[Qemu-devel] [PATCH 16/17] tcg/s390: Use softmmu fast path for unaligned accesses

[Richard Henderson <rth@twiddle.net>]

Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/s390/tcg-target.c | 26 +++++++++++++++++++++-----
 1 file changed, 21 insertions(+), 5 deletions(-)

diff --git a/tcg/s390/tcg-target.c b/tcg/s390/tcg-target.c
index 96c3d65..be51c8b 100644
--- a/tcg/s390/tcg-target.c
+++ b/tcg/s390/tcg-target.c
@@ -1504,20 +1504,36 @@ QEMU_BUILD_BUG_ON(offsetof(CPUArchState, tlb_table[NB_MMU_MODES - 1][1])
 static TCGReg tcg_out_tlb_read(TCGContext* s, TCGReg addr_reg, TCGMemOp opc,
                                int mem_index, bool is_ld)
 {
-    TCGMemOp s_bits = opc & MO_SIZE;
-    uint64_t tlb_mask = TARGET_PAGE_MASK | ((1 << s_bits) - 1);
-    int ofs;
+    int s_mask = (1 << (opc & MO_SIZE)) - 1;
+    int ofs, a_off;
+    uint64_t tlb_mask;
+
+    /* For aligned accesses, we check the first byte and include the alignment
+       bits within the address.  For unaligned access, we check that we don't
+       cross pages using the address of the last byte of the access.  */
+    if ((opc & MO_AMASK) == MO_ALIGN || s_mask == 0) {
+        a_off = 0;
+        tlb_mask = TARGET_PAGE_MASK | s_mask;
+    } else {
+        a_off = s_mask;
+        tlb_mask = TARGET_PAGE_MASK;
+    }
 
     if (facilities & FACILITY_GEN_INST_EXT) {
         tcg_out_risbg(s, TCG_REG_R2, addr_reg,
                       64 - CPU_TLB_BITS - CPU_TLB_ENTRY_BITS,
                       63 - CPU_TLB_ENTRY_BITS,
                       64 + CPU_TLB_ENTRY_BITS - TARGET_PAGE_BITS, 1);
-        tgen_andi_risbg(s, TCG_REG_R3, addr_reg, tlb_mask);
+        if (a_off) {
+            tcg_out_insn(s, RX, LA, TCG_REG_R3, addr_reg, TCG_REG_NONE, a_off);
+            tgen_andi(s, TCG_TYPE_TL, TCG_REG_R3, tlb_mask);
+        } else {
+            tgen_andi_risbg(s, TCG_REG_R3, addr_reg, tlb_mask);
+        }
     } else {
         tcg_out_sh64(s, RSY_SRLG, TCG_REG_R2, addr_reg, TCG_REG_NONE,
                      TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS);
-        tcg_out_movi(s, TCG_TYPE_TL, TCG_REG_R3, addr_reg);
+        tcg_out_insn(s, RX, LA, TCG_REG_R3, addr_reg, TCG_REG_NONE, a_off);
         tgen_andi(s, TCG_TYPE_I64, TCG_REG_R2,
                   (CPU_TLB_SIZE - 1) << CPU_TLB_ENTRY_BITS);
         tgen_andi(s, TCG_TYPE_TL, TCG_REG_R3, tlb_mask);
-- 
2.4.3