fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-09 02:28:10 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

jay/register_allocate: start using accumulators

Browse source 
this lets us lower away 8<-->2 copies/swaps in a faster, more straightforward
way by (ab)using accumulators. I think as an edge case this plays nicely enough
with my plans to profit from accs for normal fma-heavy code.

SIMD16:

   Totals:
   Instrs: 2761525 -> 2758108 (-0.12%)
   CodeSize: 44222384 -> 44167168 (-0.12%)

   Totals from 33 (1.25% of 2647) affected shaders:
   Instrs: 422130 -> 418713 (-0.81%)
   CodeSize: 6713680 -> 6658464 (-0.82%)

SIMD32:

   Totals:
   Instrs: 4911601 -> 4691895 (-4.47%)
   CodeSize: 79553984 -> 76010880 (-4.45%)

   Totals from 947 (35.78% of 2647) affected shaders:
   Instrs: 4143501 -> 3923795 (-5.30%)
   CodeSize: 67174592 -> 63631488 (-5.27%)

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41064>
This commit is contained in:
Alyssa Rosenzweig 2026-04-20 11:04:59 -04:00 • committed by Marge Bot
parent 53c1c076a8
commit 4c5ad7a832
1 changed files with 19 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

24
src/intel/compiler/jay/jay_register_allocate.c
View file

@ -374,11 +374,14 @@ add_copy(struct util_dynarray *copies, jay_reg dst, jay_reg src)
static jay_def
push_temp(jay_builder *b, jay_reg reg, bool stride4)
{
jay_def tmp = def_from_reg(reg);
jay_def tmp = reg == NO_REG ? jay_null() : def_from_reg(reg);
if (stride4 && jay_def_stride(b->shader, tmp) != JAY_STRIDE_4) {
if (jay_is_null(tmp) ||
(stride4 && jay_def_stride(b->shader, tmp) != JAY_STRIDE_4)) {
/* Put accumulators down the float pipe - it's still a raw move. */
jay_def new = def_from_reg(0);
jay_MOV(b, tmp, new);
jay_MOV(b, jay_bare_reg(ACCUM, 0), new)->type = JAY_TYPE_F32;
tmp = new;
}
@ -388,8 +391,8 @@ push_temp(jay_builder *b, jay_reg reg, bool stride4)
static void
pop_temp(jay_builder *b, struct jay_temp_regs t, jay_def temp)
{
if (temp.file == GPR && temp.reg != t.gpr) {
jay_MOV(b, temp, def_from_reg(t.gpr));
if (temp.file == GPR && make_reg(GPR, temp.reg) != t.gpr) {
jay_MOV(b, temp, jay_bare_reg(ACCUM, 0))->type = JAY_TYPE_F32;
}
}
@ -410,6 +413,17 @@ mov(jay_builder *b, jay_def dst, jay_def src, struct jay_temp_regs temps)
assert(temps.ugpr != NO_REG && "ensured by the spill limit");
jay_MOV(b, def_from_reg(temps.ugpr), src);
jay_MOV(b, dst, def_from_reg(temps.ugpr));
} else if (dst.file == GPR &&
src.file == GPR &&
jay_def_stride(b->shader, dst) !=
jay_def_stride(b->shader, src) &&
jay_def_stride(b->shader, dst) != JAY_STRIDE_4 &&
jay_def_stride(b->shader, src) != JAY_STRIDE_4) {
jay_def temp = push_temp(b, temps.gpr, true /* stride4 */);
jay_MOV(b, temp, src);
jay_MOV(b, dst, temp);
pop_temp(b, temps, temp);
} else {
jay_MOV(b, dst, src);
}