[PULL 17/38] target/i386: improve code generation for BT

[Paolo Bonzini]

Because BT does not write back to the source operand, it can modify it to
ensure that one of the operands of TSTNE is a constant (after either gen_BT
or the optimizer's constant propagation).  This produces better and more
optimizable TCG ops.  For example, the sequence

  movl $0x60013f, %ebx
  btl %ecx, %ebx

becomes just

  and_i32 tmp1,ecx,$0x1f                   dead: 1 2  pref=0xffff
  shr_i32 tmp0,$0x60013f,tmp1              dead: 1 2  pref=0xffff
  and_i32 tmp16,tmp0,$0x1                  dead: 1  pref=0xbf80

On s390x, it can use four instructions to isolate bit 0 of 0x60013f >> (ecx & 
31):

  nilf     %r12, 0x1f
  lgfi     %r11, 0x60013f
  srlk     %r12, %r11, 0(%r12)
  nilf     %r12, 1

Previously, it used five instructions to build 1 << (ecx & 31) and compute
TSTEQ, and also needed two more to construct the result of setcond:

  nilf     %r12, 0x1f
  lghi     %r11, 1
  sllk     %r12, %r11, 0(%r12)
  lgfi     %r9, 0x60013f
  nrk      %r0, %r12, %r9
  lghi     %r12, 0
  locghilh %r12, 1

Signed-off-by: Paolo Bonzini <pbonz...@redhat.com>
---
 target/i386/tcg/emit.c.inc | 36 ++++++++++++++++++++++++++++--------
 1 file changed, 28 insertions(+), 8 deletions(-)

diff --git a/target/i386/tcg/emit.c.inc b/target/i386/tcg/emit.c.inc
index 785ff63f2ac..5c115429350 100644
--- a/target/i386/tcg/emit.c.inc
+++ b/target/i386/tcg/emit.c.inc
@@ -1443,8 +1443,9 @@ static TCGv gen_bt_mask(DisasContext *s, X86DecodedInsn 
*decode)
     return mask;
 }
 
-/* Expects truncated bit index in s->T1, 1 << s->T1 in MASK.  */
-static void gen_bt_flags(DisasContext *s, X86DecodedInsn *decode, TCGv src, 
TCGv mask)
+/* Expects truncated bit index in COUNT, 1 << COUNT in MASK.  */
+static void gen_bt_flags(DisasContext *s, X86DecodedInsn *decode, TCGv src,
+                         TCGv count, TCGv mask)
 {
     TCGv cf;
 
@@ -1467,15 +1468,34 @@ static void gen_bt_flags(DisasContext *s, 
X86DecodedInsn *decode, TCGv src, TCGv
         decode->cc_src = tcg_temp_new();
         decode->cc_dst = cpu_cc_dst;
         decode->cc_op = CC_OP_SARB + cc_op_size(s->cc_op);
-        tcg_gen_shr_tl(decode->cc_src, src, s->T1);
+        tcg_gen_shr_tl(decode->cc_src, src, count);
     }
 }
 
 static void gen_BT(DisasContext *s, X86DecodedInsn *decode)
 {
-    TCGv mask = gen_bt_mask(s, decode);
+    TCGv count = s->T1;
+    TCGv mask;
 
-    gen_bt_flags(s, decode, s->T0, mask);
+    /*
+     * Try to ensure that the rhs of the TSTNE condition is a constant (and a
+     * power of two), as that is more readily available on most TCG backends.
+     *
+     * For immediate bit number gen_bt_mask()'s output is already a constant;
+     * for register bit number, shift the source right and check bit 0.
+     */
+    if (decode->e.op2 == X86_TYPE_I) {
+        mask = gen_bt_mask(s, decode);
+    } else {
+        MemOp ot = decode->op[1].ot;
+
+        tcg_gen_andi_tl(s->T1, s->T1, (8 << ot) - 1);
+        tcg_gen_shr_tl(s->T0, s->T0, s->T1);
+
+        count = tcg_constant_tl(0);
+        mask = tcg_constant_tl(1);
+    }
+    gen_bt_flags(s, decode, s->T0, count, mask);
 }
 
 static void gen_BTC(DisasContext *s, X86DecodedInsn *decode)
@@ -1491,7 +1511,7 @@ static void gen_BTC(DisasContext *s, X86DecodedInsn 
*decode)
         tcg_gen_xor_tl(s->T0, s->T0, mask);
     }
 
-    gen_bt_flags(s, decode, old, mask);
+    gen_bt_flags(s, decode, old, s->T1, mask);
 }
 
 static void gen_BTR(DisasContext *s, X86DecodedInsn *decode)
@@ -1509,7 +1529,7 @@ static void gen_BTR(DisasContext *s, X86DecodedInsn 
*decode)
         tcg_gen_andc_tl(s->T0, s->T0, mask);
     }
 
-    gen_bt_flags(s, decode, old, mask);
+    gen_bt_flags(s, decode, old, s->T1, mask);
 }
 
 static void gen_BTS(DisasContext *s, X86DecodedInsn *decode)
@@ -1525,7 +1545,7 @@ static void gen_BTS(DisasContext *s, X86DecodedInsn 
*decode)
         tcg_gen_or_tl(s->T0, s->T0, mask);
     }
 
-    gen_bt_flags(s, decode, old, mask);
+    gen_bt_flags(s, decode, old, s->T1, mask);
 }
 
 static void gen_BZHI(DisasContext *s, X86DecodedInsn *decode)
-- 
2.47.1