diff options
| author | Georgios Pinitas <georgios.pinitas@arm.com> | 2018-12-03 14:30:05 +0000 |
|---|---|---|
| committer | Georgios Pinitas <georgios.pinitas@arm.com> | 2019-01-14 17:53:22 +0000 |
| commit | 5a5945387e70f62e6e1e95a177fae261d7570443 (patch) | |
| tree | ff8bd61c2e071b5a0b923f4a0d1bef72486435e9 /arm_compute/core/NEON/wrapper/intrinsics/bsl.h | |
| parent | dea2d2d58fe3a742e6f66fe50befbe0044e15ad1 (diff) | |
| download | ComputeLibrary-5a5945387e70f62e6e1e95a177fae261d7570443.tar.gz | |
COMPMID-1809: Remove padding in NEGEMMConvolutionLayer 64-bit path.
Change-Id: I1806591a2c73a1f057f13d8c6107d7b9796a82c8
Reviewed-on: https://review.mlplatform.org/370
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Diffstat (limited to 'arm_compute/core/NEON/wrapper/intrinsics/bsl.h')
| -rw-r--r-- | arm_compute/core/NEON/wrapper/intrinsics/bsl.h | 40 |
1 files changed, 20 insertions, 20 deletions
diff --git a/arm_compute/core/NEON/wrapper/intrinsics/bsl.h b/arm_compute/core/NEON/wrapper/intrinsics/bsl.h index 9831b4b842..38f9d5f171 100644 --- a/arm_compute/core/NEON/wrapper/intrinsics/bsl.h +++ b/arm_compute/core/NEON/wrapper/intrinsics/bsl.h @@ -30,32 +30,32 @@ namespace arm_compute { namespace wrapper { -#define VBSL_IMPL(vctype, vtype, prefix, postfix) \ - inline vtype vbsl(const vctype &a, const vtype &b, const vtype &c) \ - { \ - return prefix##_##postfix(a, b, c); \ +#define VBSL_IMPL(stype, vtype, ctype, prefix, postfix) \ + inline vtype vbsl(const ctype &a, const vtype &b, const vtype &c) \ + { \ + return prefix##_##postfix(a, b, c); \ } -VBSL_IMPL(uint8x8_t, uint8x8_t, vbsl, u8) -VBSL_IMPL(uint8x8_t, int8x8_t, vbsl, s8) -VBSL_IMPL(uint16x4_t, uint16x4_t, vbsl, u16) -VBSL_IMPL(uint16x4_t, int16x4_t, vbsl, s16) -VBSL_IMPL(uint32x2_t, uint32x2_t, vbsl, u32) -VBSL_IMPL(uint32x2_t, int32x2_t, vbsl, s32) -VBSL_IMPL(uint32x2_t, float32x2_t, vbsl, f32) +VBSL_IMPL(uint8_t, uint8x8_t, uint8x8_t, vbsl, u8) +VBSL_IMPL(int8_t, int8x8_t, uint8x8_t, vbsl, s8) +VBSL_IMPL(uint16_t, uint16x4_t, uint16x4_t, vbsl, u16) +VBSL_IMPL(int16_t, int16x4_t, uint16x4_t, vbsl, s16) +VBSL_IMPL(uint32_t, uint32x2_t, uint32x2_t, vbsl, u32) +VBSL_IMPL(int32_t, int32x2_t, uint32x2_t, vbsl, s32) +VBSL_IMPL(float32x2_t, float32x2_t, uint32x2_t, vbsl, f32) #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC -VBSL_IMPL(uint16x4_t, float16x4_t, vbsl, f16) +VBSL_IMPL(float16x4_t, float16x4_t, uint16x4_t, vbsl, f16) #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC -VBSL_IMPL(uint8x16_t, uint8x16_t, vbslq, u8) -VBSL_IMPL(uint8x16_t, int8x16_t, vbslq, s8) -VBSL_IMPL(uint16x8_t, uint16x8_t, vbslq, u16) -VBSL_IMPL(uint16x8_t, int16x8_t, vbslq, s16) -VBSL_IMPL(uint32x4_t, uint32x4_t, vbslq, u32) -VBSL_IMPL(uint32x4_t, int32x4_t, vbslq, s32) -VBSL_IMPL(uint32x4_t, float32x4_t, vbslq, f32) +VBSL_IMPL(uint8_t, uint8x16_t, uint8x16_t, vbslq, u8) +VBSL_IMPL(int8_t, int8x16_t, uint8x16_t, vbslq, s8) +VBSL_IMPL(uint16_t, uint16x8_t, uint16x8_t, vbslq, u16) +VBSL_IMPL(int16_t, int16x8_t, uint16x8_t, vbslq, s16) +VBSL_IMPL(uint32_t, uint32x4_t, uint32x4_t, vbslq, u32) +VBSL_IMPL(int32_t, int32x4_t, uint32x4_t, vbslq, s32) +VBSL_IMPL(float32x4_t, float32x4_t, uint32x4_t, vbslq, f32) #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC -VBSL_IMPL(uint16x8_t, float16x8_t, vbslq, f16) +VBSL_IMPL(float16x8_t, float16x8_t, uint16x8_t, vbslq, f16) #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC #undef VBSL_IMPL |