Message boards :
Number crunching :
The Highest Ranked SETI AMD Host is a MAC: Time for a STOCK MAC APP?
Message board moderation
Previous · 1 . . . 4 · 5 · 6 · 7 · 8 · 9 · 10 . . . 21 · Next
Author | Message |
---|---|
Brent Norman Send message Joined: 1 Dec 99 Posts: 2786 Credit: 685,657,289 RAC: 835 |
If you make a blank .h file, wouldn't the compiler point to what line is missing info, thus why it is being called? Just a thought. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
I don't know the first thing about building OpenCL apps for Mac with XCode, but I'll note that MinGW/GCC CPU builds for Windows don't include sqldefs.h so don't look for int8.h. And even if sqldefs.h were included, CLIENT is defined so none of the sqldefs.h content would have any effect. Best I can tell, the problem is that sqldefs.h is referenced in the Project file here; https://setisvn.ssl.berkeley.edu/trac/browser/branches/sah_v7_opt/AKv8/mac_build/seti_boinc.xcodeproj/project.pbxproj DDA5ABB807D16B7F00318A98 /* schema_master.cpp */ = {isa = PBXFileReference; fileEncoding = 30; lastKnownFileType = sourcecode.cpp.cpp; name = schema_master.cpp; path = ../db/schema_master.cpp; sourceTree = SOURCE_ROOT; }; Apparently since Xcode sees those DataBase files in the seti_boinc.xcodeproj file, it demands them and All the #includes to be there. There are about four Header Files that have sqldefs.h on their Includes list. So, since they are listed, they must be present...according to Xcode. I tried removing the #include sqldefs.h from the other files and received more Errors with every #include I removed. It would appear the correct method would be to construct a seti_boinc.xcodeproj file that only contains the files you intend to use. There are probably a few dozen places to add a configuration line, so, without specifying a location telling someone to add a config line isn't very helpful :-) google.com/search?q=how+to+use+apple+xcode |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Another day, another set of problems. If you want to see what Xcode is working with, just look at the Project file here; https://setisvn.ssl.berkeley.edu/trac/browser/branches/sah_v7_opt/AKv8/mac_build/seti_boinc.xcodeproj/project.pbxproj All those Files and settings are loaded into Xcode. You can easily Delete the PPC targets leaving you with just the i386 listings. Pay close attention to the OTHER_CFLAGS and OTHER_LDFLAGS sections (there are a few), that is where those Configuration lines that people keep mentioning go. AFAIK the -disable-server setting has absolutely No effect in Xcode. I made a duplicate header file and named it int8.h, so far that issue hasn't arisen again...yet. I still have the target OS set for OS 10.7. Now I'm looking at a new set of RED Errors, the RED ones are Show Stoppers; /Users/Tom/sah_v7_opt/AKv8/client/analyzeFuncs.h /Users/Tom/sah_v7_opt/AKv8/client/analyzeFuncs.h:77:9: Unknown type name '__m128' /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:69:10: In file included from /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:69: /Users/Tom/sah_v7_opt/AKv8/client/analyzeFuncs.h:78:9: Unknown type name '__m128i' /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:69:10: In file included from /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:69: /Users/Tom/sah_v7_opt/AKv8/client/analyzeFuncs.h:79:9: Unknown type name '__m128i' /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:69:10: In file included from /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:69: /Users/Tom/sah_v7_opt/AKv8/client/analyzeFuncs.h:80:9: Unknown type name '__m128i' /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:69:10: In file included from /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:69: /Users/Tom/sah_v7_opt/AKv8/client/analyzeFuncs.h:81:9: Unknown type name '__m128d' /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:69:10: In file included from /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:69: /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:3571:27: Use of undeclared identifier 'tempPulsePoT' /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:3590:24: Use of undeclared identifier 'tempPulsePoT' My guess is those issues are named and declared in some other file that hasn't been loaded into Xcode. Xcode is not listing any missing files though. So, unless someone knows which file to load, this may be hard to fix. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Nice. Thanks to Google I found the Unknown type name '__m128' means SSEx is not being recognized. Strange, there were numerous SSEx flags already set...just look at the above link to the Project file. So, I replaced the other flags with -DUSE_SSE3 in the Custom Compiler flags and added All the SSEx flags in the Preprocessor Macros. Now it doesn't give the __m128 errors. I still have 2 Errors left; /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:3571:27: Use of undeclared identifier 'tempPulsePoT' /Users/Tom/sah_v7_opt/AKv8/client/analyzePoT.cpp:3590:24: Use of undeclared identifier 'tempPulsePoT' #if 0 fprintf(stderr,"FindTriplets called for: icfft=%d, PulsePotLen=%d, offset=%d,ThisPoT=%d\n", analysis_state.icfft,PulsePoTLen,TOffset,ThisPoT); #endif retval = find_triplets(tempPulsePoT, PulsePoTLen, (float)PoTInfo.TripletThresh, TOffset, ThisPoT); if (retval) SETIERROR(retval,"from find_triplets()"); } #if 0//_WIN32 //Jason: Added BOINCASSERT(_CrtCheckMemory()); #endif #if USE_OPENCL || USE_CUDA if(need_pulse_cpu_processing && ThisPoT>=PulseSearchBinStart && ThisPoT<PulseSearchBinStop){ #if OCL_VERBOSE fprintf(stderr,"Doing find_pulse() on CPU,ThisPoT=%d,PulsePoTNum=%d\n",ThisPoT,PulsePoTNum); #endif #else if(!SkipPulse) { #endif retval = find_pulse(tempPulsePoT, PulsePoTLen, (float)PoTInfo.PulseThresh, TOffset, ThisPoT ); if (retval) SETIERROR(retval,"from find_pulse()"); Hmmm, what is a tempPulsePoT... |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
Nice. It's a pointer to the array of floats to be analyzed. Stripped of some unused and/or commented out code: float *tempPulsePoT = PulsePoT; if ((TOffset) & 0x3) { //R: data not aligned tempPulsePoT = PulsePoT; memcpy(PulsePoT, &tPowerSpectrum[(ThisPoT) * PoTLen + TOffset], PulsePoTLen*sizeof(float)); } else { //R:data already aligned tempPulsePoT = tPowerSpectrum + (ThisPoT) * PoTLen + TOffset; } However, unoptimized code without USE_I386_OPTIMIZATIONS defined would miss that type declaration and give the error you're seeing. Joe |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Thanks Joe. That's 2 in a row where the Old settings in the Old Program file are NOT working with the Newer Xcode. The flag USE_I386_OPTIMIZATIONS was in every one of the Preprocessor Macros boxes for each target. I added -DUSE_I386_OPTIMIZATIONS in the Custom Compiler Flag boxes and the Errors went away. That's NOT good. It means I can't trust any of these existing settings from the Old Program file at SVN. On the other hand, maybe this will fix the problem with Joe's existing build. The main problem is the Lower Angle Range tasks use a Large amount of CPU time. On my Harpertown Xeon running 6 of the Low Angle Range tasks on my ATI cards will bring the CPU to it's knees and make the computer annoying to use. Now for the Next Error. Yes now there is a New one; /Users/Tom/sah_v7_opt/src/CUInfo.cpp /Users/Tom/sah_v7_opt/src/CUInfo.cpp:1:10: 'cuda_runtime_api.h' file not found Another missing file :-( I have downloaded most of the files on the Repository, about 5 GBs. This file isn't to be found. I don't remember seeing any Header files available to download at nVidia either. Any idea where this file may be found? |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Down to the end, I hope. So, I found the nVidia files, then had to find all the *includes from sqldefs.h. After that the "Unknown type name '__m128x'" started popping up again along with a few others. I tried about everything I could and then decided to reboot into Yosemite and try it with Xcode 6.1.1. I'm receiving the same Errors in 10.10.2 & Xcode 6.1.1 as I was with 10.8.5 & Xcode 5.1.1. So, I'm going to have to say there might be something wrong with those 2 files. There might be more after these 2, but I won't know until these 2 are solved. Right now the show is stopped with the Errors; /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/analyzeFuncs.h /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/analyzeFuncs.h:77:9: Unknown type name '__m128' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:35:10: In file included from /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:35: /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/analyzeFuncs.h:78:9: Unknown type name '__m128i' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:35:10: In file included from /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:35: /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/analyzeFuncs.h:79:9: Unknown type name '__m128i' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:35:10: In file included from /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:35: /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/analyzeFuncs.h:80:9: Unknown type name '__m128i' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:35:10: In file included from /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:35: /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/analyzeFuncs.h:81:9: Unknown type name '__m128d' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:35:10: In file included from /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:35: /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:123:14: Use of undeclared identifier '_mm_load_ss' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:124:21: Use of undeclared identifier '_mm_rcp_ss' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:127:47: Use of undeclared identifier '_mm_set_ps' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:59:22: Expanded from macro 'ONE' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:133:10: Use of undeclared identifier '_mm_cvtss_f32' |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Hmmm. I simply added a few lines to lcgamm.cpp and the previous Errors disappeared. #include "diagnostics.h" #define ITMAX 1000 #define EPS 1e-6 #define FP_EPS 1.0e-35 #define RFP_EPS 1.0e35 #define _mm_load_ss #define _mm_set_ps #define _mm_cvtss_f32 Unfortunately all the old Errors were replaced with one new Error; /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/AKv8/client/lcgamm.cpp:127:10: Cannot initialize a variable of type 'vFloat' (aka 'int') with an rvalue of type 'float *' inline float f_recip(float f) { // obtain estimate float estimate = __fres(f); // one round of Newton-Raphson return __fmadds(__fnmsubs(estimate, f, 1.0), estimate, estimate); } #elif defined(__SSE__) inline float f_recip(float f) { // obtain estimate vFloat v = _mm_load_ss(&f); vFloat estimate = _mm_rcp_ss(v); // one round of Newton-Raphson estimate = _mm_add_ss(_mm_mul_ss(_mm_sub_ss(ONE, _mm_mul_ss(estimate, v)), estimate), estimate); Sounds like some unknown math problem to me. This line is the problem, vFloat v = _mm_load_ss(&f); Anyone have a clue? |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Just a little example of what is missing. The current Mac seti_boinc-hd5xxx build does Not have SSE in it's wisdom file. Note the differences; r2549_IntelRXeonRCPUE5462280GHz_x64.wisdom (fftw-3.3.4 fftwf_wisdom #x08ac4c16 #x457005cc #xea102cf7 #xd7ff9038 (fftwf_codelet_n2bv_16_sse2 0 #x11bdd #x11bdd #x0 #xc0ad9768 #xd03118b4 #x1bfbab73 #xbca89889) (fftwf_codelet_n2bv_32_sse2 0 #x11bdd #x11bdd #x0 #x7915b129 #x48243e82 #xe00b783f #xe6fb8a03) (fftwf_dft_buffered_register 0 #x11bdd #x11bdd #x0 #x9e007033 #x2492644c #x2214c74b #x113de887) (fftwf_codelet_r2cf_8 0 #x10048 #x10048 #x0 #x284a237d #xf517855a #xf53005ee #x7baffc58) (fftwf_codelet_t2bv_64_sse2 0 #x11bdd #x11bdd #x0 #xd86f0fcc #x804581bb #x5ccfdfba #x14108c49) (fftwf_codelet_t3bv_8_sse2 0 #x11048 #x11048 #x0 #xa33f803b #x1d3d7036 #x2bd76bc9 #x1385cd33) (fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #xccfabeb1 #xf40d2a94 #x49797e4b #xbd47afeb) (fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #xa4e6fa81 #x7328a0a8 #x8668f723 #x37559e56) (fftwf_codelet_n1bv_64_sse2 0 #x11bdd #x11bdd #x0 #x4f60953d #xafb92cdb #x925373c3 #xb6cbce04) (fftwf_dft_vrank_geq1_register 0 #x11bdd #x11bdd #x0 #xc1e8d4e9 #x6f05b677 #x9c1c71ea #x515c6898) (fftwf_codelet_t1bv_8_sse2 0 #x11048 #x11048 #x0 #xd86f0fcc #x804581bb #x5ccfdfba #x14108c49) (fftwf_codelet_t2bv_8_sse2 0 #x11bdd #x11bdd #x0 #x3994c561 #x9164720f #xed58ceec #xe4bd2a71) (fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #x4a69f5b3 #x02a11518 #xed90863e #xef4fd116) (fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #x6178ec3b #x6be0f5bf #xc979cda6 #x2530a126) (fftwf_dft_vrank_geq1_register 0 #x11bdd #x11bdd #x0 #xcea6cb1a #x51cfc7d9 #x69c751ad #xaaf36f5b) (fftwf_codelet_hf_64 1 #x11bcd #x11bdd #x0 #xd7ba8a81 #x269d1e2f #x8c825f0b #x2f53c532) (fftwf_rdft_rank0_register 1 #x11048 #x11048 #x0 #x5d55fa8c #x29023783 #xbe843406 #xf9ac51c6) (fftwf_dft_buffered_register 0 #x11048 #x11048 #x0 #x9e007033 #x2492644c #x2214c74b #x113de887) (fftwf_codelet_t2bv_16_sse2 0 #x11bdd #x11bdd #x0 #xb9da9e1a #xb5ad84a3 #xd98a6c23 #x906c1255) (fftwf_codelet_n1bv_128_sse2 0 #x11bdd #x11bdd #x0 #x08a9e380 #x8855f11e #xbdec2b71 #x97c190d9) (fftwf_codelet_t2bv_16_sse2 0 #x11bdd #x11bdd #x0 #x8d9655ca #x438f5e9a #x75a3c7bc #xf7459970) (fftwf_codelet_t2bv_8_sse2 0 #x11048 #x11048 #x0 #x3063aedc #x25188f39 #x688e67b7 #x6380e1e0) r2696_IntelRXeonRCPUE5462280GHz_x64.wisdom (fftw-3.3.4 fftwf_wisdom #x08ac4c16 #x457005cc #xea102cf7 #xd7ff9038 (fftwf_dft_buffered_register 0 #x1040 #x1040 #x0 #x94055af3 #x47c04613 #x7c55788b #x735b3c18) (fftwf_dft_vrank_geq1_register 0 #x10bdd #x10bdd #x0 #xb5397ecb #x7b5616a2 #x644a8cb4 #xb4eac954) (fftwf_codelet_n2bv_32_sse2 0 #x40 #x40 #x0 #x7915b129 #x48243e82 #xe00b783f #xe6fb8a03) (fftwf_dft_buffered_register 0 #x1040 #x1040 #x0 #xf03cffb1 #xd63eb7df #x82dbdfd2 #xaf9e4be5) (fftwf_codelet_n1fv_16_sse2 0 #x10bdd #x10bdd #x0 #x22b43c01 #x2b5ed86e #x403cb186 #xc48cdc6f) (fftwf_rdft_rank0_register 0 #x1040 #x1040 #x0 #x9d3aaef3 #x396b4e63 #x59e4d810 #xb358ae2b) (fftwf_dft_nop_register 0 #x1040 #x1040 #x0 #xb768271c #xc8bdfdf8 #x0408e36b #xdd0ce04d) (fftwf_codelet_t2fv_8_sse2 0 #x10bdd #x10bdd #x0 #x54496af0 #xf9e2197e #x8f4afac4 #x7887038c) (fftwf_codelet_t2fv_8_sse2 0 #x40 #x40 #x0 #x8f86ac3f #x7cc3d8db #xc7865ecc #xea8b825e) (fftwf_rdft_rank0_register 0 #x1040 #x1040 #x0 #x676222b4 #x3bf90562 #x3ecf6426 #x676e17b6) (fftwf_dft_nop_register 0 #x11bdd #x11bdd #x0 #x2b4c0371 #x0254c533 #x65be4826 #x5beb5520) (fftwf_codelet_t2bv_16_sse2 0 #x40 #x40 #x0 #x71b35416 #x033191ee #x08427f48 #x060813e0) r2760_IntelRXeonRCPUE5462280GHz_x64.wisdom aka seti_boinc-hd5xxx (fftw-3.1.1a fftwf_wisdom (fftwf_codelet_t1_4 0 #x11bdd #x11bdd #x0 #x0358088d #xfecc6614 #x11e87e27 #x653c489f) (fftwf_dft_vrank_geq1_register 0 #x10048 #x10048 #x0 #x0b43b7c1 #x2f19edd4 #xf37a80a9 #x0077d1bd) (fftwf_dft_buffered_register 0 #x11bdd #x11bdd #x0 #x66f34087 #xf3ba91de #x56cec9eb #x0f42b68b) (fftwf_codelet_t1_4 0 #x11bdd #x11bdd #x0 #x3e32d5bf #x36c33f81 #xabdcc9d6 #xfcbff80b) (fftwf_codelet_n1_8 0 #x11bdd #x11bdd #x0 #x1d4c3a72 #x43a9797f #x9cb757db #x02f92729) (fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #xeb26511e #x291a984c #x817e07be #xa5e9e72d) (fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #x84792c48 #xca2640fc #x9a4724a1 #x80562141) (fftwf_codelet_t1_4 0 #x11048 #x11048 #x0 #xe95d8ea4 #x531619a7 #xde7a6c76 #x19250a6a) (fftwf_dft_vrank_geq1_register 0 #x10bdd #x10bdd #x0 #x92ea1485 #xdba319d1 #x326c8d9c #x685359b6) (fftwf_codelet_t1_4 0 #x11048 #x11048 #x0 #xb3d76b03 #x0e4c780f #x9a903cb6 #xbece216d) (fftwf_dft_vrank_geq1_register 0 #x10bdd #x10bdd #x0 #x4e52a122 #x2ecd8e21 #xf9c60f6b #x7d027594) (fftwf_dft_buffered_register 0 #x11048 #x11048 #x0 #x66f34087 #xf3ba91de #x56cec9eb #x0f42b68b) (fftwf_codelet_t1_4 0 #x10048 #x10048 #x0 #x135a2631 #x8cdcaea8 #x606d4378 #xd93bac8a) (fftwf_codelet_t1_4 0 #x10bdd #x10bdd #x0 #xcc9c3616 #x55222121 #x42d246a8 #x4d1dcd73) (fftwf_codelet_t1_4 0 #x10bdd #x10bdd #x0 #x6aef7b94 #x58bd3ac1 #xa44ab2c3 #xe4b76b99) (fftwf_dft_nop_register 0 #x11bdd #x11bdd #x0 #x92754d7b #xb30bc08a #xb5ce559c #x92c5cfea) (fftwf_rdft_rank0_register 2 #x11bdd #x11bdd #x0 #xcc837b2d #x9cbeb56a #x35e24005 #x66525bc4) (fftwf_codelet_t1_4 0 #x10bdd #x10bdd #x0 #x1ccca46b #x965a97b1 #x6668d9c0 #x4d4f52eb) (fftwf_dft_vrank_geq1_register 0 #x10048 #x10048 #x0 #xd5e38ffd #x25e18bc2 #x4c95d441 #x455a4d76) (fftwf_codelet_t1_4 0 #x11048 #x11048 #x0 #x6540d8a5 #x04b882ff #xf860fca7 #xb07e7d0d) (fftwf_dft_vrank_geq1_register 0 #x10048 #x10048 #x0 #x9d92cb00 #xc3591c0f #xd1599b41 #x12b9c20b) (fftwf_codelet_t1_4 0 #x11048 #x11048 #x0 #x4bd84776 #xc7f5dabe #x52a6bbde #xfc239881) No SSE present in seti_boinc-hd5xxx. Please tell me that's not an Alpha version of fftw-3.1.1. There seems to be a problem with SSEx on Macs. Back when I was testing an App for Urs there was a problem with SSE. Right now the __m128x Errors keep coming and going in Xcode suggesting SSEx is not being seen at times. |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
Hmmm. I simply added a few lines to lcgamm.cpp and the previous Errors disappeared. Those empty definitions for _mm_* are not right, those are intrinsic SIMD functions which are defined in xmmintrin.h so apparently the build isn't getting that. Under other build systems that isn't loaded directly, there's an intrin.h or similar which picks up the more specific SIMD related header files. Unfortunately all the old Errors were replaced with one new Error; A vFloat is essentially the same as a __m128 (see near the beginning of analyzeFuncs.h). The dual naming is because Alex Kan worked with PPC Macs first, and vFloat is what a vector of floats was called for Altivec. The __m128 name is Intel terminology for the same thing. The "Cannot initialize a variable of type 'vFloat' (aka 'int')..." indicates that equivalence has not been established. It seems there are multiple header files which aren't being picked up. I have no idea why. Joe |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Those empty definitions for _mm_* are not right, those are intrinsic SIMD functions which are defined in xmmintrin.h so apparently the build isn't getting that. Under other build systems that isn't loaded directly, there's an intrin.h or similar which picks up the more specific SIMD related header files... I just searched my Mac for xmmintrin.h and the only copies on the machine are in the Old Developer folder for 10.6. There are not any copies of xmmintrin.h in the 5 GBs of SVN Repository files I downloaded the other day. So, it would seem xmmintrin.h is not in the Repository. I'll add a copy to the build and see what happens. !!!!!!!!!!! Hey, we've moved on! Now lcgamm.cpp says; * $Id: lcgamm.cpp,v 1.9.2.6 2005/12/01 00:00:35 korpela Exp $ */ #include "sah_config.h" #include "intrinsics.h" #include "analyzeFuncs.h" #include "xmmintrin.h" #if defined(_WIN32) && !defined(__MINGW32__) #include <crtdbg.h> #endif #include <cstdio> #include <cmath> //#include <Accelerate/Accelerate.h> //#include <ppc_intrinsics.h> #include "diagnostics.h" #define ITMAX 1000 #define EPS 1e-6 #define FP_EPS 1.0e-35 #define RFP_EPS 1.0e35 #if defined(__VEC__) || defined(__SSE__) //#define ZERO ((vFloat) {0.0f, 0.0f, 0.0f, 0.0f}) //#define ONE ((vFloat) {1.0f, 1.0f, 1.0f, 1.0f}) //#define TWO ((vFloat) {2.0f, 2.0f, 2.0f, 2.0f}) #define ZERO (_mm_set_ps(0.0f,0.0f,0.0f,0.0f)) #define ONE (_mm_set_ps(1.0f, 1.0f, 1.0f, 1.0f)) #define TWO (_mm_set_ps(2.0f, 2.0f, 2.0f, 2.0f)) #endif All the Old Errors are Gone....but New ones are listed now. /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/src/cuda_runtime_api.h /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/src/cuda_runtime_api.h:32:16: Unknown type name 'size_t' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/src/CUInfo.cpp:1:10: In file included from /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/src/CUInfo.cpp:1: /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/src/cuda_runtime_api.h:33:16: Unknown type name 'size_t' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/src/CUInfo.cpp:1:10: In file included from /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/src/CUInfo.cpp:1: /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/src/CUInfo.cpp /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/src/CUInfo.cpp:37:12: Use of undeclared identifier 'cudaGetDeviceCount' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/src/CUInfo.cpp:45:9: Unknown type name 'cudaDeviceProp' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/src/CUInfo.cpp:74:11: Use of undeclared identifier 'AquireExecutionSlot' /Volumes/Mountain Lion/Users/Tom/sah_v7_opt/src/CUInfo.cpp:75:4: Use of undeclared identifier 'host' I thought these were supposed to be OpenCL Apps, why am I messing with all these CUDA files? It's as bad as having to deal with all the server files. I guess those CUDA files I tracked down are not up to par. Wonderful. At least the SSE problems are looking better. Any idea where I can find the same fftw-3.3.4 that's in the other Apps? Thanks. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Should not be. It's system include. Check your include paths why that file did not include from its original place. W/o that your resulting build will be incompatible with other configs. System-level includes should be taken from their original places. Direct copying into project directory allowable only for troubleshooting.
Optimized souurces support many build paths for numerous device types. It's important to chose needed path and configure it right. Check your defines. USE_OPENCL defines OpenCL path, USE_CUDA - CUDA one. Also, some configuration required on project/makefile level too. Not all source files should be included in every project config.
http://fftw.org/ |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
...Should not be. It's system include. Check your include paths why that file did not include from its original place. I'm afraid it may not be so simple. The only place xmmintrin.h was found was in the Developer folder I just installed off the 10.6 DVD the other day. It was not found in the Updated 10.6.8 system, 10.8.5 or 10.10.2. If I hadn't of installed it from the 10.6.0 DVD onto my Lion system it wouldn't have been on my Mac at all. 10.6 was the last OSX to support PPC, apparently it wasn't included in later OS versions. It also appears xmmintrin.h was removed from 10.6.x at some point. That may be a problem. My flags do not contain USE_CUDA, they do contain; -USE_OPENCL -USE_OPENCL_HD5xxx -USE_OPENCL_NV -USE_OPENCL_INTEL My guess is the Header files have the CUDA files as *includes, as they have the server files included. The Repository header CUInfo.cpp calls for the first CUDA file cuda_runtime_api.h, then it snowballs from there. In fact, it looks like all the Errors are on CUInfo.cpp. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
You may want to look at this. It's the intrinsics.h file from the SVN Repository that's *included in Numerous Repository files, https://setisvn.ssl.berkeley.edu/trac/browser/branches/sah_v7_opt/src/intrinsics.h; #elif USE_SSE #pragma "---Usage of SSE instruction set selected---" #define __SSE__ 1 #if ( _GNU_SOURCE || __APPLE__ || __FreeBSD__ || __MINGW32__ ) #include <xmmintrin.h> #include "SSE2_emulate.h" #include "SSE3_emulate.h" //#if __INTEL_COMPILER #else #include <xmmintrin.h> //#elif !_MSC_VER //#error Need MS VC++ or ICC to build //#else // VC++ //#include <intrin.h> //#endif #if _MSC_VER == 1600 //R: VS2010 intrin.h include is incompatible with emulation #else #include "SSE2_emulate.h" #include "SSE3_emulate.h" That pretty much appears to be an #include <xmmintrin.h>, but xmmintrin.h doesn't appear to be in the Repository or on any recent OSX systems....I could be wrong. I don't know why Xcode hasn't flagged it as a missing file yet, but until I installed it a few hours ago it was missing. I suppose maybe once I had gotten through all these missing server and CUDA files Xcode might have flagged it at that point. BTW, these server and CUDA files aren't in my systems either, should I stop tracking them down and adding them to my build? Here is a section from xmmintrin.h; /* The Intel API is flexible enough that we must allow aliasing with other vector types, and their scalar components. */ typedef float __m128 __attribute__ ((__vector_size__ (16), __may_alias__)); /* Internal data types for implementing the intrinsics. */ typedef float __v4sf __attribute__ ((__vector_size__ (16))); /* Create a selector for use with the SHUFPS instruction. */ #define _MM_SHUFFLE(fp3,fp2,fp1,fp0) \ (((fp3) << 6) | ((fp2) << 4) | ((fp1) << 2) | (fp0)) /* Constants for use with _mm_prefetch. */ enum _mm_hint { _MM_HINT_T0 = 3, _MM_HINT_T1 = 2, _MM_HINT_T2 = 1, _MM_HINT_NTA = 0 }; /* Bits in the MXCSR. */ #define _MM_EXCEPT_MASK 0x003f #define _MM_EXCEPT_INVALID 0x0001 #define _MM_EXCEPT_DENORM 0x0002 #define _MM_EXCEPT_DIV_ZERO 0x0004 #define _MM_EXCEPT_OVERFLOW 0x0008 #define _MM_EXCEPT_UNDERFLOW 0x0010 #define _MM_EXCEPT_INEXACT 0x0020 #define _MM_MASK_MASK 0x1f80 #define _MM_MASK_INVALID 0x0080 #define _MM_MASK_DENORM 0x0100 #define _MM_MASK_DIV_ZERO 0x0200 #define _MM_MASK_OVERFLOW 0x0400 #define _MM_MASK_UNDERFLOW 0x0800 #define _MM_MASK_INEXACT 0x1000 #define _MM_ROUND_MASK 0x6000 #define _MM_ROUND_NEAREST 0x0000 #define _MM_ROUND_DOWN 0x2000 #define _MM_ROUND_UP 0x4000 #define _MM_ROUND_TOWARD_ZERO 0x6000 #define _MM_FLUSH_ZERO_MASK 0x8000 #define _MM_FLUSH_ZERO_ON 0x8000 #define _MM_FLUSH_ZERO_OFF 0x0000 I'm starting to understand why Mr Fox is MIA. ;-) |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Then you have wrong config. -USE_OPENCL_HD5xxx -USE_OPENCL_NV -USE_OPENCL_INTEL are mutually exclusive. Also, what SSE level define do you use? perhaps you should go with USE_SSSE3. |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
Since I was able to steal away a few hours I'd thought I'd give this another shot. Installed contents of that zip into the beta folder of Seti to try it out. Created a app_info.xml <app_info> <app> <name>setiathome_v7</name> </app> <file_info> <name>seti_boinc-intel</name> <executable/> </file_info> <file_info> <name>MultiBeam_Kernels_r2760.cl</name> </file_info> <file_info> <name>MultiBeam_Kernels_additional_r2760.cl</name> </file_info> <app_version> <app_name>setiathome_v7</app_name> <platform>x86_64-apple-darwin</platform> <version_num>703</version_num> <plan_class>opencl_intel_gpu_sah</plan_class> <avg_ncpus>0.04</avg_ncpus> <max_ncpus>0.1</max_ncpus> <coproc> <type>intel_gpu</type> <count>1</count> </coproc> <file_ref> <file_name>seti_boinc-intel</file_name> <main_program/> </file_ref> <file_ref> <file_name>MultiBeam_Kernels_r2760.cl</file_name> </file_ref> <file_ref> <file_name>MultiBeam_Kernels_additional_r2760.cl</file_name> </file_ref> </app_version> </app_info> All that gets me is Mon Feb 9 18:12:31 2015 | SETI@home Beta Test | update requested by user Mon Feb 9 18:12:34 2015 | SETI@home Beta Test | Sending scheduler request: Requested by user. Mon Feb 9 18:12:34 2015 | SETI@home Beta Test | Not requesting tasks: don't need Mon Feb 9 18:12:36 2015 | SETI@home Beta Test | Scheduler request completed I know the CPU apps work because if I undo all of this it will start to download and work. Anyone see anything that jumps out at them? I even removed that section that says <main_program/> didn't make a difference. Zalster |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
All that gets me is You didn't ask for work because you didn't need any, set sched_op_debug (it's a really useful flag, all my hosts have it set) and you'll see which device Boinc is asking work for, and for how much: 09/02/2015 21:09:18 | SETI@home | [sched_op] Starting scheduler request 09/02/2015 21:09:18 | SETI@home | Sending scheduler request: To fetch work. 09/02/2015 21:09:18 | SETI@home | Requesting new tasks for NVIDIA GPU and AMD/ATI GPU 09/02/2015 21:09:18 | SETI@home | [sched_op] CPU work request: 0.00 seconds; 0.00 devices 09/02/2015 21:09:18 | SETI@home | [sched_op] NVIDIA GPU work request: 207059.79 seconds; 0.00 devices 09/02/2015 21:09:18 | SETI@home | [sched_op] AMD/ATI GPU work request: 142175.62 seconds; 0.00 devices 09/02/2015 21:09:29 | SETI@home | Scheduler request completed: got 0 new tasks 09/02/2015 21:09:29 | SETI@home | [sched_op] Server version 705 09/02/2015 21:09:29 | SETI@home | No tasks sent 09/02/2015 21:09:29 | SETI@home | No tasks are available for SETI@home v7 09/02/2015 21:09:29 | SETI@home | No tasks are available for AstroPulse v7 09/02/2015 21:09:29 | SETI@home | Tasks for CPU are available, but your preferences are set to not accept them 09/02/2015 21:09:29 | SETI@home | This computer has reached a limit on tasks in progress 09/02/2015 21:09:29 | SETI@home | Project requested delay of 303 seconds 09/02/2015 21:09:29 | SETI@home | [sched_op] Deferring communication for 00:05:03 09/02/2015 21:09:29 | SETI@home | [sched_op] Reason: requested by project Claggy |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
Sorry, getting a headache reading all of this... I changed that one place in the xml so it now reads 705 from 703. Other than that nothing. Mon Feb 9 18:51:07 2015 | SETI@home Beta Test | update requested by user Mon Feb 9 18:51:11 2015 | SETI@home Beta Test | sched RPC pending: Requested by user Mon Feb 9 18:51:11 2015 | SETI@home Beta Test | [sched_op] Starting scheduler request Mon Feb 9 18:51:11 2015 | SETI@home Beta Test | Sending scheduler request: Requested by user. Mon Feb 9 18:51:11 2015 | SETI@home Beta Test | Requesting new tasks for Intel GPU Mon Feb 9 18:51:11 2015 | SETI@home Beta Test | [sched_op] CPU work request: 0.00 seconds; 0.00 devices Mon Feb 9 18:51:11 2015 | SETI@home Beta Test | [sched_op] NVIDIA GPU work request: 0.00 seconds; 0.00 devices Mon Feb 9 18:51:11 2015 | SETI@home Beta Test | [sched_op] Intel GPU work request: 129600.00 seconds; 1.00 devices Mon Feb 9 18:51:12 2015 | SETI@home Beta Test | Scheduler request completed: got 0 new tasks Mon Feb 9 18:51:12 2015 | SETI@home Beta Test | [sched_op] Server version 705 Mon Feb 9 18:51:12 2015 | SETI@home Beta Test | No tasks sent Mon Feb 9 18:51:12 2015 | SETI@home Beta Test | No tasks are available for AstroPulse v7 Mon Feb 9 18:51:12 2015 | SETI@home Beta Test | Tasks for CPU are available, but your preferences are set to not accept them Mon Feb 9 18:51:12 2015 | SETI@home Beta Test | Tasks for AMD/ATI GPU are available, but your preferences are set to not accept them Mon Feb 9 18:51:12 2015 | SETI@home Beta Test | Project requested delay of 7 seconds Mon Feb 9 18:51:12 2015 | SETI@home Beta Test | [sched_op] Deferring communication for 00:00:07 Mon Feb 9 18:51:12 2015 | SETI@home Beta Test | [sched_op] Reason: requested by project Mon Feb 9 18:51:22 2015 | SETI@home Beta Test | [sched_op] Starting scheduler request Mon Feb 9 18:51:22 2015 | SETI@home Beta Test | Sending scheduler request: To fetch work. Mon Feb 9 18:51:22 2015 | SETI@home Beta Test | Requesting new tasks for Intel GPU Mon Feb 9 18:51:22 2015 | SETI@home Beta Test | [sched_op] CPU work request: 0.00 seconds; 0.00 devices Mon Feb 9 18:51:22 2015 | SETI@home Beta Test | [sched_op] NVIDIA GPU work request: 0.00 seconds; 0.00 devices Mon Feb 9 18:51:22 2015 | SETI@home Beta Test | [sched_op] Intel GPU work request: 129600.00 seconds; 1.00 devices Mon Feb 9 18:51:23 2015 | SETI@home Beta Test | Scheduler request completed: got 0 new tasks Mon Feb 9 18:51:23 2015 | SETI@home Beta Test | [sched_op] Server version 705 Mon Feb 9 18:51:23 2015 | SETI@home Beta Test | No tasks sent Mon Feb 9 18:51:23 2015 | SETI@home Beta Test | No tasks are available for AstroPulse v7 Mon Feb 9 18:51:23 2015 | SETI@home Beta Test | Tasks for CPU are available, but your preferences are set to not accept them Mon Feb 9 18:51:23 2015 | SETI@home Beta Test | Tasks for AMD/ATI GPU are available, but your preferences are set to not accept them Mon Feb 9 18:51:23 2015 | SETI@home Beta Test | Project requested delay of 7 seconds Mon Feb 9 18:51:23 2015 | SETI@home Beta Test | [sched_op] Deferring communication for 00:00:07 Mon Feb 9 18:51:23 2015 | SETI@home Beta Test | [sched_op] Reason: requested by project edit.. There is no work units on this computer currently |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
Sorry, getting a headache reading all of this... You're got Astropulse_v7 enabled in your preferences, But there's no work for that because of the lack of splitting at Seti Main or Beta, You're not had Seti_v7 tasks on that host since last September, So make sure Seti_v7 is enabled in your Seti Beta preferences. All SETI@home v7 tasks for computer 72386 Claggy |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
I do have the seti_v7 enabled , but what I'm trying to test is the GPU versions. Not really looking for APs at this time since I know those work. I have CPU turned off right now. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.