Nicolas
|
2e48fe5ec2
|
"variable" type renamed to "value"
|
2025-12-06 18:09:25 +01:00 |
Nicolas
|
19fc403d37
|
autograd updated with get_all_dag_edges_between to increase speed
|
2025-12-06 15:14:06 +01:00 |
Nicolas
|
a21970de79
|
type hint fixes
|
2025-12-06 15:13:28 +01:00 |
Nicolas
|
d526c5ddc0
|
added get_all_dag_edges_between function
|
2025-12-06 15:12:49 +01:00 |
Nicolas Kruse
|
959d80b082
|
shape property added and __getitem__ of matrix extended
|
2025-12-05 08:28:02 +01:00 |
Nicolas Kruse
|
da92aa9e2c
|
eye function added for creating matrices
|
2025-12-04 22:38:52 +01:00 |
Nicolas Kruse
|
5daa54fafb
|
code style adjustments
|
2025-12-04 18:19:33 +01:00 |
Nicolas Kruse
|
ebb4abc5d3
|
type hints revised
|
2025-12-04 18:18:29 +01:00 |
Nicolas Kruse
|
61dc29e68b
|
added auto grad feature "grad(..)"
|
2025-12-03 17:30:38 +01:00 |
Nicolas Kruse
|
c5048980c2
|
added optimization for + 0, - 0 and * 1 operations
|
2025-12-03 17:28:49 +01:00 |
Nicolas Kruse
|
a30ee12d0f
|
fixed get_all_dag_edges prevent it emitting edges multiple time
|
2025-12-03 17:27:29 +01:00 |
Nicolas Kruse
|
cc5582ae73
|
replaced list type by Sequence to improve type hinting
|
2025-12-02 16:59:14 +01:00 |
Nicolas Kruse
|
d2df1dd3fb
|
cp.sign and cp.relu added to _math.py
|
2025-12-02 16:57:22 +01:00 |
Nicolas Kruse
|
9f77ef5642
|
optimization for float/int operations added
|
2025-12-01 16:44:49 +01:00 |
Nicolas Kruse
|
8fe51a2e45
|
benchmark script added
|
2025-12-01 08:29:39 +01:00 |
Nicolas Kruse
|
e14092a2c3
|
matrix class added
|
2025-12-01 00:13:38 +01:00 |
Nicolas Kruse
|
29f5a26278
|
helping functions separated for vectors and matrices
|
2025-12-01 00:13:08 +01:00 |
Nicolas Kruse
|
bb4472eccb
|
variable type hints updated
|
2025-12-01 00:12:22 +01:00 |
Nicolas Kruse
|
a8eeea874b
|
median, mean and argsort added for vector class
|
2025-11-27 17:21:33 +01:00 |
Nicolas Kruse
|
054ce6d507
|
__iter__ type annotation fixed for vector class
|
2025-11-27 17:20:46 +01:00 |
Nicolas Kruse
|
89e8efb864
|
read_value function extended to work with vectors
|
2025-11-27 17:20:07 +01:00 |
Nicolas Kruse
|
ad78c4089c
|
type annotations for iif function fixed
|
2025-11-27 17:19:25 +01:00 |
Nicolas Kruse
|
99a880861a
|
typing fixed, variable[bool] replaced by variable[int]
|
2025-11-27 12:50:53 +01:00 |
Nicolas
|
77924a5655
|
Aux functions are now copied with there whole section to alow for branches without relocation inside of sections.
|
2025-11-24 16:40:37 +01:00 |
Nicolas Kruse
|
1a062760f7
|
detect_process_arch() updated to work for armv7
|
2025-11-24 16:40:37 +01:00 |
Nicolas
|
4c4e83a18e
|
R_ARM_JUMP24 & R_ARM_CALL patching added
|
2025-11-24 16:40:37 +01:00 |
Nicolas
|
d5786d2957
|
addend retrieval removed from patch_arm32_abs (moved to pelfy)
|
2025-11-24 16:40:37 +01:00 |
Nicolas
|
3b1c8f0e72
|
detection for different arm architectures added
|
2025-11-24 16:40:37 +01:00 |
Nicolas
|
96a1a153d4
|
PATCH_OBJECT_ARM32_ABS patching added to support 32 bit ARM
|
2025-11-24 16:40:37 +01:00 |
Nicolas Kruse
|
4049928139
|
vector functions extended
|
2025-11-19 10:01:30 +01:00 |
Nicolas Kruse
|
ebb3b89e69
|
math functions extended for vector. switched acos function.
|
2025-11-17 08:33:29 +01:00 |
Nicolas Kruse
|
700d46e404
|
Compiler performance improved, caching added for get_stencil_code function
|
2025-11-14 22:47:59 +01:00 |
Nicolas
|
20a8dcd1a2
|
Single argument functions do not need a dummy argument anymore
|
2025-11-14 17:28:05 +01:00 |
Nicolas
|
0c2f2812db
|
caching for stencil_database.get_relocations added
|
2025-11-14 16:23:51 +01:00 |
Nicolas
|
ab589449a3
|
Added _ADD_ABS_LO12_NC relocation for ARM; fixed addend for R_386_PC32 and R_386_32; fixed get_return_function_type and strip_function to work with stencils that do the passing on call not at the end
|
2025-11-14 11:12:36 +01:00 |
Nicolas
|
0f464d74a5
|
ENTRY_POINT output in runner fixed
|
2025-11-14 11:12:36 +01:00 |
Nicolas
|
b037e4e3fb
|
get_42 test function updated
|
2025-11-14 11:12:36 +01:00 |
Nicolas
|
0242e32713
|
Patching offset fixed in Patch aux function loop, debug prints removed
|
2025-11-14 11:12:36 +01:00 |
Nicolas
|
ee4d22265c
|
python pow function fixed and changed to musl implementation
|
2025-11-14 11:12:36 +01:00 |
Nicolas
|
5b77ce236f
|
atan2 python function fixed
|
2025-11-14 11:12:36 +01:00 |
Nicolas
|
8413eecdd4
|
debugging printfs added
|
2025-11-14 11:12:36 +01:00 |
Nicolas
|
6416df1335
|
allowing branching in stencils by patching all passing-on-jump instructions and stripping only the last jump instruction if its the last instruction in the function
|
2025-11-11 15:54:11 +01:00 |
Nicolas
|
f345d02120
|
fixed aux function alignment in get_aux_function_mem_layout
|
2025-11-11 09:11:12 +01:00 |
Nicolas
|
b8613f3d98
|
alignment of section fixed
|
2025-11-10 20:08:17 +01:00 |
Nicolas
|
9f211dfda3
|
selection of default architecture changed from system to current process (new function detect_process_arch added)
|
2025-11-10 20:06:48 +01:00 |
Nicolas
|
aa0c75afc6
|
x64 32 bit support added
|
2025-11-09 16:28:39 +01:00 |
Nicolas
|
b6242e12d7
|
index error in the aarch64 fix solved
|
2025-11-07 16:37:20 +01:00 |
Nicolas
|
7c77c42b80
|
issue with wrong results on aarch64 fixed, by guarding registers for the write op
|
2025-11-07 16:37:20 +01:00 |
Nicolas Kruse
|
8572d44bf9
|
tests for inverse trig and log functions added
|
2025-11-07 16:37:20 +01:00 |
Nicolas Kruse
|
d1935a34f8
|
inverse trig functions and log, exp and pow function added
|
2025-11-07 16:37:20 +01:00 |