We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
1 parent 92fc2f5 commit f69b126Copy full SHA for f69b126
lite/backends/arm/math/conv3x3s1_depthwise_int8.cc
@@ -151,8 +151,11 @@ namespace math {
151
"st1 {v14.8b}, [%[ptr_out1]], #8 \n"
152
153
#define MID_COMPUTE_S1 \
154
+ "movi v10.4s, #0x0\n" \
155
+ "movi v11.4s, #0x0\n" \
156
"movi v12.4s, #0x0\n" \
157
"movi v13.4s, #0x0\n" \
158
+ "movi v21.4s, #0x0\n" \
159
"cmp %[cnt], #1 \n" \
160
"blt 3f \n" \
161
"1: \n" \
@@ -335,8 +338,11 @@ namespace math {
335
338
"st1 {v18.8b}, [%[ptr_out1]], #8 \n"
336
339
337
340
#define RIGHT_COMPUTE_S1 \
- "movi v12.4s, #0x0 \n" \
- "movi v13.4s, #0x0 \n" \
341
342
343
+ "movi v12.4s, #0x0\n" \
344
+ "movi v13.4s, #0x0\n" \
345
346
"bne 1b \n" \
347
"3: \n" \
348
"ld1 {v20.16b}, [%[vmask]] \n" \
0 commit comments