aboutsummaryrefslogtreecommitdiff
path: root/src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave1VL_block2_fp16_fp16.hpp
diff options
context:
space:
mode:
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave1VL_block2_fp16_fp16.hpp')
-rw-r--r--src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave1VL_block2_fp16_fp16.hpp56
1 files changed, 28 insertions, 28 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave1VL_block2_fp16_fp16.hpp b/src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave1VL_block2_fp16_fp16.hpp
index 0e0e4e462c..982cfa6d40 100644
--- a/src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave1VL_block2_fp16_fp16.hpp
+++ b/src/core/NEON/kernels/arm_gemm/indirect-interleaves/sme_interleave1VL_block2_fp16_fp16.hpp
@@ -22,7 +22,7 @@
* SOFTWARE.
*/
-#if defined(ARM_COMPUTE_ENABLE_SME)
+#if defined(__ARM_FEATURE_SVE)
template <>
void interleave_block<1, 2, VLType::SME, false>(
@@ -63,11 +63,11 @@ void interleave_block<1, 2, VLType::SME, false>(
"mov x12, #0x0\n"
"cbz x28, 2f\n"
"1:" // K loop: Charge: Loop
- ".inst 0x25286141 // psel p1.h, p8.h/Z, p10.h[w12]\n"
- ".inst 0x25686140 // psel p0.h, p8.h/Z, p10.h[w12, #2]\n"
- ".inst 0xe0560700 // ld1h { za0h.h[x12] }, p1/Z, [x24, x22, LSL #1]\n"
+ ".inst 0x25286143 // psel p3.h, p8.h/Z, p10.h[w12]\n"
+ ".inst 0x25686142 // psel p2.h, p8.h/Z, p10.h[w12, #2]\n"
+ ".inst 0xe0560f00 // ld1h { za0h.h[x12] }, p3/Z, [x24, x22, LSL #1]\n"
"ldr x24, [x26, #0x0]\n"
- ".inst 0xe05602e2 // ld1h { za0h.h[x12, #2] }, p0/Z, [x23, x22, LSL #1]\n"
+ ".inst 0xe0560ae2 // ld1h { za0h.h[x12, #2] }, p2/Z, [x23, x22, LSL #1]\n"
"add x12, x12, #0x4\n"
"ldr x23, [x26, #0x8]\n"
"add x26, x26, #0x10\n"
@@ -88,20 +88,20 @@ void interleave_block<1, 2, VLType::SME, false>(
"mov x20, x9\n"
"3:" // K loop: Main loop
"whilelt p8.h, x27, %x[width]\n"
- "mov x12, #0x0\n"
+ "mov x15, #0x0\n"
"mov x14, #0x0\n"
"cbz x28, 5f\n"
"4:" // K loop: Main loop: First: Loop
- ".inst 0x25386143 // psel p3.h, p8.h/Z, p10.h[w12, #1]\n"
- ".inst 0x25786142 // psel p2.h, p8.h/Z, p10.h[w12, #3]\n"
+ ".inst 0x253b6143 // psel p3.h, p8.h/Z, p10.h[w15, #1]\n"
+ ".inst 0x257b6142 // psel p2.h, p8.h/Z, p10.h[w15, #3]\n"
".inst 0x252a6d21 // psel p1.h, p11.h/Z, p9.h[w14]\n"
".inst 0x253a6d20 // psel p0.h, p11.h/Z, p9.h[w14, #1]\n"
- ".inst 0xe0560f01 // ld1h { za0h.h[x12, #1] }, p3/Z, [x24, x22, LSL #1]\n"
+ ".inst 0xe0566f01 // ld1h { za0h.h[x15, #1] }, p3/Z, [x24, x22, LSL #1]\n"
"ldr x24, [x26, #0x0]\n"
- ".inst 0xe0560ae3 // ld1h { za0h.h[x12, #3] }, p2/Z, [x23, x22, LSL #1]\n"
+ ".inst 0xe0566ae3 // ld1h { za0h.h[x15, #3] }, p2/Z, [x23, x22, LSL #1]\n"
"ldr x23, [x26, #0x8]\n"
"add x26, x26, #0x10\n"
- "add x12, x12, #0x4\n"
+ "add x15, x15, #0x4\n"
".inst 0xe0bfc6a0 // st1w { za0v.s[x14] }, p1/Z, [x21, XZR, LSL #2]\n"
".inst 0xe0aac2a1 // st1w { za0v.s[x14, #1] }, p0/Z, [x21, x10, LSL #2]\n"
"add x14, x14, #0x2\n"
@@ -109,23 +109,23 @@ void interleave_block<1, 2, VLType::SME, false>(
"cmp x14, x28\n"
"blt 4b\n"
"5:" // K loop: Main loop: First: Tail
- ".inst 0x25386143 // psel p3.h, p8.h/Z, p10.h[w12, #1]\n"
- ".inst 0x25786142 // psel p2.h, p8.h/Z, p10.h[w12, #3]\n"
+ ".inst 0x253b6143 // psel p3.h, p8.h/Z, p10.h[w15, #1]\n"
+ ".inst 0x257b6142 // psel p2.h, p8.h/Z, p10.h[w15, #3]\n"
".inst 0x252a6d21 // psel p1.h, p11.h/Z, p9.h[w14]\n"
".inst 0x253a6d20 // psel p0.h, p11.h/Z, p9.h[w14, #1]\n"
"mov x26, %x[in]\n"
"whilelt p9.h, x27, %x[width]\n"
- ".inst 0xe0560f01 // ld1h { za0h.h[x12, #1] }, p3/Z, [x24, x22, LSL #1]\n"
+ ".inst 0xe0566f01 // ld1h { za0h.h[x15, #1] }, p3/Z, [x24, x22, LSL #1]\n"
"ldr x24, [x26, #0x0]\n"
"inch x27\n"
"mov x13, #0x0\n"
- ".inst 0xe0560ae3 // ld1h { za0h.h[x12, #3] }, p2/Z, [x23, x22, LSL #1]\n"
+ "whilelt p8.h, x27, %x[width]\n"
+ "mov x12, #0x0\n"
+ ".inst 0xe0566ae3 // ld1h { za0h.h[x15, #3] }, p2/Z, [x23, x22, LSL #1]\n"
"ldr x23, [x26, #0x8]\n"
"add x26, x26, #0x10\n"
"inch x22\n"
".inst 0xe0bfc6a0 // st1w { za0v.s[x14] }, p1/Z, [x21, XZR, LSL #2]\n"
- "whilelt p8.h, x27, %x[width]\n"
- "mov x12, #0x0\n"
".inst 0xe0aac2a1 // st1w { za0v.s[x14, #1] }, p0/Z, [x21, x10, LSL #2]\n"
"addvl x21, x21, #2\n"
"cbz x28, 7f\n"
@@ -172,15 +172,15 @@ void interleave_block<1, 2, VLType::SME, false>(
"mov x13, #0x0\n"
"mov x12, #0x0\n"
"9:" // K loop: Tails: Even: First
- ".inst 0x25306d21 // psel p1.s, p11.s/Z, p9.s[w12]\n"
- ".inst 0x25396140 // psel p0.h, p8.h/Z, p10.h[w13, #1]\n"
- ".inst 0xe0bf86a0 // st1w { za0v.s[x12] }, p1/Z, [x21, XZR, LSL #2]\n"
+ ".inst 0x25306d20 // psel p0.s, p11.s/Z, p9.s[w12]\n"
+ ".inst 0x25396143 // psel p3.h, p8.h/Z, p10.h[w13, #1]\n"
+ ".inst 0xe0bf82a0 // st1w { za0v.s[x12] }, p0/Z, [x21, XZR, LSL #2]\n"
+ "ldr x24, [x26, #0x0]\n"
"add x12, x12, #0x1\n"
- "addvl x21, x21, #1\n"
- "ldr x20, [x26, #0x0]\n"
- "cmp x12, x10\n"
"add x26, x26, #0x8\n"
- ".inst 0xe0562281 // ld1h { za0h.h[x13, #1] }, p0/Z, [x20, x22, LSL #1]\n"
+ "cmp x12, x10\n"
+ "addvl x21, x21, #1\n"
+ ".inst 0xe0562f01 // ld1h { za0h.h[x13, #1] }, p3/Z, [x24, x22, LSL #1]\n"
"add x13, x13, #0x2\n"
"blt 9b\n"
"whilelt p9.h, x27, %x[width]\n"
@@ -195,7 +195,7 @@ void interleave_block<1, 2, VLType::SME, false>(
"addvl x21, x21, #1\n"
"cmp x12, x11\n"
"blt 10b\n"
- "whilelt p8.h, x27, %x[width]\n"
+ "whilelt p9.h, x27, %x[width]\n"
"b 13f\n"
"11:" // K loop: Tails: Odd
"mov x12, #0x0\n"
@@ -203,16 +203,16 @@ void interleave_block<1, 2, VLType::SME, false>(
".inst 0x25306d20 // psel p0.s, p11.s/Z, p9.s[w12]\n"
".inst 0xe0bf82a0 // st1w { za0v.s[x12] }, p0/Z, [x21, XZR, LSL #2]\n"
"add x12, x12, #0x1\n"
- "addvl x21, x21, #1\n"
"cmp x12, x11\n"
+ "addvl x21, x21, #1\n"
"blt 12b\n"
"13:" // K loop: End
"mov %x[out], x21\n"
".inst 0xd503467f // SMSTOP\n"
: [out] "+&r" (out)
: [height] "r" (height), [in] "r" (in), [row_offset] "r" (row_offset), [width] "r" (width)
- : "cc", "memory", "p0", "p1", "p2", "p3", "p4", "p5", "p6", "p7", "p8", "p9", "p10", "p11", "p12", "p13", "p14", "p15", "x9", "x10", "x11", "x12", "x13", "x14", "x20", "x21", "x22", "x23", "x24", "x25", "x26", "x27", "x28", "z0", "z1", "z2", "z3", "z4", "z5", "z6", "z7", "z8", "z9", "z10", "z11", "z12", "z13", "z14", "z15", "z16", "z17", "z18", "z19", "z20", "z21", "z22", "z23", "z24", "z25", "z26", "z27", "z28", "z29", "z30", "z31"
+ : "cc", "memory", "p0", "p1", "p2", "p3", "p4", "p5", "p6", "p7", "p8", "p9", "p10", "p11", "p12", "p13", "p14", "p15", "x9", "x10", "x11", "x12", "x13", "x14", "x15", "x20", "x21", "x22", "x23", "x24", "x25", "x26", "x27", "x28", "z0", "z1", "z2", "z3", "z4", "z5", "z6", "z7", "z8", "z9", "z10", "z11", "z12", "z13", "z14", "z15", "z16", "z17", "z18", "z19", "z20", "z21", "z22", "z23", "z24", "z25", "z26", "z27", "z28", "z29", "z30", "z31"
);
}
-#endif // defined(ARM_COMPUTE_ENABLE_SME)
+#endif // defined(__ARM_FEATURE_SVE)
generated by cgit v1.2.1 (git 2.23.0) at 2024-08-08 12:05:35 +0000