From 1165669982b1ca3ca62d7091f17669d83e46e8b9 Mon Sep 17 00:00:00 2001 From: Aleksandar Markovic Date: Wed, 25 Sep 2019 14:46:08 +0200 Subject: [PATCH] target/mips: msa: Split helpers for CLT_. Achieves clearer code and slightly better performance. Signed-off-by: Aleksandar Markovic Reviewed-by: Aleksandar Rikalo Message-Id: <1569415572-19635-17-git-send-email-aleksandar.markovic@rt-rk.com> --- target/mips/helper.h | 12 ++- target/mips/msa_helper.c | 158 +++++++++++++++++++++++++++++++++++---- target/mips/translate.c | 38 ++++++++-- 3 files changed, 186 insertions(+), 22 deletions(-) diff --git a/target/mips/helper.h b/target/mips/helper.h index 32ff24b0c1..29dfcf070b 100644 --- a/target/mips/helper.h +++ b/target/mips/helper.h @@ -857,6 +857,16 @@ DEF_HELPER_4(msa_cle_u_h, void, env, i32, i32, i32) DEF_HELPER_4(msa_cle_u_w, void, env, i32, i32, i32) DEF_HELPER_4(msa_cle_u_d, void, env, i32, i32, i32) +DEF_HELPER_4(msa_clt_s_b, void, env, i32, i32, i32) +DEF_HELPER_4(msa_clt_s_h, void, env, i32, i32, i32) +DEF_HELPER_4(msa_clt_s_w, void, env, i32, i32, i32) +DEF_HELPER_4(msa_clt_s_d, void, env, i32, i32, i32) + +DEF_HELPER_4(msa_clt_u_b, void, env, i32, i32, i32) +DEF_HELPER_4(msa_clt_u_h, void, env, i32, i32, i32) +DEF_HELPER_4(msa_clt_u_w, void, env, i32, i32, i32) +DEF_HELPER_4(msa_clt_u_d, void, env, i32, i32, i32) + DEF_HELPER_4(msa_andi_b, void, env, i32, i32, i32) DEF_HELPER_4(msa_ori_b, void, env, i32, i32, i32) @@ -906,8 +916,6 @@ DEF_HELPER_5(msa_min_s_df, void, env, i32, i32, i32, i32) DEF_HELPER_5(msa_min_u_df, void, env, i32, i32, i32, i32) DEF_HELPER_5(msa_max_a_df, void, env, i32, i32, i32, i32) DEF_HELPER_5(msa_min_a_df, void, env, i32, i32, i32, i32) -DEF_HELPER_5(msa_clt_s_df, void, env, i32, i32, i32, i32) -DEF_HELPER_5(msa_clt_u_df, void, env, i32, i32, i32, i32) DEF_HELPER_5(msa_add_a_df, void, env, i32, i32, i32, i32) DEF_HELPER_5(msa_adds_a_df, void, env, i32, i32, i32, i32) DEF_HELPER_5(msa_adds_s_df, void, env, i32, i32, i32, i32) diff --git a/target/mips/msa_helper.c b/target/mips/msa_helper.c index d696791452..b658f48e04 100644 --- a/target/mips/msa_helper.c +++ b/target/mips/msa_helper.c @@ -1364,6 +1364,150 @@ void helper_msa_cle_u_d(CPUMIPSState *env, pwd->d[1] = msa_cle_u_df(DF_DOUBLE, pws->d[1], pwt->d[1]); } +static inline int64_t msa_clt_s_df(uint32_t df, int64_t arg1, int64_t arg2) +{ + return arg1 < arg2 ? -1 : 0; +} + +void helper_msa_clt_s_b(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->b[0] = msa_clt_s_df(DF_BYTE, pws->b[0], pwt->b[0]); + pwd->b[1] = msa_clt_s_df(DF_BYTE, pws->b[1], pwt->b[1]); + pwd->b[2] = msa_clt_s_df(DF_BYTE, pws->b[2], pwt->b[2]); + pwd->b[3] = msa_clt_s_df(DF_BYTE, pws->b[3], pwt->b[3]); + pwd->b[4] = msa_clt_s_df(DF_BYTE, pws->b[4], pwt->b[4]); + pwd->b[5] = msa_clt_s_df(DF_BYTE, pws->b[5], pwt->b[5]); + pwd->b[6] = msa_clt_s_df(DF_BYTE, pws->b[6], pwt->b[6]); + pwd->b[7] = msa_clt_s_df(DF_BYTE, pws->b[7], pwt->b[7]); + pwd->b[8] = msa_clt_s_df(DF_BYTE, pws->b[8], pwt->b[8]); + pwd->b[9] = msa_clt_s_df(DF_BYTE, pws->b[9], pwt->b[9]); + pwd->b[10] = msa_clt_s_df(DF_BYTE, pws->b[10], pwt->b[10]); + pwd->b[11] = msa_clt_s_df(DF_BYTE, pws->b[11], pwt->b[11]); + pwd->b[12] = msa_clt_s_df(DF_BYTE, pws->b[12], pwt->b[12]); + pwd->b[13] = msa_clt_s_df(DF_BYTE, pws->b[13], pwt->b[13]); + pwd->b[14] = msa_clt_s_df(DF_BYTE, pws->b[14], pwt->b[14]); + pwd->b[15] = msa_clt_s_df(DF_BYTE, pws->b[15], pwt->b[15]); +} + +void helper_msa_clt_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_clt_s_df(DF_HALF, pws->h[0], pwt->h[0]); + pwd->h[1] = msa_clt_s_df(DF_HALF, pws->h[1], pwt->h[1]); + pwd->h[2] = msa_clt_s_df(DF_HALF, pws->h[2], pwt->h[2]); + pwd->h[3] = msa_clt_s_df(DF_HALF, pws->h[3], pwt->h[3]); + pwd->h[4] = msa_clt_s_df(DF_HALF, pws->h[4], pwt->h[4]); + pwd->h[5] = msa_clt_s_df(DF_HALF, pws->h[5], pwt->h[5]); + pwd->h[6] = msa_clt_s_df(DF_HALF, pws->h[6], pwt->h[6]); + pwd->h[7] = msa_clt_s_df(DF_HALF, pws->h[7], pwt->h[7]); +} + +void helper_msa_clt_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_clt_s_df(DF_WORD, pws->w[0], pwt->w[0]); + pwd->w[1] = msa_clt_s_df(DF_WORD, pws->w[1], pwt->w[1]); + pwd->w[2] = msa_clt_s_df(DF_WORD, pws->w[2], pwt->w[2]); + pwd->w[3] = msa_clt_s_df(DF_WORD, pws->w[3], pwt->w[3]); +} + +void helper_msa_clt_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_clt_s_df(DF_DOUBLE, pws->d[0], pwt->d[0]); + pwd->d[1] = msa_clt_s_df(DF_DOUBLE, pws->d[1], pwt->d[1]); +} + +static inline int64_t msa_clt_u_df(uint32_t df, int64_t arg1, int64_t arg2) +{ + uint64_t u_arg1 = UNSIGNED(arg1, df); + uint64_t u_arg2 = UNSIGNED(arg2, df); + return u_arg1 < u_arg2 ? -1 : 0; +} + +void helper_msa_clt_u_b(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->b[0] = msa_clt_u_df(DF_BYTE, pws->b[0], pwt->b[0]); + pwd->b[1] = msa_clt_u_df(DF_BYTE, pws->b[1], pwt->b[1]); + pwd->b[2] = msa_clt_u_df(DF_BYTE, pws->b[2], pwt->b[2]); + pwd->b[3] = msa_clt_u_df(DF_BYTE, pws->b[3], pwt->b[3]); + pwd->b[4] = msa_clt_u_df(DF_BYTE, pws->b[4], pwt->b[4]); + pwd->b[5] = msa_clt_u_df(DF_BYTE, pws->b[5], pwt->b[5]); + pwd->b[6] = msa_clt_u_df(DF_BYTE, pws->b[6], pwt->b[6]); + pwd->b[7] = msa_clt_u_df(DF_BYTE, pws->b[7], pwt->b[7]); + pwd->b[8] = msa_clt_u_df(DF_BYTE, pws->b[8], pwt->b[8]); + pwd->b[9] = msa_clt_u_df(DF_BYTE, pws->b[9], pwt->b[9]); + pwd->b[10] = msa_clt_u_df(DF_BYTE, pws->b[10], pwt->b[10]); + pwd->b[11] = msa_clt_u_df(DF_BYTE, pws->b[11], pwt->b[11]); + pwd->b[12] = msa_clt_u_df(DF_BYTE, pws->b[12], pwt->b[12]); + pwd->b[13] = msa_clt_u_df(DF_BYTE, pws->b[13], pwt->b[13]); + pwd->b[14] = msa_clt_u_df(DF_BYTE, pws->b[14], pwt->b[14]); + pwd->b[15] = msa_clt_u_df(DF_BYTE, pws->b[15], pwt->b[15]); +} + +void helper_msa_clt_u_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_clt_u_df(DF_HALF, pws->h[0], pwt->h[0]); + pwd->h[1] = msa_clt_u_df(DF_HALF, pws->h[1], pwt->h[1]); + pwd->h[2] = msa_clt_u_df(DF_HALF, pws->h[2], pwt->h[2]); + pwd->h[3] = msa_clt_u_df(DF_HALF, pws->h[3], pwt->h[3]); + pwd->h[4] = msa_clt_u_df(DF_HALF, pws->h[4], pwt->h[4]); + pwd->h[5] = msa_clt_u_df(DF_HALF, pws->h[5], pwt->h[5]); + pwd->h[6] = msa_clt_u_df(DF_HALF, pws->h[6], pwt->h[6]); + pwd->h[7] = msa_clt_u_df(DF_HALF, pws->h[7], pwt->h[7]); +} + +void helper_msa_clt_u_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_clt_u_df(DF_WORD, pws->w[0], pwt->w[0]); + pwd->w[1] = msa_clt_u_df(DF_WORD, pws->w[1], pwt->w[1]); + pwd->w[2] = msa_clt_u_df(DF_WORD, pws->w[2], pwt->w[2]); + pwd->w[3] = msa_clt_u_df(DF_WORD, pws->w[3], pwt->w[3]); +} + +void helper_msa_clt_u_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_clt_u_df(DF_DOUBLE, pws->d[0], pwt->d[0]); + pwd->d[1] = msa_clt_u_df(DF_DOUBLE, pws->d[1], pwt->d[1]); +} + /* * Int Divide @@ -1771,18 +1915,6 @@ static inline int64_t msa_subv_df(uint32_t df, int64_t arg1, int64_t arg2) return arg1 - arg2; } -static inline int64_t msa_clt_s_df(uint32_t df, int64_t arg1, int64_t arg2) -{ - return arg1 < arg2 ? -1 : 0; -} - -static inline int64_t msa_clt_u_df(uint32_t df, int64_t arg1, int64_t arg2) -{ - uint64_t u_arg1 = UNSIGNED(arg1, df); - uint64_t u_arg2 = UNSIGNED(arg2, df); - return u_arg1 < u_arg2 ? -1 : 0; -} - static inline int64_t msa_max_s_df(uint32_t df, int64_t arg1, int64_t arg2) { return arg1 > arg2 ? arg1 : arg2; @@ -2380,8 +2512,6 @@ MSA_BINOP_DF(min_s) MSA_BINOP_DF(min_u) MSA_BINOP_DF(max_a) MSA_BINOP_DF(min_a) -MSA_BINOP_DF(clt_s) -MSA_BINOP_DF(clt_u) MSA_BINOP_DF(add_a) MSA_BINOP_DF(adds_a) MSA_BINOP_DF(adds_s) diff --git a/target/mips/translate.c b/target/mips/translate.c index 614b9e7cb3..4db87d637e 100644 --- a/target/mips/translate.c +++ b/target/mips/translate.c @@ -28578,6 +28578,38 @@ static void gen_msa_3r(CPUMIPSState *env, DisasContext *ctx) break; } break; + case OPC_CLT_S_df: + switch (df) { + case DF_BYTE: + gen_helper_msa_clt_s_b(cpu_env, twd, tws, twt); + break; + case DF_HALF: + gen_helper_msa_clt_s_h(cpu_env, twd, tws, twt); + break; + case DF_WORD: + gen_helper_msa_clt_s_w(cpu_env, twd, tws, twt); + break; + case DF_DOUBLE: + gen_helper_msa_clt_s_d(cpu_env, twd, tws, twt); + break; + } + break; + case OPC_CLT_U_df: + switch (df) { + case DF_BYTE: + gen_helper_msa_clt_u_b(cpu_env, twd, tws, twt); + break; + case DF_HALF: + gen_helper_msa_clt_u_h(cpu_env, twd, tws, twt); + break; + case DF_WORD: + gen_helper_msa_clt_u_w(cpu_env, twd, tws, twt); + break; + case DF_DOUBLE: + gen_helper_msa_clt_u_d(cpu_env, twd, tws, twt); + break; + } + break; case OPC_SLL_df: gen_helper_msa_sll_df(cpu_env, tdf, twd, tws, twt); break; @@ -28626,9 +28658,6 @@ static void gen_msa_3r(CPUMIPSState *env, DisasContext *ctx) case OPC_MAX_S_df: gen_helper_msa_max_s_df(cpu_env, tdf, twd, tws, twt); break; - case OPC_CLT_S_df: - gen_helper_msa_clt_s_df(cpu_env, tdf, twd, tws, twt); - break; case OPC_ADDS_S_df: gen_helper_msa_adds_s_df(cpu_env, tdf, twd, tws, twt); break; @@ -28647,9 +28676,6 @@ static void gen_msa_3r(CPUMIPSState *env, DisasContext *ctx) case OPC_MAX_U_df: gen_helper_msa_max_u_df(cpu_env, tdf, twd, tws, twt); break; - case OPC_CLT_U_df: - gen_helper_msa_clt_u_df(cpu_env, tdf, twd, tws, twt); - break; case OPC_ADDS_U_df: gen_helper_msa_adds_u_df(cpu_env, tdf, twd, tws, twt); break; -- GitLab