target/mips: msa: Split helpers for DOTP_S.<H|W|D>

Achieves clearer code and slightly better performance.

Reviewed-by: Aleksandar Rikalo <aleksandar.rikalo@syrmia.com>
Signed-off-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Message-Id: <20200613152133.8964-8-aleksandar.qemu.devel@gmail.com>
This commit is contained in:
Aleksandar Markovic 2020-06-13 17:21:26 +02:00
parent 0c8c76ac85
commit 165cacb65c
3 changed files with 69 additions and 14 deletions

View File

@ -1079,7 +1079,10 @@ DEF_HELPER_5(msa_subs_u_df, void, env, i32, i32, i32, i32)
DEF_HELPER_5(msa_subsus_u_df, void, env, i32, i32, i32, i32)
DEF_HELPER_5(msa_subsuu_s_df, void, env, i32, i32, i32, i32)
DEF_HELPER_5(msa_mulv_df, void, env, i32, i32, i32, i32)
DEF_HELPER_5(msa_dotp_s_df, void, env, i32, i32, i32, i32)
DEF_HELPER_4(msa_dotp_s_h, void, env, i32, i32, i32)
DEF_HELPER_4(msa_dotp_s_w, void, env, i32, i32, i32)
DEF_HELPER_4(msa_dotp_s_d, void, env, i32, i32, i32)
DEF_HELPER_5(msa_dotp_u_df, void, env, i32, i32, i32, i32)
DEF_HELPER_4(msa_dpadd_s_h, void, env, i32, i32, i32)
DEF_HELPER_4(msa_dpadd_s_w, void, env, i32, i32, i32)

View File

@ -2236,6 +2236,60 @@ void helper_msa_div_u_d(CPUMIPSState *env,
o = UNSIGNED_ODD(a, df); \
} while (0)
static inline int64_t msa_dotp_s_df(uint32_t df, int64_t arg1, int64_t arg2)
{
int64_t even_arg1;
int64_t even_arg2;
int64_t odd_arg1;
int64_t odd_arg2;
SIGNED_EXTRACT(even_arg1, odd_arg1, arg1, df);
SIGNED_EXTRACT(even_arg2, odd_arg2, arg2, df);
return (even_arg1 * even_arg2) + (odd_arg1 * odd_arg2);
}
void helper_msa_dotp_s_h(CPUMIPSState *env,
uint32_t wd, uint32_t ws, uint32_t wt)
{
wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
wr_t *pws = &(env->active_fpu.fpr[ws].wr);
wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
pwd->h[0] = msa_dotp_s_df(DF_HALF, pws->h[0], pwt->h[0]);
pwd->h[1] = msa_dotp_s_df(DF_HALF, pws->h[1], pwt->h[1]);
pwd->h[2] = msa_dotp_s_df(DF_HALF, pws->h[2], pwt->h[2]);
pwd->h[3] = msa_dotp_s_df(DF_HALF, pws->h[3], pwt->h[3]);
pwd->h[4] = msa_dotp_s_df(DF_HALF, pws->h[4], pwt->h[4]);
pwd->h[5] = msa_dotp_s_df(DF_HALF, pws->h[5], pwt->h[5]);
pwd->h[6] = msa_dotp_s_df(DF_HALF, pws->h[6], pwt->h[6]);
pwd->h[7] = msa_dotp_s_df(DF_HALF, pws->h[7], pwt->h[7]);
}
void helper_msa_dotp_s_w(CPUMIPSState *env,
uint32_t wd, uint32_t ws, uint32_t wt)
{
wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
wr_t *pws = &(env->active_fpu.fpr[ws].wr);
wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
pwd->w[0] = msa_dotp_s_df(DF_WORD, pws->w[0], pwt->w[0]);
pwd->w[1] = msa_dotp_s_df(DF_WORD, pws->w[1], pwt->w[1]);
pwd->w[2] = msa_dotp_s_df(DF_WORD, pws->w[2], pwt->w[2]);
pwd->w[3] = msa_dotp_s_df(DF_WORD, pws->w[3], pwt->w[3]);
}
void helper_msa_dotp_s_d(CPUMIPSState *env,
uint32_t wd, uint32_t ws, uint32_t wt)
{
wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
wr_t *pws = &(env->active_fpu.fpr[ws].wr);
wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
pwd->d[0] = msa_dotp_s_df(DF_DOUBLE, pws->d[0], pwt->d[0]);
pwd->d[1] = msa_dotp_s_df(DF_DOUBLE, pws->d[1], pwt->d[1]);
}
static inline int64_t msa_dpadd_s_df(uint32_t df, int64_t dest, int64_t arg1,
int64_t arg2)
{
@ -5010,17 +5064,6 @@ static inline int64_t msa_mulv_df(uint32_t df, int64_t arg1, int64_t arg2)
return arg1 * arg2;
}
static inline int64_t msa_dotp_s_df(uint32_t df, int64_t arg1, int64_t arg2)
{
int64_t even_arg1;
int64_t even_arg2;
int64_t odd_arg1;
int64_t odd_arg2;
SIGNED_EXTRACT(even_arg1, odd_arg1, arg1, df);
SIGNED_EXTRACT(even_arg2, odd_arg2, arg2, df);
return (even_arg1 * even_arg2) + (odd_arg1 * odd_arg2);
}
static inline int64_t msa_dotp_u_df(uint32_t df, int64_t arg1, int64_t arg2)
{
int64_t even_arg1;
@ -5155,7 +5198,6 @@ MSA_BINOP_DF(subs_u)
MSA_BINOP_DF(subsus_u)
MSA_BINOP_DF(subsuu_s)
MSA_BINOP_DF(mulv)
MSA_BINOP_DF(dotp_s)
MSA_BINOP_DF(dotp_u)
MSA_BINOP_DF(mul_q)

View File

@ -29394,7 +29394,17 @@ static void gen_msa_3r(CPUMIPSState *env, DisasContext *ctx)
}
break;
case OPC_DOTP_S_df:
gen_helper_msa_dotp_s_df(cpu_env, tdf, twd, tws, twt);
switch (df) {
case DF_HALF:
gen_helper_msa_dotp_s_h(cpu_env, twd, tws, twt);
break;
case DF_WORD:
gen_helper_msa_dotp_s_w(cpu_env, twd, tws, twt);
break;
case DF_DOUBLE:
gen_helper_msa_dotp_s_d(cpu_env, twd, tws, twt);
break;
}
break;
case OPC_DOTP_U_df:
gen_helper_msa_dotp_u_df(cpu_env, tdf, twd, tws, twt);