All of lore.kernel.org
 help / color / mirror / Atom feed
From: Kirill Batuzov <batuzovk@ispras.ru>
To: qemu-devel@nongnu.org
Cc: "Richard Henderson" <rth@twiddle.net>,
	"Paolo Bonzini" <pbonzini@redhat.com>,
	"Peter Crosthwaite" <crosthwaite.peter@gmail.com>,
	"Peter Maydell" <peter.maydell@linaro.org>,
	"Andrzej Zaborowski" <balrogg@gmail.com>,
	"Alex Bennée" <alex.bennee@linaro.org>,
	"Kirill Batuzov" <batuzovk@ispras.ru>
Subject: [Qemu-devel] [PATCH v2.1 06/21] tcg: use results of alias analysis in liveness analysis
Date: Thu,  2 Feb 2017 17:34:44 +0300	[thread overview]
Message-ID: <1486046099-17726-7-git-send-email-batuzovk@ispras.ru> (raw)
In-Reply-To: <1486046099-17726-1-git-send-email-batuzovk@ispras.ru>

Signed-off-by: Kirill Batuzov <batuzovk@ispras.ru>
---
 tcg/tcg.c | 61 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 61 insertions(+)

diff --git a/tcg/tcg.c b/tcg/tcg.c
index 18d97ec..27e5944 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -564,6 +564,11 @@ static intptr_t tcg_type_size(TCGType type)
     }
 }
 
+static intptr_t tcg_temp_size(const TCGTemp *tmp)
+{
+    return tcg_type_size(tmp->type);
+}
+
 int tcg_global_mem_new_internal(TCGType base_type, TCGv_ptr base,
                                 intptr_t offset, const char *name)
 {
@@ -1472,6 +1477,43 @@ static inline void tcg_la_bb_end(TCGContext *s, uint8_t *temp_state)
     }
 }
 
+/* Check if memory write completely overwrites temp's memory location.
+   If this is the case then the temp can be considered dead. */
+static int tcg_temp_overwrite(TCGContext *s, const TCGTemp *tmp,
+                               const TCGAliasInfo *ai)
+{
+    if (!(ai->alias_type & TCG_ALIAS_WRITE) || !ai->fixed_offset) {
+        return 0;
+    }
+    if (tmp->mem_base != &s->temps[GET_TCGV_PTR(s->tcg_env)]) {
+        return 0;
+    }
+    if (ai->offset > tmp->mem_offset
+        || ai->offset + ai->size < tmp->mem_offset + tcg_temp_size(tmp)) {
+            return 0;
+    }
+    return 1;
+}
+
+/* Check if memory read or write overlaps with temp's memory location.
+   If this is the case then the temp must be synced to memory. */
+static int tcg_temp_overlap(TCGContext *s, const TCGTemp *tmp,
+                            const TCGAliasInfo *ai)
+{
+    if (!ai->fixed_offset || tmp->fixed_reg) {
+        return 0;
+    }
+    if (tmp->mem_base != &s->temps[GET_TCGV_PTR(s->tcg_env)]) {
+        return 1;
+    }
+    if (ai->offset >= tmp->mem_offset + tcg_temp_size(tmp)
+        || ai->offset + ai->size <= tmp->mem_offset) {
+            return 0;
+    } else {
+        return 1;
+    }
+}
+
 /* Liveness analysis : update the opc_arg_life array to tell if a
    given input arguments is dead. Instructions updating dead
    temporaries are removed. */
@@ -1674,6 +1716,23 @@ static void liveness_pass_1(TCGContext *s, uint8_t *temp_state)
                     temp_state[arg] = TS_DEAD;
                 }
 
+                /* record if the operation uses some globals' memory location */
+                if (s->alias_info[oi].alias_type != TCG_NOT_ALIAS) {
+                    for (i = 0; i < s->nb_globals; i++) {
+                        if (tcg_temp_overwrite(s, &s->temps[i],
+                                               &s->alias_info[oi])) {
+                            temp_state[i] = TS_DEAD;
+                        } else if (tcg_temp_overlap(s, &s->temps[i],
+                                                    &s->alias_info[oi])) {
+                            if (s->alias_info[oi].alias_type & TCG_ALIAS_READ) {
+                                temp_state[i] = TS_MEM | TS_DEAD;
+                            } else if (!(temp_state[i] & TS_DEAD)) {
+                                temp_state[i] |= TS_MEM;
+                            }
+                        }
+                    }
+                }
+
                 /* if end of basic block, update */
                 if (def->flags & TCG_OPF_BB_END) {
                     tcg_la_bb_end(s, temp_state);
@@ -2622,6 +2681,8 @@ int tcg_gen_code(TCGContext *s, TranslationBlock *tb)
     s->la_time -= profile_getclock();
 #endif
 
+    tcg_alias_analysis(s);
+
     {
         uint8_t *temp_state = tcg_malloc(s->nb_temps + s->nb_indirects);
 
-- 
2.1.4

  parent reply	other threads:[~2017-02-02 14:35 UTC|newest]

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-02-02 14:34 [Qemu-devel] [PATCH v2.1 00/20] Emulate guest vector operations with host vector operations Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 01/21] tcg: add support for 128bit vector type Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 02/21] tcg: add support for 64bit " Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 03/21] tcg: support representing vector type with smaller vector or scalar types Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 04/21] tcg: add ld_v128, ld_v64, st_v128 and st_v64 opcodes Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 05/21] tcg: add simple alias analysis Kirill Batuzov
2017-02-02 14:34 ` Kirill Batuzov [this message]
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 07/21] tcg: allow globals to overlap Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 08/21] tcg: add vector addition operations Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 09/21] target/arm: support access to vector guest registers as globals Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 10/21] target/arm: use vector opcode to handle vadd.<size> instruction Kirill Batuzov
2017-02-09 13:19   ` Philippe Mathieu-Daudé
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 11/21] tcg/i386: add support for vector opcodes Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 12/21] tcg/i386: support 64-bit vector operations Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 13/21] tcg/i386: support remaining vector addition operations Kirill Batuzov
     [not found]   ` <2089cbe3-0e9b-fae2-0e35-224f2765dc28@amsat.org>
     [not found]     ` <32a902a1-e8c7-c2f7-ac66-148e02ee0b2d@amsat.org>
2017-02-21 13:29       ` Kirill Batuzov
2017-02-21 16:21         ` Alex Bennée
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 14/21] tcg: do not rely on exact values of MO_BSWAP or MO_SIGN in backend Kirill Batuzov
2017-05-05 13:59   ` Alex Bennée
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 15/21] target/aarch64: do not check for non-existent TCGMemOp Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 16/21] tcg: introduce new TCGMemOp - MO_128 Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 17/21] tcg: introduce qemu_ld_v128 and qemu_st_v128 opcodes Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 18/21] softmmu: create helpers for vector loads Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 19/21] tcg/i386: add support for qemu_ld_v128/qemu_st_v128 ops Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 20/21] target/arm: load two consecutive 64-bits vector regs as a 128-bit vector reg Kirill Batuzov
2017-02-02 14:34 ` [Qemu-devel] [PATCH v2.1 21/21] tcg/README: update README to include information about vector opcodes Kirill Batuzov
2017-02-02 15:25 ` [Qemu-devel] [PATCH v2.1 00/20] Emulate guest vector operations with host vector operations no-reply
2017-02-21 12:19 ` Kirill Batuzov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1486046099-17726-7-git-send-email-batuzovk@ispras.ru \
    --to=batuzovk@ispras.ru \
    --cc=alex.bennee@linaro.org \
    --cc=balrogg@gmail.com \
    --cc=crosthwaite.peter@gmail.com \
    --cc=pbonzini@redhat.com \
    --cc=peter.maydell@linaro.org \
    --cc=qemu-devel@nongnu.org \
    --cc=rth@twiddle.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.