diff options
Diffstat (limited to 'src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp16_nhwc_max_generic_depthfirst/generic.cpp')
| -rw-r--r-- | src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp16_nhwc_max_generic_depthfirst/generic.cpp | 32 |
1 files changed, 16 insertions, 16 deletions
diff --git a/src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp16_nhwc_max_generic_depthfirst/generic.cpp b/src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp16_nhwc_max_generic_depthfirst/generic.cpp index bbd32137c5..58ab915605 100644 --- a/src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp16_nhwc_max_generic_depthfirst/generic.cpp +++ b/src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp16_nhwc_max_generic_depthfirst/generic.cpp @@ -84,33 +84,33 @@ void sve_fp16_nhwc_max_generic_depthfirst_impl( "subs x24, x24, #0x1\n" "movprfx z23, z1\n fmax z23.h, p4/M, z23.h, z0.h\n" "ldp x21, x20, [x19, #0x10]\n" - "movprfx z18, z31\n fmax z18.h, p4/M, z18.h, z30.h\n" "add x19, x19, #0x20\n" - "fmax z22.h, p4/M, z22.h, z29.h\n" + "movprfx z18, z31\n fmax z18.h, p4/M, z18.h, z30.h\n" "ld1h { z3.h }, p3/Z, [x23, x28, LSL #1]\n" + "fmax z22.h, p4/M, z22.h, z29.h\n" "movprfx z17, z28\n fmax z17.h, p4/M, z17.h, z27.h\n" - "fmax z21.h, p4/M, z21.h, z26.h\n" "ld1h { z2.h }, p3/Z, [x22, x28, LSL #1]\n" - "fmax z16.h, p4/M, z16.h, z25.h\n" + "fmax z21.h, p4/M, z21.h, z26.h\n" "ld1h { z1.h }, p3/Z, [x21, x28, LSL #1]\n" - "fmax z20.h, p4/M, z20.h, z24.h\n" + "fmax z16.h, p4/M, z16.h, z25.h\n" "ld1h { z0.h }, p3/Z, [x20, x28, LSL #1]\n" - "fmax z19.h, p4/M, z19.h, z23.h\n" + "fmax z20.h, p4/M, z20.h, z24.h\n" "ld1h { z31.h }, p2/Z, [x23, x27, LSL #1]\n" - "fmax z18.h, p4/M, z18.h, z22.h\n" + "fmax z19.h, p4/M, z19.h, z23.h\n" "ld1h { z30.h }, p2/Z, [x22, x27, LSL #1]\n" - "fmax z17.h, p4/M, z17.h, z21.h\n" + "fmax z18.h, p4/M, z18.h, z22.h\n" "ld1h { z22.h }, p2/Z, [x21, x27, LSL #1]\n" - "fmax z16.h, p4/M, z16.h, z20.h\n" + "fmax z17.h, p4/M, z17.h, z21.h\n" "ld1h { z29.h }, p2/Z, [x20, x27, LSL #1]\n" - "fmax z7.h, p4/M, z7.h, z19.h\n" + "fmax z16.h, p4/M, z16.h, z20.h\n" "ld1h { z28.h }, p1/Z, [x23, x26, LSL #1]\n" - "fmax z6.h, p4/M, z6.h, z18.h\n" + "fmax z7.h, p4/M, z7.h, z19.h\n" "ld1h { z27.h }, p1/Z, [x22, x26, LSL #1]\n" - "fmax z5.h, p4/M, z5.h, z17.h\n" + "fmax z6.h, p4/M, z6.h, z18.h\n" "ld1h { z21.h }, p1/Z, [x21, x26, LSL #1]\n" - "fmax z4.h, p4/M, z4.h, z16.h\n" + "fmax z5.h, p4/M, z5.h, z17.h\n" "ld1h { z26.h }, p1/Z, [x20, x26, LSL #1]\n" + "fmax z4.h, p4/M, z4.h, z16.h\n" "ld1h { z16.h }, p0/Z, [x23, x25, LSL #1]\n" "ld1h { z25.h }, p0/Z, [x22, x25, LSL #1]\n" "ld1h { z20.h }, p0/Z, [x21, x25, LSL #1]\n" @@ -168,9 +168,9 @@ void sve_fp16_nhwc_max_generic_depthfirst_impl( "lsr x24, %x[n_valid_cells], #0x2\n" "cbz x24, 11f\n" "ldp x23, x22, [x19, #0x0]\n" - "subs x24, x24, #0x1\n" "ldp x21, x20, [x19, #0x10]\n" "add x19, x19, #0x20\n" + "subs x24, x24, #0x1\n" "ld1h { z3.h }, p3/Z, [x23, x28, LSL #1]\n" "ld1h { z2.h }, p3/Z, [x22, x28, LSL #1]\n" "ld1h { z1.h }, p3/Z, [x21, x28, LSL #1]\n" @@ -182,11 +182,11 @@ void sve_fp16_nhwc_max_generic_depthfirst_impl( "subs x24, x24, #0x1\n" "movprfx z23, z1\n fmax z23.h, p4/M, z23.h, z0.h\n" "ldp x21, x20, [x19, #0x10]\n" + "add x19, x19, #0x20\n" "fmax z19.h, p4/M, z19.h, z23.h\n" "ld1h { z3.h }, p3/Z, [x23, x28, LSL #1]\n" - "fmax z7.h, p4/M, z7.h, z19.h\n" "ld1h { z2.h }, p3/Z, [x22, x28, LSL #1]\n" - "add x19, x19, #0x20\n" + "fmax z7.h, p4/M, z7.h, z19.h\n" "ld1h { z1.h }, p3/Z, [x21, x28, LSL #1]\n" "ld1h { z0.h }, p3/Z, [x20, x28, LSL #1]\n" "bgt 9b\n" |