diff options
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/transforms/sve_interleave_8way_block4_8bit.hpp')
| -rw-r--r-- | src/core/NEON/kernels/arm_gemm/transforms/sve_interleave_8way_block4_8bit.hpp | 204 |
1 files changed, 102 insertions, 102 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/transforms/sve_interleave_8way_block4_8bit.hpp b/src/core/NEON/kernels/arm_gemm/transforms/sve_interleave_8way_block4_8bit.hpp index f1690baf43..a96a43cbeb 100644 --- a/src/core/NEON/kernels/arm_gemm/transforms/sve_interleave_8way_block4_8bit.hpp +++ b/src/core/NEON/kernels/arm_gemm/transforms/sve_interleave_8way_block4_8bit.hpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018 Arm Limited. + * Copyright (c) 2018 - 2019 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -67,8 +67,8 @@ inline void TransformImpl<8, 4, false, 1, 1, false>::Transform(T *out, const T * "zip1 z8.s, z0.s, z4.s\n" "zip2 z9.s, z0.s, z4.s\n" "whilelt p1.b, %[outpos], %[outwidth]\n" - "incb %[outpos], all, mul #1\n" "zip1 z0.s, z8.s, z4.s\n" + "incb %[outpos], all, mul #1\n" "zip2 z1.s, z8.s, z4.s\n" "zip1 z2.s, z9.s, z4.s\n" "zip2 z3.s, z9.s, z4.s\n" @@ -77,28 +77,28 @@ inline void TransformImpl<8, 4, false, 1, 1, false>::Transform(T *out, const T * "incb %[outpos], all, mul #1\n" "zip2 z9.s, z0.s, z4.s\n" "zip1 z10.s, z1.s, z4.s\n" - "zip2 z11.s, z1.s, z4.s\n" "st1b z8.b, p0, [%[outptr]]\n" - "zip1 z12.s, z2.s, z4.s\n" + "zip2 z11.s, z1.s, z4.s\n" "whilelt p3.b, %[outpos], %[outwidth]\n" + "zip1 z12.s, z2.s, z4.s\n" + "st1b z9.b, p1, [%[outptr], #1, MUL VL]\n" "zip2 z13.s, z2.s, z4.s\n" "incb %[outpos], all, mul #1\n" "zip1 z14.s, z3.s, z4.s\n" - "st1b z9.b, p1, [%[outptr], #1, MUL VL]\n" + "st1b z10.b, p2, [%[outptr], #2, MUL VL]\n" "zip2 z15.s, z3.s, z4.s\n" "whilelt p4.b, %[outpos], %[outwidth]\n" - "st1b z10.b, p2, [%[outptr], #2, MUL VL]\n" - "incb %[outpos], all, mul #1\n" "st1b z11.b, p3, [%[outptr], #3, MUL VL]\n" - "whilelt p5.b, %[outpos], %[outwidth]\n" "incb %[outpos], all, mul #1\n" "st1b z12.b, p4, [%[outptr], #4, MUL VL]\n" - "whilelt p6.b, %[outpos], %[outwidth]\n" + "whilelt p5.b, %[outpos], %[outwidth]\n" "incb %[outpos], all, mul #1\n" "st1b z13.b, p5, [%[outptr], #5, MUL VL]\n" - "whilelt p7.b, %[outpos], %[outwidth]\n" + "whilelt p6.b, %[outpos], %[outwidth]\n" "incb %[outpos], all, mul #1\n" "st1b z14.b, p6, [%[outptr], #6, MUL VL]\n" + "whilelt p7.b, %[outpos], %[outwidth]\n" + "incb %[outpos], all, mul #1\n" "st1b z15.b, p7, [%[outptr], #7, MUL VL]\n" "addvl %[outptr], %[outptr], #8\n" "b 1b\n" @@ -120,8 +120,8 @@ inline void TransformImpl<8, 4, false, 1, 1, false>::Transform(T *out, const T * "ld1b z1.b, p0/z, [%[inptr1], %[inpos]]\n" "incb %[inpos], all, mul #1\n" "whilelt p0.b, %[outpos], %[outwidth]\n" - "incb %[outpos], all, mul #1\n" "zip1 z8.s, z0.s, z4.s\n" + "incb %[outpos], all, mul #1\n" "zip2 z9.s, z0.s, z4.s\n" "zip1 z10.s, z1.s, z4.s\n" "zip2 z11.s, z1.s, z4.s\n" @@ -131,36 +131,36 @@ inline void TransformImpl<8, 4, false, 1, 1, false>::Transform(T *out, const T * "zip2 z1.s, z8.s, z4.s\n" "zip1 z2.s, z9.s, z4.s\n" "zip2 z3.s, z9.s, z4.s\n" - "zip1 z4.s, z10.s, z14.s\n" "whilelt p2.b, %[outpos], %[outwidth]\n" - "zip2 z5.s, z10.s, z14.s\n" + "zip1 z4.s, z10.s, z14.s\n" "incb %[outpos], all, mul #1\n" + "zip2 z5.s, z10.s, z14.s\n" "zip1 z6.s, z11.s, z14.s\n" "zip2 z7.s, z11.s, z14.s\n" + "whilelt p3.b, %[outpos], %[outwidth]\n" "zip1 z8.s, z0.s, z4.s\n" + "incb %[outpos], all, mul #1\n" "zip2 z9.s, z0.s, z4.s\n" - "whilelt p3.b, %[outpos], %[outwidth]\n" "zip1 z10.s, z1.s, z5.s\n" - "incb %[outpos], all, mul #1\n" - "zip2 z11.s, z1.s, z5.s\n" "st1b z8.b, p0, [%[outptr]]\n" - "zip1 z12.s, z2.s, z6.s\n" - "zip2 z13.s, z2.s, z6.s\n" - "zip1 z14.s, z3.s, z7.s\n" + "zip2 z11.s, z1.s, z5.s\n" "whilelt p4.b, %[outpos], %[outwidth]\n" - "zip2 z15.s, z3.s, z7.s\n" + "zip1 z12.s, z2.s, z6.s\n" "st1b z9.b, p1, [%[outptr], #1, MUL VL]\n" + "zip2 z13.s, z2.s, z6.s\n" "incb %[outpos], all, mul #1\n" + "zip1 z14.s, z3.s, z7.s\n" "st1b z10.b, p2, [%[outptr], #2, MUL VL]\n" + "zip2 z15.s, z3.s, z7.s\n" "whilelt p5.b, %[outpos], %[outwidth]\n" - "incb %[outpos], all, mul #1\n" "st1b z11.b, p3, [%[outptr], #3, MUL VL]\n" - "whilelt p6.b, %[outpos], %[outwidth]\n" "incb %[outpos], all, mul #1\n" "st1b z12.b, p4, [%[outptr], #4, MUL VL]\n" - "whilelt p7.b, %[outpos], %[outwidth]\n" + "whilelt p6.b, %[outpos], %[outwidth]\n" "incb %[outpos], all, mul #1\n" "st1b z13.b, p5, [%[outptr], #5, MUL VL]\n" + "whilelt p7.b, %[outpos], %[outwidth]\n" + "incb %[outpos], all, mul #1\n" "st1b z14.b, p6, [%[outptr], #6, MUL VL]\n" "st1b z15.b, p7, [%[outptr], #7, MUL VL]\n" "addvl %[outptr], %[outptr], #8\n" @@ -183,50 +183,50 @@ inline void TransformImpl<8, 4, false, 1, 1, false>::Transform(T *out, const T * "ld1b z1.b, p0/z, [%[inptr1], %[inpos]]\n" "ld1b z2.b, p0/z, [%[inptr2], %[inpos]]\n" "incb %[inpos], all, mul #1\n" - "whilelt p0.b, %[outpos], %[outwidth]\n" "zip1 z8.s, z0.s, z4.s\n" - "incb %[outpos], all, mul #1\n" + "whilelt p0.b, %[outpos], %[outwidth]\n" "zip2 z9.s, z0.s, z4.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z10.s, z1.s, z4.s\n" "zip2 z11.s, z1.s, z4.s\n" "zip1 z12.s, z2.s, z4.s\n" "whilelt p1.b, %[outpos], %[outwidth]\n" "zip2 z13.s, z2.s, z4.s\n" "incb %[outpos], all, mul #1\n" - "zip1 z4.s, z10.s, z14.s\n" "zip1 z0.s, z8.s, z12.s\n" "zip2 z1.s, z8.s, z12.s\n" "zip1 z2.s, z9.s, z13.s\n" "whilelt p2.b, %[outpos], %[outwidth]\n" "zip2 z3.s, z9.s, z13.s\n" "incb %[outpos], all, mul #1\n" + "zip1 z4.s, z10.s, z14.s\n" "zip2 z5.s, z10.s, z14.s\n" "zip1 z6.s, z11.s, z14.s\n" + "whilelt p3.b, %[outpos], %[outwidth]\n" "zip2 z7.s, z11.s, z14.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z8.s, z0.s, z4.s\n" - "whilelt p3.b, %[outpos], %[outwidth]\n" "zip2 z9.s, z0.s, z4.s\n" - "incb %[outpos], all, mul #1\n" "zip1 z10.s, z1.s, z5.s\n" - "st1b z8.b, p0, [%[outptr]]\n" + "whilelt p4.b, %[outpos], %[outwidth]\n" "zip2 z11.s, z1.s, z5.s\n" + "st1b z8.b, p0, [%[outptr]]\n" "zip1 z12.s, z2.s, z6.s\n" + "incb %[outpos], all, mul #1\n" "zip2 z13.s, z2.s, z6.s\n" - "whilelt p4.b, %[outpos], %[outwidth]\n" - "zip1 z14.s, z3.s, z7.s\n" "st1b z9.b, p1, [%[outptr], #1, MUL VL]\n" + "zip1 z14.s, z3.s, z7.s\n" "zip2 z15.s, z3.s, z7.s\n" - "incb %[outpos], all, mul #1\n" - "st1b z10.b, p2, [%[outptr], #2, MUL VL]\n" "whilelt p5.b, %[outpos], %[outwidth]\n" + "st1b z10.b, p2, [%[outptr], #2, MUL VL]\n" "incb %[outpos], all, mul #1\n" "st1b z11.b, p3, [%[outptr], #3, MUL VL]\n" "whilelt p6.b, %[outpos], %[outwidth]\n" "incb %[outpos], all, mul #1\n" "st1b z12.b, p4, [%[outptr], #4, MUL VL]\n" "whilelt p7.b, %[outpos], %[outwidth]\n" - "st1b z13.b, p5, [%[outptr], #5, MUL VL]\n" "incb %[outpos], all, mul #1\n" + "st1b z13.b, p5, [%[outptr], #5, MUL VL]\n" "st1b z14.b, p6, [%[outptr], #6, MUL VL]\n" "st1b z15.b, p7, [%[outptr], #7, MUL VL]\n" "addvl %[outptr], %[outptr], #8\n" @@ -256,40 +256,40 @@ inline void TransformImpl<8, 4, false, 1, 1, false>::Transform(T *out, const T * "zip1 z10.s, z1.s, z4.s\n" "zip2 z11.s, z1.s, z4.s\n" "zip1 z12.s, z2.s, z4.s\n" - "zip2 z13.s, z2.s, z4.s\n" "whilelt p1.b, %[outpos], %[outwidth]\n" - "zip1 z14.s, z3.s, z4.s\n" + "zip2 z13.s, z2.s, z4.s\n" "incb %[outpos], all, mul #1\n" + "zip1 z14.s, z3.s, z4.s\n" "zip2 z15.s, z3.s, z4.s\n" "zip1 z0.s, z8.s, z12.s\n" + "whilelt p2.b, %[outpos], %[outwidth]\n" "zip2 z1.s, z8.s, z12.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z2.s, z9.s, z13.s\n" - "whilelt p2.b, %[outpos], %[outwidth]\n" "zip2 z3.s, z9.s, z13.s\n" - "incb %[outpos], all, mul #1\n" "zip1 z4.s, z10.s, z14.s\n" + "whilelt p3.b, %[outpos], %[outwidth]\n" "zip2 z5.s, z10.s, z14.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z6.s, z11.s, z15.s\n" "zip2 z7.s, z11.s, z15.s\n" - "whilelt p3.b, %[outpos], %[outwidth]\n" "zip1 z8.s, z0.s, z4.s\n" - "incb %[outpos], all, mul #1\n" + "whilelt p4.b, %[outpos], %[outwidth]\n" "zip2 z9.s, z0.s, z4.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z10.s, z1.s, z5.s\n" - "zip2 z11.s, z1.s, z5.s\n" "st1b z8.b, p0, [%[outptr]]\n" + "zip2 z11.s, z1.s, z5.s\n" "zip1 z12.s, z2.s, z6.s\n" - "whilelt p4.b, %[outpos], %[outwidth]\n" + "whilelt p5.b, %[outpos], %[outwidth]\n" "zip2 z13.s, z2.s, z6.s\n" - "incb %[outpos], all, mul #1\n" - "zip1 z14.s, z3.s, z7.s\n" "st1b z9.b, p1, [%[outptr], #1, MUL VL]\n" + "zip1 z14.s, z3.s, z7.s\n" + "incb %[outpos], all, mul #1\n" "zip2 z15.s, z3.s, z7.s\n" - "whilelt p5.b, %[outpos], %[outwidth]\n" "st1b z10.b, p2, [%[outptr], #2, MUL VL]\n" - "incb %[outpos], all, mul #1\n" - "st1b z11.b, p3, [%[outptr], #3, MUL VL]\n" "whilelt p6.b, %[outpos], %[outwidth]\n" + "st1b z11.b, p3, [%[outptr], #3, MUL VL]\n" "incb %[outpos], all, mul #1\n" "st1b z12.b, p4, [%[outptr], #4, MUL VL]\n" "whilelt p7.b, %[outpos], %[outwidth]\n" @@ -320,45 +320,45 @@ inline void TransformImpl<8, 4, false, 1, 1, false>::Transform(T *out, const T * "incb %[inpos], all, mul #1\n" "zip1 z10.s, z1.s, z5.s\n" "whilelt p0.b, %[outpos], %[outwidth]\n" - "zip2 z11.s, z1.s, z5.s\n" - "incb %[outpos], all, mul #1\n" "zip1 z8.s, z0.s, z4.s\n" + "incb %[outpos], all, mul #1\n" "zip2 z9.s, z0.s, z4.s\n" + "zip2 z11.s, z1.s, z5.s\n" "zip1 z12.s, z2.s, z5.s\n" - "zip2 z13.s, z2.s, z5.s\n" "whilelt p1.b, %[outpos], %[outwidth]\n" - "zip1 z14.s, z3.s, z5.s\n" + "zip2 z13.s, z2.s, z5.s\n" "incb %[outpos], all, mul #1\n" + "zip1 z14.s, z3.s, z5.s\n" "zip2 z15.s, z3.s, z5.s\n" "zip1 z0.s, z8.s, z12.s\n" + "whilelt p2.b, %[outpos], %[outwidth]\n" "zip2 z1.s, z8.s, z12.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z2.s, z9.s, z13.s\n" - "whilelt p2.b, %[outpos], %[outwidth]\n" "zip2 z3.s, z9.s, z13.s\n" - "incb %[outpos], all, mul #1\n" "zip1 z4.s, z10.s, z14.s\n" + "whilelt p3.b, %[outpos], %[outwidth]\n" "zip2 z5.s, z10.s, z14.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z6.s, z11.s, z15.s\n" "zip2 z7.s, z11.s, z15.s\n" - "whilelt p3.b, %[outpos], %[outwidth]\n" "zip1 z8.s, z0.s, z4.s\n" - "incb %[outpos], all, mul #1\n" + "whilelt p4.b, %[outpos], %[outwidth]\n" "zip2 z9.s, z0.s, z4.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z10.s, z1.s, z5.s\n" - "zip2 z11.s, z1.s, z5.s\n" "st1b z8.b, p0, [%[outptr]]\n" + "zip2 z11.s, z1.s, z5.s\n" "zip1 z12.s, z2.s, z6.s\n" - "whilelt p4.b, %[outpos], %[outwidth]\n" + "whilelt p5.b, %[outpos], %[outwidth]\n" "zip2 z13.s, z2.s, z6.s\n" - "incb %[outpos], all, mul #1\n" - "zip1 z14.s, z3.s, z7.s\n" "st1b z9.b, p1, [%[outptr], #1, MUL VL]\n" + "zip1 z14.s, z3.s, z7.s\n" + "incb %[outpos], all, mul #1\n" "zip2 z15.s, z3.s, z7.s\n" - "whilelt p5.b, %[outpos], %[outwidth]\n" "st1b z10.b, p2, [%[outptr], #2, MUL VL]\n" - "incb %[outpos], all, mul #1\n" - "st1b z11.b, p3, [%[outptr], #3, MUL VL]\n" "whilelt p6.b, %[outpos], %[outwidth]\n" + "st1b z11.b, p3, [%[outptr], #3, MUL VL]\n" "incb %[outpos], all, mul #1\n" "st1b z12.b, p4, [%[outptr], #4, MUL VL]\n" "whilelt p7.b, %[outpos], %[outwidth]\n" @@ -395,40 +395,40 @@ inline void TransformImpl<8, 4, false, 1, 1, false>::Transform(T *out, const T * "zip2 z9.s, z0.s, z4.s\n" "zip1 z10.s, z1.s, z5.s\n" "zip2 z11.s, z1.s, z5.s\n" - "zip2 z13.s, z2.s, z6.s\n" "whilelt p1.b, %[outpos], %[outwidth]\n" - "zip1 z14.s, z3.s, z6.s\n" + "zip2 z13.s, z2.s, z6.s\n" "incb %[outpos], all, mul #1\n" + "zip1 z14.s, z3.s, z6.s\n" "zip2 z15.s, z3.s, z6.s\n" "zip1 z0.s, z8.s, z12.s\n" + "whilelt p2.b, %[outpos], %[outwidth]\n" "zip2 z1.s, z8.s, z12.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z2.s, z9.s, z13.s\n" - "whilelt p2.b, %[outpos], %[outwidth]\n" "zip2 z3.s, z9.s, z13.s\n" - "incb %[outpos], all, mul #1\n" "zip1 z4.s, z10.s, z14.s\n" + "whilelt p3.b, %[outpos], %[outwidth]\n" "zip2 z5.s, z10.s, z14.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z6.s, z11.s, z15.s\n" "zip2 z7.s, z11.s, z15.s\n" - "whilelt p3.b, %[outpos], %[outwidth]\n" "zip1 z8.s, z0.s, z4.s\n" - "incb %[outpos], all, mul #1\n" + "whilelt p4.b, %[outpos], %[outwidth]\n" "zip2 z9.s, z0.s, z4.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z10.s, z1.s, z5.s\n" - "zip2 z11.s, z1.s, z5.s\n" "st1b z8.b, p0, [%[outptr]]\n" + "zip2 z11.s, z1.s, z5.s\n" "zip1 z12.s, z2.s, z6.s\n" - "whilelt p4.b, %[outpos], %[outwidth]\n" + "whilelt p5.b, %[outpos], %[outwidth]\n" "zip2 z13.s, z2.s, z6.s\n" - "incb %[outpos], all, mul #1\n" - "zip1 z14.s, z3.s, z7.s\n" "st1b z9.b, p1, [%[outptr], #1, MUL VL]\n" + "zip1 z14.s, z3.s, z7.s\n" + "incb %[outpos], all, mul #1\n" "zip2 z15.s, z3.s, z7.s\n" - "whilelt p5.b, %[outpos], %[outwidth]\n" "st1b z10.b, p2, [%[outptr], #2, MUL VL]\n" - "incb %[outpos], all, mul #1\n" - "st1b z11.b, p3, [%[outptr], #3, MUL VL]\n" "whilelt p6.b, %[outpos], %[outwidth]\n" + "st1b z11.b, p3, [%[outptr], #3, MUL VL]\n" "incb %[outpos], all, mul #1\n" "st1b z12.b, p4, [%[outptr], #4, MUL VL]\n" "whilelt p7.b, %[outpos], %[outwidth]\n" @@ -459,47 +459,47 @@ inline void TransformImpl<8, 4, false, 1, 1, false>::Transform(T *out, const T * "ld1b z5.b, p0/z, [%[inptr5], %[inpos]]\n" "ld1b z6.b, p0/z, [%[inptr6], %[inpos]]\n" "incb %[inpos], all, mul #1\n" - "zip1 z14.s, z3.s, z7.s\n" - "whilelt p0.b, %[outpos], %[outwidth]\n" "zip1 z8.s, z0.s, z4.s\n" - "incb %[outpos], all, mul #1\n" + "whilelt p0.b, %[outpos], %[outwidth]\n" "zip2 z9.s, z0.s, z4.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z10.s, z1.s, z5.s\n" "zip2 z11.s, z1.s, z5.s\n" "zip1 z12.s, z2.s, z6.s\n" "whilelt p1.b, %[outpos], %[outwidth]\n" "zip2 z13.s, z2.s, z6.s\n" "incb %[outpos], all, mul #1\n" + "zip1 z14.s, z3.s, z7.s\n" "zip2 z15.s, z3.s, z7.s\n" "zip1 z0.s, z8.s, z12.s\n" + "whilelt p2.b, %[outpos], %[outwidth]\n" "zip2 z1.s, z8.s, z12.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z2.s, z9.s, z13.s\n" - "whilelt p2.b, %[outpos], %[outwidth]\n" "zip2 z3.s, z9.s, z13.s\n" - "incb %[outpos], all, mul #1\n" "zip1 z4.s, z10.s, z14.s\n" + "whilelt p3.b, %[outpos], %[outwidth]\n" "zip2 z5.s, z10.s, z14.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z6.s, z11.s, z15.s\n" "zip2 z7.s, z11.s, z15.s\n" - "whilelt p3.b, %[outpos], %[outwidth]\n" "zip1 z8.s, z0.s, z4.s\n" - "incb %[outpos], all, mul #1\n" + "whilelt p4.b, %[outpos], %[outwidth]\n" "zip2 z9.s, z0.s, z4.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z10.s, z1.s, z5.s\n" - "zip2 z11.s, z1.s, z5.s\n" "st1b z8.b, p0, [%[outptr]]\n" + "zip2 z11.s, z1.s, z5.s\n" "zip1 z12.s, z2.s, z6.s\n" - "whilelt p4.b, %[outpos], %[outwidth]\n" + "whilelt p5.b, %[outpos], %[outwidth]\n" "zip2 z13.s, z2.s, z6.s\n" - "incb %[outpos], all, mul #1\n" - "zip1 z14.s, z3.s, z7.s\n" "st1b z9.b, p1, [%[outptr], #1, MUL VL]\n" + "zip1 z14.s, z3.s, z7.s\n" + "incb %[outpos], all, mul #1\n" "zip2 z15.s, z3.s, z7.s\n" - "whilelt p5.b, %[outpos], %[outwidth]\n" "st1b z10.b, p2, [%[outptr], #2, MUL VL]\n" - "incb %[outpos], all, mul #1\n" - "st1b z11.b, p3, [%[outptr], #3, MUL VL]\n" "whilelt p6.b, %[outpos], %[outwidth]\n" + "st1b z11.b, p3, [%[outptr], #3, MUL VL]\n" "incb %[outpos], all, mul #1\n" "st1b z12.b, p4, [%[outptr], #4, MUL VL]\n" "whilelt p7.b, %[outpos], %[outwidth]\n" @@ -538,40 +538,40 @@ inline void TransformImpl<8, 4, false, 1, 1, false>::Transform(T *out, const T * "zip1 z10.s, z1.s, z5.s\n" "zip2 z11.s, z1.s, z5.s\n" "zip1 z12.s, z2.s, z6.s\n" - "zip2 z13.s, z2.s, z6.s\n" "whilelt p1.b, %[outpos], %[outwidth]\n" - "zip1 z14.s, z3.s, z7.s\n" + "zip2 z13.s, z2.s, z6.s\n" "incb %[outpos], all, mul #1\n" + "zip1 z14.s, z3.s, z7.s\n" "zip2 z15.s, z3.s, z7.s\n" "zip1 z0.s, z8.s, z12.s\n" + "whilelt p2.b, %[outpos], %[outwidth]\n" "zip2 z1.s, z8.s, z12.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z2.s, z9.s, z13.s\n" - "whilelt p2.b, %[outpos], %[outwidth]\n" "zip2 z3.s, z9.s, z13.s\n" - "incb %[outpos], all, mul #1\n" "zip1 z4.s, z10.s, z14.s\n" + "whilelt p3.b, %[outpos], %[outwidth]\n" "zip2 z5.s, z10.s, z14.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z6.s, z11.s, z15.s\n" "zip2 z7.s, z11.s, z15.s\n" - "whilelt p3.b, %[outpos], %[outwidth]\n" "zip1 z8.s, z0.s, z4.s\n" - "incb %[outpos], all, mul #1\n" + "whilelt p4.b, %[outpos], %[outwidth]\n" "zip2 z9.s, z0.s, z4.s\n" + "incb %[outpos], all, mul #1\n" "zip1 z10.s, z1.s, z5.s\n" - "zip2 z11.s, z1.s, z5.s\n" "st1b z8.b, p0, [%[outptr]]\n" + "zip2 z11.s, z1.s, z5.s\n" "zip1 z12.s, z2.s, z6.s\n" - "whilelt p4.b, %[outpos], %[outwidth]\n" + "whilelt p5.b, %[outpos], %[outwidth]\n" "zip2 z13.s, z2.s, z6.s\n" - "incb %[outpos], all, mul #1\n" - "zip1 z14.s, z3.s, z7.s\n" "st1b z9.b, p1, [%[outptr], #1, MUL VL]\n" + "zip1 z14.s, z3.s, z7.s\n" + "incb %[outpos], all, mul #1\n" "zip2 z15.s, z3.s, z7.s\n" - "whilelt p5.b, %[outpos], %[outwidth]\n" "st1b z10.b, p2, [%[outptr], #2, MUL VL]\n" - "incb %[outpos], all, mul #1\n" - "st1b z11.b, p3, [%[outptr], #3, MUL VL]\n" "whilelt p6.b, %[outpos], %[outwidth]\n" + "st1b z11.b, p3, [%[outptr], #3, MUL VL]\n" "incb %[outpos], all, mul #1\n" "st1b z12.b, p4, [%[outptr], #4, MUL VL]\n" "whilelt p7.b, %[outpos], %[outwidth]\n" |