mirror/bazzite
1
0
Fork 0
mirror of https://github.com/ublue-os/bazzite.git synced 2025-01-01 03:21:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

chore: Update patch for RADV RT

Browse Source
This commit is contained in:
Kyle Gospodnetich 2024-01-05 12:24:00 -08:00
parent 566bb401ee
commit dc2fa7929f
1 changed files with 11 additions and 32 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

43
spec_files/mesa/26105.patch
View File

@ -1,26 +1,8 @@
From 1e3541728b63cd635aac7211b76798edfa8c34eb Mon Sep 17 00:00:00 2001
From: Friedrich Vock <friedrich.vock@gmx.de>
Date: Tue, 7 Nov 2023 22:28:44 +0100
Subject: [PATCH] radv,aco: Convert 1D ray launches to 2D
Because we use unaligned dispatches, 1D launches only use 8 threads per
wave. Converting to 2D and fixing up launch IDs in the prolog
significantly increases occupancy.
Gives ~30% uplift in Ghostwire Tokyo.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26105>
---
.../compiler/aco_instruction_selection.cpp | 43 ++++++++++-
src/amd/compiler/aco_interface.h | 3 +
src/amd/vulkan/radv_cmd_buffer.c | 76 +++++++++++++++----
3 files changed, 103 insertions(+), 19 deletions(-)
diff --git a/src/amd/compiler/aco_instruction_selection.cpp b/src/amd/compiler/aco_instruction_selection.cpp
index 4bcf85f8ed644..bdf81dff3589d 100644
index 561f3cc02e0..ebc54a900e3 100644
--- a/src/amd/compiler/aco_instruction_selection.cpp
+++ b/src/amd/compiler/aco_instruction_selection.cpp
@@ -12534,7 +12534,8 @@ select_rt_prolog(Program* program, ac_shader_config* config,
@@ -12526,7 +12526,8 @@ select_rt_prolog(Program* program, ac_shader_config* config,
*/
PhysReg out_uniform_shader_addr = get_arg_reg(out_args, out_args->rt.uniform_shader_addr);
PhysReg out_launch_size_x = get_arg_reg(out_args, out_args->rt.launch_size);
@ -30,7 +12,7 @@ index 4bcf85f8ed644..bdf81dff3589d 100644
PhysReg out_launch_ids[3];
for (unsigned i = 0; i < 3; i++)
out_launch_ids[i] = get_arg_reg(out_args, out_args->rt.launch_id).advance(i * 4);
@@ -12542,9 +12543,13 @@ select_rt_prolog(Program* program, ac_shader_config* config,
@@ -12534,9 +12535,13 @@ select_rt_prolog(Program* program, ac_shader_config* config,
PhysReg out_record_ptr = get_arg_reg(out_args, out_args->rt.shader_record);
/* Temporaries: */
@ -47,7 +29,7 @@ index 4bcf85f8ed644..bdf81dff3589d 100644
/* Confirm some assumptions about register aliasing */
assert(in_ring_offsets == out_uniform_shader_addr);
@@ -12618,6 +12623,36 @@ select_rt_prolog(Program* program, ac_shader_config* config,
@@ -12610,6 +12615,36 @@ select_rt_prolog(Program* program, ac_shader_config* config,
bld.vop1(aco_opcode::v_mov_b32, Definition(out_record_ptr.advance(4), v1),
Operand(tmp_raygen_sbt.advance(4), s1));
@ -85,21 +67,21 @@ index 4bcf85f8ed644..bdf81dff3589d 100644
bld.sop1(aco_opcode::s_setpc_b64, Operand(out_uniform_shader_addr, s2));
diff --git a/src/amd/compiler/aco_interface.h b/src/amd/compiler/aco_interface.h
index 85c270ba19983..15e5398416b3c 100644
index 8f35e18b5b0..9d2c1dbb2af 100644
--- a/src/amd/compiler/aco_interface.h
+++ b/src/amd/compiler/aco_interface.h
@@ -34,6 +34,9 @@
@@ -32,6 +32,9 @@
extern "C" {
#endif
+/* Special launch size to indicate this dispatch is a 1D dispatch converted into a 2D one */
+#define ACO_RT_CONVERTED_2D_LAUNCH_SIZE -1u
+
typedef struct nir_shader nir_shader;
struct ac_shader_config;
struct aco_shader_info;
struct aco_vs_prolog_info;
diff --git a/src/amd/vulkan/radv_cmd_buffer.c b/src/amd/vulkan/radv_cmd_buffer.c
index dab2230d778a5..81d553cf32fe7 100644
index cdede679552..c7dd4b216d4 100644
--- a/src/amd/vulkan/radv_cmd_buffer.c
+++ b/src/amd/vulkan/radv_cmd_buffer.c
@@ -42,6 +42,8 @@
@ -111,7 +93,7 @@ index dab2230d778a5..81d553cf32fe7 100644
#include "util/fast_idiv_by_const.h"
enum {
@@ -9963,7 +9965,26 @@ enum radv_rt_mode {
@@ -10003,7 +10005,26 @@ enum radv_rt_mode {
};
static void
@ -139,7 +121,7 @@ index dab2230d778a5..81d553cf32fe7 100644
enum radv_rt_mode mode)
{
if (cmd_buffer->device->instance->debug_flags & RADV_DEBUG_NO_RT)
@@ -9984,34 +10005,43 @@ radv_trace_rays(struct radv_cmd_buffer *cmd_buffer, const VkTraceRaysIndirectCom
@@ -10024,34 +10045,43 @@ radv_trace_rays(struct radv_cmd_buffer *cmd_buffer, const VkTraceRaysIndirectCom
cmd_buffer->compute_scratch_size_per_wave_needed =
MAX2(cmd_buffer->compute_scratch_size_per_wave_needed, scratch_bytes_per_wave);
@ -197,7 +179,7 @@ index dab2230d778a5..81d553cf32fe7 100644
} else
info.va = launch_size_va;
@@ -10045,6 +10075,22 @@ radv_trace_rays(struct radv_cmd_buffer *cmd_buffer, const VkTraceRaysIndirectCom
@@ -10085,6 +10115,22 @@ radv_trace_rays(struct radv_cmd_buffer *cmd_buffer, const VkTraceRaysIndirectCom
assert(cmd_buffer->cs->cdw <= cdw_max);
radv_dispatch(cmd_buffer, &info, pipeline, rt_prolog, VK_PIPELINE_BIND_POINT_RAY_TRACING_KHR);
@ -220,6 +202,3 @@ index dab2230d778a5..81d553cf32fe7 100644
}
VKAPI_ATTR void VKAPI_CALL
--
GitLab