X compilebench: Add type annotations to pystone_pyx.pyx

This brings in ~8x speedup with tcc and ~14x speedup for gccO2 compared to what
we already had for pystone.py already compiled by Cython.

However Cython compilation time is 1.7x slower and compilation time for gcc*
also raises significantly. For tcc it is "only" 30% more for compilation, but
still the total tcc compilation time is very very small.

name                    old time/op    new time/op    delta
pycompile+pyimport        13.0ms ± 0%    13.0ms ± 0%       ~     (all equal)
pyimport                  12.0ms ± 0%    12.0ms ± 0%       ~     (all equal)
cython/py                  980ms ± 1%     982ms ± 1%       ~     (p=0.666 n=10+10)
cython/pyx                 983ms ± 1%    1692ms ± 1%    +72.13%  (p=0.000 n=9+10)
gcc/compile+link/py        603ms ± 0%     603ms ± 0%       ~     (p=0.766 n=8+10)
gcc/compile+linkO2/py      2.15s ± 0%     2.15s ± 0%       ~     (p=0.749 n=10+10)
gcc/compile+link/pyx       603ms ± 0%    1106ms ± 0%    +83.57%  (p=0.000 n=9+10)
gcc/compile+linkO2/pyx     2.15s ± 0%     3.65s ± 0%    +69.87%  (p=0.000 n=10+10)
tcc/compile+link/py       36.5ms ± 1%    36.0ms ± 0%     -1.37%  (p=0.022 n=10+9)
tcc/compile+link/pyx      36.7ms ± 2%    47.5ms ± 3%    +29.43%  (p=0.000 n=10+10)

name                    old pystone/s  new pystone/s  delta
pystone/py                  227k ± 2%      228k ± 1%       ~     (p=0.182 n=10+9)
pystone/py/gcc              220k ± 3%      221k ± 4%       ~     (p=0.796 n=10+10)
pystone/py/gccO2            433k ± 2%      431k ± 5%       ~     (p=1.000 n=10+10)
pystone/pyx/gcc             221k ± 1%     1855k ± 1%   +737.67%  (p=0.000 n=9+8)
pystone/pyx/gccO2           435k ± 1%     6239k ± 2%  +1335.52%  (p=0.000 n=10+10)
pystone/py/tcc              219k ± 2%      219k ± 0%       ~     (p=0.315 n=10+10)
pystone/pyx/tcc             222k ± 2%     1826k ± 0%   +723.78%  (p=0.000 n=9+9)

t/compilebench/pystone_pyx.pyx

@@ -40,24 +40,29 @@ from time import clock
 
 __version__ = "1.1"
 
-[Ident1, Ident2, Ident3, Ident4, Ident5] = range(1, 6)
+cdef int Ident1=1, Ident2=2, Ident3=3, Ident4=4, Ident5=5
 
-class Record:
+cdef class Record:
+    cdef Record PtrComp
+    cdef int    Discr
+    cdef int    EnumComp
+    cdef int    IntComp
+    cdef str    StringComp
 
-    def __init__(self, PtrComp = None, Discr = 0, EnumComp = 0,
-                       IntComp = 0, StringComp = 0):
+    def __init__(self, Record PtrComp = None, int Discr = 0, int EnumComp = 0,
+                       int IntComp = 0, str StringComp = None):
         self.PtrComp = PtrComp
         self.Discr = Discr
         self.EnumComp = EnumComp
         self.IntComp = IntComp
         self.StringComp = StringComp
 
-    def copy(self):
+    cdef Record copy(self):
         return Record(self.PtrComp, self.Discr, self.EnumComp,
                       self.IntComp, self.StringComp)
 
-TRUE = 1
-FALSE = 0
+cdef bint TRUE = 1
+cdef bint FALSE = 0
 
 def main(loops=LOOPS):
     benchtime, stones = pystones(loops)
@@ -69,14 +74,21 @@ def main(loops=LOOPS):
 def pystones(loops=LOOPS):
     return Proc0(loops)
 
-IntGlob = 0
-BoolGlob = FALSE
-Char1Glob = '\0'
-Char2Glob = '\0'
-Array1Glob = [0]*51
-Array2Glob = map(lambda x: x[:], [Array1Glob]*51)
-PtrGlb = None
-PtrGlbNext = None
+cdef int    IntGlob = 0
+cdef bint   BoolGlob = FALSE
+cdef char   Char1Glob = '\0'
+cdef char   Char2Glob = '\0'
+
+cdef int[51] _Array1Glob_X
+cdef int[:]  Array1Glob = _Array1Glob_X
+Array1Glob[:] = 0
+
+cdef int[51][51] _Array2Glob_X
+cdef int[:,:]    Array2Glob = _Array2Glob_X
+Array2Glob[:,:] = 0
+
+cdef Record PtrGlb = None
+cdef Record PtrGlbNext = None
 
 def Proc0(loops=LOOPS):
     global IntGlob
@@ -105,6 +117,10 @@ def Proc0(loops=LOOPS):
 
     starttime = clock()
 
+    cdef int  IntLoc1, IntLoc2, IntLoc3
+    cdef int  EnumLoc
+    cdef char CharIndex
+
     for i in range(loops):
         Proc5()
         Proc4()
@@ -121,9 +137,9 @@ def Proc0(loops=LOOPS):
         PtrGlb = Proc1(PtrGlb)
         CharIndex = 'A'
         while CharIndex <= Char2Glob:
-            if EnumLoc == Func1(CharIndex, 'C'):
+            if EnumLoc == Func1(CharIndex, <char>'C'):
                 EnumLoc = Proc6(Ident1)
-            CharIndex = chr(ord(CharIndex)+1)
+            CharIndex = CharIndex+1
         IntLoc3 = IntLoc2 * IntLoc1
         IntLoc2 = IntLoc3 / IntLoc1
         IntLoc2 = 7 * (IntLoc3 - IntLoc2) - IntLoc1
@@ -136,7 +152,7 @@ def Proc0(loops=LOOPS):
         loopsPerBenchtime = (loops / benchtime)
     return benchtime, loopsPerBenchtime
 
-def Proc1(PtrParIn):
+cdef Record Proc1(Record PtrParIn):
     PtrParIn.PtrComp = NextRecord = PtrGlb.copy()
     PtrParIn.IntComp = 5
     NextRecord.IntComp = PtrParIn.IntComp
@@ -152,8 +168,9 @@ def Proc1(PtrParIn):
     NextRecord.PtrComp = None
     return PtrParIn
 
-def Proc2(IntParIO):
-    IntLoc = IntParIO + 10
+cdef int Proc2(int IntParIO):
+    cdef int IntLoc = IntParIO + 10
+    cdef int EnumLoc = 0 # XXX =0: can be used initialized due to bug in original pystone.py
     while 1:
         if Char1Glob == 'A':
             IntLoc = IntLoc - 1
@@ -163,7 +180,7 @@ def Proc2(IntParIO):
             break
     return IntParIO
 
-def Proc3(PtrParOut):
+cdef Record Proc3(Record PtrParOut):
     global IntGlob
 
     if PtrGlb is not None:
@@ -173,21 +190,21 @@ def Proc3(PtrParOut):
     PtrGlb.IntComp = Proc7(10, IntGlob)
     return PtrParOut
 
-def Proc4():
+cdef Proc4():
     global Char2Glob
 
-    BoolLoc = Char1Glob == 'A'
+    cdef bint BoolLoc = (Char1Glob == 'A')
     BoolLoc = BoolLoc or BoolGlob
     Char2Glob = 'B'
 
-def Proc5():
+cdef Proc5():
     global Char1Glob
     global BoolGlob
 
     Char1Glob = 'A'
     BoolGlob = FALSE
 
-def Proc6(EnumParIn):
+cdef int Proc6(int EnumParIn):
     EnumParOut = EnumParIn
     if not Func3(EnumParIn):
         EnumParOut = Ident4
@@ -206,15 +223,15 @@ def Proc6(EnumParIn):
         EnumParOut = Ident3
     return EnumParOut
 
-def Proc7(IntParI1, IntParI2):
+cdef int Proc7(int IntParI1, int IntParI2):
     IntLoc = IntParI1 + 2
     IntParOut = IntParI2 + IntLoc
     return IntParOut
 
-def Proc8(Array1Par, Array2Par, IntParI1, IntParI2):
+cdef Proc8(int[:] Array1Par, int[:,:] Array2Par, int IntParI1, int IntParI2):
     global IntGlob
 
-    IntLoc = IntParI1 + 5
+    cdef int IntLoc = IntParI1 + 5
     Array1Par[IntLoc] = IntParI2
     Array1Par[IntLoc+1] = Array1Par[IntLoc]
     Array1Par[IntLoc+30] = IntLoc
@@ -224,7 +241,7 @@ def Proc8(Array1Par, Array2Par, IntParI1, IntParI2):
     Array2Par[IntLoc+20][IntLoc] = Array1Par[IntLoc]
     IntGlob = 5
 
-def Func1(CharPar1, CharPar2):
+cdef int Func1(char CharPar1, char CharPar2):
     CharLoc1 = CharPar1
     CharLoc2 = CharLoc1
     if CharLoc2 != CharPar2:
@@ -232,10 +249,11 @@ def Func1(CharPar1, CharPar2):
     else:
         return Ident2
 
-def Func2(StrParI1, StrParI2):
-    IntLoc = 1
+cdef bint Func2(StrParI1, StrParI2):
+    cdef int IntLoc = 1
+    cdef char CharLoc
     while IntLoc <= 1:
-        if Func1(StrParI1[IntLoc], StrParI2[IntLoc+1]) == Ident1:
+        if Func1((<char*>StrParI1)[IntLoc], (<char*>StrParI2)[IntLoc+1]) == Ident1:
             CharLoc = 'A'
             IntLoc = IntLoc + 1
     if CharLoc >= 'W' and CharLoc <= 'Z':
@@ -249,7 +267,7 @@ def Func2(StrParI1, StrParI2):
         else:
             return FALSE
 
-def Func3(EnumParIn):
+cdef bint Func3(int EnumParIn):
     EnumLoc = EnumParIn
     if EnumLoc == Ident3: return TRUE
     return FALSE

t/compilebench/y.txt

+Removing pystone.pyc
+Removing pystone_pypyx.so
+Removing pystone_pyx.so
+Benchmarkpycompile+pyimport 1 0.013 s/op
+Benchmarkpycompile+pyimport 1 0.013 s/op
+Benchmarkpycompile+pyimport 1 0.013 s/op
+Benchmarkpycompile+pyimport 1 0.013 s/op
+Benchmarkpycompile+pyimport 1 0.013 s/op
+Benchmarkpycompile+pyimport 1 0.013 s/op
+Benchmarkpycompile+pyimport 1 0.013 s/op
+Benchmarkpycompile+pyimport 1 0.013 s/op
+Benchmarkpycompile+pyimport 1 0.013 s/op
+Benchmarkpycompile+pyimport 1 0.013 s/op
+Benchmarkpyimport 1 0.012 s/op
+Benchmarkpyimport 1 0.012 s/op
+Benchmarkpyimport 1 0.012 s/op
+Benchmarkpyimport 1 0.012 s/op
+Benchmarkpyimport 1 0.012 s/op
+Benchmarkpyimport 1 0.012 s/op
+Benchmarkpyimport 1 0.012 s/op
+Benchmarkpyimport 1 0.012 s/op
+Benchmarkpyimport 1 0.012 s/op
+Benchmarkpyimport 1 0.012 s/op
+Benchmarkcython/py 1 0.992 s/op
+Benchmarkcython/py 1 0.979 s/op
+Benchmarkcython/py 1 0.979 s/op
+Benchmarkcython/py 1 0.987 s/op
+Benchmarkcython/py 1 0.976 s/op
+Benchmarkcython/py 1 0.982 s/op
+Benchmarkcython/py 1 0.981 s/op
+Benchmarkcython/py 1 0.982 s/op
+Benchmarkcython/py 1 0.984 s/op
+Benchmarkcython/py 1 0.974 s/op
+Benchmarkcython/pyx 1 1.689 s/op
+Benchmarkcython/pyx 1 1.690 s/op
+Benchmarkcython/pyx 1 1.693 s/op
+Benchmarkcython/pyx 1 1.686 s/op
+Benchmarkcython/pyx 1 1.684 s/op
+Benchmarkcython/pyx 1 1.701 s/op
+Benchmarkcython/pyx 1 1.692 s/op
+Benchmarkcython/pyx 1 1.699 s/op
+Benchmarkcython/pyx 1 1.688 s/op
+Benchmarkcython/pyx 1 1.693 s/op
+Benchmarkpystone/py 1 230803 pystone/s
+Benchmarkpystone/py 1 227862 pystone/s
+Benchmarkpystone/py 1 228006 pystone/s
+Benchmarkpystone/py 1 226811 pystone/s
+Benchmarkpystone/py 1 227605 pystone/s
+Benchmarkpystone/py 1 228955 pystone/s
+Benchmarkpystone/py 1 228692 pystone/s
+Benchmarkpystone/py 1 228915 pystone/s
+Benchmarkpystone/py 1 224686 pystone/s
+Benchmarkpystone/py 1 228180 pystone/s
+CC=gcc
+Benchmarkgcc/compile+link/py 1 0.601 s/op
+Benchmarkgcc/compile+link/py 1 0.603 s/op
+Benchmarkgcc/compile+link/py 1 0.602 s/op
+Benchmarkgcc/compile+link/py 1 0.605 s/op
+Benchmarkgcc/compile+link/py 1 0.602 s/op
+Benchmarkgcc/compile+link/py 1 0.602 s/op
+Benchmarkgcc/compile+link/py 1 0.605 s/op
+Benchmarkgcc/compile+link/py 1 0.604 s/op
+Benchmarkgcc/compile+link/py 1 0.602 s/op
+Benchmarkgcc/compile+link/py 1 0.605 s/op
+Benchmarkpystone/py/gcc 1 223406 pystone/s
+Benchmarkpystone/py/gcc 1 222159 pystone/s
+Benchmarkpystone/py/gcc 1 224375 pystone/s
+Benchmarkpystone/py/gcc 1 218221 pystone/s
+Benchmarkpystone/py/gcc 1 217240 pystone/s
+Benchmarkpystone/py/gcc 1 221248 pystone/s
+Benchmarkpystone/py/gcc 1 220844 pystone/s
+Benchmarkpystone/py/gcc 1 212688 pystone/s
+Benchmarkpystone/py/gcc 1 225562 pystone/s
+Benchmarkpystone/py/gcc 1 224123 pystone/s
+Benchmarkgcc/compile+linkO2/py 1 2.154 s/op
+Benchmarkgcc/compile+linkO2/py 1 2.155 s/op
+Benchmarkgcc/compile+linkO2/py 1 2.154 s/op
+Benchmarkgcc/compile+linkO2/py 1 2.152 s/op
+Benchmarkgcc/compile+linkO2/py 1 2.154 s/op
+Benchmarkgcc/compile+linkO2/py 1 2.157 s/op
+Benchmarkgcc/compile+linkO2/py 1 2.152 s/op
+Benchmarkgcc/compile+linkO2/py 1 2.151 s/op
+Benchmarkgcc/compile+linkO2/py 1 2.151 s/op
+Benchmarkgcc/compile+linkO2/py 1 2.148 s/op
+Benchmarkpystone/py/gccO2 1 419953 pystone/s
+Benchmarkpystone/py/gccO2 1 439344 pystone/s
+Benchmarkpystone/py/gccO2 1 434461 pystone/s
+Benchmarkpystone/py/gccO2 1 436087 pystone/s
+Benchmarkpystone/py/gccO2 1 425876 pystone/s
+Benchmarkpystone/py/gccO2 1 437032 pystone/s
+Benchmarkpystone/py/gccO2 1 438912 pystone/s
+Benchmarkpystone/py/gccO2 1 439116 pystone/s
+Benchmarkpystone/py/gccO2 1 431831 pystone/s
+Benchmarkpystone/py/gccO2 1 408117 pystone/s
+Benchmarkgcc/compile+link/pyx 1 1.108 s/op
+Benchmarkgcc/compile+link/pyx 1 1.109 s/op
+Benchmarkgcc/compile+link/pyx 1 1.107 s/op
+Benchmarkgcc/compile+link/pyx 1 1.108 s/op
+Benchmarkgcc/compile+link/pyx 1 1.107 s/op
+Benchmarkgcc/compile+link/pyx 1 1.105 s/op
+Benchmarkgcc/compile+link/pyx 1 1.106 s/op
+Benchmarkgcc/compile+link/pyx 1 1.105 s/op
+Benchmarkgcc/compile+link/pyx 1 1.104 s/op
+Benchmarkgcc/compile+link/pyx 1 1.106 s/op
+Benchmarkpystone/pyx/gcc 1 1.84631e+06 pystone/s
+Benchmarkpystone/pyx/gcc 1 1.86937e+06 pystone/s
+Benchmarkpystone/pyx/gcc 1 1.79565e+06 pystone/s
+Benchmarkpystone/pyx/gcc 1 1.84863e+06 pystone/s
+Benchmarkpystone/pyx/gcc 1 1.85832e+06 pystone/s
+Benchmarkpystone/pyx/gcc 1 1.83851e+06 pystone/s
+Benchmarkpystone/pyx/gcc 1 1.8485e+06 pystone/s
+Benchmarkpystone/pyx/gcc 1 1.75482e+06 pystone/s
+Benchmarkpystone/pyx/gcc 1 1.87392e+06 pystone/s
+Benchmarkpystone/pyx/gcc 1 1.85846e+06 pystone/s
+Benchmarkgcc/compile+linkO2/pyx 1 3.647 s/op
+Benchmarkgcc/compile+linkO2/pyx 1 3.658 s/op
+Benchmarkgcc/compile+linkO2/pyx 1 3.657 s/op
+Benchmarkgcc/compile+linkO2/pyx 1 3.655 s/op
+Benchmarkgcc/compile+linkO2/pyx 1 3.654 s/op
+Benchmarkgcc/compile+linkO2/pyx 1 3.648 s/op
+Benchmarkgcc/compile+linkO2/pyx 1 3.649 s/op
+Benchmarkgcc/compile+linkO2/pyx 1 3.672 s/op
+Benchmarkgcc/compile+linkO2/pyx 1 3.648 s/op
+Benchmarkgcc/compile+linkO2/pyx 1 3.660 s/op
+Benchmarkpystone/pyx/gccO2 1 6.29089e+06 pystone/s
+Benchmarkpystone/pyx/gccO2 1 6.27195e+06 pystone/s
+Benchmarkpystone/pyx/gccO2 1 6.2751e+06 pystone/s
+Benchmarkpystone/pyx/gccO2 1 6.24454e+06 pystone/s
+Benchmarkpystone/pyx/gccO2 1 6.27904e+06 pystone/s
+Benchmarkpystone/pyx/gccO2 1 6.222e+06 pystone/s
+Benchmarkpystone/pyx/gccO2 1 6.24376e+06 pystone/s
+Benchmarkpystone/pyx/gccO2 1 6.14402e+06 pystone/s
+Benchmarkpystone/pyx/gccO2 1 6.13422e+06 pystone/s
+Benchmarkpystone/pyx/gccO2 1 6.28852e+06 pystone/s
+CC=tcc
+Benchmarktcc/compile+link/py 1 0.036 s/op
+Benchmarktcc/compile+link/py 1 0.038 s/op
+Benchmarktcc/compile+link/py 1 0.036 s/op
+Benchmarktcc/compile+link/py 1 0.036 s/op
+Benchmarktcc/compile+link/py 1 0.036 s/op
+Benchmarktcc/compile+link/py 1 0.036 s/op
+Benchmarktcc/compile+link/py 1 0.036 s/op
+Benchmarktcc/compile+link/py 1 0.036 s/op
+Benchmarktcc/compile+link/py 1 0.036 s/op
+Benchmarktcc/compile+link/py 1 0.036 s/op
+Benchmarkpystone/py/tcc 1 219283 pystone/s
+Benchmarkpystone/py/tcc 1 218130 pystone/s
+Benchmarkpystone/py/tcc 1 218993 pystone/s
+Benchmarkpystone/py/tcc 1 218747 pystone/s
+Benchmarkpystone/py/tcc 1 217537 pystone/s
+Benchmarkpystone/py/tcc 1 218471 pystone/s
+Benchmarkpystone/py/tcc 1 218670 pystone/s
+Benchmarkpystone/py/tcc 1 218044 pystone/s
+Benchmarkpystone/py/tcc 1 219547 pystone/s
+Benchmarkpystone/py/tcc 1 218365 pystone/s
+Benchmarktcc/compile+link/pyx 1 0.047 s/op
+Benchmarktcc/compile+link/pyx 1 0.048 s/op
+Benchmarktcc/compile+link/pyx 1 0.047 s/op
+Benchmarktcc/compile+link/pyx 1 0.047 s/op
+Benchmarktcc/compile+link/pyx 1 0.049 s/op
+Benchmarktcc/compile+link/pyx 1 0.047 s/op
+Benchmarktcc/compile+link/pyx 1 0.048 s/op
+Benchmarktcc/compile+link/pyx 1 0.047 s/op
+Benchmarktcc/compile+link/pyx 1 0.047 s/op
+Benchmarktcc/compile+link/pyx 1 0.048 s/op
+Benchmarkpystone/pyx/tcc 1 1.82695e+06 pystone/s
+Benchmarkpystone/pyx/tcc 1 1.82715e+06 pystone/s
+Benchmarkpystone/pyx/tcc 1 1.82963e+06 pystone/s
+Benchmarkpystone/pyx/tcc 1 1.82989e+06 pystone/s
+Benchmarkpystone/pyx/tcc 1 1.80343e+06 pystone/s
+Benchmarkpystone/pyx/tcc 1 1.8205e+06 pystone/s
+Benchmarkpystone/pyx/tcc 1 1.81772e+06 pystone/s
+Benchmarkpystone/pyx/tcc 1 1.82615e+06 pystone/s
+Benchmarkpystone/pyx/tcc 1 1.82956e+06 pystone/s
+Benchmarkpystone/pyx/tcc 1 1.83056e+06 pystone/s

[Kirill Smelkov]

.