mirror of
https://gitlab.freedesktop.org/mesa/mesa.git
synced 2026-05-02 01:28:07 +02:00
110f14d4b4
The idea is that for repeated use of the same uniform, we could avoid loading it on each consumer. The results look pretty good. total instructions in shared programs: 6413571 -> 6521464 (1.68%) total threads in shared programs: 154214 -> 154000 (-0.14%) total uniforms in shared programs: 2393604 -> 2119629 (-11.45%) total spills in shared programs: 4960 -> 4984 (0.48%) total fills in shared programs: 6350 -> 6418 (1.07%) Once we do scheduling at the NIR level, the register pressure (and thus also instructions) issues we see here will drop back down.
254 lines
8.8 KiB
C
254 lines
8.8 KiB
C
/*
|
|
* Copyright © 2014 Broadcom
|
|
*
|
|
* Permission is hereby granted, free of charge, to any person obtaining a
|
|
* copy of this software and associated documentation files (the "Software"),
|
|
* to deal in the Software without restriction, including without limitation
|
|
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
|
|
* and/or sell copies of the Software, and to permit persons to whom the
|
|
* Software is furnished to do so, subject to the following conditions:
|
|
*
|
|
* The above copyright notice and this permission notice (including the next
|
|
* paragraph) shall be included in all copies or substantial portions of the
|
|
* Software.
|
|
*
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
|
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
|
|
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
|
|
* IN THE SOFTWARE.
|
|
*/
|
|
|
|
/**
|
|
* @file v3d_opt_copy_propagation.c
|
|
*
|
|
* This implements simple copy propagation for VIR without control flow.
|
|
*
|
|
* For each temp, it keeps a qreg of which source it was MOVed from, if it
|
|
* was. If we see that used later, we can just reuse the source value, since
|
|
* we know we don't have control flow, and we have SSA for our values so
|
|
* there's no killing to worry about.
|
|
*/
|
|
|
|
#include "v3d_compiler.h"
|
|
|
|
static bool
|
|
is_copy_mov(struct qinst *inst)
|
|
{
|
|
if (!inst)
|
|
return false;
|
|
|
|
if (inst->qpu.type != V3D_QPU_INSTR_TYPE_ALU ||
|
|
(inst->qpu.alu.mul.op != V3D_QPU_M_FMOV &&
|
|
inst->qpu.alu.mul.op != V3D_QPU_M_MOV)) {
|
|
return false;
|
|
}
|
|
|
|
if (inst->dst.file != QFILE_TEMP)
|
|
return false;
|
|
|
|
if (inst->src[0].file != QFILE_TEMP)
|
|
return false;
|
|
|
|
if (inst->qpu.alu.add.output_pack != V3D_QPU_PACK_NONE ||
|
|
inst->qpu.alu.mul.output_pack != V3D_QPU_PACK_NONE) {
|
|
return false;
|
|
}
|
|
|
|
if (inst->qpu.flags.ac != V3D_QPU_COND_NONE ||
|
|
inst->qpu.flags.mc != V3D_QPU_COND_NONE) {
|
|
return false;
|
|
}
|
|
|
|
switch (inst->src[0].file) {
|
|
case QFILE_MAGIC:
|
|
/* No copy propagating from R3/R4/R5 -- the MOVs from those
|
|
* are there to register allocate values produced into R3/4/5
|
|
* to other regs (though hopefully r3/4/5).
|
|
*/
|
|
switch (inst->src[0].index) {
|
|
case V3D_QPU_WADDR_R3:
|
|
case V3D_QPU_WADDR_R4:
|
|
case V3D_QPU_WADDR_R5:
|
|
return false;
|
|
default:
|
|
break;
|
|
}
|
|
break;
|
|
|
|
case QFILE_REG:
|
|
switch (inst->src[0].index) {
|
|
case 0:
|
|
case 1:
|
|
case 2:
|
|
/* MOVs from rf0/1/2 are only to track the live
|
|
* intervals for W/centroid W/Z.
|
|
*/
|
|
return false;
|
|
}
|
|
break;
|
|
|
|
default:
|
|
break;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
static bool
|
|
vir_has_unpack(struct qinst *inst, int chan)
|
|
{
|
|
assert(chan == 0 || chan == 1);
|
|
|
|
if (vir_is_add(inst)) {
|
|
if (chan == 0)
|
|
return inst->qpu.alu.add.a_unpack != V3D_QPU_UNPACK_NONE;
|
|
else
|
|
return inst->qpu.alu.add.b_unpack != V3D_QPU_UNPACK_NONE;
|
|
} else {
|
|
if (chan == 0)
|
|
return inst->qpu.alu.mul.a_unpack != V3D_QPU_UNPACK_NONE;
|
|
else
|
|
return inst->qpu.alu.mul.b_unpack != V3D_QPU_UNPACK_NONE;
|
|
}
|
|
}
|
|
|
|
static bool
|
|
try_copy_prop(struct v3d_compile *c, struct qinst *inst, struct qinst **movs)
|
|
{
|
|
bool debug = false;
|
|
bool progress = false;
|
|
|
|
for (int i = 0; i < vir_get_nsrc(inst); i++) {
|
|
if (inst->src[i].file != QFILE_TEMP)
|
|
continue;
|
|
|
|
/* We have two ways of finding MOVs we can copy propagate
|
|
* from. One is if it's an SSA def: then we can reuse it from
|
|
* any block in the program, as long as its source is also an
|
|
* SSA def. Alternatively, if it's in the "movs" array
|
|
* tracked within the block, then we know the sources for it
|
|
* haven't been changed since we saw the instruction within
|
|
* our block.
|
|
*/
|
|
struct qinst *mov = movs[inst->src[i].index];
|
|
if (!mov) {
|
|
if (!is_copy_mov(c->defs[inst->src[i].index]))
|
|
continue;
|
|
mov = c->defs[inst->src[i].index];
|
|
|
|
if (mov->src[0].file == QFILE_TEMP &&
|
|
!c->defs[mov->src[0].index])
|
|
continue;
|
|
}
|
|
|
|
if (vir_has_unpack(mov, 0)) {
|
|
/* Make sure that the meaning of the unpack
|
|
* would be the same between the two
|
|
* instructions.
|
|
*/
|
|
if (v3d_qpu_unpacks_f32(&inst->qpu) !=
|
|
v3d_qpu_unpacks_f32(&mov->qpu) ||
|
|
v3d_qpu_unpacks_f16(&inst->qpu) !=
|
|
v3d_qpu_unpacks_f16(&mov->qpu)) {
|
|
continue;
|
|
}
|
|
|
|
/* No composing the unpacks. */
|
|
if (vir_has_unpack(inst, i))
|
|
continue;
|
|
|
|
/* these ops can't represent abs. */
|
|
if (mov->qpu.alu.mul.a_unpack == V3D_QPU_UNPACK_ABS) {
|
|
switch (inst->qpu.alu.add.op) {
|
|
case V3D_QPU_A_VFPACK:
|
|
case V3D_QPU_A_FROUND:
|
|
case V3D_QPU_A_FTRUNC:
|
|
case V3D_QPU_A_FFLOOR:
|
|
case V3D_QPU_A_FCEIL:
|
|
case V3D_QPU_A_FDX:
|
|
case V3D_QPU_A_FDY:
|
|
case V3D_QPU_A_FTOIN:
|
|
case V3D_QPU_A_FTOIZ:
|
|
case V3D_QPU_A_FTOUZ:
|
|
case V3D_QPU_A_FTOC:
|
|
continue;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
if (debug) {
|
|
fprintf(stderr, "Copy propagate: ");
|
|
vir_dump_inst(c, inst);
|
|
fprintf(stderr, "\n");
|
|
}
|
|
|
|
inst->src[i] = mov->src[0];
|
|
if (vir_has_unpack(mov, 0)) {
|
|
enum v3d_qpu_input_unpack unpack = mov->qpu.alu.mul.a_unpack;
|
|
|
|
vir_set_unpack(inst, i, unpack);
|
|
}
|
|
|
|
if (debug) {
|
|
fprintf(stderr, "to: ");
|
|
vir_dump_inst(c, inst);
|
|
fprintf(stderr, "\n");
|
|
}
|
|
|
|
progress = true;
|
|
}
|
|
|
|
return progress;
|
|
}
|
|
|
|
static void
|
|
apply_kills(struct v3d_compile *c, struct qinst **movs, struct qinst *inst)
|
|
{
|
|
if (inst->dst.file != QFILE_TEMP)
|
|
return;
|
|
|
|
for (int i = 0; i < c->num_temps; i++) {
|
|
if (movs[i] &&
|
|
(movs[i]->dst.index == inst->dst.index ||
|
|
(movs[i]->src[0].file == QFILE_TEMP &&
|
|
movs[i]->src[0].index == inst->dst.index))) {
|
|
movs[i] = NULL;
|
|
}
|
|
}
|
|
}
|
|
|
|
bool
|
|
vir_opt_copy_propagate(struct v3d_compile *c)
|
|
{
|
|
bool progress = false;
|
|
struct qinst **movs;
|
|
|
|
movs = ralloc_array(c, struct qinst *, c->num_temps);
|
|
if (!movs)
|
|
return false;
|
|
|
|
vir_for_each_block(block, c) {
|
|
/* The MOVs array tracks only available movs within the
|
|
* block.
|
|
*/
|
|
memset(movs, 0, sizeof(struct qinst *) * c->num_temps);
|
|
|
|
vir_for_each_inst(inst, block) {
|
|
progress = try_copy_prop(c, inst, movs) || progress;
|
|
|
|
apply_kills(c, movs, inst);
|
|
|
|
if (is_copy_mov(inst))
|
|
movs[inst->dst.index] = inst;
|
|
}
|
|
}
|
|
|
|
ralloc_free(movs);
|
|
|
|
return progress;
|
|
}
|