OpenHarmony/third_party_mesa3d
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

intel/compiler: Combine nir_emit_{ssbo,shared}_atomic into one helper

Browse Source 
These are basically identical save for:
- shared has surface hardcoded to SLM rather than an SSBO index
- shared has to handle adding the 'base' const_index (SSBO have none)
- the NIR source index for data is shifted by one

It's not worth copy and pasting the entire function for this.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Rohan Garg <rohan.garg@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20604>
This commit is contained in:
Kenneth Graunke
2023-01-09 16:23:08 -08:00
committed by Marge Bot
parent b84939c678
commit 03ddde1230
2 changed files with 31 additions and 79 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
src/intel/compiler/brw_fs.h
View File

@@ -359,10 +359,9 @@ public:
nir_intrinsic_instr *instr);
void nir_emit_tes_intrinsic(const brw::fs_builder &bld,
nir_intrinsic_instr *instr);
void nir_emit_ssbo_atomic(const brw::fs_builder &bld,
nir_intrinsic_instr *instr);
void nir_emit_shared_atomic(const brw::fs_builder &bld,
nir_intrinsic_instr *instr);
void nir_emit_surface_atomic(const brw::fs_builder &bld,
nir_intrinsic_instr *instr,
fs_reg surface);
void nir_emit_global_atomic(const brw::fs_builder &bld,
nir_intrinsic_instr *instr);
void nir_emit_global_atomic_float(const brw::fs_builder &bld,

103
src/intel/compiler/brw_fs_nir.cpp
View File

@@ -3782,7 +3782,7 @@ fs_visitor::nir_emit_cs_intrinsic(const fs_builder &bld,
case nir_intrinsic_shared_atomic_fmin:
case nir_intrinsic_shared_atomic_fmax:
case nir_intrinsic_shared_atomic_fcomp_swap:
nir_emit_shared_atomic(bld, instr);
nir_emit_surface_atomic(bld, instr, brw_imm_ud(GFX7_BTI_SLM));
break;
case nir_intrinsic_load_shared: {
@@ -5055,7 +5055,8 @@ fs_visitor::nir_emit_intrinsic(const fs_builder &bld, nir_intrinsic_instr *instr
case nir_intrinsic_ssbo_atomic_fmin:
case nir_intrinsic_ssbo_atomic_fmax:
case nir_intrinsic_ssbo_atomic_fcomp_swap:
nir_emit_ssbo_atomic(bld, instr);
nir_emit_surface_atomic(bld, instr,
get_nir_ssbo_intrinsic_index(bld, instr));
break;
case nir_intrinsic_get_ssbo_size: {
@@ -5951,12 +5952,15 @@ expand_to_32bit(const fs_builder &bld, const fs_reg &src)
}
void
fs_visitor::nir_emit_ssbo_atomic(const fs_builder &bld,
nir_intrinsic_instr *instr)
fs_visitor::nir_emit_surface_atomic(const fs_builder &bld,
nir_intrinsic_instr *instr,
fs_reg surface)
{
enum lsc_opcode op = lsc_aop_for_nir_intrinsic(instr);
int num_data = lsc_op_num_data_values(op);
bool shared = surface.file == IMM && surface.ud == GFX7_BTI_SLM;
/* The BTI untyped atomic messages only support 32-bit atomics. If you
* just look at the big table of messages in the Vol 7 of the SKL PRM, they
* appear to exist. However, if you look at Vol 2a, there are no message
@@ -5974,21 +5978,37 @@ fs_visitor::nir_emit_ssbo_atomic(const fs_builder &bld,
dest = get_nir_dest(instr->dest);
fs_reg srcs[SURFACE_LOGICAL_NUM_SRCS];
srcs[SURFACE_LOGICAL_SRC_SURFACE] = get_nir_ssbo_intrinsic_index(bld, instr);
srcs[SURFACE_LOGICAL_SRC_ADDRESS] = get_nir_src(instr->src[1]);
srcs[SURFACE_LOGICAL_SRC_SURFACE] = surface;
srcs[SURFACE_LOGICAL_SRC_IMM_DIMS] = brw_imm_ud(1);
srcs[SURFACE_LOGICAL_SRC_IMM_ARG] = brw_imm_ud(op);
srcs[SURFACE_LOGICAL_SRC_ALLOW_SAMPLE_MASK] = brw_imm_ud(1);
if (shared) {
/* SLM - Get the offset */
if (nir_src_is_const(instr->src[0])) {
srcs[SURFACE_LOGICAL_SRC_ADDRESS] =
brw_imm_ud(nir_intrinsic_base(instr) +
nir_src_as_uint(instr->src[0]));
} else {
srcs[SURFACE_LOGICAL_SRC_ADDRESS] = vgrf(glsl_type::uint_type);
bld.ADD(srcs[SURFACE_LOGICAL_SRC_ADDRESS],
retype(get_nir_src(instr->src[0]), BRW_REGISTER_TYPE_UD),
brw_imm_ud(nir_intrinsic_base(instr)));
}
} else {
/* SSBOs */
srcs[SURFACE_LOGICAL_SRC_ADDRESS] = get_nir_src(instr->src[1]);
}
fs_reg data;
if (num_data >= 1)
data = expand_to_32bit(bld, get_nir_src(instr->src[2]));
data = expand_to_32bit(bld, get_nir_src(instr->src[shared ? 1 : 2]));
if (num_data >= 2) {
fs_reg tmp = bld.vgrf(data.type, 2);
fs_reg sources[2] = {
data,
expand_to_32bit(bld, get_nir_src(instr->src[3]))
expand_to_32bit(bld, get_nir_src(instr->src[shared ? 2 : 3]))
};
bld.LOAD_PAYLOAD(tmp, sources, 2, 0);
data = tmp;
@@ -6018,73 +6038,6 @@ fs_visitor::nir_emit_ssbo_atomic(const fs_builder &bld,
}
}
void
fs_visitor::nir_emit_shared_atomic(const fs_builder &bld,
nir_intrinsic_instr *instr)
{
enum lsc_opcode op = lsc_aop_for_nir_intrinsic(instr);
int num_data = lsc_op_num_data_values(op);
fs_reg dest;
if (nir_intrinsic_infos[instr->intrinsic].has_dest)
dest = get_nir_dest(instr->dest);
fs_reg srcs[SURFACE_LOGICAL_NUM_SRCS];
srcs[SURFACE_LOGICAL_SRC_SURFACE] = brw_imm_ud(GFX7_BTI_SLM);
srcs[SURFACE_LOGICAL_SRC_IMM_DIMS] = brw_imm_ud(1);
srcs[SURFACE_LOGICAL_SRC_IMM_ARG] = brw_imm_ud(op);
srcs[SURFACE_LOGICAL_SRC_ALLOW_SAMPLE_MASK] = brw_imm_ud(1);
fs_reg data;
if (num_data >= 1)
data = expand_to_32bit(bld, get_nir_src(instr->src[1]));
if (num_data >= 2) {
fs_reg tmp = bld.vgrf(data.type, 2);
fs_reg sources[2] = {
data,
expand_to_32bit(bld, get_nir_src(instr->src[2]))
};
bld.LOAD_PAYLOAD(tmp, sources, 2, 0);
data = tmp;
}
srcs[SURFACE_LOGICAL_SRC_DATA] = data;
/* Get the offset */
if (nir_src_is_const(instr->src[0])) {
srcs[SURFACE_LOGICAL_SRC_ADDRESS] =
brw_imm_ud(nir_intrinsic_base(instr) +
nir_src_as_uint(instr->src[0]));
} else {
srcs[SURFACE_LOGICAL_SRC_ADDRESS] = vgrf(glsl_type::uint_type);
bld.ADD(srcs[SURFACE_LOGICAL_SRC_ADDRESS],
retype(get_nir_src(instr->src[0]), BRW_REGISTER_TYPE_UD),
brw_imm_ud(nir_intrinsic_base(instr)));
}
/* Emit the actual atomic operation operation */
switch (nir_dest_bit_size(instr->dest)) {
case 16: {
fs_reg dest32 = bld.vgrf(BRW_REGISTER_TYPE_UD);
bld.emit(SHADER_OPCODE_UNTYPED_ATOMIC_LOGICAL,
retype(dest32, dest.type),
srcs, SURFACE_LOGICAL_NUM_SRCS);
bld.MOV(retype(dest, BRW_REGISTER_TYPE_UW),
retype(dest32, BRW_REGISTER_TYPE_UD));
break;
}
case 32:
case 64:
bld.emit(SHADER_OPCODE_UNTYPED_ATOMIC_LOGICAL,
dest, srcs, SURFACE_LOGICAL_NUM_SRCS);
break;
default:
unreachable("Unsupported bit size");
}
}
void
fs_visitor::nir_emit_global_atomic(const fs_builder &bld,
nir_intrinsic_instr *instr)