search

codezonediitj / adaboost

Implementation of AdaBoost Algorithm
Other
9 stars 16 forks source link

Matrix multiplication using Shared Memory #30

Closed Tanvi141 closed 4 years ago

Tanvi141 commented 4 years ago

References to other Issues or PRs or Relevant literature

Fixes #8

Brief description of what is fixed or changed

Made changes to multiply_gpu method to implement logic of tiling using shared memory. Corner conditions included in case matrix size does not exactly match size of tile.

czgdp1807 commented 4 years ago

Could you test your code using this example?

Tanvi141 commented 4 years ago

Tested the particular example, working properly.

[==========] Running 3 tests from 1 test case.
[----------] Global test environment set-up.
[----------] 3 tests from Cuda
[ RUN      ] Cuda.VectorGPU
[       OK ] Cuda.VectorGPU (62 ms)
[ RUN      ] Cuda.MatrixGPU
[       OK ] Cuda.MatrixGPU (39 ms)
[ RUN      ] Cuda.MatricesGPU
83.000000 63.000000 37.000000 75.000000 [       OK ] Cuda.MatricesGPU (312 ms)
[----------] 3 tests from Cuda (413 ms total)

[----------] Global test environment tear-down
[==========] 3 tests from 1 test case ran. (413 ms total)
[  PASSED  ] 3 tests.

Also updated required changed suggested above

Tanvi141 commented 4 years ago

Hi, made required changes. Please see test_cuda reports;

[==========] Running 3 tests from 1 test case.
[----------] Global test environment set-up.
[----------] 3 tests from Cuda
[ RUN      ] Cuda.VectorGPU
[       OK ] Cuda.VectorGPU (252 ms)
[ RUN      ] Cuda.MatrixGPU
[       OK ] Cuda.MatrixGPU (40 ms)
[ RUN      ] Cuda.MatricesGPU
[       OK ] Cuda.MatricesGPU (342 ms)
[----------] 3 tests from Cuda (634 ms total)

[----------] Global test environment tear-down
[==========] 3 tests from 1 test case ran. (634 ms total)
[  PASSED  ] 3 tests.