Implement BPF_ALU+BPF_MOD-BPF_K when pc->k is a power of 2. Get rid of divt

and divw arguments in emit_moddiv(), they're accessible via the pc argument.
This commit is contained in:
alnsn 2014-11-20 19:18:52 +00:00
parent b59f66e17c
commit e7d5850261

View File

@ -1,4 +1,4 @@
/* $NetBSD: bpfjit.c,v 1.34 2014/11/20 14:35:01 alnsn Exp $ */
/* $NetBSD: bpfjit.c,v 1.35 2014/11/20 19:18:52 alnsn Exp $ */
/*-
* Copyright (c) 2011-2014 Alexander Nasonov.
@ -31,9 +31,9 @@
#include <sys/cdefs.h>
#ifdef _KERNEL
__KERNEL_RCSID(0, "$NetBSD: bpfjit.c,v 1.34 2014/11/20 14:35:01 alnsn Exp $");
__KERNEL_RCSID(0, "$NetBSD: bpfjit.c,v 1.35 2014/11/20 19:18:52 alnsn Exp $");
#else
__RCSID("$NetBSD: bpfjit.c,v 1.34 2014/11/20 14:35:01 alnsn Exp $");
__RCSID("$NetBSD: bpfjit.c,v 1.35 2014/11/20 19:18:52 alnsn Exp $");
#endif
#include <sys/types.h>
@ -1087,25 +1087,43 @@ emit_msh(struct sljit_compiler *compiler, bpfjit_hint_t hints,
return SLJIT_SUCCESS;
}
/*
* Emit code for A = A / k or A = A % k when k is a power of 2.
* @pc BPF_DIV or BPF_MOD instruction.
*/
static int
emit_pow2_division(struct sljit_compiler *compiler, uint32_t k)
emit_pow2_moddiv(struct sljit_compiler *compiler, const struct bpf_insn *pc)
{
int shift = 0;
uint32_t k = pc->k;
int status = SLJIT_SUCCESS;
while (k > 1) {
k >>= 1;
shift++;
}
BJ_ASSERT(k != 0 && (k & (k - 1)) == 0);
BJ_ASSERT(k == 1 && shift < 32);
if (shift != 0) {
if (BPF_OP(pc->code) == BPF_MOD) {
status = sljit_emit_op2(compiler,
SLJIT_LSHR|SLJIT_INT_OP,
SLJIT_AND,
BJ_AREG, 0,
BJ_AREG, 0,
SLJIT_IMM, shift);
SLJIT_IMM, k - 1);
} else {
int shift = 0;
/*
* Do shift = __builtin_ctz(k).
* The loop is slower, but that's ok.
*/
while (k > 1) {
k >>= 1;
shift++;
}
if (shift != 0) {
status = sljit_emit_op2(compiler,
SLJIT_LSHR|SLJIT_INT_OP,
BJ_AREG, 0,
BJ_AREG, 0,
SLJIT_IMM, shift);
}
}
return status;
@ -1128,15 +1146,15 @@ modulus(sljit_uw x, sljit_uw y)
#endif
/*
* Emit code for A = A / div or A = A % div
* divt,divw are either SLJIT_IMM,pc->k or BJ_XREG,0.
* Emit code for A = A / div or A = A % div.
* @pc BPF_DIV or BPF_MOD instruction.
*/
static int
emit_moddiv(struct sljit_compiler *compiler,
const struct bpf_insn *pc, int divt, sljit_sw divw)
emit_moddiv(struct sljit_compiler *compiler, const struct bpf_insn *pc)
{
int status;
const bool div = BPF_OP(pc->code) == BPF_DIV;
const bool xreg = BPF_SRC(pc->code) == BPF_X;
#if BJ_XREG == SLJIT_RETURN_REG || \
BJ_XREG == SLJIT_SCRATCH_REG1 || \
@ -1157,7 +1175,8 @@ emit_moddiv(struct sljit_compiler *compiler,
status = sljit_emit_op1(compiler,
SLJIT_MOV,
SLJIT_SCRATCH_REG2, 0,
divt, divw);
xreg ? BJ_XREG : SLJIT_IMM,
xreg ? 0 : (uint32_t)pc->k);
if (status != SLJIT_SUCCESS)
return status;
@ -1940,17 +1959,14 @@ generate_insn_code(struct sljit_compiler *compiler, bpfjit_hint_t hints,
}
if (src == BPF_X) {
status = emit_moddiv(compiler, pc, BJ_XREG, 0);
status = emit_moddiv(compiler, pc);
if (status != SLJIT_SUCCESS)
goto fail;
} else if (pc->k != 0) {
/* XXX: We can do better here for MOD */
if ((pc->k & (pc->k - 1)) || op == BPF_MOD) {
status = emit_moddiv(compiler, pc,
SLJIT_IMM, (uint32_t)pc->k);
if (pc->k & (pc->k - 1)) {
status = emit_moddiv(compiler, pc);
} else {
status = emit_pow2_division(compiler,
(uint32_t)pc->k);
status = emit_pow2_moddiv(compiler, pc);
}
if (status != SLJIT_SUCCESS)
goto fail;