Add Intel encodings for brz.b1 and brnz.b1.

Use these encodings to test trapz.b1 and trapnz.b1.

When a b1 value is stored in a register, only the low 8 bits are valid.
This is so we can use the various setCC instructions to generate the b1
registers.

cranelift/filetests/isa/intel/legalize-custom.cton

@@ -28,3 +28,27 @@ ebb0(v1: i32):
     ; check: $new:
     ; nextln: return
 }
+
+function %cond_trap_b1(i32) {
+ebb0(v1: i32):
+    v2 = icmp_imm eq v1, 6
+    trapz v2
+    return
+    ; check: $ebb0($v1: i32):
+    ; check: brnz $v2, $(new=$EBB)
+    ; nextln: trap
+    ; check: $new:
+    ; nextln: return
+}
+
+function %cond_trap2_b1(i32) {
+ebb0(v1: i32):
+    v2 = icmp_imm eq v1, 6
+    trapnz v2
+    return
+    ; check: $ebb0($v1: i32):
+    ; check: brz $v2, $(new=$EBB)
+    ; nextln: trap
+    ; check: $new:
+    ; nextln: return
+}

lib/cretonne/meta/isa/intel/encodings.py

@@ -22,12 +22,14 @@ except ImportError:
 
 I32.legalize_type(
         default=narrow,
+        b1=expand,
         i32=intel_expand,
         f32=expand,
         f64=expand)
 
 I64.legalize_type(
         default=narrow,
+        b1=expand,
         i32=intel_expand,
         i64=intel_expand,
         f32=expand,
@@ -238,6 +240,16 @@ I64.enc(base.jump, *r.jmpd(0xe9))
 enc_i32_i64(base.brz, r.tjccb, 0x74)
 enc_i32_i64(base.brnz, r.tjccb, 0x75)
 
+# Branch on a b1 value in a register only looks at the low 8 bits. See also
+# bint encodings below.
+I32.enc(base.brz.b1, *r.t8jccb_abcd(0x74))
+I64.enc(base.brz.b1, *r.t8jccb_abcd.rex(0x74))
+I64.enc(base.brz.b1, *r.t8jccb_abcd(0x74))
+I32.enc(base.brnz.b1, *r.t8jccb_abcd(0x75))
+I64.enc(base.brnz.b1, *r.t8jccb_abcd.rex(0x75))
+I64.enc(base.brnz.b1, *r.t8jccb_abcd(0x75))
+
+
 #
 # Trap as ud2
 #

lib/cretonne/meta/isa/intel/recipes.py

@@ -515,6 +515,22 @@ tjccb = TailRecipe(
         disp1(destination, func, sink);
         ''')
 
+# 8-bit test-and-branch.
+#
+# Same as tjccb, but only looks at the low 8 bits of the register, for b1
+# types.
+t8jccb_abcd = TailRecipe(
+        't8jccb_abcd', Branch, size=1 + 2, ins=ABCD, outs=(),
+        branch_range=(2, 8),
+        emit='''
+        // test8 r, r.
+        PUT_OP(0x84, rex2(in_reg0, in_reg0), sink);
+        modrm_rr(in_reg0, in_reg0, sink);
+        // Jcc instruction.
+        sink.put1(bits as u8);
+        disp1(destination, func, sink);
+        ''')
+
 # Comparison that produces a `b1` result in a GPR.
 #
 # This is a macro of a `cmp` instruction followed by a `setCC` instruction.