Doc: Adopt CSV CIE 1931 blackbody reference

Replace the in-tree Wyman analytic CIE fit with a repository CIE 1931
2-degree LUT derived from the 360-830 nm, 1 nm-linear CSV reference. The
GRBBLUT3 table stores 1024 uniform log(T) XYZ nodes over
[670.146556, 101408.88] K with four-point cubic Lagrange interpolation, and
the same CSV reference supplies a three-term Rayleigh-Jeans form above the
table and an inverse-temperature/log-XYZ crossover with channel-specific
endpoint forms below it. The loader validates the header and payload
checksum and rejects legacy formats and runtime generation.

Initialize the backend in the renderer, frame test, and PSF capture tool so
the new default is active wherever colors are produced.

Makefile

@@ -29,7 +29,7 @@ else
 IMAGE_EXT := ppm
 endif
 
-COMMON_SOURCES := $(filter-out src/main.c src/spacetime_minkowski.c src/spacetime_schwarzschild.c,$(wildcard src/*.c))
+COMMON_SOURCES := $(filter-out src/main.c src/dummy_psf.c src/spacetime_minkowski.c src/spacetime_schwarzschild.c,$(wildcard src/*.c))
 PROVIDER_SOURCE := src/spacetime_$(SPACETIME).c
 BUILD_DIR := build/$(BUILD_TYPE)
 TARGET_BASENAME := $(SPACETIME)_sky
@@ -56,8 +56,11 @@ RENDER_LINKER := $(CC)
 else ifeq ($(PSF_BACKEND),hip)
 RENDER_LINKER := $(HIPCC)
 BUILD_CPPFLAGS += -DPSF_BACKEND_HIP
+else ifeq ($(PSF_BACKEND),dummy)
+RENDER_LINKER := $(CC)
+BUILD_CPPFLAGS += -DPSF_BACKEND_DUMMY
 else
-$(error Unknown PSF_BACKEND '$(PSF_BACKEND)'; choose cpu or hip)
+$(error Unknown PSF_BACKEND '$(PSF_BACKEND)'; choose cpu, hip, or dummy)
 endif
 
 ifeq ($(SPACETIME),minkowski)
@@ -80,10 +83,15 @@ endif
 
 ifeq ($(PSF_BACKEND),hip)
 TARGET := $(BUILD_DIR)/$(TARGET_BASENAME)_hip
+else ifeq ($(PSF_BACKEND),dummy)
+TARGET := $(BUILD_DIR)/$(TARGET_BASENAME)_dummy
 else
 TARGET := $(BUILD_DIR)/$(TARGET_BASENAME)
 endif
 RENDER_SOURCES := $(COMMON_SOURCES) $(PROVIDER_SOURCE) src/main.c
+ifeq ($(PSF_BACKEND),dummy)
+RENDER_SOURCES += src/dummy_psf.c
+endif
 RENDER_OBJECTS := $(patsubst %.c,$(OBJECT_DIR)/$(HDR_BUILD_TAG)/%.o,$(RENDER_SOURCES))
 ifeq ($(PSF_BACKEND),hip)
 HIP_PSF_OBJECT := $(OBJECT_DIR)/$(HDR_BUILD_TAG)/src/hip_psf.o

assets/CIE_xyz_1931_2deg.csv

@@ -0,0 +1,471 @@
+360,0.000129900000,0.0000039170000,0.000606100000
+361,0.000145847000,0.0000043935810,0.000680879200
+362,0.000163802100,0.0000049296040,0.000765145600
+363,0.000184003700,0.0000055321360,0.000860012400
+364,0.000206690200,0.0000062082450,0.000966592800
+365,0.000232100000,0.0000069650000,0.001086000000
+366,0.000260728000,0.0000078132190,0.001220586000
+367,0.000293075000,0.0000087673360,0.001372729000
+368,0.000329388000,0.0000098398440,0.001543579000
+369,0.000369914000,0.0000110432300,0.001734286000
+370,0.000414900000,0.0000123900000,0.001946000000
+371,0.000464158700,0.0000138864100,0.002177777000
+372,0.000518986000,0.0000155572800,0.002435809000
+373,0.000581854000,0.0000174429600,0.002731953000
+374,0.000655234700,0.0000195837500,0.003078064000
+375,0.000741600000,0.0000220200000,0.003486000000
+376,0.000845029600,0.0000248396500,0.003975227000
+377,0.000964526800,0.0000280412600,0.004540880000
+378,0.001094949000,0.0000315310400,0.005158320000
+379,0.001231154000,0.0000352152100,0.005802907000
+380,0.001368000000,0.0000390000000,0.006450001000
+381,0.001502050000,0.0000428264000,0.007083216000
+382,0.001642328000,0.0000469146000,0.007745488000
+383,0.001802382000,0.0000515896000,0.008501152000
+384,0.001995757000,0.0000571764000,0.009414544000
+385,0.002236000000,0.0000640000000,0.010549990000
+386,0.002535385000,0.0000723442100,0.011965800000
+387,0.002892603000,0.0000822122400,0.013655870000
+388,0.003300829000,0.0000935081600,0.015588050000
+389,0.003753236000,0.0001061361000,0.017730150000
+390,0.004243000000,0.0001200000000,0.020050010000
+391,0.004762389000,0.0001349840000,0.022511360000
+392,0.005330048000,0.0001514920000,0.025202880000
+393,0.005978712000,0.0001702080000,0.028279720000
+394,0.006741117000,0.0001918160000,0.031897040000
+395,0.007650000000,0.0002170000000,0.036210000000
+396,0.008751373000,0.0002469067000,0.041437710000
+397,0.010028880000,0.0002812400000,0.047503720000
+398,0.011421700000,0.0003185200000,0.054119880000
+399,0.012869010000,0.0003572667000,0.060998030000
+400,0.014310000000,0.0003960000000,0.067850010000
+401,0.015704430000,0.0004337147000,0.074486320000
+402,0.017147440000,0.0004730240000,0.081361560000
+403,0.018781220000,0.0005178760000,0.089153640000
+404,0.020748010000,0.0005722187000,0.098540480000
+405,0.023190000000,0.0006400000000,0.110200000000
+406,0.026207360000,0.0007245600000,0.124613300000
+407,0.029782480000,0.0008255000000,0.141701700000
+408,0.033880920000,0.0009411600000,0.161303500000
+409,0.038468240000,0.0010698800000,0.183256800000
+410,0.043510000000,0.0012100000000,0.207400000000
+411,0.048995600000,0.0013620910000,0.233692100000
+412,0.055022600000,0.0015307520000,0.262611400000
+413,0.061718800000,0.0017203680000,0.294774600000
+414,0.069212000000,0.0019353230000,0.330798500000
+415,0.077630000000,0.0021800000000,0.371300000000
+416,0.086958110000,0.0024548000000,0.416209100000
+417,0.097176720000,0.0027640000000,0.465464200000
+418,0.108406300000,0.0031178000000,0.519694800000
+419,0.120767200000,0.0035264000000,0.579530300000
+420,0.134380000000,0.0040000000000,0.645600000000
+421,0.149358200000,0.0045462400000,0.718483800000
+422,0.165395700000,0.0051593200000,0.796713300000
+423,0.181983100000,0.0058292800000,0.877845900000
+424,0.198611000000,0.0065461600000,0.959439000000
+425,0.214770000000,0.0073000000000,1.039050100000
+426,0.230186800000,0.0080865070000,1.115367300000
+427,0.244879700000,0.0089087200000,1.188497100000
+428,0.258777300000,0.0097676800000,1.258123300000
+429,0.271807900000,0.0106644300000,1.323929600000
+430,0.283900000000,0.0116000000000,1.385600000000
+431,0.294943800000,0.0125731700000,1.442635200000
+432,0.304896500000,0.0135827200000,1.494803500000
+433,0.313787300000,0.0146296800000,1.542190300000
+434,0.321645400000,0.0157150900000,1.584880700000
+435,0.328500000000,0.0168400000000,1.622960000000
+436,0.334351300000,0.0180073600000,1.656404800000
+437,0.339210100000,0.0192144800000,1.685295900000
+438,0.343121300000,0.0204539200000,1.709874500000
+439,0.346129600000,0.0217182400000,1.730382100000
+440,0.348280000000,0.0230000000000,1.747060000000
+441,0.349599900000,0.0242946100000,1.760044600000
+442,0.350147400000,0.0256102400000,1.769623300000
+443,0.350013000000,0.0269585700000,1.776263700000
+444,0.349287000000,0.0283512500000,1.780433400000
+445,0.348060000000,0.0298000000000,1.782600000000
+446,0.346373300000,0.0313108300000,1.782968200000
+447,0.344262400000,0.0328836800000,1.781699800000
+448,0.341808800000,0.0345211200000,1.779198200000
+449,0.339094100000,0.0362257100000,1.775867100000
+450,0.336200000000,0.0380000000000,1.772110000000
+451,0.333197700000,0.0398466700000,1.768258900000
+452,0.330041100000,0.0417680000000,1.764039000000
+453,0.326635700000,0.0437660000000,1.758943800000
+454,0.322886800000,0.0458426700000,1.752466300000
+455,0.318700000000,0.0480000000000,1.744100000000
+456,0.314025100000,0.0502436800000,1.733559500000
+457,0.308884000000,0.0525730400000,1.720858100000
+458,0.303290400000,0.0549805600000,1.705936900000
+459,0.297257900000,0.0574587200000,1.688737200000
+460,0.290800000000,0.0600000000000,1.669200000000
+461,0.283970100000,0.0626019700000,1.647528700000
+462,0.276721400000,0.0652775200000,1.623412700000
+463,0.268917800000,0.0680420800000,1.596022300000
+464,0.260422700000,0.0709110900000,1.564528000000
+465,0.251100000000,0.0739000000000,1.528100000000
+466,0.240847500000,0.0770160000000,1.486111400000
+467,0.229851200000,0.0802664000000,1.439521500000
+468,0.218407200000,0.0836668000000,1.389879900000
+469,0.206811500000,0.0872328000000,1.338736200000
+470,0.195360000000,0.0909800000000,1.287640000000
+471,0.184213600000,0.0949175500000,1.237422300000
+472,0.173327300000,0.0990458400000,1.187824300000
+473,0.162688100000,0.1033674000000,1.138761100000
+474,0.152283300000,0.1078846000000,1.090148000000
+475,0.142100000000,0.1126000000000,1.041900000000
+476,0.132178600000,0.1175320000000,0.994197600000
+477,0.122569600000,0.1226744000000,0.947347300000
+478,0.113275200000,0.1279928000000,0.901453100000
+479,0.104297900000,0.1334528000000,0.856619300000
+480,0.095640000000,0.1390200000000,0.812950100000
+481,0.087299550000,0.1446764000000,0.770517300000
+482,0.079308040000,0.1504693000000,0.729444800000
+483,0.071717760000,0.1564619000000,0.689913600000
+484,0.064580990000,0.1627177000000,0.652104900000
+485,0.057950010000,0.1693000000000,0.616200000000
+486,0.051862110000,0.1762431000000,0.582328600000
+487,0.046281520000,0.1835581000000,0.550416200000
+488,0.041150880000,0.1912735000000,0.520337600000
+489,0.036412830000,0.1994180000000,0.491967300000
+490,0.032010000000,0.2080200000000,0.465180000000
+491,0.027917200000,0.2171199000000,0.439924600000
+492,0.024144400000,0.2267345000000,0.416183600000
+493,0.020687000000,0.2368571000000,0.393882200000
+494,0.017540400000,0.2474812000000,0.372945900000
+495,0.014700000000,0.2586000000000,0.353300000000
+496,0.012161790000,0.2701849000000,0.334857800000
+497,0.009919960000,0.2822939000000,0.317552100000
+498,0.007967240000,0.2950505000000,0.301337500000
+499,0.006296346000,0.3085780000000,0.286168600000
+500,0.004900000000,0.3230000000000,0.272000000000
+501,0.003777173000,0.3384021000000,0.258817100000
+502,0.002945320000,0.3546858000000,0.246483800000
+503,0.002424880000,0.3716986000000,0.234771800000
+504,0.002236293000,0.3892875000000,0.223453300000
+505,0.002400000000,0.4073000000000,0.212300000000
+506,0.002925520000,0.4256299000000,0.201169200000
+507,0.003836560000,0.4443096000000,0.190119600000
+508,0.005174840000,0.4633944000000,0.179225400000
+509,0.006982080000,0.4829395000000,0.168560800000
+510,0.009300000000,0.5030000000000,0.158200000000
+511,0.012149490000,0.5235693000000,0.148138300000
+512,0.015535880000,0.5445120000000,0.138375800000
+513,0.019477520000,0.5656900000000,0.128994200000
+514,0.023992770000,0.5869653000000,0.120075100000
+515,0.029100000000,0.6082000000000,0.111700000000
+516,0.034814850000,0.6293456000000,0.103904800000
+517,0.041120160000,0.6503068000000,0.096667480000
+518,0.047985040000,0.6708752000000,0.089982720000
+519,0.055378610000,0.6908424000000,0.083845310000
+520,0.063270000000,0.7100000000000,0.078249990000
+521,0.071635010000,0.7281852000000,0.073208990000
+522,0.080462240000,0.7454636000000,0.068678160000
+523,0.089739960000,0.7619694000000,0.064567840000
+524,0.099456450000,0.7778368000000,0.060788350000
+525,0.109600000000,0.7932000000000,0.057250010000
+526,0.120167400000,0.8081104000000,0.053904350000
+527,0.131114500000,0.8224962000000,0.050746640000
+528,0.142367900000,0.8363068000000,0.047752760000
+529,0.153854200000,0.8494916000000,0.044898590000
+530,0.165500000000,0.8620000000000,0.042160000000
+531,0.177257100000,0.8738108000000,0.039507280000
+532,0.189140000000,0.8849624000000,0.036935640000
+533,0.201169400000,0.8954936000000,0.034458360000
+534,0.213365800000,0.9054432000000,0.032088720000
+535,0.225749900000,0.9148501000000,0.029840000000
+536,0.238320900000,0.9237348000000,0.027711810000
+537,0.251066800000,0.9320924000000,0.025694440000
+538,0.263992200000,0.9399226000000,0.023787160000
+539,0.277101700000,0.9472252000000,0.021989250000
+540,0.290400000000,0.9540000000000,0.020300000000
+541,0.303891200000,0.9602561000000,0.018718050000
+542,0.317572600000,0.9660074000000,0.017240360000
+543,0.331438400000,0.9712606000000,0.015863640000
+544,0.345482800000,0.9760225000000,0.014584610000
+545,0.359700000000,0.9803000000000,0.013400000000
+546,0.374083900000,0.9840924000000,0.012307230000
+547,0.388639600000,0.9874182000000,0.011301880000
+548,0.403378400000,0.9903128000000,0.010377920000
+549,0.418311500000,0.9928116000000,0.009529306000
+550,0.433449900000,0.9949501000000,0.008749999000
+551,0.448795300000,0.9967108000000,0.008035200000
+552,0.464336000000,0.9980983000000,0.007381600000
+553,0.480064000000,0.9991120000000,0.006785400000
+554,0.495971300000,0.9997482000000,0.006242800000
+555,0.512050100000,1.0000000000000,0.005749999000
+556,0.528295900000,0.9998567000000,0.005303600000
+557,0.544691600000,0.9993046000000,0.004899800000
+558,0.561209400000,0.9983255000000,0.004534200000
+559,0.577821500000,0.9968987000000,0.004202400000
+560,0.594500000000,0.9950000000000,0.003900000000
+561,0.611220900000,0.9926005000000,0.003623200000
+562,0.627975800000,0.9897426000000,0.003370600000
+563,0.644760200000,0.9864444000000,0.003141400000
+564,0.661569700000,0.9827241000000,0.002934800000
+565,0.678400000000,0.9786000000000,0.002749999000
+566,0.695239200000,0.9740837000000,0.002585200000
+567,0.712058600000,0.9691712000000,0.002438600000
+568,0.728828400000,0.9638568000000,0.002309400000
+569,0.745518800000,0.9581349000000,0.002196800000
+570,0.762100000000,0.9520000000000,0.002100000000
+571,0.778543200000,0.9454504000000,0.002017733000
+572,0.794825600000,0.9384992000000,0.001948200000
+573,0.810926400000,0.9311628000000,0.001889800000
+574,0.826824800000,0.9234576000000,0.001840933000
+575,0.842500000000,0.9154000000000,0.001800000000
+576,0.857932500000,0.9070064000000,0.001766267000
+577,0.873081600000,0.8982772000000,0.001737800000
+578,0.887894400000,0.8892048000000,0.001711200000
+579,0.902318100000,0.8797816000000,0.001683067000
+580,0.916300000000,0.8700000000000,0.001650001000
+581,0.929799500000,0.8598613000000,0.001610133000
+582,0.942798400000,0.8493920000000,0.001564400000
+583,0.955277600000,0.8386220000000,0.001513600000
+584,0.967217900000,0.8275813000000,0.001458533000
+585,0.978600000000,0.8163000000000,0.001400000000
+586,0.989385600000,0.8047947000000,0.001336667000
+587,0.999548800000,0.7930820000000,0.001270000000
+588,1.009089200000,0.7811920000000,0.001205000000
+589,1.018006400000,0.7691547000000,0.001146667000
+590,1.026300000000,0.7570000000000,0.001100000000
+591,1.033982700000,0.7447541000000,0.001068800000
+592,1.040986000000,0.7324224000000,0.001049400000
+593,1.047188000000,0.7200036000000,0.001035600000
+594,1.052466700000,0.7074965000000,0.001021200000
+595,1.056700000000,0.6949000000000,0.001000000000
+596,1.059794400000,0.6822192000000,0.000968640000
+597,1.061799200000,0.6694716000000,0.000929920000
+598,1.062806800000,0.6566744000000,0.000886880000
+599,1.062909600000,0.6438448000000,0.000842560000
+600,1.062200000000,0.6310000000000,0.000800000000
+601,1.060735200000,0.6181555000000,0.000760960000
+602,1.058443600000,0.6053144000000,0.000723680000
+603,1.055224400000,0.5924756000000,0.000685920000
+604,1.050976800000,0.5796379000000,0.000645440000
+605,1.045600000000,0.5668000000000,0.000600000000
+606,1.039036900000,0.5539611000000,0.000547866700
+607,1.031360800000,0.5411372000000,0.000491600000
+608,1.022666200000,0.5283528000000,0.000435400000
+609,1.013047700000,0.5156323000000,0.000383466700
+610,1.002600000000,0.5030000000000,0.000340000000
+611,0.991367500000,0.4904688000000,0.000307253300
+612,0.979331400000,0.4780304000000,0.000283160000
+613,0.966491600000,0.4656776000000,0.000265440000
+614,0.952847900000,0.4534032000000,0.000251813300
+615,0.938400000000,0.4412000000000,0.000240000000
+616,0.923194000000,0.4290800000000,0.000229546700
+617,0.907244000000,0.4170360000000,0.000220640000
+618,0.890502000000,0.4050320000000,0.000211960000
+619,0.872920000000,0.3930320000000,0.000202186700
+620,0.854449900000,0.3810000000000,0.000190000000
+621,0.835084000000,0.3689184000000,0.000174213300
+622,0.814946000000,0.3568272000000,0.000155640000
+623,0.794186000000,0.3447768000000,0.000135960000
+624,0.772954000000,0.3328176000000,0.000116853300
+625,0.751400000000,0.3210000000000,0.000100000000
+626,0.729583600000,0.3093381000000,0.000086133330
+627,0.707588800000,0.2978504000000,0.000074600000
+628,0.685602200000,0.2865936000000,0.000065000000
+629,0.663810400000,0.2756245000000,0.000056933330
+630,0.642400000000,0.2650000000000,0.000049999990
+631,0.621514900000,0.2547632000000,0.000044160000
+632,0.601113800000,0.2448896000000,0.000039480000
+633,0.581105200000,0.2353344000000,0.000035720000
+634,0.561397700000,0.2260528000000,0.000032640000
+635,0.541900000000,0.2170000000000,0.000030000000
+636,0.522599500000,0.2081616000000,0.000027653330
+637,0.503546400000,0.1995488000000,0.000025560000
+638,0.484743600000,0.1911552000000,0.000023640000
+639,0.466193900000,0.1829744000000,0.000021813330
+640,0.447900000000,0.1750000000000,0.000020000000
+641,0.429861300000,0.1672235000000,0.000018133330
+642,0.412098000000,0.1596464000000,0.000016200000
+643,0.394644000000,0.1522776000000,0.000014200000
+644,0.377533300000,0.1451259000000,0.000012133330
+645,0.360800000000,0.1382000000000,0.000010000000
+646,0.344456300000,0.1315003000000,0.000007733333
+647,0.328516800000,0.1250248000000,0.000005400000
+648,0.313019200000,0.1187792000000,0.000003200000
+649,0.298001100000,0.1127691000000,0.000001333333
+650,0.283500000000,0.1070000000000,0.000000000000
+651,0.269544800000,0.1014762000000,0.000000000000
+652,0.256118400000,0.0961886400000,0.000000000000
+653,0.243189600000,0.0911229600000,0.000000000000
+654,0.230727200000,0.0862648500000,0.000000000000
+655,0.218700000000,0.0816000000000,0.000000000000
+656,0.207097100000,0.0771206400000,0.000000000000
+657,0.195923200000,0.0728255200000,0.000000000000
+658,0.185170800000,0.0687100800000,0.000000000000
+659,0.174832300000,0.0647697600000,0.000000000000
+660,0.164900000000,0.0610000000000,0.000000000000
+661,0.155366700000,0.0573962100000,0.000000000000
+662,0.146230000000,0.0539550400000,0.000000000000
+663,0.137490000000,0.0506737600000,0.000000000000
+664,0.129146700000,0.0475496500000,0.000000000000
+665,0.121200000000,0.0445800000000,0.000000000000
+666,0.113639700000,0.0417587200000,0.000000000000
+667,0.106465000000,0.0390849600000,0.000000000000
+668,0.099690440000,0.0365638400000,0.000000000000
+669,0.093330610000,0.0342004800000,0.000000000000
+670,0.087400000000,0.0320000000000,0.000000000000
+671,0.081900960000,0.0299626100000,0.000000000000
+672,0.076804280000,0.0280766400000,0.000000000000
+673,0.072077120000,0.0263293600000,0.000000000000
+674,0.067686640000,0.0247080500000,0.000000000000
+675,0.063600000000,0.0232000000000,0.000000000000
+676,0.059806850000,0.0218007700000,0.000000000000
+677,0.056282160000,0.0205011200000,0.000000000000
+678,0.052971040000,0.0192810800000,0.000000000000
+679,0.049818610000,0.0181206900000,0.000000000000
+680,0.046770000000,0.0170000000000,0.000000000000
+681,0.043784050000,0.0159037900000,0.000000000000
+682,0.040875360000,0.0148371800000,0.000000000000
+683,0.038072640000,0.0138106800000,0.000000000000
+684,0.035404610000,0.0128347800000,0.000000000000
+685,0.032900000000,0.0119200000000,0.000000000000
+686,0.030564190000,0.0110683100000,0.000000000000
+687,0.028380560000,0.0102733900000,0.000000000000
+688,0.026344840000,0.0095333110000,0.000000000000
+689,0.024452750000,0.0088461570000,0.000000000000
+690,0.022700000000,0.0082100000000,0.000000000000
+691,0.021084290000,0.0076237810000,0.000000000000
+692,0.019599880000,0.0070854240000,0.000000000000
+693,0.018237320000,0.0065914760000,0.000000000000
+694,0.016987170000,0.0061384850000,0.000000000000
+695,0.015840000000,0.0057230000000,0.000000000000
+696,0.014790640000,0.0053430590000,0.000000000000
+697,0.013831320000,0.0049957960000,0.000000000000
+698,0.012948680000,0.0046764040000,0.000000000000
+699,0.012129200000,0.0043800750000,0.000000000000
+700,0.011359160000,0.0041020000000,0.000000000000
+701,0.010629350000,0.0038384530000,0.000000000000
+702,0.009938846000,0.0035890990000,0.000000000000
+703,0.009288422000,0.0033542190000,0.000000000000
+704,0.008678854000,0.0031340930000,0.000000000000
+705,0.008110916000,0.0029290000000,0.000000000000
+706,0.007582388000,0.0027381390000,0.000000000000
+707,0.007088746000,0.0025598760000,0.000000000000
+708,0.006627313000,0.0023932440000,0.000000000000
+709,0.006195408000,0.0022372750000,0.000000000000
+710,0.005790346000,0.0020910000000,0.000000000000
+711,0.005409826000,0.0019535870000,0.000000000000
+712,0.005052583000,0.0018245800000,0.000000000000
+713,0.004717512000,0.0017035800000,0.000000000000
+714,0.004403507000,0.0015901870000,0.000000000000
+715,0.004109457000,0.0014840000000,0.000000000000
+716,0.003833913000,0.0013844960000,0.000000000000
+717,0.003575748000,0.0012912680000,0.000000000000
+718,0.003334342000,0.0012040920000,0.000000000000
+719,0.003109075000,0.0011227440000,0.000000000000
+720,0.002899327000,0.0010470000000,0.000000000000
+721,0.002704348000,0.0009765896000,0.000000000000
+722,0.002523020000,0.0009111088000,0.000000000000
+723,0.002354168000,0.0008501332000,0.000000000000
+724,0.002196616000,0.0007932384000,0.000000000000
+725,0.002049190000,0.0007400000000,0.000000000000
+726,0.001910960000,0.0006900827000,0.000000000000
+727,0.001781438000,0.0006433100000,0.000000000000
+728,0.001660110000,0.0005994960000,0.000000000000
+729,0.001546459000,0.0005584547000,0.000000000000
+730,0.001439971000,0.0005200000000,0.000000000000
+731,0.001340042000,0.0004839136000,0.000000000000
+732,0.001246275000,0.0004500528000,0.000000000000
+733,0.001158471000,0.0004183452000,0.000000000000
+734,0.001076430000,0.0003887184000,0.000000000000
+735,0.000999949300,0.0003611000000,0.000000000000
+736,0.000928735800,0.0003353835000,0.000000000000
+737,0.000862433200,0.0003114404000,0.000000000000
+738,0.000800750300,0.0002891656000,0.000000000000
+739,0.000743396000,0.0002684539000,0.000000000000
+740,0.000690078600,0.0002492000000,0.000000000000
+741,0.000640515600,0.0002313019000,0.000000000000
+742,0.000594502100,0.0002146856000,0.000000000000
+743,0.000551864600,0.0001992884000,0.000000000000
+744,0.000512429000,0.0001850475000,0.000000000000
+745,0.000476021300,0.0001719000000,0.000000000000
+746,0.000442453600,0.0001597781000,0.000000000000
+747,0.000411511700,0.0001486044000,0.000000000000
+748,0.000382981400,0.0001383016000,0.000000000000
+749,0.000356649100,0.0001287925000,0.000000000000
+750,0.000332301100,0.0001200000000,0.000000000000
+751,0.000309758600,0.0001118595000,0.000000000000
+752,0.000288887100,0.0001043224000,0.000000000000
+753,0.000269539400,0.0000973356000,0.000000000000
+754,0.000251568200,0.0000908458700,0.000000000000
+755,0.000234826100,0.0000848000000,0.000000000000
+756,0.000219171000,0.0000791466700,0.000000000000
+757,0.000204525800,0.0000738580000,0.000000000000
+758,0.000190840500,0.0000689160000,0.000000000000
+759,0.000178065400,0.0000643026700,0.000000000000
+760,0.000166150500,0.0000600000000,0.000000000000
+761,0.000155023600,0.0000559818700,0.000000000000
+762,0.000144621900,0.0000522256000,0.000000000000
+763,0.000134909800,0.0000487184000,0.000000000000
+764,0.000125852000,0.0000454474700,0.000000000000
+765,0.000117413000,0.0000424000000,0.000000000000
+766,0.000109551500,0.0000395610400,0.000000000000
+767,0.000102224500,0.0000369151200,0.000000000000
+768,0.000095394450,0.0000344486800,0.000000000000
+769,0.000089023900,0.0000321481600,0.000000000000
+770,0.000083075270,0.0000300000000,0.000000000000
+771,0.000077512690,0.0000279912500,0.000000000000
+772,0.000072313040,0.0000261135600,0.000000000000
+773,0.000067457780,0.0000243602400,0.000000000000
+774,0.000062928440,0.0000227246100,0.000000000000
+775,0.000058706520,0.0000212000000,0.000000000000
+776,0.000054770280,0.0000197785500,0.000000000000
+777,0.000051099180,0.0000184528500,0.000000000000
+778,0.000047676540,0.0000172168700,0.000000000000
+779,0.000044485670,0.0000160645900,0.000000000000
+780,0.000041509940,0.0000149900000,0.000000000000
+781,0.000038733240,0.0000139872800,0.000000000000
+782,0.000036142030,0.0000130515500,0.000000000000
+783,0.000033723520,0.0000121781800,0.000000000000
+784,0.000031464870,0.0000113625400,0.000000000000
+785,0.000029353260,0.0000106000000,0.000000000000
+786,0.000027375730,0.0000098858770,0.000000000000
+787,0.000025524330,0.0000092173040,0.000000000000
+788,0.000023793760,0.0000085923620,0.000000000000
+789,0.000022178700,0.0000080091330,0.000000000000
+790,0.000020673830,0.0000074657000,0.000000000000
+791,0.000019272260,0.0000069595670,0.000000000000
+792,0.000017966400,0.0000064879950,0.000000000000
+793,0.000016749910,0.0000060486990,0.000000000000
+794,0.000015616480,0.0000056393960,0.000000000000
+795,0.000014559770,0.0000052578000,0.000000000000
+796,0.000013573870,0.0000049017710,0.000000000000
+797,0.000012654360,0.0000045697200,0.000000000000
+798,0.000011797230,0.0000042601940,0.000000000000
+799,0.000010998440,0.0000039717390,0.000000000000
+800,0.000010253980,0.0000037029000,0.000000000000
+801,0.000009559646,0.0000034521630,0.000000000000
+802,0.000008912044,0.0000032183020,0.000000000000
+803,0.000008308358,0.0000030003000,0.000000000000
+804,0.000007745769,0.0000027971390,0.000000000000
+805,0.000007221456,0.0000026078000,0.000000000000
+806,0.000006732475,0.0000024312200,0.000000000000
+807,0.000006276423,0.0000022665310,0.000000000000
+808,0.000005851304,0.0000021130130,0.000000000000
+809,0.000005455118,0.0000019699430,0.000000000000
+810,0.000005085868,0.0000018366000,0.000000000000
+811,0.000004741466,0.0000017122300,0.000000000000
+812,0.000004420236,0.0000015962280,0.000000000000
+813,0.000004120783,0.0000014880900,0.000000000000
+814,0.000003841716,0.0000013873140,0.000000000000
+815,0.000003581652,0.0000012934000,0.000000000000
+816,0.000003339127,0.0000012058200,0.000000000000
+817,0.000003112949,0.0000011241430,0.000000000000
+818,0.000002902121,0.0000010480090,0.000000000000
+819,0.000002705645,0.0000009770578,0.000000000000
+820,0.000002522525,0.0000009109300,0.000000000000
+821,0.000002351726,0.0000008492513,0.000000000000
+822,0.000002192415,0.0000007917212,0.000000000000
+823,0.000002043902,0.0000007380904,0.000000000000
+824,0.000001905497,0.0000006881098,0.000000000000
+825,0.000001776509,0.0000006415300,0.000000000000
+826,0.000001656215,0.0000005980895,0.000000000000
+827,0.000001544022,0.0000005575746,0.000000000000
+828,0.000001439440,0.0000005198080,0.000000000000
+829,0.000001341977,0.0000004846123,0.000000000000
+830,0.000001251141,0.0000004518100,0.000000000000

assets/blackbody/README.md

@@ -0,0 +1,40 @@
+# CIE 1931 blackbody LUT
+
+`cie1931_2deg_xyz_1024.grbblut` is the sole repository LUT. It contains 1,024
+XYZ nodes uniformly spaced in `log(T)` over `[670.146556, 101408.88] K`.
+Wolfram integrates the Planck spectrum against the piecewise-linear 1 nm data
+in `assets/CIE_xyz_1931_2deg.csv` at 40-digit working precision and rounds each
+stored value once to little-endian IEEE binary64.
+
+The 160-byte `GRBBLUT3` header records the CIE 1931 2-degree, 360--830 nm,
+1 nm-linear reference definition, node count, temperature limits, physical
+constants, generator description, and payload FNV-1a checksum. The C loader
+validates the header, exact payload size, and checksum, and rejects trailing
+bytes. Runtime table generation and the former `GRBBLUT1`/`GRBBLUT2` formats are
+not supported. The evaluator uses four-point cubic Lagrange interpolation in
+`log(T)`.
+
+- file size: 24,736 bytes
+- payload FNV-1a: `0e39867b0e70a809`
+- file SHA-256:
+  `850446663212fe488c88fca3ac2f54d6f8119a04801dda84fc1a62acdec37d63`
+- all 1,023 interval midpoints versus independent Gauss-16 integration:
+  max XYZ-channel relative `5.4633e-5`, no negative XYZ
+
+The LUT interpolation acceptance limit is `1e-4`. A separate `5e-4` overall
+relative error budget accounts for the estimated error of linearly
+interpolating the 1 nm CIE samples; the LUT's much smaller internal error must
+not be mistaken for physical-reference accuracy.
+
+Outside the stored range, the same CSV reference supplies both models. Above
+101408.88 K the evaluator uses three integrated Rayleigh--Jeans terms
+`A T+B+C/T` (maximum validated XYZ-channel relative error `1.6635e-5`
+through `10*Tmax`). Below 670.146556 K it uses a six-segment, degree-eight
+inverse-temperature/log-XYZ Chebyshev--Lobatto crossover down to 4.097321 K
+and channel-specific endpoint forms below that. Their maximum sampled XYZ
+relative errors are `8.16e-5` and `6.69e-5`.
+
+The Wolfram generation/validation scripts that produced this table, together
+with the bounded performance and accuracy evidence, are preserved on the
+experiment branch `codex/blackbody-cost-experiment` (tag
+`blackbody-cost-experiment-2026-09-12`) and are not part of the mainline tree.

benchmarks/dummy_psf_chunks_2026-09-11/.gitignore

@@ -0,0 +1 @@
+__pycache__/

benchmarks/dummy_psf_chunks_2026-09-11/README.md

@@ -0,0 +1,84 @@
+# Production chunk geometry diagnostic (2026-09-11)
+
+## Scope
+
+This run measures the event stream that the production HIP sink would receive
+for the supplied Schwarzschild Galactic-center render. `PSF_BACKEND=dummy`
+keeps the real imported lens map, all-sky prefetch, OpenMP dynamic triangle
+scheduling, inverse mapping, redshift/color/flux work, PSF cache eligibility,
+and 16,384-event per-worker chunk boundaries. It does not initialize HIP,
+allocate a full HDR framebuffer, splat pixels, or write PNG/FITS output.
+
+The run used repository revision `11e703e33dffb1ff168ba11dd922e68b448116b9`
+plus the uncommitted dummy-backend work. The complete stdout/stderr, executable
+and input hashes, process resource use, and protected-output hashes are in
+[`full_catalog.log`](full_catalog.log). An initial harness-only failure before
+the renderer started is retained as
+[`full_catalog_missing_time.log`](full_catalog_missing_time.log).
+
+## Build and run
+
+```sh
+make -j1 PSF_BACKEND=dummy SPACETIME=schwarzschild ENABLE_HDR=1 backend
+python3 benchmarks/dummy_psf_chunks_2026-09-11/run.py
+```
+
+The runner fixes `OMP_NUM_THREADS=16` and `OMP_DYNAMIC=FALSE`, uses a 600-second
+timeout and a process lock, refuses to overwrite its log, and hashes both the
+named PNG and companion FITS before and after the run. Its renderer arguments
+are exactly the user-supplied command except that the `_hip` executable is
+replaced by `_dummy`; the exact expanded command is the first line of the log.
+
+Inputs and protected outputs:
+
+- lens map SHA-256:
+  `905f960aea2d276f9586c79fba545a72cf742311e9e7e18faea293835c68bf28`
+- existing PNG SHA-256 before/after:
+  `3e01feb911e60acce545a6a86870ca05f96250022e874eb8097bec88a8879bfd`
+- existing FITS SHA-256 before/after:
+  `2408e62dcc22f147da284d87af9a2b0fdd0050eda2dabc23b6ec2554b2f1d38e`
+
+Both protected files retained the same size and nanosecond mtime as well as the
+same hash. No HIP backend was run.
+
+## Result
+
+The catalog prefetch loaded 343,363,141 stars for 130,882 lens triangles. The
+producer classified 598,264,924 cache events, with zero direct fallbacks,
+wing-clipped events, or min-Y discards. Sixteen workers produced 36,524 chunks:
+36,508 full chunks and one final partial chunk per worker.
+
+| full-chunk metric | min | p10 | p25 | p50 | p75 | p90 | p99 | max |
+|---|---:|---:|---:|---:|---:|---:|---:|---:|
+| event-producing triangles | 1 | 1 | 1 | 2 | 3 | 10 | 31 | 59 |
+| occupied 32px center tiles | 1 | 1 | 1 | 1 | 2 | 5 | 30 | 45 |
+| event bbox diagonal (px) | 0.199 | 0.742 | 3.250 | 12.651 | 57.018 | 429.306 | 3759.261 | 3834.557 |
+| triangle-centroid bbox diagonal (px) | 0 | 0 | 0 | 9.618 | 49.477 | 424.149 | 3744.061 | 3824.075 |
+| maximum successive centroid jump (px) | 0 | 0 | 0 | 8.724 | 25.157 | 144.000 | 3744.034 | 3792.034 |
+
+All 36,508 full chunks satisfy the replay prototype's current selector:
+`events >= 8192 && events / occupied_32px_tiles >= 32`.
+
+Catalog prefetch took 55.469 s. Producer classification and chunk accounting
+took 232.340 s; total process wall time was 288.872 s. These are diagnostic CPU
+costs and do not estimate HIP tile-kernel speed.
+
+## Interpretation and limits
+
+The measured production event order strongly supports the hypothesis that a
+full worker chunk is usually filled by very few image-plane triangles: the
+median is two and 75% contain at most three. Center-tile occupancy is even more
+concentrated. This is favorable for a tile16 path because the same output pixels
+receive many contributions that can be reduced before the final double HDR
+write.
+
+The rare tail matters: p99 chunks can contain a few clusters separated by
+nearly the full 4K image width. “Few triangles” therefore does not imply one
+small contiguous bounding rectangle. A production implementation should use a
+sparse occupied-tile list, retain the atomic fallback, and measure the complete
+streaming sink including binning and synchronization.
+
+This is one Schwarzschild Galactic-center frame with one 16-worker dynamic
+schedule. Worker-to-triangle assignment is nondeterministic, so exact quantiles
+may vary between runs. The result establishes event geometry for this input; it
+does not by itself establish single-BH/BBH universality or a HIP speedup.

benchmarks/dummy_psf_chunks_2026-09-11/full_catalog.log

@@ -0,0 +1,235 @@
+COMMAND OMP_NUM_THREADS=16 OMP_DYNAMIC=FALSE /home/wyj/Code/C/GR_4d_raytracing/build/Release/schwarzschild_sky_dummy --verbose --psf-relative-tail 1e-8 --psf-min-y 0 --max-cache-psf-flux 1e8 --all-sky-catalog assets/2mass/processed/all_sky --exposure 1e13 --psf-fwhm-pixels 2.7 --psf-moffat-beta 4.5 --catalog-load-workers 4 --hdr-output --lens-map-input output/lens/schwarzschild_galactic_center_R100_45deg_16_4_4k_.grlens --output output/imgs/2mass_galactic_center_schwarzschild_R100_4k_1e13_beta4.5_16_4-full-gpu.png
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 c71a843a6064cd730b4694c5fe4e61249a995106526da0fdca97bc625c7db470
+LENS_SHA256 905f960aea2d276f9586c79fba545a72cf742311e9e7e18faea293835c68bf28
+PROTECTED_BEFORE output/imgs/2mass_galactic_center_schwarzschild_R100_4k_1e13_beta4.5_16_4-full-gpu.png sha256=3e01feb911e60acce545a6a86870ca05f96250022e874eb8097bec88a8879bfd size=13072572 mtime_ns=1788744728751407011
+PROTECTED_BEFORE output/imgs/2mass_galactic_center_schwarzschild_R100_4k_1e13_beta4.5_16_4-full-gpu_HDR.fits sha256=2408e62dcc22f147da284d87af9a2b0fdd0050eda2dabc23b6ec2554b2f1d38e size=99535680 mtime_ns=1788744726380022561
+Dummy PSF backend: --output and --hdr-output are accepted for command parity but no image files will be written.
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.678 s
+Frame 0: finding and prefetching catalog tiles...
+Catalog prefetch: 64440 requested, 64440 newly loaded (343363141 stars), 0 unavailable in 55.469 s; 4 loader workers
+Frame 0: catalog prefetch finished (64440 candidate tiles).
+Frame 0: splatting 130882 lens triangles...
+Frame 0: splat worker 2/16 started.
+Frame 0: splat worker 1/16 started.
+Frame 0: splat worker 4/16 started.
+Frame 0: splat worker 3/16 started.
+Frame 0: splat worker 5/16 started.
+Frame 0: splat worker 10/16 started.
+Frame 0: splat worker 7/16 started.
+Frame 0: splat worker 8/16 started.
+Frame 0: splat worker 13/16 started.
+Frame 0: splat worker 11/16 started.
+Frame 0: splat worker 6/16 started.
+Frame 0: splat worker 12/16 started.
+Frame 0: splat worker 9/16 started.
+Frame 0: splat worker 14/16 started.
+Frame 0: splat worker 16/16 started.
+Frame 0: splat worker 15/16 started.
+Frame 0: splat worker 1/16 reached 8 local triangles.
+Frame 0: splat worker 6/16 reached 8 local triangles.
+Frame 0: splat worker 10/16 reached 8 local triangles.
+Frame 0: splat worker 14/16 reached 8 local triangles.
+Frame 0: splat worker 12/16 reached 8 local triangles.
+Frame 0: splat worker 2/16 reached 8 local triangles.
+Frame 0: splat worker 5/16 reached 8 local triangles.
+Frame 0: splat worker 3/16 reached 8 local triangles.
+Frame 0: splat worker 7/16 reached 8 local triangles.
+Frame 0: splat worker 15/16 reached 8 local triangles.
+Frame 0: splat worker 8/16 reached 8 local triangles.
+Frame 0: splat worker 4/16 reached 8 local triangles.
+Frame 0: splat worker 11/16 reached 8 local triangles.
+Frame 0: splat worker 13/16 reached 8 local triangles.
+Frame 0: splat worker 9/16 reached 8 local triangles.
+Frame 0: splat worker 16/16 reached 8 local triangles.
+Frame 0: splat worker 12/16 reached 16 local triangles.
+Frame 0: splat worker 4/16 reached 16 local triangles.
+Frame 0: splat worker 14/16 reached 16 local triangles.
+Frame 0: splat worker 6/16 reached 16 local triangles.
+Frame 0: splat worker 1/16 reached 16 local triangles.
+Frame 0: splat worker 8/16 reached 16 local triangles.
+Frame 0: splat worker 3/16 reached 16 local triangles.
+Frame 0: splat worker 7/16 reached 16 local triangles.
+Frame 0: splat worker 2/16 reached 16 local triangles.
+Frame 0: splat worker 15/16 reached 16 local triangles.
+Frame 0: splat worker 5/16 reached 16 local triangles.
+Frame 0: splat worker 11/16 reached 16 local triangles.
+Frame 0: splat worker 13/16 reached 16 local triangles.
+Frame 0: splat worker 9/16 reached 16 local triangles.
+Frame 0: splat worker 16/16 reached 16 local triangles.
+Frame 0: splat worker 12/16 reached 32 local triangles.
+Frame 0: splat worker 4/16 reached 32 local triangles.
+Frame 0: splat worker 7/16 reached 32 local triangles.
+Frame 0: splat worker 6/16 reached 32 local triangles.
+Frame 0: splat worker 8/16 reached 32 local triangles.
+Frame 0: splat worker 2/16 reached 32 local triangles.
+Frame 0: splat worker 3/16 reached 32 local triangles.
+Frame 0: splat worker 11/16 reached 32 local triangles.
+Frame 0: splat worker 9/16 reached 32 local triangles.
+Frame 0: splat worker 1/16 reached 32 local triangles.
+Frame 0: splat worker 15/16 reached 32 local triangles.
+Frame 0: splat worker 16/16 reached 32 local triangles.
+Frame 0: splat worker 14/16 reached 32 local triangles.
+Frame 0: splat worker 5/16 reached 32 local triangles.
+Frame 0: splat worker 13/16 reached 32 local triangles.
+Frame 0: splat worker 10/16 reached 16 local triangles.
+Frame 0: splat worker 10/16 reached 32 local triangles.
+Frame 0: splat worker 4/16 reached 64 local triangles.
+Frame 0: splat worker 6/16 reached 64 local triangles.
+Frame 0: splat worker 8/16 reached 64 local triangles.
+Frame 0: splat worker 2/16 reached 64 local triangles.
+Frame 0: splat worker 3/16 reached 64 local triangles.
+Frame 0: splat worker 12/16 reached 64 local triangles.
+Frame 0: splat worker 11/16 reached 64 local triangles.
+Frame 0: splat worker 15/16 reached 64 local triangles.
+Frame 0: splat worker 16/16 reached 64 local triangles.
+Frame 0: splat worker 14/16 reached 64 local triangles.
+Frame 0: splat worker 5/16 reached 64 local triangles.
+Frame 0: splat worker 9/16 reached 64 local triangles.
+Frame 0: splat worker 1/16 reached 64 local triangles.
+Frame 0: splat worker 7/16 reached 64 local triangles.
+Frame 0: splat worker 13/16 reached 64 local triangles.
+Frame 0: splat worker 10/16 reached 64 local triangles.
+Frame 0: splat worker 4/16 reached 128 local triangles.
+Frame 0: splat worker 12/16 reached 128 local triangles.
+Frame 0: splat worker 11/16 reached 128 local triangles.
+Frame 0: splat worker 6/16 reached 128 local triangles.
+Frame 0: splat worker 15/16 reached 128 local triangles.
+Frame 0: splat worker 2/16 reached 128 local triangles.
+Frame 0: splat worker 5/16 reached 128 local triangles.
+Frame 0: splat worker 8/16 reached 128 local triangles.
+Frame 0: splat worker 3/16 reached 128 local triangles.
+Frame 0: splat worker 16/16 reached 128 local triangles.
+Frame 0: splat worker 14/16 reached 128 local triangles.
+Frame 0: splat worker 7/16 reached 128 local triangles.
+Frame 0: splat worker 1/16 reached 128 local triangles.
+Frame 0: splat worker 13/16 reached 128 local triangles.
+Frame 0: splat worker 9/16 reached 128 local triangles.
+Frame 0: splat worker 10/16 reached 128 local triangles.
+Frame 0: splat worker 13/16 reached 256 local triangles.
+Frame 0: splat worker 15/16 reached 256 local triangles.
+Frame 0: splat worker 16/16 reached 256 local triangles.
+Frame 0: splat worker 8/16 reached 256 local triangles.
+Frame 0: splat worker 11/16 reached 256 local triangles.
+Frame 0: splat worker 6/16 reached 256 local triangles.
+Frame 0: splat worker 5/16 reached 256 local triangles.
+Frame 0: splat worker 3/16 reached 256 local triangles.
+Frame 0: splat worker 1/16 reached 256 local triangles.
+Frame 0: splat worker 2/16 reached 256 local triangles.
+Frame 0: splat worker 7/16 reached 256 local triangles.
+Frame 0: splat worker 4/16 reached 256 local triangles.
+Frame 0: splat worker 12/16 reached 256 local triangles.
+Frame 0: splat worker 10/16 reached 256 local triangles.
+Frame 0: splat worker 14/16 reached 256 local triangles.
+Frame 0: splat worker 9/16 reached 256 local triangles.
+Frame 0: splat worker 8/16 reached 512 local triangles.
+Frame 0: splat worker 12/16 reached 512 local triangles.
+Frame 0: splat worker 16/16 reached 512 local triangles.
+Frame 0: splat worker 5/16 reached 512 local triangles.
+Frame 0: splat worker 13/16 reached 512 local triangles.
+Frame 0: splat worker 3/16 reached 512 local triangles.
+Frame 0: splat worker 7/16 reached 512 local triangles.
+Frame 0: splat worker 11/16 reached 512 local triangles.
+Frame 0: splat worker 1/16 reached 512 local triangles.
+Frame 0: splat worker 15/16 reached 512 local triangles.
+Frame 0: splat worker 2/16 reached 512 local triangles.
+Frame 0: splat worker 10/16 reached 512 local triangles.
+Frame 0: splat worker 4/16 reached 512 local triangles.
+Frame 0: splat worker 6/16 reached 512 local triangles.
+Frame 0: splat worker 14/16 reached 512 local triangles.
+Frame 0: splat worker 9/16 reached 512 local triangles.
+Frame 0: splat worker 8/16 reached 1024 local triangles.
+Frame 0: splat worker 7/16 reached 1024 local triangles.
+Frame 0: splat worker 12/16 reached 1024 local triangles.
+Frame 0: splat worker 3/16 reached 1024 local triangles.
+Frame 0: splat worker 16/16 reached 1024 local triangles.
+Frame 0: splat worker 11/16 reached 1024 local triangles.
+Frame 0: splat worker 5/16 reached 1024 local triangles.
+Frame 0: splat worker 14/16 reached 1024 local triangles.
+Frame 0: splat worker 15/16 reached 1024 local triangles.
+Frame 0: splat worker 4/16 reached 1024 local triangles.
+Frame 0: splat worker 9/16 reached 1024 local triangles.
+Frame 0: splat worker 13/16 reached 1024 local triangles.
+Frame 0: splat worker 6/16 reached 1024 local triangles.
+Frame 0: splat worker 1/16 reached 1024 local triangles.
+Frame 0: splat worker 10/16 reached 1024 local triangles.
+Frame 0: splat worker 2/16 reached 1024 local triangles.
+Frame 0: splat worker 8/16 reached 2048 local triangles.
+Frame 0: splat worker 12/16 reached 2048 local triangles.
+Frame 0: splat worker 16/16 reached 2048 local triangles.
+Frame 0: splat worker 14/16 reached 2048 local triangles.
+Frame 0: splat worker 7/16 reached 2048 local triangles.
+Frame 0: splat worker 3/16 reached 2048 local triangles.
+Frame 0: splat worker 4/16 reached 2048 local triangles.
+Frame 0: splat worker 1/16 reached 2048 local triangles.
+Frame 0: splat worker 11/16 reached 2048 local triangles.
+Frame 0: splat worker 5/16 reached 2048 local triangles.
+Frame 0: splat worker 13/16 reached 2048 local triangles.
+Frame 0: splat worker 15/16 reached 2048 local triangles.
+Frame 0: splat worker 6/16 reached 2048 local triangles.
+Frame 0: splat worker 2/16 reached 2048 local triangles.
+Frame 0: splat worker 9/16 reached 2048 local triangles.
+Frame 0: splat worker 10/16 reached 2048 local triangles.
+Frame 0: splat worker 3/16 reached 4096 local triangles.
+Frame 0: splat worker 7/16 reached 4096 local triangles.
+Frame 0: splat worker 16/16 reached 4096 local triangles.
+Frame 0: splat worker 14/16 reached 4096 local triangles.
+Frame 0: splat worker 9/16 reached 4096 local triangles.
+Frame 0: splat worker 4/16 reached 4096 local triangles.
+Frame 0: splat worker 6/16 reached 4096 local triangles.
+Frame 0: splat worker 8/16 reached 4096 local triangles.
+Frame 0: splat worker 13/16 reached 4096 local triangles.
+Frame 0: splat worker 5/16 reached 4096 local triangles.
+Frame 0: splat worker 10/16 reached 4096 local triangles.
+Frame 0: splat worker 15/16 reached 4096 local triangles.
+Frame 0: splat worker 11/16 reached 4096 local triangles.
+Frame 0: splat worker 12/16 reached 4096 local triangles.
+Frame 0: splat worker 1/16 reached 4096 local triangles.
+Frame 0: splat worker 2/16 reached 4096 local triangles.
+Frame 0: splat worker 3/16 reached 8192 local triangles.
+Frame 0: splat worker 7/16 reached 8192 local triangles.
+Frame 0: splat worker 16/16 reached 8192 local triangles.
+Frame 0: splat worker 13/16 reached 8192 local triangles.
+Frame 0: splat worker 9/16 reached 8192 local triangles.
+Frame 0: splat worker 4/16 reached 8192 local triangles.
+Frame 0: splat worker 8/16 reached 8192 local triangles.
+Frame 0: splat worker 2/16 finished after 7916 local triangles.
+Frame 0: splat worker 8/16 finished after 8212 local triangles.
+Frame 0: splat worker 6/16 finished after 8150 local triangles.
+Frame 0: splat worker 1/16 finished after 7784 local triangles.
+Frame 0: splat worker 4/16 finished after 8227 local triangles.
+Frame 0: splat worker 15/16 finished after 7878 local triangles.
+Frame 0: splat worker 14/16 finished after 8186 local triangles.
+Frame 0: splat worker 11/16 finished after 7962 local triangles.
+Frame 0: splat worker 9/16 finished after 8276 local triangles.
+Frame 0: splat worker 16/16 finished after 8477 local triangles.
+Frame 0: splat worker 3/16 finished after 8967 local triangles.
+Frame 0: splat worker 7/16 finished after 8576 local triangles.
+Frame 0: splat worker 10/16 finished after 7772 local triangles.
+Frame 0: splat worker 13/16 finished after 8295 local triangles.
+Frame 0: splat worker 12/16 finished after 8013 local triangles.
+Frame 0: splat worker 5/16 finished after 8191 local triangles.
+Dummy PSF diagnostic only: no HDR or image was accumulated/written.
+Dummy PSF chunks: total=36524 full=36508 partial=16 events=598264924 capacity=16384 selector_tile=32px
+Dummy PSF all events: min=263.000 p10=16384.000 p25=16384.000 p50=16384.000 p75=16384.000 p90=16384.000 p99=16384.000 max=16384.000
+Dummy PSF all distinct_triangles: min=1.000 p10=1.000 p25=1.000 p50=2.000 p75=4.000 p90=10.000 p99=31.000 max=59.000
+Dummy PSF all occupied_32px_tiles: min=1.000 p10=1.000 p25=1.000 p50=1.000 p75=2.000 p90=5.000 p99=30.000 max=45.000
+Dummy PSF all event_bbox_diagonal_px: min=0.199 p10=0.743 p25=3.252 p50=12.667 p75=57.123 p90=432.049 p99=3759.261 max=3834.557
+Dummy PSF all triangle_centroid_bbox_diagonal_px: min=0.000 p10=0.000 p25=0.000 p50=9.621 p75=49.817 p90=425.680 p99=3744.061 max=3824.075
+Dummy PSF all maximum_successive_triangle_jump_px: min=0.000 p10=0.000 p25=0.000 p50=8.750 p75=25.157 p90=144.000 p99=3744.034 max=3792.034
+Dummy PSF all adaptive criterion: 36516/36524 chunks tile16-eligible (events>=8192 and events/occupied_32px_tiles>=32)
+Dummy PSF full events: min=16384.000 p10=16384.000 p25=16384.000 p50=16384.000 p75=16384.000 p90=16384.000 p99=16384.000 max=16384.000
+Dummy PSF full distinct_triangles: min=1.000 p10=1.000 p25=1.000 p50=2.000 p75=3.000 p90=10.000 p99=31.000 max=59.000
+Dummy PSF full occupied_32px_tiles: min=1.000 p10=1.000 p25=1.000 p50=1.000 p75=2.000 p90=5.000 p99=30.000 max=45.000
+Dummy PSF full event_bbox_diagonal_px: min=0.199 p10=0.742 p25=3.250 p50=12.651 p75=57.018 p90=429.306 p99=3759.261 max=3834.557
+Dummy PSF full triangle_centroid_bbox_diagonal_px: min=0.000 p10=0.000 p25=0.000 p50=9.618 p75=49.477 p90=424.149 p99=3744.061 max=3824.075
+Dummy PSF full maximum_successive_triangle_jump_px: min=0.000 p10=0.000 p25=0.000 p50=8.724 p75=25.157 p90=144.000 p99=3744.034 max=3792.034
+Dummy PSF full adaptive criterion: 36508/36508 chunks tile16-eligible (events>=8192 and events/occupied_32px_tiles>=32)
+Dummy PSF producers: 16 workers; classification/chunk wall 232.340 s
+Frame 0: catalog classification finished in 232.3 s; reporting chunk statistics...
+Dummy PSF classified 598264924 images from 343363141 catalog stars; no HDR, PNG, or PPM was written.
+PSF splats: cached 598264924, cached wing-clipped 0, direct fallbacks 0, discarded below min-Y 0
+PROCESS_WALL 288.871615309 EXIT 0
+PROCESS_USER 3815.986955000 PROCESS_SYSTEM 12.140569000 MAX_RSS_KIB 13769936
+PROTECTED_AFTER output/imgs/2mass_galactic_center_schwarzschild_R100_4k_1e13_beta4.5_16_4-full-gpu.png sha256=3e01feb911e60acce545a6a86870ca05f96250022e874eb8097bec88a8879bfd size=13072572 mtime_ns=1788744728751407011 unchanged=yes
+PROTECTED_AFTER output/imgs/2mass_galactic_center_schwarzschild_R100_4k_1e13_beta4.5_16_4-full-gpu_HDR.fits sha256=2408e62dcc22f147da284d87af9a2b0fdd0050eda2dabc23b6ec2554b2f1d38e size=99535680 mtime_ns=1788744726380022561 unchanged=yes

benchmarks/dummy_psf_chunks_2026-09-11/full_catalog_missing_time.log

@@ -0,0 +1,6 @@
+COMMAND OMP_NUM_THREADS=16 OMP_DYNAMIC=FALSE /usr/bin/time -v /home/wyj/Code/C/GR_4d_raytracing/build/Release/schwarzschild_sky_dummy --verbose --psf-relative-tail 1e-8 --psf-min-y 0 --max-cache-psf-flux 1e8 --all-sky-catalog assets/2mass/processed/all_sky --exposure 1e13 --psf-fwhm-pixels 2.7 --psf-moffat-beta 4.5 --catalog-load-workers 4 --hdr-output --lens-map-input output/lens/schwarzschild_galactic_center_R100_45deg_16_4_4k_.grlens --output output/imgs/2mass_galactic_center_schwarzschild_R100_4k_1e13_beta4.5_16_4-full-gpu.png
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 c71a843a6064cd730b4694c5fe4e61249a995106526da0fdca97bc625c7db470
+LENS_SHA256 905f960aea2d276f9586c79fba545a72cf742311e9e7e18faea293835c68bf28
+PROTECTED_BEFORE output/imgs/2mass_galactic_center_schwarzschild_R100_4k_1e13_beta4.5_16_4-full-gpu.png sha256=3e01feb911e60acce545a6a86870ca05f96250022e874eb8097bec88a8879bfd size=13072572 mtime_ns=1788744728751407011
+PROTECTED_BEFORE output/imgs/2mass_galactic_center_schwarzschild_R100_4k_1e13_beta4.5_16_4-full-gpu_HDR.fits sha256=2408e62dcc22f147da284d87af9a2b0fdd0050eda2dabc23b6ec2554b2f1d38e size=99535680 mtime_ns=1788744726380022561

benchmarks/dummy_psf_chunks_2026-09-11/run.py

@@ -0,0 +1,99 @@
+#!/usr/bin/env python3
+"""Run the fixed full-catalog dummy diagnostic once without touching images."""
+
+import fcntl
+import hashlib
+import os
+from pathlib import Path
+import resource
+import shlex
+import signal
+import subprocess
+import time
+
+ROOT = Path(__file__).resolve().parents[2]
+LOG = Path(__file__).with_name("full_catalog.log")
+BINARY = ROOT / "build/Release/schwarzschild_sky_dummy"
+LENS = ROOT / "output/lens/schwarzschild_galactic_center_R100_45deg_16_4_4k_.grlens"
+PNG = ROOT / "output/imgs/2mass_galactic_center_schwarzschild_R100_4k_1e13_beta4.5_16_4-full-gpu.png"
+FITS = ROOT / "output/imgs/2mass_galactic_center_schwarzschild_R100_4k_1e13_beta4.5_16_4-full-gpu_HDR.fits"
+
+
+def digest(path):
+    return hashlib.sha256(path.read_bytes()).hexdigest()
+
+
+def main():
+    if LOG.exists():
+        raise RuntimeError(f"refusing to overwrite {LOG}")
+    for path in (BINARY, LENS, PNG, FITS):
+        if not path.is_file():
+            raise RuntimeError(f"required file is missing: {path}")
+    protected_before = {
+        path: (digest(path), path.stat().st_size, path.stat().st_mtime_ns)
+        for path in (PNG, FITS)
+    }
+    command = [
+        str(BINARY), "--verbose",
+        "--psf-relative-tail", "1e-8", "--psf-min-y", "0",
+        "--max-cache-psf-flux", "1e8",
+        "--all-sky-catalog", "assets/2mass/processed/all_sky",
+        "--exposure", "1e13", "--psf-fwhm-pixels", "2.7",
+        "--psf-moffat-beta", "4.5", "--catalog-load-workers", "4",
+        "--hdr-output", "--lens-map-input",
+        "output/lens/schwarzschild_galactic_center_R100_45deg_16_4_4k_.grlens",
+        "--output",
+        "output/imgs/2mass_galactic_center_schwarzschild_R100_4k_1e13_beta4.5_16_4-full-gpu.png",
+    ]
+    guard = open("/tmp/gr_dummy_psf_chunks.lock", "w")
+    fcntl.flock(guard, fcntl.LOCK_EX | fcntl.LOCK_NB)
+    env = dict(os.environ, OMP_NUM_THREADS="16", OMP_DYNAMIC="FALSE")
+    with LOG.open("x") as out:
+        out.write("COMMAND OMP_NUM_THREADS=16 OMP_DYNAMIC=FALSE " +
+                  shlex.join(command) + "\n")
+        out.write("GIT " + subprocess.check_output(
+            ["git", "rev-parse", "HEAD"], cwd=ROOT, text=True))
+        out.write("BINARY_SHA256 " + digest(BINARY) + "\n")
+        out.write("LENS_SHA256 " + digest(LENS) + "\n")
+        for path, state in protected_before.items():
+            out.write(
+                f"PROTECTED_BEFORE {path.relative_to(ROOT)} sha256={state[0]} "
+                f"size={state[1]} mtime_ns={state[2]}\n")
+        out.flush()
+        start = time.monotonic()
+        usage_start = resource.getrusage(resource.RUSAGE_CHILDREN)
+        process = subprocess.Popen(command, cwd=ROOT, env=env, stdout=out,
+                                   stderr=subprocess.STDOUT,
+                                   start_new_session=True)
+        try:
+            returncode = process.wait(timeout=600)
+        except subprocess.TimeoutExpired:
+            os.killpg(process.pid, signal.SIGKILL)
+            process.wait()
+            out.write("TIMEOUT killed and reaped process group\n")
+            raise
+        out.write(
+            f"PROCESS_WALL {time.monotonic() - start:.9f} EXIT {returncode}\n")
+        usage_end = resource.getrusage(resource.RUSAGE_CHILDREN)
+        out.write(
+            f"PROCESS_USER {usage_end.ru_utime - usage_start.ru_utime:.9f} "
+            f"PROCESS_SYSTEM {usage_end.ru_stime - usage_start.ru_stime:.9f} "
+            f"MAX_RSS_KIB {usage_end.ru_maxrss}\n")
+    protected_after = {
+        path: (digest(path), path.stat().st_size, path.stat().st_mtime_ns)
+        for path in (PNG, FITS)
+    }
+    if protected_after != protected_before:
+        raise RuntimeError("a protected render output changed")
+    with LOG.open("a") as out:
+        for path, state in protected_after.items():
+            out.write(
+                f"PROTECTED_AFTER {path.relative_to(ROOT)} sha256={state[0]} "
+                f"size={state[1]} mtime_ns={state[2]} unchanged=yes\n")
+    if returncode:
+        raise RuntimeError(f"dummy diagnostic exited {returncode}")
+    print(LOG)
+
+
+if __name__ == "__main__":
+    main()

benchmarks/hip_psf_replay_2026-09-07.md

@@ -13,6 +13,54 @@ RX 9070 上，65,536 个真实银心视场事件的累积回放中位数从 104.
 一个通用生产默认算法；后续集成应先考虑分布选择及生产流式 producer 的端到端对照。
 没有运行完整全天渲染，也没有推断新的全天加速比。
 
+## 2026-09-11 自适应选择后续实验
+
+在提交 `11e703e33dffb1ff168ba11dd922e68b448116b9` 之上增加了仅用于回放的
+`adaptive` 模式，production sink 仍未改变。选择器逐个 16,384-event chunk 扫描
+32×32 中心 tile 占用；chunk 至少有 8,192 events、且平均每个占用中心 tile 至少
+32 events 时选择 tile16，否则选择现有 atomic。该阈值是基于本页固定输入提出的实验
+假设，不是跨场景或跨设备参数。
+
+选择扫描不构建 support 引用，四个样本的中位成本为 0.037--0.455 ms。tile 路径使用
+独立的有界 event device buffer，不复用 production 双 staging slot，因此同一 stream 内
+可以按 chunk 混合 atomic/tile；mixed fixture 实际覆盖 2 个 tile chunk、1 个 atomic chunk
+和中途 direct fallback。所有额外 scratch 仍在 replay sink 创建时一次分配，未进入 renderer。
+
+同一新 binary（SHA256
+`121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151`）串行运行，
+每项三次；表中为包含选择、分桶、上传、kernel、合并、下载和清理的 `replay_wall`
+中位数：
+
+| 65,536-event 输入 | paired atomic | paired tile16 | adaptive | adaptive 选择 |
+| --- | ---: | ---: | ---: | ---: |
+| 银心密集 | 95.018 ms | 67.354 ms | 67.834 ms | 0.177 ms |
+| 同贡献分散 | 78.203 ms | 99.693 ms | 80.092 ms | 0.455 ms |
+
+自适应路径在密集输入相对 paired atomic 快约 1.40×，距固定 tile16 中位数 0.7%；在分散
+输入保留 atomic，距其 paired 中位数 2.4%。另外两个真实输入的自适应中位数为强放大
+21.617 ms、普通区域 18.347 ms，分别选择 1/1 tile chunk。银心选择 4/4 tile chunks，
+分散对照选择 4/4 atomic chunks。此结果支持继续研究 production 流式集成，但尚未包含
+OpenMP producer 与 sink 锁竞争，也没有授权用本回放结果宣称完整全天加速。
+
+全部 adaptive/mixed 检查通过原有 finite、double HDR 和 RGB/Y 通量阈值；最坏
+`max_abs=5.96856e-11`、`max_rel=5.95958e-13`、RGB 通量相对差
+`<=2.83233e-14`、Y `<=1.15036e-14`。`make -j1 test` 与 production HIP primitive
+也通过。第一次在受限沙箱内运行明确失败为 `no ROCm-capable device is detected`，随后只在
+获准的主机 GPU 环境执行；失败记录保留，没有删除或覆盖。
+
+原始证据：
+
+- 构建：[adaptive_build_host.log](hip_psf_replay_2026-09-07/adaptive_build_host.log)；
+- adaptive 三次：`adaptive_{center_65536,dispersed_65536,lensed_16384,ordinary_12765}_host_{0,1,2}.log`；
+- paired 基线：`paired_{center,dispersed}_{atomic,16}_{0,1,2}.log`；
+- mixed、CPU 与 HIP 回归：`adaptive_boundary_mixed_host_0.log`、
+  `adaptive_cpu_regression.log`、`adaptive_hip_primitive_host.log`；
+- 受限环境失败：`adaptive_center_65536_0.log`。
+
+每份日志均保留完整命令、45/90 s timeout、Git hash、binary hash（适用时）、程序原始输出、
+process wall 和 exit status。所有性能进程均由现有独占 runner 串行执行，未与其他 renderer
+或 profiler 并发。
+
 实现入口：
 
 - [capture_psf.c](../tests/capture_psf.c)：调用真实 producer 的测试专用 consumer，捕获或诊断，不生成 HDR。

benchmarks/hip_psf_replay_2026-09-07/adaptive_boundary_mixed_host_0.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/boundary.events 32771 adaptive mixed
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.691 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=32771 frame=128x96 mode=adaptive distribution=fixture-mixed-direct contributions=142133685 CPU_reference=0.271268845 create=0.036700010
+complete_replay=0.060321093 replay_wall=0.023621082 select=0.000166893 bin=0.002514124 upload=0.000438220 prepare=0.000985861 accumulation=0.016374943 merge=0.000088215 download=0.000132840 cleanup=0.000163078 events_s=1387362.339 contributions_s=6017238463.474 center_tiles=26 tile_chunks=2 atomic_chunks=1 refs=829409 tasks=3275 scratch_used_peak=25249440 scratch_device_reserved=184549376 RSS_KiB=319696 max_abs=5.9685589803848416e-11 max_rel=5.9595772318269279e-13 flux_R=2.8323267191886195e-14 flux_G=-5.5281251977629701e-15 flux_B=9.6507636840237455e-15 flux_Y=1.1503592402127278e-14
+PROCESS_WALL 1.117062220 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_build_host.log

@@ -0,0 +1,7 @@
+COMMAND OMP_NUM_THREADS=16 make -B -j1 PSF_BACKEND=hip 'HIP_CXXFLAGS=--offload-arch=gfx1201 -std=c++17 -O2 -Wall -Wextra -Wpedantic' hip-psf-replay
+TIMEOUT 60s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+cc -DENABLE_PNG -DFRAME_PSF_EVENT_SINK=1 -DPSF_BACKEND_HIP -DSPACETIME_MINKOWSKI  -std=c11 -march=native -pipe -Wall -Wextra -Wpedantic -O2 -DNDEBUG -fopenmp -Isrc -MMD -MP -c src/optics.c -o build/Release/obj/minkowski/standard_sink1_hip/src/optics.o
+mkdir -p build/Release
+hipcc --offload-arch=gfx1201 -std=c++17 -O2 -Wall -Wextra -Wpedantic -fopenmp -Isrc -x hip tests/replay_psf.hip -x none build/Release/obj/minkowski/standard_sink1_hip/src/optics.o -lm -lpng  -o build/Release/replay_psf
+PROCESS_WALL 1.266911305 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_center_65536_0.log

@@ -0,0 +1,7 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 adaptive
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+no ROCm-capable device is detected
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.667 s
+PROCESS_WALL 1.667466220 EXIT 1

benchmarks/hip_psf_replay_2026-09-07/adaptive_center_65536_host_0.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 adaptive
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.679 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=adaptive distribution=captured contributions=432818979 CPU_reference=0.824280024 create=0.050632000
+complete_replay=0.121458054 replay_wall=0.070826054 select=0.000172138 bin=0.007817984 upload=0.001097918 prepare=0.001653194 accumulation=0.051779747 merge=0.000582457 download=0.004020136 cleanup=0.000195980 events_s=925309.214 contributions_s=6111013629.616 center_tiles=35 tile_chunks=4 atomic_chunks=0 refs=2973775 tasks=11935 scratch_used_peak=35382764 scratch_device_reserved=184549376 RSS_KiB=713084 max_abs=2.7105054312137611e-19 max_rel=1.9611487209714066e-14 flux_R=-2.8081518355845397e-17 flux_G=4.7080671715585728e-17 flux_B=5.9271432925574641e-17 flux_Y=2.0330567515243999e-17
+PROCESS_WALL 1.968497664 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_center_65536_host_1.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 adaptive
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.662 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=adaptive distribution=captured contributions=432818979 CPU_reference=0.812095881 create=0.036637783
+complete_replay=0.104471922 replay_wall=0.067834139 select=0.000179052 bin=0.007610083 upload=0.001078129 prepare=0.001619816 accumulation=0.051049709 merge=0.000552416 download=0.004829631 cleanup=0.000203133 events_s=966121.205 contributions_s=6380547998.522 center_tiles=35 tile_chunks=4 atomic_chunks=0 refs=2973775 tasks=11935 scratch_used_peak=35382764 scratch_device_reserved=184549376 RSS_KiB=713056 max_abs=2.7105054312137611e-19 max_rel=1.9611487209714066e-14 flux_R=-2.8081518355845397e-17 flux_G=4.7080671715585728e-17 flux_B=5.9271432925574641e-17 flux_Y=2.0330567515243999e-17
+PROCESS_WALL 1.918472017 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_center_65536_host_2.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 adaptive
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.664 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=adaptive distribution=captured contributions=432818979 CPU_reference=0.806693077 create=0.036690950
+complete_replay=0.103733778 replay_wall=0.067042828 select=0.000176907 bin=0.007581234 upload=0.001075745 prepare=0.001624107 accumulation=0.051224947 merge=0.000552177 download=0.004083938 cleanup=0.000195026 events_s=977524.403 contributions_s=6455858060.497 center_tiles=35 tile_chunks=4 atomic_chunks=0 refs=2973775 tasks=11935 scratch_used_peak=35382764 scratch_device_reserved=184549376 RSS_KiB=713112 max_abs=2.7105054312137611e-19 max_rel=1.9611487209714066e-14 flux_R=-2.8081518355845397e-17 flux_G=4.7080671715585728e-17 flux_B=5.9271432925574641e-17 flux_Y=2.0330567515243999e-17
+PROCESS_WALL 1.918304047 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_cpu_regression.log

@@ -0,0 +1,38 @@
+COMMAND OMP_NUM_THREADS=16 make -j1 test
+TIMEOUT 90s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+make SPACETIME=minkowski ENABLE_HDR=1 backend
+make[1]: Entering directory '/home/wyj/Code/C/GR_4d_raytracing'
+cc -O2 -DNDEBUG -fopenmp build/Release/obj/minkowski/hdr_sink1_cpu/src/catalog.o build/Release/obj/minkowski/hdr_sink1_cpu/src/frame.o build/Release/obj/minkowski/hdr_sink1_cpu/src/geodesic.o build/Release/obj/minkowski/hdr_sink1_cpu/src/lens_map.o build/Release/obj/minkowski/hdr_sink1_cpu/src/movie.o build/Release/obj/minkowski/hdr_sink1_cpu/src/observer.o build/Release/obj/minkowski/hdr_sink1_cpu/src/observer_track.o build/Release/obj/minkowski/hdr_sink1_cpu/src/optics.o build/Release/obj/minkowski/hdr_sink1_cpu/src/ray.o build/Release/obj/minkowski/hdr_sink1_cpu/src/spacetime_common.o build/Release/obj/minkowski/hdr_sink1_cpu/src/spacetime_minkowski.o build/Release/obj/minkowski/hdr_sink1_cpu/src/main.o -lm -lpng -lcfitsio -o build/Release/minkowski_sky
+make[1]: Leaving directory '/home/wyj/Code/C/GR_4d_raytracing'
+mkdir -p /tmp/gr_psf_event_sink_reference
+OMP_NUM_THREADS=16 build/Release/minkowski_sky --catalog assets/sky_grid_5deg.csv --output /tmp/gr_psf_event_sink_reference/minkowski_ra1_dec1_640x360.png --hdr-output --width 640 --height 360 --fov-deg 30 --look-ra-deg 1 --look-dec-deg 1 --exposure 0.1 --observer-radius 30 --observer-velocity 0 0 0 --psf-fwhm-pixels 2.7 --psf-moffat-beta 4.5 --max-magnification 1e300 --max-cache-psf-flux 1 --psf-relative-tail 1e-8 --psf-min-y 0 --coarse-cell-pixels 16 --refine-max-level 0 --refine-angle-abs-deg 0.001 --refine-angle-rel 0.1 --refine-jacobian-min 1e-3 --refine-min-edge-pixels 0.5 --refine-min-area-pixels2 0.25 --catalog-load-workers 4
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.667 s
+Rendered 35 images from 5654 catalog stars to /tmp/gr_psf_event_sink_reference/minkowski_ra1_dec1_640x360.png (ok)
+PSF splats: cached 35, cached wing-clipped 0, direct fallbacks 0, discarded below min-Y 0
+python3 scripts/fits_floatdiff.py tests/data/psf_event_sink_reference/minkowski_ra1_dec1_640x360_HDR.fits /tmp/gr_psf_event_sink_reference/minkowski_ra1_dec1_640x360_HDR.fits
+floatdiff: shape=(640, 360, 3) samples=691200 mismatches=0 max_abs=0 max_rel=0 worst_sample=0
+make SPACETIME=schwarzschild ENABLE_HDR=1 backend
+make[1]: Entering directory '/home/wyj/Code/C/GR_4d_raytracing'
+cc -O2 -DNDEBUG -fopenmp build/Release/obj/schwarzschild/hdr_sink1_cpu/src/catalog.o build/Release/obj/schwarzschild/hdr_sink1_cpu/src/frame.o build/Release/obj/schwarzschild/hdr_sink1_cpu/src/geodesic.o build/Release/obj/schwarzschild/hdr_sink1_cpu/src/lens_map.o build/Release/obj/schwarzschild/hdr_sink1_cpu/src/movie.o build/Release/obj/schwarzschild/hdr_sink1_cpu/src/observer.o build/Release/obj/schwarzschild/hdr_sink1_cpu/src/observer_track.o build/Release/obj/schwarzschild/hdr_sink1_cpu/src/optics.o build/Release/obj/schwarzschild/hdr_sink1_cpu/src/ray.o build/Release/obj/schwarzschild/hdr_sink1_cpu/src/spacetime_common.o build/Release/obj/schwarzschild/hdr_sink1_cpu/src/spacetime_schwarzschild.o build/Release/obj/schwarzschild/hdr_sink1_cpu/src/main.o -lm -lpng -lcfitsio -o build/Release/schwarzschild_sky
+make[1]: Leaving directory '/home/wyj/Code/C/GR_4d_raytracing'
+OMP_NUM_THREADS=16 build/Release/schwarzschild_sky --catalog assets/sky_grid_5deg.csv --output /tmp/gr_psf_event_sink_reference/schwarzschild_ra1_dec1_fov60_640x360.png --hdr-output --width 640 --height 360 --fov-deg 60 --look-ra-deg 1 --look-dec-deg 1 --exposure 0.1 --observer-radius 30 --observer-velocity 0 0 0 --psf-fwhm-pixels 2.7 --psf-moffat-beta 4.5 --max-magnification 1e300 --max-cache-psf-flux 1 --psf-relative-tail 1e-8 --psf-min-y 0 --coarse-cell-pixels 16 --refine-max-level 0 --refine-angle-abs-deg 0.001 --refine-angle-rel 0.1 --refine-jacobian-min 1e-3 --refine-min-edge-pixels 0.5 --refine-min-area-pixels2 0.25 --catalog-load-workers 4
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.665 s
+Rendered 5206 images from 5654 catalog stars to /tmp/gr_psf_event_sink_reference/schwarzschild_ra1_dec1_fov60_640x360.png (ok)
+PSF splats: cached 5206, cached wing-clipped 0, direct fallbacks 0, discarded below min-Y 0
+python3 scripts/fits_floatdiff.py tests/data/psf_event_sink_reference/schwarzschild_ra1_dec1_fov60_640x360_HDR.fits /tmp/gr_psf_event_sink_reference/schwarzschild_ra1_dec1_fov60_640x360_HDR.fits
+floatdiff: shape=(640, 360, 3) samples=691200 mismatches=0 max_abs=0 max_rel=0 worst_sample=0
+./build/Release/test_observer_minkowski
+coordinate-camera regression passed
+./build/Release/test_observer_schwarzschild
+coordinate-camera regression passed
+./build/Release/test_geodesic
+./build/Release/test_frame
+./build/Release/test_schwarzschild
+./build/Release/test_observer_track
+./build/Release/test_catalog_prefetch
+python3 tests/test_camera_cli.py build/Release
+minkowski: CLI checks passed; single/movie PNG identical, map max error 0
+schwarzschild: movie lens map identical with 1, 4 and 16 threads
+schwarzschild: CLI checks passed; single/movie PNG identical, map max error 4.44e-16
+PROCESS_WALL 26.018675067 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_dispersed_65536_host_0.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 adaptive disperse
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.682 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=adaptive distribution=synthetic-phase-preserving-dispersal contributions=432818979 CPU_reference=1.491110802 create=0.037611008
+complete_replay=0.117703199 replay_wall=0.080092192 select=0.000455141 bin=0.000000000 upload=0.000191523 prepare=0.000000000 accumulation=0.075263899 merge=0.000000000 download=0.004099738 cleanup=0.000207186 events_s=818257.044 contributions_s=5404009677.272 center_tiles=26986 tile_chunks=0 atomic_chunks=4 refs=0 tasks=0 scratch_used_peak=0 scratch_device_reserved=184549376 RSS_KiB=705516 max_abs=3.3881317890172014e-21 max_rel=8.4356460581299069e-16 flux_R=1.2382136141425945e-18 flux_G=1.811942792627668e-18 flux_B=1.5829277137138293e-18 flux_Y=1.5978644712371554e-18
+PROCESS_WALL 2.620178348 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_dispersed_65536_host_1.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 adaptive disperse
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.733 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=adaptive distribution=synthetic-phase-preserving-dispersal contributions=432818979 CPU_reference=1.488748074 create=0.038790941
+complete_replay=0.121706009 replay_wall=0.082915068 select=0.000465870 bin=0.000000000 upload=0.000181281 prepare=0.000000000 accumulation=0.077964972 merge=0.000000000 download=0.004236780 cleanup=0.000192881 events_s=790399.162 contributions_s=5220028049.767 center_tiles=26986 tile_chunks=0 atomic_chunks=4 refs=0 tasks=0 scratch_used_peak=0 scratch_device_reserved=184549376 RSS_KiB=705940 max_abs=3.3881317890172014e-21 max_rel=8.4356460581299069e-16 flux_R=1.377382799700594e-18 flux_G=1.939309838343059e-18 flux_B=2.0425953472603454e-18 flux_Y=1.7396938046841301e-18
+PROCESS_WALL 2.720085292 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_dispersed_65536_host_2.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 adaptive disperse
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.671 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=adaptive distribution=synthetic-phase-preserving-dispersal contributions=432818979 CPU_reference=1.462161779 create=0.037038088
+complete_replay=0.115585089 replay_wall=0.078547001 select=0.000451088 bin=0.000000000 upload=0.000188241 prepare=0.000000000 accumulation=0.073739567 merge=0.000000000 download=0.004098460 cleanup=0.000214100 events_s=834353.944 contributions_s=5510318333.269 center_tiles=26986 tile_chunks=0 atomic_chunks=4 refs=0 tasks=0 scratch_used_peak=0 scratch_device_reserved=184549376 RSS_KiB=705560 max_abs=3.3881317890172014e-21 max_rel=8.4356460581299069e-16 flux_R=1.4160413575526434e-18 flux_G=1.6738012163520145e-18 flux_B=2.0518958448496421e-18 flux_Y=1.5923904683351809e-18
+PROCESS_WALL 2.568827678 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_hip_primitive_host.log

@@ -0,0 +1,6 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/test_hip_psf
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 39cd23b6d9aef423ba97ba0085b9a411a47a49e4b66cf59197f471de419ef9b8
+HIP PSF cache comparison: max_abs=2.8421709430404007e-14 max_rel=7.5319184269262976e-16
+PROCESS_WALL 0.816928134 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_lensed_16384_host_0.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/lensed_16384.events 16384 adaptive
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.714 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=16384 frame=3840x2160 mode=adaptive distribution=captured contributions=107949448 CPU_reference=0.213827133 create=0.041162014
+complete_replay=0.062693119 replay_wall=0.021531105 select=0.000054836 bin=0.003102064 upload=0.000283003 prepare=0.000669003 accumulation=0.012633085 merge=0.000146866 download=0.004084578 cleanup=0.000169992 events_s=760945.617 contributions_s=5013651077.913 center_tiles=359 tile_chunks=1 atomic_chunks=0 refs=740784 tasks=5251 scratch_used_peak=31785268 scratch_device_reserved=184549376 RSS_KiB=710652 max_abs=5.3290705182007514e-15 max_rel=3.8619817275125971e-15 flux_R=-5.5779263000586243e-18 flux_G=-2.9489947183383731e-19 flux_B=4.3765760168230462e-18 flux_Y=-2.1608907287060103e-18
+PROCESS_WALL 1.268419138 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_lensed_16384_host_1.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/lensed_16384.events 16384 adaptive
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.667 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=16384 frame=3840x2160 mode=adaptive distribution=captured contributions=107949448 CPU_reference=0.214478970 create=0.036618948
+complete_replay=0.058495998 replay_wall=0.021877050 select=0.000056028 bin=0.003196001 upload=0.000288010 prepare=0.000689983 accumulation=0.012602091 merge=0.000144958 download=0.004350181 cleanup=0.000192165 events_s=748912.660 contributions_s=4934369397.489 center_tiles=359 tile_chunks=1 atomic_chunks=0 refs=740784 tasks=5251 scratch_used_peak=31785268 scratch_device_reserved=184549376 RSS_KiB=710384 max_abs=5.3290705182007514e-15 max_rel=3.8619817275125971e-15 flux_R=-5.5779263000586243e-18 flux_G=-2.9489947183383731e-19 flux_B=4.3765760168230462e-18 flux_Y=-2.1608907287060103e-18
+PROCESS_WALL 1.216200115 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_lensed_16384_host_2.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/lensed_16384.events 16384 adaptive
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.664 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=16384 frame=3840x2160 mode=adaptive distribution=captured contributions=107949448 CPU_reference=0.214956999 create=0.038650990
+complete_replay=0.060267925 replay_wall=0.021616936 select=0.000054121 bin=0.003103971 upload=0.000282049 prepare=0.000669956 accumulation=0.012650013 merge=0.000149012 download=0.004124364 cleanup=0.000187874 events_s=757924.259 contributions_s=4993744226.675 center_tiles=359 tile_chunks=1 atomic_chunks=0 refs=740784 tasks=5251 scratch_used_peak=31785268 scratch_device_reserved=184549376 RSS_KiB=710600 max_abs=5.3290705182007514e-15 max_rel=3.8619817275125971e-15 flux_R=-5.5779263000586243e-18 flux_G=-2.9489947183383731e-19 flux_B=4.3765760168230462e-18 flux_Y=-2.1608907287060103e-18
+PROCESS_WALL 1.216335608 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_ordinary_12765_host_0.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/ordinary_16384.events 12765 adaptive
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.683 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=12765 frame=3840x2160 mode=adaptive distribution=captured contributions=84287857 CPU_reference=0.161495924 create=0.036530018
+complete_replay=0.054877043 replay_wall=0.018347025 select=0.000042915 bin=0.001914978 upload=0.000241995 prepare=0.000617027 accumulation=0.009811878 merge=0.000136137 download=0.005169831 cleanup=0.000195026 events_s=695753.129 contributions_s=4594088544.521 center_tiles=8 tile_chunks=1 atomic_chunks=0 refs=575476 tasks=2331 scratch_used_peak=27265820 scratch_device_reserved=184549376 RSS_KiB=708332 max_abs=2.1371793224034263e-15 max_rel=1.6186841366251043e-14 flux_R=1.3070869891010463e-16 flux_G=2.5758793226204442e-16 flux_B=1.7069617471680846e-16 flux_Y=2.1355251009923229e-16
+PROCESS_WALL 1.166123833 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_ordinary_12765_host_1.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/ordinary_16384.events 12765 adaptive
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.679 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=12765 frame=3840x2160 mode=adaptive distribution=captured contributions=84287857 CPU_reference=0.164183140 create=0.036640882
+complete_replay=0.054260731 replay_wall=0.017619848 select=0.000036955 bin=0.001922131 upload=0.000246048 prepare=0.000601053 accumulation=0.010066986 merge=0.000137806 download=0.004148300 cleanup=0.000171900 events_s=724467.079 contributions_s=4783688020.331 center_tiles=8 tile_chunks=1 atomic_chunks=0 refs=575476 tasks=2331 scratch_used_peak=27265820 scratch_device_reserved=184549376 RSS_KiB=707824 max_abs=2.1371793224034263e-15 max_rel=1.6186841366251043e-14 flux_R=1.3070869891010463e-16 flux_G=2.5758793226204442e-16 flux_B=1.7069617471680846e-16 flux_Y=2.1355251009923229e-16
+PROCESS_WALL 1.167055415 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/adaptive_ordinary_12765_host_2.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/ordinary_16384.events 12765 adaptive
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.674 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=12765 frame=3840x2160 mode=adaptive distribution=captured contributions=84287857 CPU_reference=0.159490824 create=0.036442041
+complete_replay=0.056708097 replay_wall=0.020266056 select=0.000033855 bin=0.001877069 upload=0.000234127 prepare=0.000597954 accumulation=0.010081053 merge=0.000137806 download=0.004184938 cleanup=0.000163078 events_s=629870.951 contributions_s=4159065619.239 center_tiles=8 tile_chunks=1 atomic_chunks=0 refs=575476 tasks=2331 scratch_used_peak=27265820 scratch_device_reserved=184549376 RSS_KiB=707952 max_abs=2.1371793224034263e-15 max_rel=1.6186841366251043e-14 flux_R=1.3070869891010463e-16 flux_G=2.5758793226204442e-16 flux_B=1.7069617471680846e-16 flux_Y=2.1355251009923229e-16
+PROCESS_WALL 1.167496176 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/paired_center_16_0.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 16
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.783 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=16 distribution=captured contributions=432818979 CPU_reference=0.805742979 create=0.036848068
+complete_replay=0.103791952 replay_wall=0.066943884 select=0.000000000 bin=0.007535458 upload=0.001190662 prepare=0.001640081 accumulation=0.051215172 merge=0.000553846 download=0.004045417 cleanup=0.000201941 events_s=978969.193 contributions_s=6465399881.387 center_tiles=0 tile_chunks=4 atomic_chunks=0 refs=2973775 tasks=11935 scratch_used_peak=35382764 scratch_device_reserved=184549376 RSS_KiB=713288 max_abs=2.7105054312137611e-19 max_rel=1.9611487209714066e-14 flux_R=-2.8081518355845397e-17 flux_G=4.7080671715585728e-17 flux_B=5.9271432925574641e-17 flux_Y=2.0330567515243999e-17
+PROCESS_WALL 2.019127664 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/paired_center_16_1.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 16
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.668 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=16 distribution=captured contributions=432818979 CPU_reference=0.815722942 create=0.036595821
+complete_replay=0.106936932 replay_wall=0.070341110 select=0.000000000 bin=0.007705212 upload=0.001186848 prepare=0.001616001 accumulation=0.051277876 merge=0.000541925 download=0.004584385 cleanup=0.000200987 events_s=931688.451 contributions_s=6153143980.638 center_tiles=0 tile_chunks=4 atomic_chunks=0 refs=2973775 tasks=11935 scratch_used_peak=35382764 scratch_device_reserved=184549376 RSS_KiB=712508 max_abs=2.7105054312137611e-19 max_rel=1.9611487209714066e-14 flux_R=-2.8081518355845397e-17 flux_G=4.7080671715585728e-17 flux_B=5.9271432925574641e-17 flux_Y=2.0330567515243999e-17
+PROCESS_WALL 1.917609542 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/paired_center_16_2.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 16
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.670 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=16 distribution=captured contributions=432818979 CPU_reference=0.802520990 create=0.037091017
+complete_replay=0.104444981 replay_wall=0.067353964 select=0.000000000 bin=0.007662058 upload=0.001219988 prepare=0.001631737 accumulation=0.051412106 merge=0.000540018 download=0.004093417 cleanup=0.000197887 events_s=973008.807 contributions_s=6426035740.844 center_tiles=0 tile_chunks=4 atomic_chunks=0 refs=2973775 tasks=11935 scratch_used_peak=35382764 scratch_device_reserved=184549376 RSS_KiB=712736 max_abs=2.7105054312137611e-19 max_rel=1.9611487209714066e-14 flux_R=-2.8081518355845397e-17 flux_G=4.7080671715585728e-17 flux_B=5.9271432925574641e-17 flux_Y=2.0330567515243999e-17
+PROCESS_WALL 1.917963698 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/paired_center_atomic_0.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 atomic
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.687 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=atomic distribution=captured contributions=432818979 CPU_reference=0.809880018 create=0.039862156
+complete_replay=0.135050297 replay_wall=0.095188141 select=0.000000000 bin=0.000000000 upload=0.000205602 prepare=0.000000000 accumulation=0.090687719 merge=0.000000000 download=0.004133141 cleanup=0.000000000 events_s=688489.127 contributions_s=4546984267.662 center_tiles=0 tile_chunks=0 atomic_chunks=4 refs=0 tasks=0 scratch_used_peak=0 scratch_device_reserved=0 RSS_KiB=705156 max_abs=1.2874900798265365e-19 max_rel=7.4131884297116165e-15 flux_R=3.5244625165286513e-17 flux_G=-1.6770129667889592e-17 flux_B=-1.4396324706080044e-17 flux_Y=2.0737623150838269e-18
+PROCESS_WALL 1.967518921 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/paired_center_atomic_1.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 atomic
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.671 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=atomic distribution=captured contributions=432818979 CPU_reference=0.802126169 create=0.039602995
+complete_replay=0.134557009 replay_wall=0.094954014 select=0.000000000 bin=0.000000000 upload=0.000198362 prepare=0.000000000 accumulation=0.090435339 merge=0.000000000 download=0.004166579 cleanup=0.000000000 events_s=690186.727 contributions_s=4558195715.666 center_tiles=0 tile_chunks=0 atomic_chunks=4 refs=0 tasks=0 scratch_used_peak=0 scratch_device_reserved=0 RSS_KiB=705688 max_abs=1.0842021724855044e-19 max_rel=7.9583748032166079e-15 flux_R=7.6396533553465937e-18 flux_G=1.6562512796306053e-17 flux_B=1.5235277925542237e-18 flux_Y=1.2880679208252085e-17
+PROCESS_WALL 1.918794264 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/paired_center_atomic_2.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 atomic
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.676 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=atomic distribution=captured contributions=432818979 CPU_reference=0.799952984 create=0.039901972
+complete_replay=0.134919882 replay_wall=0.095017910 select=0.000000000 bin=0.000000000 upload=0.000204603 prepare=0.000000000 accumulation=0.090419817 merge=0.000000000 download=0.004226341 cleanup=0.000000000 events_s=689722.601 contributions_s=4555130490.487 center_tiles=0 tile_chunks=0 atomic_chunks=4 refs=0 tasks=0 scratch_used_peak=0 scratch_device_reserved=0 RSS_KiB=705636 max_abs=1.2874900798265365e-19 max_rel=7.9482312095117156e-15 flux_R=-5.3427317744228327e-18 flux_G=6.0837775507051433e-17 flux_B=-9.8579921222560893e-19 flux_Y=3.5056388280753197e-17
+PROCESS_WALL 1.968038807 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/paired_dispersed_16_0.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 16 disperse
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.662 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=16 distribution=synthetic-phase-preserving-dispersal contributions=432818979 CPU_reference=1.480471134 create=0.038483858
+complete_replay=0.138176918 replay_wall=0.099693060 select=0.000000000 bin=0.035099983 upload=0.001532316 prepare=0.001661777 accumulation=0.049884081 merge=0.000493050 download=0.004275622 cleanup=0.000198841 events_s=657377.756 contributions_s=4341515641.528 center_tiles=0 tile_chunks=4 atomic_chunks=0 refs=2987132 tasks=129439 scratch_used_peak=17137368 scratch_device_reserved=184549376 RSS_KiB=716772 max_abs=6.7762635780344027e-21 max_rel=2.2452117590762424e-15 flux_R=1.0178317181173255e-18 flux_G=1.4444219523404554e-18 flux_B=2.8354584230664648e-18 flux_Y=1.3331859086679348e-18
+PROCESS_WALL 2.620675482 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/paired_dispersed_16_1.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 16 disperse
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.672 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=16 distribution=synthetic-phase-preserving-dispersal contributions=432818979 CPU_reference=1.477596998 create=0.037245035
+complete_replay=0.137820959 replay_wall=0.100575924 select=0.000000000 bin=0.035495043 upload=0.001560211 prepare=0.001640797 accumulation=0.050065994 merge=0.000494242 download=0.004161540 cleanup=0.000195980 events_s=651607.238 contributions_s=4303405448.660 center_tiles=0 tile_chunks=4 atomic_chunks=0 refs=2987132 tasks=129439 scratch_used_peak=17137368 scratch_device_reserved=184549376 RSS_KiB=716808 max_abs=6.7762635780344027e-21 max_rel=2.2452117590762424e-15 flux_R=1.0178317181173255e-18 flux_G=1.4444219523404554e-18 flux_B=2.8354584230664648e-18 flux_Y=1.3331859086679348e-18
+PROCESS_WALL 2.667392709 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/paired_dispersed_16_2.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 16 disperse
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.675 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=16 distribution=synthetic-phase-preserving-dispersal contributions=432818979 CPU_reference=1.470952988 create=0.038564920
+complete_replay=0.130317926 replay_wall=0.091753006 select=0.000000000 bin=0.034946680 upload=0.001475096 prepare=0.001478195 accumulation=0.042901754 merge=0.000442028 download=0.004153220 cleanup=0.000182867 events_s=714265.427 contributions_s=4717218519.113 center_tiles=0 tile_chunks=4 atomic_chunks=0 refs=2987132 tasks=129439 scratch_used_peak=17137368 scratch_device_reserved=184549376 RSS_KiB=714840 max_abs=6.7762635780344027e-21 max_rel=2.2452117590762424e-15 flux_R=1.0178317181173255e-18 flux_G=1.4444219523404554e-18 flux_B=2.8354584230664648e-18 flux_Y=1.3331859086679348e-18
+PROCESS_WALL 2.623926037 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/paired_dispersed_atomic_0.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 atomic disperse
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.685 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=atomic distribution=synthetic-phase-preserving-dispersal contributions=432818979 CPU_reference=1.483494043 create=0.036720991
+complete_replay=0.115373135 replay_wall=0.078652143 select=0.000000000 bin=0.000000000 upload=0.000209722 prepare=0.000000000 accumulation=0.074174986 merge=0.000000000 download=0.004105619 cleanup=0.000000000 events_s=833238.576 contributions_s=5502952111.139 center_tiles=0 tile_chunks=0 atomic_chunks=4 refs=0 tasks=0 scratch_used_peak=0 scratch_device_reserved=0 RSS_KiB=705688 max_abs=3.3881317890172014e-21 max_rel=8.4356460581299069e-16 flux_R=1.4264853877147901e-18 flux_G=1.7304063360850003e-18 flux_B=2.1361348632691914e-18 flux_Y=1.6331855317630113e-18
+PROCESS_WALL 2.620796063 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/paired_dispersed_atomic_1.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 atomic disperse
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.669 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=atomic distribution=synthetic-phase-preserving-dispersal contributions=432818979 CPU_reference=1.474539995 create=0.037378073
+complete_replay=0.115452051 replay_wall=0.078073978 select=0.000000000 bin=0.000000000 upload=0.000202122 prepare=0.000000000 accumulation=0.073547550 merge=0.000000000 download=0.004160663 cleanup=0.000000000 events_s=839408.998 contributions_s=5543703391.789 center_tiles=0 tile_chunks=0 atomic_chunks=4 refs=0 tasks=0 scratch_used_peak=0 scratch_device_reserved=0 RSS_KiB=704932 max_abs=3.3881317890172014e-21 max_rel=8.4356460581299069e-16 flux_R=1.4285334484627677e-18 flux_G=1.6416822084297311e-18 flux_B=2.2250851163298864e-18 flux_Y=1.5826745498492225e-18
+PROCESS_WALL 2.620563869 EXIT 0

benchmarks/hip_psf_replay_2026-09-07/paired_dispersed_atomic_2.log

@@ -0,0 +1,8 @@
+COMMAND OMP_NUM_THREADS=16 build/Release/replay_psf benchmarks/hip_psf_replay_2026-09-07/center_65536.events 65536 atomic disperse
+TIMEOUT 45s
+GIT 11e703e33dffb1ff168ba11dd922e68b448116b9
+BINARY_SHA256 121b85ade12680929cf0f560e04485e2ad3d8a56ff9dcbba950a323f840c5151
+PSF cache ready: 64x64 phases, radius 47 px, relative tail 1e-08, tail abs 1e-06, boundary 1e-07, build 0.666 s
+device=AMD Radeon RX 9070 arch=gfx1201 events=65536 frame=3840x2160 mode=atomic distribution=synthetic-phase-preserving-dispersal contributions=432818979 CPU_reference=1.498821974 create=0.037057161
+complete_replay=0.115260124 replay_wall=0.078202963 select=0.000000000 bin=0.000000000 upload=0.000203722 prepare=0.000000000 accumulation=0.073602102 merge=0.000000000 download=0.004232413 cleanup=0.000000000 events_s=838024.515 contributions_s=5534559856.636 center_tiles=0 tile_chunks=0 atomic_chunks=4 refs=0 tasks=0 scratch_used_peak=0 scratch_device_reserved=0 RSS_KiB=705384 max_abs=3.3881317890172014e-21 max_rel=8.4356460581299069e-16 flux_R=1.4138169203080735e-18 flux_G=1.5950365444592374e-18 flux_B=2.1462694853060243e-18 flux_Y=1.5465636583759892e-18
+PROCESS_WALL 2.619888980 EXIT 0

build.md

@@ -104,6 +104,29 @@ times must not be added to GPU timings or interpreted as process CPU time.
 The two upload slots bound staging memory and allow CPU production ahead of GPU
 completion; copies and kernels still execute in order on one stream.
 
+### Diagnostic dummy PSF backend
+
+To inspect production chunk geometry without starting HIP or allocating an HDR
+framebuffer, build the diagnostic-only dummy backend:
+
+```sh
+make -j1 PSF_BACKEND=dummy SPACETIME=schwarzschild ENABLE_HDR=1 backend
+OMP_NUM_THREADS=16 OMP_DYNAMIC=FALSE \
+  ./build/Release/schwarzschild_sky_dummy --lens-map-input MAP.grlens \
+  --all-sky-catalog CATALOG_DIR --output PROTECTED_OR_UNUSED.png [normal PSF options]
+```
+
+The dummy executable requires an imported lens map and the PSF cache; it rejects
+`--psf-direct`. It runs the real OpenMP triangle scheduling, catalog query,
+inverse mapping, redshift/color/flux preparation, and cache/direct/min-Y
+classification, but records chunk statistics instead of accumulating pixels.
+`--output` and `--hdr-output` are accepted only for command parity: neither PNG
+nor FITS is created or modified. Reports separate full 16,384-event chunks from
+the final partial chunk of each worker and include event-producing triangle
+counts, occupied 32-pixel center tiles, bounding-box diagonals, and successive
+triangle-centroid jumps. This is a scheduling diagnostic, not a rendering or
+backend-performance benchmark.
+
 With both Minkowski binaries built using `ENABLE_HDR=1`,
 `python3 tests/test_hip_renderer.py build/Release/minkowski_sky build/Release/minkowski_sky_hip`
 runs small CPU/HIP fallback and min-Y comparisons sequentially. Each renderer

gpu_psf_acceleration_plan.md

@@ -26,6 +26,22 @@ production HIP 尚未替换为 tile 算法；收益与事件分布有关，不
 全天加速比。未运行完整全天 catalog。测试工具与原型已于 2026-09-11 提交为
 `7f5ec1a`；以下保留原执行计划作为验收依据。
 
+同日后续实验在 replay 中增加了按 chunk 的中心 tile 密度选择器。当前 RX 9070 配对数据中，
+65,536-event 银心输入的 adaptive/atomic/tile16 中位数为 67.834/95.018/67.354 ms，
+同贡献分散输入为 80.092/78.203/99.693 ms；选择扫描中位成本不超过 0.455 ms。
+mixed fixture 已覆盖同一 stream 内 atomic、tile 与 direct fallback 的组合。结果支持下一步
+做 production 流式集成原型，但选择器和 tile sink 目前仍只在测试程序中，未改变生产默认。
+
+随后用只计数的 `PSF_BACKEND=dummy` 跑过用户给定的 45 度 Schwarzschild 银心全天命令，
+没有启动 HIP，也没有写 HDR/PNG。598,264,924 个 cache event 中有 36,508 个满
+16,384-event chunk；满 chunk 的 event-producing triangle 数 p50/p75/p90/p99 为
+2/3/10/31，occupied 32px center tile 数为 1/2/5/30，全部满足当前 adaptive tile16
+阈值。这把“dense replay”提升成了真实 production producer 顺序的证据，支持把 adaptive
+tile16 接入 production sink 作为下一步候选。限制是 16-worker dynamic schedule 会有运行间
+波动，而且 p99 chunk 含相距约 3.7K px 的稀疏簇；实现必须保留稀疏 tile 列表和 atomic
+fallback，不能假定单一连续小矩形。详见
+[dummy chunk 全量记录](benchmarks/dummy_psf_chunks_2026-09-11/README.md)。
+
 ## 下一步执行计划（2026-09-07，原计划；有界实验已实施）
 
 ### 依据与目标

nr_spacetime_movie_renderer_design.md

@@ -993,6 +993,52 @@ CPU 按完整、经 cache 限制的 support 矩形建立 tile 引用；GPU 仍
 合并、清理和下载的完整成本及分布依赖，不能只引用无 atomic kernel 的单项耗时。
 本轮生产 HIP 路径保持原实现，尚未进行新的完整全天渲染。
 
+2026-09-11 的后续有界实验加入了按 chunk 的分布选择器：先统计 32px 中心 tile 占用，
+再在高中心密度 chunk 使用 16×16 像素归约，其余保留 atomic。RX 9070 的配对回放中，
+它在银心密集输入接近固定 tile16，同时在同贡献分散输入选择 atomic，详情见上述回放记录。
+此选择器仍仅存在于测试程序；阈值来自固定样本，尚未验证 production producer/sink 锁竞争
+或完整全天事件序列，因此不能成为 renderer 默认策略。
+
+2026-09-11 又加入了 `PSF_BACKEND=dummy` 的只统计后端。它要求导入 `.grlens`，复用
+production 的 OpenMP `schedule(dynamic,1)`、每 worker 16,384-event chunk，以及完整的
+catalog/inverse-map/颜色/flux/cache 分类，但不分配完整 HDR，也绝不写 PNG/FITS。对银心
+45 度 Schwarzschild 全天输入的一次 16-worker 全量运行中，598,264,924 个 cache event
+形成 36,508 个满 chunk；满 chunk 跨越的 event-producing triangle 数为 p50=2、p75=3、
+p90=10、p99=31、max=59，32px 中心 tile 占用为 p50=1、p75=2、p90=5、p99=30、
+max=45。全部满 chunk 都满足当前测试选择器的 tile16 条件。
+
+这个结果支持“真实宽视场下一个 worker 的满 chunk 通常集中在少数三角形/少数局部
+tile”这一调度假设，但也揭示了稀有的远距离簇：满 chunk 的三角形质心包围盒对角线
+p99 约 3744px，最大相邻三角形跳跃 p99 也约 3744px。因而不能把“少数三角形”等同于
+“全 chunk 单个连续矩形”；production tile 原型仍应按稀疏 occupied-tile 列表处理，并用
+真实流式端到端数据验证分桶成本。完整命令、hash、输出保护和原始记录见
+[dummy chunk 统计](benchmarks/dummy_psf_chunks_2026-09-11/README.md)。
+
+2026-09-12，黑洞渲染的黑体颜色计算改为以仓库内数值 LUT 为默认实现，取代原先
+`optics.c` 中的 Wyman 解析 CIE 拟合。正常区使用仓库唯一的
+`assets/blackbody/cie1931_2deg_xyz_1024.grbblut`：以 40 位精度把 Planck 谱和
+`assets/CIE_xyz_1931_2deg.csv` 的 360--830 nm、1 nm 分段线性 CIE 数据积分，生成覆盖
+`[670.146556,101408.88] K` 的 1024 个 `log(T)` 等距 XYZ 节点。`GRBBLUT3` header
+固定标识该 reference，loader 校验 magic/version/endianness/物理常数/范围/长度与 payload
+checksum，并拒绝旧的 Wyman/5 nm 表与 runtime 造表。表内使用四点三次 Lagrange 插值，
+在全部 1023 个区间中点相对独立 Gauss-16 积分的最大 XYZ channel 相对误差为
+`5.4633e-5`，没有负 XYZ。由 1 nm 线性插值自身约 `3.2e-4` 的最坏误差估计，正常区采用
+`1e-4` LUT 内部限与 `5e-4` 整体误差预算，避免继续追逐缺乏物理意义的 `1e-6`。
+
+两端使用同一 CSV reference 的表外模型。高温端使用连续积分得到的
+`A T+B+C/T` Rayleigh--Jeans 展开，在 `[Tmax,10Tmax]` 最大 XYZ 相对误差为 `1.6635e-5`。
+低温端针对 X/Y 的 830 nm 端点和 Z 的 650 nm 零端点分别展开：在 `4.097321 K` 以下使用
+crossover 锚定、在 `T -> 0` 趋于极限的端点式，其上用六段八次
+`1/T -> log(scaled XYZ)` Chebyshev--Lobatto 插值接到 LUT。低温端点段/过渡段的最大 XYZ
+相对误差分别为 `6.69e-5`/`8.16e-5`，因而三段均满足 `1e-4` 内部限；log 输出避免把极端
+低温的线性下溢误解为拟合硬下限。
+
+最终 legacy Wyman/5 nm 与 CSV LUT 的 1920x1080 double-HDR 比较显示，每通道峰值归一化
+最大绝对变化为 R `9.00e-3`、G `7.02e-3`、B `8.12e-4`；这是 reference 升级本身造成的
+预期变化，而不是 LUT 插值误差。旧 Wyman/5 nm 积分实现、可切换的 integral/fixed/LUT
+A/B 后端、有界性能证据与观测温度采样器保留在实验分支
+`codex/blackbody-cost-experiment`（tag `blackbody-cost-experiment-2026-09-12`）。
+
 
 ---
 

src/blackbody_common.c

@@ -0,0 +1,128 @@
+#include "blackbody_internal.h"
+
+#include <math.h>
+
+void blackbody_rayleigh_jeans_xyz(double temperature_K, int terms,
+                                  double xyz[3]) {
+  /* CSV-derived coefficients multiply T, 1, and T^-1. */
+  static const double coefficients[3][3] = {
+      {9.75239167492079072e3, -1.3438558214879372e8,
+       6.31051886434512943e11},
+      {9.51500076632225143e3, -1.25823411343233966e8,
+       5.58059336995288075e11},
+      {2.13884044303540753e4, -3.43402928472968325e8,
+       1.84249991169126495e12}};
+  xyz[0] = xyz[1] = xyz[2] = 0.0;
+  if (!isfinite(temperature_K) || temperature_K <= 0.0 || terms <= 0)
+    return;
+  if (terms > 3) terms = 3;
+  for (int channel = 0; channel < 3; ++channel) {
+    xyz[channel] = coefficients[channel][0] * temperature_K;
+    if (terms >= 2) xyz[channel] += coefficients[channel][1];
+    if (terms >= 3) xyz[channel] += coefficients[channel][2] / temperature_K;
+  }
+}
+
+static double chebyshev_clenshaw(const double coefficients[9], double x) {
+  double b1 = 0.0, b2 = 0.0;
+  for (int degree = 8; degree >= 1; --degree) {
+    const double b0 = 2.0 * x * b1 - b2 + coefficients[degree];
+    b2 = b1;
+    b1 = b0;
+  }
+  return coefficients[0] + x * b1 - b2;
+}
+
+int blackbody_low_hybrid_log_xyz(double temperature_K, double log_xyz[3]) {
+  static const double crossover_K = 4.09732109813541501;
+  static const double maximum_K = 670.146556;
+  static const double q_over_endpoint[3] = {
+      17334.66117474619, 17334.66117474619, 22135.028884675903};
+  static const double endpoint_delta[3] = {
+      6.27453428003507278e-6, 6.27523604457500139e-6,
+      9.72024116337713046e-5};
+  /* The 1 nm-linear endpoint makes these polynomials exactly cubic. Z has
+   * zero constant term because its 649--650 nm segment ends at zero. */
+  static const double endpoint[3][4] = {
+      {1.81136495194567529e-2, 6.61028930839653126e-5,
+       1.48916636199141013e-8, 1.27817061855715528e-12},
+      {6.54117161006302892e-3, 2.38707969128581198e-5,
+       5.37761358258185981e-9, 4.61567454126977889e-13},
+      {0.0, 1.18022769903543601e-3,
+       2.13277824065096828e-7, 1.44529622149769957e-11}};
+  static const double edges[7] = {
+      4.09732109813541501, 9.58175285991856125, 22.4073207028704393,
+      52.4004353296953668, 122.540559808645089, 286.566107776330601,
+      670.146556};
+  static const double coefficients[6][3][9] = {
+    {
+      {-2.19805342404174431,-0.427989370548201992,0.0457830231212280024,-0.00653024253368962181,0.00104792258851775496,-0.000179376495463268954,0.0000320194392828742251,-6.07285820395394865e-6,1.09768907792163728e-6},
+      {-3.21660305217802182,-0.427989317567608284,0.0457830153418609421,-0.00653024167298543334,0.00104792254862916667,-0.000179376508222333492,0.0000320194445897039005,-6.07285961285998572e-6,1.09768936710106915e-6},
+      {-3.15707121695679764,-0.842260091928597102,0.0900358213706993461,-0.0130901684838339702,0.00216325774923957118,-0.000378541811375747837,0.0000677130171872672182,-0.0000125764470117522098,2.20543446698214866e-6}
+    },{
+      {-1.31770949444839666,-0.441299064080369115,0.0486743388559153732,-0.00715733212557194619,0.00118377405373184402,-0.000208808548120484381,0.0000384098393642844184,-7.51938342275489625e-6,1.39706879054438949e-6},
+      {-2.33625922712095605,-0.441299020893064677,0.0486743393223627113,-0.00715733306978699914,0.00118377427098721058,-0.000208808583754120261,0.0000384098442083096037,-7.51938398617148892e-6,1.39706884379154304e-6},
+      {-1.41692623695155927,-0.878068580096058675,0.0982372095233778281,-0.0142171838603874451,0.00222410258776171213,-0.000361979109489253779,0.000060753034151493841,-0.0000107563963609356147,1.83055417864655942e-6}
+    },{
+      {-0.3921912858481657,-0.475886418801351195,0.0565482685596524812,-0.00894861708837536552,0.00159121925979185938,-0.000301479141093084493,0.0000595321929174356658,-0.0000125546605490984269,2.48503406235378882e-6},
+      {-1.41074105985953395,-0.475886424111238694,0.0565482743959804094,-0.00894861785536701427,0.00159121932381919629,-0.000301479143653568406,0.0000595321922909604336,-0.0000125546602334654497,2.48503397406341601e-6},
+      {0.405784497738078867,-0.915430672087484161,0.0971591394650644897,-0.0129050910771596898,0.00191184765152995853,-0.000308981746742567431,0.0000538107592092635484,-0.0000103283467419743846,1.93585445253475723e-6}
+    },{
+      {0.661994617272024548,-0.581350758628830698,0.0836597467165894418,-0.0158988753362009282,0.00336718167075529186,-0.000753662101271342622,0.000174585741711791468,-0.0000433048808785784171,9.75843572764750381e-6},
+      {-0.356555122545496873,-0.581350782933150466,0.0836597486771623816,-0.015898875146527317,0.00336718160135198675,-0.000753662094742459443,0.000174585746021139991,-0.0000433048853568368614,9.75843776186480067e-6},
+      {2.25724030901313405,-0.906149909784837897,0.0947967083167351412,-0.0147236528292987198,0.00294177345982300343,-0.000664178311762264752,0.000153812749558811817,-0.0000358188260289889306,7.26093914036807359e-6}
+    },{
+      {2.24517368371397051,-1.10137638803233863,0.260109466476170278,-0.0728062330999542422,0.0205487830596007146,-0.00547019170643925872,0.00127956756145661419,-0.000220004358600067039,0.0000138071701243527944},
+      {1.22677135347875279,-1.10165295376793399,0.260338122908468991,-0.0729732322401341771,0.0206568819007716045,-0.00553248986531515299,0.00131198334055904886,-0.000236511389192968018,0.0000196226106014253749},
+      {4.22881792094750399,-1.09273677796148737,0.167321326581346532,-0.0312233937525439919,0.00516074702621717587,-0.000696731180600597722,0.000080598581477044639,-0.000015356852062622951,5.09084046909121171e-6}
+    },{
+      {6.33932556236418281,-3.07589360106486531,0.456226630959950593,-0.00444896560449986224,-0.00803034195544535392,-0.000626248357061336987,0.000421665206117149112,0.0000469192966084690651,-0.0000234956663324828471},
+      {5.3740651556541423,-3.15361482486290811,0.488370588919019434,-0.0119565427661776158,-0.0071239164324135692,-0.000725343801153405533,0.000461709147315171223,0.0000420002316640812232,-0.0000256745877581467219},
+      {6.97759542644704118,-1.7029539661084809,0.314329784196094464,-0.0805402042898634484,0.0214126542337208624,-0.00404920923437737251,0.0000432174624337086157,0.000376147060473595442,-0.00015590604333367762}
+    }};
+
+  if (!isfinite(temperature_K) || temperature_K <= 0.0 ||
+      temperature_K > maximum_K)
+    return -1;
+  if (temperature_K <= crossover_K) {
+    for (int channel = 0; channel < 3; ++channel) {
+      double polynomial = endpoint[channel][3];
+      for (int term = 2; term >= 0; --term)
+        polynomial = endpoint[channel][term] + temperature_K * polynomial;
+      log_xyz[channel] = -q_over_endpoint[channel] / temperature_K +
+          log(temperature_K * polynomial) +
+          endpoint_delta[channel] * temperature_K / crossover_K;
+    }
+    return 0;
+  }
+
+  int segment = 0;
+  while (segment < 5 && temperature_K >= edges[segment + 1]) ++segment;
+  const double inverse_center =
+      0.5 * (1.0 / edges[segment] + 1.0 / edges[segment + 1]);
+  const double inverse_half =
+      0.5 * (1.0 / edges[segment] - 1.0 / edges[segment + 1]);
+  const double x = (1.0 / temperature_K - inverse_center) / inverse_half;
+  for (int channel = 0; channel < 3; ++channel)
+    log_xyz[channel] = -q_over_endpoint[channel] / temperature_K +
+        chebyshev_clenshaw(coefficients[segment][channel], x);
+  return 0;
+}
+
+int blackbody_low_hybrid_xyz(double temperature_K, double xyz[3]) {
+  double log_xyz[3];
+  if (blackbody_low_hybrid_log_xyz(temperature_K, log_xyz)) {
+    xyz[0] = xyz[1] = xyz[2] = 0.0;
+    return -1;
+  }
+  for (int channel = 0; channel < 3; ++channel)
+    xyz[channel] = exp(log_xyz[channel]);
+  return 0;
+}
+
+LinearRgb blackbody_xyz_to_linear_rgb(const double xyz[3]) {
+  return (LinearRgb){
+      fmax(0.0, 3.24096994 * xyz[0] - 1.53738318 * xyz[1] - 0.49861076 * xyz[2]),
+      fmax(0.0, -0.96924364 * xyz[0] + 1.87596750 * xyz[1] + 0.04155506 * xyz[2]),
+      fmax(0.0, 0.05563008 * xyz[0] - 0.20397696 * xyz[1] + 1.05697151 * xyz[2])};
+}

src/blackbody_internal.h

@@ -0,0 +1,12 @@
+#ifndef BLACKBODY_INTERNAL_H
+#define BLACKBODY_INTERNAL_H
+
+#include "optics.h"
+
+void blackbody_rayleigh_jeans_xyz(double temperature_K, int terms,
+                                  double xyz[3]);
+int blackbody_low_hybrid_log_xyz(double temperature_K, double log_xyz[3]);
+int blackbody_low_hybrid_xyz(double temperature_K, double xyz[3]);
+LinearRgb blackbody_xyz_to_linear_rgb(const double xyz[3]);
+
+#endif

src/blackbody_lut.c

@@ -0,0 +1,175 @@
+#include "blackbody_internal.h"
+
+#include <math.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <time.h>
+
+enum { BLACKBODY_LUT_VERSION_CIE1931 = 3,
+       BLACKBODY_LUT_ELEMENT_F64 = 1,
+       BLACKBODY_LUT_REFERENCE_CIE1931_1NM_LINEAR = 3 };
+
+static const char default_table_path[] =
+    "assets/blackbody/cie1931_2deg_xyz_1024.grbblut";
+
+typedef struct {
+  char magic[8];
+  uint32_t version, header_bytes, endian_marker, element_type;
+  uint32_t reference_type, wavelength_min_nm, wavelength_max_nm,
+      wavelength_step_nm;
+  uint64_t node_count;
+  double log_temperature_min, log_temperature_max;
+  double planck_h, planck_c, planck_k;
+  char generator[64];
+  uint64_t payload_checksum;
+} BlackbodyLutHeader;
+
+_Static_assert(sizeof(BlackbodyLutHeader) == 160,
+               "blackbody LUT header layout changed");
+
+typedef struct {
+  double *xyz;
+  size_t nodes;
+  double log_min, log_max, index_scale;
+  int ready;
+} BlackbodyLut;
+
+static BlackbodyLut table;
+
+static uint64_t fnv1a64(const void *data, size_t bytes) {
+  const unsigned char *p = data;
+  uint64_t hash = UINT64_C(14695981039346656037);
+  for (size_t i = 0; i < bytes; ++i) {
+    hash ^= p[i];
+    hash *= UINT64_C(1099511628211);
+  }
+  return hash;
+}
+
+static int validate_header(const BlackbodyLutHeader *h) {
+  return memcmp(h->magic, "GRBBLUT3", 8) == 0 &&
+         h->version == BLACKBODY_LUT_VERSION_CIE1931 &&
+         h->header_bytes == sizeof *h &&
+         h->endian_marker == UINT32_C(0x01020304) &&
+         h->element_type == BLACKBODY_LUT_ELEMENT_F64 &&
+         h->reference_type == BLACKBODY_LUT_REFERENCE_CIE1931_1NM_LINEAR &&
+         h->wavelength_min_nm == 360 && h->wavelength_max_nm == 830 &&
+         h->wavelength_step_nm == 1 && h->node_count >= 4 &&
+         h->node_count <= SIZE_MAX / (3 * sizeof(double)) &&
+         isfinite(h->log_temperature_min) &&
+         isfinite(h->log_temperature_max) &&
+         h->log_temperature_max > h->log_temperature_min &&
+         h->planck_h == 6.62607015e-34 && h->planck_c == 299792458.0 &&
+         h->planck_k == 1.380649e-23;
+}
+
+static int read_table(const char *path) {
+  FILE *file = fopen(path, "rb");
+  BlackbodyLutHeader h;
+  if (file == NULL || fread(&h, sizeof h, 1, file) != 1 ||
+      !validate_header(&h)) {
+    if (file != NULL) fclose(file);
+    return -1;
+  }
+  double *xyz = malloc(3 * (size_t)h.node_count * sizeof *xyz);
+  const size_t count = 3 * (size_t)h.node_count;
+  unsigned char extra;
+  if (xyz == NULL || fread(xyz, sizeof *xyz, count, file) != count ||
+      fread(&extra, 1, 1, file) != 0 || ferror(file) ||
+      fnv1a64(xyz, count * sizeof *xyz) != h.payload_checksum) {
+    free(xyz); fclose(file); return -1;
+  }
+  fclose(file);
+  table = (BlackbodyLut){.xyz = xyz, .nodes = (size_t)h.node_count,
+                         .log_min = h.log_temperature_min,
+                         .log_max = h.log_temperature_max, .ready = 1};
+  table.index_scale = (table.nodes - 1) / (table.log_max - table.log_min);
+  return 0;
+}
+
+int blackbody_backend_init(const char *table_path, size_t lut_nodes,
+                           double temperature_min_K,
+                           double temperature_max_K,
+                           const char *write_table_path, FILE *report) {
+  struct timespec start, finish;
+  timespec_get(&start, TIME_UTC);
+  if (table.ready) return -1;
+  if (lut_nodes != 0 || isfinite(temperature_min_K) ||
+      isfinite(temperature_max_K) || write_table_path != NULL)
+    return -1;
+  const char *path = table_path != NULL ? table_path : default_table_path;
+  if (read_table(path)) return -1;
+  timespec_get(&finish, TIME_UTC);
+  if (report != NULL)
+    fprintf(report,
+            "Blackbody LUT: %zu CIE 1931 2-deg 1 nm-linear XYZ nodes, "
+            "T=[%.9g, %.9g] K, loaded %s in %.6f s; "
+            "payload fnv1a64=%016llx\n",
+            table.nodes,
+            exp(table.log_min), exp(table.log_max),
+            path,
+            finish.tv_sec - start.tv_sec +
+                1e-9 * (finish.tv_nsec - start.tv_nsec),
+            (unsigned long long)fnv1a64(table.xyz,
+                                       3 * table.nodes * sizeof *table.xyz));
+  return 0;
+}
+
+void blackbody_backend_destroy(void) {
+  free(table.xyz);
+  table = (BlackbodyLut){0};
+}
+
+const char *blackbody_backend_name(void) { return "lut"; }
+
+LinearRgb blackbody_to_linear_rgb(double temperature_K) {
+  if (!table.ready || !isfinite(temperature_K) || temperature_K <= 0.0) {
+    return (LinearRgb){0};
+  }
+  const double log_t = log(temperature_K);
+  if (log_t < table.log_min) {
+    double xyz[3];
+    if (blackbody_low_hybrid_xyz(temperature_K, xyz))
+      return (LinearRgb){0};
+    return blackbody_xyz_to_linear_rgb(xyz);
+  }
+  if (log_t > table.log_max) {
+    double xyz[3];
+    blackbody_rayleigh_jeans_xyz(temperature_K, 3, xyz);
+    return blackbody_xyz_to_linear_rgb(xyz);
+  }
+  const double position = (log_t - table.log_min) * table.index_scale;
+  size_t low = (size_t)position;
+  if (low >= table.nodes - 1) low = table.nodes - 2;
+  const double fraction = position - low;
+  double xyz[3];
+  {
+    size_t base;
+    double x;
+    if (low == 0) {
+      base = 0;
+      x = fraction;
+    } else if (low >= table.nodes - 2) {
+      base = table.nodes - 4;
+      x = 2.0 + fraction;
+    } else {
+      base = low - 1;
+      x = 1.0 + fraction;
+    }
+    const double weights[4] = {
+        -(x - 1.0) * (x - 2.0) * (x - 3.0) / 6.0,
+        x * (x - 2.0) * (x - 3.0) / 2.0,
+        -x * (x - 1.0) * (x - 3.0) / 2.0,
+        x * (x - 1.0) * (x - 2.0) / 6.0};
+    for (int channel = 0; channel < 3; ++channel) {
+      xyz[channel] = 0.0;
+      for (int point = 0; point < 4; ++point)
+        xyz[channel] += weights[point] *
+                        table.xyz[3 * (base + (size_t)point) + channel];
+      xyz[channel] = fmax(0.0, xyz[channel]);
+    }
+  }
+  return blackbody_xyz_to_linear_rgb(xyz);
+}

src/dummy_psf.c

@@ -0,0 +1,309 @@
+#include "dummy_psf.h"
+
+#include <math.h>
+#include <omp.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+enum { DUMMY_SELECTOR_TILE = 32 };
+
+typedef struct {
+  size_t worker, events, triangles, occupied_tiles;
+  double event_bbox_diagonal, triangle_bbox_diagonal;
+  double maximum_triangle_jump;
+  int adaptive_tile16;
+} DummyChunkSample;
+
+struct DummyPsfSink {
+  DummyChunkSample *samples;
+  size_t count, capacity, event_capacity;
+  size_t total_events;
+  int width, height, tiles_x, tiles_y;
+  int failed;
+  omp_lock_t lock;
+};
+
+struct DummyPsfChunk {
+  DummyPsfSink *sink;
+  uint32_t *tile_generation;
+  uint32_t generation;
+  size_t worker, events, triangles, occupied_tiles;
+  size_t last_triangle;
+  int have_event, have_triangle;
+  double event_min_x, event_max_x, event_min_y, event_max_y;
+  double triangle_min_x, triangle_max_x, triangle_min_y, triangle_max_y;
+  double last_triangle_x, last_triangle_y, maximum_triangle_jump;
+};
+
+static int append_sample(DummyPsfSink *sink, DummyChunkSample sample) {
+  int result = 0;
+  omp_set_lock(&sink->lock);
+  if (sink->failed) {
+    result = -1;
+  } else if (sink->count == sink->capacity) {
+    const size_t capacity = sink->capacity ? 2 * sink->capacity : 1024;
+    DummyChunkSample *samples =
+        capacity < sink->capacity || capacity > SIZE_MAX / sizeof *samples
+                                    ? NULL
+                                    : realloc(sink->samples,
+                                              capacity * sizeof *samples);
+    if (samples == NULL) {
+      sink->failed = 1;
+      result = -1;
+    } else {
+      sink->samples = samples;
+      sink->capacity = capacity;
+    }
+  }
+  if (!result) {
+    sink->samples[sink->count++] = sample;
+    sink->total_events += sample.events;
+  }
+  omp_unset_lock(&sink->lock);
+  return result;
+}
+
+int dummy_psf_sink_create(DummyPsfSink **out, int width, int height,
+                          size_t event_capacity) {
+  if (out == NULL || width <= 0 || height <= 0 || event_capacity == 0)
+    return -1;
+  DummyPsfSink *sink = calloc(1, sizeof *sink);
+  if (sink == NULL)
+    return -1;
+  sink->width = width;
+  sink->height = height;
+  sink->tiles_x = (width + DUMMY_SELECTOR_TILE - 1) / DUMMY_SELECTOR_TILE;
+  sink->tiles_y = (height + DUMMY_SELECTOR_TILE - 1) / DUMMY_SELECTOR_TILE;
+  sink->event_capacity = event_capacity;
+  omp_init_lock(&sink->lock);
+  *out = sink;
+  return 0;
+}
+
+int dummy_psf_chunk_create(DummyPsfChunk **out, DummyPsfSink *sink,
+                           size_t worker_id) {
+  if (out == NULL || sink == NULL)
+    return -1;
+  DummyPsfChunk *chunk = calloc(1, sizeof *chunk);
+  if (chunk == NULL)
+    return -1;
+  const size_t tile_count = (size_t)sink->tiles_x * sink->tiles_y;
+  chunk->tile_generation = calloc(tile_count, sizeof *chunk->tile_generation);
+  if (chunk->tile_generation == NULL) {
+    free(chunk);
+    return -1;
+  }
+  chunk->sink = sink;
+  chunk->worker = worker_id;
+  chunk->generation = 1;
+  *out = chunk;
+  return 0;
+}
+
+static void reset_chunk(DummyPsfChunk *chunk) {
+  chunk->events = 0;
+  chunk->triangles = 0;
+  chunk->occupied_tiles = 0;
+  chunk->have_event = 0;
+  chunk->have_triangle = 0;
+  chunk->maximum_triangle_jump = 0.0;
+  if (++chunk->generation == 0) {
+    const size_t count = (size_t)chunk->sink->tiles_x * chunk->sink->tiles_y;
+    memset(chunk->tile_generation, 0, count * sizeof *chunk->tile_generation);
+    chunk->generation = 1;
+  }
+}
+
+int dummy_psf_chunk_flush(DummyPsfChunk *chunk) {
+  if (chunk == NULL)
+    return -1;
+  if (chunk->events == 0)
+    return 0;
+  const double event_diagonal =
+      hypot(chunk->event_max_x - chunk->event_min_x,
+            chunk->event_max_y - chunk->event_min_y);
+  const double triangle_diagonal =
+      hypot(chunk->triangle_max_x - chunk->triangle_min_x,
+            chunk->triangle_max_y - chunk->triangle_min_y);
+  const DummyChunkSample sample = {
+      .worker = chunk->worker,
+      .events = chunk->events,
+      .triangles = chunk->triangles,
+      .occupied_tiles = chunk->occupied_tiles,
+      .event_bbox_diagonal = event_diagonal,
+      .triangle_bbox_diagonal = triangle_diagonal,
+      .maximum_triangle_jump = chunk->maximum_triangle_jump,
+      .adaptive_tile16 = chunk->events >= 8192 && chunk->occupied_tiles != 0 &&
+                         chunk->events / chunk->occupied_tiles >= 32};
+  const int result = append_sample(chunk->sink, sample);
+  reset_chunk(chunk);
+  return result;
+}
+
+int dummy_psf_chunk_emit(DummyPsfChunk *chunk, const PsfCachedEvent *event,
+                         size_t triangle, double triangle_center_x,
+                         double triangle_center_y) {
+  if (chunk == NULL || event == NULL)
+    return -1;
+  if (chunk->events == chunk->sink->event_capacity &&
+      dummy_psf_chunk_flush(chunk))
+    return -1;
+
+  if (!chunk->have_event) {
+    chunk->event_min_x = chunk->event_max_x = event->x;
+    chunk->event_min_y = chunk->event_max_y = event->y;
+    chunk->have_event = 1;
+  } else {
+    chunk->event_min_x = fmin(chunk->event_min_x, event->x);
+    chunk->event_max_x = fmax(chunk->event_max_x, event->x);
+    chunk->event_min_y = fmin(chunk->event_min_y, event->y);
+    chunk->event_max_y = fmax(chunk->event_max_y, event->y);
+  }
+
+  const int pixel_x = (int)floor(event->x);
+  const int pixel_y = (int)floor(event->y);
+  if (pixel_x >= 0 && pixel_x < chunk->sink->width && pixel_y >= 0 &&
+      pixel_y < chunk->sink->height) {
+    const size_t tile = (size_t)(pixel_y / DUMMY_SELECTOR_TILE) *
+                            chunk->sink->tiles_x +
+                        pixel_x / DUMMY_SELECTOR_TILE;
+    if (chunk->tile_generation[tile] != chunk->generation) {
+      chunk->tile_generation[tile] = chunk->generation;
+      ++chunk->occupied_tiles;
+    }
+  }
+
+  if (!chunk->have_triangle || triangle != chunk->last_triangle) {
+    if (!chunk->have_triangle) {
+      chunk->triangle_min_x = chunk->triangle_max_x = triangle_center_x;
+      chunk->triangle_min_y = chunk->triangle_max_y = triangle_center_y;
+      chunk->have_triangle = 1;
+    } else {
+      chunk->triangle_min_x = fmin(chunk->triangle_min_x, triangle_center_x);
+      chunk->triangle_max_x = fmax(chunk->triangle_max_x, triangle_center_x);
+      chunk->triangle_min_y = fmin(chunk->triangle_min_y, triangle_center_y);
+      chunk->triangle_max_y = fmax(chunk->triangle_max_y, triangle_center_y);
+      chunk->maximum_triangle_jump =
+          fmax(chunk->maximum_triangle_jump,
+               hypot(triangle_center_x - chunk->last_triangle_x,
+                     triangle_center_y - chunk->last_triangle_y));
+    }
+    chunk->last_triangle = triangle;
+    chunk->last_triangle_x = triangle_center_x;
+    chunk->last_triangle_y = triangle_center_y;
+    ++chunk->triangles;
+  }
+  ++chunk->events;
+  return 0;
+}
+
+void dummy_psf_chunk_destroy(DummyPsfChunk *chunk) {
+  if (chunk == NULL)
+    return;
+  (void)dummy_psf_chunk_flush(chunk);
+  free(chunk->tile_generation);
+  free(chunk);
+}
+
+static int compare_double(const void *left, const void *right) {
+  const double a = *(const double *)left;
+  const double b = *(const double *)right;
+  return (a > b) - (a < b);
+}
+
+static double percentile(const double *values, size_t count, double fraction) {
+  if (count == 0)
+    return 0.0;
+  const double position = fraction * (count - 1);
+  const size_t low = (size_t)floor(position);
+  const size_t high = low + 1 < count ? low + 1 : low;
+  return values[low] + (values[high] - values[low]) * (position - low);
+}
+
+static void report_group(const DummyPsfSink *sink, int full_only,
+                         const char *label) {
+  size_t count = 0, adaptive = 0;
+  for (size_t i = 0; i < sink->count; ++i)
+    if (!full_only || sink->samples[i].events == sink->event_capacity)
+      ++count;
+  if (count == 0) {
+    fprintf(stderr, "Dummy PSF %s chunks: none\n", label);
+    return;
+  }
+  double *events = count > SIZE_MAX / (6 * sizeof *events)
+                       ? NULL
+                       : malloc(6 * count * sizeof *events);
+  if (events == NULL) {
+    fputs("Dummy PSF report allocation failed\n", stderr);
+    return;
+  }
+  double *triangles = events + count;
+  double *tiles = triangles + count;
+  double *event_distance = tiles + count;
+  double *triangle_distance = event_distance + count;
+  double *maximum_jump = triangle_distance + count;
+  size_t next = 0;
+  for (size_t i = 0; i < sink->count; ++i) {
+    const DummyChunkSample *sample = &sink->samples[i];
+    if (full_only && sample->events != sink->event_capacity)
+      continue;
+    events[next] = sample->events;
+    triangles[next] = sample->triangles;
+    tiles[next] = sample->occupied_tiles;
+    event_distance[next] = sample->event_bbox_diagonal;
+    triangle_distance[next] = sample->triangle_bbox_diagonal;
+    maximum_jump[next] = sample->maximum_triangle_jump;
+    adaptive += sample->adaptive_tile16;
+    ++next;
+  }
+  double *series[] = {events, triangles, tiles, event_distance,
+                      triangle_distance, maximum_jump};
+  const char *names[] = {"events", "distinct_triangles", "occupied_32px_tiles",
+                         "event_bbox_diagonal_px",
+                         "triangle_centroid_bbox_diagonal_px",
+                         "maximum_successive_triangle_jump_px"};
+  for (size_t field = 0; field < 6; ++field) {
+    qsort(series[field], count, sizeof **series, compare_double);
+    fprintf(stderr,
+            "Dummy PSF %s %s: min=%.3f p10=%.3f p25=%.3f p50=%.3f "
+            "p75=%.3f p90=%.3f p99=%.3f max=%.3f\n",
+            label, names[field], series[field][0],
+            percentile(series[field], count, 0.10),
+            percentile(series[field], count, 0.25),
+            percentile(series[field], count, 0.50),
+            percentile(series[field], count, 0.75),
+            percentile(series[field], count, 0.90),
+            percentile(series[field], count, 0.99), series[field][count - 1]);
+  }
+  fprintf(stderr,
+          "Dummy PSF %s adaptive criterion: %zu/%zu chunks tile16-eligible "
+          "(events>=8192 and events/occupied_32px_tiles>=32)\n",
+          label, adaptive, count);
+  free(events);
+}
+
+void dummy_psf_sink_report(const DummyPsfSink *sink) {
+  if (sink == NULL)
+    return;
+  size_t full = 0;
+  for (size_t i = 0; i < sink->count; ++i)
+    full += sink->samples[i].events == sink->event_capacity;
+  fprintf(stderr,
+          "Dummy PSF diagnostic only: no HDR or image was accumulated/written.\n"
+          "Dummy PSF chunks: total=%zu full=%zu partial=%zu events=%zu "
+          "capacity=%zu selector_tile=%dpx\n",
+          sink->count, full, sink->count - full, sink->total_events,
+          sink->event_capacity, DUMMY_SELECTOR_TILE);
+  report_group(sink, 0, "all");
+  report_group(sink, 1, "full");
+}
+
+void dummy_psf_sink_destroy(DummyPsfSink *sink) {
+  if (sink == NULL)
+    return;
+  omp_destroy_lock(&sink->lock);
+  free(sink->samples);
+  free(sink);
+}

src/dummy_psf.h

@@ -0,0 +1,25 @@
+#ifndef DUMMY_PSF_H
+#define DUMMY_PSF_H
+
+#include "optics.h"
+
+#include <stddef.h>
+
+typedef struct DummyPsfSink DummyPsfSink;
+typedef struct DummyPsfChunk DummyPsfChunk;
+
+/* Diagnostic-only consumer for the production PSF event stream. It preserves
+ * per-worker chunk boundaries but never accumulates or writes an HDR image. */
+int dummy_psf_sink_create(DummyPsfSink **out, int width, int height,
+                          size_t event_capacity);
+int dummy_psf_chunk_create(DummyPsfChunk **out, DummyPsfSink *sink,
+                           size_t worker_id);
+int dummy_psf_chunk_emit(DummyPsfChunk *chunk, const PsfCachedEvent *event,
+                         size_t triangle, double triangle_center_x,
+                         double triangle_center_y);
+int dummy_psf_chunk_flush(DummyPsfChunk *chunk);
+void dummy_psf_chunk_destroy(DummyPsfChunk *chunk);
+void dummy_psf_sink_report(const DummyPsfSink *sink);
+void dummy_psf_sink_destroy(DummyPsfSink *sink);
+
+#endif

src/frame.c

@@ -3,6 +3,9 @@
 #ifdef PSF_BACKEND_HIP
 #include "hip_psf.h"
 #endif
+#ifdef PSF_BACKEND_DUMMY
+#include "dummy_psf.h"
+#endif
 
 #include "optics.h"
 
@@ -138,6 +141,11 @@ typedef struct {
   double submit_seconds;
   char hip_message[256];
   HipPsfTiming hip_timing;
+#endif
+#ifdef PSF_BACKEND_DUMMY
+  DummyPsfSink *dummy;
+  DummyPsfChunk *dummy_chunk;
+  int borrowed_dummy;
 #endif
   int failed;
 } PsfEventSink;
@@ -155,6 +163,16 @@ static int psf_event_sink_init(PsfEventSink *sink, double *hdr, int width,
   *sink = (PsfEventSink){.hdr = hdr, .width = width, .height = height,
                          .cache = cache};
   if (cache != NULL) {
+#ifdef PSF_BACKEND_DUMMY
+    if (dummy_psf_sink_create(&sink->dummy, width, height,
+                              PSF_EVENT_SINK_CAPACITY)) {
+      fputs("Dummy PSF backend initialization failed\n", stderr);
+      dummy_psf_sink_destroy(sink->dummy);
+      sink->dummy = NULL;
+      sink->failed = 1;
+      return -1;
+    }
+#else
     sink->events = malloc(PSF_EVENT_SINK_CAPACITY * sizeof *sink->events);
     if (sink->events == NULL) {
 #ifdef PSF_BACKEND_HIP
@@ -174,6 +192,7 @@ static int psf_event_sink_init(PsfEventSink *sink, double *hdr, int width,
       sink->events = NULL;
       return -1;
     }
+#endif
 #endif
   }
   return 0;
@@ -182,6 +201,10 @@ static int psf_event_sink_init(PsfEventSink *sink, double *hdr, int width,
 static int psf_event_sink_flush_unlocked(PsfEventSink *sink) {
   if (sink->failed)
     return -1;
+#ifdef PSF_BACKEND_DUMMY
+  if (sink->dummy_chunk != NULL)
+    return dummy_psf_chunk_flush(sink->dummy_chunk);
+#endif
 #ifdef PSF_BACKEND_HIP
   if (sink->hip != NULL) {
     if (hip_psf_sink_submit(sink->hip, sink->events, sink->count,
@@ -233,6 +256,7 @@ static int psf_event_sink_finish_for_cpu(PsfEventSink *sink) {
   return 0;
 }
 
+#ifndef PSF_BACKEND_DUMMY
 static int psf_event_sink_resume_gpu(PsfEventSink *sink) {
 #ifdef PSF_BACKEND_HIP
   if (sink->hip != NULL && hip_psf_sink_load_hdr(sink->hip, sink->hdr,
@@ -246,8 +270,22 @@ static int psf_event_sink_resume_gpu(PsfEventSink *sink) {
 #endif
   return 0;
 }
+#endif
 
 static int psf_event_sink_destroy(PsfEventSink *sink) {
+#ifdef PSF_BACKEND_DUMMY
+  int dummy_result = sink->dummy_chunk != NULL
+                         ? dummy_psf_chunk_flush(sink->dummy_chunk)
+                         : 0;
+  dummy_psf_chunk_destroy(sink->dummy_chunk);
+  sink->dummy_chunk = NULL;
+  if (!sink->borrowed_dummy) {
+    dummy_psf_sink_report(sink->dummy);
+    dummy_psf_sink_destroy(sink->dummy);
+    sink->dummy = NULL;
+  }
+  return dummy_result;
+#endif
 #ifdef PSF_BACKEND_HIP
   if (sink->borrowed_hip) {
     const int result = psf_event_sink_flush(sink);
@@ -267,10 +305,22 @@ static int psf_event_sink_destroy(PsfEventSink *sink) {
   return result;
 }
 
-#if FRAME_PSF_EVENT_SINK || defined(PSF_BACKEND_HIP)
-static void psf_event_sink_emit(PsfEventSink *sink, const PsfCachedEvent *event) {
+#if FRAME_PSF_EVENT_SINK || defined(PSF_BACKEND_HIP) || defined(PSF_BACKEND_DUMMY)
+static void psf_event_sink_emit(PsfEventSink *sink, const PsfCachedEvent *event,
+                                size_t triangle, double triangle_center_x,
+                                double triangle_center_y) {
   if (sink->failed)
     return;
+#ifdef PSF_BACKEND_DUMMY
+  if (dummy_psf_chunk_emit(sink->dummy_chunk, event, triangle,
+                           triangle_center_x, triangle_center_y))
+    sink->failed = 1;
+  return;
+#else
+  (void)triangle;
+  (void)triangle_center_x;
+  (void)triangle_center_y;
+#endif
   if (sink->events == NULL) {
     splat_prepared_cached_event(sink->hdr, sink->width, sink->height, event, sink->cache);
     return;
@@ -1129,6 +1179,9 @@ static int splat_catalog_tile(const Star *stars, size_t count,
 #endif
     if (direct_fallback == 1) {
       /* Exclude every other submit and fallback until HDR is back on device. */
+#ifdef PSF_BACKEND_DUMMY
+      if (psf_event_sink_flush(context->event_sink)) return -1;
+#else
 #ifdef PSF_BACKEND_HIP
       const double submit_start = omp_get_wtime();
       if (context->event_sink->hip_lock)
@@ -1147,9 +1200,17 @@ static int splat_catalog_tile(const Star *stars, size_t count,
       context->event_sink->submit_seconds += omp_get_wtime() - submit_start;
 #endif
       if (failed) return -1;
+#endif
     } else if (direct_fallback != 3) {
-#if FRAME_PSF_EVENT_SINK || defined(PSF_BACKEND_HIP)
-      psf_event_sink_emit(context->event_sink, &event);
+#if FRAME_PSF_EVENT_SINK || defined(PSF_BACKEND_HIP) || defined(PSF_BACKEND_DUMMY)
+      const double triangle_center_x =
+          (context->vertex[0]->image_x + context->vertex[1]->image_x +
+           context->vertex[2]->image_x) / 3.0;
+      const double triangle_center_y =
+          (context->vertex[0]->image_y + context->vertex[1]->image_y +
+           context->vertex[2]->image_y) / 3.0;
+      psf_event_sink_emit(context->event_sink, &event, context->triangle_index,
+                          triangle_center_x, triangle_center_y);
 #else
       splat_prepared_cached_event(context->hdr, context->width, context->height,
                                   &event, context->psf_cache);
@@ -1319,6 +1380,83 @@ size_t frame_splat_catalog(const FrameLensMesh *mesh,
     progress->callback(progress->context, FRAME_SPLAT_PROGRESS_BEGIN, 0,
                        mesh->triangle_count);
 
+#ifdef PSF_BACKEND_DUMMY
+  PsfEventSink owner;
+  if (psf_event_sink_init(&owner, hdr, width, height, psf_cache)) return SIZE_MAX;
+  size_t dummy_images = 0, dummy_direct = 0, dummy_clipped = 0,
+         dummy_discarded = 0;
+  int dummy_failed = 0, dummy_workers = 0;
+  const double dummy_start = omp_get_wtime();
+#ifdef GR_DEBUG
+  double dummy_max_magnification = 0.0;
+  size_t dummy_clamped = 0;
+#pragma omp parallel reduction(+ : dummy_images, dummy_direct, dummy_clipped, dummy_discarded, dummy_clamped) reduction(max : dummy_failed, dummy_max_magnification)
+#else
+#pragma omp parallel reduction(+ : dummy_images, dummy_direct, dummy_clipped, dummy_discarded) reduction(max : dummy_failed)
+#endif
+  {
+#pragma omp single
+    dummy_workers = omp_get_num_threads();
+    const size_t worker = (size_t)omp_get_thread_num();
+    PsfEventSink local = {.hdr = hdr, .width = width, .height = height,
+        .cache = psf_cache, .dummy = owner.dummy, .borrowed_dummy = 1};
+    if (dummy_psf_chunk_create(&local.dummy_chunk, owner.dummy, worker)) {
+      fputs("Dummy PSF worker chunk initialization failed\n", stderr);
+      local.failed = 1;
+    }
+    size_t completed = 0, next_report = 8;
+    if (progress && progress->worker_callback)
+      progress->worker_callback(progress->context, worker,
+                                (size_t)dummy_workers, 0, 0);
+#pragma omp for schedule(dynamic, 1) nowait
+    for (size_t t = 0; t < mesh->triangle_count; ++t) {
+      if (local.failed) continue;
+      const CatalogSplatStats stats = splat_catalog_triangles(
+          mesh, catalog, hdr, width, height, exposure, psf, psf_cache,
+          max_magnification, max_cache_psf_flux, psf_relative_tail, psf_min_y,
+          t, t + 1, &local);
+      dummy_images += stats.images;
+      dummy_direct += stats.direct_fallbacks;
+      dummy_clipped += stats.cached_wing_clipped;
+      dummy_discarded += stats.discarded_below_min_y;
+      dummy_failed |= stats.failed;
+#ifdef GR_DEBUG
+      dummy_max_magnification = fmax(dummy_max_magnification,
+                                     stats.max_raw_magnification);
+      dummy_clamped += stats.magnification_clamped_triangles;
+#endif
+      ++completed;
+      if (progress && progress->worker_callback && completed == next_report) {
+        progress->worker_callback(progress->context, worker,
+                                  (size_t)dummy_workers, completed, 0);
+        if (next_report <= SIZE_MAX / 2) next_report *= 2;
+      }
+    }
+    if (psf_event_sink_destroy(&local)) dummy_failed = 1;
+    if (progress && progress->worker_callback)
+      progress->worker_callback(progress->context, worker,
+                                (size_t)dummy_workers, completed, 1);
+  }
+  if (psf_event_sink_destroy(&owner)) dummy_failed = 1;
+  fprintf(stderr,
+          "Dummy PSF producers: %d workers; classification/chunk wall %.3f s\n",
+          dummy_workers, omp_get_wtime() - dummy_start);
+  copy_psf_splat_stats(psf_stats, (CatalogSplatStats){
+      .images = dummy_images, .direct_fallbacks = dummy_direct,
+      .cached_wing_clipped = dummy_clipped,
+      .discarded_below_min_y = dummy_discarded,
+#ifdef GR_DEBUG
+      .max_raw_magnification = dummy_max_magnification,
+      .magnification_clamped_triangles = dummy_clamped,
+#endif
+  });
+  if (dummy_failed) return SIZE_MAX;
+  if (progress != NULL && progress->callback != NULL)
+    progress->callback(progress->context, FRAME_SPLAT_PROGRESS_END,
+                       mesh->triangle_count, mesh->triangle_count);
+  return dummy_images;
+#endif
+
 #ifdef PSF_BACKEND_HIP
   /* CPU workers own only bounded event chunks. A single device cache/HDR is
    * borrowed under a coarse submission lock, never duplicated per worker. */

src/main.c

@@ -12,6 +12,7 @@
 #include <math.h>
 #include <omp.h>
 #include <stdio.h>
+#include <stdint.h>
 #include <stdlib.h>
 #include <string.h>
 #include <sys/stat.h>
@@ -52,6 +53,7 @@ typedef struct {
   double slab_duration;
   double minkowski_proper_acceleration;
   int catalog_load_workers;
+  const char *blackbody_table_path;
   RefinementConfig refinement;
 } Settings;
 
@@ -319,6 +321,8 @@ static int parse_args(int argc, char **argv, Settings *s,
                                   &s->minkowski_proper_acceleration)) {
     } else if (!strcmp(argv[i], "--catalog-load-workers") && i + 1 < argc &&
                !parse_int(argv[++i], &s->catalog_load_workers)) {
+    } else if (!strcmp(argv[i], "--blackbody-table") && i + 1 < argc) {
+      s->blackbody_table_path = argv[++i];
     } else if (!strcmp(argv[i], "--write-minkowski-accel-track") &&
                i + 1 < argc)
       s->write_minkowski_accel_track_path = argv[++i];
@@ -377,6 +381,10 @@ static void print_help(const char *program) {
         "  --psf-min-y Y                  Skip events below this linear HDR luminance (default: 0, disabled)\n"
         "  --psf-direct                   Disable the PSF lookup cache (default: disabled)\n"
         "  --catalog-load-workers N       All-sky catalog loader workers (default: 4)\n"
+        "  --blackbody-table FILE         Explicit GRBBLUT3 table\n"
+        "                                (default: assets/blackbody/cie1931_2deg_xyz_1024.grbblut)\n",
+        stdout);
+  fputs(
         "\nAdaptive lens mesh:\n"
         "  --coarse-cell-pixels N         Initial mesh cell size in pixels (default: 16)\n"
         "  --refine-max-level N           Maximum refinement level (default: 0)\n"
@@ -445,8 +453,13 @@ static void report_splat_progress(void *context, FrameSplatProgressStage stage,
             progress->frame_id, total);
     break;
   case FRAME_SPLAT_PROGRESS_END:
+#ifdef PSF_BACKEND_DUMMY
+    fprintf(stderr, "Frame %zu: catalog classification finished in %.1f s; reporting chunk statistics...\n",
+            progress->frame_id, omp_get_wtime() - progress->splat_start);
+#else
     fprintf(stderr, "Frame %zu: catalog splatting finished in %.1f s; writing image...\n",
             progress->frame_id, omp_get_wtime() - progress->splat_start);
+#endif
     break;
   }
 }
@@ -922,7 +935,15 @@ static int render_lens_map(const Settings *s, StarCatalog *catalog) {
       }
       output_path = movie_path;
     }
+#ifdef PSF_BACKEND_DUMMY
+    (void)output_path;
+#endif
+#ifdef PSF_BACKEND_DUMMY
+    /* The dummy backend classifies real events but never touches pixels. */
+    double *hdr = calloc(1, sizeof *hdr);
+#else
     double *hdr = calloc((size_t)map.width * map.height * 3, sizeof *hdr);
+#endif
     if (hdr == NULL) { result = -1; break; }
     CatalogPrefetchStats prefetch = {0};
     PsfSplatStats psf_stats = {0};
@@ -945,7 +966,12 @@ static int render_lens_map(const Settings *s, StarCatalog *catalog) {
       break;
     }
     if (s->draw_mesh)
+#ifdef PSF_BACKEND_DUMMY
+      fputs("Dummy PSF backend ignores --draw-mesh.\n", stderr);
+#else
       frame_draw_mesh(&map.frames[i].mesh, hdr, map.width, map.height, 0.5, 0.5);
+#endif
+#ifndef PSF_BACKEND_DUMMY
 #ifdef ENABLE_HDR_OUTPUT
     if (s->write_hdr_output &&
         write_hdr_fits(s->hdr_output_path, hdr, map.width, map.height,
@@ -960,6 +986,13 @@ static int render_lens_map(const Settings *s, StarCatalog *catalog) {
             images, catalog->count, output_path, write_result == 0 ? "ok" : "write failed");
     psf_kernel_cache_report(&s->psf_cache, &psf_stats, stderr);
     if (write_result) { result = -1; break; }
+#else
+    free(hdr);
+    fprintf(stderr,
+            "Dummy PSF classified %zu images from %zu catalog stars; no HDR, PNG, or PPM was written.\n",
+            images, catalog->count);
+    psf_kernel_cache_report(&s->psf_cache, &psf_stats, stderr);
+#endif
   }
   lens_map_destroy(&map);
   return result;
@@ -1026,6 +1059,20 @@ int main(int argc, char **argv) {
     fputs("--lens-map-input and --lens-map-output are mutually exclusive.\n", stderr);
     return 2;
   }
+#ifdef PSF_BACKEND_DUMMY
+  if (settings.lens_map_input_path == NULL) {
+    fputs("The dummy PSF backend requires --lens-map-input and never traces or writes images.\n",
+          stderr);
+    return 2;
+  }
+  if (settings.psf_direct) {
+    fputs("The dummy PSF backend measures cache-event chunks and does not support --psf-direct.\n",
+          stderr);
+    return 2;
+  }
+  fputs("Dummy PSF backend: --output and --hdr-output are accepted for command parity but no image files will be written.\n",
+        stderr);
+#endif
 #ifdef ENABLE_HDR_OUTPUT
   if (settings.frames_dir != NULL && settings.write_hdr_output) {
     fputs("--hdr-output is available only for a single-frame render.\n", stderr);
@@ -1072,14 +1119,37 @@ int main(int argc, char **argv) {
     }
     fprintf(stderr, "Created test catalog: %s\n", settings.catalog_path);
   }
+  if (blackbody_backend_init(settings.blackbody_table_path,
+                             0, NAN, NAN, NULL, stderr)) {
+    fprintf(stderr,
+            "Blackbody backend '%s' initialization failed; the LUT loads the "
+            "repository CIE GRBBLUT3 table by default or a valid explicit "
+            "table.\n",
+            blackbody_backend_name());
+    catalog_destroy(&catalog);
+    spacetime_destroy(&spacetime);
+    return 1;
+  }
+  fprintf(stderr, "Blackbody backend: %s\n", blackbody_backend_name());
   if (!settings.psf_direct && psf_kernel_cache_init(&settings.psf_cache, &settings.psf,
-                                                     settings.psf_relative_tail))
+                                                     settings.psf_relative_tail)) {
+#ifdef PSF_BACKEND_DUMMY
+    fputs("PSF cache construction failed; dummy chunk statistics are unavailable.\n",
+          stderr);
+    catalog_destroy(&catalog);
+    spacetime_destroy(&spacetime);
+    blackbody_backend_destroy();
+    return 1;
+#else
     fputs("PSF cache construction failed; using direct evaluator.\n", stderr);
+#endif
+  }
   psf_kernel_cache_report_ready(&settings.psf_cache, stderr);
   if (settings.lens_map_input_path != NULL) {
     const int result = render_lens_map(&settings, &catalog);
     catalog_destroy(&catalog);
     psf_kernel_cache_destroy(&settings.psf_cache);
+    blackbody_backend_destroy();
     return result == 0 ? 0 : 1;
   }
   int result = settings.frames_dir != NULL
@@ -1089,5 +1159,6 @@ int main(int argc, char **argv) {
   spacetime_destroy(&spacetime);
   catalog_destroy(&catalog);
   psf_kernel_cache_destroy(&settings.psf_cache);
+  blackbody_backend_destroy();
   return result == 0 ? 0 : 1;
 }

src/optics.c

@@ -21,68 +21,6 @@ static double clamp(double value, double low, double high)
     return value < low ? low : value > high ? high : value;
 }
 
-/* Wyman, Sloan, and Shirley (2013), Eq. 4: analytic fits to the CIE 1931
- * 2-degree color-matching functions. Wavelength is in nanometres. */
-static void cie_1931_2deg(double wavelength_nm, double xyz[3])
-{
-    const double x1 = (wavelength_nm - 442.0) *
-                      (wavelength_nm < 442.0 ? 0.0624 : 0.0374);
-    const double x2 = (wavelength_nm - 599.8) *
-                      (wavelength_nm < 599.8 ? 0.0264 : 0.0323);
-    const double x3 = (wavelength_nm - 501.1) *
-                      (wavelength_nm < 501.1 ? 0.0490 : 0.0382);
-    const double y1 = (wavelength_nm - 568.8) *
-                      (wavelength_nm < 568.8 ? 0.0213 : 0.0247);
-    const double y2 = (wavelength_nm - 530.9) *
-                      (wavelength_nm < 530.9 ? 0.0613 : 0.0322);
-    const double z1 = (wavelength_nm - 437.0) *
-                      (wavelength_nm < 437.0 ? 0.0845 : 0.0278);
-    const double z2 = (wavelength_nm - 459.0) *
-                      (wavelength_nm < 459.0 ? 0.0385 : 0.0725);
-    xyz[0] = 0.362 * exp(-0.5 * x1 * x1) +
-             1.056 * exp(-0.5 * x2 * x2) - 0.065 * exp(-0.5 * x3 * x3);
-    xyz[1] = 0.821 * exp(-0.5 * y1 * y1) +
-             0.286 * exp(-0.5 * y2 * y2);
-    xyz[2] = 1.217 * exp(-0.5 * z1 * z1) +
-             0.681 * exp(-0.5 * z2 * z2);
-}
-
-static double planck_radiance_lambda(double wavelength_m, double temperature_K)
-{
-    const double h = 6.62607015e-34;
-    const double c = 299792458.0;
-    const double k = 1.380649e-23;
-    const double exponent = h * c / (wavelength_m * k * temperature_K);
-    return 2.0 * h * c * c /
-           (pow(wavelength_m, 5.0) * expm1(exponent));
-}
-
-LinearRgb blackbody_to_linear_rgb(double temperature_K)
-{
-    /* Integrate Planck spectral radiance from 380 to 780 nm into CIE XYZ,
-     * then transform XYZ to linear sRGB. Results are W m^-2 sr^-1 in each
-     * linear-primary channel, before the catalog amplitude and exposure. */
-    double xyz[3] = {0.0, 0.0, 0.0};
-    const double wavelength_step_m = 5e-9;
-    if (!isfinite(temperature_K) || temperature_K <= 0.0)
-        return (LinearRgb){0.0, 0.0, 0.0};
-    for (int wavelength_nm = 380; wavelength_nm <= 780; wavelength_nm += 5) {
-        double matching[3];
-        const double radiance =
-            planck_radiance_lambda(wavelength_nm * 1e-9, temperature_K);
-        cie_1931_2deg(wavelength_nm, matching);
-        for (int channel = 0; channel < 3; ++channel)
-            xyz[channel] += radiance * matching[channel] * wavelength_step_m;
-    }
-    return (LinearRgb){
-        clamp(3.24096994 * xyz[0] - 1.53738318 * xyz[1] - 0.49861076 * xyz[2],
-              0.0, INFINITY),
-        clamp(-0.96924364 * xyz[0] + 1.87596750 * xyz[1] + 0.04155506 * xyz[2],
-              0.0, INFINITY),
-        clamp(0.05563008 * xyz[0] - 0.20397696 * xyz[1] + 1.05697151 * xyz[2],
-              0.0, INFINITY)};
-}
-
 enum {
     PSF_PHASE_RESOLUTION = 64,
     PSF_QUADRATURE_ORDER = 4,

src/optics.h

@@ -50,6 +50,16 @@ extern "C" {
 /* Integrate a Planck spectrum into absolute linear-sRGB spectral radiance
  * (W m^-2 sr^-1), before catalog amplitude and display exposure. */
 LinearRgb blackbody_to_linear_rgb(double temperature_K);
+/* Backend initialization happens before renderer workers start. Integral and
+ * fixed backends reject LUT-only options. The LUT backend loads the sole
+ * repository CIE table by default, or validates an explicit equivalent file;
+ * legacy runtime-generation arguments are rejected. */
+int blackbody_backend_init(const char *table_path, size_t lut_nodes,
+                           double temperature_min_K,
+                           double temperature_max_K,
+                           const char *write_table_path, FILE *report);
+void blackbody_backend_destroy(void);
+const char *blackbody_backend_name(void);
 int psf_kernel_cache_init(PsfKernelCache *cache,
                           const PointSpreadFunction *psf,
                           double relative_tail_fraction);

tests/capture_psf.c

@@ -1,6 +1,7 @@
 /* Standalone bounded diagnostic. Include the production producer so its
  * private query/mapping path stays identical; renderer builds omit the hook. */
 #include "optics.h"
+#include <math.h>
 #include <stdint.h>
 #include <stdio.h>
 #include <stdlib.h>
@@ -100,6 +101,7 @@ int main(int argc, char **argv) {
   if (lens_map_read(argv[1],&map) || map.frame_count!=1 ||
       last>map.frames[0].mesh.triangle_count || catalog_load_csv(&catalog,argv[2]) ||
       catalog.count>32768) return 2;
+  if (blackbody_backend_init(NULL, 0, NAN, NAN, NULL, stderr)) return 1;
   const int indexed=!strcmp(argv[6],"--diagnose-indexed") || !strcmp(argv[6],"--diagnose-indexed-serial");
   if(indexed && index_subset(&catalog))return 1;
   const char *kind=indexed ? "indexed-subset" : "memory";
@@ -126,7 +128,7 @@ int main(int argc, char **argv) {
     }
     printf("DIAGNOSTIC ONLY, no HDR: workers=%d catalog=%s stars=%zu triangles=[%zu,%zu) events=%zu cached=%zu wing=%zu direct=%zu discarded=%zu checksum=%016llx producer_wall=%.9f capped_or_failed=%d\n",
       workers,kind,catalog.count,first,last,total_events,totals[0]+totals[2],totals[2],totals[1],totals[3],(unsigned long long)total_checksum,omp_get_wtime()-start,failed);
-    psf_kernel_cache_destroy(&cache);catalog_destroy(&catalog);lens_map_destroy(&map);
+    psf_kernel_cache_destroy(&cache);catalog_destroy(&catalog);lens_map_destroy(&map);blackbody_backend_destroy();
     return failed ? 1 : 0;
   }
   PsfEventSink sink={.cache=&cache};
@@ -146,6 +148,6 @@ int main(int argc, char **argv) {
     }
     if (ferror(f) || fclose(f)) return 1;
   }
-  psf_kernel_cache_destroy(&cache);catalog_destroy(&catalog);lens_map_destroy(&map);
+  psf_kernel_cache_destroy(&cache);catalog_destroy(&catalog);lens_map_destroy(&map);blackbody_backend_destroy();
   return 0;
 }

tests/replay_psf.hip

@@ -102,11 +102,13 @@ static size_t contributions(const std::vector<PsfCachedEvent>& events,int width,
   return n;
 }
 int main(int argc,char **argv) {
-  if(argc<4 || argc>5) {puts("Usage: replay_psf INPUT.events EVENTS(1..65536) atomic|16|32 [disperse|mixed]\nExternal timeout <=45s; sequential processes only.");return 2;}
+  if(argc<4 || argc>5) {puts("Usage: replay_psf INPUT.events EVENTS(1..65536) atomic|16|32|adaptive [disperse|mixed]\nExternal timeout <=45s; sequential processes only.");return 2;}
   char *end=nullptr;const long requested=strtol(argv[2],&end,10);
   if(!*argv[2] || *end || requested<1 || requested>65536) return 2;
-  const int tile=!strcmp(argv[3],"atomic") ? 0 : !strcmp(argv[3],"16") ? 16 : !strcmp(argv[3],"32") ? 32 : -1;
-  if(tile<0 || (argc==5 && strcmp(argv[4],"disperse") && strcmp(argv[4],"mixed")))return 2;
+  const bool adaptive=!strcmp(argv[3],"adaptive");
+  const int requested_tile=!strcmp(argv[3],"atomic") ? 0 : !strcmp(argv[3],"16") ? 16 :
+                           !strcmp(argv[3],"32") ? 32 : adaptive ? 16 : -1;
+  if(requested_tile<0 || (argc==5 && strcmp(argv[4],"disperse") && strcmp(argv[4],"mixed")))return 2;
   FILE *f=fopen(argv[1],"r");if(!f){perror(argv[1]);return 2;}
   int width,height;size_t count;PointSpreadFunction psf;double tail;
   if(fscanf(f,"PSFEVENTS1 %d %d %zu %lf %lf %lf",&width,&height,&count,&psf.fwhm_pixels,&psf.moffat_beta,&tail)!=6 ||
@@ -153,8 +155,10 @@ int main(int argc,char **argv) {
   const double create=omp_get_wtime()-start;
   hipDeviceProp_t prop;check(hipGetDeviceProperties(&prop,0));
   printf("device=%s arch=%s events=%zu frame=%dx%d mode=%s distribution=%s contributions=%zu CPU_reference=%.9f create=%.9f\n",prop.name,prop.gcnArchName,events.size(),width,height,argv[3],mixed?"fixture-mixed-direct":argc==5?"synthetic-phase-preserving-dispersal":"captured",work,cpu_time,create);fflush(stdout);
-  double bin=0,upload=0,prepare=0,kernel=0,merge=0,clear=0;size_t peak_scratch=0,ref_total=0,task_total=0;
+  double select=0,bin=0,upload=0,prepare=0,kernel=0,merge=0,clear=0;
+  size_t peak_scratch=0,ref_total=0,task_total=0,center_tiles=0,tile_chunks=0,atomic_chunks=0;
   Prepared *prepared=nullptr;int *bounds=nullptr;
+  PsfCachedEvent *tile_events=nullptr;
   unsigned *drefs=nullptr,*dstarts=nullptr;TileTask *dtasks=nullptr;double *partial=nullptr;
   auto direct_boundary=[&]() {
     if(hip_psf_sink_finish(s,gpu.data(),message,sizeof message)) {fprintf(stderr,"%s\n",message);exit(1);}
@@ -162,8 +166,9 @@ int main(int argc,char **argv) {
     if(hip_psf_sink_load_hdr(s,gpu.data(),message,sizeof message)) {fprintf(stderr,"%s\n",message);exit(1);}
   };
   const double replay_start=omp_get_wtime();
-  if(tile) {
+  if(requested_tile) {
     // Fixed hard bounds; no full-frame event list or per-chunk allocation on GPU.
+    check(hipMalloc(&tile_events,16384*sizeof(PsfCachedEvent)));
     check(hipMalloc(&prepared,16384*sizeof(Prepared)));
     check(hipMalloc(&bounds,16384*(size_t)(2*cache.radius_pixels+1)*2*sizeof(int)));
     check(hipMalloc(&drefs,32*1024*1024));check(hipMalloc(&dtasks,2*1024*1024));
@@ -171,7 +176,32 @@ int main(int argc,char **argv) {
   }
   for(size_t offset=0;offset<events.size();offset+=16384) {
     const size_t n=std::min((size_t)16384,events.size()-offset);
-    if(!tile) {if(hip_psf_sink_submit(s,events.data()+offset,n,message,sizeof message)) {fprintf(stderr,"%s\n",message);return 1;}if(mixed && offset==0)direct_boundary();continue;}
+    int tile=requested_tile;
+    if(adaptive) {
+      start=omp_get_wtime();
+      const int selector_tile=32,nx=(width+selector_tile-1)/selector_tile;
+      const int ny=(height+selector_tile-1)/selector_tile;
+      std::vector<unsigned char> occupied((size_t)nx*ny,0);
+      size_t count_occupied=0;
+      for(size_t i=0;i<n;++i) {
+        const int x=(int)floor(events[offset+i].x),y=(int)floor(events[offset+i].y);
+        if(x<0 || x>=width || y<0 || y>=height)continue;
+        unsigned char &mark=occupied[(size_t)(y/selector_tile)*nx+x/selector_tile];
+        if(!mark) {mark=1;++count_occupied;}
+      }
+      /* This threshold is deliberately experimental. The captured dense and
+       * lensed chunks have >=45 events per occupied center tile, while the
+       * equal-work dispersed control has about 2.4. Small chunks retain the
+       * lower-overhead production atomic path. */
+      tile=n>=8192 && count_occupied && n/count_occupied>=32 ? 16 : 0;
+      center_tiles+=count_occupied;select+=omp_get_wtime()-start;
+    }
+    if(!tile) {
+      ++atomic_chunks;
+      if(hip_psf_sink_submit(s,events.data()+offset,n,message,sizeof message)) {fprintf(stderr,"%s\n",message);return 1;}
+      if(mixed && offset==0)direct_boundary();continue;
+    }
+    ++tile_chunks;
     start=omp_get_wtime();
     const int nx=(width+tile-1)/tile,ny=(height+tile-1)/tile;
     std::vector<std::vector<unsigned>> lists(nx*ny);
@@ -208,13 +238,13 @@ int main(int argc,char **argv) {
     peak_scratch=std::max(peak_scratch,bytes+refs.size()*sizeof(unsigned)+tasks.size()*sizeof(TileTask)+starts.size()*sizeof(unsigned)+n*(sizeof(Prepared)+(size_t)(2*cache.radius_pixels+1)*2*sizeof(int)));
     ref_total+=refs.size();task_total+=tasks.size();bin+=omp_get_wtime()-start;
     start=omp_get_wtime();
-    check(hipMemcpyAsync(s->slots[0].device,events.data()+offset,n*sizeof(PsfCachedEvent),hipMemcpyHostToDevice,s->stream));
+    check(hipMemcpyAsync(tile_events,events.data()+offset,n*sizeof(PsfCachedEvent),hipMemcpyHostToDevice,s->stream));
     if(!refs.empty())check(hipMemcpyAsync(drefs,refs.data(),refs.size()*sizeof(unsigned),hipMemcpyHostToDevice,s->stream));
     if(!tasks.empty())check(hipMemcpyAsync(dtasks,tasks.data(),tasks.size()*sizeof(TileTask),hipMemcpyHostToDevice,s->stream));
     check(hipMemcpyAsync(dstarts,starts.data(),starts.size()*sizeof(unsigned),hipMemcpyHostToDevice,s->stream));
     upload+=sync_time(s,start);
     start=omp_get_wtime();
-    hipLaunchKernelGGL(prepare_tiles,dim3((n*32+127)/128),dim3(128),0,s->stream,s->slots[0].device,n,s->phase_resolution,s->radius_pixels,s->max_radius_pixels,prepared,bounds);
+    hipLaunchKernelGGL(prepare_tiles,dim3((n*32+127)/128),dim3(128),0,s->stream,tile_events,n,s->phase_resolution,s->radius_pixels,s->max_radius_pixels,prepared,bounds);
     check(hipGetLastError());prepare+=sync_time(s,start);
     start=omp_get_wtime();
     if(!tasks.empty())hipLaunchKernelGGL(tile_partial,dim3(tasks.size()),dim3(tile*tile),0,s->stream,prepared,bounds,drefs,dtasks,tile,nx,width,height,s->weights,s->phase_resolution,s->radius_pixels,partial,s->hdr);
@@ -225,10 +255,10 @@ int main(int argc,char **argv) {
     // Every partial pixel is written by exactly one thread: no memset required.
   }
   if(hip_psf_sink_finish(s,gpu.data(),message,sizeof message)){fprintf(stderr,"%s\n",message);return 1;}
-  if(tile) {start=omp_get_wtime();check(hipFree(prepared));check(hipFree(bounds));check(hipFree(drefs));check(hipFree(dtasks));check(hipFree(dstarts));check(hipFree(partial));clear=omp_get_wtime()-start;}
+  if(requested_tile) {start=omp_get_wtime();check(hipFree(tile_events));check(hipFree(prepared));check(hipFree(bounds));check(hipFree(drefs));check(hipFree(dtasks));check(hipFree(dstarts));check(hipFree(partial));clear=omp_get_wtime()-start;}
   const double wall=omp_get_wtime()-replay_start;
   HipPsfTiming timing={};hip_psf_sink_get_timing(s,&timing);
-  if(!tile){upload=timing.upload_seconds;kernel=timing.kernel_seconds;}
+  upload+=timing.upload_seconds;kernel+=timing.kernel_seconds;
   double max_abs=0,max_rel=0;long double sums[3]={},diffs[3]={};bool pass=true;
   for(size_t i=0;i<values;++i) {
     if(!std::isfinite(cpu[i]) || !std::isfinite(gpu[i]))pass=false;
@@ -238,7 +268,7 @@ int main(int argc,char **argv) {
   }
   long double ysum=.2126L*sums[0]+.7152L*sums[1]+.0722L*sums[2],ydiff=.2126L*diffs[0]+.7152L*diffs[1]+.0722L*diffs[2];
   struct rusage usage;getrusage(RUSAGE_SELF,&usage);
-  printf("complete_replay=%.9f replay_wall=%.9f bin=%.9f upload=%.9f prepare=%.9f accumulation=%.9f merge=%.9f download=%.9f cleanup=%.9f events_s=%.3f contributions_s=%.3f refs=%zu tasks=%zu scratch_used_peak=%zu scratch_device_reserved=%zu RSS_KiB=%ld max_abs=%.17g max_rel=%.17g flux_R=%.17Lg flux_G=%.17Lg flux_B=%.17Lg flux_Y=%.17Lg\n",create+wall,wall,bin,upload,prepare,kernel,merge,timing.download_seconds,clear,events.size()/wall,work/wall,ref_total,task_total,peak_scratch,tile?162*1024*1024+131072+16384*(sizeof(Prepared)+(size_t)(2*cache.radius_pixels+1)*2*sizeof(int)):0,usage.ru_maxrss,max_abs,max_rel,sums[0]?diffs[0]/sums[0]:0,sums[1]?diffs[1]/sums[1]:0,sums[2]?diffs[2]/sums[2]:0,ysum?ydiff/ysum:0);
+  printf("complete_replay=%.9f replay_wall=%.9f select=%.9f bin=%.9f upload=%.9f prepare=%.9f accumulation=%.9f merge=%.9f download=%.9f cleanup=%.9f events_s=%.3f contributions_s=%.3f center_tiles=%zu tile_chunks=%zu atomic_chunks=%zu refs=%zu tasks=%zu scratch_used_peak=%zu scratch_device_reserved=%zu RSS_KiB=%ld max_abs=%.17g max_rel=%.17g flux_R=%.17Lg flux_G=%.17Lg flux_B=%.17Lg flux_Y=%.17Lg\n",create+wall,wall,select,bin,upload,prepare,kernel,merge,timing.download_seconds,clear,events.size()/wall,work/wall,center_tiles,tile_chunks,atomic_chunks,ref_total,task_total,peak_scratch,requested_tile?163*1024*1024+131072+16384*(sizeof(Prepared)+(size_t)(2*cache.radius_pixels+1)*2*sizeof(int)):0,usage.ru_maxrss,max_abs,max_rel,sums[0]?diffs[0]/sums[0]:0,sums[1]?diffs[1]/sums[1]:0,sums[2]?diffs[2]/sums[2]:0,ysum?ydiff/ysum:0);
   for(int c=0;c<3;++c)if(fabsl(diffs[c])>1e-10L*fmaxl(fabsl(sums[c]),1e-30L))pass=false;
   if(fabsl(ydiff)>1e-10L*fmaxl(fabsl(ysum),1e-30L))pass=false;
   hip_psf_sink_destroy(s);psf_kernel_cache_destroy(&cache);

tests/test_frame.c

@@ -74,7 +74,8 @@ int main(void) {
   FrameLensMesh mesh = {0};
   double *hdr = calloc((size_t)width * height * 3, sizeof *hdr);
   int result = 1;
-  if (hdr == NULL || spacetime_create_minkowski(&spacetime, 10.0) ||
+  if (blackbody_backend_init(NULL, 0, NAN, NAN, NULL, stderr) ||
+      hdr == NULL || spacetime_create_minkowski(&spacetime, 10.0) ||
       frame_lens_mesh_build_coarse(&mesh, width, height, 20, 30.0) ||
       frame_lens_mesh_trace(&mesh, &spacetime, &observer, &trace))
     goto done;
@@ -507,6 +508,7 @@ int main(void) {
 done:
   frame_lens_mesh_destroy(&mesh);
   spacetime_destroy(&spacetime);
+  blackbody_backend_destroy();
   free(hdr);
   return result;
 }