From: Taylor Simpson <tsimpson@quicinc.com>
To: qemu-devel@nongnu.org
Cc: tsimpson@quicinc.com, richard.henderson@linaro.org,
philmd@linaro.org, peter.maydell@linaro.org, bcain@quicinc.com,
quic_mathbern@quicinc.com, stefanha@redhat.com, ale@rev.ng,
anjo@rev.ng, quic_mliebel@quicinc.com
Subject: [PULL 01/11] Hexagon (translate.c): avoid redundant PC updates on COF
Date: Wed, 19 Apr 2023 20:26:24 -0700 [thread overview]
Message-ID: <20230420032634.105311-2-tsimpson@quicinc.com> (raw)
In-Reply-To: <20230420032634.105311-1-tsimpson@quicinc.com>
From: Matheus Tavares Bernardino <quic_mathbern@quicinc.com>
When there is a conditional change of flow or an endloop instruction, we
preload HEX_REG_PC with ctx->next_PC at gen_start_packet(). Nonetheless,
we still generate TCG code to do this update again at gen_goto_tb() when
the condition for the COF is not met, thus producing redundant
instructions. This can be seen with the following packet:
0x004002e4: 0x5c20d000 { if (!P0) jump:t PC+0 }
Which generates this TCG code:
---- 004002e4
-> mov_i32 pc,$0x4002e8
and_i32 loc9,p0,$0x1
mov_i32 branch_taken,loc9
add_i32 pkt_cnt,pkt_cnt,$0x2
add_i32 insn_cnt,insn_cnt,$0x2
brcond_i32 branch_taken,$0x0,ne,$L1
goto_tb $0x0
mov_i32 pc,$0x4002e4
exit_tb $0x7fb0c36e5200
set_label $L1
goto_tb $0x1
-> mov_i32 pc,$0x4002e8
exit_tb $0x7fb0c36e5201
set_label $L0
exit_tb $0x7fb0c36e5203
Note that even after optimizations, the redundant PC update is still
present:
---- 004002e4
-> mov_i32 pc,$0x4002e8 sync: 0 dead: 0 1 pref=0xffff
mov_i32 branch_taken,$0x1 sync: 0 dead: 0 1 pref=0xffff
add_i32 pkt_cnt,pkt_cnt,$0x2 sync: 0 dead: 0 1 pref=0xffff
add_i32 insn_cnt,insn_cnt,$0x2 sync: 0 dead: 0 1 2 pref=0xffff
goto_tb $0x1
-> mov_i32 pc,$0x4002e8 sync: 0 dead: 0 1 pref=0xffff
exit_tb $0x7fb0c36e5201
set_label $L0
exit_tb $0x7fb0c36e5203
With this patch, the second redundant update is properly discarded.
Note that we need the additional "move_to_pc" flag instead of just
avoiding the update whenever `dest == ctx->next_PC`, as that could
potentially skip updates from a COF with met condition, whose
ctx->branch_dest just happens to be equal to ctx->next_PC.
Signed-off-by: Matheus Tavares Bernardino <quic_mathbern@quicinc.com>
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
Reviewed-by: Taylor Simpson <tsimpson@quicinc.com>
Message-Id: <fc059153c3f0526d97b7f13450c02b276b0908e1.1679519341.git.quic_mathbern@quicinc.com>
---
target/hexagon/translate.c | 21 +++++++++++++--------
1 file changed, 13 insertions(+), 8 deletions(-)
diff --git a/target/hexagon/translate.c b/target/hexagon/translate.c
index 665476ab48..58d638f734 100644
--- a/target/hexagon/translate.c
+++ b/target/hexagon/translate.c
@@ -128,14 +128,19 @@ static bool use_goto_tb(DisasContext *ctx, target_ulong dest)
return translator_use_goto_tb(&ctx->base, dest);
}
-static void gen_goto_tb(DisasContext *ctx, int idx, target_ulong dest)
+static void gen_goto_tb(DisasContext *ctx, int idx, target_ulong dest, bool
+ move_to_pc)
{
if (use_goto_tb(ctx, dest)) {
tcg_gen_goto_tb(idx);
- tcg_gen_movi_tl(hex_gpr[HEX_REG_PC], dest);
+ if (move_to_pc) {
+ tcg_gen_movi_tl(hex_gpr[HEX_REG_PC], dest);
+ }
tcg_gen_exit_tb(ctx->base.tb, idx);
} else {
- tcg_gen_movi_tl(hex_gpr[HEX_REG_PC], dest);
+ if (move_to_pc) {
+ tcg_gen_movi_tl(hex_gpr[HEX_REG_PC], dest);
+ }
tcg_gen_lookup_and_goto_ptr();
}
}
@@ -150,11 +155,11 @@ static void gen_end_tb(DisasContext *ctx)
if (ctx->branch_cond != TCG_COND_ALWAYS) {
TCGLabel *skip = gen_new_label();
tcg_gen_brcondi_tl(ctx->branch_cond, hex_branch_taken, 0, skip);
- gen_goto_tb(ctx, 0, ctx->branch_dest);
+ gen_goto_tb(ctx, 0, ctx->branch_dest, true);
gen_set_label(skip);
- gen_goto_tb(ctx, 1, ctx->next_PC);
+ gen_goto_tb(ctx, 1, ctx->next_PC, false);
} else {
- gen_goto_tb(ctx, 0, ctx->branch_dest);
+ gen_goto_tb(ctx, 0, ctx->branch_dest, true);
}
} else if (ctx->is_tight_loop &&
pkt->insn[pkt->num_insns - 1].opcode == J2_endloop0) {
@@ -165,9 +170,9 @@ static void gen_end_tb(DisasContext *ctx)
TCGLabel *skip = gen_new_label();
tcg_gen_brcondi_tl(TCG_COND_LEU, hex_gpr[HEX_REG_LC0], 1, skip);
tcg_gen_subi_tl(hex_gpr[HEX_REG_LC0], hex_gpr[HEX_REG_LC0], 1);
- gen_goto_tb(ctx, 0, ctx->base.tb->pc);
+ gen_goto_tb(ctx, 0, ctx->base.tb->pc, true);
gen_set_label(skip);
- gen_goto_tb(ctx, 1, ctx->next_PC);
+ gen_goto_tb(ctx, 1, ctx->next_PC, false);
} else {
tcg_gen_lookup_and_goto_ptr();
}
--
2.25.1
next prev parent reply other threads:[~2023-04-20 3:29 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-04-20 3:26 [PULL 00/11] Hexagon queue Taylor Simpson
2023-04-20 3:26 ` Taylor Simpson [this message]
2023-04-20 3:26 ` [PULL 02/11] Use f-strings in python scripts Taylor Simpson
2023-04-20 3:26 ` [PULL 03/11] Use black code style for " Taylor Simpson
2023-04-20 3:26 ` [PULL 04/11] Hexagon (target/hexagon) Remove redundant/unused macros Taylor Simpson
2023-04-20 3:26 ` [PULL 05/11] Hexagon (target/hexagon) Merge arguments to probe_pkt_scalar_hvx_stores Taylor Simpson
2023-04-20 3:26 ` [PULL 06/11] Hexagon (target/hexagon) Add overrides for count trailing zeros/ones Taylor Simpson
2023-04-20 3:26 ` [PULL 07/11] Hexagon (target/hexagon) Updates to USR should use get_result_gpr Taylor Simpson
2023-04-20 3:26 ` [PULL 08/11] Hexagon (tests/tcg/hexagon) Move HVX test infra to header file Taylor Simpson
2023-04-20 3:26 ` [PULL 09/11] Hexagon (target/hexagon) Remove unused slot variable in helpers Taylor Simpson
2023-04-20 3:26 ` [PULL 10/11] Hexagon (target/hexagon) Add overrides for cache/sync/barrier instructions Taylor Simpson
2023-04-20 3:26 ` [PULL 11/11] Hexagon (target/hexagon) Additional instructions handled by idef-parser Taylor Simpson
2023-04-21 8:20 ` [PULL 00/11] Hexagon queue Richard Henderson
2023-04-21 15:06 ` Taylor Simpson
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=20230420032634.105311-2-tsimpson@quicinc.com \
--to=tsimpson@quicinc.com \
--cc=ale@rev.ng \
--cc=anjo@rev.ng \
--cc=bcain@quicinc.com \
--cc=peter.maydell@linaro.org \
--cc=philmd@linaro.org \
--cc=qemu-devel@nongnu.org \
--cc=quic_mathbern@quicinc.com \
--cc=quic_mliebel@quicinc.com \
--cc=richard.henderson@linaro.org \
--cc=stefanha@redhat.com \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).