From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:53520) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1cxYqu-0002Vo-No for qemu-devel@nongnu.org; Mon, 10 Apr 2017 08:55:29 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1cxYqt-0003Ai-Jq for qemu-devel@nongnu.org; Mon, 10 Apr 2017 08:55:28 -0400 Received: from mail-wr0-x22f.google.com ([2a00:1450:400c:c0c::22f]:35923) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1cxYqt-0003AM-DI for qemu-devel@nongnu.org; Mon, 10 Apr 2017 08:55:27 -0400 Received: by mail-wr0-x22f.google.com with SMTP id c55so74646367wrc.3 for ; Mon, 10 Apr 2017 05:55:27 -0700 (PDT) From: =?UTF-8?q?Alex=20Benn=C3=A9e?= Date: Mon, 10 Apr 2017 13:55:22 +0100 Message-Id: <20170410125524.21008-10-alex.bennee@linaro.org> In-Reply-To: <20170410125524.21008-1-alex.bennee@linaro.org> References: <20170410125524.21008-1-alex.bennee@linaro.org> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Subject: [Qemu-devel] [PULL 09/11] cpu-exec: update icount after each TB_EXIT List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: peter.maydell@linaro.org Cc: qemu-devel@nongnu.org, =?UTF-8?q?Alex=20Benn=C3=A9e?= , Paolo Bonzini , Peter Crosthwaite , Richard Henderson There is no particular reason we shouldn't update the global system icount time as we exit each TranslationBlock run. This ensures the main-loop doesn't have to wait until we exit to the outer loop for executed instructions to be credited to timer_state. The prepare_icount_for_run function is slightly tweaked to match the logic we run in cpu_loop_exec_tb. Based on Paolo's original suggestion. Signed-off-by: Alex Bennée diff --git a/cpu-exec.c b/cpu-exec.c index 748cb66bca..63a56d0407 100644 --- a/cpu-exec.c +++ b/cpu-exec.c @@ -600,13 +600,13 @@ static inline void cpu_loop_exec_tb(CPUState *cpu, TranslationBlock *tb, /* Instruction counter expired. */ assert(use_icount); #ifndef CONFIG_USER_ONLY - if (cpu->icount_extra) { - /* Refill decrementer and continue execution. */ - cpu->icount_extra += insns_left; - insns_left = MIN(0xffff, cpu->icount_extra); - cpu->icount_extra -= insns_left; - cpu->icount_decr.u16.low = insns_left; - } else { + /* Ensure global icount has gone forward */ + cpu_update_icount(cpu); + /* Refill decrementer and continue execution. */ + insns_left = MIN(0xffff, cpu->icount_budget); + cpu->icount_decr.u16.low = insns_left; + cpu->icount_extra = cpu->icount_budget - insns_left; + if (!cpu->icount_extra) { /* Execute any remaining instructions, then let the main loop * handle the next event. */ diff --git a/cpus.c b/cpus.c index a5125d7167..9c8bd2c991 100644 --- a/cpus.c +++ b/cpus.c @@ -1211,8 +1211,7 @@ static void handle_icount_deadline(void) static void prepare_icount_for_run(CPUState *cpu) { if (use_icount) { - int64_t count; - int decr; + int insns_left; /* These should always be cleared by process_icount_data after * each vCPU execution. However u16.high can be raised @@ -1221,17 +1220,10 @@ static void prepare_icount_for_run(CPUState *cpu) g_assert(cpu->icount_decr.u16.low == 0); g_assert(cpu->icount_extra == 0); - - count = tcg_get_icount_limit(); - - /* To calculate what we have executed so far we need to know - * what we originally budgeted to run this cycle */ - cpu->icount_budget = count; - - decr = (count > 0xffff) ? 0xffff : count; - count -= decr; - cpu->icount_decr.u16.low = decr; - cpu->icount_extra = count; + cpu->icount_budget = tcg_get_icount_limit(); + insns_left = MIN(0xffff, cpu->icount_budget); + cpu->icount_decr.u16.low = insns_left; + cpu->icount_extra = cpu->icount_budget - insns_left; } } -- 2.11.0