From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:49491) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1dJCc3-0006GU-Gw for qemu-devel@nongnu.org; Fri, 09 Jun 2017 01:37:36 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1dJCc0-0001eW-Ln for qemu-devel@nongnu.org; Fri, 09 Jun 2017 01:37:35 -0400 Received: from mail-qt0-x243.google.com ([2607:f8b0:400d:c0d::243]:34729) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1dJCc0-0001eP-HL for qemu-devel@nongnu.org; Fri, 09 Jun 2017 01:37:32 -0400 Received: by mail-qt0-x243.google.com with SMTP id o21so13030893qtb.1 for ; Thu, 08 Jun 2017 22:37:32 -0700 (PDT) Sender: Richard Henderson From: Richard Henderson Date: Thu, 8 Jun 2017 22:37:19 -0700 Message-Id: <20170609053719.26251-8-rth@twiddle.net> In-Reply-To: <20170609053719.26251-1-rth@twiddle.net> References: <20170609053719.26251-1-rth@twiddle.net> Subject: [Qemu-devel] [PATCH v5 7/7] tcg/arm: Use ldr (literal) for goto_tb List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: cota@braap.org The new placement of the TB means that we can use one insn to load the goto_tb destination directly from the TB. Signed-off-by: Richard Henderson --- tcg/arm/tcg-target.inc.c | 23 ++++++++++++++++++----- 1 file changed, 18 insertions(+), 5 deletions(-) diff --git a/tcg/arm/tcg-target.inc.c b/tcg/arm/tcg-target.inc.c index 18708b1..b640fb9 100644 --- a/tcg/arm/tcg-target.inc.c +++ b/tcg/arm/tcg-target.inc.c @@ -1669,14 +1669,27 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, } break; case INDEX_op_goto_tb: - tcg_debug_assert(s->tb_jmp_insn_offset == 0); { /* Indirect jump method */ - intptr_t ptr = (intptr_t)(s->tb_jmp_target_addr + args[0]); - tcg_out_movi32(s, COND_AL, TCG_REG_R0, ptr & ~0xfff); - tcg_out_ld32_12(s, COND_AL, TCG_REG_PC, TCG_REG_R0, ptr & 0xfff); + intptr_t ptr, dif, dil; + TCGReg base = TCG_REG_PC; + + tcg_debug_assert(s->tb_jmp_insn_offset == 0); + ptr = (intptr_t)(s->tb_jmp_target_addr + args[0]); + dif = ptr - ((intptr_t)s->code_ptr + 8); + dil = sextract32(dif, 0, 12); + if (dif != dil) { + /* The TB is close, but outside the 12 bits addressable by + the load. We can extend this to 20 bits with a sub of a + shifted immediate from pc. In the vastly unlikely event + the code requires more than 1MB, we'll use 2 insns and + be no worse off. */ + base = TCG_REG_R0; + tcg_out_movi32(s, COND_AL, base, ptr - dil); + } + tcg_out_ld32_12(s, COND_AL, TCG_REG_PC, base, dil); + s->tb_jmp_reset_offset[args[0]] = tcg_current_code_size(s); } - s->tb_jmp_reset_offset[args[0]] = tcg_current_code_size(s); break; case INDEX_op_goto_ptr: tcg_out_bx(s, COND_AL, args[0]); -- 2.9.4