py: Compress load-int, load-fast, store-fast, unop, binop bytecodes.

There is a lot potential in compress bytecodes and make more use of the
coding space.  This patch introduces "multi" bytecodes which have their
argument included in the bytecode (by addition).

UNARY_OP and BINARY_OP now no longer take a 1 byte argument for the
opcode.  Rather, the opcode is included in the first byte itself.

LOAD_FAST_[0,1,2] and STORE_FAST_[0,1,2] are removed in favour of their
multi versions, which can take an argument between 0 and 15 inclusive.
The majority of LOAD_FAST/STORE_FAST codes fit in this range and so this
saves a byte for each of these.

LOAD_CONST_SMALL_INT_MULTI is used to load small ints between -16 and 47
inclusive.  Such ints are quite common and now only need 1 byte to
store, and now have much faster decoding.

In all this patch saves about 2% RAM for typically bytecode (1.8% on
64-bit test, 2.5% on pyboard test).  It also reduces the binary size
(because bytecodes are simplified) and doesn't harm performance.

1 files changed, 75 insertions, 77 deletions

diff --git a/py/bc0.h b/py/bc0.h
index b8e07cbc5..0b1091759 100644
--- a/py/bc0.h
+++ b/py/bc0.h
@@ -31,92 +31,90 @@
 #define MP_BC_LOAD_CONST_NONE    (0x11)
 #define MP_BC_LOAD_CONST_TRUE    (0x12)
 #define MP_BC_LOAD_CONST_ELLIPSIS    (0x13)
-#define MP_BC_LOAD_CONST_SMALL_INT   (0x14) // 24-bit, in excess
+#define MP_BC_LOAD_CONST_SMALL_INT   (0x14) // signed var-int
 #define MP_BC_LOAD_CONST_INT     (0x15) // qstr
 #define MP_BC_LOAD_CONST_DEC     (0x16) // qstr
-#define MP_BC_LOAD_CONST_BYTES   (0x18) // qstr
-#define MP_BC_LOAD_CONST_STRING  (0x19) // qstr
-#define MP_BC_LOAD_NULL          (0x1a)
+#define MP_BC_LOAD_CONST_BYTES   (0x17) // qstr
+#define MP_BC_LOAD_CONST_STRING  (0x18) // qstr
+#define MP_BC_LOAD_NULL          (0x19)
 
-#define MP_BC_LOAD_FAST_0        (0x20)
-#define MP_BC_LOAD_FAST_1        (0x21)
-#define MP_BC_LOAD_FAST_2        (0x22)
-#define MP_BC_LOAD_FAST_N        (0x23) // uint
-#define MP_BC_LOAD_DEREF         (0x25) // uint
-#define MP_BC_LOAD_NAME          (0x26) // qstr
-#define MP_BC_LOAD_GLOBAL        (0x27) // qstr
-#define MP_BC_LOAD_ATTR          (0x28) // qstr
-#define MP_BC_LOAD_METHOD        (0x29) // qstr
-#define MP_BC_LOAD_BUILD_CLASS   (0x2a)
-#define MP_BC_LOAD_SUBSCR        (0x2b)
+#define MP_BC_LOAD_FAST_N        (0x1a) // uint
+#define MP_BC_LOAD_DEREF         (0x1b) // uint
+#define MP_BC_LOAD_NAME          (0x1c) // qstr
+#define MP_BC_LOAD_GLOBAL        (0x1d) // qstr
+#define MP_BC_LOAD_ATTR          (0x1e) // qstr
+#define MP_BC_LOAD_METHOD        (0x1f) // qstr
+#define MP_BC_LOAD_BUILD_CLASS   (0x20)
+#define MP_BC_LOAD_SUBSCR        (0x21)
 
-#define MP_BC_STORE_FAST_0       (0x30)
-#define MP_BC_STORE_FAST_1       (0x31)
-#define MP_BC_STORE_FAST_2       (0x32)
-#define MP_BC_STORE_FAST_N       (0x33) // uint
-#define MP_BC_STORE_DEREF        (0x34) // uint
-#define MP_BC_STORE_NAME         (0x35) // qstr
-#define MP_BC_STORE_GLOBAL       (0x36) // qstr
-#define MP_BC_STORE_ATTR         (0x37) // qstr
-#define MP_BC_STORE_SUBSCR       (0x38)
+#define MP_BC_STORE_FAST_N       (0x22) // uint
+#define MP_BC_STORE_DEREF        (0x23) // uint
+#define MP_BC_STORE_NAME         (0x24) // qstr
+#define MP_BC_STORE_GLOBAL       (0x25) // qstr
+#define MP_BC_STORE_ATTR         (0x26) // qstr
+#define MP_BC_STORE_SUBSCR       (0x27)
 
-#define MP_BC_DELETE_FAST        (0x39) // uint
-#define MP_BC_DELETE_DEREF       (0x3a) // uint
-#define MP_BC_DELETE_NAME        (0x3b) // qstr
-#define MP_BC_DELETE_GLOBAL      (0x3c) // qstr
+#define MP_BC_DELETE_FAST        (0x28) // uint
+#define MP_BC_DELETE_DEREF       (0x29) // uint
+#define MP_BC_DELETE_NAME        (0x2a) // qstr
+#define MP_BC_DELETE_GLOBAL      (0x2b) // qstr
 
-#define MP_BC_DUP_TOP            (0x40)
-#define MP_BC_DUP_TOP_TWO        (0x41)
-#define MP_BC_POP_TOP            (0x42)
-#define MP_BC_ROT_TWO            (0x43)
-#define MP_BC_ROT_THREE          (0x44)
+#define MP_BC_DUP_TOP            (0x30)
+#define MP_BC_DUP_TOP_TWO        (0x31)
+#define MP_BC_POP_TOP            (0x32)
+#define MP_BC_ROT_TWO            (0x33)
+#define MP_BC_ROT_THREE          (0x34)
 
-#define MP_BC_JUMP               (0x45) // rel byte code offset, 16-bit signed, in excess
-#define MP_BC_POP_JUMP_IF_TRUE   (0x46) // rel byte code offset, 16-bit signed, in excess
-#define MP_BC_POP_JUMP_IF_FALSE  (0x47) // rel byte code offset, 16-bit signed, in excess
-#define MP_BC_JUMP_IF_TRUE_OR_POP    (0x48) // rel byte code offset, 16-bit signed, in excess
-#define MP_BC_JUMP_IF_FALSE_OR_POP   (0x49) // rel byte code offset, 16-bit signed, in excess
-#define MP_BC_SETUP_WITH         (0x4d) // rel byte code offset, 16-bit unsigned
-#define MP_BC_WITH_CLEANUP       (0x4e)
-#define MP_BC_SETUP_EXCEPT       (0x4f) // rel byte code offset, 16-bit unsigned
-#define MP_BC_SETUP_FINALLY      (0x50) // rel byte code offset, 16-bit unsigned
-#define MP_BC_END_FINALLY        (0x51)
-#define MP_BC_GET_ITER           (0x52)
-#define MP_BC_FOR_ITER           (0x53) // rel byte code offset, 16-bit unsigned
-#define MP_BC_POP_BLOCK          (0x54)
-#define MP_BC_POP_EXCEPT         (0x55)
-#define MP_BC_UNWIND_JUMP        (0x56) // rel byte code offset, 16-bit signed, in excess; then a byte
+#define MP_BC_JUMP               (0x35) // rel byte code offset, 16-bit signed, in excess
+#define MP_BC_POP_JUMP_IF_TRUE   (0x36) // rel byte code offset, 16-bit signed, in excess
+#define MP_BC_POP_JUMP_IF_FALSE  (0x37) // rel byte code offset, 16-bit signed, in excess
+#define MP_BC_JUMP_IF_TRUE_OR_POP    (0x38) // rel byte code offset, 16-bit signed, in excess
+#define MP_BC_JUMP_IF_FALSE_OR_POP   (0x39) // rel byte code offset, 16-bit signed, in excess
+#define MP_BC_SETUP_WITH         (0x3d) // rel byte code offset, 16-bit unsigned
+#define MP_BC_WITH_CLEANUP       (0x3e)
+#define MP_BC_SETUP_EXCEPT       (0x3f) // rel byte code offset, 16-bit unsigned
+#define MP_BC_SETUP_FINALLY      (0x40) // rel byte code offset, 16-bit unsigned
+#define MP_BC_END_FINALLY        (0x41)
+#define MP_BC_GET_ITER           (0x42)
+#define MP_BC_FOR_ITER           (0x43) // rel byte code offset, 16-bit unsigned
+#define MP_BC_POP_BLOCK          (0x44)
+#define MP_BC_POP_EXCEPT         (0x45)
+#define MP_BC_UNWIND_JUMP        (0x46) // rel byte code offset, 16-bit signed, in excess; then a byte
 
-#define MP_BC_NOT                (0x60)
-#define MP_BC_UNARY_OP           (0x61) // byte
-#define MP_BC_BINARY_OP          (0x62) // byte
+#define MP_BC_NOT                (0x47)
 
-#define MP_BC_BUILD_TUPLE        (0x70) // uint
-#define MP_BC_BUILD_LIST         (0x71) // uint
-#define MP_BC_LIST_APPEND        (0x72) // uint
-#define MP_BC_BUILD_MAP          (0x73) // uint
-#define MP_BC_STORE_MAP          (0x74)
-#define MP_BC_MAP_ADD            (0x75) // uint
-#define MP_BC_BUILD_SET          (0x76) // uint
-#define MP_BC_SET_ADD            (0x77) // uint
-#define MP_BC_BUILD_SLICE        (0x78) // uint
-#define MP_BC_UNPACK_SEQUENCE    (0x79) // uint
-#define MP_BC_UNPACK_EX          (0x7a) // uint
+#define MP_BC_BUILD_TUPLE        (0x50) // uint
+#define MP_BC_BUILD_LIST         (0x51) // uint
+#define MP_BC_LIST_APPEND        (0x52) // uint
+#define MP_BC_BUILD_MAP          (0x53) // uint
+#define MP_BC_STORE_MAP          (0x54)
+#define MP_BC_MAP_ADD            (0x55) // uint
+#define MP_BC_BUILD_SET          (0x56) // uint
+#define MP_BC_SET_ADD            (0x57) // uint
+#define MP_BC_BUILD_SLICE        (0x58) // uint
+#define MP_BC_UNPACK_SEQUENCE    (0x59) // uint
+#define MP_BC_UNPACK_EX          (0x5a) // uint
 
-#define MP_BC_RETURN_VALUE       (0x80)
-#define MP_BC_RAISE_VARARGS      (0x81) // byte
-#define MP_BC_YIELD_VALUE        (0x82)
-#define MP_BC_YIELD_FROM         (0x83)
+#define MP_BC_RETURN_VALUE       (0x5b)
+#define MP_BC_RAISE_VARARGS      (0x5c) // byte
+#define MP_BC_YIELD_VALUE        (0x5d)
+#define MP_BC_YIELD_FROM         (0x5e)
 
-#define MP_BC_MAKE_FUNCTION         (0x90) // uint
-#define MP_BC_MAKE_FUNCTION_DEFARGS (0x91) // uint
-#define MP_BC_MAKE_CLOSURE          (0x92) // uint
-#define MP_BC_MAKE_CLOSURE_DEFARGS  (0x93) // uint
-#define MP_BC_CALL_FUNCTION         (0x94) // uint
-#define MP_BC_CALL_FUNCTION_VAR_KW  (0x95) // uint
-#define MP_BC_CALL_METHOD           (0x96) // uint
-#define MP_BC_CALL_METHOD_VAR_KW    (0x97) // uint
+#define MP_BC_MAKE_FUNCTION         (0x60) // uint
+#define MP_BC_MAKE_FUNCTION_DEFARGS (0x61) // uint
+#define MP_BC_MAKE_CLOSURE          (0x62) // uint
+#define MP_BC_MAKE_CLOSURE_DEFARGS  (0x63) // uint
+#define MP_BC_CALL_FUNCTION         (0x64) // uint
+#define MP_BC_CALL_FUNCTION_VAR_KW  (0x65) // uint
+#define MP_BC_CALL_METHOD           (0x66) // uint
+#define MP_BC_CALL_METHOD_VAR_KW    (0x67) // uint
 
-#define MP_BC_IMPORT_NAME        (0xe0) // qstr
-#define MP_BC_IMPORT_FROM        (0xe1) // qstr
-#define MP_BC_IMPORT_STAR        (0xe2)
+#define MP_BC_IMPORT_NAME        (0x68) // qstr
+#define MP_BC_IMPORT_FROM        (0x69) // qstr
+#define MP_BC_IMPORT_STAR        (0x6a)
+
+#define MP_BC_LOAD_CONST_SMALL_INT_MULTI (0x70) // + N(64)
+#define MP_BC_LOAD_FAST_MULTI            (0xb0) // + N(16)
+#define MP_BC_STORE_FAST_MULTI           (0xc0) // + N(16)
+#define MP_BC_UNARY_OP_MULTI             (0xd0) // + op(5)
+#define MP_BC_BINARY_OP_MULTI            (0xd5) // + op(35)