diff options
author | Gian Marco Iodice <gianmarco.iodice@arm.com> | 2021-11-25 15:47:37 +0000 |
---|---|---|
committer | Gian Marco Iodice <gianmarco.iodice@arm.com> | 2021-11-29 09:23:04 +0000 |
commit | 56d55123527b5bb84a5c3516f161dd4438cdc7d8 (patch) | |
tree | baa3928802cb63d3a2cdbd75a75a84e31f706a22 /src/core/CL/cl_kernels/nhwc/dwc_native_fp_nhwc.cl | |
parent | bd2942d7c701a664421ce8ef7145f97b7163201a (diff) | |
download | ComputeLibrary-56d55123527b5bb84a5c3516f161dd4438cdc7d8.tar.gz |
Use loop unrolling only when the kernel height is less than 5
- In the dwc_native_fp_nhwc.cl, loop unrolling should only be enabled
when kernel height is less than 5.
- No performance regression experimented
- The patch reduces the compilation time required for the kernel
Resolves COMPMID-4887
Change-Id: I93188b9764cf7d1ad34ac164694f6f1fd37a90e8
Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6744
Reviewed-by: Giorgio Arena <giorgio.arena@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/CL/cl_kernels/nhwc/dwc_native_fp_nhwc.cl')
-rw-r--r-- | src/core/CL/cl_kernels/nhwc/dwc_native_fp_nhwc.cl | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/src/core/CL/cl_kernels/nhwc/dwc_native_fp_nhwc.cl b/src/core/CL/cl_kernels/nhwc/dwc_native_fp_nhwc.cl index 41da4fff10..4f57a81e7b 100644 --- a/src/core/CL/cl_kernels/nhwc/dwc_native_fp_nhwc.cl +++ b/src/core/CL/cl_kernels/nhwc/dwc_native_fp_nhwc.cl @@ -134,7 +134,7 @@ __kernel void dwc_native_fp_nhwc( c[i].v = 0; }) -#if _IWEI_HEIGHT <= 5 +#if _IWEI_HEIGHT < 5 LOOP_UNROLLING(int, yk, 0, 1, _IWEI_HEIGHT, #else // _IWEI_HEIGHT <= 5 for(int yk = 0; yk < _IWEI_HEIGHT; yk++) @@ -165,7 +165,7 @@ __kernel void dwc_native_fp_nhwc( }) }) } -#if _IWEI_HEIGHT <= 5 +#if _IWEI_HEIGHT < 5 ) #endif // _IWEI_HEIGHT <= 5 |