| Message ID | 20240112213812.173521-11-dbarboza@ventanamicro.com |
|---|---|
| State | New |
| Headers | show |
| Series | target/riscv: add 'cpu->cfg.vlenb', remove 'cpu->cfg.vlen' | expand |
On 1/13/24 08:38, Daniel Henrique Barboza wrote: > MAXSZ() returns the value in bytes. Now that we have access to vlenb > we don't need to compensate with a '-3' in the scale to use 'vlen'. > > MAXSZ() now works as follows: > > LMUL lmul=scale MAXSZ (vlenb >> -scale) > 1 0 vlenb > 2 1 vlenb << 1 > 4 2 vlenb << 2 > 8 3 vlenb << 3 Um, what? > static inline uint32_t MAXSZ(DisasContext *s) > { > - int scale = s->lmul - 3; > - return s->cfg_ptr->vlen >> -scale; > + return s->cfg_ptr->vlenb >> -s->lmul; There's no left-shift here, either before or after. Bug? r~
On 1/12/24 19:45, Richard Henderson wrote: > On 1/13/24 08:38, Daniel Henrique Barboza wrote: >> MAXSZ() returns the value in bytes. Now that we have access to vlenb >> we don't need to compensate with a '-3' in the scale to use 'vlen'. >> >> MAXSZ() now works as follows: >> >> LMUL lmul=scale MAXSZ (vlenb >> -scale) >> 1 0 vlenb >> 2 1 vlenb << 1 >> 4 2 vlenb << 2 >> 8 3 vlenb << 3 > > Um, what? > >> static inline uint32_t MAXSZ(DisasContext *s) >> { >> - int scale = s->lmul - 3; >> - return s->cfg_ptr->vlen >> -scale; >> + return s->cfg_ptr->vlenb >> -s->lmul; > > There's no left-shift here, either before or after. > Bug? Yeah, I got lost in that explanation. I'll simplify it to be just a simple case of adjusting the right shift to use vlenb like in patches 8 and 9. Thanks, Daniel > > > r~
diff --git a/target/riscv/insn_trans/trans_rvv.c.inc b/target/riscv/insn_trans/trans_rvv.c.inc index d743675262..804cfd6c7f 100644 --- a/target/riscv/insn_trans/trans_rvv.c.inc +++ b/target/riscv/insn_trans/trans_rvv.c.inc @@ -1160,12 +1160,11 @@ GEN_LDST_WHOLE_TRANS(vs8r_v, 8, 1, true) /* * MAXSZ returns the maximum vector size can be operated in bytes, * which is used in GVEC IR when vl_eq_vlmax flag is set to true - * to accerlate vector operation. + * to accelerate vector operation. */ static inline uint32_t MAXSZ(DisasContext *s) { - int scale = s->lmul - 3; - return s->cfg_ptr->vlen >> -scale; + return s->cfg_ptr->vlenb >> -s->lmul; } static bool opivv_check(DisasContext *s, arg_rmrr *a)
MAXSZ() returns the value in bytes. Now that we have access to vlenb we don't need to compensate with a '-3' in the scale to use 'vlen'. MAXSZ() now works as follows: LMUL lmul=scale MAXSZ (vlenb >> -scale) 1 0 vlenb 2 1 vlenb << 1 4 2 vlenb << 2 8 3 vlenb << 3 1/8 -3 vlenb >> 3 1/4 -2 vlenb >> 2 1/2 -1 vlenb >> 1 Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> --- target/riscv/insn_trans/trans_rvv.c.inc | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)