experiments
/
STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

no dotprod outside of linux 

PiperOrigin-RevId: 254183421
This commit is contained in:
Benoit Jacob 2019-06-20 06:17:52 -07:00 committed by TensorFlower Gardener
parent 3bbf556706
commit 8052ffb5b6
1 changed files with 6 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
tensorflow/lite/kernels/internal/optimized/depthwiseconv_uint8_3x3_filter.h
View File

@ -7209,6 +7209,7 @@ struct KernelMacroBlock<DepthwiseConvImplementation::kUseNeon3x3DotProduct,
#define DC_KERNEL_NO_MULT_25 "25"
#define DC_KERNEL_NO_MULT_26 "26"
#ifdef __linux__
asm volatile(
// Compiled code used block of 288 for spill out of total stack of 448.
// However, two 4-byte spills were sneaked in to #360 and #364.
@ -7834,6 +7835,7 @@ struct KernelMacroBlock<DepthwiseConvImplementation::kUseNeon3x3DotProduct,
"x5", "x6", "x7", "x8", "x9", "x10", "x11", "x12", "x13", "x14", "x15",
"x16", "x17", "x19", "x21", "x22", "x23", "x24", "x25", "x26", "x27",
"x28", "x29", "x30");
#endif // __linux__
} // NOLINT(readability/fn_size) Manually unrolled.
#undef DC_KERNEL_NO_MULT_1
@ -7906,6 +7908,7 @@ struct KernelMacroBlock<DepthwiseConvImplementation::kUseNeon3x3DotProduct,
#define DC_KERNEL_NO_MULT_STRIDE_18 "18"
#define DC_KERNEL_NO_MULT_STRIDE_19 "19"
#ifdef __linux__
asm volatile(
// Compiled code used block of 48 for spill out of total stack of 208.
// However, an 8-byte spill was sneaked in to #120.
@ -8257,6 +8260,7 @@ struct KernelMacroBlock<DepthwiseConvImplementation::kUseNeon3x3DotProduct,
"x5", "x6", "x7", "x8", "x9", "x10", "x11", "x12", "x13", "x14", "x15",
"x16", "x17", "x19", "x21", "x22", "x23", "x24", "x25", "x26", "x27",
"x28", "x29", "x30");
#endif // __linux__
} // NOLINT(readability/fn_size) Manually unrolled.
#undef DC_KERNEL_NO_MULT_STRIDE_1
@ -8325,6 +8329,7 @@ struct KernelMacroBlock<DepthwiseConvImplementation::kUseNeon3x3DotProduct,
#define DC_KERNEL_MULT_21 "21"
#define DC_KERNEL_MULT_22 "22"
#ifdef __linux__
asm volatile(
// Compiled code used block of 160 for spill out of total stack of 288.
// However, an 8-byte spill was sneaked in to #168.
@ -8886,6 +8891,7 @@ struct KernelMacroBlock<DepthwiseConvImplementation::kUseNeon3x3DotProduct,
"x5", "x6", "x7", "x8", "x9", "x10", "x11", "x12", "x13", "x14", "x15",
"x16", "x17", "x19", "x20", "x21", "x22", "x23", "x24", "x25", "x26",
"x27", "x28", "x29", "x30");
#endif // __linux__
} // NOLINT(readability/fn_size) Manually unrolled.
#undef DC_KERNEL_MULT_1