lightning.git - Portable just-in-time compiler library

diff --git a/check/Makefile.am b/check/Makefile.am
index 583bb12..f1155d7 100644
--- a/check/Makefile.am
+++ b/check/Makefile.am

@@ -82,6 +82,7 @@ EXTRA_DIST = \

alu_rsh.tst alu_rsh.ok \

alu_com.tst alu_com.ok \

alu_neg.tst alu_neg.ok \

+ movzr.tst movzr.ok \

fop_abs.tst fop_abs.ok \

fop_sqrt.tst fop_sqrt.ok \

varargs.tst varargs.ok \

@@ -123,6 +124,7 @@ base_TESTS = \

alu_and alu_or alu_xor \

alu_lsh alu_rsh \

alu_com alu_neg \

+ movzr \

fop_abs fop_sqrt \

varargs stack \

clobber carry call \

@@ -151,6 +153,7 @@ x87_TESTS = \

alu_and.x87 alu_or.x87 alu_xor.x87 \

alu_lsh.x87 alu_rsh.x87 \

alu_com.x87 alu_neg.x87 \

+ movzr.x87 \

fop_abs.x87 fop_sqrt.x87 \

varargs.x87 stack.x87 \

clobber.x87 carry.x87 call.x87 \

@@ -174,6 +177,7 @@ x87_nodata_TESTS = \

alu_and.x87.nodata alu_or.x87.nodata alu_xor.x87.nodata \

alu_lsh.x87.nodata alu_rsh.x87.nodata \

alu_com.x87.nodata alu_neg.x87.nodata \

+ movzr.x87.nodata \

fop_abs.x87.nodata fop_sqrt.x87.nodata \

varargs.x87.nodata stack.x87.nodata \

clobber.x87.nodata carry.x87.nodata call.x87.nodata \

@@ -199,6 +203,7 @@ arm_TESTS = \

alu_and.arm alu_or.arm alu_xor.arm \

alu_lsh.arm alu_rsh.arm \

alu_com.arm alu_neg.arm \

+ movzr.arm \

fop_abs.arm fop_sqrt.arm \

varargs.arm stack.arm \

clobber.arm carry.arm call.arm \

@@ -224,6 +229,7 @@ swf_TESTS = \

alu_and.swf alu_or.swf alu_xor.swf \

alu_lsh.swf alu_rsh.swf \

alu_com.swf alu_neg.swf \

+ movzr.swf \

fop_abs.swf fop_sqrt.swf \

varargs.swf stack.swf \

clobber.swf carry.swf call.swf \

@@ -247,6 +253,7 @@ arm_swf_TESTS = \

alu_and.arm.swf alu_or.arm.swf alu_xor.arm.swf \

alu_lsh.arm.swf alu_rsh.arm.swf \

alu_com.arm.swf alu_neg.arm.swf \

+ movzr.arm.swf \

fop_abs.arm.swf fop_sqrt.arm.swf \

varargs.arm.swf stack.arm.swf \

clobber.arm.swf carry.arm.swf call.arm.swf \

@@ -271,6 +278,7 @@ arm4_swf_TESTS = \

alu_and.arm4.swf alu_or.arm4.swf alu_xor.arm4.swf \

alu_lsh.arm4.swf alu_rsh.arm4.swf \

alu_com.arm4.swf alu_neg.arm4.swf \

+ movzr.arm4.swf \

fop_abs.arm4.swf fop_sqrt.arm4.swf \

varargs.arm4.swf stack.arm4.swf \

clobber.arm4.swf carry.arm4.swf call.arm4.swf \

@@ -298,6 +306,7 @@ nodata_TESTS = \

alu_and.nodata alu_or.nodata alu_xor.nodata \

alu_lsh.nodata alu_rsh.nodata \

alu_com.nodata alu_neg.nodata \

+ movzr.nodata \

fop_abs.nodata fop_sqrt.nodata \

varargs.nodata stack.nodata \

clobber.nodata carry.nodata call.nodata \

diff --git a/check/lightning.c b/check/lightning.c
index 4ce30d5..9798a5b 100644
--- a/check/lightning.c
+++ b/check/lightning.c

@@ -328,6 +328,7 @@ static void htonr_ui(void); static void ntohr_ui(void);

static void htonr_ul(void); static void ntohr_ul(void);

#endif

static void htonr(void); static void ntohr(void);

+static void movnr(void); static void movzr(void);

static void ldr_c(void); static void ldi_c(void);

static void ldr_uc(void); static void ldi_uc(void);

static void ldr_s(void); static void ldi_s(void);

@@ -642,6 +643,7 @@ static instr_t instr_vector[] = {

entry(htonr_ul), entry(ntohr_ul),

#endif

entry(htonr), entry(ntohr),

+ entry(movnr), entry(movzr),

entry(ldr_c), entry(ldi_c),

entry(ldr_uc), entry(ldi_uc),

entry(ldr_s), entry(ldi_s),

@@ -1490,6 +1492,7 @@ entry_ir_ir(htonr_ui) entry_ir_ir(ntohr_ui)

entry_ir_ir(htonr_ul) entry_ir_ir(ntohr_ul)

#endif

entry_ir_ir(htonr) entry_ir_ir(ntohr)

+entry_ir_ir_ir(movnr) entry_ir_ir_ir(movzr)

entry_ir_ir(ldr_c) entry_ir_pm(ldi_c)

entry_ir_ir(ldr_uc) entry_ir_pm(ldi_uc)

entry_ir_ir(ldr_s) entry_ir_pm(ldi_s)

diff --git a/check/movzr.ok b/check/movzr.ok
new file mode 100644
index 0000000..9766475
--- /dev/null
+++ b/check/movzr.ok

@@ -0,0 +1 @@

+ok

diff --git a/check/movzr.tst b/check/movzr.tst
new file mode 100644
index 0000000..baa3ff8
--- /dev/null
+++ b/check/movzr.tst

@@ -0,0 +1,62 @@

+.data 8

+ok:

+.c "ok\n"

+

+#define CMOVR(N, T, OP, I0, I1, V, R0, R1, R2) \

+ movi %R1 I0 \

+ movi %R2 I1 \

+ movi %R0 V \

+ OP##r##T %R0 %R1 %R2 \

+ beqi OP##T##N##r_##R0##R1##R2 %R0 V \

+ calli @abort \

+OP##T##N##r_##R0##R1##R2:

+

+/* reg0 = reg1 op reg0 */

+#define CMOVR1(N, T, OP, I0, I1, V, R0, R1, R2) \

+ movi %R0 I1 \

+ movi %R1 I0 \

+ movi %R2 V \

+ OP##r##T %R0 %R1 %R0 \

+ beqr OP##T##N##r_1##R0##R1##R2 %R0 %R2 \

+ calli @abort \

+OP##T##N##r_1##R0##R1##R2:

+

+#define TEST_CMOV1(N, OP, I0, I1, V, R0, R1, R2) \

+ CMOVR(N, , OP, I0, I1, V, R0, R1, R2) \

+ CMOVR1(N, , OP, I0, I1, V, R0, R1, R2) \

+

+#define TEST_CMOV(N, OP, I0, I1, V) \

+ TEST_CMOV1(N, OP, I0, I1, V, v0, v1, v2) \

+ TEST_CMOV1(N, OP, I0, I1, V, v0, v1, r0) \

+ TEST_CMOV1(N, OP, I0, I1, V, v0, v1, r1) \

+ TEST_CMOV1(N, OP, I0, I1, V, v0, v1, r2) \

+ TEST_CMOV1(N, OP, I0, I1, V, v1, v2, r1) \

+ TEST_CMOV1(N, OP, I0, I1, V, v1, v2, r2) \

+ TEST_CMOV1(N, OP, I0, I1, V, v2, r0, r1) \

+ TEST_CMOV1(N, OP, I0, I1, V, v2, r0, r2) \

+ TEST_CMOV1(N, OP, I0, I1, V, r0, r1, r2)

+

+#define MOVZR(N, I0, I1, V) TEST_CMOV(N, movz, I0, I1, V)

+#define MOVNR(N, I0, I1, V) TEST_CMOV(N, movn, I0, I1, V)

+

+.code

+ prolog

+

+ MOVZR(0, 0x0, 0x0, 0x0)

+ MOVZR(1, 0xf7de, 0x0, 0xf7de)

+

+ MOVZR(2, 0x0, 0xdead, 0xdead)

+ MOVZR(3, 0xf7de, 0xdead, 0xdead)

+

+ MOVNR(0, 0x0, 0x0, 0x0)

+ MOVNR(1, 0xf7de, 0x0, 0x0)

+

+ MOVNR(2, 0x0, 0xdead, 0x0)

+ MOVNR(3, 0xf7de, 0xdead, 0xf7de)

+

+ prepare

+ pushargi ok

+ ellipsis

+ finishi @printf

+ ret

+ epilog

diff --git a/doc/body.texi b/doc/body.texi
index 5593dad..48bfb9e 100644
--- a/doc/body.texi
+++ b/doc/body.texi

@@ -248,6 +248,8 @@ lshr O1 = O2 << O3

lshi O1 = O2 << O3

rshr _u O1 = O2 >> O3@footnote{The sign bit is propagated unless using the @code{_u} modifier.}

rshi _u O1 = O2 >> O3@footnote{The sign bit is propagated unless using the @code{_u} modifier.}

+movzr O1 = O3 ? O1 : O2

+movnr O1 = O3 ? O2 : O1

@end example

@item Four operand binary ALU operations

diff --git a/include/lightning.h.in b/include/lightning.h.in
index e1d8a0a..422fc13 100644
--- a/include/lightning.h.in
+++ b/include/lightning.h.in

@@ -891,6 +891,10 @@ typedef enum {

#define jit_movr_d_w(u, v) jit_new_node_ww(jit_code_movr_d_w, u, v)

#define jit_movi_d_w(u, v) jit_new_node_wd(jit_code_movi_d_w, u, v)

+#define jit_movnr(u,v,w) jit_new_node_www(jit_code_movnr,u,v,w)

+#define jit_movzr(u,v,w) jit_new_node_www(jit_code_movzr,u,v,w)

+ jit_code_movnr, jit_code_movzr,

+

jit_code_last_code

} jit_code_t;

diff --git a/lib/jit_aarch64-cpu.c b/lib/jit_aarch64-cpu.c
index 5829464..53698b0 100644
--- a/lib/jit_aarch64-cpu.c
+++ b/lib/jit_aarch64-cpu.c

@@ -290,6 +290,7 @@ typedef union {

# define A64_CBNZ 0x35000000

# define A64_B_C 0x54000000

# define A64_CSINC 0x1a800400

+# define A64_CSSEL 0x1a800000

# define A64_REV 0xdac00c00

# define A64_UDIV 0x1ac00800

# define A64_SDIV 0x1ac00c00

@@ -461,6 +462,7 @@ typedef union {

# define LDPI_PRE(Rt,Rt2,Rn,Simm7) oxxx7(A64_LDP_PRE|XS,Rt,Rt2,Rn,Simm7)

# define STPI_POS(Rt,Rt2,Rn,Simm7) oxxx7(A64_STP_POS|XS,Rt,Rt2,Rn,Simm7)

# define CSET(Rd,Cc) CSINC(Rd,XZR_REGNO,XZR_REGNO,Cc)

+# define CSEL(Rd,Rn,Rm,Cc) oxxxc(A64_CSSEL|XS,Rd,Rn,Rm,Cc)

# define B(Simm26) o26(A64_B,Simm26)

# define BL(Simm26) o26(A64_BL,Simm26)

# define BR(Rn) o_x_(A64_BR,Rn)

@@ -572,6 +574,10 @@ static void _rshi(jit_state_t*,jit_int32_t,jit_int32_t,jit_word_t);

# define rshr_u(r0,r1,r2) LSR(r0,r1,r2)

# define rshi_u(r0,r1,i0) _rshi_u(_jit,r0,r1,i0)

static void _rshi_u(jit_state_t*,jit_int32_t,jit_int32_t,jit_word_t);

+# define movnr(r0,r1,r2) _movnr(_jit,r0,r1,r2)

+static void _movnr(jit_state_t*,jit_int32_t,jit_int32_t,jit_int32_t);

+# define movzr(r0,r1,r2) _movzr(_jit,r0,r1,r2)

+static void _movzr(jit_state_t*,jit_int32_t,jit_int32_t,jit_int32_t);

# define negr(r0,r1) NEG(r0,r1)

# define comr(r0,r1) MVN(r0,r1)

# define andr(r0,r1,r2) AND(r0,r1,r2)

@@ -1376,6 +1382,20 @@ _rshi_u(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1, jit_word_t i0)

}

static void

+_movnr(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1, jit_int32_t r2)

+{

+ CMPI(r2, 0);

+ CSEL(r0, r0, r1, CC_NE);

+}

+

+static void

+_movzr(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1, jit_int32_t r2)

+{

+ CMPI(r2, 0);

+ CSEL(r0, r0, r1, CC_EQ);

+}

+

+static void

_andi(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1, jit_word_t i0)

{

jit_int32_t reg;

diff --git a/lib/jit_aarch64-sz.c b/lib/jit_aarch64-sz.c
index 7e22e0e..4fa7a42 100644
--- a/lib/jit_aarch64-sz.c
+++ b/lib/jit_aarch64-sz.c

@@ -399,4 +399,6 @@

0, /* movi_d_ww */

0, /* movr_d_w */

0, /* movi_d_w */

+ 8, /* movnr */

+ 8, /* movzr */

#endif /* __WORDSIZE */

diff --git a/lib/jit_aarch64.c b/lib/jit_aarch64.c
index 5b2ff49..369408c 100644
--- a/lib/jit_aarch64.c
+++ b/lib/jit_aarch64.c

@@ -1135,6 +1135,8 @@ _emit_code(jit_state_t *_jit)

case_rr(ext, _i);

case_rr(ext, _ui);

case_rr(mov,);

+ case_rrr(movn,);

+ case_rrr(movz,);

case jit_code_movi:

if (node->flag & jit_flag_node) {

temp = node->v.n;

diff --git a/lib/jit_arm-cpu.c b/lib/jit_arm-cpu.c
index f0f5111..c4a550d 100644
--- a/lib/jit_arm-cpu.c
+++ b/lib/jit_arm-cpu.c

@@ -843,6 +843,10 @@ static void _movr(jit_state_t*,jit_int32_t,jit_int32_t);

static void _movi(jit_state_t*,jit_int32_t,jit_word_t);

# define movi_p(r0,i0) _movi_p(_jit,r0,i0)

static jit_word_t _movi_p(jit_state_t*,jit_int32_t,jit_word_t);

+# define movnr(r0,r1,r2) _movnr(_jit,r0,r1,r2)

+static void _movnr(jit_state_t*,jit_int32_t,jit_int32_t,jit_int32_t);

+# define movzr(r0,r1,r2) _movzr(_jit,r0,r1,r2)

+static void _movzr(jit_state_t*,jit_int32_t,jit_int32_t,jit_int32_t);

# define comr(r0,r1) _comr(_jit,r0,r1)

static void _comr(jit_state_t*,jit_int32_t,jit_int32_t);

# define negr(r0,r1) _negr(_jit,r0,r1)

@@ -1583,6 +1587,35 @@ _movi_p(jit_state_t *_jit, jit_int32_t r0, jit_word_t i0)

}

static void

+_movznr(jit_state_t *_jit, int ct, jit_int32_t r0,

+ jit_int32_t r1, jit_int32_t r2)

+{

+ if (jit_thumb_p()) {

+ if (r2 < 7)

+ T1_CMPI(r2, 0);

+ else

+ T2_CMPI(r2, 0);

+ IT(ct);

+ T1_MOV(r0, r1);

+ } else {

+ CMPI(r2, 0);

+ CC_MOV(ct, r0, r1);

+ }

+}

+

+static void

+_movnr(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1, jit_int32_t r2)

+{

+ _movznr(_jit, ARM_CC_NE, r0, r1, r2);

+}

+

+static void

+_movzr(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1, jit_int32_t r2)

+{

+ _movznr(_jit, ARM_CC_EQ, r0, r1, r2);

+}

+

+static void

_comr(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1)

{

if (jit_thumb_p()) {

diff --git a/lib/jit_arm-sz.c b/lib/jit_arm-sz.c
index 9f0d012..6368c13 100644
--- a/lib/jit_arm-sz.c
+++ b/lib/jit_arm-sz.c

@@ -804,5 +804,7 @@

12, /* movi_d_ww */

0, /* movr_d_w */

0, /* movi_d_w */

+ 8, /* movnr */

+ 8, /* movzr */

#endif /* __ARM_PCS_VFP */

#endif /* __WORDSIZE */

diff --git a/lib/jit_arm.c b/lib/jit_arm.c
index 783fa90..051f84d 100644
--- a/lib/jit_arm.c
+++ b/lib/jit_arm.c

@@ -1503,6 +1503,8 @@ _emit_code(jit_state_t *_jit)

case_rr(ext, _s);

case_rr(ext, _us);

case_rr(mov,);

+ case_rrr(movn,);

+ case_rrr(movz,);

case jit_code_movi:

if (node->flag & jit_flag_node) {

temp = node->v.n;

diff --git a/lib/jit_mips-cpu.c b/lib/jit_mips-cpu.c
index d844c26..6541a7b 100644
--- a/lib/jit_mips-cpu.c
+++ b/lib/jit_mips-cpu.c

@@ -391,6 +391,7 @@ static void _nop(jit_state_t*,jit_int32_t);

# define JR(r0) hrrrit(MIPS_SPECIAL,r0,0,0,0,MIPS_JR)

# endif

# define J(i0) hi(MIPS_J,i0)

+# define MOVN(rd,rs,rt) hrrrit(0,rs,rt,rd,0,MIPS_MOVN)

# define MOVZ(rd,rs,rt) hrrrit(0,rs,rt,rd,0,MIPS_MOVZ)

# define comr(r0,r1) xori(r0,r1,-1)

# define negr(r0,r1) subr(r0,_ZERO_REGNO,r1)

@@ -506,6 +507,8 @@ static void _movr(jit_state_t*,jit_int32_t,jit_int32_t);

static void _movi(jit_state_t*,jit_int32_t,jit_word_t);

# define movi_p(r0,i0) _movi_p(_jit,r0,i0)

static jit_word_t _movi_p(jit_state_t*,jit_int32_t,jit_word_t);

+# define movnr(r0,r1,r2) MOVN(r0, r1, r2)

+# define movzr(r0,r1,r2) MOVZ(r0, r1, r2)

# define ldr_c(r0,r1) LB(r0,0,r1)

# define ldi_c(r0,i0) _ldi_c(_jit,r0,i0)

static void _ldi_c(jit_state_t*,jit_int32_t,jit_word_t);

diff --git a/lib/jit_mips-sz.c b/lib/jit_mips-sz.c
index 613aa00..b33fef2 100644
--- a/lib/jit_mips-sz.c
+++ b/lib/jit_mips-sz.c

@@ -1207,4 +1207,6 @@

0, /* movi_d_ww */

4, /* movr_d_w */

12, /* movi_d_w */

+ 4, /* movnr */

+ 4, /* movzr */

#endif /* __WORDSIZE */

diff --git a/lib/jit_mips.c b/lib/jit_mips.c
index dafade8..5ffad2b 100644
--- a/lib/jit_mips.c
+++ b/lib/jit_mips.c

@@ -1428,6 +1428,8 @@ _emit_code(jit_state_t *_jit)

case_rr(ext, _i);

case_rr(ext, _ui);

#endif

+ case_rrr(movn,);

+ case_rrr(movz,);

case_rr(mov,);

case jit_code_movi:

if (node->flag & jit_flag_node) {

diff --git a/lib/jit_ppc-cpu.c b/lib/jit_ppc-cpu.c
index 9e99771..0046a05 100644
--- a/lib/jit_ppc-cpu.c
+++ b/lib/jit_ppc-cpu.c

@@ -505,6 +505,10 @@ static void _nop(jit_state_t*,jit_int32_t);

static void _movr(jit_state_t*,jit_int32_t,jit_int32_t);

# define movi(r0,i0) _movi(_jit,r0,i0)

static void _movi(jit_state_t*,jit_int32_t,jit_word_t);

+# define movnr(r0,r1,r2) _movnr(_jit,r0,r1,r2)

+static void _movnr(jit_state_t*,jit_int32_t,jit_int32_t,jit_int32_t);

+# define movzr(r0,r1,r2) _movzr(_jit,r0,r1,r2)

+static void _movzr(jit_state_t*,jit_int32_t,jit_int32_t,jit_int32_t);

# define movi_p(r0,i0) _movi_p(_jit,r0,i0)

static jit_word_t _movi_p(jit_state_t*,jit_int32_t,jit_word_t);

# define negr(r0,r1) NEG(r0,r1)

@@ -1120,6 +1124,22 @@ _movi(jit_state_t *_jit, jit_int32_t r0, jit_word_t i0)

}

}

+static void

+_movnr(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1, jit_int32_t r2)

+{

+ CMPWI(r2, 0);

+ BEQ(8);

+ MR(r0, r1);

+}

+

+static void

+_movzr(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1, jit_int32_t r2)

+{

+ CMPWI(r2, 0);

+ BNE(8);

+ MR(r0, r1);

+}

+

static jit_word_t

_movi_p(jit_state_t *_jit, jit_int32_t r0, jit_word_t i0)

{

diff --git a/lib/jit_ppc-sz.c b/lib/jit_ppc-sz.c
index 788ac45..28251b4 100644
--- a/lib/jit_ppc-sz.c
+++ b/lib/jit_ppc-sz.c

@@ -401,6 +401,8 @@

0, /* movi_d_ww */

0, /* movr_d_w */

0, /* movi_d_w */

+ 12, /* movnr */

+ 12, /* movzr */

#endif /* _CALL_SYV */

#endif /* __BYTE_ORDER */

#endif /* __powerpc__ */

@@ -809,6 +811,8 @@

0, /* movi_d_ww */

0, /* movr_d_w */

0, /* movi_d_w */

+ 12, /* movnr */

+ 12, /* movzr */

#endif /* _CALL_AIX */

#endif /* __BYTEORDER */

#endif /* __powerpc__ */

@@ -1216,6 +1220,8 @@

0, /* movi_d_ww */

0, /* movr_d_w */

0, /* movi_d_w */

+ 12, /* movnr */

+ 12, /* movzr */

#endif /* __BYTEORDER */

#endif /* __powerpc__ */

#endif /* __WORDSIZE */

@@ -1622,6 +1628,8 @@

0, /* movi_d_ww */

0, /* movr_d_w */

0, /* movi_d_w */

+ 12, /* movnr */

+ 12, /* movzr */

#endif /* __BYTE_ORDER */

#endif /* __powerpc__ */

#endif /* __WORDSIZE */

diff --git a/lib/jit_ppc.c b/lib/jit_ppc.c
index 0826f4e..d05d4b1 100644
--- a/lib/jit_ppc.c
+++ b/lib/jit_ppc.c

@@ -1358,6 +1358,8 @@ _emit_code(jit_state_t *_jit)

# endif

case_rr(neg,);

case_rr(com,);

+ case_rrr(movn,);

+ case_rrr(movz,);

case_rr(mov,);

case jit_code_movi:

if (node->flag & jit_flag_node) {

diff --git a/lib/jit_x86-cpu.c b/lib/jit_x86-cpu.c
index 547f36c..6dcf672 100644
--- a/lib/jit_x86-cpu.c
+++ b/lib/jit_x86-cpu.c

@@ -369,6 +369,10 @@ static void _movcr_u(jit_state_t*,jit_int32_t,jit_int32_t);

static void _movsr(jit_state_t*,jit_int32_t,jit_int32_t);

# define movsr_u(r0, r1) _movsr_u(_jit, r0, r1)

static void _movsr_u(jit_state_t*,jit_int32_t,jit_int32_t);

+#define movnr(r0, r1, r2) _movnr(_jit, r0, r1, r2)

+static void _movnr(jit_state_t*, jit_int32_t, jit_int32_t, jit_int32_t);

+#define movzr(r0, r1, r2) _movzr(_jit, r0, r1, r2)

+static void _movzr(jit_state_t*, jit_int32_t, jit_int32_t, jit_int32_t);

# if __X64 && !__X64_32

# define movir(r0, r1) _movir(_jit, r0, r1)

static void _movir(jit_state_t*,jit_int32_t,jit_int32_t);

@@ -698,6 +702,7 @@ static void _patch_at(jit_state_t*, jit_node_t*, jit_word_t, jit_word_t);

# define ffsl(l) __builtin_ffsl(l)

# endif

# endif

+# define jit_cmov_p() jit_cpu.cmov

#endif

#if CODE

@@ -2213,6 +2218,32 @@ _movsr_u(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1)

mrm(0x03, r7(r0), r7(r1));

}

+static void

+_movnr(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1, jit_int32_t r2)

+{

+ assert(jit_cmov_p());

+

+ testr(r2, r2);

+

+ rex(0, WIDE, r0, _NOREG, r1);

+ ic(0x0f);

+ ic(0x45);

+ mrm(0x03, r7(r0), r7(r1));

+}

+

+static void

+_movzr(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1, jit_int32_t r2)

+{

+ assert(jit_cmov_p());

+

+ testr(r2, r2);

+

+ rex(0, WIDE, r0, _NOREG, r1);

+ ic(0x0f);

+ ic(0x44);

+ mrm(0x03, r7(r0), r7(r1));

+}

+

#if __X64

static void

_movir(jit_state_t *_jit, jit_int32_t r0, jit_int32_t r1)

diff --git a/lib/jit_x86-sz.c b/lib/jit_x86-sz.c
index 663b840..2cf8880 100644
--- a/lib/jit_x86-sz.c
+++ b/lib/jit_x86-sz.c

@@ -399,6 +399,8 @@

0, /* movi_d_ww */

0, /* movr_d_w */

0, /* movi_d_w */

+ 7, /* movnr */

+ 7, /* movzr */

#endif

#if __X64

@@ -802,6 +804,8 @@

0, /* movi_d_ww */

0, /* movr_d_w */

0, /* movi_d_w */

+ 7, /* movnr */

+ 7, /* movzr */

#else

# if __X64_32

@@ -1204,6 +1208,8 @@

0, /* movi_d_ww */

0, /* movr_d_w */

0, /* movi_d_w */

+ 7, /* movnr */

+ 7, /* movzr */

# else

#define JIT_INSTR_MAX 115

@@ -1605,6 +1611,8 @@

0, /* movi_d_ww */

0, /* movr_d_w */

0, /* movi_d_w */

+ 7, /* movnr */

+ 7, /* movzr */

#endif /* __CYGWIN__ || _WIN32 */

# endif /* __X64_32 */

#endif /* __X64 */

diff --git a/lib/jit_x86.c b/lib/jit_x86.c
index 7dd900e..133ee39 100644
--- a/lib/jit_x86.c
+++ b/lib/jit_x86.c

@@ -1674,6 +1674,8 @@ _emit_code(jit_state_t *_jit)

case_rrw(gt, _u);

case_rrr(ne,);

case_rrw(ne,);

+ case_rrr(movn,);

+ case_rrr(movz,);

case_rr(mov,);

case jit_code_movi:

if (node->flag & jit_flag_node) {

diff --git a/lib/lightning.c b/lib/lightning.c
index 0ed663d..15fc23d 100644
--- a/lib/lightning.c
+++ b/lib/lightning.c

@@ -1435,6 +1435,7 @@ _jit_classify(jit_state_t *_jit, jit_code_t code)

case jit_code_unordi_d:

mask = jit_cc_a0_reg|jit_cc_a0_chg|jit_cc_a1_reg|jit_cc_a2_dbl;

break;

+ case jit_code_movnr: case jit_code_movzr:

case jit_code_addr: case jit_code_addxr: case jit_code_addcr:

case jit_code_subr: case jit_code_subxr: case jit_code_subcr:

case jit_code_mulr: case jit_code_divr: case jit_code_divr_u: