[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[RFC v3 26/71] target/riscv: rvv-1.0: update vext_max_elems() for load/s
|
From: |
frank . chang |
|
Subject: |
[RFC v3 26/71] target/riscv: rvv-1.0: update vext_max_elems() for load/store insns |
|
Date: |
Thu, 6 Aug 2020 18:46:23 +0800 |
From: Frank Chang <frank.chang@sifive.com>
Unlike other vector instructions, load/store vector instructions return
the maximum vector size calculated with EMUL.
For other vector instructions, return VLMAX as the maximum vector size.
Signed-off-by: Frank Chang <frank.chang@sifive.com>
---
target/riscv/insn_trans/trans_rvv.inc.c | 10 +++-
target/riscv/internals.h | 5 +-
target/riscv/vector_helper.c | 61 ++++++++++++++++++-------
3 files changed, 56 insertions(+), 20 deletions(-)
diff --git a/target/riscv/insn_trans/trans_rvv.inc.c
b/target/riscv/insn_trans/trans_rvv.inc.c
index c7e094b6e5b..725f36fcfcc 100644
--- a/target/riscv/insn_trans/trans_rvv.inc.c
+++ b/target/riscv/insn_trans/trans_rvv.inc.c
@@ -609,7 +609,7 @@ static bool ldst_us_trans(uint32_t vd, uint32_t rs1,
uint32_t data,
/*
* As simd_desc supports at most 256 bytes, and in this implementation,
- * the max vector group length is 2048 bytes. So split it into two parts.
+ * the max vector group length is 1024 bytes. So split it into two parts.
*
* The first part is vlen in bytes, encoded in maxsz of simd_desc.
* The second part is lmul, encoded in data of simd_desc.
@@ -653,6 +653,7 @@ static bool ld_us_op(DisasContext *s, arg_r2nfvm *a,
uint8_t seq)
data = FIELD_DP32(data, VDATA, VM, a->vm);
data = FIELD_DP32(data, VDATA, LMUL, s->lmul);
+ data = FIELD_DP32(data, VDATA, SEW, s->sew);
data = FIELD_DP32(data, VDATA, NF, a->nf);
return ldst_us_trans(a->rd, a->rs1, data, fn, s, false);
}
@@ -689,6 +690,7 @@ static bool st_us_op(DisasContext *s, arg_r2nfvm *a,
uint8_t seq)
data = FIELD_DP32(data, VDATA, VM, a->vm);
data = FIELD_DP32(data, VDATA, LMUL, s->lmul);
+ data = FIELD_DP32(data, VDATA, SEW, s->sew);
data = FIELD_DP32(data, VDATA, NF, a->nf);
return ldst_us_trans(a->rd, a->rs1, data, fn, s, true);
}
@@ -763,6 +765,7 @@ static bool ld_stride_op(DisasContext *s, arg_rnfvm *a,
uint8_t seq)
data = FIELD_DP32(data, VDATA, VM, a->vm);
data = FIELD_DP32(data, VDATA, LMUL, s->lmul);
+ data = FIELD_DP32(data, VDATA, SEW, s->sew);
data = FIELD_DP32(data, VDATA, NF, a->nf);
return ldst_stride_trans(a->rd, a->rs1, a->rs2, data, fn, s, false);
}
@@ -791,6 +794,7 @@ static bool st_stride_op(DisasContext *s, arg_rnfvm *a,
uint8_t seq)
data = FIELD_DP32(data, VDATA, VM, a->vm);
data = FIELD_DP32(data, VDATA, LMUL, s->lmul);
+ data = FIELD_DP32(data, VDATA, SEW, s->sew);
data = FIELD_DP32(data, VDATA, NF, a->nf);
fn = fns[seq];
if (fn == NULL) {
@@ -889,6 +893,7 @@ static bool ld_index_op(DisasContext *s, arg_rnfvm *a,
uint8_t seq)
data = FIELD_DP32(data, VDATA, VM, a->vm);
data = FIELD_DP32(data, VDATA, LMUL, s->lmul);
+ data = FIELD_DP32(data, VDATA, SEW, s->sew);
data = FIELD_DP32(data, VDATA, NF, a->nf);
return ldst_index_trans(a->rd, a->rs1, a->rs2, data, fn, s, false);
}
@@ -940,6 +945,7 @@ static bool st_index_op(DisasContext *s, arg_rnfvm *a,
uint8_t seq)
data = FIELD_DP32(data, VDATA, VM, a->vm);
data = FIELD_DP32(data, VDATA, LMUL, s->lmul);
+ data = FIELD_DP32(data, VDATA, SEW, s->sew);
data = FIELD_DP32(data, VDATA, NF, a->nf);
return ldst_index_trans(a->rd, a->rs1, a->rs2, data, fn, s, true);
}
@@ -1005,6 +1011,7 @@ static bool ldff_op(DisasContext *s, arg_r2nfvm *a,
uint8_t seq)
data = FIELD_DP32(data, VDATA, VM, a->vm);
data = FIELD_DP32(data, VDATA, LMUL, s->lmul);
+ data = FIELD_DP32(data, VDATA, SEW, s->sew);
data = FIELD_DP32(data, VDATA, NF, a->nf);
return ldff_trans(a->rd, a->rs1, data, fn, s);
}
@@ -1189,6 +1196,7 @@ static bool amo_op(DisasContext *s, arg_rwdvm *a, uint8_t
seq)
data = FIELD_DP32(data, VDATA, VM, a->vm);
data = FIELD_DP32(data, VDATA, LMUL, s->lmul);
+ data = FIELD_DP32(data, VDATA, SEW, s->sew);
data = FIELD_DP32(data, VDATA, WD, a->wd);
return amo_trans(a->rd, a->rs1, a->rs2, data, fn, s);
}
diff --git a/target/riscv/internals.h b/target/riscv/internals.h
index bca48297dab..4fb683a7399 100644
--- a/target/riscv/internals.h
+++ b/target/riscv/internals.h
@@ -24,8 +24,9 @@
/* share data between vector helpers and decode code */
FIELD(VDATA, VM, 0, 1)
FIELD(VDATA, LMUL, 1, 3)
-FIELD(VDATA, NF, 4, 4)
-FIELD(VDATA, WD, 4, 1)
+FIELD(VDATA, SEW, 4, 3)
+FIELD(VDATA, NF, 7, 4)
+FIELD(VDATA, WD, 7, 1)
/* float point classify helpers */
target_ulong fclass_h(uint64_t frs1);
diff --git a/target/riscv/vector_helper.c b/target/riscv/vector_helper.c
index a58051ded93..77f62c86e02 100644
--- a/target/riscv/vector_helper.c
+++ b/target/riscv/vector_helper.c
@@ -17,6 +17,7 @@
*/
#include "qemu/osdep.h"
+#include "qemu/host-utils.h"
#include "cpu.h"
#include "exec/memop.h"
#include "exec/exec-all.h"
@@ -98,6 +99,11 @@ static inline uint32_t vext_vm(uint32_t desc)
return FIELD_EX32(simd_data(desc), VDATA, VM);
}
+static inline uint32_t vext_sew(uint32_t desc)
+{
+ return FIELD_EX32(simd_data(desc), VDATA, SEW);
+}
+
/*
* Encode LMUL to lmul as following:
* LMUL vlmul lmul
@@ -122,14 +128,35 @@ static uint32_t vext_wd(uint32_t desc)
}
/*
- * Get vector group length in bytes. Its range is [64, 2048].
+ * Get the maximum number of elements can be operated.
*
- * As simd_desc support at most 256, the max vlen is 512 bits.
- * So vlen in bytes is encoded as maxsz.
+ * Use ctzl() to get log2(esz) and log2(vlenb)
+ * so that we can use shifts for all arithmetics.
*/
-static inline uint32_t vext_maxsz(uint32_t desc)
+static inline uint32_t vext_max_elems(uint32_t desc, uint32_t esz, bool
is_ldst)
{
- return simd_maxsz(desc) << vext_lmul(desc);
+ /*
+ * As simd_desc support at most 256 bytes, the max vlen is 256 bits.
+ * so vlen in bytes (vlenb) is encoded as maxsz.
+ */
+ uint32_t vlenb = simd_maxsz(desc);
+
+ if (is_ldst) {
+ /*
+ * Vector load/store instructions have the EEW encoded
+ * directly in the instructions. The maximum vector size is
+ * calculated with EMUL rather than LMUL.
+ */
+ uint32_t eew = ctzl(esz);
+ uint32_t sew = vext_sew(desc);
+ uint32_t lmul = vext_lmul(desc);
+ int32_t emul = eew - sew + lmul;
+ uint32_t emul_r = emul < 0 ? 0 : emul;
+ return 1 << (ctzl(vlenb) + emul_r - ctzl(esz));
+ } else {
+ /* Return VLMAX */
+ return 1 << (ctzl(vlenb) + vext_lmul(desc) - ctzl(esz));
+ }
}
/*
@@ -224,7 +251,7 @@ vext_ldst_stride(void *vd, void *v0, target_ulong base,
{
uint32_t i, k;
uint32_t nf = vext_nf(desc);
- uint32_t vlmax = vext_maxsz(desc) / esz;
+ uint32_t max_elems = vext_max_elems(desc, esz, true);
/* probe every access*/
for (i = 0; i < env->vl; i++) {
@@ -241,7 +268,7 @@ vext_ldst_stride(void *vd, void *v0, target_ulong base,
}
while (k < nf) {
target_ulong addr = base + stride * i + k * esz;
- ldst_elem(env, addr, i + k * vlmax, vd, ra);
+ ldst_elem(env, addr, i + k * max_elems, vd, ra);
k++;
}
}
@@ -289,7 +316,7 @@ vext_ldst_us(void *vd, target_ulong base, CPURISCVState
*env, uint32_t desc,
{
uint32_t i, k;
uint32_t nf = vext_nf(desc);
- uint32_t vlmax = vext_maxsz(desc) / esz;
+ uint32_t max_elems = vext_max_elems(desc, esz, true);
/* probe every access */
probe_pages(env, base, env->vl * nf * esz, ra, access_type);
@@ -298,7 +325,7 @@ vext_ldst_us(void *vd, target_ulong base, CPURISCVState
*env, uint32_t desc,
k = 0;
while (k < nf) {
target_ulong addr = base + (i * nf + k) * esz;
- ldst_elem(env, addr, i + k * vlmax, vd, ra);
+ ldst_elem(env, addr, i + k * max_elems, vd, ra);
k++;
}
}
@@ -379,7 +406,7 @@ vext_ldst_index(void *vd, void *v0, target_ulong base,
uint32_t i, k;
uint32_t nf = vext_nf(desc);
uint32_t vm = vext_vm(desc);
- uint32_t vlmax = vext_maxsz(desc) / esz;
+ uint32_t max_elems = vext_max_elems(desc, esz, true);
/* probe every access*/
for (i = 0; i < env->vl; i++) {
@@ -397,7 +424,7 @@ vext_ldst_index(void *vd, void *v0, target_ulong base,
}
while (k < nf) {
abi_ptr addr = get_index_addr(base, i, vs2) + k * esz;
- ldst_elem(env, addr, i + k * vlmax, vd, ra);
+ ldst_elem(env, addr, i + k * max_elems, vd, ra);
k++;
}
}
@@ -467,7 +494,7 @@ vext_ldff(void *vd, void *v0, target_ulong base,
uint32_t i, k, vl = 0;
uint32_t nf = vext_nf(desc);
uint32_t vm = vext_vm(desc);
- uint32_t vlmax = vext_maxsz(desc) / esz;
+ uint32_t max_elems = vext_max_elems(desc, esz, true);
target_ulong addr, offset, remain;
/* probe every access*/
@@ -518,7 +545,7 @@ ProbeSuccess:
}
while (k < nf) {
target_ulong addr = base + (i * nf + k) * esz;
- ldst_elem(env, addr, i + k * vlmax, vd, ra);
+ ldst_elem(env, addr, i + k * max_elems, vd, ra);
k++;
}
}
@@ -1226,7 +1253,7 @@ void HELPER(NAME)(void *vd, void *v0, target_ulong s1,
\
void *vs2, CPURISCVState *env, uint32_t desc) \
{ \
uint32_t vl = env->vl; \
- uint32_t vlmax = vext_maxsz(desc) / sizeof(ETYPE); \
+ uint32_t vlmax = vext_max_elems(desc, sizeof(ETYPE), false);\
uint32_t i; \
\
for (i = 0; i < vl; i++) { \
@@ -3887,7 +3914,7 @@ void HELPER(NAME)(void *vd, void *v0, uint64_t s1, void
*vs2, \
{ \
uint32_t vm = vext_vm(desc); \
uint32_t vl = env->vl; \
- uint32_t vlmax = vext_maxsz(desc) / sizeof(ETYPE); \
+ uint32_t vlmax = vext_max_elems(desc, sizeof(ETYPE), false); \
uint32_t i; \
\
for (i = 0; i < vl; i++) { \
@@ -4692,7 +4719,7 @@ GEN_VEXT_VSLIDE1DOWN_VX(vslide1down_vx_d, uint64_t, H8)
void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2, \
CPURISCVState *env, uint32_t desc) \
{ \
- uint32_t vlmax = env_archcpu(env)->cfg.vlen; \
+ uint32_t vlmax = vext_max_elems(desc, sizeof(ETYPE), false); \
uint32_t vm = vext_vm(desc); \
uint32_t vl = env->vl; \
uint32_t index, i; \
@@ -4720,7 +4747,7 @@ GEN_VEXT_VRGATHER_VV(vrgather_vv_d, uint64_t, H8)
void HELPER(NAME)(void *vd, void *v0, target_ulong s1, void *vs2, \
CPURISCVState *env, uint32_t desc) \
{ \
- uint32_t vlmax = env_archcpu(env)->cfg.vlen; \
+ uint32_t vlmax = vext_max_elems(desc, sizeof(ETYPE), false); \
uint32_t vm = vext_vm(desc); \
uint32_t vl = env->vl; \
uint32_t index = s1, i; \
--
2.17.1
- [RFC v3 19/71] target/riscv: rvv-1.0: configure instructions, (continued)
- [RFC v3 19/71] target/riscv: rvv-1.0: configure instructions, frank . chang, 2020/08/06
- [RFC v3 18/71] target/riscv: rvv-1.0: apply nanbox helper in opfvf_trans, frank . chang, 2020/08/06
- [RFC v3 20/71] target/riscv: rvv-1.0: stride load and store instructions, frank . chang, 2020/08/06
- [RFC v3 22/71] target/riscv: rvv-1.0: fix address index overflow bug of indexed load/store insns, frank . chang, 2020/08/06
- [RFC v3 21/71] target/riscv: rvv-1.0: index load and store instructions, frank . chang, 2020/08/06
- [RFC v3 23/71] target/riscv: rvv-1.0: fault-only-first unit stride load, frank . chang, 2020/08/06
- [RFC v3 24/71] target/riscv: rvv-1.0: amo operations, frank . chang, 2020/08/06
- [RFC v3 25/71] target/riscv: rvv-1.0: load/store whole register instructions, frank . chang, 2020/08/06
- [RFC v3 26/71] target/riscv: rvv-1.0: update vext_max_elems() for load/store insns,
frank . chang <=
- Re: [RFC v3 26/71] target/riscv: rvv-1.0: update vext_max_elems() for load/store insns, Richard Henderson, 2020/08/06
- Re: [RFC v3 26/71] target/riscv: rvv-1.0: update vext_max_elems() for load/store insns, Frank Chang, 2020/08/13
- Re: [RFC v3 26/71] target/riscv: rvv-1.0: update vext_max_elems() for load/store insns, Richard Henderson, 2020/08/14
- Re: [RFC v3 26/71] target/riscv: rvv-1.0: update vext_max_elems() for load/store insns, Frank Chang, 2020/08/15
- Re: [RFC v3 26/71] target/riscv: rvv-1.0: update vext_max_elems() for load/store insns, Richard Henderson, 2020/08/15
- Re: [RFC v3 26/71] target/riscv: rvv-1.0: update vext_max_elems() for load/store insns, Frank Chang, 2020/08/15
- Re: [RFC v3 26/71] target/riscv: rvv-1.0: update vext_max_elems() for load/store insns, Frank Chang, 2020/08/15
[RFC v3 28/71] target/riscv: rvv-1.0: floating-point square-root instruction, frank . chang, 2020/08/06
[RFC v3 27/71] target/riscv: rvv-1.0: take fractional LMUL into vector max elements calculation, frank . chang, 2020/08/06
[RFC v3 29/71] target/riscv: rvv-1.0: floating-point classify instructions, frank . chang, 2020/08/06