fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-05 16:08:04 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

aco: set compr for fp16 exports

Browse source 
Obviously this didn't affect correctness. Not sure about performance.

It also changes enabled_channels to match radeonsi.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: f29c81f863 ("aco: use VOP2 for v_cvt_pkrtz_f16_f32 if possible")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9459>
(cherry picked from commit 341dd9d834)
This commit is contained in:
Rhys Perry 2021-03-05 20:01:05 +00:00 • committed by Dylan Baker
parent 60c959b3a2
commit 334a5437a9
2 changed files with 10 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
.pick_status.json
View file

@ -247,7 +247,7 @@
"description": "aco: set compr for fp16 exports",
"nominated": true,
"nomination_type": 1,
"resolution": 0,
"resolution": 1,
"master_sha": null,
"because_sha": "f29c81f863c9879a6a87724cbdae1e1818f3f6b4"
},

13
src/amd/compiler/aco_instruction_selection.cpp
View file

@ -10364,6 +10364,7 @@ static bool export_fs_mrt_color(isel_context *ctx, int slot)
unsigned target, col_format;
unsigned enabled_channels = 0;
aco_opcode compr_op = (aco_opcode)0;
bool compr = false;
slot -= FRAG_RESULT_DATA0;
target = V_008DFC_SQ_EXP_MRT + slot;
@ -10418,7 +10419,7 @@ static bool export_fs_mrt_color(isel_context *ctx, int slot)
for (int i = 0; i < 2; i++) {
bool enabled = (write_mask >> (i*2)) & 0x3;
if (enabled) {
enabled_channels |= 1 << i;
enabled_channels |= 0x3 << (i*2);
if (is_16bit) {
values[i] = bld.pseudo(aco_opcode::p_create_vector, bld.def(v1),
values[i*2].isUndefined() ? Operand(v2b) : values[i*2],
@ -10436,6 +10437,9 @@ static bool export_fs_mrt_color(isel_context *ctx, int slot)
values[i] = Operand(v1);
}
}
values[2] = Operand(v1);
values[3] = Operand(v1);
compr = true;
break;
case V_028714_SPI_SHADER_UNORM16_ABGR:
@ -10522,7 +10526,7 @@ static bool export_fs_mrt_color(isel_context *ctx, int slot)
/* check if at least one of the values to be compressed is enabled */
bool enabled = (write_mask >> (i*2)) & 0x3;
if (enabled) {
enabled_channels |= 1 << (i*2);
enabled_channels |= 0x3 << (i*2);
values[i] = bld.vop3(compr_op, bld.def(v1),
values[i*2].isUndefined() ? Operand(0u) : values[i*2],
values[i*2+1].isUndefined() ? Operand(0u): values[i*2+1]);
@ -10532,13 +10536,14 @@ static bool export_fs_mrt_color(isel_context *ctx, int slot)
}
values[2] = Operand(v1);
values[3] = Operand(v1);
} else {
compr = true;
} else if (!compr) {
for (int i = 0; i < 4; i++)
values[i] = enabled_channels & (1 << i) ? values[i] : Operand(v1);
}
bld.exp(aco_opcode::exp, values[0], values[1], values[2], values[3],
enabled_channels, target, (bool) compr_op);
enabled_channels, target, compr);
return true;
}