From: Aurelien Jarno <aurelien@aurel32.net>
To: Richard Henderson <rth@twiddle.net>
Cc: qemu-devel@nongnu.org, aliguori@amazon.com
Subject: Re: [Qemu-devel] [PATCH 03/20] tcg-ia64: Handle constant calls
Date: Sun, 3 Nov 2013 23:56:03 +0100 [thread overview]
Message-ID: <20131103225603.GA21176@ohm.rr44.fr> (raw)
In-Reply-To: <1383250929-10288-4-git-send-email-rth@twiddle.net>
On Thu, Oct 31, 2013 at 01:21:52PM -0700, Richard Henderson wrote:
> Using only indirect calls results in 3 bundles (one to load the
> descriptor address), and 4 stop bits. By looking through the
> descriptor to the constants, we can perform the call with 2
> bundles and only 1 stop bit.
>
> Signed-off-by: Richard Henderson <rth@twiddle.net>
> ---
> tcg/ia64/tcg-target.c | 38 +++++++++++++++++++++++++++++++++++---
> 1 file changed, 35 insertions(+), 3 deletions(-)
>
> diff --git a/tcg/ia64/tcg-target.c b/tcg/ia64/tcg-target.c
> index e1b8b7d..b19e298 100644
> --- a/tcg/ia64/tcg-target.c
> +++ b/tcg/ia64/tcg-target.c
> @@ -226,6 +226,7 @@ enum {
> OPC_BR_CALL_SPTK_MANY_B5 = 0x02100001000ull,
> OPC_BR_RET_SPTK_MANY_B4 = 0x00108001100ull,
> OPC_BRL_SPTK_MANY_X3 = 0x18000001000ull,
> + OPC_BRL_CALL_SPTK_MANY_X4 = 0x1a000001000ull,
> OPC_CMP_LT_A6 = 0x18000000000ull,
> OPC_CMP_LTU_A6 = 0x1a000000000ull,
> OPC_CMP_EQ_A6 = 0x1c000000000ull,
> @@ -584,6 +585,8 @@ static inline uint64_t tcg_opc_l3(uint64_t imm)
> return (imm & 0x07fffffffff00000ull) >> 18;
> }
>
> +#define tcg_opc_l4 tcg_opc_l3
> +
> static inline uint64_t tcg_opc_m1(int qp, uint64_t opc, int r1, int r3)
> {
> return opc
> @@ -668,6 +671,15 @@ static inline uint64_t tcg_opc_x3(int qp, uint64_t opc, uint64_t imm)
> | (qp & 0x3f);
> }
>
> +static inline uint64_t tcg_opc_x4(int qp, uint64_t opc, int b1, uint64_t imm)
> +{
> + return opc
> + | ((imm & 0x0800000000000000ull) >> 23) /* i */
> + | ((imm & 0x00000000000fffffull) << 13) /* imm20b */
> + | ((b1 & 0x7) << 6)
> + | (qp & 0x3f);
> +}
> +
>
> /*
> * Relocations
> @@ -893,7 +905,23 @@ static void tcg_out_br(TCGContext *s, int label_index)
> }
> }
>
> -static inline void tcg_out_call(TCGContext *s, TCGArg addr)
> +static inline void tcg_out_calli(TCGContext *s, uintptr_t addr)
> +{
> + /* Look through the function descriptor. */
> + uintptr_t disp, *desc = (uintptr_t *)addr;
> + tcg_out_bundle(s, mlx,
> + INSN_NOP_M,
> + tcg_opc_l2 (desc[1]),
> + tcg_opc_x2 (TCG_REG_P0, OPC_MOVL_X2, TCG_REG_R1, desc[1]));
> + disp = (desc[0] - (uintptr_t)s->code_ptr) >> 4;
> + tcg_out_bundle(s, mLX,
> + INSN_NOP_M,
> + tcg_opc_l4 (disp),
> + tcg_opc_x4 (TCG_REG_P0, OPC_BRL_CALL_SPTK_MANY_X4,
> + TCG_REG_B0, disp));
> +}
> +
> +static inline void tcg_out_callr(TCGContext *s, TCGReg addr)
> {
> tcg_out_bundle(s, MmI,
> tcg_opc_m1 (TCG_REG_P0, OPC_LD8_M1, TCG_REG_R2, addr),
> @@ -2004,7 +2032,11 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
> tcg_out_br(s, args[0]);
> break;
> case INDEX_op_call:
> - tcg_out_call(s, args[0]);
> + if (likely(const_args[0])) {
> + tcg_out_calli(s, args[0]);
> + } else {
> + tcg_out_callr(s, args[0]);
> + }
> break;
> case INDEX_op_goto_tb:
> tcg_out_goto_tb(s, args[0]);
> @@ -2257,7 +2289,7 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
>
> static const TCGTargetOpDef ia64_op_defs[] = {
> { INDEX_op_br, { } },
> - { INDEX_op_call, { "r" } },
> + { INDEX_op_call, { "ri" } },
> { INDEX_op_exit_tb, { } },
> { INDEX_op_goto_tb, { } },
>
Acked-by: Aurelien Jarno <aurelien@aurel32.net>
--
Aurelien Jarno GPG: 1024D/F1BCDB73
aurelien@aurel32.net http://www.aurel32.net
next prev parent reply other threads:[~2013-11-03 23:28 UTC|newest]
Thread overview: 39+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-10-31 20:21 [Qemu-devel] [PATCH 00/20] tcg improvements for ia64 Richard Henderson
2013-10-31 20:21 ` [Qemu-devel] [PATCH 01/20] tcg-ia64: Use TCGMemOp within qemu_ldst routines Richard Henderson
2013-11-03 22:55 ` Aurelien Jarno
2013-11-06 22:05 ` Aurelien Jarno
2013-10-31 20:21 ` [Qemu-devel] [PATCH 02/20] tcg-ia64: Use shortcuts for nop insns Richard Henderson
2013-11-03 22:55 ` Aurelien Jarno
2013-10-31 20:21 ` [Qemu-devel] [PATCH 03/20] tcg-ia64: Handle constant calls Richard Henderson
2013-11-03 22:56 ` Aurelien Jarno [this message]
2013-10-31 20:21 ` [Qemu-devel] [PATCH 04/20] tcg-ia64: Simplify brcond Richard Henderson
2013-11-03 22:56 ` Aurelien Jarno
2013-11-03 23:34 ` Richard Henderson
2013-11-04 4:24 ` Richard Henderson
2013-11-06 22:05 ` Aurelien Jarno
2013-11-06 22:04 ` Aurelien Jarno
2013-10-31 20:21 ` [Qemu-devel] [PATCH 05/20] tcg-ia64: Move AREG0 to R32 Richard Henderson
2013-11-06 22:33 ` Aurelien Jarno
2013-10-31 20:21 ` [Qemu-devel] [PATCH 06/20] tcg-ia64: Avoid unnecessary stop bit in tcg_out_alu Richard Henderson
2013-11-06 22:33 ` Aurelien Jarno
2013-10-31 20:21 ` [Qemu-devel] [PATCH 07/20] tcg-ia64: Use ADDS for small addition Richard Henderson
2013-11-06 22:34 ` Aurelien Jarno
2013-10-31 20:21 ` [Qemu-devel] [PATCH 08/20] tcg-ia64: Use SUB_A3 and ADDS_A4 for subtraction Richard Henderson
2013-11-06 22:34 ` Aurelien Jarno
2013-10-31 20:21 ` [Qemu-devel] [PATCH 09/20] tcg-ia64: Use A3 form of logical operations Richard Henderson
2013-11-06 22:34 ` Aurelien Jarno
2013-10-31 20:21 ` [Qemu-devel] [PATCH 10/20] tcg-ia64: Introduce tcg_opc_mov_a Richard Henderson
2013-11-06 22:34 ` Aurelien Jarno
2013-10-31 20:22 ` [Qemu-devel] [PATCH 11/20] tcg-ia64: Introduce tcg_opc_movi_a Richard Henderson
2013-11-06 22:34 ` Aurelien Jarno
2013-10-31 20:22 ` [Qemu-devel] [PATCH 12/20] tcg-ia64: Introduce tcg_opc_ext_i Richard Henderson
2013-11-06 22:34 ` Aurelien Jarno
2013-10-31 20:22 ` [Qemu-devel] [PATCH 13/20] tcg-ia64: Introduce tcg_opc_bswap64_i Richard Henderson
2013-11-06 22:34 ` Aurelien Jarno
2013-10-31 20:22 ` [Qemu-devel] [PATCH 14/20] tcg-ia64: Re-bundle the tlb load Richard Henderson
2013-10-31 20:22 ` [Qemu-devel] [PATCH 15/20] tcg-ia64: Move bswap for store into " Richard Henderson
2013-10-31 20:22 ` [Qemu-devel] [PATCH 16/20] tcg-ia64: Move tlb addend load into tlb read Richard Henderson
2013-10-31 20:22 ` [Qemu-devel] [PATCH 17/20] tcg-ia64: Reduce code duplication in tcg_out_qemu_ld Richard Henderson
2013-10-31 20:22 ` [Qemu-devel] [PATCH 18/20] tcg-ia64: Convert to new ldst helpers Richard Henderson
2013-10-31 20:22 ` [Qemu-devel] [PATCH 19/20] tcg-ia64: Move part of softmmu slow path out of line Richard Henderson
2013-10-31 20:22 ` [Qemu-devel] [PATCH 20/20] tcg-ia64: Optimize small arguments to exit_tb 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=20131103225603.GA21176@ohm.rr44.fr \
--to=aurelien@aurel32.net \
--cc=aliguori@amazon.com \
--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 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).