From: Richard Henderson <richard.henderson@linaro.org>
To: Song Gao <gaosong@loongson.cn>, qemu-devel@nongnu.org
Cc: peter.maydell@linaro.org, yangxiaojuan@loongson.cn,
david@redhat.com, bin.meng@windriver.com,
mark.cave-ayland@ilande.co.uk, aleksandar.rikalo@syrmia.com,
jcmvbkbc@gmail.com, tsimpson@quicinc.com,
alistair.francis@wdc.com, edgar.iglesias@gmail.com,
chenhuacai@gmail.com, philmd@redhat.com, atar4qemu@gmail.com,
thuth@redhat.com, ehabkost@redhat.com, groug@kaod.org,
maobibo@loongson.cn, mrolnik@gmail.com, shorne@gmail.com,
alex.bennee@linaro.org, david@gibson.dropbear.id.au,
kbastian@mail.uni-paderborn.de, crwulff@gmail.com,
laurent@vivier.eu, palmer@dabbelt.com, pbonzini@redhat.com,
aurelien@aurel32.net
Subject: Re: [PATCH v4 11/21] target/loongarch: Add floating point comparison instruction translation
Date: Sun, 5 Sep 2021 11:24:25 +0200 [thread overview]
Message-ID: <9481b350-9718-d8fc-11c5-69ff6f409220@linaro.org> (raw)
In-Reply-To: <1630586467-22463-12-git-send-email-gaosong@loongson.cn>
On 9/2/21 2:40 PM, Song Gao wrote:
> +/* fcmp_cXXX_s */
> +uint64_t helper_fcmp_c_s(CPULoongArchState *env, uint64_t fj,
> + uint64_t fk, uint32_t flags)
> +{
> + uint32_t t0, t1;
> + uint64_t cmp = 0;
> + t0 = (uint32_t)fj;
> + t1 = (uint32_t)fk;
> +
> + if (flags) {
> + if (flags & FCMP_LT) {
> + cmp |= float32_lt_quiet(t0, t1, &env->fp_status);
> + }
> + if (flags & FCMP_EQ) {
> + cmp |= float32_eq_quiet(t0, t1, &env->fp_status);
> + }
> + if (flags & FCMP_GT) {
> + cmp |= float32_lt_quiet(t1, t0, &env->fp_status);
> + }
> + if (flags & FCMP_UN) {
> + cmp |= float32_unordered_quiet(t1, t0, &env->fp_status);
> + }
> + } else {
> + cmp = (float32_unordered_quiet(t1, t0, &env->fp_status), 0);
> + }
This is silly, especially the form of the last. You want
FloatRelation cmp = float32_unordered_quiet(t1, t0, &env->fp_status);
bool ret;
switch (cmp) {
case float_relation_less:
ret = (flags & FCMP_LT);
break;
case float_relation_equal:
ret = (flags & FCMP_EQ);
break;
case float_relation_greater:
ret = (flags & FCMP_GT);
break;
case float_relation_unordered:
ret = (flags & FCMP_UN);
break;
default:
g_assert_not_reached();
}
return ret;
And of course the switch can be split out to a common subroutine for use with the other 3
comparison helpers.
> +static bool trans_fcmp_cond_s(DisasContext *ctx, arg_fcmp_cond_s *a)
> +{
> + TCGv var = tcg_temp_new();
> + TCGv_i32 flags = NULL;
> +
> + switch (a->fcond) {
> + /* caf */
> + case 0:
> + flags = tcg_constant_i32(0);
> + gen_helper_fcmp_c_s(var, cpu_env, cpu_fpr[a->fj],
> + cpu_fpr[a->fk], flags);
> + break;
> + /* saf */
> + case 1:
> + flags = tcg_constant_i32(0);
> + gen_helper_fcmp_s_s(var, cpu_env, cpu_fpr[a->fj],
> + cpu_fpr[a->fk], flags);
> + break;
Hmm. Perhaps put the integer flags into a table?
fn = (a->fcond & 1 ? gen_helper_fcmp_s_s : gen_helper_fcmp_c_s);
flags = fcond_table[a->fcond >> 1];
fn(var, cpu_env, cpu_fpr[a->fj], cpu_fpr[a->fk],
tcg_constant_i32(flags));
> +static bool trans_fcmp_cond_d(DisasContext *ctx, arg_fcmp_cond_d *a)
You'd get to share the table with this function.
> --- a/target/loongarch/translate.h
> +++ b/target/loongarch/translate.h
> @@ -11,6 +11,11 @@
>
> #include "exec/translator.h"
>
> +#define FCMP_LT 0x0001 /* fp0 < fp1 */
> +#define FCMP_EQ 0x0010 /* fp0 = fp1 */
> +#define FCMP_GT 0x0100 /* fp1 < fp0 */
> +#define FCMP_UN 0x1000 /* unordered */
Hmm. Better in internals.h? I don't think you want to include translate.h into fpu_helper.c.
r~
next prev parent reply other threads:[~2021-09-05 9:26 UTC|newest]
Thread overview: 46+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-09-02 12:40 [PATCH v4 00/21] Add LoongArch linux-user emulation support Song Gao
2021-09-02 12:40 ` [PATCH v4 01/21] target/loongarch: Add README Song Gao
2021-09-02 12:40 ` [PATCH v4 02/21] target/loongarch: Add core definition Song Gao
2021-09-04 9:44 ` Richard Henderson
2021-09-02 12:40 ` [PATCH v4 03/21] target/loongarch: Add main translation routines Song Gao
2021-09-04 9:45 ` Richard Henderson
2021-09-02 12:40 ` [PATCH v4 04/21] target/loongarch: Add fixed point arithmetic instruction translation Song Gao
2021-09-04 11:04 ` Richard Henderson
2021-09-07 12:36 ` Song Gao
2021-09-02 12:40 ` [PATCH v4 05/21] target/loongarch: Add fixed point shift " Song Gao
2021-09-04 11:17 ` Richard Henderson
2021-09-02 12:40 ` [PATCH v4 06/21] target/loongarch: Add fixed point bit " Song Gao
2021-09-04 12:57 ` Richard Henderson
2021-09-02 12:40 ` [PATCH v4 07/21] target/loongarch: Add fixed point load/store " Song Gao
2021-09-04 13:03 ` Richard Henderson
2021-09-02 12:40 ` [PATCH v4 08/21] target/loongarch: Add fixed point atomic " Song Gao
2021-09-04 13:14 ` Richard Henderson
2021-09-02 12:40 ` [PATCH v4 09/21] target/loongarch: Add fixed point extra " Song Gao
2021-09-05 8:39 ` Richard Henderson
2021-09-02 12:40 ` [PATCH v4 10/21] target/loongarch: Add floating point arithmetic " Song Gao
2021-09-05 9:08 ` Richard Henderson
2021-09-02 12:40 ` [PATCH v4 11/21] target/loongarch: Add floating point comparison " Song Gao
2021-09-05 9:24 ` Richard Henderson [this message]
2021-09-02 12:40 ` [PATCH v4 12/21] target/loongarch: Add floating point conversion " Song Gao
2021-09-05 9:29 ` Richard Henderson
2021-09-02 12:40 ` [PATCH v4 13/21] target/loongarch: Add floating point move " Song Gao
2021-09-05 9:38 ` Richard Henderson
2021-09-05 9:45 ` Richard Henderson
2021-09-02 12:41 ` [PATCH v4 14/21] target/loongarch: Add floating point load/store " Song Gao
2021-09-05 9:46 ` Richard Henderson
2021-09-02 12:41 ` [PATCH v4 15/21] target/loongarch: Add branch " Song Gao
2021-09-05 9:49 ` Richard Henderson
2021-09-02 12:41 ` [PATCH v4 16/21] target/loongarch: Add disassembler Song Gao
2021-09-02 12:41 ` [PATCH v4 17/21] LoongArch Linux User Emulation Song Gao
2021-09-05 10:04 ` Richard Henderson
2021-09-08 9:50 ` Song Gao
2021-09-10 12:52 ` Richard Henderson
2021-09-11 5:58 ` Song Gao
2021-09-12 12:38 ` Richard Henderson
2021-09-02 12:41 ` [PATCH v4 18/21] default-configs: Add loongarch linux-user support Song Gao
2021-09-02 12:41 ` [PATCH v4 19/21] target/loongarch: Add target build suport Song Gao
2021-09-05 10:05 ` Richard Henderson
2021-09-02 12:41 ` [PATCH v4 20/21] target/loongarch: 'make check-tcg' support Song Gao
2021-09-05 10:06 ` Richard Henderson
2021-09-02 12:41 ` [PATCH v4 21/21] scripts: add loongarch64 binfmt config Song Gao
2021-09-05 10:08 ` 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=9481b350-9718-d8fc-11c5-69ff6f409220@linaro.org \
--to=richard.henderson@linaro.org \
--cc=aleksandar.rikalo@syrmia.com \
--cc=alex.bennee@linaro.org \
--cc=alistair.francis@wdc.com \
--cc=atar4qemu@gmail.com \
--cc=aurelien@aurel32.net \
--cc=bin.meng@windriver.com \
--cc=chenhuacai@gmail.com \
--cc=crwulff@gmail.com \
--cc=david@gibson.dropbear.id.au \
--cc=david@redhat.com \
--cc=edgar.iglesias@gmail.com \
--cc=ehabkost@redhat.com \
--cc=gaosong@loongson.cn \
--cc=groug@kaod.org \
--cc=jcmvbkbc@gmail.com \
--cc=kbastian@mail.uni-paderborn.de \
--cc=laurent@vivier.eu \
--cc=maobibo@loongson.cn \
--cc=mark.cave-ayland@ilande.co.uk \
--cc=mrolnik@gmail.com \
--cc=palmer@dabbelt.com \
--cc=pbonzini@redhat.com \
--cc=peter.maydell@linaro.org \
--cc=philmd@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=shorne@gmail.com \
--cc=thuth@redhat.com \
--cc=tsimpson@quicinc.com \
--cc=yangxiaojuan@loongson.cn \
/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).