Re: [PATCH v3 04/14] tcg/riscv: Add riscv vset{i}vli support

[Richard Henderson <richard.henderson@linaro.org>]

On 9/9/24 19:46, LIU Zhiwei wrote:
>>     lmul = type - riscv_lg2_vlenb;
>>     if (lmul < -3) {
>>         /* Host VLEN >= 1024 bits. */
>>         vlmul = VLMUL_M1;
> I am not sure if we should use VLMUL_MF8,

Perhaps.  See below.

>>     } else if (lmul < 3) {
>>         /* 1/8 ... 1 ... 8 */
>>         vlmul = lmul & 7;
>>         lmul_eq_avl = true;
>>     } else {
>>         /* Guaranteed by Zve64x. */
>>         g_assert_not_reached();
>>     }
>>
>>     avl = tcg_type_size(type) >> vsew;
>>     vtype = encode_vtype(true, true, vsew, vlmul);
>>
>>     if (avl < 32) {
>>         insn = encode_i(OPC_VSETIVLI, TCG_REG_ZERO, avl, vtype);
> Which may benifit here? we usually use  lmul as smallest as we can for macro ops split.

lmul is unchanged, just explicitly setting AVL as well.
The "benefit" is that AVL is visible in the disassembly,
and that we are able to discard the result.

There doesn't appear to be a down side.  Is there one?

>>     } else if (lmul_eq_avl) {
>>         /* rd != 0 and rs1 == 0 uses vlmax */
>>         insn = encode_i(OPC_VSETVLI, TCG_REG_TMP0, TCG_REG_ZERO, vtype);

As opposed to here, where we must clobber a register.
It is a scratch reg, sure, and probably affects nothing
in any microarch which does register renaming.

>>     } else {
>>         tcg_out_opc_imm(s, OPC_ADDI, TCG_REG_TMP0, TCG_REG_ZERO, avl);
>>         insn = encode_i(OPC_VSETVLI, TCG_REG_ZERO, TCG_REG_TMP0, vtype);
> And perhaps here.

Here, lmul does *not* equal avl, and so we must set it, and because of non-use of VSETIVLI 
we also know that it does not fit in uimm5.

But here's a follow-up question regarding current micro-architectures:

   How much benefit is there from adjusting LMUL < 1, or AVL < VLMAX?

For instance, on other hosts with 128-bit vectors, we also promise support for 64-bit 
registers, just so we can support guests which have 64-bit vector operations.  In existing 
hosts (x86, ppc, s390x, loongarch) we accept that the host instruction will operate on all 
128-bits; we simply ignore half of any result.

Thus the question becomes: can we minimize the number of vset* instructions by bounding 
minimal lmul to 1 (or whatever) and always leaving avl as the full register?  If so, the 
only vset* changes are for SEW changes, or for load/store that are smaller than V*1REG64.


r~