diff --git a/lib/libc/riscv/string/Makefile b/lib/libc/riscv/string/Makefile new file mode 100644 --- /dev/null +++ b/lib/libc/riscv/string/Makefile @@ -0,0 +1,2 @@ +MDSRCS+= \ + strnlen.S diff --git a/lib/libc/riscv/string/strnlen.S b/lib/libc/riscv/string/strnlen.S new file mode 100644 --- /dev/null +++ b/lib/libc/riscv/string/strnlen.S @@ -0,0 +1,144 @@ +/*- + * SPDX-License-Identifier: BSD-2-Clause + * + * Copyright (c) 2024 Strahinja Stanisic + */ + +#include + +/* + * a0 - const void* b + * a1 - int c; + * a2 - size_t len; + */ +ENTRY(strnlen) + /* + * a0 - const char* s; + * a1 - size_t maxlen; + * a2 - uint64_t* ptr; + * a3 - char iter[8]; + * a4 - uint64_t* end_align; + * a5 - uint64_t* end_unroll; + */ + + beqz a1, .Lnot_found + + /* ptr = s & ~0b111 */ + /* t0 = 0x0101010101010101 */ + /* t1 = 0x8080808080808080 */ + /* end_align = (s + maxlen + 7) & ~0b111 */ + /* mask_start = t0 >> ((-s.value) << 3) */ + add a4, a0, a1 + li t0, 0x01010101 + addi a4, a4, 7 + slli t1, t0, 32 + neg t2, a0 + andi a4, a4, ~0b111 + or t0, t0, t1 + slli t2, t2, 3 + andi a2, a0, ~0b111 + slli t1, t0, 7 + srl t2, t0, t2 + + /* if pointer is aligned skip to loop */ + beq a0, a2, .Lskip_start + + /* iter = *ptr */ + ld a3, (a2) + + /* iter = iter | mask_start */ + or a3, a3, t2 + + /* has_zero */ + not t2, a3 + sub a3, a3, t0 + and t2, t2, t1 + and a3, a3, t2 + + addi a2, a2, 8 + bnez a3, .Lfind_zero + +.Lskip_start: + /* end_unroll */ + sub t2, a4, a2 + andi t2, t2, ~0b1111 + add a5, a2, t2 + + /* while (ptr != end_unroll) */ + beq a2, a5, .Lskip_loop +.Lloop: + ld a3, (a2) + ld a6, 8(a2) + + /* has_zero */ + not t2, a3 + not t3, a6 + sub a3, a3, t0 + sub a6, a6, t0 + and t2, t2, t1 + and t3, t3, t1 + and a3, a3, t2 + and a6, a6, t3 + + addi a2, a2, 8 + bnez a3, .Lfind_zero + + mv a3, a6 + + addi a2, a2, 8 + bnez a3, .Lfind_zero + + bne a2, a5, .Lloop + +.Lskip_loop: + + beq a2, a4, .Lnot_found + + ld a3, (a2) + + /* has_zero */ + not t2, a3 + sub a3, a3, t0 + and t2, t2, t1 + and a3, a3, t2 + + + addi a2, a2, 8 + beqz a3, .Lnot_found + +.Lfind_zero: + + /* move ptr back */ + addi a2, a2, -8 + + /* isolate lowest set bit */ + neg t0, a3 + and a3, a3, t0 + + li t0, 0x0001020304050607 + srli a3, a3, 7 + + /* lowest set bit is 2^(8*k) + * multiplying by it shifts the idx array in t0 by k bytes to the left */ + mul a3, a3, t0 + + /* highest byte contains idx of first zero */ + srli a3, a3, 56 + + /* zero_idx */ + sub a2, a2, a0 + add a2, a2, a3 + + /* min(zero_idx, maxlen) */ + sub a2, a2, a1 + srai t1, a2, 63 + and a2, a2, t1 + add a0, a1, a2 + + ret + +.Lnot_found: + mv a0, a1 + ret + +END(strnlen) diff --git a/share/man/man7/simd.7 b/share/man/man7/simd.7 --- a/share/man/man7/simd.7 +++ b/share/man/man7/simd.7 @@ -50,7 +50,7 @@ .Pp Enhanced functions are present for the following architectures: .Bl -column FUNCTION_________ aarch64_ arm_ amd64_ i386_ ppc64_ -offset indent -.It Em FUNCTION Ta Em AARCH64 Ta Em ARM Ta Em AMD64 Ta Em I386 Ta Em PPC64 +.It Em FUNCTION Ta Em AARCH64 Ta Em ARM Ta Em AMD64 Ta Em I386 Ta Em PPC64 Ta Em RISC-V .It bcmp Ta Ta Ta S1 Ta S .It bcopy Ta Ta S Ta S Ta S Ta SV .It bzero Ta Ta S Ta S Ta S @@ -80,7 +80,7 @@ .It strncat Ta Ta Ta S1 .It strncmp Ta S Ta S Ta S1 Ta S .It strncpy Ta Ta Ta S1 Ta Ta S2 -.It strnlen Ta A Ta Ta S1 +.It strnlen Ta A Ta Ta S1 Ta Ta Ta S .It strrchr Ta A Ta Ta S1 Ta S .It strpbrk Ta Ta Ta S2 .It strsep Ta Ta Ta S2