marfrit
d66f22f333
First QPU IDCT8 kernel running and bit-exact on V3D 7.1 via Mesa
v3dv compute. Five iterations through a Phase 7→Phase 4' loopback;
production kernel is v4.
New files:
- src/v3d_runner.{c,h} — reusable Vulkan compute plumbing (instance,
V3D device picker, HOST_VISIBLE|COHERENT
SSBOs with mmap, compute pipeline from .spv,
enables storageBuffer{8,16}BitAccess)
- src/v3d_idct8.comp — VP9 8x8 DCT_DCT IDCT add, v4 production:
256 invocations/WG, 2 blocks/subgroup
(no idle lanes), uint8 dst SSBO (race-free
per phase5 finding 5), unrolled writes
(no chained ternary), oob-flag pattern
(barrier-safe per phase5 finding 7)
- tests/bench_v3d_idct.c — M1' bit-exact gate + M2 throughput vs C ref
- docs/phase7.md — full iteration journey + decision verdict
CMakeLists.txt updated to build the new shader, library, and bench
when DAEDALUS_BUILD_VULKAN=ON.
Iteration record (1920x1088 luma, 32640 blocks/dispatch, N=3):
ver change R ns/block
v1 first-light 0.230 533
v2 kill ternary + 2-blocks-per-sg 0.474 258
v3 per-pass scope oN 0.481 254 (noise)
v4 WG 64 -> 256 invocations 0.947 129
v5 packed uint32 coeff reads 0.938 130 (noise, reverted)
v4 final N=3 0.918 +/- 0.033
Bit-exactness 100.0000% across all iterations (10000-block sample
on 128x128, 32640-block sample on 1080p) against both the C
reference (tests/vp9_idct8_ref.c) and the vendored FFmpeg NEON
ff_vp9_idct_idct_8x8_add_neon.
Key learning over the Phase 5 review's prediction model: the
chained ternary was NOT a spill killer on V3D 7.1 (shaderdb
showed 0:0 spills:fills even in v1). The actual lever was
workgroup-size-driven latency hiding — going from 64 to 256
invocations doubled throughput with the same compiled code
(270 inst, 2 threads, 21 max-temps, 0 spills) because the
v3dv scheduler had 4x more in-flight work to overlap TMU
latency.
Verdict per phase1.md decision rules: YELLOW band (0.5 <= R < 1.0)
by a wide margin, near GREEN boundary. Phase 1 YELLOW rule:
add M4 (concurrent CPU+QPU throughput) before honest-close or
continue. M4 is the next measurement, not more shader tuning —
at R = 0.92 with all 4 A76 cores still 100% free for other work,
the question is whether the system aggregate beats pure 4-core
NEON.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
129 lines
4.6 KiB
CMake
129 lines
4.6 KiB
CMake
# daedalus-fourier — Phase 3 baseline + (later) Phase 6 implementation.
|
|
#
|
|
# Builds:
|
|
# bench_neon_idct — NEON throughput baseline (Phase 3 M3) +
|
|
# bit-exact correctness gate (Phase 1 M1).
|
|
# bench_vulkan_dispatch — Vulkan compute dispatch-overhead baseline (M5).
|
|
#
|
|
# Linkage note: bench_neon_idct statically links the vendored
|
|
# FFmpeg n7.1.3 NEON snapshot (LGPL-2.1+); see
|
|
# external/ffmpeg-snapshot/PROVENANCE.md.
|
|
|
|
cmake_minimum_required(VERSION 3.20)
|
|
project(daedalus-fourier C ASM)
|
|
|
|
set(CMAKE_C_STANDARD 11)
|
|
set(CMAKE_C_STANDARD_REQUIRED ON)
|
|
|
|
if (NOT CMAKE_BUILD_TYPE)
|
|
set(CMAKE_BUILD_TYPE Release)
|
|
endif()
|
|
|
|
if (NOT CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64")
|
|
message(FATAL_ERROR
|
|
"daedalus-fourier targets aarch64 (Pi 5 / BCM2712). "
|
|
"Cross-compile not yet wired.")
|
|
endif()
|
|
|
|
add_compile_options(-Wall -Wextra -Wno-unused-parameter)
|
|
|
|
# ---- Vendored FFmpeg snapshot (LGPL-2.1+) -----------------------------------
|
|
|
|
set(FFSNAP ${CMAKE_SOURCE_DIR}/external/ffmpeg-snapshot)
|
|
|
|
# Assembly preamble (config.h shim + FFmpeg's asm helpers) used by the
|
|
# vendored .S file. -I flags expose:
|
|
# - FFSNAP/ so `#include "config.h"` finds our shim
|
|
# - FFSNAP/libavcodec/aarch64/ so `#include "neon.S"` finds the helper
|
|
# - FFSNAP/ so `#include "libavutil/aarch64/asm.S"`
|
|
# resolves against the vendored copy
|
|
set(FFASM_FLAGS
|
|
-I${FFSNAP}
|
|
-I${FFSNAP}/libavcodec/aarch64
|
|
-I${FFSNAP}
|
|
)
|
|
|
|
set(FFASM_SOURCES
|
|
${FFSNAP}/libavcodec/aarch64/vp9itxfm_neon.S
|
|
)
|
|
|
|
# Tell CMake/gas to preprocess .S sources.
|
|
set_source_files_properties(${FFASM_SOURCES} PROPERTIES
|
|
COMPILE_OPTIONS "${FFASM_FLAGS}"
|
|
LANGUAGE ASM)
|
|
|
|
# ---- NEON baseline microbench ----------------------------------------------
|
|
|
|
add_executable(bench_neon_idct
|
|
tests/bench_neon_idct.c
|
|
tests/vp9_idct8_ref.c
|
|
${FFASM_SOURCES}
|
|
)
|
|
target_compile_options(bench_neon_idct PRIVATE -O3 -march=armv8-a+simd)
|
|
# bench_neon_idct doesn't need vulkan/drm — pure CPU baseline.
|
|
|
|
# ---- Vulkan dispatch-overhead microbench (next chunk) ----------------------
|
|
# Stub: written in a follow-up step. Toggle ON with -DDAEDALUS_BUILD_VULKAN=ON
|
|
# once tests/bench_vulkan_dispatch.c exists.
|
|
|
|
option(DAEDALUS_BUILD_VULKAN "Build Vulkan compute-dispatch microbench" ON)
|
|
|
|
if (DAEDALUS_BUILD_VULKAN)
|
|
find_package(Vulkan REQUIRED)
|
|
|
|
# Compile GLSL compute shaders to SPIR-V via glslangValidator.
|
|
# The binary loads them at runtime from the build dir (cwd-relative).
|
|
find_program(GLSLANG_VALIDATOR
|
|
NAMES glslangValidator glslang
|
|
REQUIRED)
|
|
|
|
set(NOOP_SPV ${CMAKE_BINARY_DIR}/noop.spv)
|
|
add_custom_command(
|
|
OUTPUT ${NOOP_SPV}
|
|
COMMAND ${GLSLANG_VALIDATOR} -V -o ${NOOP_SPV}
|
|
${CMAKE_SOURCE_DIR}/tests/shaders/noop.comp
|
|
DEPENDS ${CMAKE_SOURCE_DIR}/tests/shaders/noop.comp
|
|
COMMENT "glslang: noop.comp -> noop.spv"
|
|
VERBATIM
|
|
)
|
|
|
|
set(IDCT8_SPV ${CMAKE_BINARY_DIR}/v3d_idct8.spv)
|
|
add_custom_command(
|
|
OUTPUT ${IDCT8_SPV}
|
|
COMMAND ${GLSLANG_VALIDATOR} -V --target-env vulkan1.3
|
|
-o ${IDCT8_SPV}
|
|
${CMAKE_SOURCE_DIR}/src/v3d_idct8.comp
|
|
DEPENDS ${CMAKE_SOURCE_DIR}/src/v3d_idct8.comp
|
|
COMMENT "glslang: v3d_idct8.comp -> v3d_idct8.spv"
|
|
VERBATIM
|
|
)
|
|
add_custom_target(daedalus_shaders ALL DEPENDS ${NOOP_SPV} ${IDCT8_SPV})
|
|
|
|
# v3d_runner — reusable Vulkan plumbing.
|
|
add_library(v3d_runner STATIC src/v3d_runner.c)
|
|
target_include_directories(v3d_runner PUBLIC src)
|
|
target_link_libraries(v3d_runner PUBLIC Vulkan::Vulkan)
|
|
target_compile_options(v3d_runner PRIVATE -O2)
|
|
|
|
add_executable(bench_vulkan_dispatch tests/bench_vulkan_dispatch.c)
|
|
add_dependencies(bench_vulkan_dispatch daedalus_shaders)
|
|
target_link_libraries(bench_vulkan_dispatch PRIVATE Vulkan::Vulkan)
|
|
target_compile_options(bench_vulkan_dispatch PRIVATE -O2)
|
|
|
|
add_executable(bench_v3d_idct
|
|
tests/bench_v3d_idct.c
|
|
tests/vp9_idct8_ref.c
|
|
)
|
|
add_dependencies(bench_v3d_idct daedalus_shaders)
|
|
target_link_libraries(bench_v3d_idct PRIVATE v3d_runner Vulkan::Vulkan)
|
|
target_compile_options(bench_v3d_idct PRIVATE -O2)
|
|
endif()
|
|
|
|
# ---- Summary ----------------------------------------------------------------
|
|
|
|
message(STATUS "daedalus-fourier build configured for ${CMAKE_SYSTEM_PROCESSOR}")
|
|
message(STATUS " FFmpeg snapshot: ${FFSNAP}")
|
|
message(STATUS " Build type: ${CMAKE_BUILD_TYPE}")
|
|
message(STATUS " Targets: bench_neon_idct"
|
|
"$<$<BOOL:${DAEDALUS_BUILD_VULKAN}>:; bench_vulkan_dispatch>")
|