When testing the sign bit or equality to zero of a partial register, it
is useful to use a single TSTEQ or TSTNE operation.  It can also be used
to test the parity flag, using bit 0 of the population count.

Do not do this for target_ulong-sized values however; the optimizer would
produce a comparison against zero anyway, and it avoids shifts by 64
which are undefined behavior.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 target/i386/tcg/translate.c | 28 ++++++++++++++++++++--------
 target/i386/tcg/emit.c.inc  |  5 ++---
 2 files changed, 22 insertions(+), 11 deletions(-)

diff --git a/target/i386/tcg/translate.c b/target/i386/tcg/translate.c
index 051ffb5e1fd..4735f084d40 100644
--- a/target/i386/tcg/translate.c
+++ b/target/i386/tcg/translate.c
@@ -928,11 +928,21 @@ typedef struct CCPrepare {
     bool no_setcond;
 } CCPrepare;
 
+static CCPrepare gen_prepare_sign_nz(TCGv src, MemOp size)
+{
+    if (size == MO_TL) {
+        return (CCPrepare) { .cond = TCG_COND_LT, .reg = src, .mask = -1 };
+    } else {
+        return (CCPrepare) { .cond = TCG_COND_TSTNE, .reg = src, .mask = -1,
+                             .imm = 1ull << ((8 << size) - 1) };
+    }
+}
+
 /* compute eflags.C to reg */
 static CCPrepare gen_prepare_eflags_c(DisasContext *s, TCGv reg)
 {
     TCGv t0, t1;
-    int size, shift;
+    MemOp size;
 
     switch (s->cc_op) {
     case CC_OP_SUBB ... CC_OP_SUBQ:
@@ -967,9 +977,7 @@ static CCPrepare gen_prepare_eflags_c(DisasContext *s, TCGv reg)
     case CC_OP_SHLB ... CC_OP_SHLQ:
         /* (CC_SRC >> (DATA_BITS - 1)) & 1 */
         size = s->cc_op - CC_OP_SHLB;
-        shift = (8 << size) - 1;
-        return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
-                             .mask = (target_ulong)1 << shift };
+        return gen_prepare_sign_nz(cpu_cc_src, size);
 
     case CC_OP_MULB ... CC_OP_MULQ:
         return (CCPrepare) { .cond = TCG_COND_NE,
@@ -1029,8 +1037,7 @@ static CCPrepare gen_prepare_eflags_s(DisasContext *s, TCGv reg)
     default:
         {
             MemOp size = (s->cc_op - CC_OP_ADDB) & 3;
-            TCGv t0 = gen_ext_tl(reg, cpu_cc_dst, size, true);
-            return (CCPrepare) { .cond = TCG_COND_LT, .reg = t0, .mask = -1 };
+            return gen_prepare_sign_nz(cpu_cc_dst, size);
         }
     }
 }
@@ -1077,8 +1084,13 @@ static CCPrepare gen_prepare_eflags_z(DisasContext *s, TCGv reg)
     default:
         {
             MemOp size = (s->cc_op - CC_OP_ADDB) & 3;
-            TCGv t0 = gen_ext_tl(reg, cpu_cc_dst, size, false);
-            return (CCPrepare) { .cond = TCG_COND_EQ, .reg = t0, .mask = -1 };
+            if (size == MO_TL) {
+                return (CCPrepare) { .cond = TCG_COND_EQ, .reg = cpu_cc_dst,
+                                     .mask = -1 };
+            } else {
+                return (CCPrepare) { .cond = TCG_COND_TSTEQ, .reg = cpu_cc_dst,
+                                     .mask = -1, .imm = (1ull << (8 << size)) - 1 };
+            }
         }
     }
 }
diff --git a/target/i386/tcg/emit.c.inc b/target/i386/tcg/emit.c.inc
index 6bcf88ecd71..0e00f6635dd 100644
--- a/target/i386/tcg/emit.c.inc
+++ b/target/i386/tcg/emit.c.inc
@@ -1209,7 +1209,7 @@ static void gen_CMPccXADD(DisasContext *s, CPUX86State *env, X86DecodedInsn *dec
         [JCC_Z] = TCG_COND_EQ,
         [JCC_BE] = TCG_COND_LEU,
         [JCC_S] = TCG_COND_LT,  /* test sign bit by comparing against 0 */
-        [JCC_P] = TCG_COND_EQ,  /* even parity - tests low bit of popcount */
+        [JCC_P] = TCG_COND_TSTEQ,  /* even parity - tests low bit of popcount */
         [JCC_L] = TCG_COND_LT,
         [JCC_LE] = TCG_COND_LE,
     };
@@ -1260,8 +1260,7 @@ static void gen_CMPccXADD(DisasContext *s, CPUX86State *env, X86DecodedInsn *dec
     case JCC_P:
         tcg_gen_ext8u_tl(s->tmp0, s->T0);
         tcg_gen_ctpop_tl(s->tmp0, s->tmp0);
-        tcg_gen_andi_tl(s->tmp0, s->tmp0, 1);
-        cmp_lhs = s->tmp0, cmp_rhs = tcg_constant_tl(0);
+        cmp_lhs = s->tmp0, cmp_rhs = tcg_constant_tl(1);
         break;
 
     case JCC_S:
-- 
2.45.0