diff options
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave2VL_fp32_fp32.hpp')
-rw-r--r-- | src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave2VL_fp32_fp32.hpp | 152 |
1 files changed, 75 insertions, 77 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave2VL_fp32_fp32.hpp b/src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave2VL_fp32_fp32.hpp index 949e003598..ac4b1b5086 100644 --- a/src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave2VL_fp32_fp32.hpp +++ b/src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave2VL_fp32_fp32.hpp @@ -22,16 +22,14 @@ * SOFTWARE. */ -#if defined(__ARM_FEATURE_SVE) +#if defined(ARM_COMPUTE_ENABLE_SME) template <> void interleave_block<2, 1, VLType::SME, false>( float * &out, const float * const *in, - size_t width, size_t height, size_t row_offset, bool first + size_t width, size_t height, size_t row_offset, bool ) { - ARM_COMPUTE_UNUSED(first); - __asm__ __volatile__( ".inst 0xd503477f // SMSTART ZA\n" "mov x22, %x[width]\n" @@ -55,12 +53,12 @@ void interleave_block<2, 1, VLType::SME, false>( "ldr x25, [x11, #0x8]\n" "and x24, x22, #0x1\n" // odd_tail = bool(n_passes & 0x1) "csel x15, x15, x16, NE\n" - "ldr x23, [x9, #0x8]\n" + "ldr x21, [x9, #0x8]\n" "ptrue p13.s\n" "whilelt p12.s, XZR, %x[height]\n" "whilelt p11.s, x16, %x[height]\n" - "mov x22, %x[row_offset]\n" - "mov x21, %x[out]\n" + "mov x23, %x[row_offset]\n" + "mov x22, %x[out]\n" "whilelt p10.s, x13, %x[width]\n" "whilelt p9.s, x13, %x[width]\n" "whilelt p8.s, x13, %x[width]\n" @@ -71,39 +69,39 @@ void interleave_block<2, 1, VLType::SME, false>( "1:" // K loop: Charge: Loop ".inst 0x25306581 // psel p1.s, p9.s/Z, p12.s[w12]\n" ".inst 0x25306160 // psel p0.s, p8.s/Z, p11.s[w12]\n" - ".inst 0xe0960540 // ld1w { za0h.s[x12] }, p1/Z, [x10, x22, LSL #2]\n" + ".inst 0xe0970540 // ld1w { za0h.s[x12] }, p1/Z, [x10, x23, LSL #2]\n" "ldr x10, [x11, #0x0]\n" - ".inst 0xe0960364 // ld1w { za1h.s[x12] }, p0/Z, [x27, x22, LSL #2]\n" + ".inst 0xe0970364 // ld1w { za1h.s[x12] }, p0/Z, [x27, x23, LSL #2]\n" ".inst 0x25706581 // psel p1.s, p9.s/Z, p12.s[w12, #1]\n" ".inst 0x25706160 // psel p0.s, p8.s/Z, p11.s[w12, #1]\n" "ldr x27, [x9, #0x0]\n" - ".inst 0xe0960721 // ld1w { za0h.s[x12, #1] }, p1/Z, [x25, x22, LSL #2]\n" + ".inst 0xe0970721 // ld1w { za0h.s[x12, #1] }, p1/Z, [x25, x23, LSL #2]\n" "ldr x25, [x11, #0x8]\n" "add x11, x11, #0x10\n" - ".inst 0xe09602e5 // ld1w { za1h.s[x12, #1] }, p0/Z, [x23, x22, LSL #2]\n" + ".inst 0xe09702a5 // ld1w { za1h.s[x12, #1] }, p0/Z, [x21, x23, LSL #2]\n" "add x12, x12, #0x2\n" "cmp x12, x14\n" - "ldr x23, [x9, #0x8]\n" + "ldr x21, [x9, #0x8]\n" "add x9, x9, #0x10\n" "blt 1b\n" "2:" // K loop: Charge: End ".inst 0x25306581 // psel p1.s, p9.s/Z, p12.s[w12]\n" ".inst 0x25306160 // psel p0.s, p8.s/Z, p11.s[w12]\n" - ".inst 0xe0960540 // ld1w { za0h.s[x12] }, p1/Z, [x10, x22, LSL #2]\n" - ".inst 0xe0960364 // ld1w { za1h.s[x12] }, p0/Z, [x27, x22, LSL #2]\n" + ".inst 0xe0970540 // ld1w { za0h.s[x12] }, p1/Z, [x10, x23, LSL #2]\n" + ".inst 0xe0970364 // ld1w { za1h.s[x12] }, p0/Z, [x27, x23, LSL #2]\n" ".inst 0x25706581 // psel p1.s, p9.s/Z, p12.s[w12, #1]\n" ".inst 0x25706160 // psel p0.s, p8.s/Z, p11.s[w12, #1]\n" "mov x11, %x[in]\n" "add x9, %x[in], x16, LSL #3\n" - ".inst 0xe0960721 // ld1w { za0h.s[x12, #1] }, p1/Z, [x25, x22, LSL #2]\n" + ".inst 0xe0970721 // ld1w { za0h.s[x12, #1] }, p1/Z, [x25, x23, LSL #2]\n" "ldr x10, [x11, #0x0]\n" - ".inst 0xe09602e5 // ld1w { za1h.s[x12, #1] }, p0/Z, [x23, x22, LSL #2]\n" + ".inst 0xe09702a5 // ld1w { za1h.s[x12, #1] }, p0/Z, [x21, x23, LSL #2]\n" "ldr x27, [x9, #0x0]\n" - "incw x22\n" + "incw x23\n" "incw x13\n" "ldr x25, [x11, #0x8]\n" "add x11, x11, #0x10\n" - "ldr x23, [x9, #0x8]\n" + "ldr x21, [x9, #0x8]\n" "add x9, x9, #0x10\n" "cbz x20, 8f\n" "mov x20, x20\n" @@ -115,59 +113,59 @@ void interleave_block<2, 1, VLType::SME, false>( "4:" // K loop: Main loop: First: Loop ".inst 0x25306581 // psel p1.s, p9.s/Z, p12.s[w12]\n" ".inst 0x25306160 // psel p0.s, p8.s/Z, p11.s[w12]\n" - ".inst 0xe0960548 // ld1w { za2h.s[x12] }, p1/Z, [x10, x22, LSL #2]\n" + ".inst 0xe0970548 // ld1w { za2h.s[x12] }, p1/Z, [x10, x23, LSL #2]\n" "ldr x10, [x11, #0x0]\n" - ".inst 0xe096036c // ld1w { za3h.s[x12] }, p0/Z, [x27, x22, LSL #2]\n" + ".inst 0xe097036c // ld1w { za3h.s[x12] }, p0/Z, [x27, x23, LSL #2]\n" ".inst 0x25706580 // psel p0.s, p9.s/Z, p12.s[w12, #1]\n" ".inst 0x25706162 // psel p2.s, p8.s/Z, p11.s[w12, #1]\n" "ldr x27, [x9, #0x0]\n" ".inst 0x25307541 // psel p1.s, p13.s/Z, p10.s[w12]\n" - ".inst 0xe0960329 // ld1w { za2h.s[x12, #1] }, p0/Z, [x25, x22, LSL #2]\n" + ".inst 0xe0970329 // ld1w { za2h.s[x12, #1] }, p0/Z, [x25, x23, LSL #2]\n" "ldr x25, [x11, #0x8]\n" ".inst 0x25307540 // psel p0.s, p13.s/Z, p10.s[w12]\n" - ".inst 0xe0960aed // ld1w { za3h.s[x12, #1] }, p2/Z, [x23, x22, LSL #2]\n" - "ldr x23, [x9, #0x8]\n" - ".inst 0xe0bf86a0 // st1w { za0v.s[x12] }, p1/Z, [x21, XZR, LSL #2]\n" + ".inst 0xe0970aad // ld1w { za3h.s[x12, #1] }, p2/Z, [x21, x23, LSL #2]\n" + "ldr x21, [x9, #0x8]\n" + ".inst 0xe0bf86c0 // st1w { za0v.s[x12] }, p1/Z, [x22, XZR, LSL #2]\n" ".inst 0x25707541 // psel p1.s, p13.s/Z, p10.s[w12, #1]\n" - ".inst 0xe0b082a4 // st1w { za1v.s[x12] }, p0/Z, [x21, x16, LSL #2]\n" + ".inst 0xe0b082c4 // st1w { za1v.s[x12] }, p0/Z, [x22, x16, LSL #2]\n" ".inst 0x25707540 // psel p0.s, p13.s/Z, p10.s[w12, #1]\n" "add x11, x11, #0x10\n" - ".inst 0xe0bc86a1 // st1w { za0v.s[x12, #1] }, p1/Z, [x21, x28, LSL #2]\n" + ".inst 0xe0bc86c1 // st1w { za0v.s[x12, #1] }, p1/Z, [x22, x28, LSL #2]\n" "add x9, x9, #0x10\n" - ".inst 0xe0ba82a5 // st1w { za1v.s[x12, #1] }, p0/Z, [x21, x26, LSL #2]\n" + ".inst 0xe0ba82c5 // st1w { za1v.s[x12, #1] }, p0/Z, [x22, x26, LSL #2]\n" "add x12, x12, #0x2\n" "cmp x12, x14\n" - "addvl x21, x21, #4\n" + "addvl x22, x22, #4\n" "blt 4b\n" "5:" // K loop: Main loop: First: Tail ".inst 0x25306581 // psel p1.s, p9.s/Z, p12.s[w12]\n" ".inst 0x25306160 // psel p0.s, p8.s/Z, p11.s[w12]\n" - ".inst 0xe0960548 // ld1w { za2h.s[x12] }, p1/Z, [x10, x22, LSL #2]\n" - ".inst 0xe096036c // ld1w { za3h.s[x12] }, p0/Z, [x27, x22, LSL #2]\n" + ".inst 0xe0970548 // ld1w { za2h.s[x12] }, p1/Z, [x10, x23, LSL #2]\n" + ".inst 0xe097036c // ld1w { za3h.s[x12] }, p0/Z, [x27, x23, LSL #2]\n" "mov x11, %x[in]\n" "add x9, %x[in], x16, LSL #3\n" "ldr x10, [x11, #0x0]\n" ".inst 0x25706580 // psel p0.s, p9.s/Z, p12.s[w12, #1]\n" ".inst 0x25706161 // psel p1.s, p8.s/Z, p11.s[w12, #1]\n" - ".inst 0xe0960329 // ld1w { za2h.s[x12, #1] }, p0/Z, [x25, x22, LSL #2]\n" + ".inst 0xe0970329 // ld1w { za2h.s[x12, #1] }, p0/Z, [x25, x23, LSL #2]\n" "ldr x27, [x9, #0x0]\n" ".inst 0x25307540 // psel p0.s, p13.s/Z, p10.s[w12]\n" - ".inst 0xe09606ed // ld1w { za3h.s[x12, #1] }, p1/Z, [x23, x22, LSL #2]\n" + ".inst 0xe09706ad // ld1w { za3h.s[x12, #1] }, p1/Z, [x21, x23, LSL #2]\n" "ldr x25, [x11, #0x8]\n" ".inst 0x25307542 // psel p2.s, p13.s/Z, p10.s[w12]\n" - "ldr x23, [x9, #0x8]\n" - ".inst 0xe0bf82a0 // st1w { za0v.s[x12] }, p0/Z, [x21, XZR, LSL #2]\n" + "ldr x21, [x9, #0x8]\n" + ".inst 0xe0bf82c0 // st1w { za0v.s[x12] }, p0/Z, [x22, XZR, LSL #2]\n" ".inst 0x25707541 // psel p1.s, p13.s/Z, p10.s[w12, #1]\n" ".inst 0x25707540 // psel p0.s, p13.s/Z, p10.s[w12, #1]\n" - ".inst 0xe0b08aa4 // st1w { za1v.s[x12] }, p2/Z, [x21, x16, LSL #2]\n" + ".inst 0xe0b08ac4 // st1w { za1v.s[x12] }, p2/Z, [x22, x16, LSL #2]\n" "whilelt p10.s, x13, %x[width]\n" "incw x13\n" - ".inst 0xe0bc86a1 // st1w { za0v.s[x12, #1] }, p1/Z, [x21, x28, LSL #2]\n" + ".inst 0xe0bc86c1 // st1w { za0v.s[x12, #1] }, p1/Z, [x22, x28, LSL #2]\n" "add x11, x11, #0x10\n" "add x9, x9, #0x10\n" - ".inst 0xe0ba82a5 // st1w { za1v.s[x12, #1] }, p0/Z, [x21, x26, LSL #2]\n" - "addvl x21, x21, #4\n" - "incw x22\n" + ".inst 0xe0ba82c5 // st1w { za1v.s[x12, #1] }, p0/Z, [x22, x26, LSL #2]\n" + "addvl x22, x22, #4\n" + "incw x23\n" "whilelt p9.s, x13, %x[width]\n" "whilelt p8.s, x13, %x[width]\n" "mov x12, #0x0\n" @@ -175,60 +173,60 @@ void interleave_block<2, 1, VLType::SME, false>( "6:" // K loop: Main loop: Second: Loop ".inst 0x25306581 // psel p1.s, p9.s/Z, p12.s[w12]\n" ".inst 0x25306160 // psel p0.s, p8.s/Z, p11.s[w12]\n" - ".inst 0xe0960540 // ld1w { za0h.s[x12] }, p1/Z, [x10, x22, LSL #2]\n" + ".inst 0xe0970540 // ld1w { za0h.s[x12] }, p1/Z, [x10, x23, LSL #2]\n" "ldr x10, [x11, #0x0]\n" - ".inst 0xe0960364 // ld1w { za1h.s[x12] }, p0/Z, [x27, x22, LSL #2]\n" + ".inst 0xe0970364 // ld1w { za1h.s[x12] }, p0/Z, [x27, x23, LSL #2]\n" ".inst 0x25706580 // psel p0.s, p9.s/Z, p12.s[w12, #1]\n" ".inst 0x25706162 // psel p2.s, p8.s/Z, p11.s[w12, #1]\n" "ldr x27, [x9, #0x0]\n" ".inst 0x25307541 // psel p1.s, p13.s/Z, p10.s[w12]\n" - ".inst 0xe0960321 // ld1w { za0h.s[x12, #1] }, p0/Z, [x25, x22, LSL #2]\n" + ".inst 0xe0970321 // ld1w { za0h.s[x12, #1] }, p0/Z, [x25, x23, LSL #2]\n" "ldr x25, [x11, #0x8]\n" ".inst 0x25307540 // psel p0.s, p13.s/Z, p10.s[w12]\n" - ".inst 0xe0960ae5 // ld1w { za1h.s[x12, #1] }, p2/Z, [x23, x22, LSL #2]\n" - "ldr x23, [x9, #0x8]\n" - ".inst 0xe0bf86a8 // st1w { za2v.s[x12] }, p1/Z, [x21, XZR, LSL #2]\n" + ".inst 0xe0970aa5 // ld1w { za1h.s[x12, #1] }, p2/Z, [x21, x23, LSL #2]\n" + "ldr x21, [x9, #0x8]\n" + ".inst 0xe0bf86c8 // st1w { za2v.s[x12] }, p1/Z, [x22, XZR, LSL #2]\n" ".inst 0x25707541 // psel p1.s, p13.s/Z, p10.s[w12, #1]\n" - ".inst 0xe0b082ac // st1w { za3v.s[x12] }, p0/Z, [x21, x16, LSL #2]\n" + ".inst 0xe0b082cc // st1w { za3v.s[x12] }, p0/Z, [x22, x16, LSL #2]\n" ".inst 0x25707540 // psel p0.s, p13.s/Z, p10.s[w12, #1]\n" "add x11, x11, #0x10\n" - ".inst 0xe0bc86a9 // st1w { za2v.s[x12, #1] }, p1/Z, [x21, x28, LSL #2]\n" + ".inst 0xe0bc86c9 // st1w { za2v.s[x12, #1] }, p1/Z, [x22, x28, LSL #2]\n" "add x9, x9, #0x10\n" - ".inst 0xe0ba82ad // st1w { za3v.s[x12, #1] }, p0/Z, [x21, x26, LSL #2]\n" + ".inst 0xe0ba82cd // st1w { za3v.s[x12, #1] }, p0/Z, [x22, x26, LSL #2]\n" "add x12, x12, #0x2\n" "cmp x12, x14\n" - "addvl x21, x21, #4\n" + "addvl x22, x22, #4\n" "blt 6b\n" "7:" // K loop: Main loop: Second: Tail ".inst 0x25306581 // psel p1.s, p9.s/Z, p12.s[w12]\n" ".inst 0x25306160 // psel p0.s, p8.s/Z, p11.s[w12]\n" - ".inst 0xe0960540 // ld1w { za0h.s[x12] }, p1/Z, [x10, x22, LSL #2]\n" - ".inst 0xe0960364 // ld1w { za1h.s[x12] }, p0/Z, [x27, x22, LSL #2]\n" + ".inst 0xe0970540 // ld1w { za0h.s[x12] }, p1/Z, [x10, x23, LSL #2]\n" + ".inst 0xe0970364 // ld1w { za1h.s[x12] }, p0/Z, [x27, x23, LSL #2]\n" "mov x11, %x[in]\n" "add x9, %x[in], x16, LSL #3\n" "ldr x10, [x11, #0x0]\n" ".inst 0x25706580 // psel p0.s, p9.s/Z, p12.s[w12, #1]\n" ".inst 0x25706161 // psel p1.s, p8.s/Z, p11.s[w12, #1]\n" - ".inst 0xe0960321 // ld1w { za0h.s[x12, #1] }, p0/Z, [x25, x22, LSL #2]\n" + ".inst 0xe0970321 // ld1w { za0h.s[x12, #1] }, p0/Z, [x25, x23, LSL #2]\n" "ldr x27, [x9, #0x0]\n" ".inst 0x25307540 // psel p0.s, p13.s/Z, p10.s[w12]\n" - ".inst 0xe09606e5 // ld1w { za1h.s[x12, #1] }, p1/Z, [x23, x22, LSL #2]\n" + ".inst 0xe09706a5 // ld1w { za1h.s[x12, #1] }, p1/Z, [x21, x23, LSL #2]\n" "ldr x25, [x11, #0x8]\n" ".inst 0x25307542 // psel p2.s, p13.s/Z, p10.s[w12]\n" - "ldr x23, [x9, #0x8]\n" - ".inst 0xe0bf82a8 // st1w { za2v.s[x12] }, p0/Z, [x21, XZR, LSL #2]\n" + "ldr x21, [x9, #0x8]\n" + ".inst 0xe0bf82c8 // st1w { za2v.s[x12] }, p0/Z, [x22, XZR, LSL #2]\n" ".inst 0x25707541 // psel p1.s, p13.s/Z, p10.s[w12, #1]\n" ".inst 0x25707540 // psel p0.s, p13.s/Z, p10.s[w12, #1]\n" - ".inst 0xe0b08aac // st1w { za3v.s[x12] }, p2/Z, [x21, x16, LSL #2]\n" + ".inst 0xe0b08acc // st1w { za3v.s[x12] }, p2/Z, [x22, x16, LSL #2]\n" "whilelt p10.s, x13, %x[width]\n" "subs x20, x20, #0x1\n" - ".inst 0xe0bc86a9 // st1w { za2v.s[x12, #1] }, p1/Z, [x21, x28, LSL #2]\n" + ".inst 0xe0bc86c9 // st1w { za2v.s[x12, #1] }, p1/Z, [x22, x28, LSL #2]\n" "add x11, x11, #0x10\n" "add x9, x9, #0x10\n" - ".inst 0xe0ba82ad // st1w { za3v.s[x12, #1] }, p0/Z, [x21, x26, LSL #2]\n" - "addvl x21, x21, #4\n" + ".inst 0xe0ba82cd // st1w { za3v.s[x12, #1] }, p0/Z, [x22, x26, LSL #2]\n" + "addvl x22, x22, #4\n" "incw x13\n" - "incw x22\n" + "incw x23\n" "bgt 3b\n" "8:" // K loop: Tails "cbnz x24, 11f\n" @@ -238,48 +236,48 @@ void interleave_block<2, 1, VLType::SME, false>( "mov x12, #0x0\n" "9:" // K loop: Tails: Even: First ".inst 0x25307540 // psel p0.s, p13.s/Z, p10.s[w12]\n" - ".inst 0xe0bf82a0 // st1w { za0v.s[x12] }, p0/Z, [x21, XZR, LSL #2]\n" + ".inst 0xe0bf82c0 // st1w { za0v.s[x12] }, p0/Z, [x22, XZR, LSL #2]\n" ".inst 0x25307540 // psel p0.s, p13.s/Z, p10.s[w12]\n" - ".inst 0xe0b082a4 // st1w { za1v.s[x12] }, p0/Z, [x21, x16, LSL #2]\n" - "ldr x10, [x11, #0x0]\n" + ".inst 0xe0b082c4 // st1w { za1v.s[x12] }, p0/Z, [x22, x16, LSL #2]\n" + "ldr x21, [x11, #0x0]\n" ".inst 0x25306581 // psel p1.s, p9.s/Z, p12.s[w12]\n" ".inst 0x25306160 // psel p0.s, p8.s/Z, p11.s[w12]\n" - "ldr x27, [x11, x16, LSL #0x3]\n" - ".inst 0xe0960548 // ld1w { za2h.s[x12] }, p1/Z, [x10, x22, LSL #2]\n" + "ldr x20, [x11, x16, LSL #0x3]\n" + ".inst 0xe09706a8 // ld1w { za2h.s[x12] }, p1/Z, [x21, x23, LSL #2]\n" "add x11, x11, #0x8\n" - "addvl x21, x21, #2\n" - ".inst 0xe096036c // ld1w { za3h.s[x12] }, p0/Z, [x27, x22, LSL #2]\n" + "addvl x22, x22, #2\n" + ".inst 0xe097028c // ld1w { za3h.s[x12] }, p0/Z, [x20, x23, LSL #2]\n" "add x12, x12, #0x1\n" "cmp x12, x16\n" "blt 9b\n" "whilelt p10.s, x13, %x[width]\n" - "whilelt p9.s, x13, %x[width]\n" + "whilelt p8.s, x13, %x[width]\n" "whilelt p8.s, x13, %x[width]\n" "mov x12, #0x0\n" "10:" // K loop: Tails: Even: Second ".inst 0x25307540 // psel p0.s, p13.s/Z, p10.s[w12]\n" - ".inst 0xe0bf82a8 // st1w { za2v.s[x12] }, p0/Z, [x21, XZR, LSL #2]\n" + ".inst 0xe0bf82c8 // st1w { za2v.s[x12] }, p0/Z, [x22, XZR, LSL #2]\n" ".inst 0x25307540 // psel p0.s, p13.s/Z, p10.s[w12]\n" - ".inst 0xe0b082ac // st1w { za3v.s[x12] }, p0/Z, [x21, x16, LSL #2]\n" + ".inst 0xe0b082cc // st1w { za3v.s[x12] }, p0/Z, [x22, x16, LSL #2]\n" "add x12, x12, #0x1\n" "cmp x12, x15\n" - "addvl x21, x21, #2\n" + "addvl x22, x22, #2\n" "blt 10b\n" - "whilelt p10.s, x13, %x[width]\n" + "whilelt p8.s, x13, %x[width]\n" "b 13f\n" "11:" // K loop: Tails: Odd "mov x12, #0x0\n" "12:" // K loop: Tails: Odd: Loop ".inst 0x25307540 // psel p0.s, p13.s/Z, p10.s[w12]\n" - ".inst 0xe0bf82a0 // st1w { za0v.s[x12] }, p0/Z, [x21, XZR, LSL #2]\n" + ".inst 0xe0bf82c0 // st1w { za0v.s[x12] }, p0/Z, [x22, XZR, LSL #2]\n" ".inst 0x25307540 // psel p0.s, p13.s/Z, p10.s[w12]\n" - ".inst 0xe0b082a4 // st1w { za1v.s[x12] }, p0/Z, [x21, x16, LSL #2]\n" + ".inst 0xe0b082c4 // st1w { za1v.s[x12] }, p0/Z, [x22, x16, LSL #2]\n" "add x12, x12, #0x1\n" "cmp x12, x15\n" - "addvl x21, x21, #2\n" + "addvl x22, x22, #2\n" "blt 12b\n" "13:" // K loop: End - "mov %x[out], x21\n" + "mov %x[out], x22\n" ".inst 0xd503467f // SMSTOP\n" : [out] "+&r" (out) : [height] "r" (height), [in] "r" (in), [row_offset] "r" (row_offset), [width] "r" (width) @@ -287,4 +285,4 @@ void interleave_block<2, 1, VLType::SME, false>( ); } -#endif // defined(__ARM_FEATURE_SVE) +#endif // defined(ARM_COMPUTE_ENABLE_SME) |