eden-emu/eden
24
4
Fork 15
Code Issues 139 Pull requests 53 Projects 1 Releases Packages Wiki Activity Actions

[VK] [DO NOT MERGE] Force subgroup size to 32 and disable demote
All checks were successful
eden-license / license-header (pull_request) Successful in 28s
Details

Browse source
This commit is contained in:
JPikachu 2025-08-30 01:40:55 +01:00
parent 57fbdd516e
commit 744b6717ab
5 changed files with 32 additions and 17 deletions
Download patch file Download diff file Expand all files Collapse all files

30
src/shader_recompiler/backend/spirv/emit_spirv_warp.cpp
View file

@ -1,3 +1,6 @@
// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
@ -37,6 +40,7 @@ Id WarpExtract(EmitContext& ctx, Id value) {
Id LoadMask(EmitContext& ctx, Id mask) {
const Id value{ctx.OpLoad(ctx.U32[4], mask)};
return ctx.OpCompositeExtract(ctx.U32[1], value, 0U);
if (!ctx.profile.warp_size_potentially_larger_than_guest) {
return ctx.OpCompositeExtract(ctx.U32[1], value, 0U);
}
@ -74,6 +78,7 @@ Id SelectValue(EmitContext& ctx, Id in_range, Id value, Id src_thread_id) {
}
Id AddPartitionBase(EmitContext& ctx, Id thread_id) {
return thread_id;
const Id partition_idx{ctx.OpShiftRightLogical(ctx.U32[1], GetThreadId(ctx), ctx.Const(5u))};
const Id partition_base{ctx.OpShiftLeftLogical(ctx.U32[1], partition_idx, ctx.Const(5u))};
return ctx.OpIAdd(ctx.U32[1], thread_id, partition_base);
@ -82,6 +87,7 @@ Id AddPartitionBase(EmitContext& ctx, Id thread_id) {
Id EmitLaneId(EmitContext& ctx) {
const Id id{GetThreadId(ctx)};
return id;
if (!ctx.profile.warp_size_potentially_larger_than_guest) {
return id;
}
@ -166,13 +172,15 @@ Id EmitShuffleIndex(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id cla
const Id lhs{ctx.OpBitwiseAnd(ctx.U32[1], index, not_seg_mask)};
Id src_thread_id{ctx.OpBitwiseOr(ctx.U32[1], lhs, min_thread_id)};
const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};
const Id same_lane{ctx.OpINotEqual(ctx.U1, src_thread_id, thread_id)};
const Id is_value_being_shuffled{ctx.OpLogicalAnd(ctx.U1, in_range, same_lane)};
if (ctx.profile.warp_size_potentially_larger_than_guest) {
src_thread_id = AddPartitionBase(ctx, src_thread_id);
}
SetInBoundsFlag(inst, in_range);
return SelectValue(ctx, in_range, value, src_thread_id);
SetInBoundsFlag(inst, is_value_being_shuffled);
return SelectValue(ctx, is_value_being_shuffled, value, src_thread_id);
}
Id EmitShuffleUp(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
@ -181,13 +189,15 @@ Id EmitShuffleUp(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
Id src_thread_id{ctx.OpISub(ctx.U32[1], thread_id, index)};
const Id in_range{ctx.OpSGreaterThanEqual(ctx.U1, src_thread_id, max_thread_id)};
const Id same_lane{ctx.OpINotEqual(ctx.U1, src_thread_id, thread_id)};
const Id is_value_being_shuffled{ctx.OpLogicalAnd(ctx.U1, in_range, same_lane)};
if (ctx.profile.warp_size_potentially_larger_than_guest) {
src_thread_id = AddPartitionBase(ctx, src_thread_id);
}
SetInBoundsFlag(inst, in_range);
return SelectValue(ctx, in_range, value, src_thread_id);
SetInBoundsFlag(inst, is_value_being_shuffled);
return SelectValue(ctx, is_value_being_shuffled, value, src_thread_id);
}
Id EmitShuffleDown(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
@ -196,13 +206,15 @@ Id EmitShuffleDown(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clam
const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
Id src_thread_id{ctx.OpIAdd(ctx.U32[1], thread_id, index)};
const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};
const Id same_lane{ctx.OpINotEqual(ctx.U1, src_thread_id, thread_id)};
const Id is_value_being_shuffled{ctx.OpLogicalAnd(ctx.U1, in_range, same_lane)};
if (ctx.profile.warp_size_potentially_larger_than_guest) {
src_thread_id = AddPartitionBase(ctx, src_thread_id);
}
SetInBoundsFlag(inst, in_range);
return SelectValue(ctx, in_range, value, src_thread_id);
SetInBoundsFlag(inst, is_value_being_shuffled);
return SelectValue(ctx, is_value_being_shuffled, value, src_thread_id);
}
Id EmitShuffleButterfly(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
@ -211,13 +223,15 @@ Id EmitShuffleButterfly(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id
const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
Id src_thread_id{ctx.OpBitwiseXor(ctx.U32[1], thread_id, index)};
const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};
const Id same_lane{ctx.OpINotEqual(ctx.U1, src_thread_id, thread_id)};
const Id is_value_being_shuffled{ctx.OpLogicalAnd(ctx.U1, in_range, same_lane)};
if (ctx.profile.warp_size_potentially_larger_than_guest) {
src_thread_id = AddPartitionBase(ctx, src_thread_id);
}
SetInBoundsFlag(inst, in_range);
return SelectValue(ctx, in_range, value, src_thread_id);
SetInBoundsFlag(inst, is_value_being_shuffled);
return SelectValue(ctx, is_value_being_shuffled, value, src_thread_id);
}
Id EmitFSwizzleAdd(EmitContext& ctx, Id op_a, Id op_b, Id swizzle) {