On 8/25/22 15:13, Paolo Bonzini wrote:
> @@ -3112,6 +3165,7 @@ static void gen_sse(CPUX86State *env, DisasContext *s, int b,
> {
> int b1, op1_offset, op2_offset, is_xmm, val;
> int modrm, mod, rm, reg;
> + struct SSEOpHelper_table1 sse_op;
> SSEFunc_0_epp sse_fn_epp;
> SSEFunc_0_eppi sse_fn_eppi;
> SSEFunc_0_ppi sse_fn_ppi;
> @@ -3127,8 +3181,10 @@ static void gen_sse(CPUX86State *env, DisasContext *s, int b,
> b1 = 3;
> else
> b1 = 0;
> - sse_fn_epp = sse_op_table1[b][b1];
> - if (!sse_fn_epp) {
> + sse_op = sse_op_table1[b];
> + sse_fn_epp = sse_op.op[b1];
> + if ((sse_op.flags & (SSE_OPF_SPECIAL | SSE_OPF_3DNOW)) == 0
> + && !sse_fn_epp) {
Do we really want a 5 word structure copy into the local stack frame here?
Surely better as
const struct SSEOpHelper_table1 *sse_op;
...
sse_op = &sse_op_table1[b];
I'll split it to two variables, one for the flags and one for fn[b1]. Op6 and op7 will just be pointers instead. Rebasing the AVX patches (I want to keep them up to date even if I end up replacing the decoder) was easier than I thought.
Paolo