From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([140.186.70.92]:53152) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Qpx9q-00042g-Ag for qemu-devel@nongnu.org; Sun, 07 Aug 2011 02:48:23 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1Qpx9o-0006Mj-PD for qemu-devel@nongnu.org; Sun, 07 Aug 2011 02:48:22 -0400 Received: from v220110690675601.yourvserver.net ([78.47.199.172]:57285) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Qpx9o-0006MR-Bv for qemu-devel@nongnu.org; Sun, 07 Aug 2011 02:48:20 -0400 Message-ID: <4E3E3530.8000501@mail.berlios.de> Date: Sun, 07 Aug 2011 08:48:16 +0200 From: Stefan Weil MIME-Version: 1.0 References: In-Reply-To: Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH] TCG: fix copy propagation List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Blue Swirl Cc: TeLeMan , Kenneth Salerno , qemu-devel , zhur@ispras.ru, Kirill Batuzov Am 06.08.2011 23:26, schrieb Blue Swirl: > Copy propagation introduced in 22613af4a6d9602001e6d0e7b6d98aa40aa018dc > considered only global registers. However, register temps and stack > allocated locals must be handled differently because register temps > don't survive across brcond. > > Fix by propagating only within same class of temps. > > Signed-off-by: Blue Swirl > --- > tcg/optimize.c | 15 +++++++++------ > tcg/tcg.h | 5 +++++ > 2 files changed, 14 insertions(+), 6 deletions(-) > > diff --git a/tcg/optimize.c b/tcg/optimize.c > index a3bfa5e..7eb5eb1 100644 > --- a/tcg/optimize.c > +++ b/tcg/optimize.c > @@ -185,12 +185,15 @@ static int op_to_movi(int op) > } > } > > -static void tcg_opt_gen_mov(TCGArg *gen_args, TCGArg dst, TCGArg src, > - int nb_temps, int nb_globals) > +static void tcg_opt_gen_mov(TCGContext *s, TCGArg *gen_args, TCGArg dst, > + TCGArg src, int nb_temps, int nb_globals) > { > reset_temp(dst, nb_temps, nb_globals); > assert(temps[src].state != TCG_TEMP_COPY); > - if (src>= nb_globals) { > + /* Don't try to copy if one of temps is a global or either one > + is local and another is register */ > + if (src>= nb_globals&& dst>= nb_globals&& > + tcg_arg_is_local(s, src) == tcg_arg_is_local(s, dst)) { > assert(temps[src].state != TCG_TEMP_CONST); > if (temps[src].state != TCG_TEMP_HAS_COPY) { > temps[src].state = TCG_TEMP_HAS_COPY; > @@ -474,7 +477,7 @@ static TCGArg *tcg_constant_folding(TCGContext *s, > uint16_t *tcg_opc_ptr, > gen_opc_buf[op_index] = INDEX_op_nop; > } else { > gen_opc_buf[op_index] = op_to_mov(op); > - tcg_opt_gen_mov(gen_args, args[0], args[1], > + tcg_opt_gen_mov(s, gen_args, args[0], args[1], > nb_temps, nb_globals); > gen_args += 2; > args += 3; > @@ -500,7 +503,7 @@ static TCGArg *tcg_constant_folding(TCGContext *s, > uint16_t *tcg_opc_ptr, > gen_opc_buf[op_index] = INDEX_op_nop; > } else { > gen_opc_buf[op_index] = op_to_mov(op); > - tcg_opt_gen_mov(gen_args, args[0], args[1], nb_temps, > + tcg_opt_gen_mov(s, gen_args, args[0], args[1], nb_temps, > nb_globals); > gen_args += 2; > args += 3; > @@ -523,7 +526,7 @@ static TCGArg *tcg_constant_folding(TCGContext *s, > uint16_t *tcg_opc_ptr, > break; > } > if (temps[args[1]].state != TCG_TEMP_CONST) { > - tcg_opt_gen_mov(gen_args, args[0], args[1], > + tcg_opt_gen_mov(s, gen_args, args[0], args[1], > nb_temps, nb_globals); > gen_args += 2; > args += 2; > diff --git a/tcg/tcg.h b/tcg/tcg.h > index e76f9af..e2a7095 100644 > --- a/tcg/tcg.h > +++ b/tcg/tcg.h > @@ -410,6 +410,11 @@ static inline TCGv_i64 tcg_temp_local_new_i64(void) > void tcg_temp_free_i64(TCGv_i64 arg); > char *tcg_get_arg_str_i64(TCGContext *s, char *buf, int buf_size, > TCGv_i64 arg); > > +static inline bool tcg_arg_is_local(TCGContext *s, TCGArg arg) > +{ > + return s->temps[arg].temp_local; > +} > + > #if defined(CONFIG_DEBUG_TCG) > /* If you call tcg_clear_temp_count() at the start of a section of > * code which is not supposed to leak any TCG temporaries, then This fixes qemu-system-x86_64 and qemu-system-mips64(el) on 32 bit hosts. Tested-by: Stefan Weil