cumo 0.1.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +7 -0
- data/.gitignore +27 -0
- data/.travis.yml +5 -0
- data/3rd_party/mkmf-cu/.gitignore +36 -0
- data/3rd_party/mkmf-cu/Gemfile +3 -0
- data/3rd_party/mkmf-cu/LICENSE +21 -0
- data/3rd_party/mkmf-cu/README.md +36 -0
- data/3rd_party/mkmf-cu/Rakefile +11 -0
- data/3rd_party/mkmf-cu/bin/mkmf-cu-nvcc +4 -0
- data/3rd_party/mkmf-cu/lib/mkmf-cu.rb +32 -0
- data/3rd_party/mkmf-cu/lib/mkmf-cu/cli.rb +80 -0
- data/3rd_party/mkmf-cu/lib/mkmf-cu/nvcc.rb +157 -0
- data/3rd_party/mkmf-cu/mkmf-cu.gemspec +16 -0
- data/3rd_party/mkmf-cu/test/test_mkmf-cu.rb +67 -0
- data/CODE_OF_CONDUCT.md +46 -0
- data/Gemfile +8 -0
- data/LICENSE.txt +82 -0
- data/README.md +252 -0
- data/Rakefile +43 -0
- data/bench/broadcast_fp32.rb +138 -0
- data/bench/cumo_bench.rb +193 -0
- data/bench/numo_bench.rb +138 -0
- data/bench/reduction_fp32.rb +117 -0
- data/bin/console +14 -0
- data/bin/setup +8 -0
- data/cumo.gemspec +32 -0
- data/ext/cumo/cuda/cublas.c +278 -0
- data/ext/cumo/cuda/driver.c +421 -0
- data/ext/cumo/cuda/memory_pool.cpp +185 -0
- data/ext/cumo/cuda/memory_pool_impl.cpp +308 -0
- data/ext/cumo/cuda/memory_pool_impl.hpp +370 -0
- data/ext/cumo/cuda/memory_pool_impl_test.cpp +554 -0
- data/ext/cumo/cuda/nvrtc.c +207 -0
- data/ext/cumo/cuda/runtime.c +167 -0
- data/ext/cumo/cumo.c +148 -0
- data/ext/cumo/depend.erb +58 -0
- data/ext/cumo/extconf.rb +179 -0
- data/ext/cumo/include/cumo.h +25 -0
- data/ext/cumo/include/cumo/compat.h +23 -0
- data/ext/cumo/include/cumo/cuda/cublas.h +153 -0
- data/ext/cumo/include/cumo/cuda/cumo_thrust.hpp +187 -0
- data/ext/cumo/include/cumo/cuda/cumo_thrust_complex.hpp +79 -0
- data/ext/cumo/include/cumo/cuda/driver.h +22 -0
- data/ext/cumo/include/cumo/cuda/memory_pool.h +28 -0
- data/ext/cumo/include/cumo/cuda/nvrtc.h +22 -0
- data/ext/cumo/include/cumo/cuda/runtime.h +40 -0
- data/ext/cumo/include/cumo/indexer.h +238 -0
- data/ext/cumo/include/cumo/intern.h +142 -0
- data/ext/cumo/include/cumo/intern_fwd.h +38 -0
- data/ext/cumo/include/cumo/intern_kernel.h +6 -0
- data/ext/cumo/include/cumo/narray.h +429 -0
- data/ext/cumo/include/cumo/narray_kernel.h +149 -0
- data/ext/cumo/include/cumo/ndloop.h +95 -0
- data/ext/cumo/include/cumo/reduce_kernel.h +126 -0
- data/ext/cumo/include/cumo/template.h +158 -0
- data/ext/cumo/include/cumo/template_kernel.h +77 -0
- data/ext/cumo/include/cumo/types/bit.h +40 -0
- data/ext/cumo/include/cumo/types/bit_kernel.h +34 -0
- data/ext/cumo/include/cumo/types/complex.h +402 -0
- data/ext/cumo/include/cumo/types/complex_kernel.h +414 -0
- data/ext/cumo/include/cumo/types/complex_macro.h +382 -0
- data/ext/cumo/include/cumo/types/complex_macro_kernel.h +186 -0
- data/ext/cumo/include/cumo/types/dcomplex.h +46 -0
- data/ext/cumo/include/cumo/types/dcomplex_kernel.h +13 -0
- data/ext/cumo/include/cumo/types/dfloat.h +47 -0
- data/ext/cumo/include/cumo/types/dfloat_kernel.h +14 -0
- data/ext/cumo/include/cumo/types/float_def.h +34 -0
- data/ext/cumo/include/cumo/types/float_def_kernel.h +39 -0
- data/ext/cumo/include/cumo/types/float_macro.h +191 -0
- data/ext/cumo/include/cumo/types/float_macro_kernel.h +158 -0
- data/ext/cumo/include/cumo/types/int16.h +24 -0
- data/ext/cumo/include/cumo/types/int16_kernel.h +23 -0
- data/ext/cumo/include/cumo/types/int32.h +24 -0
- data/ext/cumo/include/cumo/types/int32_kernel.h +19 -0
- data/ext/cumo/include/cumo/types/int64.h +24 -0
- data/ext/cumo/include/cumo/types/int64_kernel.h +19 -0
- data/ext/cumo/include/cumo/types/int8.h +24 -0
- data/ext/cumo/include/cumo/types/int8_kernel.h +19 -0
- data/ext/cumo/include/cumo/types/int_macro.h +67 -0
- data/ext/cumo/include/cumo/types/int_macro_kernel.h +48 -0
- data/ext/cumo/include/cumo/types/real_accum.h +486 -0
- data/ext/cumo/include/cumo/types/real_accum_kernel.h +101 -0
- data/ext/cumo/include/cumo/types/robj_macro.h +80 -0
- data/ext/cumo/include/cumo/types/robj_macro_kernel.h +0 -0
- data/ext/cumo/include/cumo/types/robject.h +27 -0
- data/ext/cumo/include/cumo/types/robject_kernel.h +7 -0
- data/ext/cumo/include/cumo/types/scomplex.h +46 -0
- data/ext/cumo/include/cumo/types/scomplex_kernel.h +13 -0
- data/ext/cumo/include/cumo/types/sfloat.h +48 -0
- data/ext/cumo/include/cumo/types/sfloat_kernel.h +14 -0
- data/ext/cumo/include/cumo/types/uint16.h +25 -0
- data/ext/cumo/include/cumo/types/uint16_kernel.h +20 -0
- data/ext/cumo/include/cumo/types/uint32.h +25 -0
- data/ext/cumo/include/cumo/types/uint32_kernel.h +20 -0
- data/ext/cumo/include/cumo/types/uint64.h +25 -0
- data/ext/cumo/include/cumo/types/uint64_kernel.h +20 -0
- data/ext/cumo/include/cumo/types/uint8.h +25 -0
- data/ext/cumo/include/cumo/types/uint8_kernel.h +20 -0
- data/ext/cumo/include/cumo/types/uint_macro.h +58 -0
- data/ext/cumo/include/cumo/types/uint_macro_kernel.h +38 -0
- data/ext/cumo/include/cumo/types/xint_macro.h +169 -0
- data/ext/cumo/include/cumo/types/xint_macro_kernel.h +88 -0
- data/ext/cumo/narray/SFMT-params.h +97 -0
- data/ext/cumo/narray/SFMT-params19937.h +46 -0
- data/ext/cumo/narray/SFMT.c +620 -0
- data/ext/cumo/narray/SFMT.h +167 -0
- data/ext/cumo/narray/array.c +638 -0
- data/ext/cumo/narray/data.c +961 -0
- data/ext/cumo/narray/gen/cogen.rb +56 -0
- data/ext/cumo/narray/gen/cogen_kernel.rb +58 -0
- data/ext/cumo/narray/gen/def/bit.rb +37 -0
- data/ext/cumo/narray/gen/def/dcomplex.rb +39 -0
- data/ext/cumo/narray/gen/def/dfloat.rb +37 -0
- data/ext/cumo/narray/gen/def/int16.rb +36 -0
- data/ext/cumo/narray/gen/def/int32.rb +36 -0
- data/ext/cumo/narray/gen/def/int64.rb +36 -0
- data/ext/cumo/narray/gen/def/int8.rb +36 -0
- data/ext/cumo/narray/gen/def/robject.rb +37 -0
- data/ext/cumo/narray/gen/def/scomplex.rb +39 -0
- data/ext/cumo/narray/gen/def/sfloat.rb +37 -0
- data/ext/cumo/narray/gen/def/uint16.rb +36 -0
- data/ext/cumo/narray/gen/def/uint32.rb +36 -0
- data/ext/cumo/narray/gen/def/uint64.rb +36 -0
- data/ext/cumo/narray/gen/def/uint8.rb +36 -0
- data/ext/cumo/narray/gen/erbpp2.rb +346 -0
- data/ext/cumo/narray/gen/narray_def.rb +268 -0
- data/ext/cumo/narray/gen/spec.rb +425 -0
- data/ext/cumo/narray/gen/tmpl/accum.c +86 -0
- data/ext/cumo/narray/gen/tmpl/accum_binary.c +121 -0
- data/ext/cumo/narray/gen/tmpl/accum_binary_kernel.cu +61 -0
- data/ext/cumo/narray/gen/tmpl/accum_index.c +119 -0
- data/ext/cumo/narray/gen/tmpl/accum_index_kernel.cu +66 -0
- data/ext/cumo/narray/gen/tmpl/accum_kernel.cu +12 -0
- data/ext/cumo/narray/gen/tmpl/alloc_func.c +107 -0
- data/ext/cumo/narray/gen/tmpl/allocate.c +37 -0
- data/ext/cumo/narray/gen/tmpl/aref.c +66 -0
- data/ext/cumo/narray/gen/tmpl/aref_cpu.c +50 -0
- data/ext/cumo/narray/gen/tmpl/aset.c +56 -0
- data/ext/cumo/narray/gen/tmpl/binary.c +162 -0
- data/ext/cumo/narray/gen/tmpl/binary2.c +70 -0
- data/ext/cumo/narray/gen/tmpl/binary2_kernel.cu +15 -0
- data/ext/cumo/narray/gen/tmpl/binary_kernel.cu +31 -0
- data/ext/cumo/narray/gen/tmpl/binary_s.c +45 -0
- data/ext/cumo/narray/gen/tmpl/binary_s_kernel.cu +15 -0
- data/ext/cumo/narray/gen/tmpl/bincount.c +181 -0
- data/ext/cumo/narray/gen/tmpl/cast.c +44 -0
- data/ext/cumo/narray/gen/tmpl/cast_array.c +13 -0
- data/ext/cumo/narray/gen/tmpl/class.c +9 -0
- data/ext/cumo/narray/gen/tmpl/class_kernel.cu +6 -0
- data/ext/cumo/narray/gen/tmpl/clip.c +121 -0
- data/ext/cumo/narray/gen/tmpl/coerce_cast.c +10 -0
- data/ext/cumo/narray/gen/tmpl/complex_accum_kernel.cu +129 -0
- data/ext/cumo/narray/gen/tmpl/cond_binary.c +68 -0
- data/ext/cumo/narray/gen/tmpl/cond_binary_kernel.cu +18 -0
- data/ext/cumo/narray/gen/tmpl/cond_unary.c +46 -0
- data/ext/cumo/narray/gen/tmpl/cum.c +50 -0
- data/ext/cumo/narray/gen/tmpl/each.c +47 -0
- data/ext/cumo/narray/gen/tmpl/each_with_index.c +70 -0
- data/ext/cumo/narray/gen/tmpl/ewcomp.c +79 -0
- data/ext/cumo/narray/gen/tmpl/ewcomp_kernel.cu +19 -0
- data/ext/cumo/narray/gen/tmpl/extract.c +22 -0
- data/ext/cumo/narray/gen/tmpl/extract_cpu.c +26 -0
- data/ext/cumo/narray/gen/tmpl/extract_data.c +53 -0
- data/ext/cumo/narray/gen/tmpl/eye.c +105 -0
- data/ext/cumo/narray/gen/tmpl/eye_kernel.cu +19 -0
- data/ext/cumo/narray/gen/tmpl/fill.c +52 -0
- data/ext/cumo/narray/gen/tmpl/fill_kernel.cu +29 -0
- data/ext/cumo/narray/gen/tmpl/float_accum_kernel.cu +106 -0
- data/ext/cumo/narray/gen/tmpl/format.c +62 -0
- data/ext/cumo/narray/gen/tmpl/format_to_a.c +49 -0
- data/ext/cumo/narray/gen/tmpl/frexp.c +38 -0
- data/ext/cumo/narray/gen/tmpl/gemm.c +203 -0
- data/ext/cumo/narray/gen/tmpl/init_class.c +20 -0
- data/ext/cumo/narray/gen/tmpl/init_module.c +12 -0
- data/ext/cumo/narray/gen/tmpl/inspect.c +21 -0
- data/ext/cumo/narray/gen/tmpl/lib.c +50 -0
- data/ext/cumo/narray/gen/tmpl/lib_kernel.cu +24 -0
- data/ext/cumo/narray/gen/tmpl/logseq.c +102 -0
- data/ext/cumo/narray/gen/tmpl/logseq_kernel.cu +31 -0
- data/ext/cumo/narray/gen/tmpl/map_with_index.c +98 -0
- data/ext/cumo/narray/gen/tmpl/median.c +66 -0
- data/ext/cumo/narray/gen/tmpl/minmax.c +47 -0
- data/ext/cumo/narray/gen/tmpl/module.c +9 -0
- data/ext/cumo/narray/gen/tmpl/module_kernel.cu +1 -0
- data/ext/cumo/narray/gen/tmpl/new_dim0.c +15 -0
- data/ext/cumo/narray/gen/tmpl/new_dim0_kernel.cu +8 -0
- data/ext/cumo/narray/gen/tmpl/poly.c +50 -0
- data/ext/cumo/narray/gen/tmpl/pow.c +97 -0
- data/ext/cumo/narray/gen/tmpl/pow_kernel.cu +29 -0
- data/ext/cumo/narray/gen/tmpl/powint.c +17 -0
- data/ext/cumo/narray/gen/tmpl/qsort.c +212 -0
- data/ext/cumo/narray/gen/tmpl/rand.c +168 -0
- data/ext/cumo/narray/gen/tmpl/rand_norm.c +121 -0
- data/ext/cumo/narray/gen/tmpl/real_accum_kernel.cu +75 -0
- data/ext/cumo/narray/gen/tmpl/seq.c +112 -0
- data/ext/cumo/narray/gen/tmpl/seq_kernel.cu +43 -0
- data/ext/cumo/narray/gen/tmpl/set2.c +57 -0
- data/ext/cumo/narray/gen/tmpl/sort.c +48 -0
- data/ext/cumo/narray/gen/tmpl/sort_index.c +111 -0
- data/ext/cumo/narray/gen/tmpl/store.c +41 -0
- data/ext/cumo/narray/gen/tmpl/store_array.c +187 -0
- data/ext/cumo/narray/gen/tmpl/store_array_kernel.cu +58 -0
- data/ext/cumo/narray/gen/tmpl/store_bit.c +86 -0
- data/ext/cumo/narray/gen/tmpl/store_bit_kernel.cu +66 -0
- data/ext/cumo/narray/gen/tmpl/store_from.c +81 -0
- data/ext/cumo/narray/gen/tmpl/store_from_kernel.cu +58 -0
- data/ext/cumo/narray/gen/tmpl/store_kernel.cu +3 -0
- data/ext/cumo/narray/gen/tmpl/store_numeric.c +9 -0
- data/ext/cumo/narray/gen/tmpl/to_a.c +43 -0
- data/ext/cumo/narray/gen/tmpl/unary.c +132 -0
- data/ext/cumo/narray/gen/tmpl/unary2.c +60 -0
- data/ext/cumo/narray/gen/tmpl/unary_kernel.cu +72 -0
- data/ext/cumo/narray/gen/tmpl/unary_ret2.c +34 -0
- data/ext/cumo/narray/gen/tmpl/unary_s.c +86 -0
- data/ext/cumo/narray/gen/tmpl/unary_s_kernel.cu +58 -0
- data/ext/cumo/narray/gen/tmpl_bit/allocate.c +24 -0
- data/ext/cumo/narray/gen/tmpl_bit/aref.c +54 -0
- data/ext/cumo/narray/gen/tmpl_bit/aref_cpu.c +57 -0
- data/ext/cumo/narray/gen/tmpl_bit/aset.c +56 -0
- data/ext/cumo/narray/gen/tmpl_bit/binary.c +98 -0
- data/ext/cumo/narray/gen/tmpl_bit/bit_count.c +64 -0
- data/ext/cumo/narray/gen/tmpl_bit/bit_count_cpu.c +88 -0
- data/ext/cumo/narray/gen/tmpl_bit/bit_count_kernel.cu +76 -0
- data/ext/cumo/narray/gen/tmpl_bit/bit_reduce.c +133 -0
- data/ext/cumo/narray/gen/tmpl_bit/each.c +48 -0
- data/ext/cumo/narray/gen/tmpl_bit/each_with_index.c +70 -0
- data/ext/cumo/narray/gen/tmpl_bit/extract.c +30 -0
- data/ext/cumo/narray/gen/tmpl_bit/extract_cpu.c +29 -0
- data/ext/cumo/narray/gen/tmpl_bit/fill.c +69 -0
- data/ext/cumo/narray/gen/tmpl_bit/format.c +64 -0
- data/ext/cumo/narray/gen/tmpl_bit/format_to_a.c +51 -0
- data/ext/cumo/narray/gen/tmpl_bit/inspect.c +21 -0
- data/ext/cumo/narray/gen/tmpl_bit/mask.c +136 -0
- data/ext/cumo/narray/gen/tmpl_bit/none_p.c +14 -0
- data/ext/cumo/narray/gen/tmpl_bit/store_array.c +108 -0
- data/ext/cumo/narray/gen/tmpl_bit/store_bit.c +70 -0
- data/ext/cumo/narray/gen/tmpl_bit/store_from.c +60 -0
- data/ext/cumo/narray/gen/tmpl_bit/to_a.c +47 -0
- data/ext/cumo/narray/gen/tmpl_bit/unary.c +81 -0
- data/ext/cumo/narray/gen/tmpl_bit/where.c +90 -0
- data/ext/cumo/narray/gen/tmpl_bit/where2.c +95 -0
- data/ext/cumo/narray/index.c +880 -0
- data/ext/cumo/narray/kwargs.c +153 -0
- data/ext/cumo/narray/math.c +142 -0
- data/ext/cumo/narray/narray.c +1948 -0
- data/ext/cumo/narray/ndloop.c +2105 -0
- data/ext/cumo/narray/rand.c +45 -0
- data/ext/cumo/narray/step.c +474 -0
- data/ext/cumo/narray/struct.c +886 -0
- data/lib/cumo.rb +3 -0
- data/lib/cumo/cuda.rb +11 -0
- data/lib/cumo/cuda/compile_error.rb +36 -0
- data/lib/cumo/cuda/compiler.rb +161 -0
- data/lib/cumo/cuda/device.rb +47 -0
- data/lib/cumo/cuda/link_state.rb +31 -0
- data/lib/cumo/cuda/module.rb +40 -0
- data/lib/cumo/cuda/nvrtc_program.rb +27 -0
- data/lib/cumo/linalg.rb +12 -0
- data/lib/cumo/narray.rb +2 -0
- data/lib/cumo/narray/extra.rb +1278 -0
- data/lib/erbpp.rb +294 -0
- data/lib/erbpp/line_number.rb +137 -0
- data/lib/erbpp/narray_def.rb +381 -0
- data/numo-narray-version +1 -0
- data/run.gdb +7 -0
- metadata +353 -0
@@ -0,0 +1,158 @@
|
|
1
|
+
#ifndef CUMO_FLOAT_MACRO_KERNEL_H
|
2
|
+
#define CUMO_FLOAT_MACRO_KERNEL_H
|
3
|
+
|
4
|
+
#include "float_def_kernel.h"
|
5
|
+
|
6
|
+
extern double round(double);
|
7
|
+
extern double log2(double);
|
8
|
+
extern double exp2(double);
|
9
|
+
#ifdef HAVE_EXP10
|
10
|
+
extern double exp10(double);
|
11
|
+
#else
|
12
|
+
extern double pow(double, double);
|
13
|
+
#endif
|
14
|
+
|
15
|
+
#define m_zero 0.0
|
16
|
+
#define m_one 1.0
|
17
|
+
|
18
|
+
//#define m_num_to_data(x) NUM2DBL(x)
|
19
|
+
//#define m_data_to_num(x) rb_float_new(x)
|
20
|
+
|
21
|
+
#define m_from_double(x) (x)
|
22
|
+
#define m_from_real(x) (x)
|
23
|
+
#define m_from_sint(x) (x)
|
24
|
+
#define m_from_int32(x) (x)
|
25
|
+
#define m_from_int64(x) (x)
|
26
|
+
#define m_from_uint32(x) (x)
|
27
|
+
#define m_from_uint64(x) (x)
|
28
|
+
|
29
|
+
#define m_add(x,y) ((x)+(y))
|
30
|
+
#define m_sub(x,y) ((x)-(y))
|
31
|
+
#define m_mul(x,y) ((x)*(y))
|
32
|
+
#define m_div(x,y) ((x)/(y))
|
33
|
+
#define m_div_check(x,y) ((y)==0)
|
34
|
+
#define m_mod(x,y) fmod(x,y)
|
35
|
+
#define m_divmod(x,y,a,b) {a=(x)/(y); b=m_mod(x,y);}
|
36
|
+
#define m_pow(x,y) pow(x,y)
|
37
|
+
#define m_pow_int(x,y) pow_int(x,y)
|
38
|
+
|
39
|
+
#define m_abs(x) fabs(x)
|
40
|
+
#define m_minus(x) (-(x))
|
41
|
+
#define m_reciprocal(x) (1/(x))
|
42
|
+
#define m_square(x) ((x)*(x))
|
43
|
+
#define m_floor(x) floor(x)
|
44
|
+
#define m_round(x) round(x)
|
45
|
+
#define m_ceil(x) ceil(x)
|
46
|
+
#define m_trunc(x) trunc(x)
|
47
|
+
#define m_rint(x) rint(x)
|
48
|
+
#define m_sign(x) (((x)==0) ? 0.0:(((x)>0) ? 1.0:(((x)<0) ? -1.0:(x))))
|
49
|
+
#define m_copysign(x,y) copysign(x,y)
|
50
|
+
#define m_signbit(x) signbit(x)
|
51
|
+
#define m_modf(x,y,z) {double d; y=modf(x,&d); z=d;}
|
52
|
+
|
53
|
+
#define m_eq(x,y) ((x)==(y))
|
54
|
+
#define m_ne(x,y) ((x)!=(y))
|
55
|
+
#define m_gt(x,y) ((x)>(y))
|
56
|
+
#define m_ge(x,y) ((x)>=(y))
|
57
|
+
#define m_lt(x,y) ((x)<(y))
|
58
|
+
#define m_le(x,y) ((x)<=(y))
|
59
|
+
|
60
|
+
#define m_isnan(x) isnan(x)
|
61
|
+
#define m_isinf(x) isinf(x)
|
62
|
+
#define m_isposinf(x) (isinf(x) && signbit(x)==0)
|
63
|
+
#define m_isneginf(x) (isinf(x) && signbit(x))
|
64
|
+
#define m_isfinite(x) isfinite(x)
|
65
|
+
|
66
|
+
#define m_mulsum_init 0
|
67
|
+
|
68
|
+
#define m_sprintf(s,x) sprintf(s,"%g",x)
|
69
|
+
|
70
|
+
/*
|
71
|
+
#define cmp_prnan(a,b) \
|
72
|
+
((qsort_cast(a)==qsort_cast(b)) ? 0 : \
|
73
|
+
(qsort_cast(a) > qsort_cast(b)) ? 1 : -1)
|
74
|
+
|
75
|
+
#define cmp_ignan(a,b) \
|
76
|
+
(m_isnan(qsort_cast(a)) ? (m_isnan(qsort_cast(b)) ? 0 : 1) : \
|
77
|
+
(m_isnan(qsort_cast(b)) ? -1 : \
|
78
|
+
((qsort_cast(a)==qsort_cast(b)) ? 0 : \
|
79
|
+
(qsort_cast(a) > qsort_cast(b)) ? 1 : -1)))
|
80
|
+
|
81
|
+
#define cmpgt_prnan(a,b) \
|
82
|
+
(qsort_cast(a) > qsort_cast(b))
|
83
|
+
|
84
|
+
#define cmpgt_ignan(a,b) \
|
85
|
+
((m_isnan(qsort_cast(a)) && !m_isnan(qsort_cast(b))) || \
|
86
|
+
(qsort_cast(a) > qsort_cast(b)))
|
87
|
+
*/
|
88
|
+
|
89
|
+
#define m_sqrt(x) sqrt(x)
|
90
|
+
#define m_cbrt(x) cbrt(x)
|
91
|
+
#define m_log(x) log(x)
|
92
|
+
#define m_log2(x) log2(x)
|
93
|
+
#define m_log10(x) log10(x)
|
94
|
+
#define m_exp(x) exp(x)
|
95
|
+
#define m_exp2(x) exp2(x)
|
96
|
+
#ifdef HAVE_EXP10
|
97
|
+
#define m_exp10(x) exp10(x)
|
98
|
+
#else
|
99
|
+
#define m_exp10(x) pow(10, x)
|
100
|
+
#endif
|
101
|
+
#define m_expm1(x) expm1(x)
|
102
|
+
#define m_log1p(x) log1p(x)
|
103
|
+
|
104
|
+
#define m_sin(x) sin(x)
|
105
|
+
#define m_cos(x) cos(x)
|
106
|
+
#define m_tan(x) tan(x)
|
107
|
+
#define m_asin(x) asin(x)
|
108
|
+
#define m_acos(x) acos(x)
|
109
|
+
#define m_atan(x) atan(x)
|
110
|
+
#define m_sinh(x) sinh(x)
|
111
|
+
#define m_cosh(x) cosh(x)
|
112
|
+
#define m_tanh(x) tanh(x)
|
113
|
+
#define m_asinh(x) asinh(x)
|
114
|
+
#define m_acosh(x) acosh(x)
|
115
|
+
#define m_atanh(x) atanh(x)
|
116
|
+
#define m_atan2(x,y) atan2(x,y)
|
117
|
+
#define m_hypot(x,y) hypot(x,y)
|
118
|
+
#define m_sinc(x) (sin(x)/(x))
|
119
|
+
|
120
|
+
#define m_erf(x) erf(x)
|
121
|
+
#define m_erfc(x) erfc(x)
|
122
|
+
#define m_ldexp(x,y) ldexp(x,y)
|
123
|
+
#define m_frexp(x,exp) frexp(x,exp)
|
124
|
+
|
125
|
+
/* only internal use (called by pow_int) */
|
126
|
+
__host__ __device__ static inline dtype pow_positive_int(dtype x, int p)
|
127
|
+
{
|
128
|
+
dtype r=1;
|
129
|
+
switch(p) {
|
130
|
+
case 0: return 1;
|
131
|
+
case 1: return x;
|
132
|
+
case 2: return x*x;
|
133
|
+
case 3: return x*x*x;
|
134
|
+
case 4: x=x*x; return x*x;
|
135
|
+
}
|
136
|
+
if (p>64) return pow(x,p);
|
137
|
+
while (p) {
|
138
|
+
if (p&1) r *= x;
|
139
|
+
x *= x;
|
140
|
+
p >>= 1;
|
141
|
+
}
|
142
|
+
return r;
|
143
|
+
}
|
144
|
+
|
145
|
+
__host__ __device__ static inline dtype pow_int(dtype x, int p)
|
146
|
+
{
|
147
|
+
if (p<0) return 1/pow_positive_int(x, -p);
|
148
|
+
return pow_positive_int(x, p);
|
149
|
+
}
|
150
|
+
|
151
|
+
__host__ __device__ static inline dtype f_seq(dtype x, dtype y, double c)
|
152
|
+
{
|
153
|
+
return x + y * c;
|
154
|
+
}
|
155
|
+
|
156
|
+
#include "real_accum_kernel.h"
|
157
|
+
|
158
|
+
#endif // CUMO_FLOAT_MACRO_KERNEL_H
|
@@ -0,0 +1,24 @@
|
|
1
|
+
typedef int16_t dtype;
|
2
|
+
typedef int16_t rtype;
|
3
|
+
#define cT cumo_cInt16
|
4
|
+
#define cRT cT
|
5
|
+
|
6
|
+
#define m_num_to_data(x) ((dtype)NUM2INT(x))
|
7
|
+
#define m_data_to_num(x) INT2NUM((int)(x))
|
8
|
+
#define m_extract(x) INT2NUM((int)*(dtype*)(x))
|
9
|
+
#define m_sprintf(s,x) sprintf(s,"%d",(int)(x))
|
10
|
+
|
11
|
+
#ifndef INT16_MIN
|
12
|
+
#define INT16_MIN (-32767-1)
|
13
|
+
#endif
|
14
|
+
#ifndef INT16_MAX
|
15
|
+
#define INT16_MAX (32767)
|
16
|
+
#endif
|
17
|
+
|
18
|
+
#define DATA_MIN INT16_MIN
|
19
|
+
#define DATA_MAX INT16_MAX
|
20
|
+
|
21
|
+
#define M_MIN m_data_to_num(INT16_MIN)
|
22
|
+
#define M_MAX m_data_to_num(INT16_MAX)
|
23
|
+
|
24
|
+
#include "int_macro.h"
|
@@ -0,0 +1,23 @@
|
|
1
|
+
#ifndef CUMO_INT16_KERNEL_H
|
2
|
+
#define CUMO_INT16_KERNEL_H
|
3
|
+
|
4
|
+
typedef int16_t dtype;
|
5
|
+
typedef int16_t rtype;
|
6
|
+
#define cT cumo_cInt16
|
7
|
+
#define cRT cT
|
8
|
+
|
9
|
+
#define m_sprintf(s,x) sprintf(s,"%d",(int)(x))
|
10
|
+
|
11
|
+
#ifndef INT16_MIN
|
12
|
+
#define INT16_MIN (-32767-1)
|
13
|
+
#endif
|
14
|
+
#ifndef INT16_MAX
|
15
|
+
#define INT16_MAX (32767)
|
16
|
+
#endif
|
17
|
+
|
18
|
+
#define DATA_MIN INT16_MIN
|
19
|
+
#define DATA_MAX INT16_MAX
|
20
|
+
|
21
|
+
#include "int_macro_kernel.h"
|
22
|
+
|
23
|
+
#endif // CUMO_INT16_KERNEL_H
|
@@ -0,0 +1,24 @@
|
|
1
|
+
typedef int32_t dtype;
|
2
|
+
typedef int32_t rtype;
|
3
|
+
#define cT cumo_cInt32
|
4
|
+
#define cRT cT
|
5
|
+
|
6
|
+
#define m_num_to_data(x) ((dtype)NUM2INT32(x))
|
7
|
+
#define m_data_to_num(x) INT322NUM((int32_t)(x))
|
8
|
+
#define m_extract(x) INT322NUM((int32_t)*(dtype*)(x))
|
9
|
+
#define m_sprintf(s,x) sprintf(s,"%"PRId32,(int32_t)(x))
|
10
|
+
|
11
|
+
#ifndef INT32_MIN
|
12
|
+
#define INT32_MIN (-2147483647-1)
|
13
|
+
#endif
|
14
|
+
#ifndef INT32_MAX
|
15
|
+
#define INT32_MAX (2147483647)
|
16
|
+
#endif
|
17
|
+
|
18
|
+
#define DATA_MIN INT32_MIN
|
19
|
+
#define DATA_MAX INT32_MAX
|
20
|
+
|
21
|
+
#define M_MIN m_data_to_num(INT32_MIN)
|
22
|
+
#define M_MAX m_data_to_num(INT32_MAX)
|
23
|
+
|
24
|
+
#include "int_macro.h"
|
@@ -0,0 +1,19 @@
|
|
1
|
+
#ifndef CUMO_INT32_KERNEL_H
|
2
|
+
#define CUMO_INT32_KERNEL_H
|
3
|
+
|
4
|
+
typedef int32_t dtype;
|
5
|
+
typedef int32_t rtype;
|
6
|
+
|
7
|
+
#ifndef INT32_MIN
|
8
|
+
#define INT32_MIN (-2147483647-1)
|
9
|
+
#endif
|
10
|
+
#ifndef INT32_MAX
|
11
|
+
#define INT32_MAX (2147483647)
|
12
|
+
#endif
|
13
|
+
|
14
|
+
#define DATA_MIN INT32_MIN
|
15
|
+
#define DATA_MAX INT32_MAX
|
16
|
+
|
17
|
+
#include "int_macro_kernel.h"
|
18
|
+
|
19
|
+
#endif // CUMO_INT32_KERNEL_H
|
@@ -0,0 +1,24 @@
|
|
1
|
+
typedef int64_t dtype;
|
2
|
+
typedef int64_t rtype;
|
3
|
+
#define cT cumo_cInt64
|
4
|
+
#define cRT cT
|
5
|
+
|
6
|
+
#define m_num_to_data(x) ((dtype)NUM2INT64(x))
|
7
|
+
#define m_data_to_num(x) INT642NUM((int64_t)(x))
|
8
|
+
#define m_extract(x) INT642NUM((int64_t)*(dtype*)(x))
|
9
|
+
#define m_sprintf(s,x) sprintf(s,"%"PRId64,(int64_t)(x))
|
10
|
+
|
11
|
+
#ifndef INT64_MIN
|
12
|
+
#define INT64_MIN (-9223372036854775807l-1)
|
13
|
+
#endif
|
14
|
+
#ifndef INT64_MAX
|
15
|
+
#define INT64_MAX (9223372036854775807l)
|
16
|
+
#endif
|
17
|
+
|
18
|
+
#define DATA_MIN INT64_MIN
|
19
|
+
#define DATA_MAX INT64_MAX
|
20
|
+
|
21
|
+
#define M_MIN m_data_to_num(INT64_MIN)
|
22
|
+
#define M_MAX m_data_to_num(INT64_MAX)
|
23
|
+
|
24
|
+
#include "int_macro.h"
|
@@ -0,0 +1,19 @@
|
|
1
|
+
#ifndef CUMO_INT64_KERNEL_H
|
2
|
+
#define CUMO_INT64_KERNEL_H
|
3
|
+
|
4
|
+
typedef int64_t dtype;
|
5
|
+
typedef int64_t rtype;
|
6
|
+
|
7
|
+
#ifndef INT64_MIN
|
8
|
+
#define INT64_MIN (-9223372036854775807l-1)
|
9
|
+
#endif
|
10
|
+
#ifndef INT64_MAX
|
11
|
+
#define INT64_MAX (9223372036854775807l)
|
12
|
+
#endif
|
13
|
+
|
14
|
+
#define DATA_MIN INT64_MIN
|
15
|
+
#define DATA_MAX INT64_MAX
|
16
|
+
|
17
|
+
#include "int_macro_kernel.h"
|
18
|
+
|
19
|
+
#endif // CUMO_INT64_KERNEL_H
|
@@ -0,0 +1,24 @@
|
|
1
|
+
typedef int8_t dtype;
|
2
|
+
typedef int8_t rtype;
|
3
|
+
#define cT cumo_cInt8
|
4
|
+
#define cRT cT
|
5
|
+
|
6
|
+
#define m_num_to_data(x) ((dtype)NUM2INT(x))
|
7
|
+
#define m_data_to_num(x) INT2NUM((int)(x))
|
8
|
+
#define m_extract(x) INT2NUM((int)*(dtype*)(x))
|
9
|
+
#define m_sprintf(s,x) sprintf(s,"%d",(int)(x))
|
10
|
+
|
11
|
+
#ifndef INT8_MIN
|
12
|
+
#define INT8_MIN (-127-1)
|
13
|
+
#endif
|
14
|
+
#ifndef INT8_MAX
|
15
|
+
#define INT8_MAX (127)
|
16
|
+
#endif
|
17
|
+
|
18
|
+
#define DATA_MIN INT8_MIN
|
19
|
+
#define DATA_MAX INT8_MAX
|
20
|
+
|
21
|
+
#define M_MIN INT2FIX(INT8_MIN)
|
22
|
+
#define M_MAX INT2FIX(INT8_MAX)
|
23
|
+
|
24
|
+
#include "int_macro.h"
|
@@ -0,0 +1,19 @@
|
|
1
|
+
#ifndef CUMO_INT8_KERNEL_H
|
2
|
+
#define CUMO_INT8_KERNEL_H
|
3
|
+
|
4
|
+
typedef int8_t dtype;
|
5
|
+
typedef int8_t rtype;
|
6
|
+
|
7
|
+
#ifndef INT8_MIN
|
8
|
+
#define INT8_MIN (-127-1)
|
9
|
+
#endif
|
10
|
+
#ifndef INT8_MAX
|
11
|
+
#define INT8_MAX (127)
|
12
|
+
#endif
|
13
|
+
|
14
|
+
#define DATA_MIN INT8_MIN
|
15
|
+
#define DATA_MAX INT8_MAX
|
16
|
+
|
17
|
+
#include "int_macro_kernel.h"
|
18
|
+
|
19
|
+
#endif // CUMO_INT8_KERNEL_H
|
@@ -0,0 +1,67 @@
|
|
1
|
+
#include "xint_macro.h"
|
2
|
+
|
3
|
+
#define m_sign(x) (((x)==0) ? 0 : (((x)>0) ? 1 : -1))
|
4
|
+
|
5
|
+
static inline dtype m_abs(dtype x) {
|
6
|
+
if (x==DATA_MIN) {
|
7
|
+
rb_raise(nary_eValueError, "cannot convert the minimum integer");
|
8
|
+
}
|
9
|
+
return (x<0)?-x:x;
|
10
|
+
}
|
11
|
+
|
12
|
+
static inline dtype int_reciprocal(dtype x) {
|
13
|
+
switch (x) {
|
14
|
+
case 1:
|
15
|
+
return 1;
|
16
|
+
case -1:
|
17
|
+
return -1;
|
18
|
+
case 0:
|
19
|
+
rb_raise(rb_eZeroDivError, "divided by 0");
|
20
|
+
default:
|
21
|
+
return 0;
|
22
|
+
}
|
23
|
+
}
|
24
|
+
|
25
|
+
/*
|
26
|
+
static dtype pow_int(dtype x, int p)
|
27
|
+
{
|
28
|
+
dtype r = m_one;
|
29
|
+
switch(p) {
|
30
|
+
case 0: return 1;
|
31
|
+
case 1: return x;
|
32
|
+
case 2: return x*x;
|
33
|
+
case 3: return x*x*x;
|
34
|
+
}
|
35
|
+
if (p<0) return 0;
|
36
|
+
while (p) {
|
37
|
+
if (p&1) r *= x;
|
38
|
+
x *= x;
|
39
|
+
p >>= 1;
|
40
|
+
}
|
41
|
+
return r;
|
42
|
+
}
|
43
|
+
*/
|
44
|
+
|
45
|
+
static inline int64_t f_sum(size_t n, char *p, ssize_t stride)
|
46
|
+
{
|
47
|
+
int64_t x,y=0;
|
48
|
+
size_t i=n;
|
49
|
+
for (; i--;) {
|
50
|
+
x = *(dtype*)p;
|
51
|
+
y += x;
|
52
|
+
p += stride;
|
53
|
+
}
|
54
|
+
return y;
|
55
|
+
}
|
56
|
+
|
57
|
+
static inline int64_t f_prod(size_t n, char *p, ssize_t stride)
|
58
|
+
{
|
59
|
+
int64_t x,y=1;
|
60
|
+
size_t i=n;
|
61
|
+
for (; i--;) {
|
62
|
+
x = *(dtype*)p;
|
63
|
+
y *= x;
|
64
|
+
p += stride;
|
65
|
+
}
|
66
|
+
return y;
|
67
|
+
}
|
@@ -0,0 +1,48 @@
|
|
1
|
+
#ifndef CUMO_INT_MACRO_KERNEL_H
|
2
|
+
#define CUMO_INT_MACRO_KERNEL_H
|
3
|
+
|
4
|
+
#include "xint_macro_kernel.h"
|
5
|
+
|
6
|
+
#define m_sign(x) (((x)==0) ? 0 : (((x)>0) ? 1 : -1))
|
7
|
+
|
8
|
+
__host__ __device__ static inline dtype m_abs(dtype x) {
|
9
|
+
// TODO(sonots): How to handle in CUDA kernel?
|
10
|
+
// if (x==DATA_MIN) {
|
11
|
+
// rb_raise(nary_eValueError, "cannot convert the minimum integer");
|
12
|
+
// }
|
13
|
+
return (x<0)?-x:x;
|
14
|
+
}
|
15
|
+
|
16
|
+
__host__ __device__ static inline dtype int_reciprocal(dtype x) {
|
17
|
+
switch (x) {
|
18
|
+
case 1:
|
19
|
+
return 1;
|
20
|
+
case -1:
|
21
|
+
return -1;
|
22
|
+
case 0:
|
23
|
+
return 0; // as CUDA kernel 1/0 results in 0.
|
24
|
+
//rb_raise(rb_eZeroDivError, "divided by 0");
|
25
|
+
default:
|
26
|
+
return 0;
|
27
|
+
}
|
28
|
+
}
|
29
|
+
|
30
|
+
__device__ static dtype pow_int(dtype x, int p)
|
31
|
+
{
|
32
|
+
dtype r = m_one;
|
33
|
+
switch(p) {
|
34
|
+
case 0: return 1;
|
35
|
+
case 1: return x;
|
36
|
+
case 2: return x*x;
|
37
|
+
case 3: return x*x*x;
|
38
|
+
}
|
39
|
+
if (p<0) return 0;
|
40
|
+
while (p) {
|
41
|
+
if (p&1) r *= x;
|
42
|
+
x *= x;
|
43
|
+
p >>= 1;
|
44
|
+
}
|
45
|
+
return r;
|
46
|
+
}
|
47
|
+
|
48
|
+
#endif // CUMO_INT_MACRO_KERNEL_H
|