qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Richard Henderson <rth@twiddle.net>
To: qemu-devel@nongnu.org
Cc: aurelien@aurel32.net, Richard Henderson <rth@twiddle.net>
Subject: [Qemu-devel] [PATCH 04/19] tcg-ia64: Simplify brcond
Date: Thu,  5 Sep 2013 23:50:26 -0700	[thread overview]
Message-ID: <1378450242-27080-5-git-send-email-rth@twiddle.net> (raw)
In-Reply-To: <1378450242-27080-1-git-send-email-rth@twiddle.net>

There was a misconception that a stop bit is required between a compare
and the branch that uses the predicate set by the compare.  This lead to
the usage of an extra bundle in which to perform the compare.  The extra
bundle left room for constants to be loaded for use with the compare insn.

If we pack the compare and the branch together in the same bundle, then
there's no longer any room for non-zero constants.  At which point we
can eliminate half the function by not handling them.

Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 tcg/ia64/tcg-target.c | 42 +++++++++---------------------------------
 1 file changed, 9 insertions(+), 33 deletions(-)

diff --git a/tcg/ia64/tcg-target.c b/tcg/ia64/tcg-target.c
index 6708844..413e00f 100644
--- a/tcg/ia64/tcg-target.c
+++ b/tcg/ia64/tcg-target.c
@@ -1442,38 +1442,16 @@ static inline uint64_t tcg_opc_cmp_a(int qp, TCGCond cond, TCGArg arg1,
     }
 }
 
-static inline void tcg_out_brcond(TCGContext *s, TCGCond cond, TCGArg arg1,
-                                  int const_arg1, TCGArg arg2, int const_arg2,
-                                  int label_index, int cmp4)
+static inline void tcg_out_brcond(TCGContext *s, TCGCond cond, TCGReg arg1,
+                                  TCGReg arg2, int label_index, int cmp4)
 {
     TCGLabel *l = &s->labels[label_index];
-    uint64_t opc1, opc2;
 
-    if (const_arg1 && arg1 != 0) {
-        opc1 = tcg_opc_a5(TCG_REG_P0, OPC_ADDL_A5, TCG_REG_R2,
-                          arg1, TCG_REG_R0);
-        arg1 = TCG_REG_R2;
-    } else {
-        opc1 = INSN_NOP_M;
-    }
-
-    if (const_arg2 && arg2 != 0) {
-        opc2 = tcg_opc_a5(TCG_REG_P0, OPC_ADDL_A5, TCG_REG_R3,
-                          arg2, TCG_REG_R0);
-        arg2 = TCG_REG_R3;
-    } else {
-        opc2 = INSN_NOP_I;
-    }
-
-    tcg_out_bundle(s, mII,
-                   opc1,
-                   opc2,
-                   tcg_opc_cmp_a(TCG_REG_P0, cond, arg1, arg2, cmp4));
-    tcg_out_bundle(s, mmB,
-                   INSN_NOP_M,
+    tcg_out_bundle(s, miB,
                    INSN_NOP_M,
-                   tcg_opc_b1 (TCG_REG_P6, OPC_BR_DPTK_FEW_B1,
-                               get_reloc_pcrel21b(s->code_ptr + 2)));
+                   tcg_opc_cmp_a(TCG_REG_P0, cond, arg1, arg2, cmp4),
+                   tcg_opc_b1(TCG_REG_P6, OPC_BR_DPTK_FEW_B1,
+                              get_reloc_pcrel21b(s->code_ptr + 2)));
 
     if (l->has_value) {
         reloc_pcrel21b((s->code_ptr - 16) + 2, l->u.value);
@@ -2222,12 +2200,10 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
         break;
 
     case INDEX_op_brcond_i32:
-        tcg_out_brcond(s, args[2], args[0], const_args[0],
-                       args[1], const_args[1], args[3], 1);
+        tcg_out_brcond(s, args[2], args[0], args[1], args[3], 1);
         break;
     case INDEX_op_brcond_i64:
-        tcg_out_brcond(s, args[2], args[0], const_args[0],
-                       args[1], const_args[1], args[3], 0);
+        tcg_out_brcond(s, args[2], args[0], args[1], args[3], 0);
         break;
     case INDEX_op_setcond_i32:
         tcg_out_setcond(s, args[3], args[0], args[1], args[2], 1);
@@ -2331,7 +2307,7 @@ static const TCGTargetOpDef ia64_op_defs[] = {
     { INDEX_op_bswap16_i32, { "r", "rZ" } },
     { INDEX_op_bswap32_i32, { "r", "rZ" } },
 
-    { INDEX_op_brcond_i32, { "rI", "rI" } },
+    { INDEX_op_brcond_i32, { "rZ", "rZ" } },
     { INDEX_op_setcond_i32, { "r", "rZ", "rZ" } },
     { INDEX_op_movcond_i32, { "r", "rZ", "rZ", "rI", "rI" } },
 
-- 
1.8.3.1

  parent reply	other threads:[~2013-09-06  6:51 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-09-06  6:50 [Qemu-devel] [PATCH 00/19] tcg-ia64 fixes and improvements Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 01/19] tcg-ia64: Use TCGMemOp within qemu_ldst routines Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 02/19] tcg-ia64: Use shortcuts for nop insns Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 03/19] tcg-ia64: Handle constant calls Richard Henderson
2013-09-06  6:50 ` Richard Henderson [this message]
2013-09-06  6:50 ` [Qemu-devel] [PATCH 05/19] tcg-ia64: Move AREG0 to R32 Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 06/19] tcg-ia64: Avoid unnecessary stop bit in tcg_out_alu Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 07/19] tcg-ia64: Use ADDS for small addition Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 08/19] tcg-ia64: Use SUB_A3 and ADDS_A4 for subtraction Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 09/19] tcg-ia64: Use A3 form of logical operations Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 10/19] tcg-ia64 Introduce tcg_opc_mov_a Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 11/19] tcg-ia64 Introduce tcg_opc_movi_a Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 12/19] tcg-ia64 Introduce tcg_opc_ext_i Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 13/19] tcg-ia64 Introduce tcg_opc_bswap64_i Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 14/19] tcg-ia64: Re-bundle the tlb load Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 15/19] tcg-ia64: Move bswap for store into " Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 16/19] tcg-ia64: Move tlb addend load into tlb read Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 17/19] tcg-i64: Reduce code duplication in tcg_out_qemu_ld Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 18/19] tcg-ia64: Convert to new ldst helpers Richard Henderson
2013-09-06  6:50 ` [Qemu-devel] [PATCH 19/19] tcg-ia64: Move part of softmmu slow path out of line Richard Henderson

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=1378450242-27080-5-git-send-email-rth@twiddle.net \
    --to=rth@twiddle.net \
    --cc=aurelien@aurel32.net \
    --cc=qemu-devel@nongnu.org \
    /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).