From patchwork Sat Oct 6 12:30:21 2012 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Paolo Bonzini X-Patchwork-Id: 189688 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Received: from lists.gnu.org (lists.gnu.org [208.118.235.17]) (using TLSv1 with cipher AES256-SHA (256/256 bits)) (Client did not present a certificate) by ozlabs.org (Postfix) with ESMTPS id 4414A2C0333 for ; Sat, 6 Oct 2012 22:32:28 +1000 (EST) Received: from localhost ([::1]:41735 helo=lists.gnu.org) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1TKTYQ-0008IW-BH for incoming@patchwork.ozlabs.org; Sat, 06 Oct 2012 08:32:26 -0400 Received: from eggs.gnu.org ([208.118.235.92]:49787) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1TKTWn-0004dh-UJ for qemu-devel@nongnu.org; Sat, 06 Oct 2012 08:30:47 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1TKTWm-0000MY-Dm for qemu-devel@nongnu.org; Sat, 06 Oct 2012 08:30:45 -0400 Received: from mail-wi0-f175.google.com ([209.85.212.175]:33758) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1TKTWm-00007X-3g for qemu-devel@nongnu.org; Sat, 06 Oct 2012 08:30:44 -0400 Received: by mail-wi0-f175.google.com with SMTP id hq4so1421103wib.10 for ; Sat, 06 Oct 2012 05:30:43 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=sender:from:to:subject:date:message-id:x-mailer:in-reply-to :references; bh=eHDqATgVNZ0ZfQVCrxjiuxGQvp1Qy7y36MEYSastbGw=; b=a43FCKpEnrf7LrbYAadUYzvcytoYhNDed6b9wpQZvbVHC9D4+3lT0QQcF1YCcPtpwo AalL3ZmCVOX5v6nCPMRkpTzGmwKDgD6GtGaVBYqBlx9XwOFU8ZfKr8FocBViTZXGvSsh FY+lIobPU97WyYnF/JovdqWS/SA1WTf0NVbzuuW5eh9yHlA6cMIP57muI2rNaDWaL22W JmcNuXuaeiAhTKsPOcMFBIhX6Nd0MbIH7Uty2dxBIA+Wh71FaYm70mxWXaQwBTvbN4yq KoT0WMOpefMvtazC5kCjaUEbd0pfxfOvz7yDPRskKfVNqoYPhqA69e/SOn8ZCzhH7CtP RmaQ== Received: by 10.216.211.106 with SMTP id v84mr7236855weo.145.1349526643687; Sat, 06 Oct 2012 05:30:43 -0700 (PDT) Received: from yakj.lan (93-34-169-1.ip50.fastwebnet.it. [93.34.169.1]) by mx.google.com with ESMTPS id m14sm7950701wie.8.2012.10.06.05.30.42 (version=TLSv1/SSLv3 cipher=OTHER); Sat, 06 Oct 2012 05:30:42 -0700 (PDT) From: Paolo Bonzini To: qemu-devel@nongnu.org Date: Sat, 6 Oct 2012 14:30:21 +0200 Message-Id: <1349526621-13939-15-git-send-email-pbonzini@redhat.com> X-Mailer: git-send-email 1.7.12.1 In-Reply-To: <1349526621-13939-1-git-send-email-pbonzini@redhat.com> References: <1349526621-13939-1-git-send-email-pbonzini@redhat.com> X-detected-operating-system: by eggs.gnu.org: Genre and OS details not recognized. X-Received-From: 209.85.212.175 Subject: [Qemu-devel] [PATCH 14/14] i386: optimize setcc instructions X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.14 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org Sender: qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org Reconstruct the arguments for complex conditions involving CC_OP_SUBx (BE, L, LE). In the others do it via setcond and gen_setcc_slow (which is not that slow in many cases). Signed-off-by: Paolo Bonzini --- target-i386/translate.c | 93 +++++++++++++++++++------------------------------ 1 file modificato, 36 inserzioni(+), 57 rimozioni(-) diff --git a/target-i386/translate.c b/target-i386/translate.c index 342b9ec..92e8291 100644 --- a/target-i386/translate.c +++ b/target-i386/translate.c @@ -1063,55 +1063,55 @@ static inline void gen_setcc_slow(DisasContext *s, int jcc_op, TCGv reg, bool in } } -/* return true if setcc_slow is not needed (WARNING: must be kept in - sync with gen_jcc1) */ -static int is_fast_jcc_case(DisasContext *s, int b) +/* perform a conditional store into register 'reg' according to jump opcode + value 'b'. In the fast case, T0 is guaranted not to be used. */ +static inline void gen_setcc1(DisasContext *s, int b, TCGv reg) { - int jcc_op; + int inv, jcc_op, size, cond; + TCGv t0; + + inv = b & 1; jcc_op = (b >> 1) & 7; + switch(s->cc_op) { - /* we optimize the cmp/jcc case */ + /* we optimize relational operators for the cmp/jcc case */ case CC_OP_SUBB: case CC_OP_SUBW: case CC_OP_SUBL: case CC_OP_SUBQ: - if (jcc_op == JCC_O || jcc_op == JCC_P) - goto slow_jcc; - break; - - /* some jumps are easy to compute */ - case CC_OP_ADDB: - case CC_OP_ADDW: - case CC_OP_ADDL: - case CC_OP_ADDQ: - - case CC_OP_LOGICB: - case CC_OP_LOGICW: - case CC_OP_LOGICL: - case CC_OP_LOGICQ: - - case CC_OP_INCB: - case CC_OP_INCW: - case CC_OP_INCL: - case CC_OP_INCQ: + size = s->cc_op - CC_OP_SUBB; + switch(jcc_op) { + case JCC_BE: + cond = inv ? TCG_COND_GTU : TCG_COND_LEU; + tcg_gen_add_tl(cpu_tmp4, cpu_cc_dst, cpu_cc_src); + gen_extu(size, cpu_tmp4); + t0 = gen_ext_tl(cpu_tmp0, cpu_cc_src, size, false); + tcg_gen_setcond_tl(cond, reg, cpu_tmp4, t0); + break; - case CC_OP_DECB: - case CC_OP_DECW: - case CC_OP_DECL: - case CC_OP_DECQ: + case JCC_L: + cond = inv ? TCG_COND_GE : TCG_COND_LT; + goto fast_jcc_l; + case JCC_LE: + cond = inv ? TCG_COND_GT : TCG_COND_LE; + fast_jcc_l: + tcg_gen_add_tl(cpu_tmp4, cpu_cc_dst, cpu_cc_src); + gen_exts(size, cpu_tmp4); + t0 = gen_ext_tl(cpu_tmp0, cpu_cc_src, size, true); + tcg_gen_setcond_tl(cond, reg, cpu_tmp4, t0); + break; - case CC_OP_SHLB: - case CC_OP_SHLW: - case CC_OP_SHLL: - case CC_OP_SHLQ: - if (jcc_op != JCC_Z && jcc_op != JCC_S) + default: goto slow_jcc; + } break; + default: slow_jcc: - return 0; + /* gen_setcc_slow actually generates good code for JC, JZ and JS */ + gen_setcc_slow(s, jcc_op, reg, inv); + break; } - return 1; } /* generate a conditional jump to label 'l1' according to jump opcode @@ -2477,28 +2477,7 @@ static inline void gen_jcc(DisasContext *s, int b, static void gen_setcc(DisasContext *s, int b) { - int inv, jcc_op, l1; - TCGv t0; - - if (is_fast_jcc_case(s, b)) { - /* nominal case: we use a jump */ - /* XXX: make it faster by adding new instructions in TCG */ - t0 = tcg_temp_local_new(); - tcg_gen_movi_tl(t0, 0); - l1 = gen_new_label(); - gen_jcc1(s, b ^ 1, l1); - tcg_gen_movi_tl(t0, 1); - gen_set_label(l1); - tcg_gen_mov_tl(cpu_T[0], t0); - tcg_temp_free(t0); - } else { - /* slow case: it is more efficient not to generate a jump, - although it is questionnable whether this optimization is - worth to */ - inv = b & 1; - jcc_op = (b >> 1) & 7; - gen_setcc_slow(s, jcc_op, cpu_T[0], inv); - } + gen_setcc1(s, b, cpu_T[0]); } static inline void gen_op_movl_T0_seg(int seg_reg)