[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Qemu-devel] [PATCH v7 17/26] tcg: Add generic vector ops for multiplica
From: |
Richard Henderson |
Subject: |
[Qemu-devel] [PATCH v7 17/26] tcg: Add generic vector ops for multiplication |
Date: |
Mon, 18 Dec 2017 09:17:49 -0800 |
Signed-off-by: Richard Henderson <address@hidden>
---
accel/tcg/tcg-runtime.h | 5 +++++
tcg/i386/tcg-target.h | 1 +
tcg/tcg-op-gvec.h | 2 ++
tcg/tcg-op.h | 1 +
tcg/tcg-opc.h | 1 +
tcg/tcg.h | 1 +
accel/tcg/tcg-runtime-gvec.c | 44 ++++++++++++++++++++++++++++++++++++++++++++
tcg/tcg-op-gvec.c | 29 +++++++++++++++++++++++++++++
tcg/tcg-op-vec.c | 22 ++++++++++++++++++++++
tcg/tcg.c | 2 ++
tcg/README | 4 ++++
11 files changed, 112 insertions(+)
diff --git a/accel/tcg/tcg-runtime.h b/accel/tcg/tcg-runtime.h
index 28abf30d76..c4a2e6b215 100644
--- a/accel/tcg/tcg-runtime.h
+++ b/accel/tcg/tcg-runtime.h
@@ -152,6 +152,11 @@ DEF_HELPER_FLAGS_4(gvec_sub16, TCG_CALL_NO_RWG, void, ptr,
ptr, ptr, i32)
DEF_HELPER_FLAGS_4(gvec_sub32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
DEF_HELPER_FLAGS_4(gvec_sub64, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(gvec_mul8, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(gvec_mul16, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(gvec_mul32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(gvec_mul64, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+
DEF_HELPER_FLAGS_3(gvec_neg8, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
DEF_HELPER_FLAGS_3(gvec_neg16, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
DEF_HELPER_FLAGS_3(gvec_neg32, TCG_CALL_NO_RWG, void, ptr, ptr, i32)
diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h
index 60d3684750..949d138c9d 100644
--- a/tcg/i386/tcg-target.h
+++ b/tcg/i386/tcg-target.h
@@ -184,6 +184,7 @@ extern bool have_avx2;
#define TCG_TARGET_HAS_uzp_vec 0
#define TCG_TARGET_HAS_trn_vec 0
#define TCG_TARGET_HAS_cmp_vec 1
+#define TCG_TARGET_HAS_mul_vec 0
#define TCG_TARGET_deposit_i32_valid(ofs, len) \
(((ofs) == 0 && (len) == 8) || ((ofs) == 8 && (len) == 8) || \
diff --git a/tcg/tcg-op-gvec.h b/tcg/tcg-op-gvec.h
index 308bdc13b4..ad5e22e1bf 100644
--- a/tcg/tcg-op-gvec.h
+++ b/tcg/tcg-op-gvec.h
@@ -134,6 +134,8 @@ void tcg_gen_gvec_add(unsigned vece, uint32_t dofs,
uint32_t aofs,
uint32_t bofs, uint32_t opsz, uint32_t clsz);
void tcg_gen_gvec_sub(unsigned vece, uint32_t dofs, uint32_t aofs,
uint32_t bofs, uint32_t opsz, uint32_t clsz);
+void tcg_gen_gvec_mul(unsigned vece, uint32_t dofs, uint32_t aofs,
+ uint32_t bofs, uint32_t opsz, uint32_t clsz);
void tcg_gen_gvec_and(unsigned vece, uint32_t dofs, uint32_t aofs,
uint32_t bofs, uint32_t opsz, uint32_t clsz);
diff --git a/tcg/tcg-op.h b/tcg/tcg-op.h
index b4f73c6048..3296a7baa5 100644
--- a/tcg/tcg-op.h
+++ b/tcg/tcg-op.h
@@ -920,6 +920,7 @@ void tcg_gen_movi_v128(TCGv_vec, uint64_t, uint64_t);
void tcg_gen_movi_v256(TCGv_vec, uint64_t, uint64_t, uint64_t, uint64_t);
void tcg_gen_add_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b);
void tcg_gen_sub_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b);
+void tcg_gen_mul_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b);
void tcg_gen_and_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b);
void tcg_gen_or_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b);
void tcg_gen_xor_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b);
diff --git a/tcg/tcg-opc.h b/tcg/tcg-opc.h
index d3fa014507..b21a30273c 100644
--- a/tcg/tcg-opc.h
+++ b/tcg/tcg-opc.h
@@ -220,6 +220,7 @@ DEF(st_vec, 0, 2, 1, IMPLVEC)
DEF(add_vec, 1, 2, 0, IMPLVEC)
DEF(sub_vec, 1, 2, 0, IMPLVEC)
+DEF(mul_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_mul_vec))
DEF(neg_vec, 1, 1, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_neg_vec))
DEF(and_vec, 1, 2, 0, IMPLVEC)
diff --git a/tcg/tcg.h b/tcg/tcg.h
index ceef8742b5..64078a2e92 100644
--- a/tcg/tcg.h
+++ b/tcg/tcg.h
@@ -185,6 +185,7 @@ typedef uint64_t TCGRegSet;
#define TCG_TARGET_HAS_uzp_vec 0
#define TCG_TARGET_HAS_trn_vec 0
#define TCG_TARGET_HAS_cmp_vec 0
+#define TCG_TARGET_HAS_mul_vec 0
#else
#define TCG_TARGET_MAYBE_vec 1
#endif
diff --git a/accel/tcg/tcg-runtime-gvec.c b/accel/tcg/tcg-runtime-gvec.c
index e0cde3216f..9406ccd769 100644
--- a/accel/tcg/tcg-runtime-gvec.c
+++ b/accel/tcg/tcg-runtime-gvec.c
@@ -141,6 +141,50 @@ void HELPER(gvec_sub64)(void *d, void *a, void *b,
uint32_t desc)
clear_high(d, oprsz, desc);
}
+void HELPER(gvec_mul8)(void *d, void *a, void *b, uint32_t desc)
+{
+ intptr_t oprsz = simd_oprsz(desc);
+ intptr_t i;
+
+ for (i = 0; i < oprsz; i += sizeof(vec8)) {
+ *(vec8 *)(d + i) = *(vec8 *)(a + i) * *(vec8 *)(b + i);
+ }
+ clear_high(d, oprsz, desc);
+}
+
+void HELPER(gvec_mul16)(void *d, void *a, void *b, uint32_t desc)
+{
+ intptr_t oprsz = simd_oprsz(desc);
+ intptr_t i;
+
+ for (i = 0; i < oprsz; i += sizeof(vec16)) {
+ *(vec16 *)(d + i) = *(vec16 *)(a + i) * *(vec16 *)(b + i);
+ }
+ clear_high(d, oprsz, desc);
+}
+
+void HELPER(gvec_mul32)(void *d, void *a, void *b, uint32_t desc)
+{
+ intptr_t oprsz = simd_oprsz(desc);
+ intptr_t i;
+
+ for (i = 0; i < oprsz; i += sizeof(vec32)) {
+ *(vec32 *)(d + i) = *(vec32 *)(a + i) * *(vec32 *)(b + i);
+ }
+ clear_high(d, oprsz, desc);
+}
+
+void HELPER(gvec_mul64)(void *d, void *a, void *b, uint32_t desc)
+{
+ intptr_t oprsz = simd_oprsz(desc);
+ intptr_t i;
+
+ for (i = 0; i < oprsz; i += sizeof(vec64)) {
+ *(vec64 *)(d + i) = *(vec64 *)(a + i) * *(vec64 *)(b + i);
+ }
+ clear_high(d, oprsz, desc);
+}
+
void HELPER(gvec_neg8)(void *d, void *a, uint32_t desc)
{
intptr_t oprsz = simd_oprsz(desc);
diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c
index 0db48bc75a..e6d885da66 100644
--- a/tcg/tcg-op-gvec.c
+++ b/tcg/tcg-op-gvec.c
@@ -1108,6 +1108,35 @@ void tcg_gen_gvec_sub(unsigned vece, uint32_t dofs,
uint32_t aofs,
tcg_gen_gvec_3(dofs, aofs, bofs, opsz, maxsz, &g[vece]);
}
+void tcg_gen_gvec_mul(unsigned vece, uint32_t dofs, uint32_t aofs,
+ uint32_t bofs, uint32_t opsz, uint32_t maxsz)
+{
+ static const GVecGen3 g[4] = {
+ { .fniv = tcg_gen_mul_vec,
+ .fno = gen_helper_gvec_mul8,
+ .opc = INDEX_op_mul_vec,
+ .vece = MO_8 },
+ { .fniv = tcg_gen_mul_vec,
+ .fno = gen_helper_gvec_mul16,
+ .opc = INDEX_op_mul_vec,
+ .vece = MO_16 },
+ { .fni4 = tcg_gen_mul_i32,
+ .fniv = tcg_gen_mul_vec,
+ .fno = gen_helper_gvec_mul32,
+ .opc = INDEX_op_mul_vec,
+ .vece = MO_32 },
+ { .fni8 = tcg_gen_mul_i64,
+ .fniv = tcg_gen_mul_vec,
+ .fno = gen_helper_gvec_mul64,
+ .opc = INDEX_op_mul_vec,
+ .prefer_i64 = TCG_TARGET_REG_BITS == 64,
+ .vece = MO_64 },
+ };
+
+ tcg_debug_assert(vece <= MO_64);
+ tcg_gen_gvec_3(dofs, aofs, bofs, opsz, maxsz, &g[vece]);
+}
+
/* Perform a vector negation using normal negation and a mask.
Compare gen_subv_mask above. */
static void gen_negv_mask(TCGv_i64 d, TCGv_i64 b, TCGv_i64 m)
diff --git a/tcg/tcg-op-vec.c b/tcg/tcg-op-vec.c
index 2c636ebbd6..667e5095cb 100644
--- a/tcg/tcg-op-vec.c
+++ b/tcg/tcg-op-vec.c
@@ -503,3 +503,25 @@ void tcg_gen_cmp_vec(TCGCond cond, unsigned vece,
tcg_expand_vec_op(INDEX_op_cmp_vec, type, vece, ri, ai, bi, cond);
}
}
+
+void tcg_gen_mul_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b)
+{
+ TCGTemp *rt = tcgv_vec_temp(r);
+ TCGTemp *at = tcgv_vec_temp(a);
+ TCGTemp *bt = tcgv_vec_temp(b);
+ TCGArg ri = temp_arg(rt);
+ TCGArg ai = temp_arg(at);
+ TCGArg bi = temp_arg(bt);
+ TCGType type = rt->base_type;
+ int can;
+
+ tcg_debug_assert(at->base_type == type);
+ tcg_debug_assert(bt->base_type == type);
+ can = tcg_can_emit_vec_op(INDEX_op_mul_vec, type, vece);
+ if (can > 0) {
+ vec_gen_3(INDEX_op_mul_vec, type, vece, ri, ai, bi);
+ } else {
+ tcg_debug_assert(can < 0);
+ tcg_expand_vec_op(INDEX_op_mul_vec, type, vece, ri, ai, bi);
+ }
+}
diff --git a/tcg/tcg.c b/tcg/tcg.c
index a85547a6d2..5608391dca 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -1404,6 +1404,8 @@ bool tcg_op_supported(TCGOpcode op)
return have_vec && TCG_TARGET_HAS_andc_vec;
case INDEX_op_orc_vec:
return have_vec && TCG_TARGET_HAS_orc_vec;
+ case INDEX_op_mul_vec:
+ return have_vec && TCG_TARGET_HAS_mul_vec;
case INDEX_op_shli_vec:
case INDEX_op_shri_vec:
case INDEX_op_sari_vec:
diff --git a/tcg/README b/tcg/README
index 18b6bbd8f1..17695ff7f6 100644
--- a/tcg/README
+++ b/tcg/README
@@ -547,6 +547,10 @@ E.g. VECL=1 -> 64 << 1 -> v128, and VECE=2 -> 1 << 2 ->
i32.
Similarly, v0 = v1 - v2.
+* mul_vec v0, v1, v2
+
+ Similarly, v0 = v1 * v2.
+
* neg_vec v0, v1
Similarly, v0 = -v1.
--
2.14.3
- Re: [Qemu-devel] [PATCH v7 04/26] target/arm: Align vector registers, (continued)
- [Qemu-devel] [PATCH v7 03/26] tcg: Allow multiple word entries into the constant pool, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 06/26] target/arm: Use vector infrastructure for aa64 mov/not/neg, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 07/26] target/arm: Use vector infrastructure for aa64 dup/movi, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 09/26] tcg: Add tcg_expand_vec_op and tcg-target.opc.h, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 01/26] tcg: Add types and basic operations for host vectors, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 11/26] target/arm: Use vector infrastructure for aa64 zip/uzp/trn/xtn, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 08/26] tcg/i386: Add vector operations, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 14/26] tcg: Add generic vector ops for comparisons, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 13/26] target/arm: Use vector infrastructure for aa64 constant shifts, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 17/26] tcg: Add generic vector ops for multiplication,
Richard Henderson <=
- [Qemu-devel] [PATCH v7 12/26] tcg: Add generic vector ops for constant shifts, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 20/26] target/arm: Use vector infrastructure for aa64 widening shifts, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 16/26] tcg/i386: Add vector operations/expansions for shift/cmp/interleave, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 02/26] tcg: Add generic vector expanders, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 10/26] tcg: Add generic vector ops for interleave, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 18/26] target/arm: Use vector infrastructure for aa64 multiplies, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 15/26] target/arm: Use vector infrastructure for aa64 compares, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 19/26] tcg: Add generic vector ops for extension, Richard Henderson, 2017/12/18
- [Qemu-devel] [PATCH v7 21/26] tcg/i386: Add vector operations/expansions for mul/extend, Richard Henderson, 2017/12/18