t)lb->raddr);
     tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP,
-                 (tcg_target_long)qemu_ld_helpers[lb->opc & 3]);
+                 (intptr_t)qemu_ld_helpers[lb->opc & 3]);
     tcg_out_callr(s, TCG_REG_TMP);
     if (lb->opc & 0x04) {
         tcg_out_sxt(s, 1, lb->opc & 3, lb->datalo_reg, TCG_REG_X0);
@@ -979,7 +979,7 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
     tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP,
                  (intptr_t)qemu_st_helpers[lb->opc & 3]);
     tcg_out_callr(s, TCG_REG_TMP);
-    tcg_out_goto(s, (tcg_target_long)lb->raddr);
+    tcg_out_goto(s, (intptr_t)lb->raddr);
 }
 
 static void add_qemu_ldst_label(TCGContext *s, int is_ld, int opc,
-- 
1.9.0

[Qemu-devel] [PATCH v4 03/25] tcg-aarch64: Use TCGType and TCGMemOp constants

[Richard Henderson]

Rather than raw constants that could mean anything.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 73 +++++++++++++++++++++++++-----------------------
 1 file changed, 38 insertions(+), 35 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 6938248..5e6d10b 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -595,7 +595,7 @@ static inline void tcg_out_mov(TCGContext *s,
                                TCGType type, TCGReg ret, TCGReg arg)
 {
     if (ret != arg) {
-        tcg_out_movr(s, type == TCG_TYPE_I64, ret, arg);
+        tcg_out_movr(s, type, ret, arg);
     }
 }
 
@@ -828,19 +828,19 @@ static inline void tcg_out_rev16(TCGContext *s, TCGType ext,
     tcg_out32(s, base | rm << 5 | rd);
 }
 
-static inline void tcg_out_sxt(TCGContext *s, TCGType ext, int s_bits,
+static inline void tcg_out_sxt(TCGContext *s, TCGType ext, TCGMemOp s_bits,
                                TCGReg rd, TCGReg rn)
 {
     /* Using ALIASes SXTB, SXTH, SXTW, of SBFM Xd, Xn, #0, #7|15|31 */
-    int bits = 8 * (1 << s_bits) - 1;
+    int bits = (8 << s_bits) - 1;
     tcg_out_sbfm(s, ext, rd, rn, 0, bits);
 }
 
-static inline void tcg_out_uxt(TCGContext *s, int s_bits,
+static inline void tcg_out_uxt(TCGContext *s, TCGMemOp s_bits,
                                TCGReg rd, TCGReg rn)
 {
     /* Using ALIASes UXTB, UXTH of UBFM Wd, Wn, #0, #7|15 */
-    int bits = 8 * (1 << s_bits) - 1;
+    int bits = (8 << s_bits) - 1;
     tcg_out_ubfm(s, 0, rd, rn, 0, bits);
 }
 
@@ -949,19 +949,21 @@ static const void * const qemu_st_helpers[4] = {
 
 static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
 {
+    TCGMemOp opc = lb->opc;
+    TCGMemOp size = opc & MO_SIZE;
+
     reloc_pc19(lb->label_ptr[0], (intptr_t)s->code_ptr);
 
-    tcg_out_movr(s, 1, TCG_REG_X0, TCG_AREG0);
-    tcg_out_movr(s, (TARGET_LONG_BITS == 64), TCG_REG_X1, lb->addrlo_reg);
+    tcg_out_movr(s, TCG_TYPE_I64, TCG_REG_X0, TCG_AREG0);
+    tcg_out_movr(s, TARGET_LONG_BITS == 64, TCG_REG_X1, lb->addrlo_reg);
     tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_X2, lb->mem_index);
     tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_X3, (intptr_t)lb->raddr);
-    tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP,
-                 (intptr_t)qemu_ld_helpers[lb->opc & 3]);
+    tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP, (intptr_t)qemu_ld_helpers[size]);
     tcg_out_callr(s, TCG_REG_TMP);
-    if (lb->opc & 0x04) {
-        tcg_out_sxt(s, 1, lb->opc & 3, lb->datalo_reg, TCG_REG_X0);
+    if (opc & MO_SIGN) {
+        tcg_out_sxt(s, TCG_TYPE_I64, size, lb->datalo_reg, TCG_REG_X0);
     } else {
-        tcg_out_movr(s, 1, lb->datalo_reg, TCG_REG_X0);
+        tcg_out_movr(s, TCG_TYPE_I64, lb->datalo_reg, TCG_REG_X0);
     }
 
     tcg_out_goto(s, (intptr_t)lb->raddr);
@@ -969,15 +971,16 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
 
 static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
 {
+    TCGMemOp size = lb->opc;
+
     reloc_pc19(lb->label_ptr[0], (intptr_t)s->code_ptr);
 
-    tcg_out_movr(s, 1, TCG_REG_X0, TCG_AREG0);
-    tcg_out_movr(s, (TARGET_LONG_BITS == 64), TCG_REG_X1, lb->addrlo_reg);
-    tcg_out_movr(s, 1, TCG_REG_X2, lb->datalo_reg);
+    tcg_out_movr(s, TCG_TYPE_I64, TCG_REG_X0, TCG_AREG0);
+    tcg_out_movr(s, TARGET_LONG_BITS == 64, TCG_REG_X1, lb->addrlo_reg);
+    tcg_out_movr(s, size == MO_64, TCG_REG_X2, lb->datalo_reg);
     tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_X3, lb->mem_index);
     tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_X4, (intptr_t)lb->raddr);
-    tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP,
-                 (intptr_t)qemu_st_helpers[lb->opc & 3]);
+    tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP, (intptr_t)qemu_st_helpers[size]);
     tcg_out_callr(s, TCG_REG_TMP);
     tcg_out_goto(s, (intptr_t)lb->raddr);
 }
@@ -1061,14 +1064,14 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data_r,
     case 1:
         tcg_out_ldst_r(s, LDST_16, LDST_LD, data_r, addr_r, off_r);
         if (TCG_LDST_BSWAP) {
-            tcg_out_rev16(s, 0, data_r, data_r);
+            tcg_out_rev16(s, TCG_TYPE_I32, data_r, data_r);
         }
         break;
     case 1 | 4:
         if (TCG_LDST_BSWAP) {
             tcg_out_ldst_r(s, LDST_16, LDST_LD, data_r, addr_r, off_r);
-            tcg_out_rev16(s, 0, data_r, data_r);
-            tcg_out_sxt(s, 1, 1, data_r, data_r);
+            tcg_out_rev16(s, TCG_TYPE_I32, data_r, data_r);
+            tcg_out_sxt(s, TCG_TYPE_I64, MO_16, data_r, data_r);
         } else {
             tcg_out_ldst_r(s, LDST_16, LDST_LD_S_X, data_r, addr_r, off_r);
         }
@@ -1076,14 +1079,14 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data_r,
     case 2:
         tcg_out_ldst_r(s, LDST_32, LDST_LD, data_r, addr_r, off_r);
         if (TCG_LDST_BSWAP) {
-            tcg_out_rev(s, 0, data_r, data_r);
+            tcg_out_rev(s, TCG_TYPE_I32, data_r, data_r);
         }
         break;
     case 2 | 4:
         if (TCG_LDST_BSWAP) {
             tcg_out_ldst_r(s, LDST_32, LDST_LD, data_r, addr_r, off_r);
-            tcg_out_rev(s, 0, data_r, data_r);
-            tcg_out_sxt(s, 1, 2, data_r, data_r);
+            tcg_out_rev(s, TCG_TYPE_I32, data_r, data_r);
+            tcg_out_sxt(s, TCG_TYPE_I64, MO_32, data_r, data_r);
         } else {
             tcg_out_ldst_r(s, LDST_32, LDST_LD_S_X, data_r, addr_r, off_r);
         }
@@ -1091,7 +1094,7 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data_r,
     case 3:
         tcg_out_ldst_r(s, LDST_64, LDST_LD, data_r, addr_r, off_r);
         if (TCG_LDST_BSWAP) {
-            tcg_out_rev(s, 1, data_r, data_r);
+            tcg_out_rev(s, TCG_TYPE_I64, data_r, data_r);
         }
         break;
     default:
@@ -1108,7 +1111,7 @@ static void tcg_out_qemu_st_direct(TCGContext *s, int opc, TCGReg data_r,
         break;
     case 1:
         if (TCG_LDST_BSWAP) {
-            tcg_out_rev16(s, 0, TCG_REG_TMP, data_r);
+            tcg_out_rev16(s, TCG_TYPE_I32, TCG_REG_TMP, data_r);
             tcg_out_ldst_r(s, LDST_16, LDST_ST, TCG_REG_TMP, addr_r, off_r);
         } else {
             tcg_out_ldst_r(s, LDST_16, LDST_ST, data_r, addr_r, off_r);
@@ -1116,7 +1119,7 @@ static void tcg_out_qemu_st_direct(TCGContext *s, int opc, TCGReg data_r,
         break;
     case 2:
         if (TCG_LDST_BSWAP) {
-            tcg_out_rev(s, 0, TCG_REG_TMP, data_r);
+            tcg_out_rev(s, TCG_TYPE_I32, TCG_REG_TMP, data_r);
             tcg_out_ldst_r(s, LDST_32, LDST_ST, TCG_REG_TMP, addr_r, off_r);
         } else {
             tcg_out_ldst_r(s, LDST_32, LDST_ST, data_r, addr_r, off_r);
@@ -1124,7 +1127,7 @@ static void tcg_out_qemu_st_direct(TCGContext *s, int opc, TCGReg data_r,
         break;
     case 3:
         if (TCG_LDST_BSWAP) {
-            tcg_out_rev(s, 1, TCG_REG_TMP, data_r);
+            tcg_out_rev(s, TCG_TYPE_I64, TCG_REG_TMP, data_r);
             tcg_out_ldst_r(s, LDST_64, LDST_ST, TCG_REG_TMP, addr_r, off_r);
         } else {
             tcg_out_ldst_r(s, LDST_64, LDST_ST, data_r, addr_r, off_r);
@@ -1547,30 +1550,30 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
         break;
     case INDEX_op_bswap16_i64:
     case INDEX_op_bswap16_i32:
-        tcg_out_rev16(s, 0, a0, a1);
+        tcg_out_rev16(s, TCG_TYPE_I32, a0, a1);
         break;
 
     case INDEX_op_ext8s_i64:
     case INDEX_op_ext8s_i32:
-        tcg_out_sxt(s, ext, 0, a0, a1);
+        tcg_out_sxt(s, ext, MO_8, a0, a1);
         break;
     case INDEX_op_ext16s_i64:
     case INDEX_op_ext16s_i32:
-        tcg_out_sxt(s, ext, 1, a0, a1);
+        tcg_out_sxt(s, ext, MO_16, a0, a1);
         break;
     case INDEX_op_ext32s_i64:
-        tcg_out_sxt(s, 1, 2, a0, a1);
+        tcg_out_sxt(s, TCG_TYPE_I64, MO_32, a0, a1);
         break;
     case INDEX_op_ext8u_i64:
     case INDEX_op_ext8u_i32:
-        tcg_out_uxt(s, 0, a0, a1);
+        tcg_out_uxt(s, MO_8, a0, a1);
         break;
     case INDEX_op_ext16u_i64:
     case INDEX_op_ext16u_i32:
-        tcg_out_uxt(s, 1, a0, a1);
+        tcg_out_uxt(s, MO_16, a0, a1);
         break;
     case INDEX_op_ext32u_i64:
-        tcg_out_movr(s, 0, a0, a1);
+        tcg_out_movr(s, TCG_TYPE_I32, a0, a1);
         break;
 
     case INDEX_op_deposit_i64:
@@ -1794,7 +1797,7 @@ static void tcg_target_qemu_prologue(TCGContext *s)
                       TCG_REG_FP, TCG_REG_LR, frame_size_callee_saved);
 
     /* FP -> callee_saved */
-    tcg_out_movr_sp(s, 1, TCG_REG_FP, TCG_REG_SP);
+    tcg_out_movr_sp(s, TCG_TYPE_I64, TCG_REG_FP, TCG_REG_SP);
 
     /* store callee-preserved regs x19..x28 using FP -> callee_saved */
     for (r = TCG_REG_X19; r <= TCG_REG_X27; r += 2) {
-- 
1.9.0

[Qemu-devel] [PATCH v4 04/25] tcg-aarch64: Use MOVN in tcg_out_movi

[Richard Henderson]

When profitable, initialize the register with MOVN instead of MOVZ,
before setting the remaining lanes with MOVK.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 63 ++++++++++++++++++++++++++++++++++++++----------
 1 file changed, 50 insertions(+), 13 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 5e6d10b..1d7612c 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -531,24 +531,61 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg rd,
                          tcg_target_long value)
 {
     AArch64Insn insn;
-
-    if (type == TCG_TYPE_I32) {
+    int i, wantinv, shift;
+    tcg_target_long svalue = value;
+    tcg_target_long ivalue = ~value;
+    tcg_target_long imask;
+
+    /* For 32-bit values, discard potential garbage in value.  For 64-bit
+       values within [2**31, 2**32-1], we can create smaller sequences by
+       interpreting this as a negative 32-bit number, while ensuring that
+       the high 32 bits are cleared by setting SF=0.  */
+    if (type == TCG_TYPE_I32 || (value & ~0xffffffffull) == 0) {
+        svalue = (int32_t)value;
         value = (uint32_t)value;
+        ivalue = (uint32_t)ivalue;
+        type = TCG_TYPE_I32;
+    }
+
+    /* Would it take fewer insns to begin with MOVN?  For the value and its
+       inverse, count the number of 16-bit lanes that are 0.  */
+    for (i = wantinv = imask = 0; i < 64; i += 16) {
+        tcg_target_long mask = 0xffffull << i;
+        if ((value & mask) == 0) {
+            wantinv -= 1;
+        }
+        if ((ivalue & mask) == 0) {
+            wantinv += 1;
+            imask |= mask;
+        }
     }
 
-    /* count trailing zeros in 16 bit steps, mapping 64 to 0. Emit the
-       first MOVZ with the half-word immediate skipping the zeros, with a shift
-       (LSL) equal to this number. Then all next instructions use MOVKs.
-       Zero the processed half-word in the value, continue until empty.
-       We build the final result 16bits at a time with up to 4 instructions,
-       but do not emit instructions for 16bit zero holes. */
+    /* If we had more 0xffff than 0x0000, invert VALUE and use MOVN.  */
     insn = I3405_MOVZ;
-    do {
-        unsigned shift = ctz64(value) & (63 & -16);
-        tcg_out_insn_3405(s, insn, shift >= 32, rd, value >> shift, shift);
+    if (wantinv > 0) {
+        value = ivalue;
+        insn = I3405_MOVN;
+    }
+
+    /* Find the lowest lane that is not 0x0000.  */
+    shift = ctz64(value) & (63 & -16);
+    tcg_out_insn_3405(s, insn, type, rd, value >> shift, shift);
+
+    if (wantinv > 0) {
+        /* Re-invert the value, so MOVK sees non-inverted bits.  */
+        value = ~value;
+        /* Clear out all the 0xffff lanes.  */
+        value ^= imask;
+    }
+    /* Clear out the lane that we just set.  */
+    value &= ~(0xffffUL << shift);
+
+    /* Iterate until all lanes have been set, and thus cleared from VALUE.  */
+    while (value) {
+        shift = ctz64(value) & (63 & -16);
+        tcg_out_insn(s, 3405, MOVK, type, rd, value >> shift, shift);
         value &= ~(0xffffUL << shift);
-        insn = I3405_MOVK;
-    } while (value);
+    }
 }
 
 static inline void tcg_out_ldst_r(TCGContext *s,
-- 
1.9.0

[Qemu-devel] [PATCH v4 05/25] tcg-aarch64: Use ORRI in tcg_out_movi

[Richard Henderson]

The subset of logical immediates that we support is quite quick to test,
and such constants are quite common to want to load.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 70 +++++++++++++++++++++++++++---------------------
 1 file changed, 39 insertions(+), 31 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 1d7612c..c1d9895 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -527,6 +527,37 @@ static void tcg_out_movr_sp(TCGContext *s, TCGType ext, TCGReg rd, TCGReg rn)
     tcg_out_insn(s, 3401, ADDI, ext, rd, rn, 0);
 }
 
+/* This function is used for the Logical (immediate) instruction group.
+   The value of LIMM must satisfy IS_LIMM.  See the comment above about
+   only supporting simplified logical immediates.  */
+static void tcg_out_logicali(TCGContext *s, AArch64Insn insn, TCGType ext,
+                             TCGReg rd, TCGReg rn, uint64_t limm)
+{
+    unsigned h, l, r, c;
+
+    assert(is_limm(limm));
+
+    h = clz64(limm);
+    l = ctz64(limm);
+    if (l == 0) {
+        r = 0;                  /* form 0....01....1 */
+        c = ctz64(~limm) - 1;
+        if (h == 0) {
+            r = clz64(~limm);   /* form 1..10..01..1 */
+            c += r;
+        }
+    } else {
+        r = 64 - l;             /* form 1....10....0 or 0..01..10..0 */
+        c = r - h - 1;
+    }
+    if (ext == TCG_TYPE_I32) {
+        r &= 31;
+        c &= 31;
+    }
+
+    tcg_out_insn_3404(s, insn, ext, rd, rn, ext, r, c);
+}
+
 static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg rd,
                          tcg_target_long value)
 {
@@ -547,6 +578,14 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg rd,
         type = TCG_TYPE_I32;
     }
 
+    /* Check for bitfield immediates.  For the benefit of 32-bit quantities,
+       use the sign-extended value.  That lets us match rotated values such
+       as 0xff0000ff with the same 64-bit logic matching 0xffffffffff0000ff. */
+    if (is_limm(svalue)) {
+        tcg_out_logicali(s, I3404_ORRI, type, rd, TCG_REG_XZR, svalue);
+        return;
+    }
+
     /* Would it take fewer insns to begin with MOVN?  For the value and its
        inverse, count the number of 16-bit lanes that are 0.  */
     for (i = wantinv = imask = 0; i < 64; i += 16) {
@@ -891,37 +930,6 @@ static void tcg_out_addsubi(TCGContext *s, int ext, TCGReg rd,
     }
 }
 
-/* This function is used for the Logical (immediate) instruction group.
-   The value of LIMM must satisfy IS_LIMM.  See the comment above about
-   only supporting simplified logical immediates.  */
-static void tcg_out_logicali(TCGContext *s, AArch64Insn insn, TCGType ext,
-                             TCGReg rd, TCGReg rn, uint64_t limm)
-{
-    unsigned h, l, r, c;
-
-    assert(is_limm(limm));
-
-    h = clz64(limm);
-    l = ctz64(limm);
-    if (l == 0) {
-        r = 0;                  /* form 0....01....1 */
-        c = ctz64(~limm) - 1;
-        if (h == 0) {
-            r = clz64(~limm);   /* form 1..10..01..1 */
-            c += r;
-        }
-    } else {
-        r = 64 - l;             /* form 1....10....0 or 0..01..10..0 */
-        c = r - h - 1;
-    }
-    if (ext == TCG_TYPE_I32) {
-        r &= 31;
-        c &= 31;
-    }
-
-    tcg_out_insn_3404(s, insn, ext, rd, rn, ext, r, c);
-}
-
 static inline void tcg_out_addsub2(TCGContext *s, int ext, TCGReg rl,
                                    TCGReg rh, TCGReg al, TCGReg ah,
                                    tcg_target_long bl, tcg_target_long bh,
-- 
1.9.0

[Qemu-devel] [PATCH v4 06/25] tcg-aarch64: Special case small constants in tcg_out_movi

[Richard Henderson]

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index c1d9895..a08f6c7 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -578,6 +578,16 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg rd,
         type = TCG_TYPE_I32;
     }
 
+    /* Speed things up by handling the common case of small positive
+       and negative values specially.  */
+    if ((value & ~0xffffull) == 0) {
+        tcg_out_insn(s, 3405, MOVZ, type, rd, value, 0);
+        return;
+    } else if ((ivalue & ~0xffffull) == 0) {
+        tcg_out_insn(s, 3405, MOVN, type, rd, ivalue, 0);
+        return;
+    }
+
     /* Check for bitfield immediates.  For the benefit of 32-bit quantities,
        use the sign-extended value.  That lets us match rotated values such
        as 0xff0000ff with the same 64-bit logic matching 0xffffffffff0000ff. */
-- 
1.9.0

[Qemu-devel] [PATCH v4 07/25] tcg-aarch64: Use adrp in tcg_out_movi

[Richard Henderson]

Loading an qemu pointer as an immediate happens often.  E.g.

- exit_tb $0x7fa8140013
+ exit_tb $0x7f81ee0013
...
- :  d2800260        mov     x0, #0x13
- :  f2b50280        movk    x0, #0xa814, lsl #16
- :  f2c00fe0        movk    x0, #0x7f, lsl #32
+ :  90ff1000        adrp    x0, 0x7f81ee0000
+ :  91004c00        add     x0, x0, #0x13

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index a08f6c7..1337a13 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -294,6 +294,10 @@ typedef enum {
     I3405_MOVZ      = 0x52800000,
     I3405_MOVK      = 0x72800000,
 
+    /* PC relative addressing instructions.  */
+    I3406_ADR       = 0x10000000,
+    I3406_ADRP      = 0x90000000,
+
     /* Add/subtract shifted register instructions (without a shift).  */
     I3502_ADD       = 0x0b000000,
     I3502_ADDS      = 0x2b000000,
@@ -457,6 +461,12 @@ static void tcg_out_insn_3405(TCGContext *s, AArch64Insn insn, TCGType ext,
     tcg_out32(s, insn | ext << 31 | shift << (21 - 4) | half << 5 | rd);
 }
 
+static void tcg_out_insn_3406(TCGContext *s, AArch64Insn insn,
+                              TCGReg rd, int64_t disp)
+{
+    tcg_out32(s, insn | (disp & 3) << 29 | (disp & 0x1ffffc) << (5 - 2) | rd);
+}
+
 /* This function is for both 3.5.2 (Add/Subtract shifted register), for
    the rare occasion when we actually want to supply a shift amount.  */
 static inline void tcg_out_insn_3502S(TCGContext *s, AArch64Insn insn,
@@ -596,6 +606,19 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg rd,
         return;
     }
 
+    /* Look for host pointer values within 4G of the PC.  This happens
+       often when loading pointers to QEMU's own data structures.  */
+    if (type == TCG_TYPE_I64) {
+        tcg_target_long disp = (value >> 12) - ((intptr_t)s->code_ptr >> 12);
+        if (disp == sextract64(disp, 0, 21)) {
+            tcg_out_insn(s, 3406, ADRP, rd, disp);
+            if (value & 0xfff) {
+                tcg_out_insn(s, 3401, ADDI, type, rd, rd, value & 0xfff);
+            }
+            return;
+        }
+    }
+
     /* Would it take fewer insns to begin with MOVN?  For the value and its
        inverse, count the number of 16-bit lanes that are 0.  */
     for (i = wantinv = imask = 0; i < 64; i += 16) {
-- 
1.9.0

[Qemu-devel] [PATCH v4 08/25] tcg-aarch64: Use symbolic names for branches

[Richard Henderson]

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 74 ++++++++++++++++++++++++++++--------------------
 1 file changed, 43 insertions(+), 31 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 1337a13..8b15d3b 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -270,6 +270,18 @@ enum aarch64_ldst_op_type { /* type of operation */
    use the section number of the architecture reference manual in which the
    instruction group is described.  */
 typedef enum {
+    /* Conditional branch (immediate).  */
+    I3202_B_C       = 0x54000000,
+
+    /* Unconditional branch (immediate).  */
+    I3206_B         = 0x14000000,
+    I3206_BL        = 0x94000000,
+
+    /* Unconditional branch (register).  */
+    I3207_BR        = 0xd61f0000,
+    I3207_BLR       = 0xd63f0000,
+    I3207_RET       = 0xd65f0000,
+
     /* Add/subtract immediate instructions.  */
     I3401_ADDI      = 0x11000000,
     I3401_ADDSI     = 0x31000000,
@@ -421,6 +433,22 @@ static inline uint32_t tcg_in32(TCGContext *s)
 #define tcg_out_insn(S, FMT, OP, ...) \
     glue(tcg_out_insn_,FMT)(S, glue(glue(glue(I,FMT),_),OP), ## __VA_ARGS__)
 
+static void tcg_out_insn_3202(TCGContext *s, AArch64Insn insn,
+                              TCGCond c, int imm19)
+{
+    tcg_out32(s, insn | tcg_cond_to_aarch64[c] | (imm19 & 0x7ffff) << 5);
+}
+
+static void tcg_out_insn_3206(TCGContext *s, AArch64Insn insn, int imm26)
+{
+    tcg_out32(s, insn | (imm26 & 0x03ffffff));
+}
+
+static void tcg_out_insn_3207(TCGContext *s, AArch64Insn insn, TCGReg rn)
+{
+    tcg_out32(s, insn | rn << 5);
+}
+
 static void tcg_out_insn_3401(TCGContext *s, AArch64Insn insn, TCGType ext,
                               TCGReg rd, TCGReg rn, uint64_t aimm)
 {
@@ -817,28 +845,24 @@ static inline void tcg_out_goto(TCGContext *s, intptr_t target)
         tcg_abort();
     }
 
-    tcg_out32(s, 0x14000000 | (offset & 0x03ffffff));
+    tcg_out_insn(s, 3206, B, offset);
 }
 
 static inline void tcg_out_goto_noaddr(TCGContext *s)
 {
-    /* We pay attention here to not modify the branch target by
-       reading from the buffer. This ensure that caches and memory are
-       kept coherent during retranslation.
-       Mask away possible garbage in the high bits for the first translation,
-       while keeping the offset bits for retranslation. */
-    uint32_t insn;
-    insn = (tcg_in32(s) & 0x03ffffff) | 0x14000000;
-    tcg_out32(s, insn);
+    /* We pay attention here to not modify the branch target by reading from
+       the buffer. This ensure that caches and memory are kept coherent during
+       retranslation.  Mask away possible garbage in the high bits for the
+       first translation, while keeping the offset bits for retranslation. */
+    uint32_t old = tcg_in32(s);
+    tcg_out_insn(s, 3206, B, old);
 }
 
 static inline void tcg_out_goto_cond_noaddr(TCGContext *s, TCGCond c)
 {
-    /* see comments in tcg_out_goto_noaddr */
-    uint32_t insn;
-    insn = tcg_in32(s) & (0x07ffff << 5);
-    insn |= 0x54000000 | tcg_cond_to_aarch64[c];
-    tcg_out32(s, insn);
+    /* See comments in tcg_out_goto_noaddr.  */
+    uint32_t old = tcg_in32(s) >> 5;
+    tcg_out_insn(s, 3202, B_C, c, old);
 }
 
 static inline void tcg_out_goto_cond(TCGContext *s, TCGCond c, intptr_t target)
@@ -850,18 +874,12 @@ static inline void tcg_out_goto_cond(TCGContext *s, TCGCond c, intptr_t target)
         tcg_abort();
     }
 
-    offset &= 0x7ffff;
-    tcg_out32(s, 0x54000000 | tcg_cond_to_aarch64[c] | offset << 5);
+    tcg_out_insn(s, 3202, B_C, c, offset);
 }
 
 static inline void tcg_out_callr(TCGContext *s, TCGReg reg)
 {
-    tcg_out32(s, 0xd63f0000 | reg << 5);
-}
-
-static inline void tcg_out_gotor(TCGContext *s, TCGReg reg)
-{
-    tcg_out32(s, 0xd61f0000 | reg << 5);
+    tcg_out_insn(s, 3207, BLR, reg);
 }
 
 static inline void tcg_out_call(TCGContext *s, intptr_t target)
@@ -872,16 +890,10 @@ static inline void tcg_out_call(TCGContext *s, intptr_t target)
         tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP, target);
         tcg_out_callr(s, TCG_REG_TMP);
     } else {
-        tcg_out32(s, 0x94000000 | (offset & 0x03ffffff));
+        tcg_out_insn(s, 3206, BL, offset);
     }
 }
 
-static inline void tcg_out_ret(TCGContext *s)
-{
-    /* emit RET { LR } */
-    tcg_out32(s, 0xd65f03c0);
-}
-
 void aarch64_tb_set_jmp_target(uintptr_t jmp_addr, uintptr_t addr)
 {
     intptr_t target = addr;
@@ -1899,7 +1911,7 @@ static void tcg_target_qemu_prologue(TCGContext *s)
 #endif
 
     tcg_out_mov(s, TCG_TYPE_PTR, TCG_AREG0, tcg_target_call_iarg_regs[0]);
-    tcg_out_gotor(s, tcg_target_call_iarg_regs[1]);
+    tcg_out_insn(s, 3207, BR, tcg_target_call_iarg_regs[1]);
 
     tb_ret_addr = s->code_ptr;
 
@@ -1917,5 +1929,5 @@ static void tcg_target_qemu_prologue(TCGContext *s)
     /* pop (FP, LR), restore SP to previous frame, return */
     tcg_out_pop_pair(s, TCG_REG_SP,
                      TCG_REG_FP, TCG_REG_LR, frame_size_callee_saved);
-    tcg_out_ret(s);
+    tcg_out_insn(s, 3207, RET, TCG_REG_LR);
 }
-- 
1.9.0

[Qemu-devel] [PATCH v4 09/25] tcg-aarch64: Create tcg_out_brcond

[Richard Henderson]

Rearrange code to put the compare and branch in the same place.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 34 ++++++++++++++--------------------
 1 file changed, 14 insertions(+), 20 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 8b15d3b..5889a98 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -865,18 +865,6 @@ static inline void tcg_out_goto_cond_noaddr(TCGContext *s, TCGCond c)
     tcg_out_insn(s, 3202, B_C, c, old);
 }
 
-static inline void tcg_out_goto_cond(TCGContext *s, TCGCond c, intptr_t target)
-{
-    intptr_t offset = (target - (intptr_t)s->code_ptr) / 4;
-
-    if (offset < -0x40000 || offset >= 0x40000) {
-        /* out of 19bit range */
-        tcg_abort();
-    }
-
-    tcg_out_insn(s, 3202, B_C, c, offset);
-}
-
 static inline void tcg_out_callr(TCGContext *s, TCGReg reg)
 {
     tcg_out_insn(s, 3207, BLR, reg);
@@ -920,17 +908,24 @@ static inline void tcg_out_goto_label(TCGContext *s, int label_index)
     }
 }
 
-static inline void tcg_out_goto_label_cond(TCGContext *s,
-                                           TCGCond c, int label_index)
+static void tcg_out_brcond(TCGContext *s, TCGMemOp ext, TCGCond c, TCGArg a,
+                           TCGArg b, bool b_const, int label)
 {
-    TCGLabel *l = &s->labels[label_index];
+    TCGLabel *l = &s->labels[label];
+    intptr_t offset;
+
+    tcg_out_cmp(s, ext, a, b, b_const);
 
     if (!l->has_value) {
-        tcg_out_reloc(s, s->code_ptr, R_AARCH64_CONDBR19, label_index, 0);
-        tcg_out_goto_cond_noaddr(s, c);
+        tcg_out_reloc(s, s->code_ptr, R_AARCH64_CONDBR19, label, 0);
+        offset = tcg_in32(s) >> 5;
     } else {
-        tcg_out_goto_cond(s, c, l->u.value);
+        offset = l->u.value - (uintptr_t)s->code_ptr;
+        offset >>= 2;
+        assert(offset >= -0x40000 && offset < 0x40000);
     }
+
+    tcg_out_insn(s, 3202, B_C, c, offset);
 }
 
 static inline void tcg_out_rev(TCGContext *s, TCGType ext,
@@ -1571,8 +1566,7 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
         a1 = (int32_t)a1;
         /* FALLTHRU */
     case INDEX_op_brcond_i64:
-        tcg_out_cmp(s, ext, a0, a1, const_args[1]);
-        tcg_out_goto_label_cond(s, a2, args[3]);
+        tcg_out_brcond(s, ext, a2, a0, a1, const_args[1], args[3]);
         break;
 
     case INDEX_op_setcond_i32:
-- 
1.9.0

[Qemu-devel] [PATCH v4 10/25] tcg-aarch64: Use CBZ and CBNZ

[Richard Henderson]

A compare and branch against zero happens at the start of
every single TB.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 26 ++++++++++++++++++++++++--
 1 file changed, 24 insertions(+), 2 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 5889a98..48a246d 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -270,6 +270,10 @@ enum aarch64_ldst_op_type { /* type of operation */
    use the section number of the architecture reference manual in which the
    instruction group is described.  */
 typedef enum {
+    /* Compare and branch (immediate).  */
+    I3201_CBZ       = 0x34000000,
+    I3201_CBNZ      = 0x35000000,
+
     /* Conditional branch (immediate).  */
     I3202_B_C       = 0x54000000,
 
@@ -433,6 +437,12 @@ static inline uint32_t tcg_in32(TCGContext *s)
 #define tcg_out_insn(S, FMT, OP, ...) \
     glue(tcg_out_insn_,FMT)(S, glue(glue(glue(I,FMT),_),OP), ## __VA_ARGS__)
 
+static void tcg_out_insn_3201(TCGContext *s, AArch64Insn insn, TCGType ext,
+                              TCGReg rt, int imm19)
+{
+    tcg_out32(s, insn | ext << 31 | (imm19 & 0x7ffff) << 5 | rt);
+}
+
 static void tcg_out_insn_3202(TCGContext *s, AArch64Insn insn,
                               TCGCond c, int imm19)
 {
@@ -913,8 +923,14 @@ static void tcg_out_brcond(TCGContext *s, TCGMemOp ext, TCGCond c, TCGArg a,
 {
     TCGLabel *l = &s->labels[label];
     intptr_t offset;
+    bool need_cmp;
 
-    tcg_out_cmp(s, ext, a, b, b_const);
+    if (b_const && b == 0 && (c == TCG_COND_EQ || c == TCG_COND_NE)) {
+        need_cmp = false;
+    } else {
+        need_cmp = true;
+        tcg_out_cmp(s, ext, a, b, b_const);
+    }
 
     if (!l->has_value) {
         tcg_out_reloc(s, s->code_ptr, R_AARCH64_CONDBR19, label, 0);
@@ -925,7 +941,13 @@ static void tcg_out_brcond(TCGContext *s, TCGMemOp ext, TCGCond c, TCGArg a,
         assert(offset >= -0x40000 && offset < 0x40000);
     }
 
-    tcg_out_insn(s, 3202, B_C, c, offset);
+    if (need_cmp) {
+        tcg_out_insn(s, 3202, B_C, c, offset);
+    } else if (c == TCG_COND_EQ) {
+        tcg_out_insn(s, 3201, CBZ, ext, a, offset);
+    } else {
+        tcg_out_insn(s, 3201, CBNZ, ext, a, offset);
+    }
 }
 
 static inline void tcg_out_rev(TCGContext *s, TCGType ext,
-- 
1.9.0

[Qemu-devel] [PATCH v4 11/25] tcg-aarch64: Reuse LR in translated code

[Richard Henderson]

It's obviously call-clobbered, but is otherwise unused.
Repurpose it as the TCG temporary.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 34 ++++++++++++++++------------------
 tcg/aarch64/tcg-target.h | 32 +++++++++++++++++---------------
 2 files changed, 33 insertions(+), 33 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 48a246d..e36909e 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -23,10 +23,7 @@ static const char * const tcg_target_reg_names[TCG_TARGET_NB_REGS] = {
     "%x0", "%x1", "%x2", "%x3", "%x4", "%x5", "%x6", "%x7",
     "%x8", "%x9", "%x10", "%x11", "%x12", "%x13", "%x14", "%x15",
     "%x16", "%x17", "%x18", "%x19", "%x20", "%x21", "%x22", "%x23",
-    "%x24", "%x25", "%x26", "%x27", "%x28",
-    "%fp", /* frame pointer */
-    "%lr", /* link register */
-    "%sp",  /* stack pointer */
+    "%x24", "%x25", "%x26", "%x27", "%x28", "%fp", "%x30", "%sp",
 };
 #endif /* NDEBUG */
 
@@ -41,16 +38,17 @@ static const int tcg_target_reg_alloc_order[] = {
     TCG_REG_X24, TCG_REG_X25, TCG_REG_X26, TCG_REG_X27,
     TCG_REG_X28, /* we will reserve this for GUEST_BASE if configured */
 
-    TCG_REG_X9, TCG_REG_X10, TCG_REG_X11, TCG_REG_X12,
-    TCG_REG_X13, TCG_REG_X14, TCG_REG_X15,
+    TCG_REG_X8, TCG_REG_X9, TCG_REG_X10, TCG_REG_X11,
+    TCG_REG_X12, TCG_REG_X13, TCG_REG_X14, TCG_REG_X15,
     TCG_REG_X16, TCG_REG_X17,
 
-    TCG_REG_X18, TCG_REG_X19, /* will not use these, see tcg_target_init */
-
     TCG_REG_X0, TCG_REG_X1, TCG_REG_X2, TCG_REG_X3,
     TCG_REG_X4, TCG_REG_X5, TCG_REG_X6, TCG_REG_X7,
 
-    TCG_REG_X8, /* will not use, see tcg_target_init */
+    /* X18 reserved by system */
+    /* X19 reserved for AREG0 */
+    /* X29 reserved as fp */
+    /* X30 reserved as temporary */
 };
 
 static const int tcg_target_call_iarg_regs[8] = {
@@ -61,13 +59,13 @@ static const int tcg_target_call_oarg_regs[1] = {
     TCG_REG_X0
 };
 
-#define TCG_REG_TMP TCG_REG_X8
+#define TCG_REG_TMP TCG_REG_X30
 
 #ifndef CONFIG_SOFTMMU
-# if defined(CONFIG_USE_GUEST_BASE)
-# define TCG_REG_GUEST_BASE TCG_REG_X28
+# ifdef CONFIG_USE_GUEST_BASE
+#  define TCG_REG_GUEST_BASE TCG_REG_X28
 # else
-# define TCG_REG_GUEST_BASE TCG_REG_XZR
+#  define TCG_REG_GUEST_BASE TCG_REG_XZR
 # endif
 #endif
 
@@ -1871,7 +1869,7 @@ static void tcg_target_init(TCGContext *s)
                      (1 << TCG_REG_X12) | (1 << TCG_REG_X13) |
                      (1 << TCG_REG_X14) | (1 << TCG_REG_X15) |
                      (1 << TCG_REG_X16) | (1 << TCG_REG_X17) |
-                     (1 << TCG_REG_X18));
+                     (1 << TCG_REG_X18) | (1 << TCG_REG_X30));
 
     tcg_regset_clear(s->reserved_regs);
     tcg_regset_set_reg(s->reserved_regs, TCG_REG_SP);
@@ -1902,13 +1900,13 @@ static void tcg_target_qemu_prologue(TCGContext *s)
     tcg_out_push_pair(s, TCG_REG_SP,
                       TCG_REG_FP, TCG_REG_LR, frame_size_callee_saved);
 
-    /* FP -> callee_saved */
+    /* Set up frame pointer for canonical unwinding.  */
     tcg_out_movr_sp(s, TCG_TYPE_I64, TCG_REG_FP, TCG_REG_SP);
 
-    /* store callee-preserved regs x19..x28 using FP -> callee_saved */
+    /* Store callee-preserved regs x19..x28.  */
     for (r = TCG_REG_X19; r <= TCG_REG_X27; r += 2) {
         int idx = (r - TCG_REG_X19) / 2 + 1;
-        tcg_out_store_pair(s, TCG_REG_FP, r, r + 1, idx);
+        tcg_out_store_pair(s, TCG_REG_SP, r, r + 1, idx);
     }
 
     /* Make stack space for TCG locals.  */
@@ -1939,7 +1937,7 @@ static void tcg_target_qemu_prologue(TCGContext *s)
        FP must be preserved, so it still points to callee_saved area */
     for (r = TCG_REG_X19; r <= TCG_REG_X27; r += 2) {
         int idx = (r - TCG_REG_X19) / 2 + 1;
-        tcg_out_load_pair(s, TCG_REG_FP, r, r + 1, idx);
+        tcg_out_load_pair(s, TCG_REG_SP, r, r + 1, idx);
     }
 
     /* pop (FP, LR), restore SP to previous frame, return */
diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h
index 988983e..faccc36 100644
--- a/tcg/aarch64/tcg-target.h
+++ b/tcg/aarch64/tcg-target.h
@@ -17,17 +17,23 @@
 #undef TCG_TARGET_STACK_GROWSUP
 
 typedef enum {
-    TCG_REG_X0, TCG_REG_X1, TCG_REG_X2, TCG_REG_X3, TCG_REG_X4,
-    TCG_REG_X5, TCG_REG_X6, TCG_REG_X7, TCG_REG_X8, TCG_REG_X9,
-    TCG_REG_X10, TCG_REG_X11, TCG_REG_X12, TCG_REG_X13, TCG_REG_X14,
-    TCG_REG_X15, TCG_REG_X16, TCG_REG_X17, TCG_REG_X18, TCG_REG_X19,
-    TCG_REG_X20, TCG_REG_X21, TCG_REG_X22, TCG_REG_X23, TCG_REG_X24,
-    TCG_REG_X25, TCG_REG_X26, TCG_REG_X27, TCG_REG_X28,
-    TCG_REG_FP,  /* frame pointer */
-    TCG_REG_LR, /* link register */
-    TCG_REG_SP,  /* stack pointer or zero register */
-    TCG_REG_XZR = TCG_REG_SP /* same register number */
-    /* program counter is not directly accessible! */
+    TCG_REG_X0, TCG_REG_X1, TCG_REG_X2, TCG_REG_X3,
+    TCG_REG_X4, TCG_REG_X5, TCG_REG_X6, TCG_REG_X7,
+    TCG_REG_X8, TCG_REG_X9, TCG_REG_X10, TCG_REG_X11,
+    TCG_REG_X12, TCG_REG_X13, TCG_REG_X14, TCG_REG_X15,
+    TCG_REG_X16, TCG_REG_X17, TCG_REG_X18, TCG_REG_X19,
+    TCG_REG_X20, TCG_REG_X21, TCG_REG_X22, TCG_REG_X23,
+    TCG_REG_X24, TCG_REG_X25, TCG_REG_X26, TCG_REG_X27,
+    TCG_REG_X28, TCG_REG_X29, TCG_REG_X30,
+
+    /* X31 is either the stack pointer or zero, depending on context.  */
+    TCG_REG_SP = 31,
+    TCG_REG_XZR = 31,
+
+    /* Aliases.  */
+    TCG_REG_FP = TCG_REG_X29,
+    TCG_REG_LR = TCG_REG_X30,
+    TCG_AREG0  = TCG_REG_X19,
 } TCGReg;
 
 #define TCG_TARGET_NB_REGS 32
@@ -92,10 +98,6 @@ typedef enum {
 #define TCG_TARGET_HAS_muluh_i64        1
 #define TCG_TARGET_HAS_mulsh_i64        1
 
-enum {
-    TCG_AREG0 = TCG_REG_X19,
-};
-
 #define TCG_TARGET_HAS_new_ldst         0
 
 static inline void flush_icache_range(uintptr_t start, uintptr_t stop)
-- 
1.9.0

[Qemu-devel] [PATCH v4 12/25] tcg-aarch64: Introduce tcg_out_insn_3314

[Richard Henderson]

Combines 4 other inline functions and tidies the prologue.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 100 ++++++++++++++++-------------------------------
 1 file changed, 33 insertions(+), 67 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index e36909e..5cffe50 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -284,6 +284,10 @@ typedef enum {
     I3207_BLR       = 0xd63f0000,
     I3207_RET       = 0xd65f0000,
 
+    /* Load/store register pair instructions.  */
+    I3314_LDP       = 0x28400000,
+    I3314_STP       = 0x28000000,
+
     /* Add/subtract immediate instructions.  */
     I3401_ADDI      = 0x11000000,
     I3401_ADDSI     = 0x31000000,
@@ -457,6 +461,20 @@ static void tcg_out_insn_3207(TCGContext *s, AArch64Insn insn, TCGReg rn)
     tcg_out32(s, insn | rn << 5);
 }
 
+static void tcg_out_insn_3314(TCGContext *s, AArch64Insn insn,
+                              TCGReg r1, TCGReg r2, TCGReg rn,
+                              tcg_target_long ofs, bool pre, bool w)
+{
+    insn |= 1u << 31; /* ext */
+    insn |= pre << 24;
+    insn |= w << 23;
+
+    assert(ofs >= -0x200 && ofs < 0x200 && (ofs & 7) == 0);
+    insn |= (ofs & (0x7f << 3)) << (15 - 3);
+
+    tcg_out32(s, insn | r2 << 10 | rn << 5 | r1);
+}
+
 static void tcg_out_insn_3401(TCGContext *s, AArch64Insn insn, TCGType ext,
                               TCGReg rd, TCGReg rn, uint64_t aimm)
 {
@@ -1292,56 +1310,6 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int opc)
 
 static uint8_t *tb_ret_addr;
 
-/* callee stack use example:
-   stp     x29, x30, [sp,#-32]!
-   mov     x29, sp
-   stp     x1, x2, [sp,#16]
-   ...
-   ldp     x1, x2, [sp,#16]
-   ldp     x29, x30, [sp],#32
-   ret
-*/
-
-/* push r1 and r2, and alloc stack space for a total of
-   alloc_n elements (1 element=16 bytes, must be between 1 and 31. */
-static inline void tcg_out_push_pair(TCGContext *s, TCGReg addr,
-                                     TCGReg r1, TCGReg r2, int alloc_n)
-{
-    /* using indexed scaled simm7 STP 0x28800000 | (ext) | 0x01000000 (pre-idx)
-       | alloc_n * (-1) << 16 | r2 << 10 | addr << 5 | r1 */
-    assert(alloc_n > 0 && alloc_n < 0x20);
-    alloc_n = (-alloc_n) & 0x3f;
-    tcg_out32(s, 0xa9800000 | alloc_n << 16 | r2 << 10 | addr << 5 | r1);
-}
-
-/* dealloc stack space for a total of alloc_n elements and pop r1, r2.  */
-static inline void tcg_out_pop_pair(TCGContext *s, TCGReg addr,
-                                    TCGReg r1, TCGReg r2, int alloc_n)
-{
-    /* using indexed scaled simm7 LDP 0x28c00000 | (ext) | nothing (post-idx)
-       | alloc_n << 16 | r2 << 10 | addr << 5 | r1 */
-    assert(alloc_n > 0 && alloc_n < 0x20);
-    tcg_out32(s, 0xa8c00000 | alloc_n << 16 | r2 << 10 | addr << 5 | r1);
-}
-
-static inline void tcg_out_store_pair(TCGContext *s, TCGReg addr,
-                                      TCGReg r1, TCGReg r2, int idx)
-{
-    /* using register pair offset simm7 STP 0x29000000 | (ext)
-       | idx << 16 | r2 << 10 | addr << 5 | r1 */
-    assert(idx > 0 && idx < 0x20);
-    tcg_out32(s, 0xa9000000 | idx << 16 | r2 << 10 | addr << 5 | r1);
-}
-
-static inline void tcg_out_load_pair(TCGContext *s, TCGReg addr,
-                                     TCGReg r1, TCGReg r2, int idx)
-{
-    /* using register pair offset simm7 LDP 0x29400000 | (ext)
-       | idx << 16 | r2 << 10 | addr << 5 | r1 */
-    assert(idx > 0 && idx < 0x20);
-    tcg_out32(s, 0xa9400000 | idx << 16 | r2 << 10 | addr << 5 | r1);
-}
-
 static void tcg_out_op(TCGContext *s, TCGOpcode opc,
                        const TCGArg args[TCG_MAX_OP_ARGS],
                        const int const_args[TCG_MAX_OP_ARGS])
@@ -1887,33 +1855,32 @@ static void tcg_target_qemu_prologue(TCGContext *s)
     TCGReg r;
 
     /* save pairs             (FP, LR) and (X19, X20) .. (X27, X28) */
-    frame_size_callee_saved = (1) + (TCG_REG_X28 - TCG_REG_X19) / 2 + 1;
+    frame_size_callee_saved = 16 + (TCG_REG_X28 - TCG_REG_X19 + 1) * 8;
 
     /* frame size requirement for TCG local variables */
     frame_size_tcg_locals = TCG_STATIC_CALL_ARGS_SIZE
         + CPU_TEMP_BUF_NLONGS * sizeof(long)
         + (TCG_TARGET_STACK_ALIGN - 1);
     frame_size_tcg_locals &= ~(TCG_TARGET_STACK_ALIGN - 1);
-    frame_size_tcg_locals /= TCG_TARGET_STACK_ALIGN;
 
-    /* push (FP, LR) and update sp */
-    tcg_out_push_pair(s, TCG_REG_SP,
-                      TCG_REG_FP, TCG_REG_LR, frame_size_callee_saved);
+    /* Push (FP, LR) and allocate space for all saved registers.  */
+    tcg_out_insn(s, 3314, STP, TCG_REG_FP, TCG_REG_LR,
+                 TCG_REG_SP, -frame_size_callee_saved, 1, 1);
 
     /* Set up frame pointer for canonical unwinding.  */
     tcg_out_movr_sp(s, TCG_TYPE_I64, TCG_REG_FP, TCG_REG_SP);
 
     /* Store callee-preserved regs x19..x28.  */
     for (r = TCG_REG_X19; r <= TCG_REG_X27; r += 2) {
-        int idx = (r - TCG_REG_X19) / 2 + 1;
-        tcg_out_store_pair(s, TCG_REG_SP, r, r + 1, idx);
+        int ofs = (r - TCG_REG_X19 + 2) * 8;
+        tcg_out_insn(s, 3314, STP, r, r + 1, TCG_REG_SP, ofs, 1, 0);
     }
 
     /* Make stack space for TCG locals.  */
     tcg_out_insn(s, 3401, SUBI, TCG_TYPE_I64, TCG_REG_SP, TCG_REG_SP,
-                 frame_size_tcg_locals * TCG_TARGET_STACK_ALIGN);
+                 frame_size_tcg_locals);
 
-    /* inform TCG about how to find TCG locals with register, offset, size */
+    /* Inform TCG about how to find TCG locals with register, offset, size.  */
     tcg_set_frame(s, TCG_REG_SP, TCG_STATIC_CALL_ARGS_SIZE,
                   CPU_TEMP_BUF_NLONGS * sizeof(long));
 
@@ -1931,17 +1898,16 @@ static void tcg_target_qemu_prologue(TCGContext *s)
 
     /* Remove TCG locals stack space.  */
     tcg_out_insn(s, 3401, ADDI, TCG_TYPE_I64, TCG_REG_SP, TCG_REG_SP,
-                 frame_size_tcg_locals * TCG_TARGET_STACK_ALIGN);
+                 frame_size_tcg_locals);
 
-    /* restore registers x19..x28.
-       FP must be preserved, so it still points to callee_saved area */
+    /* Restore registers x19..x28.  */
     for (r = TCG_REG_X19; r <= TCG_REG_X27; r += 2) {
-        int idx = (r - TCG_REG_X19) / 2 + 1;
-        tcg_out_load_pair(s, TCG_REG_SP, r, r + 1, idx);
+        int ofs = (r - TCG_REG_X19 + 2) * 8;
+        tcg_out_insn(s, 3314, LDP, r, r + 1, TCG_REG_SP, ofs, 1, 0);
     }
 
-    /* pop (FP, LR), restore SP to previous frame, return */
-    tcg_out_pop_pair(s, TCG_REG_SP,
-                     TCG_REG_FP, TCG_REG_LR, frame_size_callee_saved);
+    /* Pop (FP, LR), restore SP to previous frame.  */
+    tcg_out_insn(s, 3314, LDP, TCG_REG_FP, TCG_REG_LR,
+                 TCG_REG_SP, frame_size_callee_saved, 0, 1);
     tcg_out_insn(s, 3207, RET, TCG_REG_LR);
 }
-- 
1.9.0

[Qemu-devel] [PATCH v4 13/25] tcg-aarch64: Implement tcg_register_jit

[Richard Henderson]

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 84 +++++++++++++++++++++++++++++++++++++++---------
 1 file changed, 69 insertions(+), 15 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 5cffe50..4414bd1 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -1848,24 +1848,29 @@ static void tcg_target_init(TCGContext *s)
     tcg_add_target_add_op_defs(aarch64_op_defs);
 }
 
+/* Saving pairs: (X19, X20) .. (X27, X28), (X29(fp), X30(lr)).  */
+#define PUSH_SIZE  ((30 - 19 + 1) * 8)
+
+#define FRAME_SIZE \
+    ((PUSH_SIZE \
+      + TCG_STATIC_CALL_ARGS_SIZE \
+      + CPU_TEMP_BUF_NLONGS * sizeof(long) \
+      + TCG_TARGET_STACK_ALIGN - 1) \
+     & ~(TCG_TARGET_STACK_ALIGN - 1))
+
+/* We're expecting a 2 byte uleb128 encoded value.  */
+QEMU_BUILD_BUG_ON(FRAME_SIZE >= (1 << 14));
+
+/* We're expecting to use a single ADDI insn.  */
+QEMU_BUILD_BUG_ON(FRAME_SIZE - PUSH_SIZE > 0xfff);
+
 static void tcg_target_qemu_prologue(TCGContext *s)
 {
-    /* NB: frame sizes are in 16 byte stack units! */
-    int frame_size_callee_saved, frame_size_tcg_locals;
     TCGReg r;
 
-    /* save pairs             (FP, LR) and (X19, X20) .. (X27, X28) */
-    frame_size_callee_saved = 16 + (TCG_REG_X28 - TCG_REG_X19 + 1) * 8;
-
-    /* frame size requirement for TCG local variables */
-    frame_size_tcg_locals = TCG_STATIC_CALL_ARGS_SIZE
-        + CPU_TEMP_BUF_NLONGS * sizeof(long)
-        + (TCG_TARGET_STACK_ALIGN - 1);
-    frame_size_tcg_locals &= ~(TCG_TARGET_STACK_ALIGN - 1);
-
     /* Push (FP, LR) and allocate space for all saved registers.  */
     tcg_out_insn(s, 3314, STP, TCG_REG_FP, TCG_REG_LR,
-                 TCG_REG_SP, -frame_size_callee_saved, 1, 1);
+                 TCG_REG_SP, -PUSH_SIZE, 1, 1);
 
     /* Set up frame pointer for canonical unwinding.  */
     tcg_out_movr_sp(s, TCG_TYPE_I64, TCG_REG_FP, TCG_REG_SP);
@@ -1878,7 +1883,7 @@ static void tcg_target_qemu_prologue(TCGContext *s)
 
     /* Make stack space for TCG locals.  */
     tcg_out_insn(s, 3401, SUBI, TCG_TYPE_I64, TCG_REG_SP, TCG_REG_SP,
-                 frame_size_tcg_locals);
+                 FRAME_SIZE - PUSH_SIZE);
 
     /* Inform TCG about how to find TCG locals with register, offset, size.  */
     tcg_set_frame(s, TCG_REG_SP, TCG_STATIC_CALL_ARGS_SIZE,
@@ -1898,7 +1903,7 @@ static void tcg_target_qemu_prologue(TCGContext *s)
 
     /* Remove TCG locals stack space.  */
     tcg_out_insn(s, 3401, ADDI, TCG_TYPE_I64, TCG_REG_SP, TCG_REG_SP,
-                 frame_size_tcg_locals);
+                 FRAME_SIZE - PUSH_SIZE);
 
     /* Restore registers x19..x28.  */
     for (r = TCG_REG_X19; r <= TCG_REG_X27; r += 2) {
@@ -1908,6 +1913,55 @@ static void tcg_target_qemu_prologue(TCGContext *s)
 
     /* Pop (FP, LR), restore SP to previous frame.  */
     tcg_out_insn(s, 3314, LDP, TCG_REG_FP, TCG_REG_LR,
-                 TCG_REG_SP, frame_size_callee_saved, 0, 1);
+                 TCG_REG_SP, PUSH_SIZE, 0, 1);
     tcg_out_insn(s, 3207, RET, TCG_REG_LR);
 }
+
+typedef struct {
+    DebugFrameCIE cie;
+    DebugFrameFDEHeader fde;
+    uint8_t fde_def_cfa[4];
+    uint8_t fde_reg_ofs[24];
+} DebugFrame;
+
+#define ELF_HOST_MACHINE EM_AARCH64
+
+static DebugFrame debug_frame = {
+    .cie.len = sizeof(DebugFrameCIE)-4, /* length after .len member */
+    .cie.id = -1,
+    .cie.version = 1,
+    .cie.code_align = 1,
+    .cie.data_align = 0x78,             /* sleb128 -8 */
+    .cie.return_column = TCG_REG_LR,
+
+    /* Total FDE size does not include the "len" member.  */
+    .fde.len = sizeof(DebugFrame) - offsetof(DebugFrame, fde.cie_offset),
+
+    .fde_def_cfa = {
+        12, TCG_REG_SP,                 /* DW_CFA_def_cfa sp, ... */
+        (FRAME_SIZE & 0x7f) | 0x80,     /* ... uleb128 FRAME_SIZE */
+        (FRAME_SIZE >> 7)
+    },
+    .fde_reg_ofs = {
+        0x80 + 28, 1,                   /* DW_CFA_offset, x28,  -8 */
+        0x80 + 27, 2,                   /* DW_CFA_offset, x27, -16 */
+        0x80 + 26, 3,                   /* DW_CFA_offset, x26, -24 */
+        0x80 + 25, 4,                   /* DW_CFA_offset, x25, -32 */
+        0x80 + 24, 5,                   /* DW_CFA_offset, x24, -40 */
+        0x80 + 23, 6,                   /* DW_CFA_offset, x23, -48 */
+        0x80 + 22, 7,                   /* DW_CFA_offset, x22, -56 */
+        0x80 + 21, 8,                   /* DW_CFA_offset, x21, -64 */
+        0x80 + 20, 9,                   /* DW_CFA_offset, x20, -72 */
+        0x80 + 19, 10,                  /* DW_CFA_offset, x1p, -80 */
+        0x80 + 30, 11,                  /* DW_CFA_offset,  lr, -88 */
+        0x80 + 29, 12,                  /* DW_CFA_offset,  fp, -96 */
+    }
+};
+
+void tcg_register_jit(void *buf, size_t buf_size)
+{
+    debug_frame.fde.func_start = (intptr_t)buf;
+    debug_frame.fde.func_len = buf_size;
+
+    tcg_register_jit_int(buf, buf_size, &debug_frame, sizeof(debug_frame));
+}
-- 
1.9.0

[Qemu-devel] [PATCH v4 14/25] tcg-aarch64: Avoid add with zero in tlb load

[Richard Henderson]

Some guest env are small enough to reach the tlb with only a 12-bit addition.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 28 +++++++++++++++++++---------
 1 file changed, 19 insertions(+), 9 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 4414bd1..5186311 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -1128,47 +1128,57 @@ static void add_qemu_ldst_label(TCGContext *s, int is_ld, int opc,
    slow path for the failure case, which will be patched later when finalizing
    the slow path. Generated code returns the host addend in X1,
    clobbers X0,X2,X3,TMP. */
-static void tcg_out_tlb_read(TCGContext *s, TCGReg addr_reg,
-            int s_bits, uint8_t **label_ptr, int mem_index, int is_read)
+static void tcg_out_tlb_read(TCGContext *s, TCGReg addr_reg, int s_bits,
+                             uint8_t **label_ptr, int mem_index, bool is_read)
 {
     TCGReg base = TCG_AREG0;
     int tlb_offset = is_read ?
         offsetof(CPUArchState, tlb_table[mem_index][0].addr_read)
         : offsetof(CPUArchState, tlb_table[mem_index][0].addr_write);
+
     /* Extract the TLB index from the address into X0.
        X0<CPU_TLB_BITS:0> =
        addr_reg<TARGET_PAGE_BITS+CPU_TLB_BITS:TARGET_PAGE_BITS> */
-    tcg_out_ubfm(s, (TARGET_LONG_BITS == 64), TCG_REG_X0, addr_reg,
+    tcg_out_ubfm(s, TARGET_LONG_BITS == 64, TCG_REG_X0, addr_reg,
                  TARGET_PAGE_BITS, TARGET_PAGE_BITS + CPU_TLB_BITS);
+
     /* Store the page mask part of the address and the low s_bits into X3.
        Later this allows checking for equality and alignment at the same time.
        X3 = addr_reg & (PAGE_MASK | ((1 << s_bits) - 1)) */
     tcg_out_logicali(s, I3404_ANDI, TARGET_LONG_BITS == 64, TCG_REG_X3,
                      addr_reg, TARGET_PAGE_MASK | ((1 << s_bits) - 1));
+
     /* Add any "high bits" from the tlb offset to the env address into X2,
        to take advantage of the LSL12 form of the ADDI instruction.
        X2 = env + (tlb_offset & 0xfff000) */
-    tcg_out_insn(s, 3401, ADDI, TCG_TYPE_I64, TCG_REG_X2, base,
-                 tlb_offset & 0xfff000);
+    if (tlb_offset & 0xfff000) {
+        tcg_out_insn(s, 3401, ADDI, TCG_TYPE_I64, TCG_REG_X2, base,
+                     tlb_offset & 0xfff000);
+        base = TCG_REG_X2;
+    }
+
     /* Merge the tlb index contribution into X2.
        X2 = X2 + (X0 << CPU_TLB_ENTRY_BITS) */
-    tcg_out_insn(s, 3502S, ADD_LSL, 1, TCG_REG_X2, TCG_REG_X2,
+    tcg_out_insn(s, 3502S, ADD_LSL, TCG_TYPE_I64, TCG_REG_X2, base,
                  TCG_REG_X0, CPU_TLB_ENTRY_BITS);
+
     /* Merge "low bits" from tlb offset, load the tlb comparator into X0.
        X0 = load [X2 + (tlb_offset & 0x000fff)] */
     tcg_out_ldst(s, TARGET_LONG_BITS == 64 ? LDST_64 : LDST_32,
-                 LDST_LD, TCG_REG_X0, TCG_REG_X2,
-                 (tlb_offset & 0xfff));
+                 LDST_LD, TCG_REG_X0, TCG_REG_X2, tlb_offset & 0xfff);
+
     /* Load the tlb addend. Do that early to avoid stalling.
        X1 = load [X2 + (tlb_offset & 0xfff) + offsetof(addend)] */
     tcg_out_ldst(s, LDST_64, LDST_LD, TCG_REG_X1, TCG_REG_X2,
                  (tlb_offset & 0xfff) + (offsetof(CPUTLBEntry, addend)) -
                  (is_read ? offsetof(CPUTLBEntry, addr_read)
                   : offsetof(CPUTLBEntry, addr_write)));
+
     /* Perform the address comparison. */
     tcg_out_cmp(s, (TARGET_LONG_BITS == 64), TCG_REG_X0, TCG_REG_X3, 0);
-    *label_ptr = s->code_ptr;
+
     /* If not equal, we jump to the slow path. */
+    *label_ptr = s->code_ptr;
     tcg_out_goto_cond_noaddr(s, TCG_COND_NE);
 }
 
-- 
1.9.0

[Qemu-devel] [PATCH v4 15/25] tcg-aarch64: Use tcg_out_call for qemu_ld/st

[Richard Henderson]

In some cases, a direct branch will be in range.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 5186311..4729d11 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -1081,8 +1081,7 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
     tcg_out_movr(s, TARGET_LONG_BITS == 64, TCG_REG_X1, lb->addrlo_reg);
     tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_X2, lb->mem_index);
     tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_X3, (intptr_t)lb->raddr);
-    tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP, (intptr_t)qemu_ld_helpers[size]);
-    tcg_out_callr(s, TCG_REG_TMP);
+    tcg_out_call(s, (intptr_t)qemu_ld_helpers[size]);
     if (opc & MO_SIGN) {
         tcg_out_sxt(s, TCG_TYPE_I64, size, lb->datalo_reg, TCG_REG_X0);
     } else {
@@ -1103,8 +1102,7 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
     tcg_out_movr(s, size == MO_64, TCG_REG_X2, lb->datalo_reg);
     tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_X3, lb->mem_index);
     tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_X4, (intptr_t)lb->raddr);
-    tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP, (intptr_t)qemu_st_helpers[size]);
-    tcg_out_callr(s, TCG_REG_TMP);
+    tcg_out_call(s, (intptr_t)qemu_st_helpers[size]);
     tcg_out_goto(s, (intptr_t)lb->raddr);
 }
 
-- 
1.9.0

[Qemu-devel] [PATCH v4 16/25] tcg-aarch64: Use ADR to pass the return address to the ld/st helpers

[Richard Henderson]

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 4729d11..5d19e27 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -1070,6 +1070,13 @@ static const void * const qemu_st_helpers[4] = {
     helper_ret_stq_mmu,
 };
 
+static inline void tcg_out_adr(TCGContext *s, TCGReg rd, uintptr_t addr)
+{
+    addr -= (uintptr_t)s->code_ptr;
+    assert(addr == sextract64(addr, 0, 21));
+    tcg_out_insn(s, 3406, ADR, rd, addr);
+}
+
 static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
 {
     TCGMemOp opc = lb->opc;
@@ -1080,7 +1087,7 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
     tcg_out_movr(s, TCG_TYPE_I64, TCG_REG_X0, TCG_AREG0);
     tcg_out_movr(s, TARGET_LONG_BITS == 64, TCG_REG_X1, lb->addrlo_reg);
     tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_X2, lb->mem_index);
-    tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_X3, (intptr_t)lb->raddr);
+    tcg_out_adr(s, TCG_REG_X3, (intptr_t)lb->raddr);
     tcg_out_call(s, (intptr_t)qemu_ld_helpers[size]);
     if (opc & MO_SIGN) {
         tcg_out_sxt(s, TCG_TYPE_I64, size, lb->datalo_reg, TCG_REG_X0);
@@ -1101,7 +1108,7 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
     tcg_out_movr(s, TARGET_LONG_BITS == 64, TCG_REG_X1, lb->addrlo_reg);
     tcg_out_movr(s, size == MO_64, TCG_REG_X2, lb->datalo_reg);
     tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_X3, lb->mem_index);
-    tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_X4, (intptr_t)lb->raddr);
+    tcg_out_adr(s, TCG_REG_X4, (intptr_t)lb->raddr);
     tcg_out_call(s, (intptr_t)qemu_st_helpers[size]);
     tcg_out_goto(s, (intptr_t)lb->raddr);
 }
-- 
1.9.0

[Qemu-devel] [PATCH v4 17/25] tcg-aarch64: Use TCGMemOp in qemu_ld/st

[Richard Henderson]

Making the bswap conditional on the memop instead of a compile-time test.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 131 +++++++++++++++++++++++------------------------
 1 file changed, 63 insertions(+), 68 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 5d19e27..68305ea 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -27,12 +27,6 @@ static const char * const tcg_target_reg_names[TCG_TARGET_NB_REGS] = {
 };
 #endif /* NDEBUG */
 
-#ifdef TARGET_WORDS_BIGENDIAN
- #define TCG_LDST_BSWAP 1
-#else
- #define TCG_LDST_BSWAP 0
-#endif
-
 static const int tcg_target_reg_alloc_order[] = {
     TCG_REG_X20, TCG_REG_X21, TCG_REG_X22, TCG_REG_X23,
     TCG_REG_X24, TCG_REG_X25, TCG_REG_X26, TCG_REG_X27,
@@ -1113,7 +1107,7 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
     tcg_out_goto(s, (intptr_t)lb->raddr);
 }
 
-static void add_qemu_ldst_label(TCGContext *s, int is_ld, int opc,
+static void add_qemu_ldst_label(TCGContext *s, int is_ld, TCGMemOp opc,
                                 TCGReg data_reg, TCGReg addr_reg,
                                 int mem_index,
                                 uint8_t *raddr, uint8_t *label_ptr)
@@ -1133,7 +1127,7 @@ static void add_qemu_ldst_label(TCGContext *s, int is_ld, int opc,
    slow path for the failure case, which will be patched later when finalizing
    the slow path. Generated code returns the host addend in X1,
    clobbers X0,X2,X3,TMP. */
-static void tcg_out_tlb_read(TCGContext *s, TCGReg addr_reg, int s_bits,
+static void tcg_out_tlb_read(TCGContext *s, TCGReg addr_reg, TCGMemOp s_bits,
                              uint8_t **label_ptr, int mem_index, bool is_read)
 {
     TCGReg base = TCG_AREG0;
@@ -1189,24 +1183,26 @@ static void tcg_out_tlb_read(TCGContext *s, TCGReg addr_reg, int s_bits,
 
 #endif /* CONFIG_SOFTMMU */
 
-static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data_r,
-                                   TCGReg addr_r, TCGReg off_r)
+static void tcg_out_qemu_ld_direct(TCGContext *s, TCGMemOp memop,
+                                   TCGReg data_r, TCGReg addr_r, TCGReg off_r)
 {
-    switch (opc) {
-    case 0:
+    const TCGMemOp bswap = memop & MO_BSWAP;
+
+    switch (memop & MO_SSIZE) {
+    case MO_UB:
         tcg_out_ldst_r(s, LDST_8, LDST_LD, data_r, addr_r, off_r);
         break;
-    case 0 | 4:
+    case MO_SB:
         tcg_out_ldst_r(s, LDST_8, LDST_LD_S_X, data_r, addr_r, off_r);
         break;
-    case 1:
+    case MO_UW:
         tcg_out_ldst_r(s, LDST_16, LDST_LD, data_r, addr_r, off_r);
-        if (TCG_LDST_BSWAP) {
+        if (bswap) {
             tcg_out_rev16(s, TCG_TYPE_I32, data_r, data_r);
         }
         break;
-    case 1 | 4:
-        if (TCG_LDST_BSWAP) {
+    case MO_SW:
+        if (bswap) {
             tcg_out_ldst_r(s, LDST_16, LDST_LD, data_r, addr_r, off_r);
             tcg_out_rev16(s, TCG_TYPE_I32, data_r, data_r);
             tcg_out_sxt(s, TCG_TYPE_I64, MO_16, data_r, data_r);
@@ -1214,14 +1210,14 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data_r,
             tcg_out_ldst_r(s, LDST_16, LDST_LD_S_X, data_r, addr_r, off_r);
         }
         break;
-    case 2:
+    case MO_UL:
         tcg_out_ldst_r(s, LDST_32, LDST_LD, data_r, addr_r, off_r);
-        if (TCG_LDST_BSWAP) {
+        if (bswap) {
             tcg_out_rev(s, TCG_TYPE_I32, data_r, data_r);
         }
         break;
-    case 2 | 4:
-        if (TCG_LDST_BSWAP) {
+    case MO_SL:
+        if (bswap) {
             tcg_out_ldst_r(s, LDST_32, LDST_LD, data_r, addr_r, off_r);
             tcg_out_rev(s, TCG_TYPE_I32, data_r, data_r);
             tcg_out_sxt(s, TCG_TYPE_I64, MO_32, data_r, data_r);
@@ -1229,9 +1225,9 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data_r,
             tcg_out_ldst_r(s, LDST_32, LDST_LD_S_X, data_r, addr_r, off_r);
         }
         break;
-    case 3:
+    case MO_Q:
         tcg_out_ldst_r(s, LDST_64, LDST_LD, data_r, addr_r, off_r);
-        if (TCG_LDST_BSWAP) {
+        if (bswap) {
             tcg_out_rev(s, TCG_TYPE_I64, data_r, data_r);
         }
         break;
@@ -1240,47 +1236,47 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, int opc, TCGReg data_r,
     }
 }
 
-static void tcg_out_qemu_st_direct(TCGContext *s, int opc, TCGReg data_r,
-                                   TCGReg addr_r, TCGReg off_r)
+static void tcg_out_qemu_st_direct(TCGContext *s, TCGMemOp memop,
+                                   TCGReg data_r, TCGReg addr_r, TCGReg off_r)
 {
-    switch (opc) {
-    case 0:
+    const TCGMemOp bswap = memop & MO_BSWAP;
+
+    switch (memop & MO_SIZE) {
+    case MO_8:
         tcg_out_ldst_r(s, LDST_8, LDST_ST, data_r, addr_r, off_r);
         break;
-    case 1:
-        if (TCG_LDST_BSWAP) {
+    case MO_16:
+        if (bswap) {
             tcg_out_rev16(s, TCG_TYPE_I32, TCG_REG_TMP, data_r);
-            tcg_out_ldst_r(s, LDST_16, LDST_ST, TCG_REG_TMP, addr_r, off_r);
-        } else {
-            tcg_out_ldst_r(s, LDST_16, LDST_ST, data_r, addr_r, off_r);
+            data_r = TCG_REG_TMP;
         }
+        tcg_out_ldst_r(s, LDST_16, LDST_ST, data_r, addr_r, off_r);
         break;
-    case 2:
-        if (TCG_LDST_BSWAP) {
+    case MO_32:
+        if (bswap) {
             tcg_out_rev(s, TCG_TYPE_I32, TCG_REG_TMP, data_r);
-            tcg_out_ldst_r(s, LDST_32, LDST_ST, TCG_REG_TMP, addr_r, off_r);
-        } else {
-            tcg_out_ldst_r(s, LDST_32, LDST_ST, data_r, addr_r, off_r);
+            data_r = TCG_REG_TMP;
         }
+        tcg_out_ldst_r(s, LDST_32, LDST_ST, data_r, addr_r, off_r);
         break;
-    case 3:
-        if (TCG_LDST_BSWAP) {
+    case MO_64:
+        if (bswap) {
             tcg_out_rev(s, TCG_TYPE_I64, TCG_REG_TMP, data_r);
-            tcg_out_ldst_r(s, LDST_64, LDST_ST, TCG_REG_TMP, addr_r, off_r);
-        } else {
-            tcg_out_ldst_r(s, LDST_64, LDST_ST, data_r, addr_r, off_r);
+            data_r = TCG_REG_TMP;
         }
+        tcg_out_ldst_r(s, LDST_64, LDST_ST, data_r, addr_r, off_r);
         break;
     default:
         tcg_abort();
     }
 }
 
-static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int opc)
+static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, TCGMemOp memop)
 {
     TCGReg addr_reg, data_reg;
 #ifdef CONFIG_SOFTMMU
-    int mem_index, s_bits;
+    int mem_index;
+    TCGMemOp s_bits;
     uint8_t *label_ptr;
 #endif
     data_reg = args[0];
@@ -1288,22 +1284,23 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int opc)
 
 #ifdef CONFIG_SOFTMMU
     mem_index = args[2];
-    s_bits = opc & 3;
+    s_bits = memop & MO_SIZE;
     tcg_out_tlb_read(s, addr_reg, s_bits, &label_ptr, mem_index, 1);
-    tcg_out_qemu_ld_direct(s, opc, data_reg, addr_reg, TCG_REG_X1);
-    add_qemu_ldst_label(s, 1, opc, data_reg, addr_reg,
+    tcg_out_qemu_ld_direct(s, memop, data_reg, addr_reg, TCG_REG_X1);
+    add_qemu_ldst_label(s, 1, memop, data_reg, addr_reg,
                         mem_index, s->code_ptr, label_ptr);
 #else /* !CONFIG_SOFTMMU */
-    tcg_out_qemu_ld_direct(s, opc, data_reg, addr_reg,
+    tcg_out_qemu_ld_direct(s, memop, data_reg, addr_reg,
                            GUEST_BASE ? TCG_REG_GUEST_BASE : TCG_REG_XZR);
 #endif /* CONFIG_SOFTMMU */
 }
 
-static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int opc)
+static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, TCGMemOp memop)
 {
     TCGReg addr_reg, data_reg;
 #ifdef CONFIG_SOFTMMU
-    int mem_index, s_bits;
+    int mem_index;
+    TCGMemOp s_bits;
     uint8_t *label_ptr;
 #endif
     data_reg = args[0];
@@ -1311,14 +1308,14 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int opc)
 
 #ifdef CONFIG_SOFTMMU
     mem_index = args[2];
-    s_bits = opc & 3;
+    s_bits = memop & MO_SIZE;
 
     tcg_out_tlb_read(s, addr_reg, s_bits, &label_ptr, mem_index, 0);
-    tcg_out_qemu_st_direct(s, opc, data_reg, addr_reg, TCG_REG_X1);
-    add_qemu_ldst_label(s, 0, opc, data_reg, addr_reg,
+    tcg_out_qemu_st_direct(s, memop, data_reg, addr_reg, TCG_REG_X1);
+    add_qemu_ldst_label(s, 0, memop, data_reg, addr_reg,
                         mem_index, s->code_ptr, label_ptr);
 #else /* !CONFIG_SOFTMMU */
-    tcg_out_qemu_st_direct(s, opc, data_reg, addr_reg,
+    tcg_out_qemu_st_direct(s, memop, data_reg, addr_reg,
                            GUEST_BASE ? TCG_REG_GUEST_BASE : TCG_REG_XZR);
 #endif /* CONFIG_SOFTMMU */
 }
@@ -1591,40 +1588,38 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
         break;
 
     case INDEX_op_qemu_ld8u:
-        tcg_out_qemu_ld(s, args, 0 | 0);
+        tcg_out_qemu_ld(s, args, MO_UB);
         break;
     case INDEX_op_qemu_ld8s:
-        tcg_out_qemu_ld(s, args, 4 | 0);
+        tcg_out_qemu_ld(s, args, MO_SB);
         break;
     case INDEX_op_qemu_ld16u:
-        tcg_out_qemu_ld(s, args, 0 | 1);
+        tcg_out_qemu_ld(s, args, MO_TEUW);
         break;
     case INDEX_op_qemu_ld16s:
-        tcg_out_qemu_ld(s, args, 4 | 1);
+        tcg_out_qemu_ld(s, args, MO_TESW);
         break;
     case INDEX_op_qemu_ld32u:
-        tcg_out_qemu_ld(s, args, 0 | 2);
+    case INDEX_op_qemu_ld32:
+        tcg_out_qemu_ld(s, args, MO_TEUL);
         break;
     case INDEX_op_qemu_ld32s:
-        tcg_out_qemu_ld(s, args, 4 | 2);
-        break;
-    case INDEX_op_qemu_ld32:
-        tcg_out_qemu_ld(s, args, 0 | 2);
+        tcg_out_qemu_ld(s, args, MO_TESL);
         break;
     case INDEX_op_qemu_ld64:
-        tcg_out_qemu_ld(s, args, 0 | 3);
+        tcg_out_qemu_ld(s, args, MO_TEQ);
         break;
     case INDEX_op_qemu_st8:
-        tcg_out_qemu_st(s, args, 0);
+        tcg_out_qemu_st(s, args, MO_UB);
         break;
     case INDEX_op_qemu_st16:
-        tcg_out_qemu_st(s, args, 1);
+        tcg_out_qemu_st(s, args, MO_TEUW);
         break;
     case INDEX_op_qemu_st32:
-        tcg_out_qemu_st(s, args, 2);
+        tcg_out_qemu_st(s, args, MO_TEUL);
         break;
     case INDEX_op_qemu_st64:
-        tcg_out_qemu_st(s, args, 3);
+        tcg_out_qemu_st(s, args, MO_TEQ);
         break;
 
     case INDEX_op_bswap32_i64:
-- 
1.9.0

[Qemu-devel] [PATCH v4 18/25] tcg-aarch64: Pass qemu_ld/st arguments directly

[Richard Henderson]

Instead of passing them the "args" array.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 49 +++++++++++++++++-------------------------------
 1 file changed, 17 insertions(+), 32 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 68305ea..3a2955f 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -1271,20 +1271,13 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGMemOp memop,
     }
 }
 
-static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, TCGMemOp memop)
+static void tcg_out_qemu_ld(TCGContext *s, TCGReg data_reg, TCGReg addr_reg,
+                            TCGMemOp memop, int mem_index)
 {
-    TCGReg addr_reg, data_reg;
 #ifdef CONFIG_SOFTMMU
-    int mem_index;
-    TCGMemOp s_bits;
+    TCGMemOp s_bits = memop & MO_SIZE;
     uint8_t *label_ptr;
-#endif
-    data_reg = args[0];
-    addr_reg = args[1];
 
-#ifdef CONFIG_SOFTMMU
-    mem_index = args[2];
-    s_bits = memop & MO_SIZE;
     tcg_out_tlb_read(s, addr_reg, s_bits, &label_ptr, mem_index, 1);
     tcg_out_qemu_ld_direct(s, memop, data_reg, addr_reg, TCG_REG_X1);
     add_qemu_ldst_label(s, 1, memop, data_reg, addr_reg,
@@ -1295,20 +1288,12 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, TCGMemOp memop)
 #endif /* CONFIG_SOFTMMU */
 }
 
-static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, TCGMemOp memop)
+static void tcg_out_qemu_st(TCGContext *s, TCGReg data_reg, TCGReg addr_reg,
+                            TCGMemOp memop, int mem_index)
 {
-    TCGReg addr_reg, data_reg;
 #ifdef CONFIG_SOFTMMU
-    int mem_index;
-    TCGMemOp s_bits;
+    TCGMemOp s_bits = memop & MO_SIZE;
     uint8_t *label_ptr;
-#endif
-    data_reg = args[0];
-    addr_reg = args[1];
-
-#ifdef CONFIG_SOFTMMU
-    mem_index = args[2];
-    s_bits = memop & MO_SIZE;
 
     tcg_out_tlb_read(s, addr_reg, s_bits, &label_ptr, mem_index, 0);
     tcg_out_qemu_st_direct(s, memop, data_reg, addr_reg, TCG_REG_X1);
@@ -1588,38 +1573,38 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
         break;
 
     case INDEX_op_qemu_ld8u:
-        tcg_out_qemu_ld(s, args, MO_UB);
+        tcg_out_qemu_ld(s, a0, a1, MO_UB, a2);
         break;
     case INDEX_op_qemu_ld8s:
-        tcg_out_qemu_ld(s, args, MO_SB);
+        tcg_out_qemu_ld(s, a0, a1, MO_SB, a2);
         break;
     case INDEX_op_qemu_ld16u:
-        tcg_out_qemu_ld(s, args, MO_TEUW);
+        tcg_out_qemu_ld(s, a0, a1, MO_TEUW, a2);
         break;
     case INDEX_op_qemu_ld16s:
-        tcg_out_qemu_ld(s, args, MO_TESW);
+        tcg_out_qemu_ld(s, a0, a1, MO_TESW, a2);
         break;
     case INDEX_op_qemu_ld32u:
     case INDEX_op_qemu_ld32:
-        tcg_out_qemu_ld(s, args, MO_TEUL);
+        tcg_out_qemu_ld(s, a0, a1, MO_TEUL, a2);
         break;
     case INDEX_op_qemu_ld32s:
-        tcg_out_qemu_ld(s, args, MO_TESL);
+        tcg_out_qemu_ld(s, a0, a1, MO_TESL, a2);
         break;
     case INDEX_op_qemu_ld64:
-        tcg_out_qemu_ld(s, args, MO_TEQ);
+        tcg_out_qemu_ld(s, a0, a1, MO_TEQ, a2);
         break;
     case INDEX_op_qemu_st8:
-        tcg_out_qemu_st(s, args, MO_UB);
+        tcg_out_qemu_st(s, a0, a1, MO_UB, a2);
         break;
     case INDEX_op_qemu_st16:
-        tcg_out_qemu_st(s, args, MO_TEUW);
+        tcg_out_qemu_st(s, a0, a1, MO_TEUW, a2);
         break;
     case INDEX_op_qemu_st32:
-        tcg_out_qemu_st(s, args, MO_TEUL);
+        tcg_out_qemu_st(s, a0, a1, MO_TEUL, a2);
         break;
     case INDEX_op_qemu_st64:
-        tcg_out_qemu_st(s, args, MO_TEQ);
+        tcg_out_qemu_st(s, a0, a1, MO_TEQ, a2);
         break;
 
     case INDEX_op_bswap32_i64:
-- 
1.9.0

[Qemu-devel] [PATCH v4 19/25] tcg-aarch64: Implement TCG_TARGET_HAS_new_ldst

[Richard Henderson]

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 89 ++++++++++++++++--------------------------------
 tcg/aarch64/tcg-target.h |  2 +-
 2 files changed, 31 insertions(+), 60 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 3a2955f..34e477d 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -1047,21 +1047,27 @@ static inline void tcg_out_addsub2(TCGContext *s, int ext, TCGReg rl,
 /* helper signature: helper_ret_ld_mmu(CPUState *env, target_ulong addr,
  *                                     int mmu_idx, uintptr_t ra)
  */
-static const void * const qemu_ld_helpers[4] = {
-    helper_ret_ldub_mmu,
-    helper_ret_lduw_mmu,
-    helper_ret_ldul_mmu,
-    helper_ret_ldq_mmu,
+static const void * const qemu_ld_helpers[16] = {
+    [MO_UB]   = helper_ret_ldub_mmu,
+    [MO_LEUW] = helper_le_lduw_mmu,
+    [MO_LEUL] = helper_le_ldul_mmu,
+    [MO_LEQ]  = helper_le_ldq_mmu,
+    [MO_BEUW] = helper_be_lduw_mmu,
+    [MO_BEUL] = helper_be_ldul_mmu,
+    [MO_BEQ]  = helper_be_ldq_mmu,
 };
 
 /* helper signature: helper_ret_st_mmu(CPUState *env, target_ulong addr,
  *                                     uintxx_t val, int mmu_idx, uintptr_t ra)
  */
-static const void * const qemu_st_helpers[4] = {
-    helper_ret_stb_mmu,
-    helper_ret_stw_mmu,
-    helper_ret_stl_mmu,
-    helper_ret_stq_mmu,
+static const void * const qemu_st_helpers[16] = {
+    [MO_UB]   = helper_ret_stb_mmu,
+    [MO_LEUW] = helper_le_stw_mmu,
+    [MO_LEUL] = helper_le_stl_mmu,
+    [MO_LEQ]  = helper_le_stq_mmu,
+    [MO_BEUW] = helper_be_stw_mmu,
+    [MO_BEUL] = helper_be_stl_mmu,
+    [MO_BEQ]  = helper_be_stq_mmu,
 };
 
 static inline void tcg_out_adr(TCGContext *s, TCGReg rd, uintptr_t addr)
@@ -1082,7 +1088,7 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
     tcg_out_movr(s, TARGET_LONG_BITS == 64, TCG_REG_X1, lb->addrlo_reg);
     tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_X2, lb->mem_index);
     tcg_out_adr(s, TCG_REG_X3, (intptr_t)lb->raddr);
-    tcg_out_call(s, (intptr_t)qemu_ld_helpers[size]);
+    tcg_out_call(s, (intptr_t)qemu_ld_helpers[opc & ~MO_SIGN]);
     if (opc & MO_SIGN) {
         tcg_out_sxt(s, TCG_TYPE_I64, size, lb->datalo_reg, TCG_REG_X0);
     } else {
@@ -1094,7 +1100,8 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
 
 static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
 {
-    TCGMemOp size = lb->opc;
+    TCGMemOp opc = lb->opc;
+    TCGMemOp size = opc & MO_SIZE;
 
     reloc_pc19(lb->label_ptr[0], (intptr_t)s->code_ptr);
 
@@ -1103,7 +1110,7 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
     tcg_out_movr(s, size == MO_64, TCG_REG_X2, lb->datalo_reg);
     tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_X3, lb->mem_index);
     tcg_out_adr(s, TCG_REG_X4, (intptr_t)lb->raddr);
-    tcg_out_call(s, (intptr_t)qemu_st_helpers[size]);
+    tcg_out_call(s, (intptr_t)qemu_st_helpers[opc]);
     tcg_out_goto(s, (intptr_t)lb->raddr);
 }
 
@@ -1572,39 +1579,13 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
         tcg_out_insn(s, 3506, CSEL, ext, a0, REG0(3), REG0(4), args[5]);
         break;
 
-    case INDEX_op_qemu_ld8u:
-        tcg_out_qemu_ld(s, a0, a1, MO_UB, a2);
-        break;
-    case INDEX_op_qemu_ld8s:
-        tcg_out_qemu_ld(s, a0, a1, MO_SB, a2);
-        break;
-    case INDEX_op_qemu_ld16u:
-        tcg_out_qemu_ld(s, a0, a1, MO_TEUW, a2);
-        break;
-    case INDEX_op_qemu_ld16s:
-        tcg_out_qemu_ld(s, a0, a1, MO_TESW, a2);
-        break;
-    case INDEX_op_qemu_ld32u:
-    case INDEX_op_qemu_ld32:
-        tcg_out_qemu_ld(s, a0, a1, MO_TEUL, a2);
-        break;
-    case INDEX_op_qemu_ld32s:
-        tcg_out_qemu_ld(s, a0, a1, MO_TESL, a2);
+    case INDEX_op_qemu_ld_i32:
+    case INDEX_op_qemu_ld_i64:
+        tcg_out_qemu_ld(s, a0, a1, a2, args[3]);
         break;
-    case INDEX_op_qemu_ld64:
-        tcg_out_qemu_ld(s, a0, a1, MO_TEQ, a2);
-        break;
-    case INDEX_op_qemu_st8:
-        tcg_out_qemu_st(s, a0, a1, MO_UB, a2);
-        break;
-    case INDEX_op_qemu_st16:
-        tcg_out_qemu_st(s, a0, a1, MO_TEUW, a2);
-        break;
-    case INDEX_op_qemu_st32:
-        tcg_out_qemu_st(s, a0, a1, MO_TEUL, a2);
-        break;
-    case INDEX_op_qemu_st64:
-        tcg_out_qemu_st(s, a0, a1, MO_TEQ, a2);
+    case INDEX_op_qemu_st_i32:
+    case INDEX_op_qemu_st_i64:
+        tcg_out_qemu_st(s, a0, a1, a2, args[3]);
         break;
 
     case INDEX_op_bswap32_i64:
@@ -1770,20 +1751,10 @@ static const TCGTargetOpDef aarch64_op_defs[] = {
     { INDEX_op_movcond_i32, { "r", "r", "rwA", "rZ", "rZ" } },
     { INDEX_op_movcond_i64, { "r", "r", "rA", "rZ", "rZ" } },
 
-    { INDEX_op_qemu_ld8u, { "r", "l" } },
-    { INDEX_op_qemu_ld8s, { "r", "l" } },
-    { INDEX_op_qemu_ld16u, { "r", "l" } },
-    { INDEX_op_qemu_ld16s, { "r", "l" } },
-    { INDEX_op_qemu_ld32u, { "r", "l" } },
-    { INDEX_op_qemu_ld32s, { "r", "l" } },
-
-    { INDEX_op_qemu_ld32, { "r", "l" } },
-    { INDEX_op_qemu_ld64, { "r", "l" } },
-
-    { INDEX_op_qemu_st8, { "l", "l" } },
-    { INDEX_op_qemu_st16, { "l", "l" } },
-    { INDEX_op_qemu_st32, { "l", "l" } },
-    { INDEX_op_qemu_st64, { "l", "l" } },
+    { INDEX_op_qemu_ld_i32, { "r", "l" } },
+    { INDEX_op_qemu_ld_i64, { "r", "l" } },
+    { INDEX_op_qemu_st_i32, { "l", "l" } },
+    { INDEX_op_qemu_st_i64, { "l", "l" } },
 
     { INDEX_op_bswap16_i32, { "r", "r" } },
     { INDEX_op_bswap32_i32, { "r", "r" } },
diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h
index faccc36..adf0261 100644
--- a/tcg/aarch64/tcg-target.h
+++ b/tcg/aarch64/tcg-target.h
@@ -98,7 +98,7 @@ typedef enum {
 #define TCG_TARGET_HAS_muluh_i64        1
 #define TCG_TARGET_HAS_mulsh_i64        1
 
-#define TCG_TARGET_HAS_new_ldst         0
+#define TCG_TARGET_HAS_new_ldst         1
 
 static inline void flush_icache_range(uintptr_t start, uintptr_t stop)
 {
-- 
1.9.0

[Qemu-devel] [PATCH v4 20/25] tcg-aarch64: Support stores of zero

[Richard Henderson]

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 35 +++++++++++++++++++----------------
 1 file changed, 19 insertions(+), 16 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 34e477d..caaf8a2 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -1253,21 +1253,21 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGMemOp memop,
         tcg_out_ldst_r(s, LDST_8, LDST_ST, data_r, addr_r, off_r);
         break;
     case MO_16:
-        if (bswap) {
+        if (bswap && data_r != TCG_REG_XZR) {
             tcg_out_rev16(s, TCG_TYPE_I32, TCG_REG_TMP, data_r);
             data_r = TCG_REG_TMP;
         }
         tcg_out_ldst_r(s, LDST_16, LDST_ST, data_r, addr_r, off_r);
         break;
     case MO_32:
-        if (bswap) {
+        if (bswap && data_r != TCG_REG_XZR) {
             tcg_out_rev(s, TCG_TYPE_I32, TCG_REG_TMP, data_r);
             data_r = TCG_REG_TMP;
         }
         tcg_out_ldst_r(s, LDST_32, LDST_ST, data_r, addr_r, off_r);
         break;
     case MO_64:
-        if (bswap) {
+        if (bswap && data_r != TCG_REG_XZR) {
             tcg_out_rev(s, TCG_TYPE_I64, TCG_REG_TMP, data_r);
             data_r = TCG_REG_TMP;
         }
@@ -1364,8 +1364,6 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
 
     case INDEX_op_ld_i32:
     case INDEX_op_ld_i64:
-    case INDEX_op_st_i32:
-    case INDEX_op_st_i64:
     case INDEX_op_ld8u_i32:
     case INDEX_op_ld8s_i32:
     case INDEX_op_ld16u_i32:
@@ -1376,13 +1374,18 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
     case INDEX_op_ld16s_i64:
     case INDEX_op_ld32u_i64:
     case INDEX_op_ld32s_i64:
+        tcg_out_ldst(s, aarch64_ldst_get_data(opc), aarch64_ldst_get_type(opc),
+                     a0, a1, a2);
+        break;
+    case INDEX_op_st_i32:
+    case INDEX_op_st_i64:
     case INDEX_op_st8_i32:
     case INDEX_op_st8_i64:
     case INDEX_op_st16_i32:
     case INDEX_op_st16_i64:
     case INDEX_op_st32_i64:
         tcg_out_ldst(s, aarch64_ldst_get_data(opc), aarch64_ldst_get_type(opc),
-                     a0, a1, a2);
+                     REG0(0), a1, a2);
         break;
 
     case INDEX_op_add_i32:
@@ -1585,7 +1588,7 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
         break;
     case INDEX_op_qemu_st_i32:
     case INDEX_op_qemu_st_i64:
-        tcg_out_qemu_st(s, a0, a1, a2, args[3]);
+        tcg_out_qemu_st(s, REG0(0), a1, a2, args[3]);
         break;
 
     case INDEX_op_bswap32_i64:
@@ -1693,13 +1696,13 @@ static const TCGTargetOpDef aarch64_op_defs[] = {
     { INDEX_op_ld32s_i64, { "r", "r" } },
     { INDEX_op_ld_i64, { "r", "r" } },
 
-    { INDEX_op_st8_i32, { "r", "r" } },
-    { INDEX_op_st16_i32, { "r", "r" } },
-    { INDEX_op_st_i32, { "r", "r" } },
-    { INDEX_op_st8_i64, { "r", "r" } },
-    { INDEX_op_st16_i64, { "r", "r" } },
-    { INDEX_op_st32_i64, { "r", "r" } },
-    { INDEX_op_st_i64, { "r", "r" } },
+    { INDEX_op_st8_i32, { "rZ", "r" } },
+    { INDEX_op_st16_i32, { "rZ", "r" } },
+    { INDEX_op_st_i32, { "rZ", "r" } },
+    { INDEX_op_st8_i64, { "rZ", "r" } },
+    { INDEX_op_st16_i64, { "rZ", "r" } },
+    { INDEX_op_st32_i64, { "rZ", "r" } },
+    { INDEX_op_st_i64, { "rZ", "r" } },
 
     { INDEX_op_add_i32, { "r", "r", "rwA" } },
     { INDEX_op_add_i64, { "r", "r", "rA" } },
@@ -1753,8 +1756,8 @@ static const TCGTargetOpDef aarch64_op_defs[] = {
 
     { INDEX_op_qemu_ld_i32, { "r", "l" } },
     { INDEX_op_qemu_ld_i64, { "r", "l" } },
-    { INDEX_op_qemu_st_i32, { "l", "l" } },
-    { INDEX_op_qemu_st_i64, { "l", "l" } },
+    { INDEX_op_qemu_st_i32, { "lZ", "l" } },
+    { INDEX_op_qemu_st_i64, { "lZ", "l" } },
 
     { INDEX_op_bswap16_i32, { "r", "r" } },
     { INDEX_op_bswap32_i32, { "r", "r" } },
-- 
1.9.0

[Qemu-devel] [PATCH v4 21/25] tcg-aarch64: Introduce tcg_out_insn_3507

[Richard Henderson]

Cleaning up the implementation of REV and REV16 at the same time.

Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 57 ++++++++++++++++++++++++++++--------------------
 1 file changed, 33 insertions(+), 24 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index caaf8a2..0846835 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -327,6 +327,11 @@ typedef enum {
     I3506_CSEL      = 0x1a800000,
     I3506_CSINC     = 0x1a800400,
 
+    /* Data-processing (1 source) instructions.  */
+    I3507_REV16     = 0x5ac00400,
+    I3507_REV32     = 0x5ac00800,
+    I3507_REV64     = 0x5ac00c00,
+
     /* Data-processing (2 source) instructions.  */
     I3508_LSLV      = 0x1ac02000,
     I3508_LSRV      = 0x1ac02400,
@@ -545,6 +550,12 @@ static void tcg_out_insn_3506(TCGContext *s, AArch64Insn insn, TCGType ext,
               | tcg_cond_to_aarch64[c] << 12);
 }
 
+static void tcg_out_insn_3507(TCGContext *s, AArch64Insn insn, TCGType ext,
+                              TCGReg rd, TCGReg rn)
+{
+    tcg_out32(s, insn | ext << 31 | rn << 5 | rd);
+}
+
 static void tcg_out_insn_3509(TCGContext *s, AArch64Insn insn, TCGType ext,
                               TCGReg rd, TCGReg rn, TCGReg rm, TCGReg ra)
 {
@@ -960,20 +971,19 @@ static void tcg_out_brcond(TCGContext *s, TCGMemOp ext, TCGCond c, TCGArg a,
     }
 }
 
-static inline void tcg_out_rev(TCGContext *s, TCGType ext,
-                               TCGReg rd, TCGReg rm)
+static inline void tcg_out_rev64(TCGContext *s, TCGReg rd, TCGReg rn)
 {
-    /* using REV 0x5ac00800 */
-    unsigned int base = ext ? 0xdac00c00 : 0x5ac00800;
-    tcg_out32(s, base | rm << 5 | rd);
+    tcg_out_insn(s, 3507, REV64, TCG_TYPE_I64, rd, rn);
 }
 
-static inline void tcg_out_rev16(TCGContext *s, TCGType ext,
-                                 TCGReg rd, TCGReg rm)
+static inline void tcg_out_rev32(TCGContext *s, TCGReg rd, TCGReg rn)
 {
-    /* using REV16 0x5ac00400 */
-    unsigned int base = ext ? 0xdac00400 : 0x5ac00400;
-    tcg_out32(s, base | rm << 5 | rd);
+    tcg_out_insn(s, 3507, REV32, TCG_TYPE_I32, rd, rn);
+}
+
+static inline void tcg_out_rev16(TCGContext *s, TCGReg rd, TCGReg rn)
+{
+    tcg_out_insn(s, 3507, REV16, TCG_TYPE_I32, rd, rn);
 }
 
 static inline void tcg_out_sxt(TCGContext *s, TCGType ext, TCGMemOp s_bits,
@@ -1205,13 +1215,13 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGMemOp memop,
     case MO_UW:
         tcg_out_ldst_r(s, LDST_16, LDST_LD, data_r, addr_r, off_r);
         if (bswap) {
-            tcg_out_rev16(s, TCG_TYPE_I32, data_r, data_r);
+            tcg_out_rev16(s, data_r, data_r);
         }
         break;
     case MO_SW:
         if (bswap) {
             tcg_out_ldst_r(s, LDST_16, LDST_LD, data_r, addr_r, off_r);
-            tcg_out_rev16(s, TCG_TYPE_I32, data_r, data_r);
+            tcg_out_rev16(s, data_r, data_r);
             tcg_out_sxt(s, TCG_TYPE_I64, MO_16, data_r, data_r);
         } else {
             tcg_out_ldst_r(s, LDST_16, LDST_LD_S_X, data_r, addr_r, off_r);
@@ -1220,13 +1230,13 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGMemOp memop,
     case MO_UL:
         tcg_out_ldst_r(s, LDST_32, LDST_LD, data_r, addr_r, off_r);
         if (bswap) {
-            tcg_out_rev(s, TCG_TYPE_I32, data_r, data_r);
+            tcg_out_rev32(s, data_r, data_r);
         }
         break;
     case MO_SL:
         if (bswap) {
             tcg_out_ldst_r(s, LDST_32, LDST_LD, data_r, addr_r, off_r);
-            tcg_out_rev(s, TCG_TYPE_I32, data_r, data_r);
+            tcg_out_rev32(s, data_r, data_r);
             tcg_out_sxt(s, TCG_TYPE_I64, MO_32, data_r, data_r);
         } else {
             tcg_out_ldst_r(s, LDST_32, LDST_LD_S_X, data_r, addr_r, off_r);
@@ -1235,7 +1245,7 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGMemOp memop,
     case MO_Q:
         tcg_out_ldst_r(s, LDST_64, LDST_LD, data_r, addr_r, off_r);
         if (bswap) {
-            tcg_out_rev(s, TCG_TYPE_I64, data_r, data_r);
+            tcg_out_rev64(s, data_r, data_r);
         }
         break;
     default:
@@ -1254,21 +1264,21 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGMemOp memop,
         break;
     case MO_16:
         if (bswap && data_r != TCG_REG_XZR) {
-            tcg_out_rev16(s, TCG_TYPE_I32, TCG_REG_TMP, data_r);
+            tcg_out_rev16(s, TCG_REG_TMP, data_r);
             data_r = TCG_REG_TMP;
         }
         tcg_out_ldst_r(s, LDST_16, LDST_ST, data_r, addr_r, off_r);
         break;
     case MO_32:
         if (bswap && data_r != TCG_REG_XZR) {
-            tcg_out_rev(s, TCG_TYPE_I32, TCG_REG_TMP, data_r);
+            tcg_out_rev32(s, TCG_REG_TMP, data_r);
             data_r = TCG_REG_TMP;
         }
         tcg_out_ldst_r(s, LDST_32, LDST_ST, data_r, addr_r, off_r);
         break;
     case MO_64:
         if (bswap && data_r != TCG_REG_XZR) {
-            tcg_out_rev(s, TCG_TYPE_I64, TCG_REG_TMP, data_r);
+            tcg_out_rev64(s, TCG_REG_TMP, data_r);
             data_r = TCG_REG_TMP;
         }
         tcg_out_ldst_r(s, LDST_64, LDST_ST, data_r, addr_r, off_r);
@@ -1591,17 +1601,16 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
         tcg_out_qemu_st(s, REG0(0), a1, a2, args[3]);
         break;
 
-    case INDEX_op_bswap32_i64:
-        /* Despite the _i64, this is a 32-bit bswap.  */
-        ext = 0;
-        /* FALLTHRU */
     case INDEX_op_bswap64_i64:
+        tcg_out_rev64(s, a0, a1);
+        break;
+    case INDEX_op_bswap32_i64:
     case INDEX_op_bswap32_i32:
-        tcg_out_rev(s, ext, a0, a1);
+        tcg_out_rev32(s, a0, a1);
         break;
     case INDEX_op_bswap16_i64:
     case INDEX_op_bswap16_i32:
-        tcg_out_rev16(s, TCG_TYPE_I32, a0, a1);
+        tcg_out_rev16(s, a0, a1);
         break;
 
     case INDEX_op_ext8s_i64:
-- 
1.9.0

[Qemu-devel] [PATCH v4 22/25] tcg-aarch64: Merge aarch64_ldst_get_data/type into tcg_out_op

[Richard Henderson]

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 115 +++++++++++++----------------------------------
 1 file changed, 32 insertions(+), 83 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 0846835..7f72df5 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -356,78 +356,6 @@ typedef enum {
     I3510_ANDS      = 0x6a000000,
 } AArch64Insn;
 
-static inline enum aarch64_ldst_op_data
-aarch64_ldst_get_data(TCGOpcode tcg_op)
-{
-    switch (tcg_op) {
-    case INDEX_op_ld8u_i32:
-    case INDEX_op_ld8s_i32:
-    case INDEX_op_ld8u_i64:
-    case INDEX_op_ld8s_i64:
-    case INDEX_op_st8_i32:
-    case INDEX_op_st8_i64:
-        return LDST_8;
-
-    case INDEX_op_ld16u_i32:
-    case INDEX_op_ld16s_i32:
-    case INDEX_op_ld16u_i64:
-    case INDEX_op_ld16s_i64:
-    case INDEX_op_st16_i32:
-    case INDEX_op_st16_i64:
-        return LDST_16;
-
-    case INDEX_op_ld_i32:
-    case INDEX_op_st_i32:
-    case INDEX_op_ld32u_i64:
-    case INDEX_op_ld32s_i64:
-    case INDEX_op_st32_i64:
-        return LDST_32;
-
-    case INDEX_op_ld_i64:
-    case INDEX_op_st_i64:
-        return LDST_64;
-
-    default:
-        tcg_abort();
-    }
-}
-
-static inline enum aarch64_ldst_op_type
-aarch64_ldst_get_type(TCGOpcode tcg_op)
-{
-    switch (tcg_op) {
-    case INDEX_op_st8_i32:
-    case INDEX_op_st16_i32:
-    case INDEX_op_st8_i64:
-    case INDEX_op_st16_i64:
-    case INDEX_op_st_i32:
-    case INDEX_op_st32_i64:
-    case INDEX_op_st_i64:
-        return LDST_ST;
-
-    case INDEX_op_ld8u_i32:
-    case INDEX_op_ld16u_i32:
-    case INDEX_op_ld8u_i64:
-    case INDEX_op_ld16u_i64:
-    case INDEX_op_ld_i32:
-    case INDEX_op_ld32u_i64:
-    case INDEX_op_ld_i64:
-        return LDST_LD;
-
-    case INDEX_op_ld8s_i32:
-    case INDEX_op_ld16s_i32:
-        return LDST_LD_S_W;
-
-    case INDEX_op_ld8s_i64:
-    case INDEX_op_ld16s_i64:
-    case INDEX_op_ld32s_i64:
-        return LDST_LD_S_X;
-
-    default:
-        tcg_abort();
-    }
-}
-
 static inline uint32_t tcg_in32(TCGContext *s)
 {
     uint32_t v = *(uint32_t *)s->code_ptr;
@@ -1372,30 +1300,51 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
         tcg_out_goto_label(s, a0);
         break;
 
-    case INDEX_op_ld_i32:
-    case INDEX_op_ld_i64:
     case INDEX_op_ld8u_i32:
-    case INDEX_op_ld8s_i32:
-    case INDEX_op_ld16u_i32:
-    case INDEX_op_ld16s_i32:
     case INDEX_op_ld8u_i64:
+        tcg_out_ldst(s, LDST_8, LDST_LD, a0, a1, a2);
+        break;
+    case INDEX_op_ld8s_i32:
+        tcg_out_ldst(s, LDST_8, LDST_LD_S_W, a0, a1, a2);
+        break;
     case INDEX_op_ld8s_i64:
+        tcg_out_ldst(s, LDST_8, LDST_LD_S_X, a0, a1, a2);
+        break;
+    case INDEX_op_ld16u_i32:
     case INDEX_op_ld16u_i64:
+        tcg_out_ldst(s, LDST_16, LDST_LD, a0, a1, a2);
+        break;
+    case INDEX_op_ld16s_i32:
+        tcg_out_ldst(s, LDST_16, LDST_LD_S_W, a0, a1, a2);
+        break;
     case INDEX_op_ld16s_i64:
+        tcg_out_ldst(s, LDST_16, LDST_LD_S_X, a0, a1, a2);
+        break;
+    case INDEX_op_ld_i32:
     case INDEX_op_ld32u_i64:
+        tcg_out_ldst(s, LDST_32, LDST_LD, a0, a1, a2);
+        break;
     case INDEX_op_ld32s_i64:
-        tcg_out_ldst(s, aarch64_ldst_get_data(opc), aarch64_ldst_get_type(opc),
-                     a0, a1, a2);
+        tcg_out_ldst(s, LDST_32, LDST_LD_S_X, a0, a1, a2);
         break;
-    case INDEX_op_st_i32:
-    case INDEX_op_st_i64:
+    case INDEX_op_ld_i64:
+        tcg_out_ldst(s, LDST_64, LDST_LD, a0, a1, a2);
+        break;
+
     case INDEX_op_st8_i32:
     case INDEX_op_st8_i64:
+        tcg_out_ldst(s, LDST_8, LDST_ST, REG0(0), a1, a2);
+        break;
     case INDEX_op_st16_i32:
     case INDEX_op_st16_i64:
+        tcg_out_ldst(s, LDST_16, LDST_ST, REG0(0), a1, a2);
+        break;
+    case INDEX_op_st_i32:
     case INDEX_op_st32_i64:
-        tcg_out_ldst(s, aarch64_ldst_get_data(opc), aarch64_ldst_get_type(opc),
-                     REG0(0), a1, a2);
+        tcg_out_ldst(s, LDST_32, LDST_ST, REG0(0), a1, a2);
+        break;
+    case INDEX_op_st_i64:
+        tcg_out_ldst(s, LDST_64, LDST_ST, REG0(0), a1, a2);
         break;
 
     case INDEX_op_add_i32:
-- 
1.9.0

[Qemu-devel] [PATCH v4 23/25] tcg-aarch64: Introduce tcg_out_insn_3312, _3310, _3313

[Richard Henderson]

Replace aarch64_ldst_op_data with AArch64LdstType, as it wasn't encoded
for the proper shift for the field and was confusing.

Merge aarch64_ldst_op_data, AArch64LdstType, and a few stray opcode bits
into a single I3312_* argument, eliminating some magic numbers from the
helper functions.

Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 176 ++++++++++++++++++++++++-----------------------
 1 file changed, 89 insertions(+), 87 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 7f72df5..3235824 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -242,19 +242,12 @@ static const enum aarch64_cond_code tcg_cond_to_aarch64[] = {
     [TCG_COND_LEU] = COND_LS,
 };
 
-/* opcodes for LDR / STR instructions with base + simm9 addressing */
-enum aarch64_ldst_op_data { /* size of the data moved */
-    LDST_8 = 0x38,
-    LDST_16 = 0x78,
-    LDST_32 = 0xb8,
-    LDST_64 = 0xf8,
-};
-enum aarch64_ldst_op_type { /* type of operation */
-    LDST_ST = 0x0,    /* store */
-    LDST_LD = 0x4,    /* load */
-    LDST_LD_S_X = 0x8,  /* load and sign-extend into Xt */
-    LDST_LD_S_W = 0xc,  /* load and sign-extend into Wt */
-};
+typedef enum {
+    LDST_ST = 0,    /* store */
+    LDST_LD = 1,    /* load */
+    LDST_LD_S_X = 2,  /* load and sign-extend into Xt */
+    LDST_LD_S_W = 3,  /* load and sign-extend into Wt */
+} AArch64LdstType;
 
 /* We encode the format of the insn into the beginning of the name, so that
    we can have the preprocessor help "typecheck" the insn vs the output
@@ -278,6 +271,28 @@ typedef enum {
     I3207_BLR       = 0xd63f0000,
     I3207_RET       = 0xd65f0000,
 
+    /* Load/store register.  Described here as 3.3.12, but the helper
+       that emits them can transform to 3.3.10 or 3.3.13.  */
+    I3312_STRB      = 0x38000000 | LDST_ST << 22 | MO_8 << 30,
+    I3312_STRH      = 0x38000000 | LDST_ST << 22 | MO_16 << 30,
+    I3312_STRW      = 0x38000000 | LDST_ST << 22 | MO_32 << 30,
+    I3312_STRX      = 0x38000000 | LDST_ST << 22 | MO_64 << 30,
+
+    I3312_LDRB      = 0x38000000 | LDST_LD << 22 | MO_8 << 30,
+    I3312_LDRH      = 0x38000000 | LDST_LD << 22 | MO_16 << 30,
+    I3312_LDRW      = 0x38000000 | LDST_LD << 22 | MO_32 << 30,
+    I3312_LDRX      = 0x38000000 | LDST_LD << 22 | MO_64 << 30,
+
+    I3312_LDRSBW    = 0x38000000 | LDST_LD_S_W << 22 | MO_8 << 30,
+    I3312_LDRSHW    = 0x38000000 | LDST_LD_S_W << 22 | MO_16 << 30,
+
+    I3312_LDRSBX    = 0x38000000 | LDST_LD_S_X << 22 | MO_8 << 30,
+    I3312_LDRSHX    = 0x38000000 | LDST_LD_S_X << 22 | MO_16 << 30,
+    I3312_LDRSWX    = 0x38000000 | LDST_LD_S_X << 22 | MO_32 << 30,
+
+    I3312_TO_I3310  = 0x00206800,
+    I3312_TO_I3313  = 0x01000000,
+
     /* Load/store register pair instructions.  */
     I3314_LDP       = 0x28400000,
     I3314_STP       = 0x28000000,
@@ -490,26 +505,25 @@ static void tcg_out_insn_3509(TCGContext *s, AArch64Insn insn, TCGType ext,
     tcg_out32(s, insn | ext << 31 | rm << 16 | ra << 10 | rn << 5 | rd);
 }
 
+static void tcg_out_insn_3310(TCGContext *s, AArch64Insn insn,
+                              TCGReg rd, TCGReg base, TCGReg regoff)
+{
+    /* Note the AArch64Insn constants above are for C3.3.12.  Adjust.  */
+    tcg_out32(s, insn | I3312_TO_I3310 | regoff << 16 | base << 5 | rd);
+}
 
-static inline void tcg_out_ldst_9(TCGContext *s,
-                                  enum aarch64_ldst_op_data op_data,
-                                  enum aarch64_ldst_op_type op_type,
-                                  TCGReg rd, TCGReg rn, intptr_t offset)
+
+static void tcg_out_insn_3312(TCGContext *s, AArch64Insn insn,
+                              TCGReg rd, TCGReg rn, intptr_t offset)
 {
-    /* use LDUR with BASE register with 9bit signed unscaled offset */
-    tcg_out32(s, op_data << 24 | op_type << 20
-              | (offset & 0x1ff) << 12 | rn << 5 | rd);
+    tcg_out32(s, insn | (offset & 0x1ff) << 12 | rn << 5 | rd);
 }
 
-/* tcg_out_ldst_12 expects a scaled unsigned immediate offset */
-static inline void tcg_out_ldst_12(TCGContext *s,
-                                   enum aarch64_ldst_op_data op_data,
-                                   enum aarch64_ldst_op_type op_type,
-                                   TCGReg rd, TCGReg rn,
-                                   tcg_target_ulong scaled_uimm)
+static void tcg_out_insn_3313(TCGContext *s, AArch64Insn insn,
+                              TCGReg rd, TCGReg rn, uintptr_t scaled_uimm)
 {
-    tcg_out32(s, (op_data | 1) << 24
-              | op_type << 20 | scaled_uimm << 10 | rn << 5 | rd);
+    /* Note the AArch64Insn constants above are for C3.3.12.  Adjust.  */
+    tcg_out32(s, insn | I3312_TO_I3313 | scaled_uimm << 10 | rn << 5 | rd);
 }
 
 /* Register to register move using ORR (shifted register with no shift). */
@@ -647,44 +661,32 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg rd,
     }
 }
 
-static inline void tcg_out_ldst_r(TCGContext *s,
-                                  enum aarch64_ldst_op_data op_data,
-                                  enum aarch64_ldst_op_type op_type,
-                                  TCGReg rd, TCGReg base, TCGReg regoff)
-{
-    /* load from memory to register using base + 64bit register offset */
-    /* using f.e. STR Wt, [Xn, Xm] 0xb8600800|(regoff << 16)|(base << 5)|rd */
-    /* the 0x6000 is for the "no extend field" */
-    tcg_out32(s, 0x00206800
-              | op_data << 24 | op_type << 20 | regoff << 16 | base << 5 | rd);
-}
+/* Define something more legible for general use.  */
+#define tcg_out_ldst_r  tcg_out_insn_3310
 
-/* solve the whole ldst problem */
-static inline void tcg_out_ldst(TCGContext *s, enum aarch64_ldst_op_data data,
-                                enum aarch64_ldst_op_type type,
-                                TCGReg rd, TCGReg rn, intptr_t offset)
+static void tcg_out_ldst(TCGContext *s, AArch64Insn insn,
+                         TCGReg rd, TCGReg rn, intptr_t offset)
 {
+    TCGMemOp size = (uint32_t)insn >> 30;
+
     if (offset >= -256 && offset < 256) {
-        tcg_out_ldst_9(s, data, type, rd, rn, offset);
+        tcg_out_insn_3312(s, insn, rd, rn, offset);
         return;
     }
 
-    if (offset >= 256) {
-        /* if the offset is naturally aligned and in range,
-           then we can use the scaled uimm12 encoding */
-        unsigned int s_bits = data >> 6;
-        if (!(offset & ((1 << s_bits) - 1))) {
-            tcg_target_ulong scaled_uimm = offset >> s_bits;
-            if (scaled_uimm <= 0xfff) {
-                tcg_out_ldst_12(s, data, type, rd, rn, scaled_uimm);
-                return;
-            }
+    /* If the offset is naturally aligned and in range, then we can
+       use the scaled uimm12 encoding */
+    if (offset >= 0 && !(offset & ((1 << size) - 1))) {
+        uintptr_t scaled_uimm = offset >> size;
+        if (scaled_uimm <= 0xfff) {
+            tcg_out_insn_3313(s, insn, rd, rn, scaled_uimm);
+            return;
         }
     }
 
-    /* worst-case scenario, move offset to temp register, use reg offset */
+    /* Worst-case scenario, move offset to temp register, use reg offset.  */
     tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP, offset);
-    tcg_out_ldst_r(s, data, type, rd, rn, TCG_REG_TMP);
+    tcg_out_ldst_r(s, insn, rd, rn, TCG_REG_TMP);
 }
 
 static inline void tcg_out_mov(TCGContext *s,
@@ -698,14 +700,14 @@ static inline void tcg_out_mov(TCGContext *s,
 static inline void tcg_out_ld(TCGContext *s, TCGType type, TCGReg arg,
                               TCGReg arg1, intptr_t arg2)
 {
-    tcg_out_ldst(s, (type == TCG_TYPE_I64) ? LDST_64 : LDST_32, LDST_LD,
+    tcg_out_ldst(s, type == TCG_TYPE_I32 ? I3312_LDRW : I3312_LDRX,
                  arg, arg1, arg2);
 }
 
 static inline void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg,
                               TCGReg arg1, intptr_t arg2)
 {
-    tcg_out_ldst(s, (type == TCG_TYPE_I64) ? LDST_64 : LDST_32, LDST_ST,
+    tcg_out_ldst(s, type == TCG_TYPE_I32 ? I3312_STRW : I3312_STRX,
                  arg, arg1, arg2);
 }
 
@@ -1108,12 +1110,12 @@ static void tcg_out_tlb_read(TCGContext *s, TCGReg addr_reg, TCGMemOp s_bits,
 
     /* Merge "low bits" from tlb offset, load the tlb comparator into X0.
        X0 = load [X2 + (tlb_offset & 0x000fff)] */
-    tcg_out_ldst(s, TARGET_LONG_BITS == 64 ? LDST_64 : LDST_32,
-                 LDST_LD, TCG_REG_X0, TCG_REG_X2, tlb_offset & 0xfff);
+    tcg_out_ldst(s, TARGET_LONG_BITS == 32 ? I3312_LDRW : I3312_LDRX,
+                 TCG_REG_X0, TCG_REG_X2, tlb_offset & 0xfff);
 
     /* Load the tlb addend. Do that early to avoid stalling.
        X1 = load [X2 + (tlb_offset & 0xfff) + offsetof(addend)] */
-    tcg_out_ldst(s, LDST_64, LDST_LD, TCG_REG_X1, TCG_REG_X2,
+    tcg_out_ldst(s, I3312_LDRX, TCG_REG_X1, TCG_REG_X2,
                  (tlb_offset & 0xfff) + (offsetof(CPUTLBEntry, addend)) -
                  (is_read ? offsetof(CPUTLBEntry, addr_read)
                   : offsetof(CPUTLBEntry, addr_write)));
@@ -1135,43 +1137,43 @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGMemOp memop,
 
     switch (memop & MO_SSIZE) {
     case MO_UB:
-        tcg_out_ldst_r(s, LDST_8, LDST_LD, data_r, addr_r, off_r);
+        tcg_out_ldst_r(s, I3312_LDRB, data_r, addr_r, off_r);
         break;
     case MO_SB:
-        tcg_out_ldst_r(s, LDST_8, LDST_LD_S_X, data_r, addr_r, off_r);
+        tcg_out_ldst_r(s, I3312_LDRSBX, data_r, addr_r, off_r);
         break;
     case MO_UW:
-        tcg_out_ldst_r(s, LDST_16, LDST_LD, data_r, addr_r, off_r);
+        tcg_out_ldst_r(s, I3312_LDRH, data_r, addr_r, off_r);
         if (bswap) {
             tcg_out_rev16(s, data_r, data_r);
         }
         break;
     case MO_SW:
         if (bswap) {
-            tcg_out_ldst_r(s, LDST_16, LDST_LD, data_r, addr_r, off_r);
+            tcg_out_ldst_r(s, I3312_LDRH, data_r, addr_r, off_r);
             tcg_out_rev16(s, data_r, data_r);
             tcg_out_sxt(s, TCG_TYPE_I64, MO_16, data_r, data_r);
         } else {
-            tcg_out_ldst_r(s, LDST_16, LDST_LD_S_X, data_r, addr_r, off_r);
+            tcg_out_ldst_r(s, I3312_LDRSHX, data_r, addr_r, off_r);
         }
         break;
     case MO_UL:
-        tcg_out_ldst_r(s, LDST_32, LDST_LD, data_r, addr_r, off_r);
+        tcg_out_ldst_r(s, I3312_LDRW, data_r, addr_r, off_r);
         if (bswap) {
             tcg_out_rev32(s, data_r, data_r);
         }
         break;
     case MO_SL:
         if (bswap) {
-            tcg_out_ldst_r(s, LDST_32, LDST_LD, data_r, addr_r, off_r);
+            tcg_out_ldst_r(s, I3312_LDRW, data_r, addr_r, off_r);
             tcg_out_rev32(s, data_r, data_r);
             tcg_out_sxt(s, TCG_TYPE_I64, MO_32, data_r, data_r);
         } else {
-            tcg_out_ldst_r(s, LDST_32, LDST_LD_S_X, data_r, addr_r, off_r);
+            tcg_out_ldst_r(s, I3312_LDRSWX, data_r, addr_r, off_r);
         }
         break;
     case MO_Q:
-        tcg_out_ldst_r(s, LDST_64, LDST_LD, data_r, addr_r, off_r);
+        tcg_out_ldst_r(s, I3312_LDRX, data_r, addr_r, off_r);
         if (bswap) {
             tcg_out_rev64(s, data_r, data_r);
         }
@@ -1188,28 +1190,28 @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGMemOp memop,
 
     switch (memop & MO_SIZE) {
     case MO_8:
-        tcg_out_ldst_r(s, LDST_8, LDST_ST, data_r, addr_r, off_r);
+        tcg_out_ldst_r(s, I3312_STRB, data_r, addr_r, off_r);
         break;
     case MO_16:
         if (bswap && data_r != TCG_REG_XZR) {
             tcg_out_rev16(s, TCG_REG_TMP, data_r);
             data_r = TCG_REG_TMP;
         }
-        tcg_out_ldst_r(s, LDST_16, LDST_ST, data_r, addr_r, off_r);
+        tcg_out_ldst_r(s, I3312_STRH, data_r, addr_r, off_r);
         break;
     case MO_32:
         if (bswap && data_r != TCG_REG_XZR) {
             tcg_out_rev32(s, TCG_REG_TMP, data_r);
             data_r = TCG_REG_TMP;
         }
-        tcg_out_ldst_r(s, LDST_32, LDST_ST, data_r, addr_r, off_r);
+        tcg_out_ldst_r(s, I3312_STRW, data_r, addr_r, off_r);
         break;
     case MO_64:
         if (bswap && data_r != TCG_REG_XZR) {
             tcg_out_rev64(s, TCG_REG_TMP, data_r);
             data_r = TCG_REG_TMP;
         }
-        tcg_out_ldst_r(s, LDST_64, LDST_ST, data_r, addr_r, off_r);
+        tcg_out_ldst_r(s, I3312_STRX, data_r, addr_r, off_r);
         break;
     default:
         tcg_abort();
@@ -1302,49 +1304,49 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
 
     case INDEX_op_ld8u_i32:
     case INDEX_op_ld8u_i64:
-        tcg_out_ldst(s, LDST_8, LDST_LD, a0, a1, a2);
+        tcg_out_ldst(s, I3312_LDRB, a0, a1, a2);
         break;
     case INDEX_op_ld8s_i32:
-        tcg_out_ldst(s, LDST_8, LDST_LD_S_W, a0, a1, a2);
+        tcg_out_ldst(s, I3312_LDRSBW, a0, a1, a2);
         break;
     case INDEX_op_ld8s_i64:
-        tcg_out_ldst(s, LDST_8, LDST_LD_S_X, a0, a1, a2);
+        tcg_out_ldst(s, I3312_LDRSBX, a0, a1, a2);
         break;
     case INDEX_op_ld16u_i32:
     case INDEX_op_ld16u_i64:
-        tcg_out_ldst(s, LDST_16, LDST_LD, a0, a1, a2);
+        tcg_out_ldst(s, I3312_LDRH, a0, a1, a2);
         break;
     case INDEX_op_ld16s_i32:
-        tcg_out_ldst(s, LDST_16, LDST_LD_S_W, a0, a1, a2);
+        tcg_out_ldst(s, I3312_LDRSHW, a0, a1, a2);
         break;
     case INDEX_op_ld16s_i64:
-        tcg_out_ldst(s, LDST_16, LDST_LD_S_X, a0, a1, a2);
+        tcg_out_ldst(s, I3312_LDRSHX, a0, a1, a2);
         break;
     case INDEX_op_ld_i32:
     case INDEX_op_ld32u_i64:
-        tcg_out_ldst(s, LDST_32, LDST_LD, a0, a1, a2);
+        tcg_out_ldst(s, I3312_LDRW, a0, a1, a2);
         break;
     case INDEX_op_ld32s_i64:
-        tcg_out_ldst(s, LDST_32, LDST_LD_S_X, a0, a1, a2);
+        tcg_out_ldst(s, I3312_LDRSWX, a0, a1, a2);
         break;
     case INDEX_op_ld_i64:
-        tcg_out_ldst(s, LDST_64, LDST_LD, a0, a1, a2);
+        tcg_out_ldst(s, I3312_LDRX, a0, a1, a2);
         break;
 
     case INDEX_op_st8_i32:
     case INDEX_op_st8_i64:
-        tcg_out_ldst(s, LDST_8, LDST_ST, REG0(0), a1, a2);
+        tcg_out_ldst(s, I3312_STRB, REG0(0), a1, a2);
         break;
     case INDEX_op_st16_i32:
     case INDEX_op_st16_i64:
-        tcg_out_ldst(s, LDST_16, LDST_ST, REG0(0), a1, a2);
+        tcg_out_ldst(s, I3312_STRH, REG0(0), a1, a2);
         break;
     case INDEX_op_st_i32:
     case INDEX_op_st32_i64:
-        tcg_out_ldst(s, LDST_32, LDST_ST, REG0(0), a1, a2);
+        tcg_out_ldst(s, I3312_STRW, REG0(0), a1, a2);
         break;
     case INDEX_op_st_i64:
-        tcg_out_ldst(s, LDST_64, LDST_ST, REG0(0), a1, a2);
+        tcg_out_ldst(s, I3312_STRX, REG0(0), a1, a2);
         break;
 
     case INDEX_op_add_i32:
-- 
1.9.0

[Qemu-devel] [PATCH v4 24/25] tcg-aarch64: Prefer unsigned offsets before signed offsets for ldst

[Richard Henderson]

The assembler seems to prefer them, perhaps we should too.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 3235824..7ff4be7 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -669,11 +669,6 @@ static void tcg_out_ldst(TCGContext *s, AArch64Insn insn,
 {
     TCGMemOp size = (uint32_t)insn >> 30;
 
-    if (offset >= -256 && offset < 256) {
-        tcg_out_insn_3312(s, insn, rd, rn, offset);
-        return;
-    }
-
     /* If the offset is naturally aligned and in range, then we can
        use the scaled uimm12 encoding */
     if (offset >= 0 && !(offset & ((1 << size) - 1))) {
@@ -684,6 +679,12 @@ static void tcg_out_ldst(TCGContext *s, AArch64Insn insn,
         }
     }
 
+    /* Small signed offsets can use the unscaled encoding.  */
+    if (offset >= -256 && offset < 256) {
+        tcg_out_insn_3312(s, insn, rd, rn, offset);
+        return;
+    }
+
     /* Worst-case scenario, move offset to temp register, use reg offset.  */
     tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP, offset);
     tcg_out_ldst_r(s, insn, rd, rn, TCG_REG_TMP);
-- 
1.9.0

[Qemu-devel] [PATCH v4 25/25] tcg-aarch64: Use tcg_out_mov in preference to tcg_out_movr

[Richard Henderson]

It's the more canonical interface.

Reviewed-by: Claudio Fontana <claudio.fontana@huawei.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/aarch64/tcg-target.c | 16 +++++++---------
 1 file changed, 7 insertions(+), 9 deletions(-)

diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 7ff4be7..73ed658 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -979,9 +979,7 @@ static inline void tcg_out_addsub2(TCGContext *s, int ext, TCGReg rl,
     }
     tcg_out_insn_3503(s, insn, ext, rh, ah, bh);
 
-    if (rl != orig_rl) {
-        tcg_out_movr(s, ext, orig_rl, rl);
-    }
+    tcg_out_mov(s, ext, orig_rl, rl);
 }
 
 #ifdef CONFIG_SOFTMMU
@@ -1025,15 +1023,15 @@ static void tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
 
     reloc_pc19(lb->label_ptr[0], (intptr_t)s->code_ptr);
 
-    tcg_out_movr(s, TCG_TYPE_I64, TCG_REG_X0, TCG_AREG0);
-    tcg_out_movr(s, TARGET_LONG_BITS == 64, TCG_REG_X1, lb->addrlo_reg);
+    tcg_out_mov(s, TCG_TYPE_I64, TCG_REG_X0, TCG_AREG0);
+    tcg_out_mov(s, TARGET_LONG_BITS == 64, TCG_REG_X1, lb->addrlo_reg);
     tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_X2, lb->mem_index);
     tcg_out_adr(s, TCG_REG_X3, (intptr_t)lb->raddr);
     tcg_out_call(s, (intptr_t)qemu_ld_helpers[opc & ~MO_SIGN]);
     if (opc & MO_SIGN) {
         tcg_out_sxt(s, TCG_TYPE_I64, size, lb->datalo_reg, TCG_REG_X0);
     } else {
-        tcg_out_movr(s, TCG_TYPE_I64, lb->datalo_reg, TCG_REG_X0);
+        tcg_out_mov(s, size == MO_64, lb->datalo_reg, TCG_REG_X0);
     }
 
     tcg_out_goto(s, (intptr_t)lb->raddr);
@@ -1046,9 +1044,9 @@ static void tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb)
 
     reloc_pc19(lb->label_ptr[0], (intptr_t)s->code_ptr);
 
-    tcg_out_movr(s, TCG_TYPE_I64, TCG_REG_X0, TCG_AREG0);
-    tcg_out_movr(s, TARGET_LONG_BITS == 64, TCG_REG_X1, lb->addrlo_reg);
-    tcg_out_movr(s, size == MO_64, TCG_REG_X2, lb->datalo_reg);
+    tcg_out_mov(s, TCG_TYPE_I64, TCG_REG_X0, TCG_AREG0);
+    tcg_out_mov(s, TARGET_LONG_BITS == 64, TCG_REG_X1, lb->addrlo_reg);
+    tcg_out_mov(s, size == MO_64, TCG_REG_X2, lb->datalo_reg);
     tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_X3, lb->mem_index);
     tcg_out_adr(s, TCG_REG_X4, (intptr_t)lb->raddr);
     tcg_out_call(s, (intptr_t)qemu_st_helpers[opc]);
-- 
1.9.0