From 9620ae01b888e7d7ffd084782ac4edaa857413b2 Mon Sep 17 00:00:00 2001 From: Matheus Ferst Date: Wed, 2 Mar 2022 06:51:38 +0100 Subject: [PATCH] tcg/tcg-op-gvec.c: Introduce tcg_gen_gvec_4i MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Following the implementation of tcg_gen_gvec_3i, add a four-vector and immediate operand expansion method. Reviewed-by: Richard Henderson Signed-off-by: Matheus Ferst Message-Id: <20220225210936.1749575-34-matheus.ferst@eldorado.org.br> Signed-off-by: Cédric Le Goater --- include/tcg/tcg-op-gvec.h | 22 ++++++ tcg/tcg-op-gvec.c | 146 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 168 insertions(+) diff --git a/include/tcg/tcg-op-gvec.h b/include/tcg/tcg-op-gvec.h index da55fed870..28cafbcc5c 100644 --- a/include/tcg/tcg-op-gvec.h +++ b/include/tcg/tcg-op-gvec.h @@ -218,6 +218,25 @@ typedef struct { bool write_aofs; } GVecGen4; +typedef struct { + /* + * Expand inline as a 64-bit or 32-bit integer. Only one of these will be + * non-NULL. + */ + void (*fni8)(TCGv_i64, TCGv_i64, TCGv_i64, TCGv_i64, int64_t); + void (*fni4)(TCGv_i32, TCGv_i32, TCGv_i32, TCGv_i32, int32_t); + /* Expand inline with a host vector type. */ + void (*fniv)(unsigned, TCGv_vec, TCGv_vec, TCGv_vec, TCGv_vec, int64_t); + /* Expand out-of-line helper w/descriptor, data in descriptor. */ + gen_helper_gvec_4 *fno; + /* The optional opcodes, if any, utilized by .fniv. */ + const TCGOpcode *opt_opc; + /* The vector element size, if applicable. */ + uint8_t vece; + /* Prefer i64 to v64. */ + bool prefer_i64; +} GVecGen4i; + void tcg_gen_gvec_2(uint32_t dofs, uint32_t aofs, uint32_t oprsz, uint32_t maxsz, const GVecGen2 *); void tcg_gen_gvec_2i(uint32_t dofs, uint32_t aofs, uint32_t oprsz, @@ -231,6 +250,9 @@ void tcg_gen_gvec_3i(uint32_t dofs, uint32_t aofs, uint32_t bofs, const GVecGen3i *); void tcg_gen_gvec_4(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t cofs, uint32_t oprsz, uint32_t maxsz, const GVecGen4 *); +void tcg_gen_gvec_4i(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t cofs, + uint32_t oprsz, uint32_t maxsz, int64_t c, + const GVecGen4i *); /* Expand a specific vector operation. */ diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c index ffe55e908f..079a761b04 100644 --- a/tcg/tcg-op-gvec.c +++ b/tcg/tcg-op-gvec.c @@ -836,6 +836,30 @@ static void expand_4_i32(uint32_t dofs, uint32_t aofs, uint32_t bofs, tcg_temp_free_i32(t0); } +static void expand_4i_i32(uint32_t dofs, uint32_t aofs, uint32_t bofs, + uint32_t cofs, uint32_t oprsz, int32_t c, + void (*fni)(TCGv_i32, TCGv_i32, TCGv_i32, TCGv_i32, + int32_t)) +{ + TCGv_i32 t0 = tcg_temp_new_i32(); + TCGv_i32 t1 = tcg_temp_new_i32(); + TCGv_i32 t2 = tcg_temp_new_i32(); + TCGv_i32 t3 = tcg_temp_new_i32(); + uint32_t i; + + for (i = 0; i < oprsz; i += 4) { + tcg_gen_ld_i32(t1, cpu_env, aofs + i); + tcg_gen_ld_i32(t2, cpu_env, bofs + i); + tcg_gen_ld_i32(t3, cpu_env, cofs + i); + fni(t0, t1, t2, t3, c); + tcg_gen_st_i32(t0, cpu_env, dofs + i); + } + tcg_temp_free_i32(t3); + tcg_temp_free_i32(t2); + tcg_temp_free_i32(t1); + tcg_temp_free_i32(t0); +} + /* Expand OPSZ bytes worth of two-operand operations using i64 elements. */ static void expand_2_i64(uint32_t dofs, uint32_t aofs, uint32_t oprsz, bool load_dest, void (*fni)(TCGv_i64, TCGv_i64)) @@ -971,6 +995,30 @@ static void expand_4_i64(uint32_t dofs, uint32_t aofs, uint32_t bofs, tcg_temp_free_i64(t0); } +static void expand_4i_i64(uint32_t dofs, uint32_t aofs, uint32_t bofs, + uint32_t cofs, uint32_t oprsz, int64_t c, + void (*fni)(TCGv_i64, TCGv_i64, TCGv_i64, TCGv_i64, + int64_t)) +{ + TCGv_i64 t0 = tcg_temp_new_i64(); + TCGv_i64 t1 = tcg_temp_new_i64(); + TCGv_i64 t2 = tcg_temp_new_i64(); + TCGv_i64 t3 = tcg_temp_new_i64(); + uint32_t i; + + for (i = 0; i < oprsz; i += 8) { + tcg_gen_ld_i64(t1, cpu_env, aofs + i); + tcg_gen_ld_i64(t2, cpu_env, bofs + i); + tcg_gen_ld_i64(t3, cpu_env, cofs + i); + fni(t0, t1, t2, t3, c); + tcg_gen_st_i64(t0, cpu_env, dofs + i); + } + tcg_temp_free_i64(t3); + tcg_temp_free_i64(t2); + tcg_temp_free_i64(t1); + tcg_temp_free_i64(t0); +} + /* Expand OPSZ bytes worth of two-operand operations using host vectors. */ static void expand_2_vec(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t oprsz, uint32_t tysz, TCGType type, @@ -1121,6 +1169,35 @@ static void expand_4_vec(unsigned vece, uint32_t dofs, uint32_t aofs, tcg_temp_free_vec(t0); } +/* + * Expand OPSZ bytes worth of four-vector operands and an immediate operand + * using host vectors. + */ +static void expand_4i_vec(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t cofs, uint32_t oprsz, + uint32_t tysz, TCGType type, int64_t c, + void (*fni)(unsigned, TCGv_vec, TCGv_vec, + TCGv_vec, TCGv_vec, int64_t)) +{ + TCGv_vec t0 = tcg_temp_new_vec(type); + TCGv_vec t1 = tcg_temp_new_vec(type); + TCGv_vec t2 = tcg_temp_new_vec(type); + TCGv_vec t3 = tcg_temp_new_vec(type); + uint32_t i; + + for (i = 0; i < oprsz; i += tysz) { + tcg_gen_ld_vec(t1, cpu_env, aofs + i); + tcg_gen_ld_vec(t2, cpu_env, bofs + i); + tcg_gen_ld_vec(t3, cpu_env, cofs + i); + fni(vece, t0, t1, t2, t3, c); + tcg_gen_st_vec(t0, cpu_env, dofs + i); + } + tcg_temp_free_vec(t3); + tcg_temp_free_vec(t2); + tcg_temp_free_vec(t1); + tcg_temp_free_vec(t0); +} + /* Expand a vector two-operand operation. */ void tcg_gen_gvec_2(uint32_t dofs, uint32_t aofs, uint32_t oprsz, uint32_t maxsz, const GVecGen2 *g) @@ -1533,6 +1610,75 @@ void tcg_gen_gvec_4(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t cofs, } } +/* Expand a vector four-operand operation. */ +void tcg_gen_gvec_4i(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t cofs, + uint32_t oprsz, uint32_t maxsz, int64_t c, + const GVecGen4i *g) +{ + const TCGOpcode *this_list = g->opt_opc ? : vecop_list_empty; + const TCGOpcode *hold_list = tcg_swap_vecop_list(this_list); + TCGType type; + uint32_t some; + + check_size_align(oprsz, maxsz, dofs | aofs | bofs | cofs); + check_overlap_4(dofs, aofs, bofs, cofs, maxsz); + + type = 0; + if (g->fniv) { + type = choose_vector_type(g->opt_opc, g->vece, oprsz, g->prefer_i64); + } + switch (type) { + case TCG_TYPE_V256: + /* + * Recall that ARM SVE allows vector sizes that are not a + * power of 2, but always a multiple of 16. The intent is + * that e.g. size == 80 would be expanded with 2x32 + 1x16. + */ + some = QEMU_ALIGN_DOWN(oprsz, 32); + expand_4i_vec(g->vece, dofs, aofs, bofs, cofs, some, + 32, TCG_TYPE_V256, c, g->fniv); + if (some == oprsz) { + break; + } + dofs += some; + aofs += some; + bofs += some; + cofs += some; + oprsz -= some; + maxsz -= some; + /* fallthru */ + case TCG_TYPE_V128: + expand_4i_vec(g->vece, dofs, aofs, bofs, cofs, oprsz, + 16, TCG_TYPE_V128, c, g->fniv); + break; + case TCG_TYPE_V64: + expand_4i_vec(g->vece, dofs, aofs, bofs, cofs, oprsz, + 8, TCG_TYPE_V64, c, g->fniv); + break; + + case 0: + if (g->fni8 && check_size_impl(oprsz, 8)) { + expand_4i_i64(dofs, aofs, bofs, cofs, oprsz, c, g->fni8); + } else if (g->fni4 && check_size_impl(oprsz, 4)) { + expand_4i_i32(dofs, aofs, bofs, cofs, oprsz, c, g->fni4); + } else { + assert(g->fno != NULL); + tcg_gen_gvec_4_ool(dofs, aofs, bofs, cofs, + oprsz, maxsz, c, g->fno); + oprsz = maxsz; + } + break; + + default: + g_assert_not_reached(); + } + tcg_swap_vecop_list(hold_list); + + if (oprsz < maxsz) { + expand_clr(dofs + oprsz, maxsz - oprsz); + } +} + /* * Expand specific vector operations. */