From 2f6997f40e404884d074a958cd24f6ff3a01381a Mon Sep 17 00:00:00 2001 From: skykongkong8 Date: Tue, 24 Dec 2024 16:43:30 +0900 Subject: [PATCH 1/2] [ trivial ] Fix typo & misspelled function param - intY -> incY (incremental index of Y) - __scopy_kernel() only works for interested SIMD register width **Self evaluation:** 1. Build test: [X]Passed [ ]Failed [ ]Skipped 2. Run test: [X]Passed [ ]Failed [ ]Skipped Signed-off-by: skykongkong8 --- nntrainer/tensor/blas_interface.cpp | 2 +- nntrainer/tensor/blas_interface.h | 8 ++++---- nntrainer/tensor/blas_neon.cpp | 2 +- 3 files changed, 6 insertions(+), 6 deletions(-) diff --git a/nntrainer/tensor/blas_interface.cpp b/nntrainer/tensor/blas_interface.cpp index 0847890b49..71671cbc3b 100644 --- a/nntrainer/tensor/blas_interface.cpp +++ b/nntrainer/tensor/blas_interface.cpp @@ -891,7 +891,7 @@ void scopy(const unsigned int N, const float *X, const int incX, float *Y, } void scopy(const unsigned int N, const uint8_t *X, const int incX, uint8_t *Y, - const int intY) { + const int incY) { #ifdef USE_NEON nntrainer::neon::copy_int8_or_int4(N, X, Y); #else diff --git a/nntrainer/tensor/blas_interface.h b/nntrainer/tensor/blas_interface.h index e99a4cedbe..75b77aa1ca 100644 --- a/nntrainer/tensor/blas_interface.h +++ b/nntrainer/tensor/blas_interface.h @@ -265,7 +265,7 @@ void scopy(const unsigned int N, const void *X, const int incX, void *Y, * @param[in] Y float * for Vector Y */ void scopy(const unsigned int N, const float *X, const int incX, float *Y, - const int intY); + const int incY); /** * @brief copy function : Y = X * @param[in] N number of elements in X @@ -273,7 +273,7 @@ void scopy(const unsigned int N, const float *X, const int incX, float *Y, * @param[in] Y uint8_t * for Vector Y */ void scopy(const unsigned int N, const uint8_t *X, const int incX, uint8_t *Y, - const int intY); + const int incY); /** * @brief copy function : Y = X * @param[in] N number of elements in X @@ -281,7 +281,7 @@ void scopy(const unsigned int N, const uint8_t *X, const int incX, uint8_t *Y, * @param[in] Y float * for Vector Y */ void scopy_int4_to_float32(const unsigned int N, const uint8_t *X, - const int incX, float *Y, const int intY); + const int incX, float *Y, const int incY); /** * @brief copy function : Y = X @@ -290,7 +290,7 @@ void scopy_int4_to_float32(const unsigned int N, const uint8_t *X, * @param[in] Y float * for Vector Y */ void scopy_int8_to_float32(const unsigned int N, const uint8_t *X, - const int incX, float *Y, const int intY); + const int incX, float *Y, const int incY); /** * @brief sdot computation : sum of all X * Y diff --git a/nntrainer/tensor/blas_neon.cpp b/nntrainer/tensor/blas_neon.cpp index 0494e96d6b..9e534e0897 100644 --- a/nntrainer/tensor/blas_neon.cpp +++ b/nntrainer/tensor/blas_neon.cpp @@ -602,7 +602,7 @@ void custom_scopy(const unsigned int N, const float *X, const int incX, : "r"(&Y[i]), "r"(&X[i]) : "v0", "memory"); #else - __scopy_kernel(N, X + i, Y + i); + __scopy_kernel(X + i, Y + i); #endif } for (unsigned int i = N4; i < N; ++i) { From 1c595bee706c836f28f539e8f0675456a80d5523 Mon Sep 17 00:00:00 2001 From: skykongkong8 Date: Thu, 26 Dec 2024 09:08:03 +0900 Subject: [PATCH 2/2] [ trivial ] Add missing header file for quantizer - With certain version of compiler, skipping including header file may cause build failure **Self evaluation:** 1. Build test: [X]Passed [ ]Failed [ ]Skipped 2. Run test: [X]Passed [ ]Failed [ ]Skipped Signed-off-by: skykongkong8 --- nntrainer/tensor/quantizer.h | 1 + 1 file changed, 1 insertion(+) diff --git a/nntrainer/tensor/quantizer.h b/nntrainer/tensor/quantizer.h index 6e8d567491..dcd6a9baed 100644 --- a/nntrainer/tensor/quantizer.h +++ b/nntrainer/tensor/quantizer.h @@ -13,6 +13,7 @@ #ifdef __cplusplus #include +#include namespace nntrainer {