Dongju Chae [Tue, 4 Feb 2020 09:57:32 +0000 (18:57 +0900)]
[Debian/Dist] Add missing dependency for sub-packages of npu-engine
This commit adds missing dependency for sub-packages of npu-engine.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 4 Feb 2020 07:52:58 +0000 (16:52 +0900)]
[Coverage] Fix unittest coverage script
This commit fixes unittest coverage script to be compatible with gcc-9
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Mon, 3 Feb 2020 06:19:32 +0000 (15:19 +0900)]
[Unittest/Fix] Remove check routines after .return_buffer()
This commit removes check routines after .return_buffer().
We cannot ensure buffer states after .return_buffer() becasue after
this function call, we will lose the ownership of the buffer.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 31 Jan 2020 07:27:57 +0000 (16:27 +0900)]
[Quantization] Do not perform quantization for 1-byte types
This commit makes that quantization is not performed for 1-byte types
such as uint8_t and int8_t because they have the same data size with
DATA_TYPE_SRNPU.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 31 Jan 2020 07:18:46 +0000 (16:18 +0900)]
[FIX] fix meson version dependency
This commit fixes meson version dependency, 0.50
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 7 Jan 2020 04:07:07 +0000 (13:07 +0900)]
[Release] NPU Engine 0.0.6 release
This commit releases a newer version of NPU Engine library, 0.0.6.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Thu, 30 Jan 2020 09:32:44 +0000 (18:32 +0900)]
[Test/Quantization] Add a unittest to test quantization
This commit adds a unittest to test quantization.
If data type is not DATA_TYPE_SRNPU, we need to perform quantization.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Thu, 30 Jan 2020 09:32:24 +0000 (18:32 +0900)]
[Test/DataManip] Add a unittest to test data manipulation
This commit adds a unittest to test data manipulation.
If a depth is not a multiple of WPL_A (i.e., 64), padding is added.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Thu, 30 Jan 2020 08:20:02 +0000 (17:20 +0900)]
[FIX] Fix some functions' descriptions
This comit fixes some functions' descriptions.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Wed, 29 Jan 2020 09:12:38 +0000 (18:12 +0900)]
[Quantization] Implement input/output quantization during data manip
This commit implements input/output quantization during data
manipulation. It requires the intensive verification for calculations.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Wed, 29 Jan 2020 04:53:55 +0000 (13:53 +0900)]
[Test/npubinfmt_v2] Enable testcases for npubinfmt v2
This commit enables testcases for npubinfmt v2. Also, it revises
the existing apptest, example_visa, to support NPU models with
npubinfmt v2.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 28 Jan 2020 07:18:11 +0000 (16:18 +0900)]
[FIX] Remove unnecessary broadcast in callback handling
This commit removes unnecessary broadcast in callback handling.
It may cause a potential deadlock situation.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 28 Jan 2020 04:42:19 +0000 (13:42 +0900)]
[TEST] Change unit/apptest codes to sync with the updated interface.
This commit changes unit/apptest codes to sync with the updated
interface.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 28 Jan 2020 04:41:04 +0000 (13:41 +0900)]
[libnpuhost] Support output_buffers with multiple tensors
This commit changes libnpuhost.h to support output_buffers
to hold multiple tensors' data (like input_tensors).
Also, it revises other related codes for multi tensors and npubinfmt v2.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Wed, 22 Jan 2020 07:29:01 +0000 (16:29 +0900)]
[Fix] Fix warnings caused by the updated gcc version
This commit fixs wraning caused by the updated gcc version.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Mon, 20 Jan 2020 09:47:07 +0000 (18:47 +0900)]
[DataManip] Implement data manipulation for BUFFER_MAPPED
This commit implements data manipulation for BUFFER_MAPPED type.
For other types, it's not efficient for NPU Engine to handle them.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Thu, 16 Jan 2020 07:14:43 +0000 (16:14 +0900)]
[Fix/Log] Omit logs for third-party libraries (e.g., libnpuvision)
This commit omits dirty logs for third-party libraries including
libnpuvision and encoder. Such logs are printed as standrad output, thus
we can redirect them to /dev/null.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Thu, 9 Jan 2020 10:16:05 +0000 (19:16 +0900)]
[Handler] Revise data preparation for multiple tensors for inference.
This commit revises data preparation for multiple tensors by adding new
handler API for feeding input data to NE buffer. It enables to feed
multiple tensors' data to the physically-contiguous NE buffer.
Also, this patch moves the impl of setup_hwmem() to ne-handler.c
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 14 Jan 2020 09:23:28 +0000 (18:23 +0900)]
[AppTest] Revise existing unittests to use allocInputBuffers()
This commit revises existing unittests to use allocInputBuffers().
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Thu, 9 Jan 2020 10:16:05 +0000 (19:16 +0900)]
[libnpuhost] revise buffer allocation/cleanup APIs for users.
This commit revises buffer allocation/cleanup APIs for users.
Existing APIs, allocNPUBuffer() and cleanNPUBuffer(), handle memory
for only a single tensor. However, to support multiple tensors, we
need new pair of APIs to allocate memory for them at the same time
by allocating a single physically-contiguous buffer to cover all
tensors (in case of BUFFER_DMABUF).
So, this patch introduces new APIs for multiple tensors.
- allocInputBuffers() & cleanInputBuffer()
Also, it renames existing APIs for single tensor.
- allocInputBuffer() & cleanInputBUffer()
- allocModelBuffer() & cleanModelBUffer()
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 10 Jan 2020 06:45:01 +0000 (15:45 +0900)]
[libnpuhost/api] add the API to set data info for input/output tensors
This commit adds the API to set data info for input/output tensors.
It defines the data layout/type for tensor layouts, and NPU Engine
will use such information to perform data manipulation and quantization.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Mon, 13 Jan 2020 05:18:52 +0000 (14:18 +0900)]
[ModelGen] Update model generator for newly added metadata info
This commit updates model generator for newly added metadata info,
quantization parameters.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 10 Jan 2020 07:25:05 +0000 (16:25 +0900)]
[npubinfmt/v2] add quantization parameters for each input/output tensor
This commit adds quantization parameters to npubinfmt.h (v2), which will
be used to quantize each input/output activation.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Mon, 13 Jan 2020 04:15:11 +0000 (13:15 +0900)]
[Bug/Ip] Handle non-zero input offset in plugin-comm-ip.c
This commit fixes the ip plugin to handle non-zero input offset.
It's valid only for npubinfmt v1. For npubinfmt v2, other PRs may
resolve this by handling per-tensor data manipulation.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Mon, 13 Jan 2020 04:08:03 +0000 (13:08 +0900)]
[Comm/Api] Add the API to obtain the metadata of target model
This commit adds the API to obtain the metadata of target model.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Mon, 13 Jan 2020 04:04:08 +0000 (13:04 +0900)]
[Testdata] Fix testcase generator to consider non-zero input offset
This commit fixes testcase generator to consider non-zero input offset.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Wed, 18 Dec 2019 07:08:30 +0000 (16:08 +0900)]
[DOC] Revise doc for .gbs.conf to remove repo.extra
This commit revises doc for .gbs.conf to remove repo.extra.
It's not necessary.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 3 Jan 2020 09:15:54 +0000 (18:15 +0900)]
[TESTGEN] Revise testcase generator for npubinfmt v2
This commit revises testcase generator for npubinfmt v2.
It considers non-contiguous tensors as testcase*_v2.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 3 Jan 2020 05:09:28 +0000 (14:09 +0900)]
[MODEL/META] Verify the magiccode in npubinfmt's metadata
This commit verifies the magiccode in npubinfmt's metadata.
Also, it updates related unitttests to use the default magiccode.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Mon, 30 Dec 2019 05:55:58 +0000 (14:55 +0900)]
[TESTGEN] Revise model generator for npubinfmt v2
This commit revises the model generator script to support npubinfmt v2.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Mon, 30 Dec 2019 06:27:38 +0000 (15:27 +0900)]
[BUILD] add the build option for NPU data manipulation
This commit adds the build option for NPU data manipulation.
As HW could provide this feature in the near future, it is enabled by
the build option.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 13 Dec 2019 05:44:43 +0000 (14:44 +0900)]
[TEST/MODELGEN] Sync with the updated npubinfmt.h
This commit fixes a script to generate test models to sync new
npubinfmt.h
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 13 Dec 2019 05:39:17 +0000 (14:39 +0900)]
[NPUBINFMT] Append per-tensor information (e.g., num, dims, emod)
This commit appends per-tensor information such as dimensions, elem
size, emod, etc to 'npubinfmt.h'. This information is valid only for
npubinfmt v2. The previous version (v1) still valid. Also, we need to
make sure to assign its magic code to specify the npubinfmt version.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 17 Dec 2019 06:35:07 +0000 (15:35 +0900)]
[Release] NPU Engine 0.0.5 Release
This commit starts NPU Engine 0.0.5 Release.
After this, 0.0.6 will include npubinfmt v2 and data manipulation.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 17 Dec 2019 06:25:35 +0000 (15:25 +0900)]
[ASR] Revise inference-related SW stack for NPU-ASR
This commit revises all inference-related SW stack for ASR.
It reuses exsiting codes using npu device type, using effective op.
Also, It adds new example apptest for NPU-ASR device.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Mon, 2 Dec 2019 05:19:44 +0000 (14:19 +0900)]
[DOC] Revise doc to describe additional info (ubuntu/testdata)
This commit revises doc to describe additional info related to ubuntu
build and testdata package.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Mon, 2 Dec 2019 05:07:39 +0000 (14:07 +0900)]
[README] Revise README.md to describe 'tools' folder
This commit revises README.md to describe 'tools' folder.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 29 Nov 2019 05:53:58 +0000 (14:53 +0900)]
[ASR] make ASR use a dummy model
This commit makes ASR use a dummy model to utilize existing codes.
In ASR device, registerModel() is not required. But, we can utilize
existing codes using a dummy model.
Also, as ASR can't resize I/O buffers on model registeration, new API
.resize_buffers() is required.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 29 Nov 2019 05:28:58 +0000 (14:28 +0900)]
[TESTDATA-GEN] Fix the input size to include esum sizes for all layers
This commit fixes the input size to include esum sizes for all layers.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Wook Song [Fri, 29 Nov 2019 05:35:11 +0000 (14:35 +0900)]
[Dist/Debian] Create a new package containing test data
This patch creates a new package containing test data separated
from the example package.
Signed-off-by: Wook Song <wook16.song@samsung.com>
MyungJoo Ham [Fri, 29 Nov 2019 00:34:06 +0000 (09:34 +0900)]
Common/Typedef: Update input mode description
There had been some confusion related with inputs for ASR.
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
Wook Song [Thu, 28 Nov 2019 04:18:41 +0000 (13:18 +0900)]
[Dist/Debian] Change prefixes of sub-packages to npu-engine-*
This patch changes the prefixes of sub-packages from libnpuengine- to
npu-engine-.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Thu, 28 Nov 2019 04:16:48 +0000 (13:16 +0900)]
[Meson] Clean code in meson build scripts
This patchs cleans code in meson build scripts.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Dongju Chae [Tue, 26 Nov 2019 11:07:28 +0000 (20:07 +0900)]
[FIX/RUN_SYNC] Fix the timing issue due to early inference completion
This commit fixes the timing issue caused when `run_sync()`'s inference is finished too
early. This patch introduces an extra variable to check whether the callback is
called or not.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 26 Nov 2019 11:00:19 +0000 (20:00 +0900)]
[BUG/HWMEM] setup_hwmem() bug fix: hwmem deactivation on return
This commit fixes the bug, which `setup_hwmem()` has an ambiguous
assumption for hwmem deactivation. Now, `setup_hwmem()` does not
deactivate hwmem, which should be handled in its caller.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 26 Nov 2019 12:53:23 +0000 (21:53 +0900)]
[LOG] Remove unnecessary logs and fix some log typos
This commit removes unnecessary logs and fixes some log typos.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Wed, 27 Nov 2019 07:44:04 +0000 (16:44 +0900)]
[CONF] Remove iniparser's duplicated loads & error messages
This commit removes iniparser's duplicated loads and error messages.
The origianl codes used to call load_conf() multiple times, generating
meaningless error messages.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Wook Song [Wed, 27 Nov 2019 03:46:41 +0000 (12:46 +0900)]
[Debian] Update version to 0.0.3
Update version to 0.0.3
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Wed, 27 Nov 2019 03:43:46 +0000 (12:43 +0900)]
[Dist/Debian] Generate multi-layer data for application test
This patch removes the tarball containing multi-layer data for
application test and makes those data be generated at the build-time.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Wed, 27 Nov 2019 02:07:01 +0000 (11:07 +0900)]
[Dist/Tizen] Generate multi-layer data for application test
This patch removes the tarball containing multi-layer test data and
makes those test data be generated at build time.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Wed, 27 Nov 2019 00:06:56 +0000 (09:06 +0900)]
[Dist/Debian] Gernerate single layer data for application test
This patch removes the tarball containing pre-build single layer model
data for application data. Instead, those test data is generated at
build time by this patch.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Tue, 26 Nov 2019 23:48:34 +0000 (08:48 +0900)]
[Dist/Tizen] Generate single layer data for application test
Instead of tarball, the sigle layer data for application test is
generated at build-time by this patch. In addition, this patch creates a
new package containing those test data.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Tue, 26 Nov 2019 04:28:05 +0000 (13:28 +0900)]
[Dist/Debian] Add omitted npu-engine.ini to the main package
This patch adds the omitted npu-engine.ini file to the main package.
Note that this file is installed in the etc directory under /opt/trinity
when the emulation mode is enabled.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Tue, 26 Nov 2019 04:04:46 +0000 (13:04 +0900)]
[Dist/Debian] Fix a warning about 'badly formatted trailer line'
This patch fixes a warning about 'badly formatted trailer line' in
the changelog file.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Tue, 26 Nov 2019 06:43:30 +0000 (15:43 +0900)]
[Meson] Replace version string of meson with >= 0.42
Since OBS does not provide meson >= 0.50, this patch replaces the
version string of meson with >=0.42.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Dongju Chae [Mon, 25 Nov 2019 07:12:27 +0000 (16:12 +0900)]
[ASR/IP] Revise getnumNPU/ASRdevice() to use common function
This commit revises getnumNPU/ASRdevice() to use common function.
Also, getASRdevice() is implemented in a similar way to the NPU device.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Mon, 25 Nov 2019 07:11:26 +0000 (16:11 +0900)]
[BUILD/ASR] Remove asr-specific codes to reuse existing one
This commit removes asr-specific codes & dependency to reuse existing
one.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Mon, 25 Nov 2019 11:51:49 +0000 (20:51 +0900)]
[TESTDATA] Bug fix on CONVE mod calculation
This commit fixes the bug on CONVE mode calculation.
As CONVE's esum uses the output size, mod calcs also should be matched.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 22 Nov 2019 02:11:41 +0000 (11:11 +0900)]
[RELEASE] Add CHANGES to maintain changelog & make 0.0.2 release
This commit adds CHANGES to maintain changelog and increases the package
version to 0.0.2.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 22 Nov 2019 11:57:24 +0000 (20:57 +0900)]
[TEST] Add ubuntu auto_test & enable example_visa as default
This commit adds ubuntu auto_test and enables example_visa as default.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 22 Nov 2019 10:34:50 +0000 (19:34 +0900)]
[CONF] Revise all source codes which have used test_mode conf
This commit revises all source codes which have used test_mode conf.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 22 Nov 2019 10:13:21 +0000 (19:13 +0900)]
[CONF] Remove test_mode configuration
This commit removes test_mode configuration cuz it's no longer used.
Now, NPU Engine runs with either NPU Emulation (Ubuntu 16.04) or real
NPU device (FastModel)
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 22 Nov 2019 08:03:56 +0000 (17:03 +0900)]
[Valgrind] Revise a way to enable valgrind to remove redundant codes
This commit revises a way to enable valgrind to remove redundant codes.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Wook Song [Fri, 22 Nov 2019 07:22:53 +0000 (16:22 +0900)]
[Tools/GenTestData] Update model_gen tool to use testdata packages
This patch updates the model_gen tool to be possible to use the file
provided by testdata packages from the decoder and libnpuvision.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Fri, 22 Nov 2019 07:32:18 +0000 (16:32 +0900)]
[Tools/GenTestData/Meson] Fix errors while building for real mode
This patch fixes errors while building the test data generation tool for
real mode.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Dongju Chae [Tue, 19 Nov 2019 07:24:19 +0000 (16:24 +0900)]
[Packaging] Remove the libnpuvision dependency for release mode
This commit removes the libnpuvision dependency for release mode.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 22 Nov 2019 04:59:17 +0000 (13:59 +0900)]
[FIX] version check fixup for some packages
This commit fixes the version checking for some packages.
Note that the version checking of gcc is not working because it's meta
package.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Wook Song [Wed, 20 Nov 2019 07:31:55 +0000 (16:31 +0900)]
[Tools/Meson] Include gen-testdata in build tree
This patch adds meson build scripts in order to include gen-testdata in
the build tree.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Wed, 20 Nov 2019 09:06:38 +0000 (18:06 +0900)]
[Dist/Tizen] Exclude tools for test data generation
This patch excludes tools for test data generation from packaging.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Wed, 20 Nov 2019 08:59:25 +0000 (17:59 +0900)]
[Tools/GenTestData] Fix a warning about sign-compare
This patch fixes the sign-compare warning in testdata_gen.cpp.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Wed, 20 Nov 2019 08:08:28 +0000 (17:08 +0900)]
[Tools/GenTestData] Fix a warning about unused-result
This patch fixes the unused-result warning in testdata_gen.cpp.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Wed, 20 Nov 2019 07:52:33 +0000 (16:52 +0900)]
[Tools/GenTestData] Remove unneeded white spaces
This patch removes unneeded white spaces in testdata_gen.cpp.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Wed, 20 Nov 2019 06:55:23 +0000 (15:55 +0900)]
[Test] Move files related to testdata generation to the tools directory
This patch moves files related to testdata generation, which are
orinally placed in ./src/test/testdata, to the tools directory, i.e.,
./tools/gen-testdata.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Thu, 21 Nov 2019 06:48:16 +0000 (15:48 +0900)]
[Dist/Debian] Add omitted build-dependency on libiniparser-dev
This patch adds the omitted build-dependency on libiniparser-dev to the
debian control file.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Dongju Chae [Wed, 20 Nov 2019 04:40:48 +0000 (13:40 +0900)]
[EMUL/BUG] Bug fix: when model_gen.py makes esum data (ESUM vs. CONVE)
This commit fixes the bug when model generator makes esum data, which should have handled differently depending on OPCODE (i.e., ESUM vs. CONVE).
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 19 Nov 2019 07:52:06 +0000 (16:52 +0900)]
[EMUL/BUG] Bug fix: Add a missing line when setting esum data
This commit adds a missing line when setting esum data.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 19 Nov 2019 07:25:45 +0000 (16:25 +0900)]
[CONF] Set the default resv mem size to 0 (Kernel CMA)
This commit change the default resv mem size to 0 (Kernel CMA).
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 19 Nov 2019 07:17:32 +0000 (16:17 +0900)]
[FIX/GEM] fix the size checking in gem_create() to prevent overflow
This commit fixes the size checking in gem_create(), which caused size
overflow due to alignment.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Wook Song [Mon, 18 Nov 2019 07:14:38 +0000 (16:14 +0900)]
[Core/NPU/Emul] Fill the function body for CONVE emulation
This patch fills the function body for emulation of the CONVE
instruction and enables the emulation using it.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Mon, 18 Nov 2019 05:01:57 +0000 (14:01 +0900)]
[Core/NPU/Emul] Fill the function body for ReLU emulation
This function fills the function body for emulation of the ReLU
instruction. Using this function, the emulation for ReLU is enabled.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Mon, 18 Nov 2019 04:42:20 +0000 (13:42 +0900)]
[Core/NPU/Emul] Fill the function body for AVGP emulation
This function fills the function body for emulation of the Average
Pooling instruction. The emulation for AVGP is enabled via this
function.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Dongju Chae [Mon, 18 Nov 2019 02:26:17 +0000 (11:26 +0900)]
[README/DOC] Revise README.md and add some documentation
This commit revises README.md and add some documentation.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Mon, 18 Nov 2019 13:01:55 +0000 (22:01 +0900)]
[BUG] Bug fix in CONV emulation and esum/weight data setup
Thix commit fixes the bug in CONV emulation and esum/weight data setup.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 19 Nov 2019 03:13:25 +0000 (12:13 +0900)]
[TESTDATA] Add a testcase for multi-layer model to test CONVE/ReLU/AVGP
This commit adds new testcase, testcase3, to test CONVE/ReLU/AVGP.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Tue, 19 Nov 2019 03:08:57 +0000 (12:08 +0900)]
[TESTDATA/FIX] Fix the model generation to use specified addresses
This commit fixes the model generation to use specified addresses
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Wook Song [Fri, 15 Nov 2019 05:17:19 +0000 (14:17 +0900)]
[Core/NPU/Emul] Add skeleton code for ReLU, AVGP, CONVE, TCNV, TCNVE
This patch adds skeleton code for running operators, ReLU, AVGP, CONVE,
TCNV, and TCNVE, in the test mode.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Fri, 15 Nov 2019 02:31:26 +0000 (11:31 +0900)]
[Core/NPU/Emul] Add a struct representing instruction format for TCNVE
This patch adds a data struct representing the instruction format for
TCNVE.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Fri, 15 Nov 2019 02:27:58 +0000 (11:27 +0900)]
[Core/NPU/Emul] Add a struct representing instruction format for CONVE
This patch adds a data struct representing the instruction format for
CONVE.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Fri, 15 Nov 2019 02:15:07 +0000 (11:15 +0900)]
[Core/NPU/Emul] Add a struct representing instruction format for TCNV
This patch adds a data struct representing the instruction format for
TCNV.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Fri, 15 Nov 2019 02:06:49 +0000 (11:06 +0900)]
[Core/NPU/Emul] Add a struct representing instruction format for AVGP
This patch adds a data struct representing the instruction format for
AVGP.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Fri, 15 Nov 2019 01:51:16 +0000 (10:51 +0900)]
[Core/NPU/Emul] Add a struct representing instruction format for RELU
This patch adds a data struct representing the instruction format for
RELU.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Dongju Chae [Thu, 14 Nov 2019 10:03:56 +0000 (19:03 +0900)]
[NPU/EMUL] Add NPU emulation engine (support CONV/MAXP/ESUM)
This commit adds NPU emulation engine.
Currently, it supports CONV/MAXP/ESUM.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Fri, 15 Nov 2019 05:19:32 +0000 (14:19 +0900)]
[TESTDATA] Add enum for trinity opcodes
This commit adds enum for trinity opcodes.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Dongju Chae [Thu, 14 Nov 2019 10:48:28 +0000 (19:48 +0900)]
[TESTDATA] Fix multi-layer testdata generation when setting their shamt
This commit fixes the bug in multi-layer testdata generation.
The shamt is differently set depending on an operation type.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Wook Song [Thu, 14 Nov 2019 08:19:57 +0000 (17:19 +0900)]
[Local/Dist/Debian] Run meson with PKG_CONFIG_PATH pointing libnpuvision
NPU Engine requires dependencies on libnpuvision, which is placed under
/opt/trinity. In order to provide those dependencies, this patch adds
PKG_CONFIG_PATH pointing when running meson.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Thu, 14 Nov 2019 07:11:00 +0000 (16:11 +0900)]
[Local/Dist/Debian] Initial debianize
This patch debianizes this NPU Engine for Ubuntu 16.04 running on
x86_64.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Wook Song [Thu, 14 Nov 2019 07:05:51 +0000 (16:05 +0900)]
[Local/Meson] Add a mode to build/run NPU Engine without FM simulator
This patch adds a mode, build_test_only_mode, to build NPU Engine on
a Ubuntu x86_64 machine and to run it without the FastModel simulator.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Dongju Chae [Tue, 12 Nov 2019 02:29:39 +0000 (11:29 +0900)]
[NPU/EMUL] Add a skeleton for NPU ioctl emulations
This commit adds a skeleton for NPU ioctl emulations. It replaces the existing NPU ioctls to dummy ones, and will eventually perform emulations (TODO).
Note that this source file is written in C++. The reason is the existing emulation codes for VISA are written in C++, and we need their C wrappers.
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>
Wook Song [Mon, 4 Nov 2019 07:52:42 +0000 (16:52 +0900)]
[Core/Conf] Make load_conf_default static
In order to avoid confusion between load_conf and load_conf_default,
this patch makes the load_conf_default function static.
Signed-off-by: Wook Song <wook16.song@samsung.com>
Dongju Chae [Mon, 11 Nov 2019 03:03:49 +0000 (12:03 +0900)]
[NPU] add a missing header, typedef.h, to npubinfmt.h
This commit adds a missing header, tpyedef.h, to npubinfmt.h
Signed-off-by: Dongju Chae <dongju.chae@samsung.com>