circuitpython/tests/cmdline/cmd_showbc.py.exp

547 lines
11 KiB
Plaintext
Raw Normal View History

File cmdline/cmd_showbc.py, code block '<module>' (descriptor: \.\+, bytecode @\.\+ bytes)
Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
########
arg names:
(N_STATE 3)
(N_EXC_STACK 0)
bc=0 line=1
########
bc=\\d\+ line=160
00 MAKE_FUNCTION \.\+
\\d\+ STORE_NAME f
\\d\+ MAKE_FUNCTION \.\+
\\d\+ STORE_NAME f
\\d\+ LOAD_CONST_SMALL_INT 1
\\d\+ BUILD_TUPLE 1
\\d\+ LOAD_NULL
\\d\+ MAKE_FUNCTION_DEFARGS \.\+
\\d\+ STORE_NAME f
\\d\+ MAKE_FUNCTION \.\+
\\d\+ STORE_NAME f
\\d\+ LOAD_BUILD_CLASS
\\d\+ MAKE_FUNCTION \.\+
\\d\+ LOAD_CONST_STRING 'Class'
\\d\+ CALL_FUNCTION n=2 nkw=0
\\d\+ STORE_NAME Class
\\d\+ DELETE_NAME Class
\\d\+ MAKE_FUNCTION \.\+
\\d\+ STORE_NAME f
\\d\+ LOAD_CONST_SMALL_INT 0
\\d\+ LOAD_CONST_STRING '*'
\\d\+ BUILD_TUPLE 1
\\d\+ IMPORT_NAME 'sys'
\\d\+ IMPORT_STAR
\\d\+ LOAD_CONST_NONE
\\d\+ RETURN_VALUE
File cmdline/cmd_showbc.py, code block 'f' (descriptor: \.\+, bytecode @\.\+ bytes)
Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
########
\.\+rg names:
(N_STATE 22)
(N_EXC_STACK 2)
(INIT_CELL 14)
(INIT_CELL 15)
(INIT_CELL 16)
bc=0 line=1
########
bc=\\d\+ line=127
00 LOAD_CONST_NONE
01 LOAD_CONST_FALSE
02 BINARY_OP 27 __add__
03 LOAD_CONST_TRUE
04 BINARY_OP 27 __add__
05 STORE_FAST 0
06 LOAD_CONST_SMALL_INT 0
07 STORE_FAST 0
08 LOAD_CONST_SMALL_INT 1000
11 STORE_FAST 0
12 LOAD_CONST_SMALL_INT -1000
15 STORE_FAST 0
16 LOAD_CONST_SMALL_INT 1
17 STORE_FAST 0
18 LOAD_CONST_SMALL_INT 1
19 LOAD_CONST_SMALL_INT 2
20 BUILD_TUPLE 2
22 STORE_DEREF 14
24 LOAD_CONST_SMALL_INT 1
25 LOAD_CONST_SMALL_INT 2
26 BUILD_LIST 2
28 STORE_FAST 1
29 LOAD_CONST_SMALL_INT 1
30 LOAD_CONST_SMALL_INT 2
31 BUILD_SET 2
33 STORE_FAST 2
34 BUILD_MAP 0
36 STORE_DEREF 15
38 BUILD_MAP 1
40 LOAD_CONST_SMALL_INT 2
41 LOAD_CONST_SMALL_INT 1
42 STORE_MAP
43 STORE_FAST 3
44 LOAD_CONST_STRING 'a'
47 STORE_FAST 4
48 LOAD_CONST_OBJ \.\+
\\d\+ STORE_FAST 5
\\d\+ LOAD_CONST_SMALL_INT 1
\\d\+ STORE_FAST 6
\\d\+ LOAD_CONST_SMALL_INT 2
\\d\+ STORE_FAST 7
\\d\+ LOAD_FAST 0
\\d\+ LOAD_DEREF 14
58 BINARY_OP 27 __add__
\\d\+ STORE_FAST 8
\\d\+ LOAD_FAST 0
\\d\+ UNARY_OP 1 __neg__
\\d\+ STORE_FAST 9
\\d\+ LOAD_FAST 0
\\d\+ UNARY_OP 3
\\d\+ STORE_FAST 10
\\d\+ LOAD_FAST 0
\\d\+ LOAD_DEREF 14
\\d\+ DUP_TOP
\\d\+ ROT_THREE
\\d\+ BINARY_OP 2 __eq__
\\d\+ JUMP_IF_FALSE_OR_POP \\d\+
\\d\+ LOAD_FAST 1
\\d\+ BINARY_OP 2 __eq__
\\d\+ JUMP \\d\+
\\d\+ ROT_TWO
\\d\+ POP_TOP
\\d\+ STORE_FAST 10
\\d\+ LOAD_FAST 0
\\d\+ LOAD_DEREF 14
\\d\+ BINARY_OP 2 __eq__
\\d\+ JUMP_IF_FALSE_OR_POP \\d\+
\\d\+ LOAD_DEREF 14
\\d\+ LOAD_FAST 1
\\d\+ BINARY_OP 2 __eq__
\\d\+ UNARY_OP 3
\\d\+ STORE_FAST 10
\\d\+ LOAD_DEREF 14
all: Remove MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE. This commit removes all parts of code associated with the existing MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE optimisation option, including the -mcache-lookup-bc option to mpy-cross. This feature originally provided a significant performance boost for Unix, but wasn't able to be enabled for MCU targets (due to frozen bytecode), and added significant extra complexity to generating and distributing .mpy files. The equivalent performance gain is now provided by the combination of MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE (which has been enabled on the unix port in the previous commit). It's hard to provide precise performance numbers, but tests have been run on a wide variety of architectures (x86-64, ARM Cortex, Aarch64, RISC-V, xtensa) and they all generally agree on the qualitative improvements seen by the combination of MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE. For example, on a "quiet" Linux x64 environment (i3-5010U @ 2.10GHz) the change from CACHE_MAP_LOOKUP_IN_BYTECODE, to LOAD_ATTR_FAST_PATH combined with MAP_LOOKUP_CACHE is: diff of scores (higher is better) N=2000 M=2000 bccache -> attrmapcache diff diff% (error%) bm_chaos.py 13742.56 -> 13905.67 : +163.11 = +1.187% (+/-3.75%) bm_fannkuch.py 60.13 -> 61.34 : +1.21 = +2.012% (+/-2.11%) bm_fft.py 113083.20 -> 114793.68 : +1710.48 = +1.513% (+/-1.57%) bm_float.py 256552.80 -> 243908.29 : -12644.51 = -4.929% (+/-1.90%) bm_hexiom.py 521.93 -> 625.41 : +103.48 = +19.826% (+/-0.40%) bm_nqueens.py 197544.25 -> 217713.12 : +20168.87 = +10.210% (+/-3.01%) bm_pidigits.py 8072.98 -> 8198.75 : +125.77 = +1.558% (+/-3.22%) misc_aes.py 17283.45 -> 16480.52 : -802.93 = -4.646% (+/-0.82%) misc_mandel.py 99083.99 -> 128939.84 : +29855.85 = +30.132% (+/-5.88%) misc_pystone.py 83860.10 -> 82592.56 : -1267.54 = -1.511% (+/-2.27%) misc_raytrace.py 21490.40 -> 22227.23 : +736.83 = +3.429% (+/-1.88%) This shows that the new optimisations are at least as good as the existing inline-bytecode-caching, and are sometimes much better (because the new ones apply caching to a wider variety of map lookups). The new optimisations can also benefit code generated by the native emitter, because they apply to the runtime rather than the generated code. The improvement for the native emitter when LOAD_ATTR_FAST_PATH and MAP_LOOKUP_CACHE are enabled is (same Linux environment as above): diff of scores (higher is better) N=2000 M=2000 native -> nat-attrmapcache diff diff% (error%) bm_chaos.py 14130.62 -> 15464.68 : +1334.06 = +9.441% (+/-7.11%) bm_fannkuch.py 74.96 -> 76.16 : +1.20 = +1.601% (+/-1.80%) bm_fft.py 166682.99 -> 168221.86 : +1538.87 = +0.923% (+/-4.20%) bm_float.py 233415.23 -> 265524.90 : +32109.67 = +13.756% (+/-2.57%) bm_hexiom.py 628.59 -> 734.17 : +105.58 = +16.796% (+/-1.39%) bm_nqueens.py 225418.44 -> 232926.45 : +7508.01 = +3.331% (+/-3.10%) bm_pidigits.py 6322.00 -> 6379.52 : +57.52 = +0.910% (+/-5.62%) misc_aes.py 20670.10 -> 27223.18 : +6553.08 = +31.703% (+/-1.56%) misc_mandel.py 138221.11 -> 152014.01 : +13792.90 = +9.979% (+/-2.46%) misc_pystone.py 85032.14 -> 105681.44 : +20649.30 = +24.284% (+/-2.25%) misc_raytrace.py 19800.01 -> 23350.73 : +3550.72 = +17.933% (+/-2.79%) In summary, compared to MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE, the new MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE options: - are simpler; - take less code size; - are faster (generally); - work with code generated by the native emitter; - can be used on embedded targets with a small and constant RAM overhead; - allow the same .mpy bytecode to run on all targets. See #7680 for further discussion. And see also #7653 for a discussion about simplifying mpy-cross options. Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
2021-09-05 22:28:06 -04:00
\\d\+ LOAD_ATTR c
\\d\+ STORE_FAST 11
\\d\+ LOAD_FAST 11
\\d\+ LOAD_DEREF 14
all: Remove MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE. This commit removes all parts of code associated with the existing MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE optimisation option, including the -mcache-lookup-bc option to mpy-cross. This feature originally provided a significant performance boost for Unix, but wasn't able to be enabled for MCU targets (due to frozen bytecode), and added significant extra complexity to generating and distributing .mpy files. The equivalent performance gain is now provided by the combination of MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE (which has been enabled on the unix port in the previous commit). It's hard to provide precise performance numbers, but tests have been run on a wide variety of architectures (x86-64, ARM Cortex, Aarch64, RISC-V, xtensa) and they all generally agree on the qualitative improvements seen by the combination of MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE. For example, on a "quiet" Linux x64 environment (i3-5010U @ 2.10GHz) the change from CACHE_MAP_LOOKUP_IN_BYTECODE, to LOAD_ATTR_FAST_PATH combined with MAP_LOOKUP_CACHE is: diff of scores (higher is better) N=2000 M=2000 bccache -> attrmapcache diff diff% (error%) bm_chaos.py 13742.56 -> 13905.67 : +163.11 = +1.187% (+/-3.75%) bm_fannkuch.py 60.13 -> 61.34 : +1.21 = +2.012% (+/-2.11%) bm_fft.py 113083.20 -> 114793.68 : +1710.48 = +1.513% (+/-1.57%) bm_float.py 256552.80 -> 243908.29 : -12644.51 = -4.929% (+/-1.90%) bm_hexiom.py 521.93 -> 625.41 : +103.48 = +19.826% (+/-0.40%) bm_nqueens.py 197544.25 -> 217713.12 : +20168.87 = +10.210% (+/-3.01%) bm_pidigits.py 8072.98 -> 8198.75 : +125.77 = +1.558% (+/-3.22%) misc_aes.py 17283.45 -> 16480.52 : -802.93 = -4.646% (+/-0.82%) misc_mandel.py 99083.99 -> 128939.84 : +29855.85 = +30.132% (+/-5.88%) misc_pystone.py 83860.10 -> 82592.56 : -1267.54 = -1.511% (+/-2.27%) misc_raytrace.py 21490.40 -> 22227.23 : +736.83 = +3.429% (+/-1.88%) This shows that the new optimisations are at least as good as the existing inline-bytecode-caching, and are sometimes much better (because the new ones apply caching to a wider variety of map lookups). The new optimisations can also benefit code generated by the native emitter, because they apply to the runtime rather than the generated code. The improvement for the native emitter when LOAD_ATTR_FAST_PATH and MAP_LOOKUP_CACHE are enabled is (same Linux environment as above): diff of scores (higher is better) N=2000 M=2000 native -> nat-attrmapcache diff diff% (error%) bm_chaos.py 14130.62 -> 15464.68 : +1334.06 = +9.441% (+/-7.11%) bm_fannkuch.py 74.96 -> 76.16 : +1.20 = +1.601% (+/-1.80%) bm_fft.py 166682.99 -> 168221.86 : +1538.87 = +0.923% (+/-4.20%) bm_float.py 233415.23 -> 265524.90 : +32109.67 = +13.756% (+/-2.57%) bm_hexiom.py 628.59 -> 734.17 : +105.58 = +16.796% (+/-1.39%) bm_nqueens.py 225418.44 -> 232926.45 : +7508.01 = +3.331% (+/-3.10%) bm_pidigits.py 6322.00 -> 6379.52 : +57.52 = +0.910% (+/-5.62%) misc_aes.py 20670.10 -> 27223.18 : +6553.08 = +31.703% (+/-1.56%) misc_mandel.py 138221.11 -> 152014.01 : +13792.90 = +9.979% (+/-2.46%) misc_pystone.py 85032.14 -> 105681.44 : +20649.30 = +24.284% (+/-2.25%) misc_raytrace.py 19800.01 -> 23350.73 : +3550.72 = +17.933% (+/-2.79%) In summary, compared to MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE, the new MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE options: - are simpler; - take less code size; - are faster (generally); - work with code generated by the native emitter; - can be used on embedded targets with a small and constant RAM overhead; - allow the same .mpy bytecode to run on all targets. See #7680 for further discussion. And see also #7653 for a discussion about simplifying mpy-cross options. Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
2021-09-05 22:28:06 -04:00
\\d\+ STORE_ATTR c
\\d\+ LOAD_DEREF 14
\\d\+ LOAD_CONST_SMALL_INT 0
\\d\+ LOAD_SUBSCR
\\d\+ STORE_FAST 12
\\d\+ LOAD_FAST 12
\\d\+ LOAD_DEREF 14
\\d\+ LOAD_CONST_SMALL_INT 0
\\d\+ STORE_SUBSCR
\\d\+ LOAD_DEREF 14
\\d\+ LOAD_CONST_SMALL_INT 0
\\d\+ DUP_TOP_TWO
\\d\+ LOAD_SUBSCR
\\d\+ LOAD_FAST 12
\\d\+ BINARY_OP 14 __iadd__
\\d\+ ROT_THREE
\\d\+ STORE_SUBSCR
\\d\+ LOAD_DEREF 14
\\d\+ LOAD_CONST_NONE
\\d\+ LOAD_CONST_NONE
\\d\+ BUILD_SLICE 2
\\d\+ LOAD_SUBSCR
\\d\+ STORE_FAST 0
\\d\+ LOAD_FAST 1
\\d\+ UNPACK_SEQUENCE 2
\\d\+ STORE_FAST 0
\\d\+ STORE_DEREF 14
\\d\+ LOAD_FAST 0
\\d\+ UNPACK_EX 1
\\d\+ STORE_FAST 0
\\d\+ STORE_FAST 0
\\d\+ LOAD_DEREF 14
\\d\+ LOAD_FAST 0
\\d\+ ROT_TWO
\\d\+ STORE_FAST 0
\\d\+ STORE_DEREF 14
\\d\+ LOAD_FAST 1
\\d\+ LOAD_DEREF 14
\\d\+ LOAD_FAST 0
\\d\+ ROT_THREE
\\d\+ ROT_TWO
\\d\+ STORE_FAST 0
\\d\+ STORE_DEREF 14
\\d\+ STORE_FAST 1
\\d\+ DELETE_FAST 0
\\d\+ LOAD_FAST 0
\\d\+ STORE_GLOBAL gl
\\d\+ DELETE_GLOBAL gl
\\d\+ LOAD_FAST 14
\\d\+ LOAD_FAST 15
\\d\+ MAKE_CLOSURE \.\+ 2
\\d\+ LOAD_FAST 2
\\d\+ GET_ITER
\\d\+ CALL_FUNCTION n=1 nkw=0
\\d\+ STORE_FAST 0
\\d\+ LOAD_FAST 14
\\d\+ LOAD_FAST 15
\\d\+ MAKE_CLOSURE \.\+ 2
\\d\+ LOAD_FAST 2
\\d\+ CALL_FUNCTION n=1 nkw=0
\\d\+ STORE_FAST 0
\\d\+ LOAD_FAST 14
\\d\+ LOAD_FAST 15
\\d\+ MAKE_CLOSURE \.\+ 2
\\d\+ LOAD_FAST 2
\\d\+ CALL_FUNCTION n=1 nkw=0
\\d\+ STORE_FAST 0
\\d\+ LOAD_FAST 0
\\d\+ CALL_FUNCTION n=0 nkw=0
\\d\+ POP_TOP
\\d\+ LOAD_FAST 0
\\d\+ LOAD_CONST_SMALL_INT 1
\\d\+ CALL_FUNCTION n=1 nkw=0
\\d\+ POP_TOP
\\d\+ LOAD_FAST 0
\\d\+ LOAD_CONST_STRING 'b'
\\d\+ LOAD_CONST_SMALL_INT 1
\\d\+ CALL_FUNCTION n=0 nkw=1
\\d\+ POP_TOP
\\d\+ LOAD_FAST 0
\\d\+ LOAD_DEREF 14
\\d\+ LOAD_NULL
\\d\+ CALL_FUNCTION_VAR_KW n=0 nkw=0
\\d\+ POP_TOP
\\d\+ LOAD_FAST 0
\\d\+ LOAD_METHOD b
\\d\+ CALL_METHOD n=0 nkw=0
\\d\+ POP_TOP
\\d\+ LOAD_FAST 0
\\d\+ LOAD_METHOD b
\\d\+ LOAD_CONST_SMALL_INT 1
\\d\+ CALL_METHOD n=1 nkw=0
\\d\+ POP_TOP
\\d\+ LOAD_FAST 0
\\d\+ LOAD_METHOD b
\\d\+ LOAD_CONST_STRING 'c'
\\d\+ LOAD_CONST_SMALL_INT 1
\\d\+ CALL_METHOD n=0 nkw=1
\\d\+ POP_TOP
\\d\+ LOAD_FAST 0
\\d\+ LOAD_METHOD b
\\d\+ LOAD_FAST 1
\\d\+ LOAD_NULL
\\d\+ CALL_METHOD_VAR_KW n=0 nkw=0
\\d\+ POP_TOP
\\d\+ LOAD_FAST 0
\\d\+ POP_JUMP_IF_FALSE \\d\+
\\d\+ LOAD_DEREF 16
\\d\+ POP_TOP
\\d\+ JUMP \\d\+
all: Remove MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE. This commit removes all parts of code associated with the existing MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE optimisation option, including the -mcache-lookup-bc option to mpy-cross. This feature originally provided a significant performance boost for Unix, but wasn't able to be enabled for MCU targets (due to frozen bytecode), and added significant extra complexity to generating and distributing .mpy files. The equivalent performance gain is now provided by the combination of MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE (which has been enabled on the unix port in the previous commit). It's hard to provide precise performance numbers, but tests have been run on a wide variety of architectures (x86-64, ARM Cortex, Aarch64, RISC-V, xtensa) and they all generally agree on the qualitative improvements seen by the combination of MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE. For example, on a "quiet" Linux x64 environment (i3-5010U @ 2.10GHz) the change from CACHE_MAP_LOOKUP_IN_BYTECODE, to LOAD_ATTR_FAST_PATH combined with MAP_LOOKUP_CACHE is: diff of scores (higher is better) N=2000 M=2000 bccache -> attrmapcache diff diff% (error%) bm_chaos.py 13742.56 -> 13905.67 : +163.11 = +1.187% (+/-3.75%) bm_fannkuch.py 60.13 -> 61.34 : +1.21 = +2.012% (+/-2.11%) bm_fft.py 113083.20 -> 114793.68 : +1710.48 = +1.513% (+/-1.57%) bm_float.py 256552.80 -> 243908.29 : -12644.51 = -4.929% (+/-1.90%) bm_hexiom.py 521.93 -> 625.41 : +103.48 = +19.826% (+/-0.40%) bm_nqueens.py 197544.25 -> 217713.12 : +20168.87 = +10.210% (+/-3.01%) bm_pidigits.py 8072.98 -> 8198.75 : +125.77 = +1.558% (+/-3.22%) misc_aes.py 17283.45 -> 16480.52 : -802.93 = -4.646% (+/-0.82%) misc_mandel.py 99083.99 -> 128939.84 : +29855.85 = +30.132% (+/-5.88%) misc_pystone.py 83860.10 -> 82592.56 : -1267.54 = -1.511% (+/-2.27%) misc_raytrace.py 21490.40 -> 22227.23 : +736.83 = +3.429% (+/-1.88%) This shows that the new optimisations are at least as good as the existing inline-bytecode-caching, and are sometimes much better (because the new ones apply caching to a wider variety of map lookups). The new optimisations can also benefit code generated by the native emitter, because they apply to the runtime rather than the generated code. The improvement for the native emitter when LOAD_ATTR_FAST_PATH and MAP_LOOKUP_CACHE are enabled is (same Linux environment as above): diff of scores (higher is better) N=2000 M=2000 native -> nat-attrmapcache diff diff% (error%) bm_chaos.py 14130.62 -> 15464.68 : +1334.06 = +9.441% (+/-7.11%) bm_fannkuch.py 74.96 -> 76.16 : +1.20 = +1.601% (+/-1.80%) bm_fft.py 166682.99 -> 168221.86 : +1538.87 = +0.923% (+/-4.20%) bm_float.py 233415.23 -> 265524.90 : +32109.67 = +13.756% (+/-2.57%) bm_hexiom.py 628.59 -> 734.17 : +105.58 = +16.796% (+/-1.39%) bm_nqueens.py 225418.44 -> 232926.45 : +7508.01 = +3.331% (+/-3.10%) bm_pidigits.py 6322.00 -> 6379.52 : +57.52 = +0.910% (+/-5.62%) misc_aes.py 20670.10 -> 27223.18 : +6553.08 = +31.703% (+/-1.56%) misc_mandel.py 138221.11 -> 152014.01 : +13792.90 = +9.979% (+/-2.46%) misc_pystone.py 85032.14 -> 105681.44 : +20649.30 = +24.284% (+/-2.25%) misc_raytrace.py 19800.01 -> 23350.73 : +3550.72 = +17.933% (+/-2.79%) In summary, compared to MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE, the new MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE options: - are simpler; - take less code size; - are faster (generally); - work with code generated by the native emitter; - can be used on embedded targets with a small and constant RAM overhead; - allow the same .mpy bytecode to run on all targets. See #7680 for further discussion. And see also #7653 for a discussion about simplifying mpy-cross options. Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
2021-09-05 22:28:06 -04:00
\\d\+ LOAD_GLOBAL y
\\d\+ POP_TOP
\\d\+ JUMP \\d\+
\\d\+ LOAD_DEREF 14
\\d\+ POP_TOP
\\d\+ LOAD_FAST 0
\\d\+ POP_JUMP_IF_TRUE \\d\+
\\d\+ JUMP \\d\+
\\d\+ LOAD_DEREF 14
\\d\+ POP_TOP
\\d\+ LOAD_FAST 0
\\d\+ POP_JUMP_IF_FALSE \\d\+
\\d\+ LOAD_FAST 0
\\d\+ JUMP_IF_TRUE_OR_POP \\d\+
\\d\+ LOAD_FAST 0
\\d\+ STORE_FAST 0
\\d\+ LOAD_DEREF 14
\\d\+ GET_ITER_STACK
\\d\+ FOR_ITER \\d\+
\\d\+ STORE_FAST 0
\\d\+ LOAD_FAST 1
\\d\+ POP_TOP
\\d\+ JUMP \\d\+
\\d\+ SETUP_FINALLY \\d\+
\\d\+ SETUP_EXCEPT \\d\+
\\d\+ JUMP \\d\+
\\d\+ JUMP \\d\+
\\d\+ LOAD_FAST 0
\\d\+ POP_JUMP_IF_TRUE \\d\+
\\d\+ POP_EXCEPT_JUMP \\d\+
\\d\+ POP_TOP
\\d\+ LOAD_DEREF 14
\\d\+ POP_TOP
\\d\+ POP_EXCEPT_JUMP \\d\+
\\d\+ END_FINALLY
\\d\+ LOAD_CONST_NONE
\\d\+ LOAD_FAST 1
\\d\+ POP_TOP
\\d\+ END_FINALLY
\\d\+ JUMP \\d\+
\\d\+ SETUP_EXCEPT \\d\+
\\d\+ UNWIND_JUMP \\d\+ 1
\\d\+ POP_EXCEPT_JUMP \\d\+
\\d\+ POP_TOP
\\d\+ POP_EXCEPT_JUMP \\d\+
\\d\+ END_FINALLY
\\d\+ LOAD_FAST 0
\\d\+ POP_JUMP_IF_TRUE \\d\+
\\d\+ LOAD_FAST 0
\\d\+ SETUP_WITH \\d\+
\\d\+ POP_TOP
\\d\+ LOAD_DEREF 14
\\d\+ POP_TOP
\\d\+ LOAD_CONST_NONE
\\d\+ WITH_CLEANUP
\\d\+ END_FINALLY
\\d\+ LOAD_CONST_SMALL_INT 1
\\d\+ STORE_DEREF 16
\\d\+ LOAD_FAST_N 16
\\d\+ MAKE_CLOSURE \.\+ 1
\\d\+ STORE_FAST 13
\\d\+ LOAD_CONST_SMALL_INT 0
\\d\+ LOAD_CONST_NONE
\\d\+ IMPORT_NAME 'a'
\\d\+ STORE_FAST 0
\\d\+ LOAD_CONST_SMALL_INT 0
\\d\+ LOAD_CONST_STRING 'b'
\\d\+ BUILD_TUPLE 1
\\d\+ IMPORT_NAME 'a'
\\d\+ IMPORT_FROM 'b'
\\d\+ STORE_DEREF 14
\\d\+ POP_TOP
\\d\+ RAISE_LAST
\\d\+ LOAD_CONST_SMALL_INT 1
\\d\+ RAISE_OBJ
\\d\+ LOAD_CONST_NONE
\\d\+ RETURN_VALUE
\\d\+ LOAD_CONST_SMALL_INT 1
\\d\+ RETURN_VALUE
File cmdline/cmd_showbc.py, code block 'f' (descriptor: \.\+, bytecode @\.\+ bytes)
Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
########
\.\+rg names:
(N_STATE 22)
(N_EXC_STACK 0)
bc=0 line=1
########
bc=\\d\+ line=133
00 LOAD_CONST_SMALL_INT 1
01 DUP_TOP
02 STORE_FAST 0
03 DUP_TOP
04 STORE_FAST 1
05 DUP_TOP
06 STORE_FAST 2
07 DUP_TOP
08 STORE_FAST 3
09 DUP_TOP
10 STORE_FAST 4
11 DUP_TOP
12 STORE_FAST 5
13 DUP_TOP
14 STORE_FAST 6
15 DUP_TOP
16 STORE_FAST 7
17 DUP_TOP
18 STORE_FAST 8
19 STORE_FAST 9
20 LOAD_CONST_SMALL_INT 2
21 DUP_TOP
22 STORE_FAST 10
23 DUP_TOP
24 STORE_FAST 11
25 DUP_TOP
26 STORE_FAST 12
27 DUP_TOP
28 STORE_FAST 13
29 DUP_TOP
30 STORE_FAST 14
31 DUP_TOP
32 STORE_FAST 15
33 DUP_TOP
34 STORE_FAST_N 16
36 DUP_TOP
37 STORE_FAST_N 17
39 DUP_TOP
40 STORE_FAST_N 18
42 STORE_FAST_N 19
44 LOAD_FAST 9
45 LOAD_FAST_N 19
47 BINARY_OP 27 __add__
48 POP_TOP
49 LOAD_CONST_NONE
50 RETURN_VALUE
File cmdline/cmd_showbc.py, code block 'f' (descriptor: \.\+, bytecode @\.\+ bytes)
Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
########
\.\+63
arg names: a
(N_STATE 5)
(N_EXC_STACK 0)
(INIT_CELL 0)
########
bc=\\d\+ line=139
00 LOAD_CONST_SMALL_INT 2
01 BUILD_TUPLE 1
03 LOAD_NULL
04 LOAD_FAST 0
05 MAKE_CLOSURE_DEFARGS \.\+ 1
\\d\+ STORE_FAST 1
\\d\+ LOAD_CONST_NONE
\\d\+ RETURN_VALUE
File cmdline/cmd_showbc.py, code block 'f' (descriptor: \.\+, bytecode @\.\+ bytes)
Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
########
\.\+63
arg names:
(N_STATE 2)
(N_EXC_STACK 0)
bc=0 line=1
bc=0 line=144
bc=3 line=145
bc=6 line=146
00 LOAD_CONST_NONE
01 YIELD_VALUE
02 POP_TOP
03 LOAD_CONST_SMALL_INT 1
04 YIELD_VALUE
05 POP_TOP
06 LOAD_CONST_SMALL_INT 1
07 GET_ITER
08 LOAD_CONST_NONE
09 YIELD_FROM
10 POP_TOP
11 LOAD_CONST_NONE
12 RETURN_VALUE
File cmdline/cmd_showbc.py, code block 'Class' (descriptor: \.\+, bytecode @\.\+ bytes)
Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
########
\.\+63
arg names:
(N_STATE 1)
(N_EXC_STACK 0)
bc=0 line=1
########
all: Remove MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE. This commit removes all parts of code associated with the existing MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE optimisation option, including the -mcache-lookup-bc option to mpy-cross. This feature originally provided a significant performance boost for Unix, but wasn't able to be enabled for MCU targets (due to frozen bytecode), and added significant extra complexity to generating and distributing .mpy files. The equivalent performance gain is now provided by the combination of MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE (which has been enabled on the unix port in the previous commit). It's hard to provide precise performance numbers, but tests have been run on a wide variety of architectures (x86-64, ARM Cortex, Aarch64, RISC-V, xtensa) and they all generally agree on the qualitative improvements seen by the combination of MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE. For example, on a "quiet" Linux x64 environment (i3-5010U @ 2.10GHz) the change from CACHE_MAP_LOOKUP_IN_BYTECODE, to LOAD_ATTR_FAST_PATH combined with MAP_LOOKUP_CACHE is: diff of scores (higher is better) N=2000 M=2000 bccache -> attrmapcache diff diff% (error%) bm_chaos.py 13742.56 -> 13905.67 : +163.11 = +1.187% (+/-3.75%) bm_fannkuch.py 60.13 -> 61.34 : +1.21 = +2.012% (+/-2.11%) bm_fft.py 113083.20 -> 114793.68 : +1710.48 = +1.513% (+/-1.57%) bm_float.py 256552.80 -> 243908.29 : -12644.51 = -4.929% (+/-1.90%) bm_hexiom.py 521.93 -> 625.41 : +103.48 = +19.826% (+/-0.40%) bm_nqueens.py 197544.25 -> 217713.12 : +20168.87 = +10.210% (+/-3.01%) bm_pidigits.py 8072.98 -> 8198.75 : +125.77 = +1.558% (+/-3.22%) misc_aes.py 17283.45 -> 16480.52 : -802.93 = -4.646% (+/-0.82%) misc_mandel.py 99083.99 -> 128939.84 : +29855.85 = +30.132% (+/-5.88%) misc_pystone.py 83860.10 -> 82592.56 : -1267.54 = -1.511% (+/-2.27%) misc_raytrace.py 21490.40 -> 22227.23 : +736.83 = +3.429% (+/-1.88%) This shows that the new optimisations are at least as good as the existing inline-bytecode-caching, and are sometimes much better (because the new ones apply caching to a wider variety of map lookups). The new optimisations can also benefit code generated by the native emitter, because they apply to the runtime rather than the generated code. The improvement for the native emitter when LOAD_ATTR_FAST_PATH and MAP_LOOKUP_CACHE are enabled is (same Linux environment as above): diff of scores (higher is better) N=2000 M=2000 native -> nat-attrmapcache diff diff% (error%) bm_chaos.py 14130.62 -> 15464.68 : +1334.06 = +9.441% (+/-7.11%) bm_fannkuch.py 74.96 -> 76.16 : +1.20 = +1.601% (+/-1.80%) bm_fft.py 166682.99 -> 168221.86 : +1538.87 = +0.923% (+/-4.20%) bm_float.py 233415.23 -> 265524.90 : +32109.67 = +13.756% (+/-2.57%) bm_hexiom.py 628.59 -> 734.17 : +105.58 = +16.796% (+/-1.39%) bm_nqueens.py 225418.44 -> 232926.45 : +7508.01 = +3.331% (+/-3.10%) bm_pidigits.py 6322.00 -> 6379.52 : +57.52 = +0.910% (+/-5.62%) misc_aes.py 20670.10 -> 27223.18 : +6553.08 = +31.703% (+/-1.56%) misc_mandel.py 138221.11 -> 152014.01 : +13792.90 = +9.979% (+/-2.46%) misc_pystone.py 85032.14 -> 105681.44 : +20649.30 = +24.284% (+/-2.25%) misc_raytrace.py 19800.01 -> 23350.73 : +3550.72 = +17.933% (+/-2.79%) In summary, compared to MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE, the new MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE options: - are simpler; - take less code size; - are faster (generally); - work with code generated by the native emitter; - can be used on embedded targets with a small and constant RAM overhead; - allow the same .mpy bytecode to run on all targets. See #7680 for further discussion. And see also #7653 for a discussion about simplifying mpy-cross options. Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
2021-09-05 22:28:06 -04:00
bc=12 line=150
00 LOAD_NAME __name__
03 STORE_NAME __module__
06 LOAD_CONST_STRING 'Class'
09 STORE_NAME __qualname__
12 LOAD_CONST_NONE
13 RETURN_VALUE
File cmdline/cmd_showbc.py, code block 'f' (descriptor: \.\+, bytecode @\.\+ bytes)
Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
########
\.\+63
arg names: self
(N_STATE 4)
(N_EXC_STACK 0)
bc=0 line=1
bc=0 line=157
all: Remove MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE. This commit removes all parts of code associated with the existing MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE optimisation option, including the -mcache-lookup-bc option to mpy-cross. This feature originally provided a significant performance boost for Unix, but wasn't able to be enabled for MCU targets (due to frozen bytecode), and added significant extra complexity to generating and distributing .mpy files. The equivalent performance gain is now provided by the combination of MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE (which has been enabled on the unix port in the previous commit). It's hard to provide precise performance numbers, but tests have been run on a wide variety of architectures (x86-64, ARM Cortex, Aarch64, RISC-V, xtensa) and they all generally agree on the qualitative improvements seen by the combination of MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE. For example, on a "quiet" Linux x64 environment (i3-5010U @ 2.10GHz) the change from CACHE_MAP_LOOKUP_IN_BYTECODE, to LOAD_ATTR_FAST_PATH combined with MAP_LOOKUP_CACHE is: diff of scores (higher is better) N=2000 M=2000 bccache -> attrmapcache diff diff% (error%) bm_chaos.py 13742.56 -> 13905.67 : +163.11 = +1.187% (+/-3.75%) bm_fannkuch.py 60.13 -> 61.34 : +1.21 = +2.012% (+/-2.11%) bm_fft.py 113083.20 -> 114793.68 : +1710.48 = +1.513% (+/-1.57%) bm_float.py 256552.80 -> 243908.29 : -12644.51 = -4.929% (+/-1.90%) bm_hexiom.py 521.93 -> 625.41 : +103.48 = +19.826% (+/-0.40%) bm_nqueens.py 197544.25 -> 217713.12 : +20168.87 = +10.210% (+/-3.01%) bm_pidigits.py 8072.98 -> 8198.75 : +125.77 = +1.558% (+/-3.22%) misc_aes.py 17283.45 -> 16480.52 : -802.93 = -4.646% (+/-0.82%) misc_mandel.py 99083.99 -> 128939.84 : +29855.85 = +30.132% (+/-5.88%) misc_pystone.py 83860.10 -> 82592.56 : -1267.54 = -1.511% (+/-2.27%) misc_raytrace.py 21490.40 -> 22227.23 : +736.83 = +3.429% (+/-1.88%) This shows that the new optimisations are at least as good as the existing inline-bytecode-caching, and are sometimes much better (because the new ones apply caching to a wider variety of map lookups). The new optimisations can also benefit code generated by the native emitter, because they apply to the runtime rather than the generated code. The improvement for the native emitter when LOAD_ATTR_FAST_PATH and MAP_LOOKUP_CACHE are enabled is (same Linux environment as above): diff of scores (higher is better) N=2000 M=2000 native -> nat-attrmapcache diff diff% (error%) bm_chaos.py 14130.62 -> 15464.68 : +1334.06 = +9.441% (+/-7.11%) bm_fannkuch.py 74.96 -> 76.16 : +1.20 = +1.601% (+/-1.80%) bm_fft.py 166682.99 -> 168221.86 : +1538.87 = +0.923% (+/-4.20%) bm_float.py 233415.23 -> 265524.90 : +32109.67 = +13.756% (+/-2.57%) bm_hexiom.py 628.59 -> 734.17 : +105.58 = +16.796% (+/-1.39%) bm_nqueens.py 225418.44 -> 232926.45 : +7508.01 = +3.331% (+/-3.10%) bm_pidigits.py 6322.00 -> 6379.52 : +57.52 = +0.910% (+/-5.62%) misc_aes.py 20670.10 -> 27223.18 : +6553.08 = +31.703% (+/-1.56%) misc_mandel.py 138221.11 -> 152014.01 : +13792.90 = +9.979% (+/-2.46%) misc_pystone.py 85032.14 -> 105681.44 : +20649.30 = +24.284% (+/-2.25%) misc_raytrace.py 19800.01 -> 23350.73 : +3550.72 = +17.933% (+/-2.79%) In summary, compared to MICROPY_OPT_CACHE_MAP_LOOKUP_IN_BYTECODE, the new MICROPY_OPT_LOAD_ATTR_FAST_PATH and MICROPY_OPT_MAP_LOOKUP_CACHE options: - are simpler; - take less code size; - are faster (generally); - work with code generated by the native emitter; - can be used on embedded targets with a small and constant RAM overhead; - allow the same .mpy bytecode to run on all targets. See #7680 for further discussion. And see also #7653 for a discussion about simplifying mpy-cross options. Signed-off-by: Jim Mussared <jim.mussared@gmail.com>
2021-09-05 22:28:06 -04:00
00 LOAD_GLOBAL super
\\d\+ LOAD_GLOBAL __class__
\\d\+ LOAD_FAST 0
\\d\+ LOAD_SUPER_METHOD f
\\d\+ CALL_METHOD n=0 nkw=0
\\d\+ POP_TOP
\\d\+ LOAD_CONST_NONE
\\d\+ RETURN_VALUE
File cmdline/cmd_showbc.py, code block '<genexpr>' (descriptor: \.\+, bytecode @\.\+ bytes)
Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
########
\.\+63
arg names: * * *
(N_STATE 9)
(N_EXC_STACK 0)
bc=0 line=1
bc=0 line=60
########
00 LOAD_NULL
01 LOAD_FAST 2
02 LOAD_NULL
03 LOAD_NULL
04 FOR_ITER 20
07 STORE_FAST 3
08 LOAD_DEREF 1
10 POP_JUMP_IF_FALSE 4
13 LOAD_DEREF 0
15 YIELD_VALUE
16 POP_TOP
17 JUMP 4
20 LOAD_CONST_NONE
21 RETURN_VALUE
File cmdline/cmd_showbc.py, code block '<listcomp>' (descriptor: \.\+, bytecode @\.\+ bytes)
Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
########
\.\+63
arg names: * * *
(N_STATE 10)
(N_EXC_STACK 0)
bc=0 line=1
bc=0 line=61
########
00 BUILD_LIST 0
02 LOAD_FAST 2
03 GET_ITER_STACK
04 FOR_ITER 20
07 STORE_FAST 3
08 LOAD_DEREF 1
10 POP_JUMP_IF_FALSE 4
13 LOAD_DEREF 0
15 STORE_COMP 20
17 JUMP 4
20 RETURN_VALUE
File cmdline/cmd_showbc.py, code block '<dictcomp>' (descriptor: \.\+, bytecode @\.\+ bytes)
Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
########
\.\+63
arg names: * * *
(N_STATE 11)
(N_EXC_STACK 0)
bc=0 line=1
########
00 BUILD_MAP 0
02 LOAD_FAST 2
03 GET_ITER_STACK
04 FOR_ITER 22
07 STORE_FAST 3
08 LOAD_DEREF 1
10 POP_JUMP_IF_FALSE 4
13 LOAD_DEREF 0
15 LOAD_DEREF 0
17 STORE_COMP 25
19 JUMP 4
22 RETURN_VALUE
File cmdline/cmd_showbc.py, code block 'closure' (descriptor: \.\+, bytecode @\.\+ bytes)
Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
########
\.\+63
arg names: *
(N_STATE 4)
(N_EXC_STACK 0)
bc=0 line=1
########
bc=\\d\+ line=114
00 LOAD_DEREF 0
02 LOAD_CONST_SMALL_INT 1
03 BINARY_OP 27 __add__
04 STORE_FAST 1
05 LOAD_CONST_SMALL_INT 1
06 STORE_DEREF 0
08 DELETE_DEREF 0
10 LOAD_CONST_NONE
11 RETURN_VALUE
File cmdline/cmd_showbc.py, code block 'f' (descriptor: \.\+, bytecode @\.\+ bytes)
Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
########
\.\+63
arg names: * b
(N_STATE 4)
(N_EXC_STACK 0)
bc=0 line=1
########
bc=\\d\+ line=140
00 LOAD_FAST 1
01 LOAD_DEREF 0
03 BINARY_OP 27 __add__
04 RETURN_VALUE
mem: total=\\d\+, current=\\d\+, peak=\\d\+
stack: \\d\+ out of \\d\+
GC: total: \\d\+, used: \\d\+, free: \\d\+
No. of 1-blocks: \\d\+, 2-blocks: \\d\+, max blk sz: \\d\+, max free sz: \\d\+