diff --git a/.gitattributes b/.gitattributes
index 487a7678f0b1a64ca20d44a865dec0842d1bc3c6..48ca44cd2d84f678ccd26442d82be637eb783cc6 100644
--- a/.gitattributes
+++ b/.gitattributes
@@ -3657,7 +3657,6 @@ RTCP/GPUProc/src/Align.h -text
RTCP/GPUProc/src/BandPass.cc -text
RTCP/GPUProc/src/BandPass.h -text
RTCP/GPUProc/src/BeamFormer/BeamFormer.cl -text
-RTCP/GPUProc/src/BeamFormer/BeamFormer.cl-0.ptx -text
RTCP/GPUProc/src/BeamFormer/BeamFormer.cl.4x3 -text
RTCP/GPUProc/src/BeamFormer/BeamFormer.cl.6x3 -text
RTCP/GPUProc/src/BeamFormer/BeamFormer.cl.bak -text
@@ -3665,16 +3664,11 @@ RTCP/GPUProc/src/BeamFormer/BeamFormer.cl.not -text
RTCP/GPUProc/src/BeamFormer/BeamFormer.cl.ok -text
RTCP/GPUProc/src/BeamFormer/BeamFormer.cl.orig -text
RTCP/GPUProc/src/BeamFormer/CoherentStokes.cl -text
-RTCP/GPUProc/src/BeamFormer/CoherentStokes.cl-0.ptx -text
RTCP/GPUProc/src/BeamFormer/CoherentStokes.cl.ok -text
RTCP/GPUProc/src/BeamFormer/Dedispersion.cl -text
-RTCP/GPUProc/src/BeamFormer/Dedispersion.cl-0.ptx -text
RTCP/GPUProc/src/BeamFormer/IncoherentStokes.cl -text
-RTCP/GPUProc/src/BeamFormer/IncoherentStokes.cl-0.ptx -text
RTCP/GPUProc/src/BeamFormer/IntToFloat.cl -text
-RTCP/GPUProc/src/BeamFormer/IntToFloat.cl-0.ptx -text
RTCP/GPUProc/src/BeamFormer/Transpose.cl -text
-RTCP/GPUProc/src/BeamFormer/Transpose.cl-0.ptx -text
RTCP/GPUProc/src/BeamletBuffer.cc -text
RTCP/GPUProc/src/BeamletBuffer.h -text
RTCP/GPUProc/src/BeamletBufferToComputeNode.cc -text
@@ -3682,14 +3676,11 @@ RTCP/GPUProc/src/BeamletBufferToComputeNode.h -text
RTCP/GPUProc/src/CL/cl.hpp -text
RTCP/GPUProc/src/CMakeLists.txt -text
RTCP/GPUProc/src/Correlator.cl -text
-RTCP/GPUProc/src/Correlator.cl-0.ptx -text
RTCP/GPUProc/src/DelayAndBandPass.cl -text
-RTCP/GPUProc/src/DelayAndBandPass.cl-0.ptx -text
RTCP/GPUProc/src/Delays.cc -text
RTCP/GPUProc/src/Delays.h -text
RTCP/GPUProc/src/FFT.cl -text
RTCP/GPUProc/src/FIR.cl -text
-RTCP/GPUProc/src/FIR.cl-0.ptx -text
RTCP/GPUProc/src/FilterBank.cc -text
RTCP/GPUProc/src/FilterBank.h -text
RTCP/GPUProc/src/InputSection.cc -text
@@ -3700,7 +3691,6 @@ RTCP/GPUProc/src/LockedRanges.h -text
RTCP/GPUProc/src/LogThread.cc -text
RTCP/GPUProc/src/LogThread.h -text
RTCP/GPUProc/src/NewCorrelator.cl -text
-RTCP/GPUProc/src/NewCorrelator.cl-0.ptx -text
RTCP/GPUProc/src/OpenCL_Support.cc -text
RTCP/GPUProc/src/OpenCL_Support.h -text
RTCP/GPUProc/src/OpenMP_Support.h -text
@@ -3714,29 +3704,22 @@ RTCP/GPUProc/src/Scheduling.cc -text
RTCP/GPUProc/src/Scheduling.h -text
RTCP/GPUProc/src/SlidingPointer.h -text
RTCP/GPUProc/src/UHEP/BeamFormer.cl -text
-RTCP/GPUProc/src/UHEP/BeamFormer.cl-0.ptx -text
RTCP/GPUProc/src/UHEP/BeamFormer.cl.4groups -text
RTCP/GPUProc/src/UHEP/BeamFormer.cl.hop -text
-RTCP/GPUProc/src/UHEP/BeamFormer.cl.hop-0.ptx -text
RTCP/GPUProc/src/UHEP/BeamFormer.cl.not -text
RTCP/GPUProc/src/UHEP/BeamFormer.cl.ok -text
RTCP/GPUProc/src/UHEP/InvFFT.cl -text
-RTCP/GPUProc/src/UHEP/InvFFT.cl-0.ptx -text
RTCP/GPUProc/src/UHEP/InvFIR.cl -text
-RTCP/GPUProc/src/UHEP/InvFIR.cl-0.ptx -text
RTCP/GPUProc/src/UHEP/InvertedStationPPFWeights.cc -text
RTCP/GPUProc/src/UHEP/InvertedStationPPFWeights.h -text
RTCP/GPUProc/src/UHEP/Transpose.cl -text
-RTCP/GPUProc/src/UHEP/Transpose.cl-0.ptx -text
RTCP/GPUProc/src/UHEP/Transpose.cl.ok -text
RTCP/GPUProc/src/UHEP/Trigger.cl -text
-RTCP/GPUProc/src/UHEP/Trigger.cl-0.ptx -text
RTCP/GPUProc/src/UHEP/Trigger.cl.8 -text
RTCP/GPUProc/src/UHEP/Trigger.cl.ok -text
RTCP/GPUProc/src/WallClockTime.h -text
RTCP/GPUProc/src/fft2.cl -text
RTCP/GPUProc/src/math.cl -text
-RTCP/GPUProc/src/octave-core -text
RTCP/GPUProc/test/77_Stations.parset -text
RTCP/GPUProc/test/AARTFAAC.parset -text
RTCP/GPUProc/test/CMakeLists.txt -text
diff --git a/RTCP/GPUProc/src/BeamFormer/BeamFormer.cl-0.ptx b/RTCP/GPUProc/src/BeamFormer/BeamFormer.cl-0.ptx
deleted file mode 100644
index 7f1d211419f4937459be9ecf7a4c5dbc50a45446..0000000000000000000000000000000000000000
--- a/RTCP/GPUProc/src/BeamFormer/BeamFormer.cl-0.ptx
+++ /dev/null
@@ -1,607 +0,0 @@
-//
-// Generated by NVIDIA NVVM Compiler
-// Compiler built on Sat Sep 29 10:12:13 2012 (1348906333)
-// Driver 304.54
-//
-
-.version 3.0
-.target sm_30, texmode_independent
-.address_size 32
-
-.extern .shared .align 16 .b8 shr_4__local[4096];
-
-.entry complexVoltages(
- .param .u32 .ptr .global .align 1 complexVoltages_param_0,
- .param .u32 .ptr .global .align 1 complexVoltages_param_1,
- .param .u32 .ptr .global .align 1 complexVoltages_param_2
-)
-{
- .reg .f32 %f<1124>;
- .reg .pred %p<13>;
- .reg .s32 %r<206>;
-
-
- ld.param.u32 %r1, [complexVoltages_param_0];
- ld.param.u32 %r3, [complexVoltages_param_2];
- // inline asm
- mov.u32 %r80, %tid.x;
- // inline asm
- // inline asm
- mov.u32 %r81, %tid.y;
- // inline asm
- // inline asm
- mov.u32 %r82, %envreg5;
- // inline asm
- // inline asm
- mov.u32 %r83, %ntid.z;
- // inline asm
- // inline asm
- mov.u32 %r84, %ctaid.z;
- // inline asm
- // inline asm
- mov.u32 %r85, %tid.z;
- // inline asm
- add.s32 %r87, %r85, %r82;
- mad.lo.s32 %r10, %r84, %r83, %r87;
- shl.b32 %r88, %r10, 10;
- shl.b32 %r89, %r81, 3;
- add.s32 %r90, %r88, %r89;
- add.s32 %r91, %r3, %r90;
- ld.global.v2.f32 {%f1028, %f1029}, [%r91];
- ld.global.v2.f32 {%f1034, %f1035}, [%r91+2097152];
- ld.global.v2.f32 {%f1040, %f1041}, [%r91+4194304];
- ld.global.v2.f32 {%f1046, %f1047}, [%r91+6291456];
- ld.global.v2.f32 {%f1052, %f1053}, [%r91+8388608];
- ld.global.v2.f32 {%f1058, %f1059}, [%r91+10485760];
- ld.global.v2.f32 {%f1064, %f1065}, [%r91+12582912];
- ld.global.v2.f32 {%f1070, %f1071}, [%r91+14680064];
- ld.global.v2.f32 {%f1076, %f1077}, [%r91+16777216];
- ld.global.v2.f32 {%f1082, %f1083}, [%r91+18874368];
- ld.global.v2.f32 {%f1088, %f1089}, [%r91+20971520];
- ld.global.v2.f32 {%f1094, %f1095}, [%r91+23068672];
- ld.global.v2.f32 {%f1100, %f1101}, [%r91+25165824];
- ld.global.v2.f32 {%f1106, %f1107}, [%r91+27262976];
- ld.global.v2.f32 {%f1112, %f1113}, [%r91+29360128];
- ld.global.v2.f32 {%f1118, %f1119}, [%r91+31457280];
- shl.b32 %r92, %r10, 16;
- shl.b32 %r93, %r81, 4;
- add.s32 %r94, %r92, %r93;
- shl.b32 %r95, %r80, 3;
- add.s32 %r96, %r94, %r95;
- add.s32 %r189, %r1, %r96;
- mov.u32 %r97, shr_4__local;
- add.s32 %r98, %r97, %r95;
- add.s32 %r12, %r98, 2048;
- mov.u32 %r182, 0;
-
-BB0_1:
- mov.u32 %r187, %r189;
- mov.u32 %r13, %r187;
- shl.b32 %r15, %r182, 4;
- // inline asm
- mov.u32 %r99, %tid.x;
- // inline asm
- // inline asm
- mov.u32 %r100, %tid.y;
- // inline asm
- shl.b32 %r17, %r100, 1;
- add.s32 %r184, %r17, %r99;
- setp.gt.u32 %p1, %r184, 255;
- @%p1 bra BB0_4;
-
- add.s32 %r19, %r99, %r17;
- and.b32 %r102, %r19, 15;
- add.s32 %r20, %r15, %r102;
- mov.u32 %r183, 0;
-
-BB0_3:
- mov.u32 %r22, %r184;
- add.s32 %r103, %r19, %r183;
- and.b32 %r104, %r103, 65520;
- shl.b32 %r105, %r104, 16;
- ld.param.u32 %r179, [complexVoltages_param_1];
- add.s32 %r106, %r179, %r105;
- shl.b32 %r107, %r10, 9;
- add.s32 %r108, %r106, %r107;
- shl.b32 %r109, %r20, 4;
- add.s32 %r110, %r108, %r109;
- shl.b32 %r111, %r22, 4;
- add.s32 %r113, %r97, %r111;
- ld.global.v4.f32 {%f1024, %f1025, %f1026, %f1027}, [%r110];
- st.shared.v4.f32 [%r113], {%f1024, %f1025, %f1026, %f1027};
- add.s32 %r23, %r22, 256;
- add.s32 %r183, %r183, 256;
- setp.gt.u32 %p2, %r22, -257;
- mov.u32 %r184, %r23;
- @%p2 bra BB0_3;
-
-BB0_4:
- mov.u32 %r186, %r12;
- bar.sync 0;
- mov.u32 %r185, 16;
- mov.u32 %r188, %r13;
-
-BB0_5:
- mov.u32 %r28, %r188;
- ld.shared.v2.f32 {%f798, %f799}, [%r186+-2048];
- mov.f32 %f1, 0f00000000;
- fma.rn.ftz.f32 %f802, %f1028, %f798, %f1;
- fma.rn.ftz.f32 %f803, %f1028, %f799, %f1;
- neg.ftz.f32 %f3, %f799;
- fma.rn.ftz.f32 %f810, %f1029, %f3, %f802;
- fma.rn.ftz.f32 %f811, %f1029, %f798, %f803;
- ld.shared.v2.f32 {%f814, %f815}, [%r186+-1792];
- fma.rn.ftz.f32 %f816, %f1034, %f814, %f810;
- fma.rn.ftz.f32 %f817, %f1034, %f815, %f811;
- neg.ftz.f32 %f6, %f815;
- fma.rn.ftz.f32 %f824, %f1035, %f6, %f816;
- fma.rn.ftz.f32 %f825, %f1035, %f814, %f817;
- ld.shared.v2.f32 {%f828, %f829}, [%r186+-1536];
- fma.rn.ftz.f32 %f830, %f1040, %f828, %f824;
- fma.rn.ftz.f32 %f831, %f1040, %f829, %f825;
- neg.ftz.f32 %f9, %f829;
- fma.rn.ftz.f32 %f838, %f1041, %f9, %f830;
- fma.rn.ftz.f32 %f839, %f1041, %f828, %f831;
- ld.shared.v2.f32 {%f842, %f843}, [%r186+-1280];
- fma.rn.ftz.f32 %f844, %f1046, %f842, %f838;
- fma.rn.ftz.f32 %f845, %f1046, %f843, %f839;
- neg.ftz.f32 %f12, %f843;
- fma.rn.ftz.f32 %f852, %f1047, %f12, %f844;
- fma.rn.ftz.f32 %f853, %f1047, %f842, %f845;
- ld.shared.v2.f32 {%f856, %f857}, [%r186+-1024];
- fma.rn.ftz.f32 %f858, %f1052, %f856, %f852;
- fma.rn.ftz.f32 %f859, %f1052, %f857, %f853;
- neg.ftz.f32 %f15, %f857;
- fma.rn.ftz.f32 %f866, %f1053, %f15, %f858;
- fma.rn.ftz.f32 %f867, %f1053, %f856, %f859;
- ld.shared.v2.f32 {%f870, %f871}, [%r186+-768];
- fma.rn.ftz.f32 %f872, %f1058, %f870, %f866;
- fma.rn.ftz.f32 %f873, %f1058, %f871, %f867;
- neg.ftz.f32 %f18, %f871;
- fma.rn.ftz.f32 %f880, %f1059, %f18, %f872;
- fma.rn.ftz.f32 %f881, %f1059, %f870, %f873;
- ld.shared.v2.f32 {%f884, %f885}, [%r186+-512];
- fma.rn.ftz.f32 %f886, %f1064, %f884, %f880;
- fma.rn.ftz.f32 %f887, %f1064, %f885, %f881;
- neg.ftz.f32 %f21, %f885;
- fma.rn.ftz.f32 %f894, %f1065, %f21, %f886;
- fma.rn.ftz.f32 %f895, %f1065, %f884, %f887;
- ld.shared.v2.f32 {%f898, %f899}, [%r186+-256];
- fma.rn.ftz.f32 %f900, %f1070, %f898, %f894;
- fma.rn.ftz.f32 %f901, %f1070, %f899, %f895;
- neg.ftz.f32 %f24, %f899;
- fma.rn.ftz.f32 %f908, %f1071, %f24, %f900;
- fma.rn.ftz.f32 %f909, %f1071, %f898, %f901;
- ld.shared.v2.f32 {%f912, %f913}, [%r186];
- fma.rn.ftz.f32 %f914, %f1076, %f912, %f908;
- fma.rn.ftz.f32 %f915, %f1076, %f913, %f909;
- neg.ftz.f32 %f27, %f913;
- fma.rn.ftz.f32 %f922, %f1077, %f27, %f914;
- fma.rn.ftz.f32 %f923, %f1077, %f912, %f915;
- ld.shared.v2.f32 {%f926, %f927}, [%r186+256];
- fma.rn.ftz.f32 %f928, %f1082, %f926, %f922;
- fma.rn.ftz.f32 %f929, %f1082, %f927, %f923;
- neg.ftz.f32 %f30, %f927;
- fma.rn.ftz.f32 %f936, %f1083, %f30, %f928;
- fma.rn.ftz.f32 %f937, %f1083, %f926, %f929;
- ld.shared.v2.f32 {%f940, %f941}, [%r186+512];
- fma.rn.ftz.f32 %f942, %f1088, %f940, %f936;
- fma.rn.ftz.f32 %f943, %f1088, %f941, %f937;
- neg.ftz.f32 %f33, %f941;
- fma.rn.ftz.f32 %f950, %f1089, %f33, %f942;
- fma.rn.ftz.f32 %f951, %f1089, %f940, %f943;
- ld.shared.v2.f32 {%f954, %f955}, [%r186+768];
- fma.rn.ftz.f32 %f956, %f1094, %f954, %f950;
- fma.rn.ftz.f32 %f957, %f1094, %f955, %f951;
- neg.ftz.f32 %f36, %f955;
- fma.rn.ftz.f32 %f964, %f1095, %f36, %f956;
- fma.rn.ftz.f32 %f965, %f1095, %f954, %f957;
- ld.shared.v2.f32 {%f968, %f969}, [%r186+1024];
- fma.rn.ftz.f32 %f970, %f1100, %f968, %f964;
- fma.rn.ftz.f32 %f971, %f1100, %f969, %f965;
- neg.ftz.f32 %f39, %f969;
- fma.rn.ftz.f32 %f978, %f1101, %f39, %f970;
- fma.rn.ftz.f32 %f979, %f1101, %f968, %f971;
- ld.shared.v2.f32 {%f982, %f983}, [%r186+1280];
- fma.rn.ftz.f32 %f984, %f1106, %f982, %f978;
- fma.rn.ftz.f32 %f985, %f1106, %f983, %f979;
- neg.ftz.f32 %f42, %f983;
- fma.rn.ftz.f32 %f992, %f1107, %f42, %f984;
- fma.rn.ftz.f32 %f993, %f1107, %f982, %f985;
- ld.shared.v2.f32 {%f996, %f997}, [%r186+1536];
- fma.rn.ftz.f32 %f998, %f1112, %f996, %f992;
- fma.rn.ftz.f32 %f999, %f1112, %f997, %f993;
- neg.ftz.f32 %f45, %f997;
- fma.rn.ftz.f32 %f1006, %f1113, %f45, %f998;
- fma.rn.ftz.f32 %f1007, %f1113, %f996, %f999;
- ld.shared.v2.f32 {%f1010, %f1011}, [%r186+1792];
- fma.rn.ftz.f32 %f1012, %f1118, %f1010, %f1006;
- fma.rn.ftz.f32 %f1013, %f1118, %f1011, %f1007;
- neg.ftz.f32 %f48, %f1011;
- fma.rn.ftz.f32 %f1020, %f1119, %f48, %f1012;
- fma.rn.ftz.f32 %f1021, %f1119, %f1010, %f1013;
- st.global.v2.f32 [%r28], {%f1020, %f1021};
- add.s32 %r29, %r28, 2048;
- add.s32 %r186, %r186, 16;
- add.s32 %r185, %r185, -1;
- setp.ne.s32 %p3, %r185, 0;
- mov.u32 %r188, %r29;
- @%p3 bra BB0_5;
-
- bar.sync 0;
- add.s32 %r182, %r182, 1;
- add.s32 %r33, %r13, 32768;
- setp.ne.s32 %p4, %r182, 2;
- mov.u32 %r189, %r33;
- @%p4 bra BB0_1;
-
- ld.param.u32 %r181, [complexVoltages_param_2];
- add.s32 %r119, %r181, %r90;
- ld.global.v2.f32 {%f702, %f703}, [%r119+33554432];
- ld.global.v2.f32 {%f708, %f709}, [%r119+35651584];
- ld.global.v2.f32 {%f714, %f715}, [%r119+37748736];
- ld.global.v2.f32 {%f720, %f721}, [%r119+39845888];
- ld.global.v2.f32 {%f726, %f727}, [%r119+41943040];
- ld.global.v2.f32 {%f732, %f733}, [%r119+44040192];
- ld.global.v2.f32 {%f738, %f739}, [%r119+46137344];
- ld.global.v2.f32 {%f744, %f745}, [%r119+48234496];
- ld.global.v2.f32 {%f750, %f751}, [%r119+50331648];
- ld.global.v2.f32 {%f756, %f757}, [%r119+52428800];
- ld.global.v2.f32 {%f762, %f763}, [%r119+54525952];
- ld.global.v2.f32 {%f768, %f769}, [%r119+56623104];
- ld.global.v2.f32 {%f774, %f775}, [%r119+58720256];
- ld.global.v2.f32 {%f780, %f781}, [%r119+60817408];
- ld.global.v2.f32 {%f786, %f787}, [%r119+62914560];
- ld.global.v2.f32 {%f792, %f793}, [%r119+65011712];
- ld.param.u32 %r176, [complexVoltages_param_0];
- add.s32 %r197, %r176, %r96;
- mov.u32 %r190, 0;
-
-BB0_8:
- mov.u32 %r195, %r197;
- mov.u32 %r36, %r195;
- shl.b32 %r38, %r190, 4;
- // inline asm
- mov.u32 %r129, %tid.x;
- // inline asm
- // inline asm
- mov.u32 %r130, %tid.y;
- // inline asm
- shl.b32 %r40, %r130, 1;
- add.s32 %r192, %r40, %r129;
- setp.gt.u32 %p5, %r192, 255;
- @%p5 bra BB0_11;
-
- add.s32 %r42, %r129, %r40;
- and.b32 %r132, %r42, 15;
- add.s32 %r43, %r38, %r132;
- mov.u32 %r191, 0;
-
-BB0_10:
- mov.u32 %r45, %r192;
- add.s32 %r133, %r42, %r191;
- shl.b32 %r134, %r43, 4;
- shl.b32 %r135, %r10, 9;
- shl.b32 %r136, %r45, 4;
- add.s32 %r138, %r97, %r136;
- and.b32 %r139, %r133, 65520;
- shl.b32 %r140, %r139, 16;
- ld.param.u32 %r178, [complexVoltages_param_1];
- add.s32 %r141, %r140, %r178;
- add.s32 %r142, %r141, %r135;
- add.s32 %r143, %r142, %r134;
- ld.global.v4.f32 {%f698, %f699, %f700, %f701}, [%r143+16777216];
- st.shared.v4.f32 [%r138], {%f698, %f699, %f700, %f701};
- add.s32 %r46, %r45, 256;
- add.s32 %r191, %r191, 256;
- setp.gt.u32 %p6, %r45, -257;
- mov.u32 %r192, %r46;
- @%p6 bra BB0_10;
-
-BB0_11:
- mov.u32 %r194, %r12;
- bar.sync 0;
- mov.u32 %r193, 16;
- mov.u32 %r196, %r36;
-
-BB0_12:
- mov.u32 %r51, %r196;
- ld.shared.v2.f32 {%f472, %f473}, [%r194+-2048];
- ld.global.v2.f32 {%f474, %f475}, [%r51];
- fma.rn.ftz.f32 %f476, %f702, %f472, %f474;
- fma.rn.ftz.f32 %f477, %f702, %f473, %f475;
- neg.ftz.f32 %f51, %f473;
- fma.rn.ftz.f32 %f484, %f703, %f51, %f476;
- fma.rn.ftz.f32 %f485, %f703, %f472, %f477;
- ld.shared.v2.f32 {%f488, %f489}, [%r194+-1792];
- fma.rn.ftz.f32 %f490, %f708, %f488, %f484;
- fma.rn.ftz.f32 %f491, %f708, %f489, %f485;
- neg.ftz.f32 %f54, %f489;
- fma.rn.ftz.f32 %f498, %f709, %f54, %f490;
- fma.rn.ftz.f32 %f499, %f709, %f488, %f491;
- ld.shared.v2.f32 {%f502, %f503}, [%r194+-1536];
- fma.rn.ftz.f32 %f504, %f714, %f502, %f498;
- fma.rn.ftz.f32 %f505, %f714, %f503, %f499;
- neg.ftz.f32 %f57, %f503;
- fma.rn.ftz.f32 %f512, %f715, %f57, %f504;
- fma.rn.ftz.f32 %f513, %f715, %f502, %f505;
- ld.shared.v2.f32 {%f516, %f517}, [%r194+-1280];
- fma.rn.ftz.f32 %f518, %f720, %f516, %f512;
- fma.rn.ftz.f32 %f519, %f720, %f517, %f513;
- neg.ftz.f32 %f60, %f517;
- fma.rn.ftz.f32 %f526, %f721, %f60, %f518;
- fma.rn.ftz.f32 %f527, %f721, %f516, %f519;
- ld.shared.v2.f32 {%f530, %f531}, [%r194+-1024];
- fma.rn.ftz.f32 %f532, %f726, %f530, %f526;
- fma.rn.ftz.f32 %f533, %f726, %f531, %f527;
- neg.ftz.f32 %f63, %f531;
- fma.rn.ftz.f32 %f540, %f727, %f63, %f532;
- fma.rn.ftz.f32 %f541, %f727, %f530, %f533;
- ld.shared.v2.f32 {%f544, %f545}, [%r194+-768];
- fma.rn.ftz.f32 %f546, %f732, %f544, %f540;
- fma.rn.ftz.f32 %f547, %f732, %f545, %f541;
- neg.ftz.f32 %f66, %f545;
- fma.rn.ftz.f32 %f554, %f733, %f66, %f546;
- fma.rn.ftz.f32 %f555, %f733, %f544, %f547;
- ld.shared.v2.f32 {%f558, %f559}, [%r194+-512];
- fma.rn.ftz.f32 %f560, %f738, %f558, %f554;
- fma.rn.ftz.f32 %f561, %f738, %f559, %f555;
- neg.ftz.f32 %f69, %f559;
- fma.rn.ftz.f32 %f568, %f739, %f69, %f560;
- fma.rn.ftz.f32 %f569, %f739, %f558, %f561;
- ld.shared.v2.f32 {%f572, %f573}, [%r194+-256];
- fma.rn.ftz.f32 %f574, %f744, %f572, %f568;
- fma.rn.ftz.f32 %f575, %f744, %f573, %f569;
- neg.ftz.f32 %f72, %f573;
- fma.rn.ftz.f32 %f582, %f745, %f72, %f574;
- fma.rn.ftz.f32 %f583, %f745, %f572, %f575;
- ld.shared.v2.f32 {%f586, %f587}, [%r194];
- fma.rn.ftz.f32 %f588, %f750, %f586, %f582;
- fma.rn.ftz.f32 %f589, %f750, %f587, %f583;
- neg.ftz.f32 %f75, %f587;
- fma.rn.ftz.f32 %f596, %f751, %f75, %f588;
- fma.rn.ftz.f32 %f597, %f751, %f586, %f589;
- ld.shared.v2.f32 {%f600, %f601}, [%r194+256];
- fma.rn.ftz.f32 %f602, %f756, %f600, %f596;
- fma.rn.ftz.f32 %f603, %f756, %f601, %f597;
- neg.ftz.f32 %f78, %f601;
- fma.rn.ftz.f32 %f610, %f757, %f78, %f602;
- fma.rn.ftz.f32 %f611, %f757, %f600, %f603;
- ld.shared.v2.f32 {%f614, %f615}, [%r194+512];
- fma.rn.ftz.f32 %f616, %f762, %f614, %f610;
- fma.rn.ftz.f32 %f617, %f762, %f615, %f611;
- neg.ftz.f32 %f81, %f615;
- fma.rn.ftz.f32 %f624, %f763, %f81, %f616;
- fma.rn.ftz.f32 %f625, %f763, %f614, %f617;
- ld.shared.v2.f32 {%f628, %f629}, [%r194+768];
- fma.rn.ftz.f32 %f630, %f768, %f628, %f624;
- fma.rn.ftz.f32 %f631, %f768, %f629, %f625;
- neg.ftz.f32 %f84, %f629;
- fma.rn.ftz.f32 %f638, %f769, %f84, %f630;
- fma.rn.ftz.f32 %f639, %f769, %f628, %f631;
- ld.shared.v2.f32 {%f642, %f643}, [%r194+1024];
- fma.rn.ftz.f32 %f644, %f774, %f642, %f638;
- fma.rn.ftz.f32 %f645, %f774, %f643, %f639;
- neg.ftz.f32 %f87, %f643;
- fma.rn.ftz.f32 %f652, %f775, %f87, %f644;
- fma.rn.ftz.f32 %f653, %f775, %f642, %f645;
- ld.shared.v2.f32 {%f656, %f657}, [%r194+1280];
- fma.rn.ftz.f32 %f658, %f780, %f656, %f652;
- fma.rn.ftz.f32 %f659, %f780, %f657, %f653;
- neg.ftz.f32 %f90, %f657;
- fma.rn.ftz.f32 %f666, %f781, %f90, %f658;
- fma.rn.ftz.f32 %f667, %f781, %f656, %f659;
- ld.shared.v2.f32 {%f670, %f671}, [%r194+1536];
- fma.rn.ftz.f32 %f672, %f786, %f670, %f666;
- fma.rn.ftz.f32 %f673, %f786, %f671, %f667;
- neg.ftz.f32 %f93, %f671;
- fma.rn.ftz.f32 %f680, %f787, %f93, %f672;
- fma.rn.ftz.f32 %f681, %f787, %f670, %f673;
- ld.shared.v2.f32 {%f684, %f685}, [%r194+1792];
- fma.rn.ftz.f32 %f686, %f792, %f684, %f680;
- fma.rn.ftz.f32 %f687, %f792, %f685, %f681;
- neg.ftz.f32 %f96, %f685;
- fma.rn.ftz.f32 %f694, %f793, %f96, %f686;
- fma.rn.ftz.f32 %f695, %f793, %f684, %f687;
- st.global.v2.f32 [%r51], {%f694, %f695};
- add.s32 %r52, %r51, 2048;
- add.s32 %r194, %r194, 16;
- add.s32 %r193, %r193, -1;
- setp.ne.s32 %p7, %r193, 0;
- mov.u32 %r196, %r52;
- @%p7 bra BB0_12;
-
- bar.sync 0;
- add.s32 %r190, %r190, 1;
- add.s32 %r56, %r36, 32768;
- setp.ne.s32 %p8, %r190, 2;
- mov.u32 %r197, %r56;
- @%p8 bra BB0_8;
-
- ld.param.u32 %r180, [complexVoltages_param_2];
- add.s32 %r149, %r180, %r90;
- ld.global.v2.f32 {%f376, %f377}, [%r149+67108864];
- ld.global.v2.f32 {%f382, %f383}, [%r149+69206016];
- ld.global.v2.f32 {%f388, %f389}, [%r149+71303168];
- ld.global.v2.f32 {%f394, %f395}, [%r149+73400320];
- ld.global.v2.f32 {%f400, %f401}, [%r149+75497472];
- ld.global.v2.f32 {%f406, %f407}, [%r149+77594624];
- ld.global.v2.f32 {%f412, %f413}, [%r149+79691776];
- ld.global.v2.f32 {%f418, %f419}, [%r149+81788928];
- ld.global.v2.f32 {%f424, %f425}, [%r149+83886080];
- ld.global.v2.f32 {%f430, %f431}, [%r149+85983232];
- ld.global.v2.f32 {%f436, %f437}, [%r149+88080384];
- ld.global.v2.f32 {%f442, %f443}, [%r149+90177536];
- ld.global.v2.f32 {%f448, %f449}, [%r149+92274688];
- ld.global.v2.f32 {%f454, %f455}, [%r149+94371840];
- ld.global.v2.f32 {%f460, %f461}, [%r149+96468992];
- ld.global.v2.f32 {%f466, %f467}, [%r149+98566144];
- ld.param.u32 %r175, [complexVoltages_param_0];
- add.s32 %r205, %r175, %r96;
- mov.u32 %r198, 0;
-
-BB0_15:
- mov.u32 %r203, %r205;
- mov.u32 %r59, %r203;
- shl.b32 %r61, %r198, 4;
- // inline asm
- mov.u32 %r159, %tid.x;
- // inline asm
- // inline asm
- mov.u32 %r160, %tid.y;
- // inline asm
- shl.b32 %r63, %r160, 1;
- add.s32 %r200, %r63, %r159;
- setp.gt.u32 %p9, %r200, 255;
- @%p9 bra BB0_18;
-
- add.s32 %r65, %r159, %r63;
- and.b32 %r162, %r65, 15;
- add.s32 %r66, %r61, %r162;
- mov.u32 %r199, 0;
-
-BB0_17:
- mov.u32 %r68, %r200;
- add.s32 %r163, %r65, %r199;
- shl.b32 %r164, %r66, 4;
- shl.b32 %r165, %r10, 9;
- shl.b32 %r166, %r68, 4;
- add.s32 %r168, %r97, %r166;
- and.b32 %r169, %r163, 65520;
- shl.b32 %r170, %r169, 16;
- ld.param.u32 %r177, [complexVoltages_param_1];
- add.s32 %r171, %r170, %r177;
- add.s32 %r172, %r171, %r165;
- add.s32 %r173, %r172, %r164;
- ld.global.v4.f32 {%f372, %f373, %f374, %f375}, [%r173+33554432];
- st.shared.v4.f32 [%r168], {%f372, %f373, %f374, %f375};
- add.s32 %r69, %r68, 256;
- add.s32 %r199, %r199, 256;
- setp.gt.u32 %p10, %r68, -257;
- mov.u32 %r200, %r69;
- @%p10 bra BB0_17;
-
-BB0_18:
- mov.u32 %r202, %r12;
- bar.sync 0;
- mov.u32 %r201, 16;
- mov.u32 %r204, %r59;
-
-BB0_19:
- mov.u32 %r74, %r204;
- ld.shared.v2.f32 {%f146, %f147}, [%r202+-2048];
- ld.global.v2.f32 {%f148, %f149}, [%r74];
- fma.rn.ftz.f32 %f150, %f376, %f146, %f148;
- fma.rn.ftz.f32 %f151, %f376, %f147, %f149;
- neg.ftz.f32 %f99, %f147;
- fma.rn.ftz.f32 %f158, %f377, %f99, %f150;
- fma.rn.ftz.f32 %f159, %f377, %f146, %f151;
- ld.shared.v2.f32 {%f162, %f163}, [%r202+-1792];
- fma.rn.ftz.f32 %f164, %f382, %f162, %f158;
- fma.rn.ftz.f32 %f165, %f382, %f163, %f159;
- neg.ftz.f32 %f102, %f163;
- fma.rn.ftz.f32 %f172, %f383, %f102, %f164;
- fma.rn.ftz.f32 %f173, %f383, %f162, %f165;
- ld.shared.v2.f32 {%f176, %f177}, [%r202+-1536];
- fma.rn.ftz.f32 %f178, %f388, %f176, %f172;
- fma.rn.ftz.f32 %f179, %f388, %f177, %f173;
- neg.ftz.f32 %f105, %f177;
- fma.rn.ftz.f32 %f186, %f389, %f105, %f178;
- fma.rn.ftz.f32 %f187, %f389, %f176, %f179;
- ld.shared.v2.f32 {%f190, %f191}, [%r202+-1280];
- fma.rn.ftz.f32 %f192, %f394, %f190, %f186;
- fma.rn.ftz.f32 %f193, %f394, %f191, %f187;
- neg.ftz.f32 %f108, %f191;
- fma.rn.ftz.f32 %f200, %f395, %f108, %f192;
- fma.rn.ftz.f32 %f201, %f395, %f190, %f193;
- ld.shared.v2.f32 {%f204, %f205}, [%r202+-1024];
- fma.rn.ftz.f32 %f206, %f400, %f204, %f200;
- fma.rn.ftz.f32 %f207, %f400, %f205, %f201;
- neg.ftz.f32 %f111, %f205;
- fma.rn.ftz.f32 %f214, %f401, %f111, %f206;
- fma.rn.ftz.f32 %f215, %f401, %f204, %f207;
- ld.shared.v2.f32 {%f218, %f219}, [%r202+-768];
- fma.rn.ftz.f32 %f220, %f406, %f218, %f214;
- fma.rn.ftz.f32 %f221, %f406, %f219, %f215;
- neg.ftz.f32 %f114, %f219;
- fma.rn.ftz.f32 %f228, %f407, %f114, %f220;
- fma.rn.ftz.f32 %f229, %f407, %f218, %f221;
- ld.shared.v2.f32 {%f232, %f233}, [%r202+-512];
- fma.rn.ftz.f32 %f234, %f412, %f232, %f228;
- fma.rn.ftz.f32 %f235, %f412, %f233, %f229;
- neg.ftz.f32 %f117, %f233;
- fma.rn.ftz.f32 %f242, %f413, %f117, %f234;
- fma.rn.ftz.f32 %f243, %f413, %f232, %f235;
- ld.shared.v2.f32 {%f246, %f247}, [%r202+-256];
- fma.rn.ftz.f32 %f248, %f418, %f246, %f242;
- fma.rn.ftz.f32 %f249, %f418, %f247, %f243;
- neg.ftz.f32 %f120, %f247;
- fma.rn.ftz.f32 %f256, %f419, %f120, %f248;
- fma.rn.ftz.f32 %f257, %f419, %f246, %f249;
- ld.shared.v2.f32 {%f260, %f261}, [%r202];
- fma.rn.ftz.f32 %f262, %f424, %f260, %f256;
- fma.rn.ftz.f32 %f263, %f424, %f261, %f257;
- neg.ftz.f32 %f123, %f261;
- fma.rn.ftz.f32 %f270, %f425, %f123, %f262;
- fma.rn.ftz.f32 %f271, %f425, %f260, %f263;
- ld.shared.v2.f32 {%f274, %f275}, [%r202+256];
- fma.rn.ftz.f32 %f276, %f430, %f274, %f270;
- fma.rn.ftz.f32 %f277, %f430, %f275, %f271;
- neg.ftz.f32 %f126, %f275;
- fma.rn.ftz.f32 %f284, %f431, %f126, %f276;
- fma.rn.ftz.f32 %f285, %f431, %f274, %f277;
- ld.shared.v2.f32 {%f288, %f289}, [%r202+512];
- fma.rn.ftz.f32 %f290, %f436, %f288, %f284;
- fma.rn.ftz.f32 %f291, %f436, %f289, %f285;
- neg.ftz.f32 %f129, %f289;
- fma.rn.ftz.f32 %f298, %f437, %f129, %f290;
- fma.rn.ftz.f32 %f299, %f437, %f288, %f291;
- ld.shared.v2.f32 {%f302, %f303}, [%r202+768];
- fma.rn.ftz.f32 %f304, %f442, %f302, %f298;
- fma.rn.ftz.f32 %f305, %f442, %f303, %f299;
- neg.ftz.f32 %f132, %f303;
- fma.rn.ftz.f32 %f312, %f443, %f132, %f304;
- fma.rn.ftz.f32 %f313, %f443, %f302, %f305;
- ld.shared.v2.f32 {%f316, %f317}, [%r202+1024];
- fma.rn.ftz.f32 %f318, %f448, %f316, %f312;
- fma.rn.ftz.f32 %f319, %f448, %f317, %f313;
- neg.ftz.f32 %f135, %f317;
- fma.rn.ftz.f32 %f326, %f449, %f135, %f318;
- fma.rn.ftz.f32 %f327, %f449, %f316, %f319;
- ld.shared.v2.f32 {%f330, %f331}, [%r202+1280];
- fma.rn.ftz.f32 %f332, %f454, %f330, %f326;
- fma.rn.ftz.f32 %f333, %f454, %f331, %f327;
- neg.ftz.f32 %f138, %f331;
- fma.rn.ftz.f32 %f340, %f455, %f138, %f332;
- fma.rn.ftz.f32 %f341, %f455, %f330, %f333;
- ld.shared.v2.f32 {%f344, %f345}, [%r202+1536];
- fma.rn.ftz.f32 %f346, %f460, %f344, %f340;
- fma.rn.ftz.f32 %f347, %f460, %f345, %f341;
- neg.ftz.f32 %f141, %f345;
- fma.rn.ftz.f32 %f354, %f461, %f141, %f346;
- fma.rn.ftz.f32 %f355, %f461, %f344, %f347;
- ld.shared.v2.f32 {%f358, %f359}, [%r202+1792];
- fma.rn.ftz.f32 %f360, %f466, %f358, %f354;
- fma.rn.ftz.f32 %f361, %f466, %f359, %f355;
- neg.ftz.f32 %f144, %f359;
- fma.rn.ftz.f32 %f368, %f467, %f144, %f360;
- fma.rn.ftz.f32 %f369, %f467, %f358, %f361;
- st.global.v2.f32 [%r74], {%f368, %f369};
- add.s32 %r75, %r74, 2048;
- add.s32 %r202, %r202, 16;
- add.s32 %r201, %r201, -1;
- setp.ne.s32 %p11, %r201, 0;
- mov.u32 %r204, %r75;
- @%p11 bra BB0_19;
-
- bar.sync 0;
- add.s32 %r198, %r198, 1;
- add.s32 %r79, %r59, 32768;
- setp.ne.s32 %p12, %r198, 2;
- mov.u32 %r205, %r79;
- @%p12 bra BB0_15;
-
- ret;
-}
-
-
diff --git a/RTCP/GPUProc/src/BeamFormer/CoherentStokes.cl-0.ptx b/RTCP/GPUProc/src/BeamFormer/CoherentStokes.cl-0.ptx
deleted file mode 100644
index 95682a21f13f73a46e51290282c947066bcf9e94..0000000000000000000000000000000000000000
Binary files a/RTCP/GPUProc/src/BeamFormer/CoherentStokes.cl-0.ptx and /dev/null differ
diff --git a/RTCP/GPUProc/src/BeamFormer/Dedispersion.cl-0.ptx b/RTCP/GPUProc/src/BeamFormer/Dedispersion.cl-0.ptx
deleted file mode 100644
index 221a542fb5fdfc1737730694aaf0921d00774dfb..0000000000000000000000000000000000000000
--- a/RTCP/GPUProc/src/BeamFormer/Dedispersion.cl-0.ptx
+++ /dev/null
@@ -1,417 +0,0 @@
-//
-// Generated by NVIDIA NVVM Compiler
-// Compiler built on Sat Sep 29 10:12:13 2012 (1348906333)
-// Driver 304.54
-//
-
-.version 3.0
-.target sm_30, texmode_independent
-.address_size 32
-
-.extern .shared .align 4 .b8 shr_2_local_DMs[512];
-
-.entry applyChirp(
- .param .u32 .ptr .global .align 1 applyChirp_param_0,
- .param .u32 .ptr .global .align 4 applyChirp_param_1,
- .param .f32 applyChirp_param_2
-)
-{
- .reg .f32 %f<225>;
- .reg .pred %p<26>;
- .reg .s32 %r<97>;
-
-
- // inline asm
- mov.u32 %r32, %tid.x;
- // inline asm
- setp.gt.s32 %p4, %r32, 127;
- mov.u32 %r92, %r32;
- @%p4 bra BB0_2;
-
-BB0_1:
- shl.b32 %r34, %r92, 2;
- ld.param.u32 %r91, [applyChirp_param_1];
- add.s32 %r35, %r91, %r34;
- ld.global.f32 %f29, [%r35];
- add.ftz.f32 %f30, %f29, %f29;
- mul.ftz.f32 %f31, %f30, 0f40490FDB;
- mul.ftz.f32 %f32, %f31, 0f596BD7E5;
- mov.u32 %r36, shr_2_local_DMs;
- add.s32 %r37, %r36, %r34;
- st.shared.f32 [%r37], %f32;
- // inline asm
- mov.u32 %r33, %ntid.x;
- // inline asm
- add.s32 %r92, %r33, %r92;
- setp.lt.s32 %p5, %r92, 128;
- @%p5 bra BB0_1;
-
-BB0_2:
- bar.sync 0;
- // inline asm
- mov.u32 %r38, %envreg3;
- // inline asm
- // inline asm
- mov.u32 %r39, %ntid.x;
- // inline asm
- // inline asm
- mov.u32 %r40, %ctaid.x;
- // inline asm
- // inline asm
- mov.u32 %r41, %tid.x;
- // inline asm
- add.s32 %r50, %r41, %r38;
- mad.lo.s32 %r51, %r40, %r39, %r50;
- // inline asm
- mov.u32 %r42, %envreg4;
- // inline asm
- // inline asm
- mov.u32 %r43, %ntid.y;
- // inline asm
- // inline asm
- mov.u32 %r44, %ctaid.y;
- // inline asm
- // inline asm
- mov.u32 %r45, %tid.y;
- // inline asm
- // inline asm
- mov.u32 %r46, %envreg5;
- // inline asm
- // inline asm
- mov.u32 %r47, %ntid.z;
- // inline asm
- // inline asm
- mov.u32 %r48, %ctaid.z;
- // inline asm
- // inline asm
- mov.u32 %r49, %tid.z;
- // inline asm
- add.s32 %r52, %r49, %r46;
- mad.lo.s32 %r53, %r48, %r47, %r52;
- cvt.rn.f32.u32 %f35, %r53;
- ld.param.f32 %f217, [applyChirp_param_2];
- add.ftz.f32 %f36, %f217, 0fC7BEBC20;
- fma.rn.ftz.f32 %f2, %f35, 0f42BEBC20, %f36;
- cvt.rn.f32.u32 %f37, %r51;
- mul.ftz.f32 %f38, %f37, 0f403EBC20;
- setp.gt.u32 %p6, %r51, 32;
- fma.rn.ftz.f32 %f39, %f37, 0f403EBC20, 0fC2BEBC20;
- selp.f32 %f3, %f39, %f38, %p6;
- div.rn.ftz.f32 %f34, %f3, 0f42334A70;
- // inline asm
- abs.f32 %f33, %f34;
- // inline asm
- setp.eq.ftz.f32 %p7, %f34, 0f3F800000;
- @%p7 bra BB0_25;
-
- setp.nan.ftz.f32 %p8, %f34, %f34;
- @%p8 bra BB0_24;
-
- mov.f32 %f6, 0fFF800000;
- mov.f32 %f220, 0f42A00000;
- mov.f32 %f44, 0f3F000000;
- mul.rn.f32 %f41, %f44, %f220;
- // inline asm
- cvt.rmi.f32.f32 %f40, %f41;
- // inline asm
- mov.f32 %f45, 0f40000000;
- mul.rn.f32 %f46, %f45, %f40;
- sub.ftz.f32 %f47, %f220, %f46;
- setp.eq.ftz.f32 %p1, %f47, 0f3F800000;
- // inline asm
- cvt.rzi.f32.f32 %f42, %f220;
- // inline asm
- setp.eq.ftz.f32 %p2, %f42, 0f42A00000;
- and.pred %p3, %p1, %p2;
- setp.eq.ftz.f32 %p9, %f33, 0f00000000;
- @%p9 bra BB0_23;
-
- setp.eq.ftz.f32 %p10, %f34, 0f7F800000;
- setp.eq.ftz.f32 %p11, %f34, %f6;
- or.pred %p12, %p10, %p11;
- @%p12 bra BB0_20;
-
- setp.geu.ftz.f32 %p13, %f34, 0f00000000;
- @%p13 bra BB0_8;
-
- mov.f32 %f49, 0f42A00000;
- // inline asm
- cvt.rzi.f32.f32 %f48, %f49;
- // inline asm
- setp.neu.ftz.f32 %p14, %f48, 0f42A00000;
- @%p14 bra BB0_19;
-
-BB0_8:
- // inline asm
- abs.f32 %f50, %f34;
- // inline asm
- mov.b32 %r18, %f50;
- shr.u32 %r54, %r18, 23;
- and.b32 %r55, %r54, 255;
- add.s32 %r93, %r55, -127;
- setp.eq.s32 %p15, %r55, 0;
- mov.f32 %f218, %f50;
- @%p15 bra BB0_9;
- bra.uni BB0_10;
-
-BB0_9:
- and.b32 %r56, %r18, -2139095041;
- or.b32 %r57, %r56, 1065353216;
- mov.b32 %f52, %r57;
- add.ftz.f32 %f53, %f52, 0fBF800000;
- mov.b32 %r58, %f53;
- shr.u32 %r59, %r58, 23;
- and.b32 %r60, %r59, 255;
- add.s32 %r93, %r60, -253;
- and.b32 %r61, %r58, -2139095041;
- or.b32 %r62, %r61, 1065353216;
- mov.b32 %f218, %r62;
-
-BB0_10:
- mov.b32 %r63, %f218;
- and.b32 %r64, %r63, -2139095041;
- or.b32 %r65, %r64, 1065353216;
- mov.b32 %f219, %r65;
- setp.gt.ftz.f32 %p16, %f219, 0f3FB504F3;
- @%p16 bra BB0_11;
- bra.uni BB0_12;
-
-BB0_11:
- mul.rn.f32 %f219, %f219, %f44;
- add.s32 %r93, %r93, 1;
-
-BB0_12:
- add.ftz.f32 %f63, %f219, 0f3F800000;
- rcp.approx.ftz.f32 %f57, %f63;
- add.ftz.f32 %f56, %f219, 0fBF800000;
- // inline asm
- mul.rz.f32 %f55, %f56, %f57;
- // inline asm
- mul.rn.f32 %f65, %f45, %f55;
- mul.rn.f32 %f66, %f65, %f65;
- mov.f32 %f67, 0f3B18F0FE;
- mul.rn.f32 %f68, %f67, %f66;
- add.ftz.f32 %f69, %f68, 0f3C4CAF63;
- mul.rn.f32 %f70, %f69, %f66;
- add.ftz.f32 %f71, %f70, 0f3DAAAABD;
- mul.rn.f32 %f72, %f71, %f66;
- mul.rn.f32 %f60, %f72, %f65;
- mov.b32 %r66, %f65;
- and.b32 %r67, %r66, -4096;
- mov.b32 %f73, %r67;
- mov.b32 %r68, %f56;
- and.b32 %r69, %r68, -4096;
- mov.b32 %f74, %r69;
- sub.ftz.f32 %f75, %f56, %f73;
- mul.rn.f32 %f76, %f45, %f75;
- sub.ftz.f32 %f77, %f56, %f74;
- mul.rn.f32 %f78, %f73, %f74;
- sub.ftz.f32 %f79, %f76, %f78;
- mul.rn.f32 %f80, %f73, %f77;
- sub.ftz.f32 %f81, %f79, %f80;
- mul.rn.f32 %f82, %f57, %f81;
- add.ftz.f32 %f83, %f73, %f82;
- sub.ftz.f32 %f84, %f83, %f73;
- sub.ftz.f32 %f85, %f82, %f84;
- add.ftz.f32 %f86, %f83, %f60;
- neg.ftz.f32 %f59, %f60;
- // inline asm
- add.rz.f32 %f58, %f59, %f60;
- // inline asm
- add.ftz.f32 %f87, %f58, %f85;
- add.ftz.f32 %f88, %f86, %f87;
- sub.ftz.f32 %f89, %f87, %f87;
- cvt.rn.f32.s32 %f90, %r93;
- mov.f32 %f91, 0f3F317200;
- mul.rn.f32 %f92, %f90, %f91;
- mov.f32 %f93, 0f35BFBE8E;
- mul.rn.f32 %f94, %f90, %f93;
- add.ftz.f32 %f95, %f92, %f88;
- sub.ftz.f32 %f96, %f88, %f88;
- add.ftz.f32 %f97, %f96, %f89;
- add.ftz.f32 %f98, %f97, %f94;
- add.ftz.f32 %f13, %f95, %f98;
- sub.ftz.f32 %f14, %f98, %f98;
- // inline asm
- abs.f32 %f61, %f220;
- // inline asm
- setp.gt.ftz.f32 %p17, %f61, 0f77F684DF;
- @%p17 bra BB0_13;
- bra.uni BB0_14;
-
-BB0_13:
- mov.f32 %f100, 0f39000000;
- mov.f32 %f101, 0f42A00000;
- mul.rn.f32 %f220, %f101, %f100;
-
-BB0_14:
- mov.f32 %f102, 0f45800800;
- mul.rn.f32 %f103, %f13, %f102;
- sub.ftz.f32 %f104, %f13, %f103;
- add.ftz.f32 %f105, %f104, %f103;
- sub.ftz.f32 %f106, %f13, %f105;
- mul.rn.f32 %f107, %f220, %f102;
- sub.ftz.f32 %f108, %f220, %f107;
- add.ftz.f32 %f109, %f108, %f107;
- sub.ftz.f32 %f110, %f220, %f109;
- mul.rn.f32 %f111, %f105, %f109;
- mul.rn.f32 %f112, %f13, %f220;
- sub.ftz.f32 %f113, %f111, %f112;
- mul.rn.f32 %f114, %f105, %f110;
- add.ftz.f32 %f115, %f113, %f114;
- mul.rn.f32 %f116, %f106, %f109;
- add.ftz.f32 %f117, %f115, %f116;
- mul.rn.f32 %f118, %f106, %f110;
- add.ftz.f32 %f119, %f117, %f118;
- mul.rn.f32 %f120, %f14, %f220;
- add.ftz.f32 %f121, %f120, %f119;
- add.ftz.f32 %f122, %f112, %f121;
- sub.ftz.f32 %f17, %f121, %f121;
- mov.f32 %f223, %f17;
- mov.f32 %f224, %f122;
- mov.b32 %r24, %f122;
- setp.eq.s32 %p18, %r24, 1118925336;
- @%p18 bra BB0_15;
- bra.uni BB0_16;
-
-BB0_15:
- add.s32 %r70, %r24, -1;
- mov.b32 %f123, %r70;
- add.ftz.f32 %f124, %f17, 0f37000000;
- mov.f32 %f223, %f124;
- mov.f32 %f224, %f123;
-
-BB0_16:
- // inline asm
- mul.f32 %f125, %f224, 0f3FB8AA3B;ex2.approx.f32 %f125, %f125;
- // inline asm
- setp.neu.ftz.f32 %p19, %f125, 0f7F800000;
- mov.f32 %f221, %f125;
- @%p19 bra BB0_17;
- bra.uni BB0_18;
-
-BB0_17:
- // inline asm
- mad.f32 %f127, %f125, %f223, %f125;
- // inline asm
- mov.f32 %f221, %f127;
-
-BB0_18:
- not.pred %p21, %p3;
- or.pred %p23, %p13, %p21;
- mov.b32 %r71, %f221;
- xor.b32 %r72, %r71, -2147483648;
- mov.b32 %f131, %r72;
- selp.f32 %f222, %f221, %f131, %p23;
- bra.uni BB0_26;
-
-BB0_19:
- mov.f32 %f222, 0f7FFFFFFF;
- bra.uni BB0_26;
-
-BB0_20:
- mov.b32 %r73, %f34;
- setp.lt.s32 %p24, %r73, 0;
- @%p24 bra BB0_22;
-
- mov.f32 %f222, 0f7F800000;
- bra.uni BB0_26;
-
-BB0_22:
- selp.f32 %f222, 0fFF800000, 0f7F800000, %p3;
- bra.uni BB0_26;
-
-BB0_23:
- mov.b32 %r74, %f34;
- and.b32 %r75, %r74, -2147483648;
- mov.b32 %f132, %r75;
- selp.f32 %f222, %f132, 0f00000000, %p3;
- bra.uni BB0_26;
-
-BB0_24:
- add.ftz.f32 %f222, %f34, 0f42A00000;
- bra.uni BB0_26;
-
-BB0_25:
- mov.f32 %f222, 0f3F800000;
-
-BB0_26:
- add.ftz.f32 %f135, %f222, 0f3F800000;
- // inline asm
- rsqrt.approx.f32 %f134, %f135;
- // inline asm
- mul.ftz.f32 %f136, %f134, 0f42000000;
- div.approx.ftz.f32 %f137, %f3, %f2;
- mul.ftz.f32 %f138, %f137, %f137;
- add.ftz.f32 %f139, %f2, %f3;
- div.approx.ftz.f32 %f28, %f138, %f139;
- add.s32 %r78, %r45, %r42;
- mad.lo.s32 %r79, %r44, %r43, %r78;
- shl.b32 %r80, %r79, 8;
- shl.b32 %r83, %r53, 8;
- add.s32 %r84, %r80, %r83;
- shl.b32 %r87, %r51, 3;
- add.s32 %r88, %r84, %r87;
- ld.param.u32 %r90, [applyChirp_param_0];
- add.s32 %r89, %r88, %r90;
- add.s32 %r95, %r89, 1048576;
- mov.u32 %r96, 128;
- mov.u32 %r94, shr_2_local_DMs;
-
-BB0_27:
- ld.global.v2.f32 {%f181, %f182}, [%r95+-1048576];
- ld.global.v2.f32 {%f183, %f184}, [%r95+-524288];
- ld.shared.f32 %f149, [%r94];
- mul.ftz.f32 %f144, %f149, %f28;
- // inline asm
- cos.approx.f32 %f141, %f144;
- // inline asm
- // inline asm
- sin.approx.f32 %f143, %f144;
- // inline asm
- mul.ftz.f32 %f187, %f141, %f136;
- mul.ftz.f32 %f188, %f143, %f136;
- mul.ftz.f32 %f152, %f187, %f181;
- neg.f32 %f155, %f188;
- fma.rn.ftz.f32 %f156, %f155, %f182, %f152;
- mul.ftz.f32 %f157, %f188, %f181;
- fma.rn.ftz.f32 %f158, %f187, %f182, %f157;
- st.global.v2.f32 [%r95+-1048576], {%f156, %f158};
- mul.ftz.f32 %f160, %f187, %f183;
- fma.rn.ftz.f32 %f162, %f155, %f184, %f160;
- mul.ftz.f32 %f163, %f188, %f183;
- fma.rn.ftz.f32 %f164, %f187, %f184, %f163;
- st.global.v2.f32 [%r95+-524288], {%f162, %f164};
- ld.global.v2.f32 {%f195, %f196}, [%r95];
- ld.global.v2.f32 {%f197, %f198}, [%r95+524288];
- ld.shared.f32 %f165, [%r94+4];
- mul.ftz.f32 %f148, %f165, %f28;
- // inline asm
- cos.approx.f32 %f145, %f148;
- // inline asm
- // inline asm
- sin.approx.f32 %f147, %f148;
- // inline asm
- mul.ftz.f32 %f201, %f145, %f136;
- mul.ftz.f32 %f202, %f147, %f136;
- mul.ftz.f32 %f168, %f201, %f195;
- neg.f32 %f171, %f202;
- fma.rn.ftz.f32 %f172, %f171, %f196, %f168;
- mul.ftz.f32 %f173, %f202, %f195;
- fma.rn.ftz.f32 %f174, %f201, %f196, %f173;
- st.global.v2.f32 [%r95], {%f172, %f174};
- mul.ftz.f32 %f176, %f201, %f197;
- fma.rn.ftz.f32 %f178, %f171, %f198, %f176;
- mul.ftz.f32 %f179, %f202, %f197;
- fma.rn.ftz.f32 %f180, %f201, %f198, %f179;
- st.global.v2.f32 [%r95+524288], {%f178, %f180};
- add.s32 %r95, %r95, 2097152;
- add.s32 %r94, %r94, 8;
- add.s32 %r96, %r96, -2;
- setp.ne.s32 %p25, %r96, 0;
- @%p25 bra BB0_27;
-
- ret;
-}
-
-
diff --git a/RTCP/GPUProc/src/BeamFormer/IncoherentStokes.cl-0.ptx b/RTCP/GPUProc/src/BeamFormer/IncoherentStokes.cl-0.ptx
deleted file mode 100644
index be0613c10bbe025940d0054bfd9e8579149984e3..0000000000000000000000000000000000000000
Binary files a/RTCP/GPUProc/src/BeamFormer/IncoherentStokes.cl-0.ptx and /dev/null differ
diff --git a/RTCP/GPUProc/src/BeamFormer/IntToFloat.cl-0.ptx b/RTCP/GPUProc/src/BeamFormer/IntToFloat.cl-0.ptx
deleted file mode 100644
index ce5c60b9ac9b08ef14ba9402e30163b4bede74bb..0000000000000000000000000000000000000000
--- a/RTCP/GPUProc/src/BeamFormer/IntToFloat.cl-0.ptx
+++ /dev/null
@@ -1,82 +0,0 @@
-//
-// Generated by NVIDIA NVVM Compiler
-// Compiler built on Sat Sep 29 10:12:13 2012 (1348906333)
-// Driver 304.54
-//
-
-.version 3.0
-.target sm_30, texmode_independent
-.address_size 32
-
-
-.entry intToFloat(
- .param .u32 .ptr .global .align 1 intToFloat_param_0,
- .param .u32 .ptr .global .align 1 intToFloat_param_1
-)
-{
- .reg .f32 %f<9>;
- .reg .pred %p<3>;
- .reg .s32 %r<25>;
- .reg .s16 %rc<9>;
-
-
- // inline asm
- mov.u32 %r7, %envreg4;
- // inline asm
- // inline asm
- mov.u32 %r8, %ntid.y;
- // inline asm
- // inline asm
- mov.u32 %r9, %ctaid.y;
- // inline asm
- // inline asm
- mov.u32 %r10, %tid.y;
- // inline asm
- add.s32 %r12, %r10, %r7;
- mad.lo.s32 %r3, %r9, %r8, %r12;
- // inline asm
- mov.u32 %r11, %tid.x;
- // inline asm
- setp.gt.u32 %p1, %r11, 65535;
- mov.u32 %r24, %r11;
- @%p1 bra BB0_2;
-
-BB0_1:
- shl.b32 %r14, %r3, 18;
- ld.param.u32 %r23, [intToFloat_param_1];
- add.s32 %r15, %r23, %r14;
- shl.b32 %r16, %r24, 2;
- add.s32 %r17, %r15, %r16;
- ld.global.v2.u8 {%rc5, %rc6}, [%r17];
- // inline asm
- cvt.rn.f32.s8 %f1, %rc5;
- // inline asm
- // inline asm
- cvt.rn.f32.s8 %f2, %rc6;
- // inline asm
- shl.b32 %r18, %r3, 20;
- ld.param.u32 %r22, [intToFloat_param_0];
- add.s32 %r19, %r22, %r18;
- shl.b32 %r20, %r24, 3;
- add.s32 %r21, %r19, %r20;
- st.global.v2.f32 [%r21], {%f1, %f2};
- ld.global.v2.u8 {%rc7, %rc8}, [%r17+2];
- // inline asm
- cvt.rn.f32.s8 %f3, %rc7;
- // inline asm
- // inline asm
- cvt.rn.f32.s8 %f4, %rc8;
- // inline asm
- st.global.v2.f32 [%r21+524288], {%f3, %f4};
- // inline asm
- mov.u32 %r13, %ntid.x;
- // inline asm
- add.s32 %r24, %r13, %r24;
- setp.lt.u32 %p2, %r24, 65536;
- @%p2 bra BB0_1;
-
-BB0_2:
- ret;
-}
-
-
diff --git a/RTCP/GPUProc/src/BeamFormer/Transpose.cl-0.ptx b/RTCP/GPUProc/src/BeamFormer/Transpose.cl-0.ptx
deleted file mode 100644
index 26a5c9cca9b954543c29abfbb6bb29d45835a803..0000000000000000000000000000000000000000
--- a/RTCP/GPUProc/src/BeamFormer/Transpose.cl-0.ptx
+++ /dev/null
@@ -1,134 +0,0 @@
-//
-// Generated by NVIDIA NVVM Compiler
-// Compiler built on Sat Sep 29 10:12:13 2012 (1348906333)
-// Driver 304.54
-//
-
-.version 3.0
-.target sm_30, texmode_independent
-.address_size 32
-
-.extern .shared .align 16 .b8 shr_3_tmp[4352];
-
-.entry transposeComplexVoltages(
- .param .u32 .ptr .global .align 1 transposeComplexVoltages_param_0,
- .param .u32 .ptr .global .align 1 transposeComplexVoltages_param_1
-)
-{
- .reg .f32 %f<65>;
- .reg .pred %p<2>;
- .reg .s32 %r<58>;
-
-
- ld.param.u32 %r26, [transposeComplexVoltages_param_0];
- ld.param.u32 %r27, [transposeComplexVoltages_param_1];
- // inline asm
- mov.u32 %r13, %envreg4;
- // inline asm
- // inline asm
- mov.u32 %r14, %ntid.y;
- // inline asm
- // inline asm
- mov.u32 %r15, %ctaid.y;
- // inline asm
- // inline asm
- mov.u32 %r16, %tid.y;
- // inline asm
- // inline asm
- mov.u32 %r17, %envreg5;
- // inline asm
- // inline asm
- mov.u32 %r18, %ntid.z;
- // inline asm
- // inline asm
- mov.u32 %r19, %ctaid.z;
- // inline asm
- // inline asm
- mov.u32 %r20, %tid.z;
- // inline asm
- // inline asm
- mov.u32 %r21, %tid.x;
- // inline asm
- // inline asm
- mov.u32 %r22, %tid.x;
- // inline asm
- shr.u32 %r28, %r22, 4;
- // inline asm
- mov.u32 %r23, %tid.x;
- // inline asm
- shr.u32 %r29, %r23, 4;
- // inline asm
- mov.u32 %r24, %tid.x;
- // inline asm
- and.b32 %r30, %r21, 15;
- mov.u32 %r31, shr_3_tmp;
- mad.lo.s32 %r32, %r30, 272, %r31;
- and.b32 %r33, %r22, -16;
- add.s32 %r1, %r32, %r33;
- mad.lo.s32 %r34, %r29, 272, %r31;
- and.b32 %r35, %r24, 15;
- shl.b32 %r36, %r35, 4;
- add.s32 %r2, %r34, %r36;
- add.s32 %r37, %r20, %r17;
- mad.lo.s32 %r38, %r19, %r18, %r37;
- shl.b32 %r39, %r38, 4;
- add.s32 %r40, %r28, %r39;
- shl.b32 %r41, %r40, 16;
- add.s32 %r42, %r16, %r13;
- mad.lo.s32 %r43, %r15, %r14, %r42;
- shl.b32 %r44, %r43, 4;
- add.s32 %r45, %r44, %r30;
- shl.b32 %r46, %r45, 4;
- add.s32 %r47, %r41, %r46;
- add.s32 %r48, %r47, %r27;
- add.s32 %r56, %r48, 4096;
- add.s32 %r49, %r29, %r44;
- shl.b32 %r50, %r49, 20;
- add.s32 %r51, %r39, %r35;
- shl.b32 %r52, %r51, 3;
- add.s32 %r53, %r50, %r52;
- add.s32 %r54, %r53, %r26;
- add.s32 %r55, %r54, 525056;
- mov.u32 %r57, 2048;
-
-BB0_1:
- add.s32 %r8, %r56, -4096;
- ld.global.v4.f32 {%f61, %f62, %f63, %f64}, [%r56+-4096];
- st.shared.v4.f32 [%r1], {%f61, %f62, %f63, %f64};
- bar.sync 0;
- ld.shared.v4.f32 {%f53, %f54, %f55, %f56}, [%r2];
- add.s32 %r9, %r55, -525056;
- st.global.v2.f32 [%r55+-525056], {%f53, %f54};
- st.global.v2.f32 [%r55+-768], {%f55, %f56};
- bar.sync 0;
- ld.global.v4.f32 {%f49, %f50, %f51, %f52}, [%r8+2048];
- st.shared.v4.f32 [%r1], {%f49, %f50, %f51, %f52};
- bar.sync 0;
- ld.shared.v4.f32 {%f41, %f42, %f43, %f44}, [%r2];
- st.global.v2.f32 [%r9+256], {%f41, %f42};
- st.global.v2.f32 [%r9+524544], {%f43, %f44};
- bar.sync 0;
- ld.global.v4.f32 {%f37, %f38, %f39, %f40}, [%r8+4096];
- st.shared.v4.f32 [%r1], {%f37, %f38, %f39, %f40};
- bar.sync 0;
- ld.shared.v4.f32 {%f29, %f30, %f31, %f32}, [%r2];
- st.global.v2.f32 [%r9+512], {%f29, %f30};
- st.global.v2.f32 [%r9+524800], {%f31, %f32};
- bar.sync 0;
- ld.global.v4.f32 {%f25, %f26, %f27, %f28}, [%r8+6144];
- st.shared.v4.f32 [%r1], {%f25, %f26, %f27, %f28};
- bar.sync 0;
- ld.shared.v4.f32 {%f17, %f18, %f19, %f20}, [%r2];
- st.global.v2.f32 [%r9+768], {%f17, %f18};
- st.global.v2.f32 [%r9+525056], {%f19, %f20};
- bar.sync 0;
- add.s32 %r56, %r56, 8192;
- add.s32 %r55, %r55, 1024;
- add.s32 %r57, %r57, -4;
- setp.ne.s32 %p1, %r57, 0;
- @%p1 bra BB0_1;
-
- ret;
-}
-
-
diff --git a/RTCP/GPUProc/src/CL/cl.hpp b/RTCP/GPUProc/src/CL/cl.hpp
index 5111f37dc8c74a540999ff6a06b9ae7ec07b6455..aa2b0bd040fd81ed57f433c242dcc1b39693f35a 100644
--- a/RTCP/GPUProc/src/CL/cl.hpp
+++ b/RTCP/GPUProc/src/CL/cl.hpp
@@ -1,4 +1,4 @@
-#include <iostream>
+//#include <iostream>
/*******************************************************************************
* Copyright (c) 2008-2011 The Khronos Group Inc.
*
@@ -2069,9 +2069,9 @@ public:
{
cl_int error;
object_ = ::clCreateBuffer(context(), flags, size, host_ptr, &error);
-if (!(flags & CL_MEM_ALLOC_HOST_PTR))
-#pragma omp critical (cout)
-std::cout << "A: Allocating " << size << " bytes" << std::endl;
+//if (!(flags & CL_MEM_ALLOC_HOST_PTR))
+//#pragma omp critical (cout)
+//std::cout << "A: Allocating " << size << " bytes" << std::endl;
detail::errHandler(error, __CREATE_BUFFER_ERR);
if (err != NULL) {
@@ -2090,9 +2090,9 @@ std::cout << "A: Allocating " << size << " bytes" << std::endl;
Context context = Context::getDefault(err);
object_ = ::clCreateBuffer(context(), flags, size, host_ptr, &error);
-if (!(flags & CL_MEM_ALLOC_HOST_PTR))
-#pragma omp critical (cout)
-std::cout << "B: Allocating " << size << " bytes" << std::endl;
+//if (!(flags & CL_MEM_ALLOC_HOST_PTR))
+//#pragma omp critical (cout)
+//std::cout << "B: Allocating " << size << " bytes" << std::endl;
detail::errHandler(error, __CREATE_BUFFER_ERR);
if (err != NULL) {
diff --git a/RTCP/GPUProc/src/Correlator.cl b/RTCP/GPUProc/src/Correlator.cl
index 9ca1ce91620cc5927211907be030ff33626164ce..5d41253f40d8ca7cabe4bc7a640ab1309ea913fd 100644
--- a/RTCP/GPUProc/src/Correlator.cl
+++ b/RTCP/GPUProc/src/Correlator.cl
@@ -1,7 +1,7 @@
#define NR_BASELINES (NR_STATIONS * (NR_STATIONS + 1) / 2)
#if NR_STATIONS == 288
-#define BLOCK_SIZE 8
+#define BLOCK_SIZE 6
#elif defined NVIDIA_CUDA && NR_SAMPLES_PER_CHANNEL % 24 == 0
#define BLOCK_SIZE 24
#else
@@ -24,7 +24,7 @@ __kernel void correlate(__global void *visibilitiesPtr,
__local float samples[4][BLOCK_SIZE][NR_STATIONS | 1]; // avoid power-of-2
uint baseline = get_global_id(0);
- uint channel = get_global_id(1);
+ uint channel = get_global_id(1) + 1;
uint stat_0 = convert_uint_rtz(sqrt(convert_float(8 * baseline + 1)) - 0.99999f) / 2;
uint stat_A = baseline - stat_0 * (stat_0 + 1) / 2;
@@ -84,7 +84,7 @@ __kernel void correlate_2x2(__global void *visibilitiesPtr,
__local float4 samples[2][BLOCK_SIZE][(NR_STATIONS + 1) / 2 | 1]; // avoid power-of-2
- uint channel = get_global_id(1);
+ uint channel = get_global_id(1) + 1;
uint block = get_global_id(0);
uint x = convert_uint_rtz(sqrt(convert_float(8 * block + 1)) - 0.99999f) / 2;
@@ -181,7 +181,7 @@ __kernel void correlate_3x3(__global void *visibilitiesPtr,
__local float4 samples[3][BLOCK_SIZE][(NR_STATIONS + 2) / 3 | 1]; // avoid power-of-2
- uint channel = get_global_id(1);
+ uint channel = get_global_id(1) + 1;
uint block = get_global_id(0);
uint x = convert_uint_rtz(sqrt(convert_float(8 * block + 1)) - 0.99999f) / 2;
@@ -338,7 +338,7 @@ __kernel void correlate_4x4(__global void *visibilitiesPtr,
__local float4 samples[4][BLOCK_SIZE][(NR_STATIONS + 3) / 4 | 1]; // avoid power-of-2
- uint channel = get_global_id(1);
+ uint channel = get_global_id(1) + 1;
uint block = get_global_id(0);
uint x = convert_uint_rtz(sqrt(convert_float(8 * block + 1)) - 0.99999f) / 2;
diff --git a/RTCP/GPUProc/src/Correlator.cl-0.ptx b/RTCP/GPUProc/src/Correlator.cl-0.ptx
deleted file mode 100644
index 37d40c5127c10f59970bebd14f4800ba1f03e8a1..0000000000000000000000000000000000000000
Binary files a/RTCP/GPUProc/src/Correlator.cl-0.ptx and /dev/null differ
diff --git a/RTCP/GPUProc/src/DelayAndBandPass.cl-0.ptx b/RTCP/GPUProc/src/DelayAndBandPass.cl-0.ptx
deleted file mode 100644
index 765b24b939aaab3b67274c4aedf221c92af81818..0000000000000000000000000000000000000000
Binary files a/RTCP/GPUProc/src/DelayAndBandPass.cl-0.ptx and /dev/null differ
diff --git a/RTCP/GPUProc/src/FIR.cl-0.ptx b/RTCP/GPUProc/src/FIR.cl-0.ptx
deleted file mode 100644
index d1a6fd1785efd5436cb75094fa168a1692b1643d..0000000000000000000000000000000000000000
Binary files a/RTCP/GPUProc/src/FIR.cl-0.ptx and /dev/null differ
diff --git a/RTCP/GPUProc/src/FilterBank.cc b/RTCP/GPUProc/src/FilterBank.cc
index beacf7cbcdf86a2243fda08c43abef4096a3b253..b33756a68ba03a56684f7a1d9025a730374293c8 100644
--- a/RTCP/GPUProc/src/FilterBank.cc
+++ b/RTCP/GPUProc/src/FilterBank.cc
@@ -174,152 +174,155 @@ void FilterBank::interpolate(const double x[], const double y[], unsigned xlen,
// but has better stopband characteristics.
void FilterBank::generate_fir_filter(unsigned n, double w, const double window[], double result[])
{
- // make sure grid is big enough for the window
- // the grid must be at least (n+1)/2
- // for all filters where the order is a power of two minus 1, grid_n = n+1;
- unsigned grid_n = nextPowerOfTwo(n + 1);
+#pragma omp critical (FFTW)
+ {
+ // make sure grid is big enough for the window
+ // the grid must be at least (n+1)/2
+ // for all filters where the order is a power of two minus 1, grid_n = n+1;
+ unsigned grid_n = nextPowerOfTwo(n + 1);
- unsigned ramp_n = 2; // grid_n/20;
+ unsigned ramp_n = 2; // grid_n/20;
- // Apply ramps to discontinuities
- // this is a low pass filter
- // maybe we can omit the "w, 0" point?
- // I did observe a small difference
- double f[] = { 0.0, w - ramp_n / grid_n / 2.0, w, w + ramp_n / grid_n / 2.0, 1.0 };
- double m[] = { 1.0, 1.0, 0.0, 0.0, 0.0 };
+ // Apply ramps to discontinuities
+ // this is a low pass filter
+ // maybe we can omit the "w, 0" point?
+ // I did observe a small difference
+ double f[] = { 0.0, w - ramp_n / grid_n / 2.0, w, w + ramp_n / grid_n / 2.0, 1.0 };
+ double m[] = { 1.0, 1.0, 0.0, 0.0, 0.0 };
- // grid is a 1-D array with grid_n+1 points. Values are 1 in filter passband, 0 otherwise
- std::vector<double> grid(grid_n + 1);
+ // grid is a 1-D array with grid_n+1 points. Values are 1 in filter passband, 0 otherwise
+ std::vector<double> grid(grid_n + 1);
- // interpolate between grid points
- interpolate(f, m, 5 /* length of f and m arrays */, grid_n + 1, &grid[0]);
+ // interpolate between grid points
+ interpolate(f, m, 5 /* length of f and m arrays */, grid_n + 1, &grid[0]);
#if 0
- std::stringstream logStr;
- logStr << "interpolated = [";
- for(unsigned i=0; i<grid_n+1; i++) {
- logStr << grid[i];
- if(i != grid_n+1-1) logStr << ", ";
- }
- logStr << "];";
- LOG_DEBUG(logStr.str());
+ std::stringstream logStr;
+ logStr << "interpolated = [";
+ for(unsigned i=0; i<grid_n+1; i++) {
+ logStr << grid[i];
+ if(i != grid_n+1-1) logStr << ", ";
+ }
+ logStr << "];";
+ LOG_DEBUG(logStr.str());
#endif
- // the grid we do an ifft on is:
- // grid appended with grid_n*2 zeros
- // appended with original grid values from indices grid_n..2, i.e., the values in reverse order
- // (note, arrays start at 1 in octave!)
- // the input for the ifft is of size 4*grid_n
- // input = [grid ; zeros(grid_n*2,1) ;grid(grid_n:-1:2)];
+ // the grid we do an ifft on is:
+ // grid appended with grid_n*2 zeros
+ // appended with original grid values from indices grid_n..2, i.e., the values in reverse order
+ // (note, arrays start at 1 in octave!)
+ // the input for the ifft is of size 4*grid_n
+ // input = [grid ; zeros(grid_n*2,1) ;grid(grid_n:-1:2)];
#if defined HAVE_FFTW3
- fftwf_complex* cinput = (fftwf_complex*) fftwf_malloc(grid_n * 4 * sizeof(fftwf_complex));
- fftwf_complex* coutput = (fftwf_complex*) fftwf_malloc(grid_n * 4 * sizeof(fftwf_complex));
+ fftwf_complex* cinput = (fftwf_complex*) fftwf_malloc(grid_n * 4 * sizeof(fftwf_complex));
+ fftwf_complex* coutput = (fftwf_complex*) fftwf_malloc(grid_n * 4 * sizeof(fftwf_complex));
#elif defined HAVE_FFTW2
- fftw_complex* cinput = (fftw_complex*) fftw_malloc(grid_n*4*sizeof(fftw_complex));
- fftw_complex* coutput = (fftw_complex*) fftw_malloc(grid_n*4*sizeof(fftw_complex));
+ fftw_complex* cinput = (fftw_complex*) fftw_malloc(grid_n*4*sizeof(fftw_complex));
+ fftw_complex* coutput = (fftw_complex*) fftw_malloc(grid_n*4*sizeof(fftw_complex));
#endif
- if (cinput == NULL || coutput == NULL) {
- THROW(GPUProcException, "cannot allocate buffers");
- }
+ if (cinput == NULL || coutput == NULL) {
+ THROW(GPUProcException, "cannot allocate buffers");
+ }
- // wipe imaginary part
- for (unsigned i = 0; i < grid_n * 4; i++) {
- fftw_imag(cinput[i]) = 0.0;
- }
+ // wipe imaginary part
+ for (unsigned i = 0; i < grid_n * 4; i++) {
+ fftw_imag(cinput[i]) = 0.0;
+ }
- // copy first part of grid
- for (unsigned i = 0; i < grid_n + 1; i++) {
- fftw_real(cinput[i]) = grid[i];
- }
+ // copy first part of grid
+ for (unsigned i = 0; i < grid_n + 1; i++) {
+ fftw_real(cinput[i]) = grid[i];
+ }
- // append zeros
- for (unsigned i = grid_n + 1; i <= grid_n * 3; i++) {
- fftw_real(cinput[i]) = 0.0;
- }
+ // append zeros
+ for (unsigned i = grid_n + 1; i <= grid_n * 3; i++) {
+ fftw_real(cinput[i]) = 0.0;
+ }
- // now append the grid in reverse order
- for (unsigned i = grid_n - 1, index = 0; i >= 1; i --, index ++) {
- fftw_real(cinput[grid_n * 3 + 1 + index]) = grid[i];
- }
+ // now append the grid in reverse order
+ for (unsigned i = grid_n - 1, index = 0; i >= 1; i --, index ++) {
+ fftw_real(cinput[grid_n * 3 + 1 + index]) = grid[i];
+ }
#if 0
- std::stringstream logStr;
- logStr << "ifft_in = [";
- for(unsigned i=0; i<grid_n*4; i++) {
- logStr << fftw_real(cinput[i]) << " " << fftw_imag(cinput[i]);
- if(i != grid_n*4-1) logStr << ", ";
- }
- logStr << "];";
- LOG_DEBUG(logStr.str());
+ std::stringstream logStr;
+ logStr << "ifft_in = [";
+ for(unsigned i=0; i<grid_n*4; i++) {
+ logStr << fftw_real(cinput[i]) << " " << fftw_imag(cinput[i]);
+ if(i != grid_n*4-1) logStr << ", ";
+ }
+ logStr << "];";
+ LOG_DEBUG(logStr.str());
#endif
#if defined HAVE_FFTW3
- fftwf_plan plan = fftwf_plan_dft_1d(grid_n * 4, cinput, coutput, FFTW_BACKWARD, FFTW_ESTIMATE);
- fftwf_execute(plan);
+ fftwf_plan plan = fftwf_plan_dft_1d(grid_n * 4, cinput, coutput, FFTW_BACKWARD, FFTW_ESTIMATE);
+ fftwf_execute(plan);
#elif defined HAVE_FFTW2
- fftw_plan plan = fftw_create_plan(grid_n * 4, FFTW_BACKWARD, FFTW_ESTIMATE);
- fftw_one(plan, cinput, coutput);
+ fftw_plan plan = fftw_create_plan(grid_n * 4, FFTW_BACKWARD, FFTW_ESTIMATE);
+ fftw_one(plan, cinput, coutput);
#endif
#if 0
- for(unsigned i=0; i<grid_n*4; i++) {
- LOG_DEBUG_STR("ifft result [" << i << "] = " << fftw_real(coutput[i]) << " " << fftw_imag(coutput[i]));
- }
+ for(unsigned i=0; i<grid_n*4; i++) {
+ LOG_DEBUG_STR("ifft result [" << i << "] = " << fftw_real(coutput[i]) << " " << fftw_imag(coutput[i]));
+ }
#endif
- // half end
- // 1 2 n+1 2(n+1) 3(n+1) 4(n+1)
- // x x x x x x x x x # last quarter
- // x x x x x x # first quarter
+ // half end
+ // 1 2 n+1 2(n+1) 3(n+1) 4(n+1)
+ // x x x x x x x x x # last quarter
+ // x x x x x x # first quarter
- // last_quarter = b([end-n+1:2:end]); # the size is only 1/8, since we skip half of the elements
- // first_quarter = b(2:2:(n+1)); # the size is only 1/8, since we skip half of the elements
+ // last_quarter = b([end-n+1:2:end]); # the size is only 1/8, since we skip half of the elements
+ // first_quarter = b(2:2:(n+1)); # the size is only 1/8, since we skip half of the elements
- unsigned index = 0;
+ unsigned index = 0;
- for (unsigned i = 4 * grid_n - n; i < 4 * grid_n; i += 2) {
- result[index] = fftw_real(coutput[i]);
- index++;
- }
+ for (unsigned i = 4 * grid_n - n; i < 4 * grid_n; i += 2) {
+ result[index] = fftw_real(coutput[i]);
+ index++;
+ }
- for (unsigned i = 1; i <= n; i += 2) {
- result[index] = fftw_real(coutput[i]);
- index++;
- }
+ for (unsigned i = 1; i <= n; i += 2) {
+ result[index] = fftw_real(coutput[i]);
+ index++;
+ }
#if defined HAVE_FFTW3
- fftwf_destroy_plan(plan);
- fftwf_free(cinput);
- fftwf_free(coutput);
+ fftwf_destroy_plan(plan);
+ fftwf_free(cinput);
+ fftwf_free(coutput);
#elif defined HAVE_FFTW2
- fftw_destroy_plan(plan);
- fftw_free(cinput);
- fftw_free(coutput);
+ fftw_destroy_plan(plan);
+ fftw_free(cinput);
+ fftw_free(coutput);
#endif
- // multiply with window
- for (unsigned i = 0; i <= n; i++) {
- result[i] *= window[i];
- }
+ // multiply with window
+ for (unsigned i = 0; i <= n; i++) {
+ result[i] *= window[i];
+ }
- // normalize
- double factor = result[n / 2];
- for (unsigned i = 0; i <= n; i++) {
- result[i] /= factor;
- }
+ // normalize
+ double factor = result[n / 2];
+ for (unsigned i = 0; i <= n; i++) {
+ result[i] /= factor;
+ }
#if 0
- std::stringstream logStr;
- logStr << "result = [";
- for(unsigned i=0; i<=n; i++) {
- logStr << result[i];
- if(i != n) logStr << ", ";
- }
- logStr << "];";
- LOG_DEBUG(logStr.str());
+ std::stringstream logStr;
+ logStr << "result = [";
+ for(unsigned i=0; i<=n; i++) {
+ logStr << result[i];
+ if(i != n) logStr << ", ";
+ }
+ logStr << "];";
+ LOG_DEBUG(logStr.str());
#endif
+ }
}
diff --git a/RTCP/GPUProc/src/NewCorrelator.cl b/RTCP/GPUProc/src/NewCorrelator.cl
index 307c70f347743ea6201b6fd2a87fad59ad62812d..6d4e6ee4bc3782018a89d6287278253fc1a59c1e 100644
--- a/RTCP/GPUProc/src/NewCorrelator.cl
+++ b/RTCP/GPUProc/src/NewCorrelator.cl
@@ -1,72 +1,124 @@
#define NR_STATIONS_PER_BLOCK 32
-#define BLOCK_SIZE 8
+#define NR_TIMES_PER_BLOCK 8
#define NR_BASELINES (NR_STATIONS * (NR_STATIONS + 1) / 2)
-typedef __global float (*CorrectedDataType)[NR_STATIONS][NR_CHANNELS][NR_SAMPLES_PER_CHANNEL][NR_POLARIZATIONS * 2];
+typedef __global float4 (*CorrectedDataType)[NR_STATIONS][NR_CHANNELS][NR_SAMPLES_PER_CHANNEL];
typedef __global float8 (*VisibilitiesType)[NR_BASELINES][NR_CHANNELS];
-#if 0
-__kernel void correlateTriangles(__global void *visibilitiesPtr,
- __global const void *correctedDataPtr
-)
+__kernel
+void correlateTriangleKernel(__global void *visibilitiesPtr,
+ __global const void *correctedDataPtr)
{
VisibilitiesType visibilities = (VisibilitiesType) visibilitiesPtr;
CorrectedDataType correctedData = (CorrectedDataType) correctedDataPtr;
- __local float4 samples[BLOCK_SIZE][NR_STATIONS_PER_BLOCK];
+ __local float4 samples[2][NR_TIMES_PER_BLOCK][NR_STATIONS_PER_BLOCK / 2 | 1];
+ uint channel = get_global_id(2) + 1;
+ uint block = get_global_id(1);
+
+#if NR_STATIONS % NR_STATIONS_PER_BLOCK == 0
+ uint firstStation = block * NR_STATIONS_PER_BLOCK;
+ uint nrStationsThisBlock = NR_STATIONS_PER_BLOCK;
+#else
+ uint lastStation = block * NR_STATIONS_PER_BLOCK + NR_STATIONS % NR_STATIONS_PER_BLOCK;
+ uint firstStation = block == 0 ? 0 : lastStation - NR_STATIONS_PER_BLOCK;
+ uint nrStationsThisBlock = lastStation - firstStation;
+#endif
- uint triangle = get_global_id(1);
- uint channel = get_global_id(2);
- uint firstStation = triangle * NR_STATIONS_PER_BLOCK;
+ uint miniBlock = get_local_id(0);
+ uint statXoffset = convert_uint_rtz(sqrt(convert_float(8 * miniBlock + 1)) - 0.99999f) / 2;
+ uint statYoffset = miniBlock - statXoffset * (statXoffset + 1) / 2;
- float4 vis_0A_r = (float4) 0, vis_0A_i = (float4) 0;
- float4 vis_0B_r = (float4) 0, vis_0B_i = (float4) 0;
- float4 vis_1A_r = (float4) 0, vis_1A_i = (float4) 0;
- float4 vis_1B_r = (float4) 0, vis_1B_i = (float4) 0;
+ statXoffset *= 2, statYoffset *= 2;
- for (uint major = 0; major < NR_SAMPLES_PER_CHANNEL; major += BLOCK_SIZE) {
+ float4 vis_0A_r = 0, vis_0A_i = 0;
+ float4 vis_0B_r = 0, vis_0B_i = 0;
+ float4 vis_1A_r = 0, vis_1A_i = 0;
+ float4 vis_1B_r = 0, vis_1B_i = 0;
+
+ bool doCorrelate = statXoffset < nrStationsThisBlock;
+
+ for (uint major = 0; major < NR_SAMPLES_PER_CHANNEL; major += NR_TIMES_PER_BLOCK) {
// load data into local memory
-#pragma unroll 1
- for (uint i = get_local_id(0); i < BLOCK_SIZE * NR_STATIONS_PER_BLOCK; i += get_local_size(0)) {
- uint time = i % BLOCK_SIZE;
- uint stat = i / BLOCK_SIZE;
+ barrier(CLK_LOCAL_MEM_FENCE);
- if (firstStation + stat < NR_STATIONS)
- samples[time][stat] = (*correctedData)[firstStation + stat][channel][major + time];
+ for (uint i = get_local_id(0); i < nrStationsThisBlock * NR_TIMES_PER_BLOCK; i += get_local_size(0)) {
+ uint time = i % NR_TIMES_PER_BLOCK;
+ uint stat = i / NR_TIMES_PER_BLOCK;
+
+ samples[stat % 2][time][stat / 2] = (*correctedData)[firstStation + stat][channel][major + time];
}
barrier(CLK_LOCAL_MEM_FENCE);
- // compute auto-correlations
+#pragma unroll 1
+ for (uint time = 0; time < NR_TIMES_PER_BLOCK; time ++) {
+ float4 sample_0, sample_1, sample_A, sample_B;
+
+ if (doCorrelate) {
+ sample_0 = samples[0][time][statYoffset / 2];
+ sample_A = samples[0][time][statXoffset / 2];
+ sample_B = samples[1][time][statXoffset / 2];
+ sample_1 = samples[1][time][statYoffset / 2];
+
+ vis_0A_r += sample_0.xxzz * sample_A.xzxz;
+ vis_0A_i += sample_0.yyww * sample_A.xzxz;
+ vis_0B_r += sample_0.xxzz * sample_B.xzxz;
+ vis_0B_i += sample_0.yyww * sample_B.xzxz;
+ vis_1A_r += sample_1.xxzz * sample_A.xzxz;
+ vis_1A_i += sample_1.yyww * sample_A.xzxz;
+ vis_1B_r += sample_1.xxzz * sample_B.xzxz;
+ vis_1B_i += sample_1.yyww * sample_B.xzxz;
- if (firstStation + get_local_id(0) < NR_STATIONS) {
- for (time = 0; time + BLOCK_SIZE; time ++) {
- float sample = samples[time][get_local_id(0)];
+ vis_0A_r += sample_0.yyww * sample_A.ywyw;
+ vis_0A_i -= sample_0.xxzz * sample_A.ywyw;
+ vis_0B_r += sample_0.yyww * sample_B.ywyw;
+ vis_0B_i -= sample_0.xxzz * sample_B.ywyw;
+ vis_1A_r += sample_1.yyww * sample_A.ywyw;
+ vis_1A_i -= sample_1.xxzz * sample_A.ywyw;
+ vis_1B_r += sample_1.yyww * sample_B.ywyw;
+ vis_1B_i -= sample_1.xxzz * sample_B.ywyw;
}
+ }
+ }
+ int statY = firstStation + statYoffset;
+ uint statX = firstStation + statXoffset;
+ uint baseline = (statX * (statX + 1) / 2) + statY;
- barrier(CLK_LOCAL_MEM_FENCE);
+ if (statXoffset < nrStationsThisBlock) {
+ (*visibilities)[baseline ][channel].even = vis_0A_r;
+ (*visibilities)[baseline ][channel].odd = vis_0A_i;
+ }
+
+ if (statXoffset < nrStationsThisBlock && statYoffset + 1 < nrStationsThisBlock) {
+ (*visibilities)[baseline + 1][channel].even = vis_1A_r;
+ (*visibilities)[baseline + 1][channel].odd = vis_1A_i;
+ }
+
+ if (statXoffset + 1 < nrStationsThisBlock) {
+ (*visibilities)[baseline + statX + 1][channel].even = vis_0B_r;
+ (*visibilities)[baseline + statX + 1][channel].odd = vis_0B_i;
+ (*visibilities)[baseline + statX + 2][channel].even = vis_1B_r;
+ (*visibilities)[baseline + statX + 2][channel].odd = vis_1B_i;
}
}
-#endif
__kernel __attribute__((reqd_work_group_size(NR_STATIONS_PER_BLOCK * NR_STATIONS_PER_BLOCK / 4, 1, 1)))
-void correlateRectangles(__global void *visibilitiesPtr,
- __global const void *correctedDataPtr
-)
+void correlateRectangleKernel(__global void *visibilitiesPtr,
+ __global const void *correctedDataPtr)
{
- VisibilitiesType visibilities = (VisibilitiesType) visibilitiesPtr;
+ VisibilitiesType visibilities = (VisibilitiesType) visibilitiesPtr;
CorrectedDataType correctedData = (CorrectedDataType) correctedDataPtr;
- __local float4 samplesX[2][BLOCK_SIZE][NR_STATIONS_PER_BLOCK / 2 | 1];
- __local float4 samplesY[2][BLOCK_SIZE][NR_STATIONS_PER_BLOCK / 2 | 1];
+ __local float4 samplesX[2][NR_TIMES_PER_BLOCK][NR_STATIONS_PER_BLOCK / 2 | 1];
+ __local float4 samplesY[2][NR_TIMES_PER_BLOCK][NR_STATIONS_PER_BLOCK / 2 | 1];
uint block = get_global_id(1);
- uint channel = get_global_id(2);
uint blockX = convert_uint_rtz(sqrt(convert_float(8 * block + 1)) - 0.99999f) / 2;
uint blockY = block - blockX * (blockX + 1) / 2;
@@ -78,35 +130,291 @@ void correlateRectangles(__global void *visibilitiesPtr,
int firstStationY = (blockY - 1) * NR_STATIONS_PER_BLOCK + NR_STATIONS % NR_STATIONS_PER_BLOCK;
#endif
- uint statXoffset = (get_local_id(0) / (NR_STATIONS_PER_BLOCK / 2));
- uint statYoffset = (get_local_id(0) % (NR_STATIONS_PER_BLOCK / 2));
+ uint statXoffset = get_local_id(0) / (NR_STATIONS_PER_BLOCK / 2);
+ uint statYoffset = get_local_id(0) % (NR_STATIONS_PER_BLOCK / 2);
+
+ float4 vis_0A_r = 0, vis_0A_i = 0;
+ float4 vis_0B_r = 0, vis_0B_i = 0;
+ float4 vis_1A_r = 0, vis_1A_i = 0;
+ float4 vis_1B_r = 0, vis_1B_i = 0;
+
+ uint loadTime = get_local_id(0) % NR_TIMES_PER_BLOCK;
+ uint loadStat = get_local_id(0) / NR_TIMES_PER_BLOCK;
+
+ bool doCorrelateLower = NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || (int) (firstStationY + 2 * statYoffset) >= 0;
+ bool doCorrelateUpper = NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || (int) (firstStationY + 2 * statYoffset) >= -1;
+ bool doLoadY = NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || (int) (firstStationY + loadStat) >= 0;
+
+ uint channel = get_global_id(2) + 1;
+
+ for (uint major = 0; major < NR_SAMPLES_PER_CHANNEL; major += NR_TIMES_PER_BLOCK) {
+ // load data into local memory
+ float4 sampleX = (*correctedData)[firstStationX + loadStat][channel][major + loadTime];
+ float4 sampleY;
+
+ if (doLoadY)
+ sampleY = (*correctedData)[firstStationY + loadStat][channel][major + loadTime];
+
+ barrier(CLK_LOCAL_MEM_FENCE);
+
+ samplesX[loadStat % 2][loadTime][loadStat / 2] = sampleX;
+
+ if (doLoadY)
+ samplesY[loadStat % 2][loadTime][loadStat / 2] = sampleY;
+
+ barrier(CLK_LOCAL_MEM_FENCE);
+
+#pragma unroll 1
+ for (uint time = 0; time < NR_TIMES_PER_BLOCK; time ++) {
+ float4 sample_0, sample_1, sample_A, sample_B;
+
+ if (doCorrelateLower) {
+ sample_0 = samplesY[0][time][statYoffset];
+ }
+
+ if (doCorrelateUpper) {
+ sample_A = samplesX[0][time][statXoffset];
+ sample_B = samplesX[1][time][statXoffset];
+ sample_1 = samplesY[1][time][statYoffset];
+ }
+
+ if (doCorrelateLower) {
+ vis_0A_r += sample_0.xxzz * sample_A.xzxz;
+ vis_0A_i += sample_0.yyww * sample_A.xzxz;
+ vis_0B_r += sample_0.xxzz * sample_B.xzxz;
+ vis_0B_i += sample_0.yyww * sample_B.xzxz;
+ vis_0A_r += sample_0.yyww * sample_A.ywyw;
+ vis_0A_i -= sample_0.xxzz * sample_A.ywyw;
+ vis_0B_r += sample_0.yyww * sample_B.ywyw;
+ vis_0B_i -= sample_0.xxzz * sample_B.ywyw;
+ }
+
+ if (doCorrelateUpper) {
+ vis_1A_r += sample_1.xxzz * sample_A.xzxz;
+ vis_1A_i += sample_1.yyww * sample_A.xzxz;
+ vis_1B_r += sample_1.xxzz * sample_B.xzxz;
+ vis_1B_i += sample_1.yyww * sample_B.xzxz;
+ vis_1A_r += sample_1.yyww * sample_A.ywyw;
+ vis_1A_i -= sample_1.xxzz * sample_A.ywyw;
+ vis_1B_r += sample_1.yyww * sample_B.ywyw;
+ vis_1B_i -= sample_1.xxzz * sample_B.ywyw;
+ }
+ }
+ }
+
+ int statY = firstStationY + 2 * statYoffset;
+ uint statX = firstStationX + 2 * statXoffset;
+ uint baseline = (statX * (statX + 1) / 2) + statY;
+
+ if (doCorrelateLower) {
+ (*visibilities)[baseline ][channel].even = vis_0A_r;
+ (*visibilities)[baseline ][channel].odd = vis_0A_i;
+ (*visibilities)[baseline + statX + 1][channel].even = vis_0B_r;
+ (*visibilities)[baseline + statX + 1][channel].odd = vis_0B_i;
+ }
+
+ if (doCorrelateUpper) {
+ (*visibilities)[baseline + 1][channel].even = vis_1A_r;
+ (*visibilities)[baseline + 1][channel].odd = vis_1A_i;
+ (*visibilities)[baseline + statX + 2][channel].even = vis_1B_r;
+ (*visibilities)[baseline + statX + 2][channel].odd = vis_1B_i;
+ }
+}
+
+////////////////////////////////////////////////////////////////////////////////
+
+void correlateTriangle(VisibilitiesType visibilities,
+ CorrectedDataType correctedData,
+ __local float4 samples[2][NR_TIMES_PER_BLOCK][NR_STATIONS_PER_BLOCK / 2 | 1],
+ uint block)
+{
+ uint channel = get_global_id(2) + 1;
+
+#if NR_STATIONS % NR_STATIONS_PER_BLOCK == 0
+ uint firstStation = block * NR_STATIONS_PER_BLOCK;
+#else
+ int firstStation = (block - 1) * NR_STATIONS_PER_BLOCK + NR_STATIONS % NR_STATIONS_PER_BLOCK;
+#endif
+
+ bool doCorrelate = false, doAutoCorrelate = false, doNearAutoCorrelate = false;
+ uint statXoffset, statYoffset;
+
+ if (get_local_id(0) < 128) {
+ uint miniBlock = get_local_id(0);
+ uint miniBlockX = convert_uint_rtz(sqrt(convert_float(8 * miniBlock + 1)) - 0.99999f) / 2;
+ uint miniBlockY = miniBlock - miniBlockX * (miniBlockX + 1) / 2;
+
+ statXoffset = 2 * miniBlockX + 2;
+ statYoffset = 2 * miniBlockY;
+ doCorrelate = statXoffset < NR_STATIONS_PER_BLOCK && (NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || (int) (firstStation + statYoffset) >= 0);
+ } else if (get_local_id(0) < 128 + NR_STATIONS_PER_BLOCK / 2) {
+ statXoffset = statYoffset = 2 * (get_local_id(0) - 128);
+ // actually, it is the visibility one right of statXoffset
+ doNearAutoCorrelate = (int) (firstStation + statXoffset) >= 0;
+ } else if (get_local_id(0) >= 192 && get_local_id(0) < 192 + NR_STATIONS_PER_BLOCK) {
+ statXoffset = statYoffset = get_local_id(0) - 192;
+ doAutoCorrelate = NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || (int) (firstStation + statYoffset) >= 0;
+ }
float4 vis_0A_r = 0, vis_0A_i = 0;
float4 vis_0B_r = 0, vis_0B_i = 0;
float4 vis_1A_r = 0, vis_1A_i = 0;
float4 vis_1B_r = 0, vis_1B_i = 0;
- for (uint major = 0; major < NR_SAMPLES_PER_CHANNEL; major += BLOCK_SIZE) {
+ uint loadTime = get_local_id(0) % NR_TIMES_PER_BLOCK;
+ uint loadStat = get_local_id(0) / NR_TIMES_PER_BLOCK;
+
+ bool doLoad = NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || (int) (firstStation + loadStat) >= 0;
+
+ for (uint major = 0; major < NR_SAMPLES_PER_CHANNEL; major += NR_TIMES_PER_BLOCK) {
// load data into local memory
- for (uint i = get_local_id(0); i < 4 * BLOCK_SIZE * NR_STATIONS_PER_BLOCK; i += NR_STATIONS_PER_BLOCK * NR_STATIONS_PER_BLOCK / 4) {
- uint p = i % 4;
- uint time = i / 4 % BLOCK_SIZE;
- uint stat = i / 4 / BLOCK_SIZE;
+ float4 sample;
+
+ if (doLoad)
+ sample = (*correctedData)[firstStation + loadStat][channel][major + loadTime];
+
+ barrier(CLK_LOCAL_MEM_FENCE);
+
+ if (doLoad)
+ samples[loadStat % 2][loadTime][loadStat / 2] = sample;
+
+ barrier(CLK_LOCAL_MEM_FENCE);
+
+#pragma unroll 1
+ for (uint time = 0; time < NR_TIMES_PER_BLOCK; time ++) {
+ if (doCorrelate) {
+ float4 sample_0 = samples[0][time][statYoffset / 2];
+ float4 sample_A = samples[0][time][statXoffset / 2];
+ float4 sample_B = samples[1][time][statXoffset / 2];
+ float4 sample_1 = samples[1][time][statYoffset / 2];
+
+ vis_0A_r += sample_0.xxzz * sample_A.xzxz;
+ vis_0A_i += sample_0.yyww * sample_A.xzxz;
+ vis_0B_r += sample_0.xxzz * sample_B.xzxz;
+ vis_0B_i += sample_0.yyww * sample_B.xzxz;
+ vis_0A_r += sample_0.yyww * sample_A.ywyw;
+ vis_0A_i -= sample_0.xxzz * sample_A.ywyw;
+ vis_0B_r += sample_0.yyww * sample_B.ywyw;
+ vis_0B_i -= sample_0.xxzz * sample_B.ywyw;
+
+ vis_1A_r += sample_1.xxzz * sample_A.xzxz;
+ vis_1A_i += sample_1.yyww * sample_A.xzxz;
+ vis_1B_r += sample_1.xxzz * sample_B.xzxz;
+ vis_1B_i += sample_1.yyww * sample_B.xzxz;
+ vis_1A_r += sample_1.yyww * sample_A.ywyw;
+ vis_1A_i -= sample_1.xxzz * sample_A.ywyw;
+ vis_1B_r += sample_1.yyww * sample_B.ywyw;
+ vis_1B_i -= sample_1.xxzz * sample_B.ywyw;
+ }
- ((__local float *) &samplesX[stat % 2][time][stat / 2])[p] = (*correctedData)[firstStationX + stat][channel][major + time][p];
+ if (doAutoCorrelate) {
+ float4 sample_0 = samples[statYoffset % 2][time][statYoffset / 2];
+ vis_0A_r.xyw += sample_0.xxz * sample_0.xzz;
+ vis_0A_i.y += sample_0.y * sample_0.z;
+ vis_0A_r.xyw += sample_0.yyw * sample_0.yww;
+ vis_0A_i.y -= sample_0.x * sample_0.w;
+ }
- if (NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || (int) (firstStationY + stat) >= 0)
- ((__local float *) &samplesY[stat % 2][time][stat / 2])[p] = (*correctedData)[firstStationY + stat][channel][major + time][p];
+ if (doNearAutoCorrelate) {
+ float4 sample_0 = samples[0][time][statYoffset / 2];
+ float4 sample_B = samples[1][time][statXoffset / 2];
+ vis_0B_r += sample_0.xxzz * sample_B.xzxz;
+ vis_0B_i += sample_0.yyww * sample_B.xzxz;
+ vis_0B_r += sample_0.yyww * sample_B.ywyw;
+ vis_0B_i -= sample_0.xxzz * sample_B.ywyw;
+ }
}
+ }
+
+ if (doAutoCorrelate) {
+ vis_0A_r.z = vis_0A_r.y;
+ vis_0A_i.z = -vis_0A_i.y;
+ }
+
+ int statY = firstStation + statYoffset;
+ uint statX = firstStation + statXoffset;
+ uint baseline = (statX * (statX + 1) / 2) + statY;
+
+ if (doCorrelate || doAutoCorrelate) {
+ (*visibilities)[baseline ][channel].even = vis_0A_r;
+ (*visibilities)[baseline ][channel].odd = vis_0A_i;
+ }
+
+ if (doCorrelate || doNearAutoCorrelate) {
+ (*visibilities)[baseline + statX + 1][channel].even = vis_0B_r;
+ (*visibilities)[baseline + statX + 1][channel].odd = vis_0B_i;
+ }
+
+ if (doCorrelate) {
+ (*visibilities)[baseline + 1][channel].even = vis_1A_r;
+ (*visibilities)[baseline + 1][channel].odd = vis_1A_i;
+ (*visibilities)[baseline + statX + 2][channel].even = vis_1B_r;
+ (*visibilities)[baseline + statX + 2][channel].odd = vis_1B_i;
+ }
+}
+
+
+void correlateTriangle2(VisibilitiesType visibilities,
+ CorrectedDataType correctedData,
+ __local float4 samples[2][NR_TIMES_PER_BLOCK][NR_STATIONS_PER_BLOCK / 2 | 1],
+ uint block
+)
+{
+ uint channel = get_global_id(2) + 1;
+
+#if NR_STATIONS % NR_STATIONS_PER_BLOCK == 0
+ uint firstStation = block * NR_STATIONS_PER_BLOCK;
+ //uint lastStation = firstStation + NR_STATIONS_PER_BLOCK;
+ uint nrStationsThisBlock = NR_STATIONS_PER_BLOCK;
+#else
+ uint lastStation = block * NR_STATIONS_PER_BLOCK + NR_STATIONS % NR_STATIONS_PER_BLOCK;
+ uint firstStation = block == 0 ? 0 : lastStation - NR_STATIONS_PER_BLOCK;
+ uint nrStationsThisBlock = lastStation - firstStation;
+#endif
+
+ uint miniBlock = get_local_id(0);
+ uint statXoffset = convert_uint_rtz(sqrt(convert_float(8 * miniBlock + 1)) - 0.99999f) / 2;
+ uint statYoffset = miniBlock - statXoffset * (statXoffset + 1) / 2;
+
+ statXoffset *= 2, statYoffset *= 2;
+
+ //bool doCorrelate = statXoffset < nrStationsThisBlock;
+
+ float4 vis_0A_r = 0, vis_0A_i = 0;
+ float4 vis_0B_r = 0, vis_0B_i = 0;
+ float4 vis_1A_r = 0, vis_1A_i = 0;
+ float4 vis_1B_r = 0, vis_1B_i = 0;
+
+ uint loadTime = get_local_id(0) % NR_TIMES_PER_BLOCK;
+ uint loadStat = get_local_id(0) / NR_TIMES_PER_BLOCK;
+
+ bool doCorrelateLeft = statXoffset < nrStationsThisBlock;
+ //bool doCorrelateRight = statXoffset + 1 < nrStationsThisBlock;
+ bool doLoad = NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || loadStat < nrStationsThisBlock;
+
+ for (uint major = 0; major < NR_SAMPLES_PER_CHANNEL; major += NR_TIMES_PER_BLOCK) {
+ // load data into local memory
+ float4 sample;
+
+ if (doLoad)
+ sample = (*correctedData)[firstStation + loadStat][channel][major + loadTime];
barrier(CLK_LOCAL_MEM_FENCE);
- if (NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || (int) (firstStationY + 2 * statYoffset) >= -1) {
- for (uint time = 0; time < BLOCK_SIZE; time ++) {
- float4 sample_0 = samplesY[0][time][statYoffset];
- float4 sample_A = samplesX[0][time][statXoffset];
- float4 sample_B = samplesX[1][time][statXoffset];
- float4 sample_1 = samplesY[1][time][statYoffset];
+ if (doLoad)
+ samples[loadStat % 2][loadTime][loadStat / 2] = sample;
+
+ barrier(CLK_LOCAL_MEM_FENCE);
+
+#pragma unroll 1
+ for (uint time = 0; time < NR_TIMES_PER_BLOCK; time ++) {
+ float4 sample_0, sample_1, sample_A, sample_B;
+
+ if (doCorrelateLeft) {
+ sample_0 = samples[0][time][statYoffset / 2];
+ sample_A = samples[0][time][statXoffset / 2];
+ sample_B = samples[1][time][statXoffset / 2];
+ sample_1 = samples[1][time][statYoffset / 2];
vis_0A_r += sample_0.xxzz * sample_A.xzxz;
vis_0A_i += sample_0.yyww * sample_A.xzxz;
@@ -127,25 +435,152 @@ void correlateRectangles(__global void *visibilitiesPtr,
vis_1B_i -= sample_1.xxzz * sample_B.ywyw;
}
}
+ }
+
+ int statY = firstStation + statYoffset;
+ uint statX = firstStation + statXoffset;
+ uint baseline = (statX * (statX + 1) / 2) + statY;
+
+ if (statXoffset < nrStationsThisBlock) {
+ (*visibilities)[baseline ][channel].even = vis_0A_r;
+ (*visibilities)[baseline ][channel].odd = vis_0A_i;
+ }
+
+ if (statXoffset < nrStationsThisBlock && statYoffset + 1 < nrStationsThisBlock) {
+ (*visibilities)[baseline + 1][channel].even = vis_1A_r;
+ (*visibilities)[baseline + 1][channel].odd = vis_1A_i;
+ }
+
+ if (statXoffset + 1 < nrStationsThisBlock) {
+ (*visibilities)[baseline + statX + 1][channel].even = vis_0B_r;
+ (*visibilities)[baseline + statX + 1][channel].odd = vis_0B_i;
+ (*visibilities)[baseline + statX + 2][channel].even = vis_1B_r;
+ (*visibilities)[baseline + statX + 2][channel].odd = vis_1B_i;
+ }
+}
+
+
+void correlateRectangle(VisibilitiesType visibilities,
+ CorrectedDataType correctedData,
+ __local float4 samplesX[2][NR_TIMES_PER_BLOCK][NR_STATIONS_PER_BLOCK / 2 | 1],
+ __local float4 samplesY[2][NR_TIMES_PER_BLOCK][NR_STATIONS_PER_BLOCK / 2 | 1],
+ uint blockX,
+ uint blockY
+)
+{
+ uint channel = get_global_id(2) + 1;
+
+#if NR_STATIONS % NR_STATIONS_PER_BLOCK == 0
+ uint firstStationX = blockX * NR_STATIONS_PER_BLOCK;
+ uint firstStationY = blockY * NR_STATIONS_PER_BLOCK;
+#else
+ uint firstStationX = (blockX - 1) * NR_STATIONS_PER_BLOCK + NR_STATIONS % NR_STATIONS_PER_BLOCK;
+ int firstStationY = (blockY - 1) * NR_STATIONS_PER_BLOCK + NR_STATIONS % NR_STATIONS_PER_BLOCK;
+#endif
+
+ uint statXoffset = get_local_id(0) / (NR_STATIONS_PER_BLOCK / 2);
+ uint statYoffset = get_local_id(0) % (NR_STATIONS_PER_BLOCK / 2);
+
+ float4 vis_0A_r = 0, vis_0A_i = 0;
+ float4 vis_0B_r = 0, vis_0B_i = 0;
+ float4 vis_1A_r = 0, vis_1A_i = 0;
+ float4 vis_1B_r = 0, vis_1B_i = 0;
+
+ uint loadTime = get_local_id(0) % NR_TIMES_PER_BLOCK;
+ uint loadStat = get_local_id(0) / NR_TIMES_PER_BLOCK;
+
+ bool doCorrelateLower = NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || (int) (firstStationY + 2 * statYoffset) >= 0;
+ bool doCorrelateUpper = NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || (int) (firstStationY + 2 * statYoffset) >= -1;
+ bool doLoadY = NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || (int) (firstStationY + loadStat) >= 0;
+
+ for (uint major = 0; major < NR_SAMPLES_PER_CHANNEL; major += NR_TIMES_PER_BLOCK) {
+ // load data into local memory
+ float4 sampleX = (*correctedData)[firstStationX + loadStat][channel][major + loadTime];
+ float4 sampleY;
+
+ if (doLoadY)
+ sampleY = (*correctedData)[firstStationY + loadStat][channel][major + loadTime];
barrier(CLK_LOCAL_MEM_FENCE);
+
+ samplesX[loadStat % 2][loadTime][loadStat / 2] = sampleX;
+
+ if (doLoadY)
+ samplesY[loadStat % 2][loadTime][loadStat / 2] = sampleY;
+
+ barrier(CLK_LOCAL_MEM_FENCE);
+
+#pragma unroll 1
+ for (uint time = 0; time < NR_TIMES_PER_BLOCK; time ++) {
+ float4 sample_0, sample_1, sample_A, sample_B;
+
+ if (doCorrelateLower) {
+ sample_0 = samplesY[0][time][statYoffset];
+ }
+
+ if (doCorrelateUpper) {
+ sample_A = samplesX[0][time][statXoffset];
+ sample_B = samplesX[1][time][statXoffset];
+ sample_1 = samplesY[1][time][statYoffset];
+ }
+
+ if (doCorrelateLower) {
+ vis_0A_r += sample_0.xxzz * sample_A.xzxz;
+ vis_0A_i += sample_0.yyww * sample_A.xzxz;
+ vis_0B_r += sample_0.xxzz * sample_B.xzxz;
+ vis_0B_i += sample_0.yyww * sample_B.xzxz;
+ vis_0A_r += sample_0.yyww * sample_A.ywyw;
+ vis_0A_i -= sample_0.xxzz * sample_A.ywyw;
+ vis_0B_r += sample_0.yyww * sample_B.ywyw;
+ vis_0B_i -= sample_0.xxzz * sample_B.ywyw;
+ }
+
+ if (doCorrelateUpper) {
+ vis_1A_r += sample_1.xxzz * sample_A.xzxz;
+ vis_1A_i += sample_1.yyww * sample_A.xzxz;
+ vis_1B_r += sample_1.xxzz * sample_B.xzxz;
+ vis_1B_i += sample_1.yyww * sample_B.xzxz;
+ vis_1A_r += sample_1.yyww * sample_A.ywyw;
+ vis_1A_i -= sample_1.xxzz * sample_A.ywyw;
+ vis_1B_r += sample_1.yyww * sample_B.ywyw;
+ vis_1B_i -= sample_1.xxzz * sample_B.ywyw;
+ }
+ }
}
int statY = firstStationY + 2 * statYoffset;
uint statX = firstStationX + 2 * statXoffset;
uint baseline = (statX * (statX + 1) / 2) + statY;
- if (NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || statY >= 0) {
- (*visibilities)[baseline ][channel].even = vis_0A_r;
- (*visibilities)[baseline ][channel].odd = vis_0A_i;
- (*visibilities)[baseline + 1][channel].even = vis_1A_r;
- (*visibilities)[baseline + 1][channel].odd = vis_1A_i;
- }
-
- if (NR_STATIONS % NR_STATIONS_PER_BLOCK == 0 || statY >= -1) {
+ if (doCorrelateLower) {
+ (*visibilities)[baseline ][channel].even = vis_0A_r;
+ (*visibilities)[baseline ][channel].odd = vis_0A_i;
(*visibilities)[baseline + statX + 1][channel].even = vis_0B_r;
(*visibilities)[baseline + statX + 1][channel].odd = vis_0B_i;
+ }
+
+ if (doCorrelateUpper) {
+ (*visibilities)[baseline + 1][channel].even = vis_1A_r;
+ (*visibilities)[baseline + 1][channel].odd = vis_1A_i;
(*visibilities)[baseline + statX + 2][channel].even = vis_1B_r;
(*visibilities)[baseline + statX + 2][channel].odd = vis_1B_i;
}
}
+
+
+__kernel __attribute__((reqd_work_group_size(NR_STATIONS_PER_BLOCK * NR_STATIONS_PER_BLOCK / 4, 1, 1)))
+void correlate(__global void *visibilitiesPtr,
+ __global const void *correctedDataPtr)
+{
+ __local float4 samplesX[2][NR_TIMES_PER_BLOCK][NR_STATIONS_PER_BLOCK / 2 | 1];
+ __local float4 samplesY[2][NR_TIMES_PER_BLOCK][NR_STATIONS_PER_BLOCK / 2 | 1];
+
+ uint block = get_global_id(1);
+ uint blockX = convert_uint_rtz(sqrt(convert_float(8 * block + 1)) - 0.99999f) / 2;
+ uint blockY = block - blockX * (blockX + 1) / 2;
+
+ if (blockX == blockY)
+ correlateTriangle2((VisibilitiesType) visibilitiesPtr, (CorrectedDataType) correctedDataPtr, samplesX, blockX);
+ else
+ correlateRectangle((VisibilitiesType) visibilitiesPtr, (CorrectedDataType) correctedDataPtr, samplesX, samplesY, blockX, blockY);
+}
diff --git a/RTCP/GPUProc/src/NewCorrelator.cl-0.ptx b/RTCP/GPUProc/src/NewCorrelator.cl-0.ptx
deleted file mode 100644
index c69f71f47ba5748d5cb342ecbed29c627f766f0f..0000000000000000000000000000000000000000
Binary files a/RTCP/GPUProc/src/NewCorrelator.cl-0.ptx and /dev/null differ
diff --git a/RTCP/GPUProc/src/RTCP.cc b/RTCP/GPUProc/src/RTCP.cc
index 0e6c8b99b6746c48388c67fd96eb40a36d5e0329..9c6a8830cdc4fcc45bba57c306ff434c9b3f56e9 100644
--- a/RTCP/GPUProc/src/RTCP.cc
+++ b/RTCP/GPUProc/src/RTCP.cc
@@ -14,11 +14,11 @@
#include <iostream>
#include <sstream>
#include <boost/multi_array.hpp>
+#include <boost/date_time/posix_time/posix_time.hpp>
#include "Align.h"
-#include "BandPass.h"
#include "ApplCommon/PosixTime.h"
-#include <boost/date_time/posix_time/posix_time.hpp>
+#include "BandPass.h"
#include "Common/LofarLogger.h"
#include "Common/SystemUtil.h"
#include "FilterBank.h"
@@ -30,6 +30,11 @@
#include "UHEP/InvertedStationPPFWeights.h"
//#include "clAmdFft/include/clAmdFft.h"
+#if defined __linux__
+#include <sched.h>
+#include <sys/time.h>
+#endif
+
namespace LOFAR {
namespace RTCP {
@@ -41,9 +46,7 @@ unsigned nrGPUs;
#define NR_TAPS 16
#define NR_STATION_FILTER_TAPS 16
-// the SAP to process (we support only one SAP for now)
-#define SAP 0
-
+#undef USE_NEW_CORRELATOR
#define USE_2X2
#undef USE_CUSTOM_FFT
#undef USE_TEST_DATA
@@ -81,6 +84,40 @@ double getTime()
return now - firstTime;
}
+#if defined __linux__
+
+inline void set_affinity(unsigned device)
+{
+#if 0
+ static const char mapping[1][12] = {
+ 0, 1, 2, 3, 8, 9, 10, 11,
+ };
+#else
+ static const char mapping[8][12] = {
+ { 0, 1, 2, 3, 4, 5, 12, 13, 14, 15, 16, 17, },
+ { 0, 1, 2, 3, 4, 5, 12, 13, 14, 15, 16, 17, },
+ { 0, 1, 2, 3, 4, 5, 12, 13, 14, 15, 16, 17, },
+ { 0, 1, 2, 3, 4, 5, 12, 13, 14, 15, 16, 17, },
+ { 6, 7, 8, 9, 10, 11, 18, 19, 20, 21, 22, 23, },
+ { 6, 7, 8, 9, 10, 11, 18, 19, 20, 21, 22, 23, },
+ { 6, 7, 8, 9, 10, 11, 18, 19, 20, 21, 22, 23, },
+ { 6, 7, 8, 9, 10, 11, 18, 19, 20, 21, 22, 23, },
+ };
+#endif
+
+ cpu_set_t set;
+
+ CPU_ZERO(&set);
+
+ for (unsigned coreIndex = 0; coreIndex < 12; coreIndex ++)
+ CPU_SET(mapping[device][coreIndex], &set);
+
+ if (sched_setaffinity(0, sizeof set, &set) < 0)
+ perror("sched_setaffinity");
+}
+
+#endif
+
class PerformanceCounter
{
@@ -200,7 +237,7 @@ cl::Program createProgram(const Parset &ps, cl::Context &context, std::vector<cl
args << " -DNR_SAMPLES_PER_CHANNEL=" << ps.nrSamplesPerChannel();
args << " -DNR_SAMPLES_PER_SUBBAND=" << ps.nrSamplesPerSubband();
args << " -DNR_BEAMS=" << ps.nrBeams();
- args << " -DNR_TABS=" << ps.nrTABs(SAP);
+ args << " -DNR_TABS=" << ps.nrTABs(0);
args << " -DNR_COHERENT_STOKES=" << ps.nrCoherentStokes();
args << " -DNR_INCOHERENT_STOKES=" << ps.nrIncoherentStokes();
args << " -DCOHERENT_STOKES_TIME_INTEGRATION_FACTOR=" << ps.coherentStokesTimeIntegrationFactor();
@@ -257,6 +294,10 @@ class Pipeline
SmartPtr<InputSection<i16complex> > inputSection16;
SmartPtr<InputSection<i8complex> > inputSection8;
SmartPtr<InputSection<i4complex> > inputSection4;
+
+#if defined USE_B7015
+ OMP_Lock hostToDeviceLock[4], deviceToHostLock[4];
+#endif
};
@@ -273,12 +314,12 @@ class CorrelatorPipeline : public Pipeline
FilterBank filterBank;
cl::Program firFilterProgram, delayAndBandPassProgram, correlatorProgram;
+#if defined USE_NEW_CORRELATOR
+ PerformanceCounter firFilterCounter, delayAndBandPassCounter, correlateTriangleCounter, correlateRectangleCounter, fftCounter;
+#else
PerformanceCounter firFilterCounter, delayAndBandPassCounter, correlatorCounter, fftCounter;
- PerformanceCounter samplesCounter, visibilitiesCounter;
-
-#if defined USE_B7015
- OMP_Lock hostToDeviceLock[4], deviceToHostLock[4];
#endif
+ PerformanceCounter samplesCounter, visibilitiesCounter;
};
@@ -332,7 +373,12 @@ CorrelatorPipeline::CorrelatorPipeline(const Parset &ps)
filterBank(true, NR_TAPS, ps.nrChannelsPerSubband(), KAISER),
firFilterCounter("FIR filter"),
delayAndBandPassCounter("delay/bp"),
+#if defined USE_NEW_CORRELATOR
+ correlateTriangleCounter("cor.triangle"),
+ correlateRectangleCounter("cor.rectangle"),
+#else
correlatorCounter("correlator"),
+#endif
fftCounter("FFT"),
samplesCounter("samples"),
visibilitiesCounter("visibilities")
@@ -348,8 +394,11 @@ CorrelatorPipeline::CorrelatorPipeline(const Parset &ps)
#pragma omp section
delayAndBandPassProgram = createProgram("DelayAndBandPass.cl");
#pragma omp section
+#if defined USE_NEW_CORRELATOR
correlatorProgram = createProgram("NewCorrelator.cl");
- //correlatorProgram = createProgram("Correlator.cl");
+#else
+ correlatorProgram = createProgram("Correlator.cl");
+#endif
}
std::cout << "compile time = " << getTime() - startTime << std::endl;
@@ -666,7 +715,7 @@ class DelayAndBandPassKernel : public Kernel
};
-#if 0
+#if !defined USE_NEW_CORRELATOR
class CorrelatorKernel : public Kernel
{
@@ -714,12 +763,13 @@ class CorrelatorKernel : public Kernel
nrThreads = (nrThreads + preferredMultiple - 1) / preferredMultiple * preferredMultiple;
//std::cout << "nrBlocks = " << nrBlocks << ", nrPasses = " << nrPasses << ", preferredMultiple = " << preferredMultiple << ", nrThreads = " << nrThreads << std::endl;
- globalWorkSize = cl::NDRange(nrPasses * nrThreads, ps.nrChannelsPerSubband());
+ unsigned nrUsableChannels = std::max(ps.nrChannelsPerSubband() - 1, 1U);
+ globalWorkSize = cl::NDRange(nrPasses * nrThreads, nrUsableChannels);
localWorkSize = cl::NDRange(nrThreads, 1);
- nrOperations = (size_t) ps.nrChannelsPerSubband() * ps.nrBaselines() * ps.nrSamplesPerChannel() * 32;
- nrBytesRead = (size_t) nrPasses * ps.nrStations() * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * NR_POLARIZATIONS * sizeof(std::complex<float>);
- nrBytesWritten = (size_t) ps.nrBaselines() * ps.nrChannelsPerSubband() * NR_POLARIZATIONS * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ nrOperations = (size_t) nrUsableChannels * ps.nrBaselines() * ps.nrSamplesPerChannel() * 32;
+ nrBytesRead = (size_t) nrPasses * ps.nrStations() * nrUsableChannels * ps.nrSamplesPerChannel() * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ nrBytesWritten = (size_t) ps.nrBaselines() * nrUsableChannels * NR_POLARIZATIONS * NR_POLARIZATIONS * sizeof(std::complex<float>);
}
};
@@ -731,7 +781,7 @@ class CorrelatorKernel : public Kernel
CorrelatorKernel(const Parset &ps, cl::CommandQueue &queue, cl::Program &program, cl::Buffer &devVisibilities, cl::Buffer &devCorrectedData)
:
#if defined USE_2X2
- Kernel(ps, program, "correlateRectangles")
+ Kernel(ps, program, "correlate")
#else
#error not implemented
#endif
@@ -739,17 +789,90 @@ class CorrelatorKernel : public Kernel
setArg(0, devVisibilities);
setArg(1, devCorrectedData);
- unsigned nrRectanglesPerSide = ((ps.nrStations() - 1) / (2 * 16));
+ unsigned nrRectanglesPerSide = (ps.nrStations() - 1) / (2 * 16);
+ unsigned nrRectangles = nrRectanglesPerSide * (nrRectanglesPerSide + 1) / 2;
+//#pragma omp critical (cout)
+ //std::cout << "nrRectangles = " << nrRectangles << std::endl;
+
+ unsigned nrBlocksPerSide = (ps.nrStations() + 2 * 16 - 1) / (2 * 16);
+ unsigned nrBlocks = nrBlocksPerSide * (nrBlocksPerSide + 1) / 2;
+//#pragma omp critical (cout)
+ //std::cout << "nrBlocks = " << nrBlocks << std::endl;
+
+ unsigned nrUsableChannels = std::max(ps.nrChannelsPerSubband() - 1, 1U);
+ globalWorkSize = cl::NDRange(16 * 16, nrBlocks, nrUsableChannels);
+ localWorkSize = cl::NDRange(16 * 16, 1, 1);
+
+ // FIXME
+ //nrOperations = (size_t) (32 * 32) * nrRectangles * nrUsableChannels * ps.nrSamplesPerChannel() * 32;
+ nrOperations = (size_t) ps.nrBaselines() * ps.nrSamplesPerSubband() * 32;
+ nrBytesRead = (size_t) (32 + 32) * nrRectangles * nrUsableChannels * ps.nrSamplesPerChannel() * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ nrBytesWritten = (size_t) (32 * 32) * nrRectangles * nrUsableChannels * NR_POLARIZATIONS * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ }
+};
+
+
+class CorrelateRectangleKernel : public Kernel
+{
+ public:
+ CorrelateRectangleKernel(const Parset &ps, cl::CommandQueue &queue, cl::Program &program, cl::Buffer &devVisibilities, cl::Buffer &devCorrectedData)
+ :
+#if defined USE_2X2
+ Kernel(ps, program, "correlateRectangleKernel")
+#else
+#error not implemented
+#endif
+ {
+ setArg(0, devVisibilities);
+ setArg(1, devCorrectedData);
+
+ unsigned nrRectanglesPerSide = (ps.nrStations() - 1) / (2 * 16);
unsigned nrRectangles = nrRectanglesPerSide * (nrRectanglesPerSide + 1) / 2;
#pragma omp critical (cout)
std::cout << "nrRectangles = " << nrRectangles << std::endl;
- globalWorkSize = cl::NDRange(16 * 16, nrRectangles, ps.nrChannelsPerSubband());
+ unsigned nrUsableChannels = std::max(ps.nrChannelsPerSubband() - 1, 1U);
+ globalWorkSize = cl::NDRange(16 * 16, nrRectangles, nrUsableChannels);
localWorkSize = cl::NDRange(16 * 16, 1, 1);
- nrOperations = (size_t) (32 * 32) * nrRectangles * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * 32;
- nrBytesRead = (size_t) (32 + 32) * nrRectangles * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * NR_POLARIZATIONS * sizeof(std::complex<float>);
- nrBytesWritten = (size_t) (32 * 32) * nrRectangles * ps.nrChannelsPerSubband() * NR_POLARIZATIONS * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ nrOperations = (size_t) (32 * 32) * nrRectangles * nrUsableChannels * ps.nrSamplesPerChannel() * 32;
+ nrBytesRead = (size_t) (32 + 32) * nrRectangles * nrUsableChannels * ps.nrSamplesPerChannel() * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ nrBytesWritten = (size_t) (32 * 32) * nrRectangles * nrUsableChannels * NR_POLARIZATIONS * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ }
+};
+
+
+class CorrelateTriangleKernel : public Kernel
+{
+ public:
+ CorrelateTriangleKernel(const Parset &ps, cl::CommandQueue &queue, cl::Program &program, cl::Buffer &devVisibilities, cl::Buffer &devCorrectedData)
+ :
+#if defined USE_2X2
+ Kernel(ps, program, "correlateTriangleKernel")
+#else
+#error not implemented
+#endif
+ {
+ setArg(0, devVisibilities);
+ setArg(1, devCorrectedData);
+
+ unsigned nrTriangles = (ps.nrStations() + 2 * 16 - 1) / (2 * 16);
+ unsigned nrMiniBlocksPerSide = 16;
+ unsigned nrMiniBlocks = nrMiniBlocksPerSide * (nrMiniBlocksPerSide + 1) / 2;
+ size_t preferredMultiple;
+ getWorkGroupInfo(queue.getInfo<CL_QUEUE_DEVICE>(), CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE, &preferredMultiple);
+ unsigned nrThreads = align(nrMiniBlocks, preferredMultiple);
+
+#pragma omp critical (cout)
+ std::cout << "nrTriangles = " << nrTriangles << ", nrMiniBlocks = " << nrMiniBlocks << ", nrThreads = " << nrThreads << std::endl;
+
+ unsigned nrUsableChannels = std::max(ps.nrChannelsPerSubband() - 1, 1U);
+ globalWorkSize = cl::NDRange(nrThreads, nrTriangles, nrUsableChannels);
+ localWorkSize = cl::NDRange(nrThreads, 1, 1);
+
+ nrOperations = (size_t) (32 * 32 / 2) * nrTriangles * nrUsableChannels * ps.nrSamplesPerChannel() * 32;
+ nrBytesRead = (size_t) 32 * nrTriangles * nrUsableChannels * ps.nrSamplesPerChannel() * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ nrBytesWritten = (size_t) (32 * 32 / 2) * nrTriangles * nrUsableChannels * NR_POLARIZATIONS * NR_POLARIZATIONS * sizeof(std::complex<float>);
}
};
@@ -815,18 +938,18 @@ class BeamFormerKernel : public Kernel
setArg(1, devCorrectedData);
setArg(2, devBeamFormerWeights);
- globalWorkSize = cl::NDRange(NR_POLARIZATIONS, ps.nrTABs(SAP), ps.nrChannelsPerSubband());
- localWorkSize = cl::NDRange(NR_POLARIZATIONS, ps.nrTABs(SAP), 1);
+ globalWorkSize = cl::NDRange(NR_POLARIZATIONS, ps.nrTABs(0), ps.nrChannelsPerSubband());
+ localWorkSize = cl::NDRange(NR_POLARIZATIONS, ps.nrTABs(0), 1);
// FIXME: nrTABs
- //queue.enqueueNDRangeKernel(*this, cl::NullRange, cl::NDRange(16, ps.nrTABs(SAP), ps.nrChannelsPerSubband()), cl::NDRange(16, ps.nrTABs(SAP), 1), 0, &event);
+ //queue.enqueueNDRangeKernel(*this, cl::NullRange, cl::NDRange(16, ps.nrTABs(0), ps.nrChannelsPerSubband()), cl::NDRange(16, ps.nrTABs(0), 1), 0, &event);
size_t count = ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * NR_POLARIZATIONS;
- size_t nrWeightsBytes = ps.nrStations() * ps.nrTABs(SAP) * ps.nrChannelsPerSubband() * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ size_t nrWeightsBytes = ps.nrStations() * ps.nrTABs(0) * ps.nrChannelsPerSubband() * NR_POLARIZATIONS * sizeof(std::complex<float>);
size_t nrSampleBytesPerPass = count * ps.nrStations() * sizeof(std::complex<float>);
- size_t nrComplexVoltagesBytesPerPass = count * ps.nrTABs(SAP) * sizeof(std::complex<float>);
+ size_t nrComplexVoltagesBytesPerPass = count * ps.nrTABs(0) * sizeof(std::complex<float>);
unsigned nrPasses = std::max((ps.nrStations() + 6) / 16, 1U);
- nrOperations = count * ps.nrStations() * ps.nrTABs(SAP) * 8;
+ nrOperations = count * ps.nrStations() * ps.nrTABs(0) * 8;
nrBytesRead = nrWeightsBytes + nrSampleBytesPerPass + (nrPasses - 1) * nrComplexVoltagesBytesPerPass;
nrBytesWritten = nrPasses * nrComplexVoltagesBytesPerPass;
}
@@ -844,14 +967,14 @@ class BeamFormerTransposeKernel : public Kernel
setArg(0, devTransposedData);
setArg(1, devComplexVoltages);
- //globalWorkSize = cl::NDRange(256, (ps.nrTABs(SAP) + 15) / 16, (ps.nrChannelsPerSubband() + 15) / 16);
- globalWorkSize = cl::NDRange(256, (ps.nrTABs(SAP) + 15) / 16, ps.nrSamplesPerChannel() / 16);
+ //globalWorkSize = cl::NDRange(256, (ps.nrTABs(0) + 15) / 16, (ps.nrChannelsPerSubband() + 15) / 16);
+ globalWorkSize = cl::NDRange(256, (ps.nrTABs(0) + 15) / 16, ps.nrSamplesPerChannel() / 16);
localWorkSize = cl::NDRange(256, 1, 1);
nrOperations = 0;
- nrBytesRead = (size_t) ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * ps.nrTABs(SAP) * NR_POLARIZATIONS * sizeof(std::complex<float>),
- //nrBytesWritten = (size_t) ps.nrTABs(SAP) * NR_POLARIZATIONS * ps.nrSamplesPerChannel() * ps.nrChannelsPerSubband() * sizeof(std::complex<float>);
- nrBytesWritten = (size_t) ps.nrTABs(SAP) * NR_POLARIZATIONS * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * sizeof(std::complex<float>);
+ nrBytesRead = (size_t) ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * ps.nrTABs(0) * NR_POLARIZATIONS * sizeof(std::complex<float>),
+ //nrBytesWritten = (size_t) ps.nrTABs(0) * NR_POLARIZATIONS * ps.nrSamplesPerChannel() * ps.nrChannelsPerSubband() * sizeof(std::complex<float>);
+ nrBytesWritten = (size_t) ps.nrTABs(0) * NR_POLARIZATIONS * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * sizeof(std::complex<float>);
}
};
@@ -871,7 +994,7 @@ class Dedispersion_FFT_Kernel
void enqueue(cl::CommandQueue &queue, PerformanceCounter &counter, clFFT_Direction direction)
{
- size_t nrFFTs = (size_t) ps.nrTABs(SAP) * NR_POLARIZATIONS * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() / ps.dedispersionFFTsize();
+ size_t nrFFTs = (size_t) ps.nrTABs(0) * NR_POLARIZATIONS * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() / ps.dedispersionFFTsize();
cl_int error = clFFT_ExecuteInterleaved(queue(), plan.plan, nrFFTs, direction, buffer(), buffer(), 0, 0, &event());
@@ -896,7 +1019,7 @@ class DedispersionForwardFFTkernel : public FFT_Kernel
public:
DedispersionForwardFFTkernel(const Parset &ps, cl::Context &context, cl::Buffer &buffer)
:
- FFT_Kernel(context, ps.dedispersionFFTsize(), ps.nrTABs(SAP) * NR_POLARIZATIONS * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() / ps.dedispersionFFTsize(), true, buffer)
+ FFT_Kernel(context, ps.dedispersionFFTsize(), ps.nrTABs(0) * NR_POLARIZATIONS * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() / ps.dedispersionFFTsize(), true, buffer)
{
ASSERT(ps.nrSamplesPerChannel() % ps.dedispersionFFTsize() == 0);
}
@@ -908,7 +1031,7 @@ class DedispersionBackwardFFTkernel : public FFT_Kernel
public:
DedispersionBackwardFFTkernel(const Parset &ps, cl::Context &context, cl::Buffer &buffer)
:
- FFT_Kernel(context, ps.dedispersionFFTsize(), ps.nrTABs(SAP) * NR_POLARIZATIONS * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() / ps.dedispersionFFTsize(), false, buffer)
+ FFT_Kernel(context, ps.dedispersionFFTsize(), ps.nrTABs(0) * NR_POLARIZATIONS * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() / ps.dedispersionFFTsize(), false, buffer)
{
ASSERT(ps.nrSamplesPerChannel() % ps.dedispersionFFTsize() == 0);
}
@@ -946,8 +1069,8 @@ class DedispersionChirpKernel : public Kernel
//std::cout << "localWorkSize = NDRange(" << fftSize / divisor << ", 1, 1))" << std::endl;
}
- nrOperations = (size_t) NR_POLARIZATIONS * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * (9 * ps.nrTABs(SAP) + 17),
- nrBytesRead = nrBytesWritten = sizeof(std::complex<float>) * ps.nrTABs(SAP) * NR_POLARIZATIONS * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel();
+ nrOperations = (size_t) NR_POLARIZATIONS * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * (9 * ps.nrTABs(0) + 17),
+ nrBytesRead = nrBytesWritten = sizeof(std::complex<float>) * ps.nrTABs(0) * NR_POLARIZATIONS * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel();
}
void enqueue(cl::CommandQueue &queue, PerformanceCounter &counter, double subbandFrequency)
@@ -970,12 +1093,12 @@ class CoherentStokesKernel : public Kernel
setArg(0, devStokesData);
setArg(1, devComplexVoltages);
- globalWorkSize = cl::NDRange(256, (ps.nrTABs(SAP) + 15) / 16, (ps.nrChannelsPerSubband() + 15) / 16);
+ globalWorkSize = cl::NDRange(256, (ps.nrTABs(0) + 15) / 16, (ps.nrChannelsPerSubband() + 15) / 16);
localWorkSize = cl::NDRange(256, 1, 1);
- nrOperations = (size_t) ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * ps.nrTABs(SAP) * (ps.nrCoherentStokes() == 1 ? 8 : 20 + 2.0 / ps.coherentStokesTimeIntegrationFactor());
- nrBytesRead = (size_t) ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * ps.nrTABs(SAP) * NR_POLARIZATIONS * sizeof(std::complex<float>);
- nrBytesWritten = (size_t) ps.nrTABs(SAP) * ps.nrCoherentStokes() * ps.nrSamplesPerChannel() / ps.coherentStokesTimeIntegrationFactor() * ps.nrChannelsPerSubband() * sizeof(float);
+ nrOperations = (size_t) ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * ps.nrTABs(0) * (ps.nrCoherentStokes() == 1 ? 8 : 20 + 2.0 / ps.coherentStokesTimeIntegrationFactor());
+ nrBytesRead = (size_t) ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * ps.nrTABs(0) * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ nrBytesWritten = (size_t) ps.nrTABs(0) * ps.nrCoherentStokes() * ps.nrSamplesPerChannel() / ps.coherentStokesTimeIntegrationFactor() * ps.nrChannelsPerSubband() * sizeof(float);
}
};
@@ -992,31 +1115,31 @@ class UHEP_BeamFormerKernel : public Kernel
setArg(2, devBeamFormerWeights);
#if 1
- globalWorkSize = cl::NDRange(NR_POLARIZATIONS, ps.nrTABs(SAP), ps.nrSubbands());
- localWorkSize = cl::NDRange(NR_POLARIZATIONS, ps.nrTABs(SAP), 1);
+ globalWorkSize = cl::NDRange(NR_POLARIZATIONS, ps.nrTABs(0), ps.nrSubbands());
+ localWorkSize = cl::NDRange(NR_POLARIZATIONS, ps.nrTABs(0), 1);
size_t count = ps.nrSubbands() * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) * NR_POLARIZATIONS;
- size_t nrWeightsBytes = ps.nrStations() * ps.nrTABs(SAP) * ps.nrSubbands() * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ size_t nrWeightsBytes = ps.nrStations() * ps.nrTABs(0) * ps.nrSubbands() * NR_POLARIZATIONS * sizeof(std::complex<float>);
size_t nrSampleBytes = count * ps.nrStations() * ps.nrBytesPerComplexSample();
- size_t nrComplexVoltagesBytesPerPass = count * ps.nrTABs(SAP) * sizeof(std::complex<float>);
+ size_t nrComplexVoltagesBytesPerPass = count * ps.nrTABs(0) * sizeof(std::complex<float>);
unsigned nrPasses = std::max((ps.nrStations() + 6) / 16, 1U);
- nrOperations = count * ps.nrStations() * ps.nrTABs(SAP) * 8;
+ nrOperations = count * ps.nrStations() * ps.nrTABs(0) * 8;
nrBytesRead = nrWeightsBytes + nrSampleBytes + (nrPasses - 1) * nrComplexVoltagesBytesPerPass;
nrBytesWritten = nrPasses * nrComplexVoltagesBytesPerPass;
#else
- ASSERT(ps.nrTABs(SAP) % 3 == 0);
+ ASSERT(ps.nrTABs(0) % 3 == 0);
ASSERT(ps.nrStations() % 6 == 0);
- unsigned nrThreads = NR_POLARIZATIONS * (ps.nrTABs(SAP) / 3) * (ps.nrStations() / 6);
+ unsigned nrThreads = NR_POLARIZATIONS * (ps.nrTABs(0) / 3) * (ps.nrStations() / 6);
globalWorkSize = cl::NDRange(nrThreads, ps.nrSubbands());
localWorkSize = cl::NDRange(nrThreads, 1);
- //globalWorkSize = cl::NDRange(ps.nrStations() / 6, ps.nrTABs(SAP) / 3, ps.nrSubbands());
- //localWorkSize = cl::NDRange(ps.nrStations() / 6, ps.nrTABs(SAP) / 3, 1);
+ //globalWorkSize = cl::NDRange(ps.nrStations() / 6, ps.nrTABs(0) / 3, ps.nrSubbands());
+ //localWorkSize = cl::NDRange(ps.nrStations() / 6, ps.nrTABs(0) / 3, 1);
size_t count = ps.nrSubbands() * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) * NR_POLARIZATIONS;
- size_t nrWeightsBytes = ps.nrStations() * ps.nrTABs(SAP) * ps.nrSubbands() * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ size_t nrWeightsBytes = ps.nrStations() * ps.nrTABs(0) * ps.nrSubbands() * NR_POLARIZATIONS * sizeof(std::complex<float>);
size_t nrSampleBytes = count * ps.nrStations() * ps.nrBytesPerComplexSample();
- size_t nrComplexVoltagesBytes = count * ps.nrTABs(SAP) * sizeof(std::complex<float>);
- nrOperations = count * ps.nrStations() * ps.nrTABs(SAP) * 8;
+ size_t nrComplexVoltagesBytes = count * ps.nrTABs(0) * sizeof(std::complex<float>);
+ nrOperations = count * ps.nrStations() * ps.nrTABs(0) * 8;
nrBytesRead = nrWeightsBytes + nrSampleBytes;
nrBytesWritten = nrComplexVoltagesBytes;
#endif
@@ -1035,12 +1158,12 @@ class UHEP_TransposeKernel : public Kernel
setArg(1, devComplexVoltages);
setArg(2, devReverseSubbandMapping);
- globalWorkSize = cl::NDRange(256, (ps.nrTABs(SAP) + 15) / 16, 512 / 16);
+ globalWorkSize = cl::NDRange(256, (ps.nrTABs(0) + 15) / 16, 512 / 16);
localWorkSize = cl::NDRange(256, 1, 1);
nrOperations = 0;
- nrBytesRead = (size_t) ps.nrSubbands() * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) * ps.nrTABs(SAP) * NR_POLARIZATIONS * sizeof(std::complex<float>);
- nrBytesWritten = (size_t) ps.nrTABs(SAP) * NR_POLARIZATIONS * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) * 512 * sizeof(std::complex<float>);
+ nrBytesRead = (size_t) ps.nrSubbands() * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) * ps.nrTABs(0) * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ nrBytesWritten = (size_t) ps.nrTABs(0) * NR_POLARIZATIONS * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) * 512 * sizeof(std::complex<float>);
}
};
@@ -1055,10 +1178,10 @@ class UHEP_InvFFT_Kernel : public Kernel
setArg(0, devFFTedData);
setArg(1, devFFTedData);
- globalWorkSize = cl::NDRange(128, ps.nrTABs(SAP) * NR_POLARIZATIONS * ps.nrSamplesPerChannel());
+ globalWorkSize = cl::NDRange(128, ps.nrTABs(0) * NR_POLARIZATIONS * ps.nrSamplesPerChannel());
localWorkSize = cl::NDRange(128, 1);
- size_t nrFFTs = (size_t) ps.nrTABs(SAP) * NR_POLARIZATIONS * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1);
+ size_t nrFFTs = (size_t) ps.nrTABs(0) * NR_POLARIZATIONS * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1);
nrOperations = nrFFTs * 5 * 1024 * 10;
nrBytesRead = nrFFTs * 512 * sizeof(std::complex<float>);
nrBytesWritten = nrFFTs * 1024 * sizeof(float);
@@ -1083,10 +1206,10 @@ class UHEP_InvFIR_Kernel : public Kernel
for (nrThreads = 1024; nrThreads > maxNrThreads; nrThreads /= 2)
;
- globalWorkSize = cl::NDRange(1024, NR_POLARIZATIONS, ps.nrTABs(SAP));
+ globalWorkSize = cl::NDRange(1024, NR_POLARIZATIONS, ps.nrTABs(0));
localWorkSize = cl::NDRange(nrThreads, 1, 1);
- size_t count = ps.nrTABs(SAP) * NR_POLARIZATIONS * 1024;
+ size_t count = ps.nrTABs(0) * NR_POLARIZATIONS * 1024;
nrOperations = count * ps.nrSamplesPerChannel() * NR_STATION_FILTER_TAPS * 2;
nrBytesRead = count * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) * sizeof(float);
nrBytesWritten = count * ps.nrSamplesPerChannel() * sizeof(float);
@@ -1104,12 +1227,12 @@ class UHEP_TriggerKernel : public Kernel
setArg(0, devTriggerInfo);
setArg(1, devInvFIRfilteredData);
- globalWorkSize = cl::NDRange(16, 16, ps.nrTABs(SAP));
+ globalWorkSize = cl::NDRange(16, 16, ps.nrTABs(0));
localWorkSize = cl::NDRange(16, 16, 1);
- nrOperations = (size_t) ps.nrTABs(SAP) * ps.nrSamplesPerChannel() * 1024 * (3 /* power */ + 2 /* window */ + 1 /* max */ + 7 /* mean/variance */);
- nrBytesRead = (size_t) ps.nrTABs(SAP) * NR_POLARIZATIONS * ps.nrSamplesPerChannel() * 1024 * sizeof(float);
- nrBytesWritten = (size_t) ps.nrTABs(SAP) * sizeof(TriggerInfo);
+ nrOperations = (size_t) ps.nrTABs(0) * ps.nrSamplesPerChannel() * 1024 * (3 /* power */ + 2 /* window */ + 1 /* max */ + 7 /* mean/variance */);
+ nrBytesRead = (size_t) ps.nrTABs(0) * NR_POLARIZATIONS * ps.nrSamplesPerChannel() * 1024 * sizeof(float);
+ nrBytesWritten = (size_t) ps.nrTABs(0) * sizeof(TriggerInfo);
}
};
@@ -1143,6 +1266,8 @@ CorrelatorWorkQueue::CorrelatorWorkQueue(CorrelatorPipeline &pipeline)
inputSamples(boost::extents[ps.nrStations()][(ps.nrSamplesPerChannel() + NR_TAPS - 1) * ps.nrChannelsPerSubband()][NR_POLARIZATIONS][ps.nrBytesPerComplexSample()], queue, CL_MEM_WRITE_ONLY, devBufferA),
visibilities(boost::extents[ps.nrBaselines()][ps.nrChannelsPerSubband()][NR_POLARIZATIONS][NR_POLARIZATIONS], queue, CL_MEM_READ_ONLY, devBufferB)
{
+ memset(inputSamples.origin(), 0, inputSamples.bytesize()); // FIXME
+ memset(visibilities.origin(), 0, visibilities.bytesize()); // FIXME
size_t firWeightsSize = ps.nrChannelsPerSubband() * NR_TAPS * sizeof(float);
devFIRweights = cl::Buffer(pipeline.context, CL_MEM_READ_ONLY, firWeightsSize);
queue.enqueueWriteBuffer(devFIRweights, CL_TRUE, 0, ps.nrChannelsPerSubband() * NR_TAPS * sizeof(float), pipeline.filterBank.getWeights().origin());
@@ -1168,7 +1293,12 @@ void CorrelatorWorkQueue::doWork()
FIR_FilterKernel firFilterKernel(ps, queue, pipeline.firFilterProgram, devFilteredData, inputSamples, devFIRweights);
Filter_FFT_Kernel fftKernel(ps, pipeline.context, devFilteredData);
DelayAndBandPassKernel delayAndBandPassKernel(ps, pipeline.delayAndBandPassProgram, devCorrectedData, devFilteredData, delaysAtBegin, delaysAfterEnd, phaseOffsets, bandPassCorrectionWeights);
+#if defined USE_NEW_CORRELATOR
+ CorrelateTriangleKernel correlateTriangleKernel(ps, queue, pipeline.correlatorProgram, visibilities, devCorrectedData);
+ CorrelateRectangleKernel correlateRectangleKernel(ps, queue, pipeline.correlatorProgram, visibilities, devCorrectedData);
+#else
CorrelatorKernel correlatorKernel(ps, queue, pipeline.correlatorProgram, visibilities, devCorrectedData);
+#endif
double startTime = ps.startTime(), currentTime, stopTime = ps.stopTime(), blockTime = ps.CNintegrationTime();
#pragma omp barrier
@@ -1176,7 +1306,7 @@ void CorrelatorWorkQueue::doWork()
double executionStartTime = getTime();
for (unsigned block = 0; (currentTime = startTime + block * blockTime) < stopTime; block ++) {
-#pragma omp single
+#pragma omp single nowait
#pragma omp critical (cout)
std::cout << "block = " << block << ", time = " << to_simple_string(from_ustime_t(currentTime)) << std::endl;
@@ -1185,17 +1315,14 @@ void CorrelatorWorkQueue::doWork()
memset(phaseOffsets.origin(), 0, phaseOffsets.bytesize());
// FIXME!!!
- if (ps.nrStations() >= 3)
- delaysAtBegin[0][2][0] = 1e-6, delaysAfterEnd[0][2][0] = 1.1e-6;
+ //if (ps.nrStations() >= 3)
+ //delaysAtBegin[0][2][0] = 1e-6, delaysAfterEnd[0][2][0] = 1.1e-6;
delaysAtBegin.hostToDevice(CL_FALSE);
delaysAfterEnd.hostToDevice(CL_FALSE);
phaseOffsets.hostToDevice(CL_FALSE);
- queue.finish();
-#pragma omp barrier
-
-#pragma omp for schedule(dynamic)
+#pragma omp for schedule(dynamic), nowait
for (unsigned subband = 0; subband < ps.nrSubbands(); subband ++) {
try {
#if defined USE_TEST_DATA
@@ -1217,7 +1344,12 @@ void CorrelatorWorkQueue::doWork()
}
delayAndBandPassKernel.enqueue(queue, pipeline.delayAndBandPassCounter, subband);
+#if defined USE_NEW_CORRELATOR
+ correlateTriangleKernel.enqueue(queue, pipeline.correlateTriangleCounter);
+ correlateRectangleKernel.enqueue(queue, pipeline.correlateRectangleCounter);
+#else
correlatorKernel.enqueue(queue, pipeline.correlatorCounter);
+#endif
queue.finish();
{
@@ -1276,11 +1408,11 @@ BeamFormerWorkQueue::BeamFormerWorkQueue(BeamFormerPipeline &pipeline)
delaysAfterEnd(boost::extents[ps.nrBeams()][ps.nrStations()][NR_POLARIZATIONS], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY),
phaseOffsets(boost::extents[ps.nrBeams()][NR_POLARIZATIONS], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY),
devCorrectedData(cl::Buffer(pipeline.context, CL_MEM_READ_WRITE, ps.nrStations() * ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * NR_POLARIZATIONS * sizeof(std::complex<float>))),
- beamFormerWeights(boost::extents[ps.nrStations()][ps.nrChannelsPerSubband()][ps.nrTABs(SAP)], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY),
- devComplexVoltages(cl::Buffer(pipeline.context, CL_MEM_READ_WRITE, ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * ps.nrTABs(SAP) * NR_POLARIZATIONS * sizeof(std::complex<float>))),
- //transposedComplexVoltages(boost::extents[ps.nrTABs(SAP)][NR_POLARIZATIONS][ps.nrSamplesPerChannel()][ps.nrChannelsPerSubband()], queue, CL_MEM_READ_ONLY, CL_MEM_READ_WRITE)
- transposedComplexVoltages(boost::extents[ps.nrTABs(SAP)][NR_POLARIZATIONS][ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel()], queue, CL_MEM_READ_ONLY, CL_MEM_READ_WRITE),
- DMs(boost::extents[ps.nrTABs(SAP)], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY)
+ beamFormerWeights(boost::extents[ps.nrStations()][ps.nrChannelsPerSubband()][ps.nrTABs(0)], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY),
+ devComplexVoltages(cl::Buffer(pipeline.context, CL_MEM_READ_WRITE, ps.nrChannelsPerSubband() * ps.nrSamplesPerChannel() * ps.nrTABs(0) * NR_POLARIZATIONS * sizeof(std::complex<float>))),
+ //transposedComplexVoltages(boost::extents[ps.nrTABs(0)][NR_POLARIZATIONS][ps.nrSamplesPerChannel()][ps.nrChannelsPerSubband()], queue, CL_MEM_READ_ONLY, CL_MEM_READ_WRITE)
+ transposedComplexVoltages(boost::extents[ps.nrTABs(0)][NR_POLARIZATIONS][ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel()], queue, CL_MEM_READ_ONLY, CL_MEM_READ_WRITE),
+ DMs(boost::extents[ps.nrTABs(0)], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY)
{
if (ps.correctBandPass()) {
BandPass::computeCorrectionFactors(bandPassCorrectionWeights.origin(), ps.nrChannelsPerSubband());
@@ -1310,7 +1442,7 @@ void BeamFormerWorkQueue::doWork()
double executionStartTime = getTime();
for (unsigned block = 0; (currentTime = startTime + block * blockTime) < stopTime; block ++) {
-#pragma omp single
+#pragma omp single nowait
#pragma omp critical (cout)
std::cout << "block = " << block << ", time = " << to_simple_string(from_ustime_t(currentTime)) << std::endl;
@@ -1326,11 +1458,8 @@ void BeamFormerWorkQueue::doWork()
delaysAfterEnd.hostToDevice(CL_FALSE);
phaseOffsets.hostToDevice(CL_FALSE);
beamFormerWeights.hostToDevice(CL_FALSE);
- queue.finish();
-
-#pragma omp barrier
-#pragma omp for schedule(dynamic)
+#pragma omp for schedule(dynamic), nowait
for (unsigned subband = 0; subband < ps.nrSubbands(); subband ++) {
try {
#if 1
@@ -1384,13 +1513,13 @@ UHEP_WorkQueue::UHEP_WorkQueue(UHEP_Pipeline &pipeline)
WorkQueue(pipeline),
pipeline(pipeline),
hostInputSamples(boost::extents[ps.nrStations()][ps.nrSubbands()][ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1][NR_POLARIZATIONS][ps.nrBytesPerComplexSample()], queue, CL_MEM_WRITE_ONLY),
- hostBeamFormerWeights(boost::extents[ps.nrStations()][ps.nrSubbands()][ps.nrTABs(SAP)], queue, CL_MEM_WRITE_ONLY),
- hostTriggerInfo(ps.nrTABs(SAP), queue, CL_MEM_READ_ONLY)
+ hostBeamFormerWeights(boost::extents[ps.nrStations()][ps.nrSubbands()][ps.nrTABs(0)], queue, CL_MEM_WRITE_ONLY),
+ hostTriggerInfo(ps.nrTABs(0), queue, CL_MEM_READ_ONLY)
{
size_t inputSamplesSize = ps.nrStations() * ps.nrSubbands() * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) * NR_POLARIZATIONS * ps.nrBytesPerComplexSample();
- size_t complexVoltagesSize = ps.nrSubbands() * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) * ps.nrTABs(SAP) * NR_POLARIZATIONS * sizeof(std::complex<float>);
- size_t transposedDataSize = ps.nrTABs(SAP) * NR_POLARIZATIONS * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) * 512 * sizeof(std::complex<float>);
- size_t invFIRfilteredDataSize = ps.nrTABs(SAP) * NR_POLARIZATIONS * ps.nrSamplesPerChannel() * 512 * sizeof(std::complex<float>);
+ size_t complexVoltagesSize = ps.nrSubbands() * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) * ps.nrTABs(0) * NR_POLARIZATIONS * sizeof(std::complex<float>);
+ size_t transposedDataSize = ps.nrTABs(0) * NR_POLARIZATIONS * (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) * 512 * sizeof(std::complex<float>);
+ size_t invFIRfilteredDataSize = ps.nrTABs(0) * NR_POLARIZATIONS * ps.nrSamplesPerChannel() * 512 * sizeof(std::complex<float>);
size_t buffer0size = std::max(inputSamplesSize, transposedDataSize);
size_t buffer1size = std::max(complexVoltagesSize, invFIRfilteredDataSize);
@@ -1398,7 +1527,7 @@ UHEP_WorkQueue::UHEP_WorkQueue(UHEP_Pipeline &pipeline)
devBuffers[0] = cl::Buffer(pipeline.context, CL_MEM_READ_WRITE, buffer0size);
devBuffers[1] = cl::Buffer(pipeline.context, CL_MEM_READ_WRITE, buffer1size);
- size_t beamFormerWeightsSize = ps.nrStations() * ps.nrSubbands() * ps.nrTABs(SAP) * sizeof(std::complex<float>);
+ size_t beamFormerWeightsSize = ps.nrStations() * ps.nrSubbands() * ps.nrTABs(0) * sizeof(std::complex<float>);
devBeamFormerWeights = cl::Buffer(pipeline.context, CL_MEM_READ_ONLY, beamFormerWeightsSize);
devInputSamples = devBuffers[0];
@@ -1409,7 +1538,7 @@ UHEP_WorkQueue::UHEP_WorkQueue(UHEP_Pipeline &pipeline)
devFFTedData = devBuffers[0];
devInvFIRfilteredData = devBuffers[1];
- devTriggerInfo = cl::Buffer(pipeline.context, CL_MEM_WRITE_ONLY, ps.nrTABs(SAP) * sizeof(TriggerInfo));
+ devTriggerInfo = cl::Buffer(pipeline.context, CL_MEM_WRITE_ONLY, ps.nrTABs(0) * sizeof(TriggerInfo));
}
@@ -1430,12 +1559,12 @@ void UHEP_WorkQueue::doWork(const float * /*delaysAtBegin*/, const float * /*del
double executionStartTime = getTime();
-#pragma omp for schedule(dynamic)
+#pragma omp for schedule(dynamic), nowait
for (unsigned block = 0; block < nrBlocks; block ++) {
try {
double currentTime = startTime + block * blockTime;
-//#pragma omp single // FIXME: why does the compiler complain here???
+//#pragma omp single nowait // FIXME: why does the compiler complain here???
#pragma omp critical (cout)
std::cout << "block = " << block << ", time = " << to_simple_string(from_ustime_t(currentTime)) << std::endl;
@@ -1607,18 +1736,21 @@ struct CorrelatorTest : public UnitTest
{
CorrelatorTest(const Parset &ps)
:
- //UnitTest(ps, "Correlator.cl")
+#if defined USE_NEW_CORRELATOR
UnitTest(ps, "NewCorrelator.cl")
+#else
+ UnitTest(ps, "Correlator.cl")
+#endif
{
if (ps.nrStations() >= 5 && ps.nrChannelsPerSubband() >= 6 && ps.nrSamplesPerChannel() >= 100) {
- MultiArraySharedBuffer<std::complex<float>, 4> inputData(boost::extents[ps.nrStations()][ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel()][NR_POLARIZATIONS], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY);
MultiArraySharedBuffer<std::complex<float>, 4> visibilities(boost::extents[ps.nrBaselines()][ps.nrChannelsPerSubband()][NR_POLARIZATIONS][NR_POLARIZATIONS], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
+ MultiArraySharedBuffer<std::complex<float>, 4> inputData(boost::extents[ps.nrStations()][ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel()][NR_POLARIZATIONS], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY);
CorrelatorKernel correlator(ps, queue, program, visibilities, inputData);
//inputData[3][5][99][1] = std::complex<float>(3, 4);
//inputData[4][5][99][1] = std::complex<float>(5, 6);
- inputData[2][5][99][1] = std::complex<float>(3, 4);
- inputData[65][5][99][1] = std::complex<float>(5, 6);
+ inputData[0][5][99][1] = std::complex<float>(3, 4);
+ inputData[2][5][99][1] = std::complex<float>(5, 6);
visibilities.hostToDevice(CL_FALSE);
inputData.hostToDevice(CL_FALSE);
@@ -1635,6 +1767,70 @@ visibilities.hostToDevice(CL_FALSE);
};
+#if defined USE_NEW_CORRELATOR
+
+struct CorrelateRectangleTest : public UnitTest
+{
+ CorrelateRectangleTest(const Parset &ps)
+ :
+ //UnitTest(ps, "Correlator.cl")
+ UnitTest(ps, "NewCorrelator.cl")
+ {
+ if (ps.nrStations() >= 5 && ps.nrChannelsPerSubband() >= 6 && ps.nrSamplesPerChannel() >= 100) {
+ MultiArraySharedBuffer<std::complex<float>, 4> visibilities(boost::extents[ps.nrBaselines()][ps.nrChannelsPerSubband()][NR_POLARIZATIONS][NR_POLARIZATIONS], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
+ MultiArraySharedBuffer<std::complex<float>, 4> inputData(boost::extents[ps.nrStations()][ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel()][NR_POLARIZATIONS], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY);
+ CorrelateRectangleKernel correlator(ps, queue, program, visibilities, inputData);
+
+ inputData[27][5][99][1] = std::complex<float>(3, 4);
+ inputData[68][5][99][1] = std::complex<float>(5, 6);
+
+visibilities.hostToDevice(CL_FALSE);
+ inputData.hostToDevice(CL_FALSE);
+ correlator.enqueue(queue, counter);
+ visibilities.deviceToHost(CL_TRUE);
+
+ //check(visibilities[5463][5][1][1], std::complex<float>(39, 2));
+ for (unsigned bl = 0; bl < ps.nrBaselines(); bl ++)
+ if (visibilities[bl][5][1][1] != std::complex<float>(0, 0))
+ std::cout << "bl = " << bl << ", visibility = " << visibilities[bl][5][1][1] << std::endl;
+ }
+ }
+};
+
+
+struct CorrelateTriangleTest : public UnitTest
+{
+ CorrelateTriangleTest(const Parset &ps)
+ :
+ //UnitTest(ps, "Correlator.cl")
+ UnitTest(ps, "NewCorrelator.cl")
+ {
+ if (ps.nrStations() >= 5 && ps.nrChannelsPerSubband() >= 6 && ps.nrSamplesPerChannel() >= 100) {
+ MultiArraySharedBuffer<std::complex<float>, 4> visibilities(boost::extents[ps.nrBaselines()][ps.nrChannelsPerSubband()][NR_POLARIZATIONS][NR_POLARIZATIONS], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
+ MultiArraySharedBuffer<std::complex<float>, 4> inputData(boost::extents[ps.nrStations()][ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel()][NR_POLARIZATIONS], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY);
+ CorrelateTriangleKernel correlator(ps, queue, program, visibilities, inputData);
+
+ //inputData[3][5][99][1] = std::complex<float>(3, 4);
+ //inputData[4][5][99][1] = std::complex<float>(5, 6);
+ inputData[0][5][99][1] = std::complex<float>(3, 4);
+ inputData[2][5][99][1] = std::complex<float>(5, 6);
+
+visibilities.hostToDevice(CL_FALSE);
+ inputData.hostToDevice(CL_FALSE);
+ correlator.enqueue(queue, counter);
+ visibilities.deviceToHost(CL_TRUE);
+
+ //check(visibilities[13][5][1][1], std::complex<float>(39, 2));
+ for (unsigned bl = 0; bl < ps.nrBaselines(); bl ++)
+ if (visibilities[bl][5][1][1] != std::complex<float>(0, 0))
+ std::cout << "bl = " << bl << ", visibility = " << visibilities[bl][5][1][1] << std::endl;
+ }
+ }
+};
+
+#endif
+
+
struct IncoherentStokesTest : public UnitTest
{
IncoherentStokesTest(const Parset &ps)
@@ -1699,10 +1895,10 @@ struct BeamFormerTest : public UnitTest
:
UnitTest(ps, "BeamFormer/BeamFormer.cl")
{
- if (ps.nrStations() >= 5 && ps.nrSamplesPerChannel() >= 13 && ps.nrChannelsPerSubband() >= 7 && ps.nrTABs(SAP) >= 6) {
+ if (ps.nrStations() >= 5 && ps.nrSamplesPerChannel() >= 13 && ps.nrChannelsPerSubband() >= 7 && ps.nrTABs(0) >= 6) {
MultiArraySharedBuffer<std::complex<float>, 4> inputData(boost::extents[ps.nrStations()][ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel()][NR_POLARIZATIONS], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY);
- MultiArraySharedBuffer<std::complex<float>, 3> beamFormerWeights(boost::extents[ps.nrStations()][ps.nrChannelsPerSubband()][ps.nrTABs(SAP)], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY);
- MultiArraySharedBuffer<std::complex<float>, 4> complexVoltages(boost::extents[ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel()][ps.nrTABs(SAP)][NR_POLARIZATIONS], queue, CL_MEM_READ_ONLY, CL_MEM_READ_WRITE);
+ MultiArraySharedBuffer<std::complex<float>, 3> beamFormerWeights(boost::extents[ps.nrStations()][ps.nrChannelsPerSubband()][ps.nrTABs(0)], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY);
+ MultiArraySharedBuffer<std::complex<float>, 4> complexVoltages(boost::extents[ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel()][ps.nrTABs(0)][NR_POLARIZATIONS], queue, CL_MEM_READ_ONLY, CL_MEM_READ_WRITE);
BeamFormerKernel beamFormer(ps, program, complexVoltages, inputData, beamFormerWeights);
inputData[4][6][12][1] = std::complex<float>(2.2, 3);
@@ -1716,7 +1912,7 @@ struct BeamFormerTest : public UnitTest
check(complexVoltages[6][12][5][1], std::complex<float>(-6.2, 23));
#if 0
- for (unsigned tab = 0; tab < ps.nrTABs(SAP); tab ++)
+ for (unsigned tab = 0; tab < ps.nrTABs(0); tab ++)
for (unsigned pol = 0; pol < NR_POLARIZATIONS; pol ++)
for (unsigned ch = 0; ch < ps.nrChannelsPerSubband(); ch ++)
for (unsigned t = 0; t < ps.nrSamplesPerChannel(); t ++)
@@ -1734,9 +1930,9 @@ struct BeamFormerTransposeTest : public UnitTest
:
UnitTest(ps, "BeamFormer/Transpose.cl")
{
- if (ps.nrChannelsPerSubband() >= 19 && ps.nrSamplesPerChannel() >= 175 && ps.nrTABs(SAP) >= 5) {
- MultiArraySharedBuffer<std::complex<float>, 4> transposedData(boost::extents[ps.nrTABs(SAP)][NR_POLARIZATIONS][ps.nrSamplesPerChannel()][ps.nrChannelsPerSubband()], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
- MultiArraySharedBuffer<std::complex<float>, 4> complexVoltages(boost::extents[ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel()][ps.nrTABs(SAP)][NR_POLARIZATIONS], queue, CL_MEM_READ_WRITE, CL_MEM_READ_ONLY);
+ if (ps.nrChannelsPerSubband() >= 19 && ps.nrSamplesPerChannel() >= 175 && ps.nrTABs(0) >= 5) {
+ MultiArraySharedBuffer<std::complex<float>, 4> transposedData(boost::extents[ps.nrTABs(0)][NR_POLARIZATIONS][ps.nrSamplesPerChannel()][ps.nrChannelsPerSubband()], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
+ MultiArraySharedBuffer<std::complex<float>, 4> complexVoltages(boost::extents[ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel()][ps.nrTABs(0)][NR_POLARIZATIONS], queue, CL_MEM_READ_WRITE, CL_MEM_READ_ONLY);
BeamFormerTransposeKernel transpose(ps, program, transposedData, complexVoltages);
complexVoltages[18][174][4][1] = std::complex<float>(24, 42);
@@ -1757,9 +1953,9 @@ struct DedispersionChirpTest : public UnitTest
:
UnitTest(ps, "BeamFormer/Dedispersion.cl")
{
- if (ps.nrTABs(SAP) > 3 && ps.nrChannelsPerSubband() > 13 && ps.nrSamplesPerChannel() / ps.dedispersionFFTsize() > 1 && ps.dedispersionFFTsize() > 77) {
- MultiArraySharedBuffer<std::complex<float>, 5> data(boost::extents[ps.nrTABs(SAP)][NR_POLARIZATIONS][ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel() / ps.dedispersionFFTsize()][ps.dedispersionFFTsize()], queue, CL_MEM_READ_WRITE, CL_MEM_READ_WRITE);
- MultiArraySharedBuffer<float, 1> DMs(boost::extents[ps.nrTABs(SAP)], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
+ if (ps.nrTABs(0) > 3 && ps.nrChannelsPerSubband() > 13 && ps.nrSamplesPerChannel() / ps.dedispersionFFTsize() > 1 && ps.dedispersionFFTsize() > 77) {
+ MultiArraySharedBuffer<std::complex<float>, 5> data(boost::extents[ps.nrTABs(0)][NR_POLARIZATIONS][ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel() / ps.dedispersionFFTsize()][ps.dedispersionFFTsize()], queue, CL_MEM_READ_WRITE, CL_MEM_READ_WRITE);
+ MultiArraySharedBuffer<float, 1> DMs(boost::extents[ps.nrTABs(0)], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
DedispersionChirpKernel dedispersionChirpKernel(ps, program, queue, data, DMs);
data[3][1][13][1][77] = std::complex<float>(2, 3);
@@ -1782,16 +1978,16 @@ struct CoherentStokesTest : public UnitTest
:
UnitTest(ps, "BeamFormer/CoherentStokes.cl")
{
- if (ps.nrChannelsPerSubband() >= 19 && ps.nrSamplesPerChannel() >= 175 && ps.nrTABs(SAP) >= 5) {
- MultiArraySharedBuffer<float, 4> stokesData(boost::extents[ps.nrTABs(SAP)][ps.nrCoherentStokes()][ps.nrSamplesPerChannel() / ps.coherentStokesTimeIntegrationFactor()][ps.nrChannelsPerSubband()], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
+ if (ps.nrChannelsPerSubband() >= 19 && ps.nrSamplesPerChannel() >= 175 && ps.nrTABs(0) >= 5) {
+ MultiArraySharedBuffer<float, 4> stokesData(boost::extents[ps.nrTABs(0)][ps.nrCoherentStokes()][ps.nrSamplesPerChannel() / ps.coherentStokesTimeIntegrationFactor()][ps.nrChannelsPerSubband()], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
#if 1
- MultiArraySharedBuffer<std::complex<float>, 4> complexVoltages(boost::extents[ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel()][ps.nrTABs(SAP)][NR_POLARIZATIONS], queue, CL_MEM_READ_WRITE, CL_MEM_READ_ONLY);
+ MultiArraySharedBuffer<std::complex<float>, 4> complexVoltages(boost::extents[ps.nrChannelsPerSubband()][ps.nrSamplesPerChannel()][ps.nrTABs(0)][NR_POLARIZATIONS], queue, CL_MEM_READ_WRITE, CL_MEM_READ_ONLY);
CoherentStokesKernel stokesKernel(ps, program, stokesData, complexVoltages);
complexVoltages[18][174][4][0] = std::complex<float>(2, 3);
complexVoltages[18][174][4][1] = std::complex<float>(4, 5);
#else
- MultiArraySharedBuffer<std::complex<float>, 4> complexVoltages(boost::extents[ps.nrTABs(SAP)][NR_POLARIZATIONS][ps.nrSamplesPerChannel()][ps.nrChannelsPerSubband()], queue, CL_MEM_READ_WRITE, CL_MEM_READ_ONLY);
+ MultiArraySharedBuffer<std::complex<float>, 4> complexVoltages(boost::extents[ps.nrTABs(0)][NR_POLARIZATIONS][ps.nrSamplesPerChannel()][ps.nrChannelsPerSubband()], queue, CL_MEM_READ_WRITE, CL_MEM_READ_ONLY);
CoherentStokesKernel stokesKernel(ps, program, stokesData, complexVoltages);
complexVoltages[18][174][4][0] = std::complex<float>(2, 3);
@@ -1815,10 +2011,10 @@ struct UHEP_BeamFormerTest : public UnitTest
:
UnitTest(ps, "UHEP/BeamFormer.cl")
{
- if (ps.nrStations() >= 5 && (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) >= 13 && ps.nrSubbands() >= 7 && ps.nrTABs(SAP) >= 6) {
+ if (ps.nrStations() >= 5 && (ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1) >= 13 && ps.nrSubbands() >= 7 && ps.nrTABs(0) >= 6) {
MultiArraySharedBuffer<char, 5> inputSamples(boost::extents[ps.nrStations()][ps.nrSubbands()][ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1][NR_POLARIZATIONS][ps.nrBytesPerComplexSample()], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY);
- MultiArraySharedBuffer<std::complex<float>, 3> beamFormerWeights(boost::extents[ps.nrStations()][ps.nrSubbands()][ps.nrTABs(SAP)], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY);
- MultiArraySharedBuffer<std::complex<float>, 4> complexVoltages(boost::extents[ps.nrSubbands()][ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1][ps.nrTABs(SAP)][NR_POLARIZATIONS], queue, CL_MEM_READ_ONLY, CL_MEM_READ_WRITE);
+ MultiArraySharedBuffer<std::complex<float>, 3> beamFormerWeights(boost::extents[ps.nrStations()][ps.nrSubbands()][ps.nrTABs(0)], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY);
+ MultiArraySharedBuffer<std::complex<float>, 4> complexVoltages(boost::extents[ps.nrSubbands()][ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1][ps.nrTABs(0)][NR_POLARIZATIONS], queue, CL_MEM_READ_ONLY, CL_MEM_READ_WRITE);
UHEP_BeamFormerKernel beamFormer(ps, program, complexVoltages, inputSamples, beamFormerWeights);
switch (ps.nrBytesPerComplexSample()) {
@@ -1851,9 +2047,9 @@ struct UHEP_TransposeTest : public UnitTest
:
UnitTest(ps, "UHEP/Transpose.cl")
{
- if (ps.nrSubbands() >= 19 && ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1 >= 175 && ps.nrTABs(SAP) >= 5) {
- MultiArraySharedBuffer<std::complex<float>, 4> transposedData(boost::extents[ps.nrTABs(SAP)][NR_POLARIZATIONS][ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1][512], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
- MultiArraySharedBuffer<std::complex<float>, 4> complexVoltages(boost::extents[ps.nrSubbands()][ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1][ps.nrTABs(SAP)][NR_POLARIZATIONS], queue, CL_MEM_READ_WRITE, CL_MEM_READ_ONLY);
+ if (ps.nrSubbands() >= 19 && ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1 >= 175 && ps.nrTABs(0) >= 5) {
+ MultiArraySharedBuffer<std::complex<float>, 4> transposedData(boost::extents[ps.nrTABs(0)][NR_POLARIZATIONS][ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1][512], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
+ MultiArraySharedBuffer<std::complex<float>, 4> complexVoltages(boost::extents[ps.nrSubbands()][ps.nrSamplesPerChannel() + NR_STATION_FILTER_TAPS - 1][ps.nrTABs(0)][NR_POLARIZATIONS], queue, CL_MEM_READ_WRITE, CL_MEM_READ_ONLY);
cl::Buffer devReverseSubbandMapping(context, CL_MEM_READ_ONLY, 512 * sizeof(int));
UHEP_TransposeKernel transpose(ps, program, transposedData, complexVoltages, devReverseSubbandMapping);
@@ -1876,9 +2072,9 @@ struct UHEP_TriggerTest : public UnitTest
:
UnitTest(ps, "UHEP/Trigger.cl")
{
- if (ps.nrTABs(SAP) >= 4 && 1024 * ps.nrSamplesPerChannel() > 100015) {
- MultiArraySharedBuffer<float, 3> inputData(boost::extents[ps.nrTABs(SAP)][NR_POLARIZATIONS][ps.nrSamplesPerChannel() * 1024], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY);
- MultiArraySharedBuffer<TriggerInfo, 1> triggerInfo(boost::extents[ps.nrTABs(SAP)], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
+ if (ps.nrTABs(0) >= 4 && 1024 * ps.nrSamplesPerChannel() > 100015) {
+ MultiArraySharedBuffer<float, 3> inputData(boost::extents[ps.nrTABs(0)][NR_POLARIZATIONS][ps.nrSamplesPerChannel() * 1024], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY);
+ MultiArraySharedBuffer<TriggerInfo, 1> triggerInfo(boost::extents[ps.nrTABs(0)], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
UHEP_TriggerKernel trigger(ps, program, triggerInfo, inputData);
inputData[3][1][100015] = 1000;
@@ -1900,7 +2096,7 @@ struct UHEP_TriggerTest : public UnitTest
struct FFT_Test : public UnitTest
{
FFT_Test(const Parset &ps)
- : UnitTest(ps, "fft2.cl")
+ : UnitTest(ps, "fft.cl")
{
MultiArraySharedBuffer<std::complex<float>, 1> in(boost::extents[8], queue, CL_MEM_WRITE_ONLY, CL_MEM_READ_ONLY);
MultiArraySharedBuffer<std::complex<float>, 1> out(boost::extents[8], queue, CL_MEM_READ_ONLY, CL_MEM_WRITE_ONLY);
@@ -1979,34 +2175,14 @@ int main(int argc, char **argv)
const char *str = getenv("NR_GPUS");
nrGPUs = str ? atoi(str) : 1;
-#if 0
- ps.nrSubbands() = 10;//488;
- ps.nrChannelsPerSubband() = 64;
- ps.nrBeams() = 1;
- ps.nrSamplesPerChannel() = 196608 / ps.nrChannelsPerSubband();
- ps.subbandBandwidth() = 195312.5;
- ps.correctBandPass() = true;
-#endif
+ profiling = false; CorrelatorPipeline(ps).doWork();
+ profiling = true; CorrelatorPipeline(ps).doWork();
- //profiling = false; CorrelatorPipeline(ps).doWork();
- //profiling = true; CorrelatorPipeline(ps).doWork();
-
- (CorrelatorTest)(ps);
+ //(CorrelatorTest)(ps);
+ //(CorrelateRectangleTest)(ps);
+ //(CorrelateTriangleTest)(ps);
#if 0
- ps.nrSubbands() = 488;
- ps.nrChannelsPerSubband() = 2048;
- ps.nrBeams() = 1;
- ps.nrTABs(SAP) = 128;
- ps.nrIncoherentStokes() = 4;
- ps.nrCoherentStokes() = 4;
- ps.incoherentStokesTimeIntegrationFactor() = 8;
- ps.coherentStokesTimeIntegrationFactor() = 8;
- ps.nrSamplesPerChannel() = 65536 / ps.nrChannelsPerSubband();//262144 / ps.nrChannelsPerSubband();
- ps.subbandBandwidth() = 195312.5;
- ps.correctBandPass() = true;
- ps.dedispersionFFTsize() = ps.nrSamplesPerChannel();
-
profiling = false; BeamFormerPipeline(ps).doWork();
profiling = true; BeamFormerPipeline(ps).doWork();
//(IncoherentStokesTest)(ps);
@@ -2018,12 +2194,6 @@ int main(int argc, char **argv)
#endif
#if 0
- ps.nrSubbands() = 488;
- ps.nrSamplesPerChannel() = 1024;
- ps.nrBeams() = 1;
- ps.subbandBandwidth() = 195312.5;
- ps.nrTABs(SAP) = 48;
-
profiling = false; UHEP_Pipeline(ps).doWork();
profiling = true; UHEP_Pipeline(ps).doWork();
//(UHEP_BeamFormerTest)(ps);
diff --git a/RTCP/GPUProc/src/UHEP/BeamFormer.cl-0.ptx b/RTCP/GPUProc/src/UHEP/BeamFormer.cl-0.ptx
deleted file mode 100644
index ea220810bb76a4961820e5b0583dc8ffe90325aa..0000000000000000000000000000000000000000
--- a/RTCP/GPUProc/src/UHEP/BeamFormer.cl-0.ptx
+++ /dev/null
@@ -1,207 +0,0 @@
-//
-// Generated by NVIDIA NVVM Compiler
-// Compiler built on Tue Feb 7 07:15:59 2012 (1328595359)
-// Driver 295.20
-//
-
-.version 3.0
-.target sm_21, texmode_independent
-.address_size 32
-
-.extern .shared .align 16 .b8 shr_2__local[2048];
-
-.entry complexVoltages(
- .param .u32 .ptr .global .align 1 complexVoltages_param_0,
- .param .u32 .ptr .global .align 1 complexVoltages_param_1,
- .param .u32 .ptr .global .align 1 complexVoltages_param_2
-)
-{
- .reg .f32 %f<173>;
- .reg .pred %p<7>;
- .reg .s32 %r<76>;
- .reg .s16 %rc<16>;
-
-
- ld.param.u32 %r29, [complexVoltages_param_2];
- // inline asm
- mov.u32 %r22, %tid.x;
- // inline asm
- // inline asm
- mov.u32 %r23, %tid.y;
- // inline asm
- // inline asm
- mov.u32 %r24, %envreg5;
- // inline asm
- // inline asm
- mov.u32 %r25, %ntid.z;
- // inline asm
- // inline asm
- mov.u32 %r26, %ctaid.z;
- // inline asm
- // inline asm
- mov.u32 %r27, %tid.z;
- // inline asm
- add.s32 %r30, %r27, %r24;
- mad.lo.s32 %r4, %r26, %r25, %r30;
- shl.b32 %r31, %r23, 4;
- mad.lo.s32 %r32, %r4, 66496, %r31;
- shl.b32 %r33, %r22, 3;
- add.s32 %r5, %r32, %r33;
- shl.b32 %r34, %r4, 5;
- shl.b32 %r35, %r23, 3;
- add.s32 %r36, %r34, %r35;
- add.s32 %r37, %r29, %r36;
- ld.global.v2.f32 {%f131, %f132}, [%r37];
- ld.global.v2.f32 {%f137, %f138}, [%r37+15616];
- ld.global.v2.f32 {%f143, %f144}, [%r37+31232];
- ld.global.v2.f32 {%f149, %f150}, [%r37+46848];
- ld.global.v2.f32 {%f155, %f156}, [%r37+62464];
- ld.global.v2.f32 {%f161, %f162}, [%r37+78080];
- ld.global.v2.f32 {%f167, %f168}, [%r37+93696];
- mov.u32 %r72, 0;
-
-BB0_1:
- shl.b32 %r40, %r72, 10;
- add.s32 %r7, %r5, %r40;
- // inline asm
- mov.u32 %r38, %tid.x;
- // inline asm
- // inline asm
- mov.u32 %r39, %tid.y;
- // inline asm
- shl.b32 %r9, %r39, 1;
- add.s32 %r74, %r9, %r38;
- setp.gt.u32 %p1, %r74, 127;
- @%p1 bra BB0_6;
-
- add.s32 %r11, %r38, %r9;
- cvt.u8.u32 %rc1, %r11;
- mov.u16 %rc15, 0;
- mov.u32 %r73, 0;
-
-BB0_3:
- add.s32 %r14, %r11, %r73;
- shl.b32 %r42, %r72, 4;
- shl.b16 %rc5, %rc15, 3;
- add.s16 %rc6, %rc1, %rc5;
- cvt.u32.u8 %r43, %rc6;
- and.b32 %r44, %r43, 15;
- add.s32 %r15, %r42, %r44;
- setp.gt.u32 %p2, %r15, 1038;
- @%p2 bra BB0_5;
-
- shr.u32 %r45, %r14, 4;
- ld.param.u32 %r71, [complexVoltages_param_1];
- mad.lo.s32 %r46, %r45, 2028128, %r71;
- mad.lo.s32 %r47, %r4, 4156, %r46;
- shl.b32 %r48, %r15, 2;
- add.s32 %r49, %r47, %r48;
- ld.global.v4.u8 {%rc11, %rc12, %rc13, %rc14}, [%r49];
- // inline asm
- cvt.rn.f32.s8 %f1, %rc11;
- // inline asm
- // inline asm
- cvt.rn.f32.s8 %f2, %rc12;
- // inline asm
- // inline asm
- cvt.rn.f32.s8 %f3, %rc13;
- // inline asm
- // inline asm
- cvt.rn.f32.s8 %f4, %rc14;
- // inline asm
- shl.b32 %r50, %r74, 4;
- mov.u32 %r51, shr_2__local;
- add.s32 %r52, %r51, %r50;
- st.shared.v4.f32 [%r52], {%f1, %f2, %f3, %f4};
-
-BB0_5:
- add.s32 %r74, %r74, 8;
- setp.lt.u32 %p3, %r74, 128;
- add.s32 %r73, %r73, 8;
- add.s16 %rc15, %rc15, 1;
- @%p3 bra BB0_3;
-
-BB0_6:
- bar.sync 0;
- mov.u32 %r56, 1039;
- shl.b32 %r57, %r72, 4;
- sub.s32 %r18, %r56, %r57;
- mov.u32 %r54, 16;
- // inline asm
- min.u32 %r53, %r54, %r18;
- // inline asm
- setp.eq.s32 %p4, %r53, 0;
- @%p4 bra BB0_9;
-
- mov.u32 %r75, 0;
-
-BB0_8:
- shl.b32 %r62, %r75, 6;
- add.s32 %r63, %r7, %r62;
- ld.param.u32 %r70, [complexVoltages_param_0];
- add.s32 %r64, %r70, %r63;
- shl.b32 %r65, %r75, 4;
- mov.u32 %r66, shr_2__local;
- add.s32 %r67, %r66, %r65;
- add.s32 %r69, %r67, %r33;
- ld.shared.v2.f32 {%f27, %f28}, [%r69];
- mov.f32 %f5, 0f00000000;
- fma.rn.ftz.f32 %f31, %f131, %f27, %f5;
- fma.rn.ftz.f32 %f32, %f131, %f28, %f5;
- neg.ftz.f32 %f7, %f28;
- fma.rn.ftz.f32 %f39, %f132, %f7, %f31;
- fma.rn.ftz.f32 %f40, %f132, %f27, %f32;
- ld.shared.v2.f32 {%f43, %f44}, [%r69+256];
- fma.rn.ftz.f32 %f45, %f137, %f43, %f39;
- fma.rn.ftz.f32 %f46, %f137, %f44, %f40;
- neg.ftz.f32 %f10, %f44;
- fma.rn.ftz.f32 %f53, %f138, %f10, %f45;
- fma.rn.ftz.f32 %f54, %f138, %f43, %f46;
- ld.shared.v2.f32 {%f57, %f58}, [%r69+512];
- fma.rn.ftz.f32 %f59, %f143, %f57, %f53;
- fma.rn.ftz.f32 %f60, %f143, %f58, %f54;
- neg.ftz.f32 %f13, %f58;
- fma.rn.ftz.f32 %f67, %f144, %f13, %f59;
- fma.rn.ftz.f32 %f68, %f144, %f57, %f60;
- ld.shared.v2.f32 {%f71, %f72}, [%r69+768];
- fma.rn.ftz.f32 %f73, %f149, %f71, %f67;
- fma.rn.ftz.f32 %f74, %f149, %f72, %f68;
- neg.ftz.f32 %f16, %f72;
- fma.rn.ftz.f32 %f81, %f150, %f16, %f73;
- fma.rn.ftz.f32 %f82, %f150, %f71, %f74;
- ld.shared.v2.f32 {%f85, %f86}, [%r69+1024];
- fma.rn.ftz.f32 %f87, %f155, %f85, %f81;
- fma.rn.ftz.f32 %f88, %f155, %f86, %f82;
- neg.ftz.f32 %f19, %f86;
- fma.rn.ftz.f32 %f95, %f156, %f19, %f87;
- fma.rn.ftz.f32 %f96, %f156, %f85, %f88;
- ld.shared.v2.f32 {%f99, %f100}, [%r69+1280];
- fma.rn.ftz.f32 %f101, %f161, %f99, %f95;
- fma.rn.ftz.f32 %f102, %f161, %f100, %f96;
- neg.ftz.f32 %f22, %f100;
- fma.rn.ftz.f32 %f109, %f162, %f22, %f101;
- fma.rn.ftz.f32 %f110, %f162, %f99, %f102;
- ld.shared.v2.f32 {%f113, %f114}, [%r69+1536];
- fma.rn.ftz.f32 %f115, %f167, %f113, %f109;
- fma.rn.ftz.f32 %f116, %f167, %f114, %f110;
- neg.ftz.f32 %f25, %f114;
- fma.rn.ftz.f32 %f123, %f168, %f25, %f115;
- fma.rn.ftz.f32 %f124, %f168, %f113, %f116;
- st.global.v2.f32 [%r64], {%f123, %f124};
- // inline asm
- min.u32 %r59, %r54, %r18;
- // inline asm
- add.s32 %r75, %r75, 1;
- setp.lt.u32 %p5, %r75, %r59;
- @%p5 bra BB0_8;
-
-BB0_9:
- bar.sync 0;
- add.s32 %r72, %r72, 1;
- setp.ne.s32 %p6, %r72, 65;
- @%p6 bra BB0_1;
-
- ret;
-}
-
-
diff --git a/RTCP/GPUProc/src/UHEP/BeamFormer.cl.hop-0.ptx b/RTCP/GPUProc/src/UHEP/BeamFormer.cl.hop-0.ptx
deleted file mode 100644
index 79498463a835875a85828f52fd7f756eb2b37aab..0000000000000000000000000000000000000000
Binary files a/RTCP/GPUProc/src/UHEP/BeamFormer.cl.hop-0.ptx and /dev/null differ
diff --git a/RTCP/GPUProc/src/UHEP/InvFFT.cl-0.ptx b/RTCP/GPUProc/src/UHEP/InvFFT.cl-0.ptx
deleted file mode 100644
index 86d5c2b091806e681f5dfe3df13f2ce5d1d68535..0000000000000000000000000000000000000000
--- a/RTCP/GPUProc/src/UHEP/InvFFT.cl-0.ptx
+++ /dev/null
@@ -1,759 +0,0 @@
-//
-// Generated by NVIDIA NVVM Compiler
-// Compiler built on Tue Feb 7 07:15:59 2012 (1328595359)
-// Driver 295.20
-//
-
-.version 3.0
-.target sm_21, texmode_independent
-.address_size 32
-
-.extern .shared .align 32 .b8 shr_3_lds[4096];
-
-.entry inv_fft(
- .param .u32 .ptr .global .align 8 inv_fft_param_0,
- .param .u32 .ptr .global .align 4 inv_fft_param_1
-)
-.reqntid 128, 1, 1
-{
- .reg .f32 %f<576>;
- .reg .pred %p<2>;
- .reg .s32 %r<212>;
-
-
- ld.param.u32 %r26, [inv_fft_param_1];
- // inline asm
- mov.u32 %r2, %tid.x;
- // inline asm
- // inline asm
- mov.u32 %r3, %envreg4;
- // inline asm
- // inline asm
- mov.u32 %r4, %ntid.y;
- // inline asm
- // inline asm
- mov.u32 %r5, %ctaid.y;
- // inline asm
- // inline asm
- mov.u32 %r6, %tid.y;
- // inline asm
- add.s32 %r27, %r6, %r3;
- mad.lo.s32 %r28, %r5, %r4, %r27;
- shl.b32 %r29, %r28, 12;
- add.s32 %r30, %r26, %r29;
- shl.b32 %r31, %r2, 3;
- add.s32 %r32, %r30, %r31;
- ld.global.v2.f32 {%f574, %f575}, [%r32];
- // inline asm
- mov.u32 %r7, %tid.x;
- // inline asm
- // inline asm
- mov.u32 %r8, %envreg4;
- // inline asm
- // inline asm
- mov.u32 %r9, %ntid.y;
- // inline asm
- // inline asm
- mov.u32 %r10, %ctaid.y;
- // inline asm
- // inline asm
- mov.u32 %r11, %tid.y;
- // inline asm
- add.s32 %r33, %r11, %r8;
- mad.lo.s32 %r34, %r10, %r9, %r33;
- shl.b32 %r35, %r34, 12;
- add.s32 %r36, %r26, %r35;
- shl.b32 %r37, %r7, 3;
- add.s32 %r38, %r37, %r36;
- ld.global.v2.f32 {%f516, %f517}, [%r38+1024];
- // inline asm
- mov.u32 %r12, %tid.x;
- // inline asm
- // inline asm
- mov.u32 %r13, %envreg4;
- // inline asm
- // inline asm
- mov.u32 %r14, %ntid.y;
- // inline asm
- // inline asm
- mov.u32 %r15, %ctaid.y;
- // inline asm
- // inline asm
- mov.u32 %r16, %tid.y;
- // inline asm
- add.s32 %r39, %r16, %r13;
- mad.lo.s32 %r40, %r15, %r14, %r39;
- shl.b32 %r41, %r40, 12;
- add.s32 %r42, %r26, %r41;
- shl.b32 %r43, %r12, 3;
- add.s32 %r44, %r43, %r42;
- ld.global.v2.f32 {%f510, %f511}, [%r44+2048];
- // inline asm
- mov.u32 %r17, %tid.x;
- // inline asm
- // inline asm
- mov.u32 %r18, %envreg4;
- // inline asm
- // inline asm
- mov.u32 %r19, %ntid.y;
- // inline asm
- // inline asm
- mov.u32 %r20, %ctaid.y;
- // inline asm
- // inline asm
- mov.u32 %r21, %tid.y;
- // inline asm
- add.s32 %r45, %r21, %r18;
- mad.lo.s32 %r46, %r20, %r19, %r45;
- shl.b32 %r47, %r46, 12;
- add.s32 %r48, %r26, %r47;
- shl.b32 %r49, %r17, 3;
- add.s32 %r50, %r49, %r48;
- ld.global.v2.f32 {%f522, %f523}, [%r50+3072];
- // inline asm
- mov.u32 %r22, %tid.x;
- // inline asm
- shl.b32 %r51, %r22, 3;
- mov.u32 %r52, shr_3_lds;
- add.s32 %r53, %r52, %r51;
- st.shared.v2.f32 [%r53], {%f574, %f575};
- // inline asm
- mov.u32 %r23, %tid.x;
- // inline asm
- shl.b32 %r54, %r23, 3;
- add.s32 %r55, %r54, %r52;
- st.shared.v2.f32 [%r55+1024], {%f516, %f517};
- // inline asm
- mov.u32 %r24, %tid.x;
- // inline asm
- shl.b32 %r56, %r24, 3;
- add.s32 %r57, %r56, %r52;
- st.shared.v2.f32 [%r57+2048], {%f510, %f511};
- // inline asm
- mov.u32 %r25, %tid.x;
- // inline asm
- shl.b32 %r58, %r25, 3;
- add.s32 %r59, %r58, %r52;
- st.shared.v2.f32 [%r59+3072], {%f522, %f523};
- bar.sync 0;
- // inline asm
- mov.u32 %r60, %tid.x;
- // inline asm
- setp.eq.s32 %p1, %r60, 0;
- @%p1 bra BB0_2;
-
- // inline asm
- mov.u32 %r61, %tid.x;
- // inline asm
- mov.u32 %r62, 512;
- sub.s32 %r63, %r62, %r61;
- shl.b32 %r64, %r63, 3;
- add.s32 %r66, %r52, %r64;
- ld.shared.v2.f32 {%f572, %f573}, [%r66];
- bra.uni BB0_3;
-
-BB0_2:
- mov.f32 %f17, 0f00000000;
- mov.f32 %f572, %f575;
- mov.f32 %f573, %f17;
- mov.f32 %f574, %f574;
- mov.f32 %f575, %f17;
-
-BB0_3:
- // inline asm
- mov.u32 %r67, %tid.x;
- // inline asm
- mov.u32 %r70, 384;
- sub.s32 %r71, %r70, %r67;
- shl.b32 %r72, %r71, 3;
- add.s32 %r74, %r52, %r72;
- ld.shared.v2.f32 {%f490, %f491}, [%r74];
- // inline asm
- mov.u32 %r68, %tid.x;
- // inline asm
- mov.u32 %r75, 256;
- sub.s32 %r76, %r75, %r68;
- shl.b32 %r77, %r76, 3;
- add.s32 %r78, %r52, %r77;
- ld.shared.v2.f32 {%f492, %f493}, [%r78];
- // inline asm
- mov.u32 %r69, %tid.x;
- // inline asm
- mov.u32 %r79, 128;
- sub.s32 %r80, %r79, %r69;
- shl.b32 %r81, %r80, 3;
- add.s32 %r82, %r52, %r81;
- neg.ftz.f32 %f20, %f573;
- neg.ftz.f32 %f22, %f491;
- neg.ftz.f32 %f24, %f493;
- ld.shared.v2.f32 {%f500, %f501}, [%r82];
- neg.ftz.f32 %f26, %f501;
- add.ftz.f32 %f504, %f574, %f572;
- add.ftz.f32 %f505, %f575, %f20;
- sub.ftz.f32 %f506, %f574, %f572;
- sub.ftz.f32 %f507, %f575, %f20;
- add.ftz.f32 %f508, %f510, %f492;
- add.ftz.f32 %f509, %f511, %f24;
- sub.ftz.f32 %f512, %f510, %f492;
- sub.ftz.f32 %f513, %f511, %f24;
- add.ftz.f32 %f514, %f516, %f490;
- add.ftz.f32 %f515, %f517, %f22;
- sub.ftz.f32 %f518, %f516, %f490;
- sub.ftz.f32 %f519, %f517, %f22;
- add.ftz.f32 %f520, %f522, %f500;
- add.ftz.f32 %f521, %f523, %f26;
- sub.ftz.f32 %f524, %f522, %f500;
- sub.ftz.f32 %f525, %f523, %f26;
- add.ftz.f32 %f526, %f504, %f508;
- add.ftz.f32 %f527, %f505, %f509;
- sub.ftz.f32 %f528, %f504, %f508;
- sub.ftz.f32 %f529, %f505, %f509;
- neg.ftz.f32 %f28, %f513;
- add.ftz.f32 %f534, %f506, %f28;
- add.ftz.f32 %f535, %f507, %f512;
- sub.ftz.f32 %f536, %f506, %f28;
- sub.ftz.f32 %f537, %f507, %f512;
- add.ftz.f32 %f538, %f514, %f520;
- add.ftz.f32 %f539, %f515, %f521;
- sub.ftz.f32 %f540, %f514, %f520;
- sub.ftz.f32 %f541, %f515, %f521;
- neg.ftz.f32 %f31, %f525;
- add.ftz.f32 %f546, %f518, %f31;
- add.ftz.f32 %f547, %f519, %f524;
- sub.ftz.f32 %f548, %f518, %f31;
- sub.ftz.f32 %f549, %f519, %f524;
- add.ftz.f32 %f446, %f526, %f538;
- add.ftz.f32 %f447, %f527, %f539;
- sub.ftz.f32 %f454, %f526, %f538;
- sub.ftz.f32 %f455, %f527, %f539;
- neg.ftz.f32 %f34, %f547;
- add.ftz.f32 %f554, %f34, %f546;
- add.ftz.f32 %f555, %f546, %f547;
- mov.f32 %f36, 0f3F3504F3;
- fma.rn.ftz.f32 %f448, %f554, %f36, %f534;
- fma.rn.ftz.f32 %f449, %f555, %f36, %f535;
- neg.f32 %f558, %f554;
- neg.f32 %f559, %f555;
- fma.rn.ftz.f32 %f456, %f558, %f36, %f534;
- fma.rn.ftz.f32 %f457, %f559, %f36, %f535;
- neg.ftz.f32 %f38, %f541;
- add.ftz.f32 %f450, %f528, %f38;
- add.ftz.f32 %f451, %f529, %f540;
- sub.ftz.f32 %f458, %f528, %f38;
- sub.ftz.f32 %f459, %f529, %f540;
- neg.ftz.f32 %f41, %f549;
- sub.ftz.f32 %f568, %f41, %f548;
- sub.ftz.f32 %f569, %f548, %f549;
- fma.rn.ftz.f32 %f452, %f568, %f36, %f536;
- fma.rn.ftz.f32 %f453, %f569, %f36, %f537;
- neg.f32 %f570, %f568;
- neg.f32 %f571, %f569;
- fma.rn.ftz.f32 %f460, %f570, %f36, %f536;
- fma.rn.ftz.f32 %f461, %f571, %f36, %f537;
- bar.sync 0;
- // inline asm
- mov.u32 %r83, %tid.x;
- // inline asm
- shl.b32 %r84, %r83, 5;
- add.s32 %r86, %r52, %r84;
- st.shared.v4.f32 [%r86+16], {%f454, %f456, %f458, %f460};
- st.shared.v4.f32 [%r86], {%f446, %f448, %f450, %f452};
- bar.sync 0;
- // inline asm
- mov.u32 %r87, %tid.x;
- // inline asm
- shl.b32 %r95, %r87, 2;
- add.s32 %r97, %r52, %r95;
- ld.shared.f32 %f51, [%r97];
- // inline asm
- mov.u32 %r88, %tid.x;
- // inline asm
- shl.b32 %r98, %r88, 2;
- add.s32 %r99, %r98, %r52;
- ld.shared.f32 %f1, [%r99+512];
- // inline asm
- mov.u32 %r89, %tid.x;
- // inline asm
- shl.b32 %r100, %r89, 2;
- add.s32 %r101, %r100, %r52;
- ld.shared.f32 %f2, [%r101+1024];
- // inline asm
- mov.u32 %r90, %tid.x;
- // inline asm
- shl.b32 %r102, %r90, 2;
- add.s32 %r103, %r102, %r52;
- ld.shared.f32 %f3, [%r103+1536];
- // inline asm
- mov.u32 %r91, %tid.x;
- // inline asm
- shl.b32 %r104, %r91, 2;
- add.s32 %r105, %r104, %r52;
- ld.shared.f32 %f4, [%r105+2048];
- // inline asm
- mov.u32 %r92, %tid.x;
- // inline asm
- shl.b32 %r106, %r92, 2;
- add.s32 %r107, %r106, %r52;
- ld.shared.f32 %f5, [%r107+2560];
- // inline asm
- mov.u32 %r93, %tid.x;
- // inline asm
- shl.b32 %r108, %r93, 2;
- add.s32 %r109, %r108, %r52;
- ld.shared.f32 %f6, [%r109+3072];
- // inline asm
- mov.u32 %r94, %tid.x;
- // inline asm
- shl.b32 %r110, %r94, 2;
- add.s32 %r111, %r110, %r52;
- ld.shared.f32 %f7, [%r111+3584];
- bar.sync 0;
- // inline asm
- mov.u32 %r112, %tid.x;
- // inline asm
- shl.b32 %r113, %r112, 5;
- add.s32 %r115, %r52, %r113;
- st.shared.v4.f32 [%r115+16], {%f455, %f457, %f459, %f461};
- st.shared.v4.f32 [%r115], {%f447, %f449, %f451, %f453};
- bar.sync 0;
- // inline asm
- mov.u32 %r116, %tid.x;
- // inline asm
- shl.b32 %r124, %r116, 2;
- add.s32 %r126, %r52, %r124;
- ld.shared.f32 %f61, [%r126];
- // inline asm
- mov.u32 %r117, %tid.x;
- // inline asm
- shl.b32 %r127, %r117, 2;
- add.s32 %r128, %r127, %r52;
- ld.shared.f32 %f8, [%r128+512];
- // inline asm
- mov.u32 %r118, %tid.x;
- // inline asm
- shl.b32 %r129, %r118, 2;
- add.s32 %r130, %r129, %r52;
- ld.shared.f32 %f9, [%r130+1024];
- // inline asm
- mov.u32 %r119, %tid.x;
- // inline asm
- shl.b32 %r131, %r119, 2;
- add.s32 %r132, %r131, %r52;
- ld.shared.f32 %f10, [%r132+1536];
- // inline asm
- mov.u32 %r120, %tid.x;
- // inline asm
- shl.b32 %r133, %r120, 2;
- add.s32 %r134, %r133, %r52;
- ld.shared.f32 %f11, [%r134+2048];
- // inline asm
- mov.u32 %r121, %tid.x;
- // inline asm
- shl.b32 %r135, %r121, 2;
- add.s32 %r136, %r135, %r52;
- ld.shared.f32 %f12, [%r136+2560];
- // inline asm
- mov.u32 %r122, %tid.x;
- // inline asm
- shl.b32 %r137, %r122, 2;
- add.s32 %r138, %r137, %r52;
- ld.shared.f32 %f13, [%r138+3072];
- // inline asm
- mov.u32 %r123, %tid.x;
- // inline asm
- shl.b32 %r139, %r123, 2;
- add.s32 %r140, %r139, %r52;
- ld.shared.f32 %f14, [%r140+3584];
- bar.sync 0;
- mov.f32 %f90, 0f40000000;
- mul.ftz.f32 %f15, %f90, 0f40490FDB;
- div.rn.ftz.f32 %f91, %f15, 0f42800000;
- // inline asm
- mov.u32 %r141, %tid.x;
- // inline asm
- and.b32 %r144, %r141, 7;
- cvt.rn.f32.u32 %f92, %r144;
- mul.ftz.f32 %f65, %f91, %f92;
- // inline asm
- cos.approx.f32 %f62, %f65;
- // inline asm
- // inline asm
- sin.approx.f32 %f64, %f65;
- // inline asm
- mul.ftz.f32 %f93, %f62, %f1;
- neg.f32 %f94, %f64;
- fma.rn.ftz.f32 %f95, %f94, %f8, %f93;
- mul.ftz.f32 %f96, %f64, %f1;
- fma.rn.ftz.f32 %f97, %f62, %f8, %f96;
- fma.rn.ftz.f32 %f67, %f91, %f92, %f65;
- // inline asm
- cos.approx.f32 %f66, %f67;
- // inline asm
- // inline asm
- sin.approx.f32 %f68, %f67;
- // inline asm
- mul.ftz.f32 %f98, %f66, %f2;
- neg.f32 %f99, %f68;
- fma.rn.ftz.f32 %f100, %f99, %f9, %f98;
- mul.ftz.f32 %f101, %f68, %f2;
- fma.rn.ftz.f32 %f102, %f66, %f9, %f101;
- mul.ftz.f32 %f73, %f65, 0f40400000;
- // inline asm
- cos.approx.f32 %f70, %f73;
- // inline asm
- // inline asm
- sin.approx.f32 %f72, %f73;
- // inline asm
- mul.ftz.f32 %f103, %f70, %f3;
- neg.f32 %f104, %f72;
- fma.rn.ftz.f32 %f105, %f104, %f10, %f103;
- mul.ftz.f32 %f106, %f72, %f3;
- fma.rn.ftz.f32 %f107, %f70, %f10, %f106;
- mul.ftz.f32 %f77, %f65, 0f40800000;
- // inline asm
- cos.approx.f32 %f74, %f77;
- // inline asm
- // inline asm
- sin.approx.f32 %f76, %f77;
- // inline asm
- mul.ftz.f32 %f108, %f74, %f4;
- neg.f32 %f109, %f76;
- fma.rn.ftz.f32 %f110, %f109, %f11, %f108;
- mul.ftz.f32 %f111, %f76, %f4;
- fma.rn.ftz.f32 %f112, %f74, %f11, %f111;
- mul.ftz.f32 %f81, %f65, 0f40A00000;
- // inline asm
- cos.approx.f32 %f78, %f81;
- // inline asm
- // inline asm
- sin.approx.f32 %f80, %f81;
- // inline asm
- mul.ftz.f32 %f113, %f78, %f5;
- neg.f32 %f114, %f80;
- fma.rn.ftz.f32 %f115, %f114, %f12, %f113;
- mul.ftz.f32 %f116, %f80, %f5;
- fma.rn.ftz.f32 %f117, %f78, %f12, %f116;
- mul.ftz.f32 %f85, %f65, 0f40C00000;
- // inline asm
- cos.approx.f32 %f82, %f85;
- // inline asm
- // inline asm
- sin.approx.f32 %f84, %f85;
- // inline asm
- mul.ftz.f32 %f118, %f82, %f6;
- neg.f32 %f119, %f84;
- fma.rn.ftz.f32 %f120, %f119, %f13, %f118;
- mul.ftz.f32 %f121, %f84, %f6;
- fma.rn.ftz.f32 %f122, %f82, %f13, %f121;
- mul.ftz.f32 %f89, %f65, 0f40E00000;
- // inline asm
- cos.approx.f32 %f86, %f89;
- // inline asm
- // inline asm
- sin.approx.f32 %f88, %f89;
- // inline asm
- mul.ftz.f32 %f123, %f86, %f7;
- neg.f32 %f124, %f88;
- fma.rn.ftz.f32 %f125, %f124, %f14, %f123;
- mul.ftz.f32 %f126, %f88, %f7;
- fma.rn.ftz.f32 %f127, %f86, %f14, %f126;
- add.ftz.f32 %f380, %f51, %f110;
- add.ftz.f32 %f381, %f61, %f112;
- sub.ftz.f32 %f384, %f51, %f110;
- sub.ftz.f32 %f385, %f61, %f112;
- add.ftz.f32 %f386, %f100, %f120;
- add.ftz.f32 %f387, %f102, %f122;
- sub.ftz.f32 %f388, %f100, %f120;
- sub.ftz.f32 %f389, %f102, %f122;
- add.ftz.f32 %f390, %f95, %f115;
- add.ftz.f32 %f391, %f97, %f117;
- sub.ftz.f32 %f392, %f95, %f115;
- sub.ftz.f32 %f393, %f97, %f117;
- add.ftz.f32 %f394, %f105, %f125;
- add.ftz.f32 %f395, %f107, %f127;
- sub.ftz.f32 %f396, %f105, %f125;
- sub.ftz.f32 %f397, %f107, %f127;
- add.ftz.f32 %f398, %f380, %f386;
- add.ftz.f32 %f399, %f381, %f387;
- sub.ftz.f32 %f400, %f380, %f386;
- sub.ftz.f32 %f401, %f381, %f387;
- neg.ftz.f32 %f129, %f389;
- add.ftz.f32 %f406, %f384, %f129;
- add.ftz.f32 %f407, %f385, %f388;
- sub.ftz.f32 %f408, %f384, %f129;
- sub.ftz.f32 %f409, %f385, %f388;
- add.ftz.f32 %f410, %f390, %f394;
- add.ftz.f32 %f411, %f391, %f395;
- sub.ftz.f32 %f412, %f390, %f394;
- sub.ftz.f32 %f413, %f391, %f395;
- neg.ftz.f32 %f132, %f397;
- add.ftz.f32 %f418, %f392, %f132;
- add.ftz.f32 %f419, %f393, %f396;
- sub.ftz.f32 %f420, %f392, %f132;
- sub.ftz.f32 %f421, %f393, %f396;
- add.ftz.f32 %f348, %f398, %f410;
- add.ftz.f32 %f349, %f399, %f411;
- sub.ftz.f32 %f356, %f398, %f410;
- sub.ftz.f32 %f357, %f399, %f411;
- neg.ftz.f32 %f135, %f419;
- add.ftz.f32 %f426, %f135, %f418;
- add.ftz.f32 %f427, %f418, %f419;
- fma.rn.ftz.f32 %f350, %f426, %f36, %f406;
- fma.rn.ftz.f32 %f351, %f427, %f36, %f407;
- neg.f32 %f430, %f426;
- neg.f32 %f431, %f427;
- fma.rn.ftz.f32 %f358, %f430, %f36, %f406;
- fma.rn.ftz.f32 %f359, %f431, %f36, %f407;
- neg.ftz.f32 %f139, %f413;
- add.ftz.f32 %f352, %f400, %f139;
- add.ftz.f32 %f353, %f401, %f412;
- sub.ftz.f32 %f360, %f400, %f139;
- sub.ftz.f32 %f361, %f401, %f412;
- neg.ftz.f32 %f142, %f421;
- sub.ftz.f32 %f440, %f142, %f420;
- sub.ftz.f32 %f441, %f420, %f421;
- fma.rn.ftz.f32 %f354, %f440, %f36, %f408;
- fma.rn.ftz.f32 %f355, %f441, %f36, %f409;
- neg.f32 %f442, %f440;
- neg.f32 %f443, %f441;
- fma.rn.ftz.f32 %f362, %f442, %f36, %f408;
- fma.rn.ftz.f32 %f363, %f443, %f36, %f409;
- // inline asm
- mov.u32 %r142, %tid.x;
- // inline asm
- shl.b32 %r145, %r142, 3;
- // inline asm
- mov.u32 %r143, %tid.x;
- // inline asm
- and.b32 %r146, %r143, 7;
- and.b32 %r147, %r145, 1073741760;
- add.s32 %r148, %r147, %r146;
- shl.b32 %r149, %r148, 2;
- add.s32 %r1, %r52, %r149;
- st.shared.f32 [%r1], %f348;
- st.shared.f32 [%r1+32], %f350;
- st.shared.f32 [%r1+64], %f352;
- st.shared.f32 [%r1+96], %f354;
- st.shared.f32 [%r1+128], %f356;
- st.shared.f32 [%r1+160], %f358;
- st.shared.f32 [%r1+192], %f360;
- st.shared.f32 [%r1+224], %f362;
- bar.sync 0;
- // inline asm
- mov.u32 %r151, %tid.x;
- // inline asm
- shl.b32 %r155, %r151, 3;
- add.s32 %r157, %r52, %r155;
- ld.shared.v2.f32 {%f344, %f345}, [%r157];
- // inline asm
- mov.u32 %r152, %tid.x;
- // inline asm
- shl.b32 %r158, %r152, 3;
- add.s32 %r159, %r158, %r52;
- ld.shared.v2.f32 {%f282, %f283}, [%r159+1024];
- // inline asm
- mov.u32 %r153, %tid.x;
- // inline asm
- shl.b32 %r160, %r153, 3;
- add.s32 %r161, %r160, %r52;
- ld.shared.v2.f32 {%f288, %f289}, [%r161+2048];
- // inline asm
- mov.u32 %r154, %tid.x;
- // inline asm
- shl.b32 %r162, %r154, 3;
- add.s32 %r163, %r162, %r52;
- ld.shared.v2.f32 {%f294, %f295}, [%r163+3072];
- bar.sync 0;
- st.shared.f32 [%r1], %f349;
- st.shared.f32 [%r1+32], %f351;
- st.shared.f32 [%r1+64], %f353;
- st.shared.f32 [%r1+96], %f355;
- st.shared.f32 [%r1+128], %f357;
- st.shared.f32 [%r1+160], %f359;
- st.shared.f32 [%r1+192], %f361;
- st.shared.f32 [%r1+224], %f363;
- bar.sync 0;
- // inline asm
- mov.u32 %r164, %tid.x;
- // inline asm
- shl.b32 %r168, %r164, 3;
- add.s32 %r170, %r52, %r168;
- ld.shared.v2.f32 {%f342, %f343}, [%r170];
- // inline asm
- mov.u32 %r165, %tid.x;
- // inline asm
- shl.b32 %r171, %r165, 3;
- add.s32 %r172, %r171, %r52;
- ld.shared.v2.f32 {%f284, %f285}, [%r172+1024];
- // inline asm
- mov.u32 %r166, %tid.x;
- // inline asm
- shl.b32 %r173, %r166, 3;
- add.s32 %r174, %r173, %r52;
- ld.shared.v2.f32 {%f290, %f291}, [%r174+2048];
- // inline asm
- mov.u32 %r167, %tid.x;
- // inline asm
- shl.b32 %r175, %r167, 3;
- add.s32 %r176, %r175, %r52;
- ld.shared.v2.f32 {%f296, %f297}, [%r176+3072];
- bar.sync 0;
- div.rn.ftz.f32 %f184, %f15, 0f43800000;
- // inline asm
- mov.u32 %r177, %tid.x;
- // inline asm
- shl.b32 %r180, %r177, 1;
- and.b32 %r181, %r180, 62;
- cvt.rn.f32.u32 %f185, %r181;
- mul.ftz.f32 %f163, %f184, %f185;
- // inline asm
- cos.approx.f32 %f160, %f163;
- // inline asm
- // inline asm
- sin.approx.f32 %f162, %f163;
- // inline asm
- mul.ftz.f32 %f187, %f160, %f282;
- neg.f32 %f189, %f162;
- fma.rn.ftz.f32 %f190, %f189, %f284, %f187;
- mul.ftz.f32 %f191, %f162, %f282;
- fma.rn.ftz.f32 %f192, %f160, %f284, %f191;
- fma.rn.ftz.f32 %f165, %f184, %f185, %f163;
- // inline asm
- cos.approx.f32 %f164, %f165;
- // inline asm
- // inline asm
- sin.approx.f32 %f166, %f165;
- // inline asm
- mul.ftz.f32 %f194, %f164, %f288;
- neg.f32 %f196, %f166;
- fma.rn.ftz.f32 %f197, %f196, %f290, %f194;
- mul.ftz.f32 %f198, %f166, %f288;
- fma.rn.ftz.f32 %f199, %f164, %f290, %f198;
- mul.ftz.f32 %f171, %f163, 0f40400000;
- // inline asm
- cos.approx.f32 %f168, %f171;
- // inline asm
- // inline asm
- sin.approx.f32 %f170, %f171;
- // inline asm
- mul.ftz.f32 %f201, %f168, %f294;
- neg.f32 %f203, %f170;
- fma.rn.ftz.f32 %f204, %f203, %f296, %f201;
- mul.ftz.f32 %f205, %f170, %f294;
- fma.rn.ftz.f32 %f206, %f168, %f296, %f205;
- fma.rn.ftz.f32 %f175, %f184, %f185, %f184;
- // inline asm
- cos.approx.f32 %f172, %f175;
- // inline asm
- // inline asm
- sin.approx.f32 %f174, %f175;
- // inline asm
- mul.ftz.f32 %f208, %f172, %f283;
- neg.f32 %f210, %f174;
- fma.rn.ftz.f32 %f211, %f210, %f285, %f208;
- mul.ftz.f32 %f212, %f174, %f283;
- fma.rn.ftz.f32 %f213, %f172, %f285, %f212;
- add.ftz.f32 %f177, %f175, %f175;
- // inline asm
- cos.approx.f32 %f176, %f177;
- // inline asm
- // inline asm
- sin.approx.f32 %f178, %f177;
- // inline asm
- mul.ftz.f32 %f215, %f176, %f289;
- neg.f32 %f217, %f178;
- fma.rn.ftz.f32 %f218, %f217, %f291, %f215;
- mul.ftz.f32 %f219, %f178, %f289;
- fma.rn.ftz.f32 %f220, %f176, %f291, %f219;
- mul.ftz.f32 %f183, %f175, 0f40400000;
- // inline asm
- cos.approx.f32 %f180, %f183;
- // inline asm
- // inline asm
- sin.approx.f32 %f182, %f183;
- // inline asm
- mul.ftz.f32 %f222, %f180, %f295;
- neg.f32 %f224, %f182;
- fma.rn.ftz.f32 %f225, %f224, %f297, %f222;
- mul.ftz.f32 %f226, %f182, %f295;
- fma.rn.ftz.f32 %f227, %f180, %f297, %f226;
- add.ftz.f32 %f306, %f344, %f197;
- add.ftz.f32 %f307, %f342, %f199;
- sub.ftz.f32 %f310, %f344, %f197;
- sub.ftz.f32 %f311, %f342, %f199;
- add.ftz.f32 %f312, %f190, %f204;
- add.ftz.f32 %f313, %f192, %f206;
- sub.ftz.f32 %f314, %f190, %f204;
- sub.ftz.f32 %f315, %f192, %f206;
- add.ftz.f32 %f260, %f306, %f312;
- add.ftz.f32 %f261, %f307, %f313;
- sub.ftz.f32 %f272, %f306, %f312;
- sub.ftz.f32 %f273, %f307, %f313;
- neg.ftz.f32 %f229, %f315;
- add.ftz.f32 %f266, %f310, %f229;
- add.ftz.f32 %f267, %f311, %f314;
- sub.ftz.f32 %f278, %f310, %f229;
- sub.ftz.f32 %f279, %f311, %f314;
- add.ftz.f32 %f320, %f345, %f218;
- add.ftz.f32 %f321, %f343, %f220;
- sub.ftz.f32 %f324, %f345, %f218;
- sub.ftz.f32 %f325, %f343, %f220;
- add.ftz.f32 %f326, %f211, %f225;
- add.ftz.f32 %f327, %f213, %f227;
- sub.ftz.f32 %f328, %f211, %f225;
- sub.ftz.f32 %f329, %f213, %f227;
- add.ftz.f32 %f262, %f320, %f326;
- add.ftz.f32 %f263, %f321, %f327;
- sub.ftz.f32 %f274, %f320, %f326;
- sub.ftz.f32 %f275, %f321, %f327;
- neg.ftz.f32 %f232, %f329;
- add.ftz.f32 %f268, %f324, %f232;
- add.ftz.f32 %f269, %f325, %f328;
- sub.ftz.f32 %f280, %f324, %f232;
- sub.ftz.f32 %f281, %f325, %f328;
- // inline asm
- mov.u32 %r178, %tid.x;
- // inline asm
- shl.b32 %r182, %r178, 2;
- // inline asm
- mov.u32 %r179, %tid.x;
- // inline asm
- and.b32 %r183, %r179, 31;
- and.b32 %r184, %r182, 536870784;
- add.s32 %r185, %r184, %r183;
- shl.b32 %r186, %r185, 3;
- add.s32 %r188, %r52, %r186;
- st.shared.v2.f32 [%r188], {%f260, %f262};
- st.shared.v2.f32 [%r188+256], {%f266, %f268};
- st.shared.v2.f32 [%r188+512], {%f272, %f274};
- st.shared.v2.f32 [%r188+768], {%f278, %f280};
- bar.sync 0;
- // inline asm
- mov.u32 %r193, %tid.x;
- // inline asm
- shl.b32 %r197, %r193, 3;
- add.s32 %r199, %r52, %r197;
- st.shared.v2.f32 [%r199], {%f261, %f263};
- // inline asm
- mov.u32 %r194, %tid.x;
- // inline asm
- shl.b32 %r200, %r194, 3;
- add.s32 %r201, %r200, %r52;
- st.shared.v2.f32 [%r201+1024], {%f267, %f269};
- // inline asm
- mov.u32 %r195, %tid.x;
- // inline asm
- shl.b32 %r202, %r195, 3;
- add.s32 %r203, %r202, %r52;
- st.shared.v2.f32 [%r203+2048], {%f273, %f275};
- // inline asm
- mov.u32 %r196, %tid.x;
- // inline asm
- shl.b32 %r204, %r196, 3;
- add.s32 %r205, %r204, %r52;
- st.shared.v2.f32 [%r205+3072], {%f279, %f281};
- bar.sync 0;
- ret;
-}
-
-
diff --git a/RTCP/GPUProc/src/UHEP/InvFIR.cl-0.ptx b/RTCP/GPUProc/src/UHEP/InvFIR.cl-0.ptx
deleted file mode 100644
index 2179b103dc77531446838556584310d90c722f7b..0000000000000000000000000000000000000000
--- a/RTCP/GPUProc/src/UHEP/InvFIR.cl-0.ptx
+++ /dev/null
@@ -1,447 +0,0 @@
-//
-// Generated by NVIDIA NVVM Compiler
-// Compiler built on Tue Feb 7 07:15:59 2012 (1328595359)
-// Driver 295.20
-//
-
-.version 3.0
-.target sm_21, texmode_independent
-.address_size 32
-
-
-.entry invFIRfilter(
- .param .u32 .ptr .global .align 1 invFIRfilter_param_0,
- .param .u32 .ptr .global .align 1 invFIRfilter_param_1,
- .param .u32 .ptr .global .align 1 invFIRfilter_param_2
-)
-{
- .reg .f32 %f<386>;
- .reg .pred %p<2>;
- .reg .s32 %r<71>;
-
-
- ld.param.u32 %r21, [invFIRfilter_param_0];
- ld.param.u32 %r22, [invFIRfilter_param_1];
- ld.param.u32 %r23, [invFIRfilter_param_2];
- // inline asm
- mov.u32 %r7, %envreg3;
- // inline asm
- // inline asm
- mov.u32 %r8, %ntid.x;
- // inline asm
- // inline asm
- mov.u32 %r9, %ctaid.x;
- // inline asm
- mul.lo.s32 %r24, %r9, %r8;
- // inline asm
- mov.u32 %r10, %tid.x;
- // inline asm
- add.s32 %r25, %r10, %r7;
- mad.lo.s32 %r26, %r9, %r8, %r25;
- // inline asm
- mov.u32 %r11, %envreg4;
- // inline asm
- // inline asm
- mov.u32 %r12, %ntid.y;
- // inline asm
- // inline asm
- mov.u32 %r13, %ctaid.y;
- // inline asm
- mul.lo.s32 %r27, %r13, %r12;
- // inline asm
- mov.u32 %r14, %tid.y;
- // inline asm
- add.s32 %r28, %r14, %r11;
- mad.lo.s32 %r29, %r13, %r12, %r28;
- // inline asm
- mov.u32 %r15, %envreg5;
- // inline asm
- // inline asm
- mov.u32 %r16, %ntid.z;
- // inline asm
- // inline asm
- mov.u32 %r17, %ctaid.z;
- // inline asm
- mul.lo.s32 %r30, %r17, %r16;
- // inline asm
- mov.u32 %r18, %tid.z;
- // inline asm
- add.s32 %r31, %r18, %r15;
- mad.lo.s32 %r32, %r17, %r16, %r31;
- shl.b32 %r33, %r26, 6;
- add.s32 %r34, %r23, %r33;
- mad.lo.s32 %r35, %r32, 8511488, %r22;
- mad.lo.s32 %r36, %r29, 4255744, %r35;
- shl.b32 %r37, %r26, 2;
- add.s32 %r38, %r36, %r37;
- ld.global.f32 %f17, [%r38];
- mov.u32 %r70, 0;
- ld.global.f32 %f18, [%r38+4096];
- ld.global.f32 %f19, [%r38+8192];
- ld.global.f32 %f20, [%r38+12288];
- ld.global.f32 %f21, [%r38+16384];
- ld.global.f32 %f22, [%r38+20480];
- ld.global.f32 %f23, [%r38+24576];
- ld.global.f32 %f24, [%r38+28672];
- ld.global.f32 %f25, [%r38+32768];
- ld.global.f32 %f26, [%r38+36864];
- ld.global.f32 %f27, [%r38+40960];
- ld.global.f32 %f28, [%r38+45056];
- ld.global.f32 %f29, [%r38+49152];
- ld.global.f32 %f30, [%r38+53248];
- ld.global.f32 %f31, [%r38+57344];
- ld.global.v4.f32 {%f367, %f368, %f369, %f370}, [%r34+48];
- ld.global.v4.f32 {%f371, %f372, %f373, %f374}, [%r34+32];
- ld.global.v4.f32 {%f375, %f376, %f377, %f378}, [%r34+16];
- ld.global.v4.f32 {%f379, %f380, %f381, %f382}, [%r34];
- shl.b32 %r39, %r30, 23;
- shl.b32 %r40, %r15, 23;
- add.s32 %r41, %r39, %r40;
- shl.b32 %r42, %r18, 23;
- add.s32 %r43, %r41, %r42;
- shl.b32 %r44, %r27, 22;
- add.s32 %r45, %r43, %r44;
- shl.b32 %r46, %r11, 22;
- add.s32 %r47, %r45, %r46;
- shl.b32 %r48, %r14, 22;
- add.s32 %r49, %r47, %r48;
- shl.b32 %r50, %r24, 2;
- add.s32 %r51, %r49, %r50;
- shl.b32 %r52, %r7, 2;
- add.s32 %r53, %r51, %r52;
- shl.b32 %r54, %r10, 2;
- add.s32 %r55, %r53, %r54;
- add.s32 %r56, %r55, %r21;
- add.s32 %r1, %r56, 32768;
- mul.lo.s32 %r57, %r15, 8511488;
- mad.lo.s32 %r58, %r30, 8511488, %r57;
- mad.lo.s32 %r59, %r18, 8511488, %r58;
- mad.lo.s32 %r60, %r27, 4255744, %r59;
- mad.lo.s32 %r61, %r11, 4255744, %r60;
- mad.lo.s32 %r62, %r14, 4255744, %r61;
- add.s32 %r63, %r62, %r50;
- add.s32 %r64, %r63, %r52;
- add.s32 %r65, %r64, %r54;
- add.s32 %r66, %r65, %r22;
- add.s32 %r2, %r66, 122880;
- mov.f32 %f347, %f25;
- mov.f32 %f348, %f26;
- mov.f32 %f349, %f27;
- mov.f32 %f350, %f28;
- mov.f32 %f335, %f21;
- mov.f32 %f336, %f22;
- mov.f32 %f337, %f23;
- mov.f32 %f338, %f24;
- mov.f32 %f323, %f17;
- mov.f32 %f324, %f18;
- mov.f32 %f325, %f19;
- mov.f32 %f326, %f20;
- mov.f32 %f383, %f29;
- mov.f32 %f384, %f30;
- mov.f32 %f385, %f31;
- mov.f32 %f362, %f32;
- mov.u32 %r69, 64;
-
-BB0_1:
- mul.ftz.f32 %f35, %f369, %f324;
- fma.rn.ftz.f32 %f36, %f370, %f323, %f35;
- fma.rn.ftz.f32 %f38, %f368, %f325, %f36;
- fma.rn.ftz.f32 %f40, %f367, %f326, %f38;
- fma.rn.ftz.f32 %f42, %f374, %f335, %f40;
- fma.rn.ftz.f32 %f44, %f373, %f336, %f42;
- fma.rn.ftz.f32 %f46, %f372, %f337, %f44;
- fma.rn.ftz.f32 %f48, %f371, %f338, %f46;
- fma.rn.ftz.f32 %f50, %f378, %f347, %f48;
- fma.rn.ftz.f32 %f52, %f377, %f348, %f50;
- fma.rn.ftz.f32 %f54, %f376, %f349, %f52;
- fma.rn.ftz.f32 %f56, %f375, %f350, %f54;
- fma.rn.ftz.f32 %f58, %f382, %f383, %f56;
- fma.rn.ftz.f32 %f60, %f381, %f384, %f58;
- fma.rn.ftz.f32 %f62, %f380, %f385, %f60;
- add.s32 %r67, %r2, %r70;
- ld.global.f32 %f63, [%r67+-57344];
- fma.rn.ftz.f32 %f64, %f379, %f63, %f62;
- add.s32 %r68, %r1, %r70;
- st.global.f32 [%r68+-32768], %f64;
- mul.ftz.f32 %f65, %f369, %f325;
- fma.rn.ftz.f32 %f66, %f370, %f324, %f65;
- fma.rn.ftz.f32 %f67, %f368, %f326, %f66;
- fma.rn.ftz.f32 %f68, %f367, %f335, %f67;
- fma.rn.ftz.f32 %f69, %f374, %f336, %f68;
- fma.rn.ftz.f32 %f70, %f373, %f337, %f69;
- fma.rn.ftz.f32 %f71, %f372, %f338, %f70;
- fma.rn.ftz.f32 %f72, %f371, %f347, %f71;
- fma.rn.ftz.f32 %f73, %f378, %f348, %f72;
- fma.rn.ftz.f32 %f74, %f377, %f349, %f73;
- fma.rn.ftz.f32 %f75, %f376, %f350, %f74;
- fma.rn.ftz.f32 %f76, %f375, %f383, %f75;
- fma.rn.ftz.f32 %f77, %f382, %f384, %f76;
- fma.rn.ftz.f32 %f78, %f381, %f385, %f77;
- ld.global.f32 %f79, [%r67+-53248];
- fma.rn.ftz.f32 %f80, %f380, %f79, %f78;
- fma.rn.ftz.f32 %f81, %f379, %f323, %f80;
- st.global.f32 [%r68+-28672], %f81;
- mul.ftz.f32 %f82, %f369, %f326;
- fma.rn.ftz.f32 %f83, %f370, %f325, %f82;
- fma.rn.ftz.f32 %f84, %f368, %f335, %f83;
- fma.rn.ftz.f32 %f85, %f367, %f336, %f84;
- fma.rn.ftz.f32 %f86, %f374, %f337, %f85;
- fma.rn.ftz.f32 %f87, %f373, %f338, %f86;
- fma.rn.ftz.f32 %f88, %f372, %f347, %f87;
- fma.rn.ftz.f32 %f89, %f371, %f348, %f88;
- fma.rn.ftz.f32 %f90, %f378, %f349, %f89;
- fma.rn.ftz.f32 %f91, %f377, %f350, %f90;
- fma.rn.ftz.f32 %f92, %f376, %f383, %f91;
- fma.rn.ftz.f32 %f93, %f375, %f384, %f92;
- fma.rn.ftz.f32 %f94, %f382, %f385, %f93;
- ld.global.f32 %f95, [%r67+-49152];
- fma.rn.ftz.f32 %f96, %f381, %f95, %f94;
- fma.rn.ftz.f32 %f97, %f380, %f323, %f96;
- fma.rn.ftz.f32 %f98, %f379, %f324, %f97;
- st.global.f32 [%r68+-24576], %f98;
- mul.ftz.f32 %f99, %f369, %f335;
- fma.rn.ftz.f32 %f100, %f370, %f326, %f99;
- fma.rn.ftz.f32 %f101, %f368, %f336, %f100;
- fma.rn.ftz.f32 %f102, %f367, %f337, %f101;
- fma.rn.ftz.f32 %f103, %f374, %f338, %f102;
- fma.rn.ftz.f32 %f104, %f373, %f347, %f103;
- fma.rn.ftz.f32 %f105, %f372, %f348, %f104;
- fma.rn.ftz.f32 %f106, %f371, %f349, %f105;
- fma.rn.ftz.f32 %f107, %f378, %f350, %f106;
- fma.rn.ftz.f32 %f108, %f377, %f383, %f107;
- fma.rn.ftz.f32 %f109, %f376, %f384, %f108;
- fma.rn.ftz.f32 %f110, %f375, %f385, %f109;
- ld.global.f32 %f111, [%r67+-45056];
- fma.rn.ftz.f32 %f112, %f382, %f111, %f110;
- fma.rn.ftz.f32 %f113, %f381, %f323, %f112;
- fma.rn.ftz.f32 %f114, %f380, %f324, %f113;
- fma.rn.ftz.f32 %f115, %f379, %f325, %f114;
- st.global.f32 [%r68+-20480], %f115;
- mul.ftz.f32 %f116, %f369, %f336;
- fma.rn.ftz.f32 %f117, %f370, %f335, %f116;
- fma.rn.ftz.f32 %f118, %f368, %f337, %f117;
- fma.rn.ftz.f32 %f119, %f367, %f338, %f118;
- fma.rn.ftz.f32 %f120, %f374, %f347, %f119;
- fma.rn.ftz.f32 %f121, %f373, %f348, %f120;
- fma.rn.ftz.f32 %f122, %f372, %f349, %f121;
- fma.rn.ftz.f32 %f123, %f371, %f350, %f122;
- fma.rn.ftz.f32 %f124, %f378, %f383, %f123;
- fma.rn.ftz.f32 %f125, %f377, %f384, %f124;
- fma.rn.ftz.f32 %f126, %f376, %f385, %f125;
- ld.global.f32 %f127, [%r67+-40960];
- fma.rn.ftz.f32 %f128, %f375, %f127, %f126;
- fma.rn.ftz.f32 %f129, %f382, %f323, %f128;
- fma.rn.ftz.f32 %f130, %f381, %f324, %f129;
- fma.rn.ftz.f32 %f131, %f380, %f325, %f130;
- fma.rn.ftz.f32 %f132, %f379, %f326, %f131;
- st.global.f32 [%r68+-16384], %f132;
- mul.ftz.f32 %f133, %f369, %f337;
- fma.rn.ftz.f32 %f134, %f370, %f336, %f133;
- fma.rn.ftz.f32 %f135, %f368, %f338, %f134;
- fma.rn.ftz.f32 %f136, %f367, %f347, %f135;
- fma.rn.ftz.f32 %f137, %f374, %f348, %f136;
- fma.rn.ftz.f32 %f138, %f373, %f349, %f137;
- fma.rn.ftz.f32 %f139, %f372, %f350, %f138;
- fma.rn.ftz.f32 %f140, %f371, %f383, %f139;
- fma.rn.ftz.f32 %f141, %f378, %f384, %f140;
- fma.rn.ftz.f32 %f142, %f377, %f385, %f141;
- ld.global.f32 %f143, [%r67+-36864];
- fma.rn.ftz.f32 %f144, %f376, %f143, %f142;
- fma.rn.ftz.f32 %f145, %f375, %f323, %f144;
- fma.rn.ftz.f32 %f146, %f382, %f324, %f145;
- fma.rn.ftz.f32 %f147, %f381, %f325, %f146;
- fma.rn.ftz.f32 %f148, %f380, %f326, %f147;
- fma.rn.ftz.f32 %f149, %f379, %f335, %f148;
- st.global.f32 [%r68+-12288], %f149;
- mul.ftz.f32 %f150, %f369, %f338;
- fma.rn.ftz.f32 %f151, %f370, %f337, %f150;
- fma.rn.ftz.f32 %f152, %f368, %f347, %f151;
- fma.rn.ftz.f32 %f153, %f367, %f348, %f152;
- fma.rn.ftz.f32 %f154, %f374, %f349, %f153;
- fma.rn.ftz.f32 %f155, %f373, %f350, %f154;
- fma.rn.ftz.f32 %f156, %f372, %f383, %f155;
- fma.rn.ftz.f32 %f157, %f371, %f384, %f156;
- fma.rn.ftz.f32 %f158, %f378, %f385, %f157;
- ld.global.f32 %f159, [%r67+-32768];
- fma.rn.ftz.f32 %f160, %f377, %f159, %f158;
- fma.rn.ftz.f32 %f161, %f376, %f323, %f160;
- fma.rn.ftz.f32 %f162, %f375, %f324, %f161;
- fma.rn.ftz.f32 %f163, %f382, %f325, %f162;
- fma.rn.ftz.f32 %f164, %f381, %f326, %f163;
- fma.rn.ftz.f32 %f165, %f380, %f335, %f164;
- fma.rn.ftz.f32 %f166, %f379, %f336, %f165;
- st.global.f32 [%r68+-8192], %f166;
- mul.ftz.f32 %f167, %f369, %f347;
- fma.rn.ftz.f32 %f168, %f370, %f338, %f167;
- fma.rn.ftz.f32 %f169, %f368, %f348, %f168;
- fma.rn.ftz.f32 %f170, %f367, %f349, %f169;
- fma.rn.ftz.f32 %f171, %f374, %f350, %f170;
- fma.rn.ftz.f32 %f172, %f373, %f383, %f171;
- fma.rn.ftz.f32 %f173, %f372, %f384, %f172;
- fma.rn.ftz.f32 %f174, %f371, %f385, %f173;
- ld.global.f32 %f175, [%r67+-28672];
- fma.rn.ftz.f32 %f176, %f378, %f175, %f174;
- fma.rn.ftz.f32 %f177, %f377, %f323, %f176;
- fma.rn.ftz.f32 %f178, %f376, %f324, %f177;
- fma.rn.ftz.f32 %f179, %f375, %f325, %f178;
- fma.rn.ftz.f32 %f180, %f382, %f326, %f179;
- fma.rn.ftz.f32 %f181, %f381, %f335, %f180;
- fma.rn.ftz.f32 %f182, %f380, %f336, %f181;
- fma.rn.ftz.f32 %f183, %f379, %f337, %f182;
- st.global.f32 [%r68+-4096], %f183;
- mul.ftz.f32 %f184, %f369, %f348;
- fma.rn.ftz.f32 %f185, %f370, %f347, %f184;
- fma.rn.ftz.f32 %f186, %f368, %f349, %f185;
- fma.rn.ftz.f32 %f187, %f367, %f350, %f186;
- fma.rn.ftz.f32 %f188, %f374, %f383, %f187;
- fma.rn.ftz.f32 %f189, %f373, %f384, %f188;
- fma.rn.ftz.f32 %f190, %f372, %f385, %f189;
- ld.global.f32 %f191, [%r67+-24576];
- fma.rn.ftz.f32 %f192, %f371, %f191, %f190;
- fma.rn.ftz.f32 %f193, %f378, %f323, %f192;
- fma.rn.ftz.f32 %f194, %f377, %f324, %f193;
- fma.rn.ftz.f32 %f195, %f376, %f325, %f194;
- fma.rn.ftz.f32 %f196, %f375, %f326, %f195;
- fma.rn.ftz.f32 %f197, %f382, %f335, %f196;
- fma.rn.ftz.f32 %f198, %f381, %f336, %f197;
- fma.rn.ftz.f32 %f199, %f380, %f337, %f198;
- fma.rn.ftz.f32 %f200, %f379, %f338, %f199;
- st.global.f32 [%r68], %f200;
- mul.ftz.f32 %f201, %f369, %f349;
- fma.rn.ftz.f32 %f202, %f370, %f348, %f201;
- fma.rn.ftz.f32 %f203, %f368, %f350, %f202;
- fma.rn.ftz.f32 %f204, %f367, %f383, %f203;
- fma.rn.ftz.f32 %f205, %f374, %f384, %f204;
- fma.rn.ftz.f32 %f206, %f373, %f385, %f205;
- ld.global.f32 %f207, [%r67+-20480];
- fma.rn.ftz.f32 %f208, %f372, %f207, %f206;
- fma.rn.ftz.f32 %f209, %f371, %f323, %f208;
- fma.rn.ftz.f32 %f210, %f378, %f324, %f209;
- fma.rn.ftz.f32 %f211, %f377, %f325, %f210;
- fma.rn.ftz.f32 %f212, %f376, %f326, %f211;
- fma.rn.ftz.f32 %f213, %f375, %f335, %f212;
- fma.rn.ftz.f32 %f214, %f382, %f336, %f213;
- fma.rn.ftz.f32 %f215, %f381, %f337, %f214;
- fma.rn.ftz.f32 %f216, %f380, %f338, %f215;
- fma.rn.ftz.f32 %f217, %f379, %f347, %f216;
- st.global.f32 [%r68+4096], %f217;
- mul.ftz.f32 %f218, %f369, %f350;
- fma.rn.ftz.f32 %f219, %f370, %f349, %f218;
- fma.rn.ftz.f32 %f220, %f368, %f383, %f219;
- fma.rn.ftz.f32 %f221, %f367, %f384, %f220;
- fma.rn.ftz.f32 %f222, %f374, %f385, %f221;
- ld.global.f32 %f223, [%r67+-16384];
- fma.rn.ftz.f32 %f224, %f373, %f223, %f222;
- fma.rn.ftz.f32 %f225, %f372, %f323, %f224;
- fma.rn.ftz.f32 %f226, %f371, %f324, %f225;
- fma.rn.ftz.f32 %f227, %f378, %f325, %f226;
- fma.rn.ftz.f32 %f228, %f377, %f326, %f227;
- fma.rn.ftz.f32 %f229, %f376, %f335, %f228;
- fma.rn.ftz.f32 %f230, %f375, %f336, %f229;
- fma.rn.ftz.f32 %f231, %f382, %f337, %f230;
- fma.rn.ftz.f32 %f232, %f381, %f338, %f231;
- fma.rn.ftz.f32 %f233, %f380, %f347, %f232;
- fma.rn.ftz.f32 %f234, %f379, %f348, %f233;
- st.global.f32 [%r68+8192], %f234;
- mul.ftz.f32 %f235, %f369, %f383;
- fma.rn.ftz.f32 %f236, %f370, %f350, %f235;
- fma.rn.ftz.f32 %f237, %f368, %f384, %f236;
- fma.rn.ftz.f32 %f238, %f367, %f385, %f237;
- ld.global.f32 %f239, [%r67+-12288];
- fma.rn.ftz.f32 %f240, %f374, %f239, %f238;
- fma.rn.ftz.f32 %f241, %f373, %f323, %f240;
- fma.rn.ftz.f32 %f242, %f372, %f324, %f241;
- fma.rn.ftz.f32 %f243, %f371, %f325, %f242;
- fma.rn.ftz.f32 %f244, %f378, %f326, %f243;
- fma.rn.ftz.f32 %f245, %f377, %f335, %f244;
- fma.rn.ftz.f32 %f246, %f376, %f336, %f245;
- fma.rn.ftz.f32 %f247, %f375, %f337, %f246;
- fma.rn.ftz.f32 %f248, %f382, %f338, %f247;
- fma.rn.ftz.f32 %f249, %f381, %f347, %f248;
- fma.rn.ftz.f32 %f250, %f380, %f348, %f249;
- fma.rn.ftz.f32 %f251, %f379, %f349, %f250;
- st.global.f32 [%r68+12288], %f251;
- mul.ftz.f32 %f252, %f369, %f384;
- fma.rn.ftz.f32 %f253, %f370, %f383, %f252;
- fma.rn.ftz.f32 %f254, %f368, %f385, %f253;
- ld.global.f32 %f255, [%r67+-8192];
- fma.rn.ftz.f32 %f256, %f367, %f255, %f254;
- fma.rn.ftz.f32 %f257, %f374, %f323, %f256;
- fma.rn.ftz.f32 %f258, %f373, %f324, %f257;
- fma.rn.ftz.f32 %f259, %f372, %f325, %f258;
- fma.rn.ftz.f32 %f260, %f371, %f326, %f259;
- fma.rn.ftz.f32 %f261, %f378, %f335, %f260;
- fma.rn.ftz.f32 %f262, %f377, %f336, %f261;
- fma.rn.ftz.f32 %f263, %f376, %f337, %f262;
- fma.rn.ftz.f32 %f264, %f375, %f338, %f263;
- fma.rn.ftz.f32 %f265, %f382, %f347, %f264;
- fma.rn.ftz.f32 %f266, %f381, %f348, %f265;
- fma.rn.ftz.f32 %f267, %f380, %f349, %f266;
- fma.rn.ftz.f32 %f268, %f379, %f350, %f267;
- st.global.f32 [%r68+16384], %f268;
- mul.ftz.f32 %f269, %f369, %f385;
- fma.rn.ftz.f32 %f270, %f370, %f384, %f269;
- ld.global.f32 %f271, [%r67+-4096];
- fma.rn.ftz.f32 %f272, %f368, %f271, %f270;
- fma.rn.ftz.f32 %f273, %f367, %f323, %f272;
- fma.rn.ftz.f32 %f274, %f374, %f324, %f273;
- fma.rn.ftz.f32 %f275, %f373, %f325, %f274;
- fma.rn.ftz.f32 %f276, %f372, %f326, %f275;
- fma.rn.ftz.f32 %f277, %f371, %f335, %f276;
- fma.rn.ftz.f32 %f278, %f378, %f336, %f277;
- fma.rn.ftz.f32 %f279, %f377, %f337, %f278;
- fma.rn.ftz.f32 %f280, %f376, %f338, %f279;
- fma.rn.ftz.f32 %f281, %f375, %f347, %f280;
- fma.rn.ftz.f32 %f282, %f382, %f348, %f281;
- fma.rn.ftz.f32 %f283, %f381, %f349, %f282;
- fma.rn.ftz.f32 %f284, %f380, %f350, %f283;
- fma.rn.ftz.f32 %f285, %f379, %f383, %f284;
- st.global.f32 [%r68+20480], %f285;
- ld.global.f32 %f286, [%r67];
- mul.ftz.f32 %f287, %f369, %f286;
- fma.rn.ftz.f32 %f288, %f370, %f385, %f287;
- fma.rn.ftz.f32 %f289, %f368, %f323, %f288;
- fma.rn.ftz.f32 %f290, %f367, %f324, %f289;
- fma.rn.ftz.f32 %f291, %f374, %f325, %f290;
- fma.rn.ftz.f32 %f292, %f373, %f326, %f291;
- fma.rn.ftz.f32 %f293, %f372, %f335, %f292;
- fma.rn.ftz.f32 %f294, %f371, %f336, %f293;
- fma.rn.ftz.f32 %f295, %f378, %f337, %f294;
- fma.rn.ftz.f32 %f296, %f377, %f338, %f295;
- fma.rn.ftz.f32 %f297, %f376, %f347, %f296;
- fma.rn.ftz.f32 %f298, %f375, %f348, %f297;
- fma.rn.ftz.f32 %f299, %f382, %f349, %f298;
- fma.rn.ftz.f32 %f300, %f381, %f350, %f299;
- fma.rn.ftz.f32 %f301, %f380, %f383, %f300;
- fma.rn.ftz.f32 %f302, %f379, %f384, %f301;
- st.global.f32 [%r68+24576], %f302;
- mul.ftz.f32 %f303, %f369, %f323;
- fma.rn.ftz.f32 %f304, %f370, %f286, %f303;
- fma.rn.ftz.f32 %f305, %f368, %f324, %f304;
- fma.rn.ftz.f32 %f306, %f367, %f325, %f305;
- fma.rn.ftz.f32 %f307, %f374, %f326, %f306;
- fma.rn.ftz.f32 %f308, %f373, %f335, %f307;
- fma.rn.ftz.f32 %f309, %f372, %f336, %f308;
- fma.rn.ftz.f32 %f310, %f371, %f337, %f309;
- fma.rn.ftz.f32 %f311, %f378, %f338, %f310;
- fma.rn.ftz.f32 %f312, %f377, %f347, %f311;
- fma.rn.ftz.f32 %f313, %f376, %f348, %f312;
- fma.rn.ftz.f32 %f314, %f375, %f349, %f313;
- fma.rn.ftz.f32 %f315, %f382, %f350, %f314;
- fma.rn.ftz.f32 %f316, %f381, %f383, %f315;
- fma.rn.ftz.f32 %f317, %f380, %f384, %f316;
- fma.rn.ftz.f32 %f318, %f379, %f385, %f317;
- st.global.f32 [%r68+28672], %f318;
- add.s32 %r70, %r70, 65536;
- add.s32 %r69, %r69, -1;
- setp.ne.s32 %p1, %r69, 0;
- mov.f32 %f383, %f383;
- mov.f32 %f384, %f384;
- mov.f32 %f385, %f385;
- mov.f32 %f366, %f286;
- @%p1 bra BB0_1;
-
- ret;
-}
-
-
diff --git a/RTCP/GPUProc/src/UHEP/Transpose.cl-0.ptx b/RTCP/GPUProc/src/UHEP/Transpose.cl-0.ptx
deleted file mode 100644
index cb8248be325fa81a44e309be3a1a84ebe64e5e75..0000000000000000000000000000000000000000
--- a/RTCP/GPUProc/src/UHEP/Transpose.cl-0.ptx
+++ /dev/null
@@ -1,142 +0,0 @@
-//
-// Generated by NVIDIA NVVM Compiler
-// Compiler built on Tue Feb 7 07:15:59 2012 (1328595359)
-// Driver 295.20
-//
-
-.version 3.0
-.target sm_21, texmode_independent
-.address_size 32
-
-.extern .shared .align 16 .b8 shr_1_tmp[4352];
-
-.entry UHEP_Transpose(
- .param .u32 .ptr .global .align 1 UHEP_Transpose_param_0,
- .param .u32 .ptr .global .align 1 UHEP_Transpose_param_1,
- .param .u32 .ptr .global .align 4 UHEP_Transpose_param_2
-)
-{
- .reg .f32 %f<30>;
- .reg .pred %p<7>;
- .reg .s32 %r<61>;
-
-
- ld.param.u32 %r25, [UHEP_Transpose_param_0];
- ld.param.u32 %r26, [UHEP_Transpose_param_1];
- ld.param.u32 %r27, [UHEP_Transpose_param_2];
- // inline asm
- mov.u32 %r12, %envreg4;
- // inline asm
- // inline asm
- mov.u32 %r13, %ntid.y;
- // inline asm
- // inline asm
- mov.u32 %r14, %ctaid.y;
- // inline asm
- // inline asm
- mov.u32 %r15, %tid.y;
- // inline asm
- add.s32 %r28, %r15, %r12;
- mad.lo.s32 %r29, %r14, %r13, %r28;
- // inline asm
- mov.u32 %r16, %envreg5;
- // inline asm
- // inline asm
- mov.u32 %r17, %ntid.z;
- // inline asm
- // inline asm
- mov.u32 %r18, %ctaid.z;
- // inline asm
- // inline asm
- mov.u32 %r19, %tid.z;
- // inline asm
- add.s32 %r30, %r19, %r16;
- mad.lo.s32 %r31, %r18, %r17, %r30;
- // inline asm
- mov.u32 %r20, %tid.x;
- // inline asm
- shl.b32 %r32, %r29, 4;
- and.b32 %r33, %r20, 15;
- add.s32 %r34, %r32, %r33;
- // inline asm
- mov.u32 %r21, %tid.x;
- // inline asm
- shr.u32 %r35, %r21, 4;
- shl.b32 %r36, %r31, 4;
- add.s32 %r37, %r36, %r35;
- shl.b32 %r38, %r37, 2;
- add.s32 %r39, %r27, %r38;
- setp.lt.u32 %p4, %r34, 4;
- ld.global.u32 %r40, [%r39];
- setp.gt.s32 %p5, %r40, -1;
- and.pred %p1, %p4, %p5;
- // inline asm
- mov.u32 %r22, %tid.x;
- // inline asm
- shr.u32 %r41, %r22, 4;
- // inline asm
- mov.u32 %r23, %tid.x;
- // inline asm
- and.b32 %r42, %r23, 15;
- add.s32 %r43, %r36, %r42;
- shl.b32 %r44, %r43, 2;
- add.s32 %r45, %r27, %r44;
- add.s32 %r46, %r41, %r32;
- setp.lt.u32 %p2, %r46, 4;
- mov.u32 %r47, shr_1_tmp;
- mad.lo.s32 %r48, %r33, 272, %r47;
- and.b32 %r49, %r21, -16;
- add.s32 %r1, %r48, %r49;
- ld.global.u32 %r50, [%r45];
- setp.gt.s32 %p3, %r50, -1;
- mad.lo.s32 %r51, %r41, 272, %r47;
- shl.b32 %r52, %r42, 4;
- add.s32 %r2, %r51, %r52;
- shl.b32 %r53, %r43, 3;
- mad.lo.s32 %r54, %r46, 8511488, %r53;
- add.s32 %r55, %r54, %r25;
- add.s32 %r60, %r55, 4255744;
- shl.b32 %r56, %r34, 4;
- mad.lo.s32 %r57, %r40, 66496, %r56;
- add.s32 %r59, %r26, %r57;
- mov.u32 %r58, 1039;
-
-BB0_1:
- add.s32 %r8, %r60, -4255744;
- @!%p1 bra BB0_3;
-
- ld.global.v4.f32 {%f22, %f23, %f24, %f25}, [%r59];
- st.shared.v4.f32 [%r1], {%f22, %f23, %f24, %f25};
-
-BB0_3:
- bar.sync 0;
- @!%p2 bra BB0_8;
-
- @%p3 bra BB0_6;
-
- mov.f32 %f1, 0f00000000;
- mov.f32 %f26, %f1;
- mov.f32 %f27, %f1;
- mov.f32 %f28, %f1;
- mov.f32 %f29, %f1;
- bra.uni BB0_7;
-
-BB0_6:
- ld.shared.v4.f32 {%f26, %f27, %f28, %f29}, [%r2];
-
-BB0_7:
- st.global.v2.f32 [%r8], {%f26, %f27};
- st.global.v2.f32 [%r8+4255744], {%f28, %f29};
-
-BB0_8:
- bar.sync 0;
- add.s32 %r60, %r60, 4096;
- add.s32 %r59, %r59, 64;
- add.s32 %r58, %r58, -1;
- setp.ne.s32 %p6, %r58, 0;
- @%p6 bra BB0_1;
-
- ret;
-}
-
-
diff --git a/RTCP/GPUProc/src/UHEP/Trigger.cl-0.ptx b/RTCP/GPUProc/src/UHEP/Trigger.cl-0.ptx
deleted file mode 100644
index ae2713f9037df945f18a988f001c6c0d0abbf29b..0000000000000000000000000000000000000000
--- a/RTCP/GPUProc/src/UHEP/Trigger.cl-0.ptx
+++ /dev/null
@@ -1,511 +0,0 @@
-//
-// Generated by NVIDIA NVVM Compiler
-// Compiler built on Tue Feb 7 07:15:59 2012 (1328595359)
-// Driver 295.20
-//
-
-.version 3.0
-.target sm_21, texmode_independent
-.address_size 32
-
-.extern .shared .align 64 .b8 shr_4_tmp[16384];
-
-.entry trigger(
- .param .u32 .ptr .global .align 1 trigger_param_0,
- .param .u32 .ptr .global .align 4 trigger_param_1
-)
-{
- .reg .f32 %f<434>;
- .reg .pred %p<21>;
- .reg .s32 %r<71>;
-
-
- ld.param.u32 %r31, [trigger_param_1];
- // inline asm
- mov.u32 %r23, %tid.x;
- // inline asm
- // inline asm
- mov.u32 %r24, %tid.y;
- // inline asm
- shl.b32 %r32, %r24, 4;
- // inline asm
- mov.u32 %r25, %envreg5;
- // inline asm
- // inline asm
- mov.u32 %r26, %ntid.z;
- // inline asm
- // inline asm
- mov.u32 %r27, %ctaid.z;
- // inline asm
- // inline asm
- mov.u32 %r28, %tid.z;
- // inline asm
- add.s32 %r33, %r28, %r25;
- mad.lo.s32 %r3, %r27, %r26, %r33;
- shl.b32 %r34, %r24, 10;
- mov.u32 %r35, shr_4_tmp;
- add.s32 %r36, %r35, %r34;
- shl.b32 %r37, %r23, 6;
- add.s32 %r4, %r36, %r37;
- add.s32 %r2, %r32, %r23;
- shl.b32 %r38, %r2, 12;
- and.b32 %r5, %r38, 16773120;
- shl.b32 %r39, %r3, 21;
- add.s32 %r40, %r23, %r39;
- shl.b32 %r41, %r24, 12;
- add.s32 %r42, %r40, %r41;
- shl.b32 %r43, %r42, 2;
- add.s32 %r70, %r31, %r43;
- shl.b32 %r44, %r24, 6;
- shl.b32 %r45, %r23, 2;
- add.s32 %r46, %r35, %r44;
- add.s32 %r47, %r46, %r45;
- add.s32 %r7, %r47, 3072;
- mov.f32 %f420, 0f00000000;
- mov.f32 %f430, %f420;
- mov.f32 %f431, %f420;
- mov.f32 %f432, %f420;
- mov.f32 %f433, %f420;
- mov.f32 %f422, %f405;
- mov.f32 %f419, %f420;
- mov.f32 %f421, %f420;
- mov.f32 %f418, %f420;
- mov.u32 %r65, 0;
- mov.u32 %r64, %r65;
- mov.f32 %f426, %f420;
- mov.f32 %f427, %f420;
- mov.f32 %f428, %f420;
- mov.f32 %f429, %f420;
- mov.f32 %f423, %f420;
- mov.f32 %f424, %f420;
- mov.f32 %f425, %f420;
-
-BB0_1:
- mov.u32 %r68, %r70;
- mov.u32 %r8, %r68;
- shl.b32 %r49, %r64, 4;
- add.s32 %r12, %r5, %r49;
- mov.u32 %r67, 16;
- mov.u32 %r66, %r7;
- mov.u32 %r69, %r8;
-
-BB0_2:
- mov.u32 %r15, %r69;
- ld.global.f32 %f40, [%r15];
- ld.global.f32 %f41, [%r15+4194304];
- mul.ftz.f32 %f42, %f41, %f41;
- fma.rn.ftz.f32 %f43, %f40, %f40, %f42;
- st.shared.f32 [%r66+-3072], %f43;
- sub.ftz.f32 %f44, %f43, %f419;
- add.ftz.f32 %f45, %f421, 0f3F800000;
- div.approx.ftz.f32 %f46, %f44, %f45;
- add.ftz.f32 %f47, %f419, %f46;
- sub.ftz.f32 %f48, %f43, %f47;
- fma.rn.ftz.f32 %f49, %f44, %f48, %f420;
- ld.global.f32 %f50, [%r15+262144];
- ld.global.f32 %f51, [%r15+4456448];
- mul.ftz.f32 %f52, %f51, %f51;
- fma.rn.ftz.f32 %f53, %f50, %f50, %f52;
- st.shared.f32 [%r66+-2048], %f53;
- add.ftz.f32 %f54, %f45, 0f3F800000;
- sub.ftz.f32 %f55, %f53, %f47;
- div.approx.ftz.f32 %f56, %f55, %f54;
- add.ftz.f32 %f57, %f47, %f56;
- sub.ftz.f32 %f58, %f53, %f57;
- fma.rn.ftz.f32 %f59, %f55, %f58, %f49;
- ld.global.f32 %f60, [%r15+524288];
- ld.global.f32 %f61, [%r15+4718592];
- mul.ftz.f32 %f62, %f61, %f61;
- fma.rn.ftz.f32 %f63, %f60, %f60, %f62;
- st.shared.f32 [%r66+-1024], %f63;
- add.ftz.f32 %f64, %f54, 0f3F800000;
- sub.ftz.f32 %f65, %f63, %f57;
- div.approx.ftz.f32 %f66, %f65, %f64;
- add.ftz.f32 %f67, %f57, %f66;
- sub.ftz.f32 %f68, %f63, %f67;
- fma.rn.ftz.f32 %f69, %f65, %f68, %f59;
- ld.global.f32 %f70, [%r15+786432];
- ld.global.f32 %f71, [%r15+4980736];
- mul.ftz.f32 %f72, %f71, %f71;
- fma.rn.ftz.f32 %f73, %f70, %f70, %f72;
- st.shared.f32 [%r66], %f73;
- add.ftz.f32 %f421, %f64, 0f3F800000;
- sub.ftz.f32 %f74, %f73, %f67;
- div.approx.ftz.f32 %f75, %f74, %f421;
- add.ftz.f32 %f419, %f67, %f75;
- sub.ftz.f32 %f76, %f73, %f419;
- fma.rn.ftz.f32 %f420, %f74, %f76, %f69;
- add.s32 %r16, %r15, 1048576;
- add.s32 %r66, %r66, 4096;
- add.s32 %r67, %r67, -4;
- setp.ne.s32 %p1, %r67, 0;
- mov.u32 %r69, %r16;
- @%p1 bra BB0_2;
-
- bar.sync 0;
- ld.shared.v4.f32 {%f258, %f259, %f260, %f261}, [%r4];
- add.ftz.f32 %f79, %f422, %f258;
- sub.ftz.f32 %f81, %f79, %f423;
- add.ftz.f32 %f83, %f81, %f259;
- sub.ftz.f32 %f85, %f83, %f424;
- add.ftz.f32 %f87, %f85, %f260;
- sub.ftz.f32 %f89, %f87, %f425;
- add.ftz.f32 %f91, %f89, %f261;
- sub.ftz.f32 %f93, %f91, %f426;
- ld.shared.v4.f32 {%f274, %f423, %f424, %f425}, [%r4+16];
- add.ftz.f32 %f95, %f93, %f274;
- sub.ftz.f32 %f97, %f95, %f427;
- add.ftz.f32 %f99, %f97, %f423;
- sub.ftz.f32 %f101, %f99, %f428;
- add.ftz.f32 %f103, %f101, %f424;
- sub.ftz.f32 %f105, %f103, %f429;
- add.ftz.f32 %f107, %f105, %f425;
- sub.ftz.f32 %f109, %f107, %f430;
- ld.shared.v4.f32 {%f426, %f427, %f428, %f429}, [%r4+32];
- add.ftz.f32 %f111, %f109, %f426;
- sub.ftz.f32 %f113, %f111, %f431;
- add.ftz.f32 %f115, %f113, %f427;
- sub.ftz.f32 %f117, %f115, %f432;
- add.ftz.f32 %f119, %f117, %f428;
- sub.ftz.f32 %f121, %f119, %f433;
- add.ftz.f32 %f123, %f121, %f429;
- sub.ftz.f32 %f124, %f123, %f258;
- ld.shared.v4.f32 {%f430, %f431, %f432, %f433}, [%r4+48];
- add.ftz.f32 %f126, %f124, %f430;
- sub.ftz.f32 %f127, %f126, %f259;
- add.ftz.f32 %f129, %f127, %f431;
- sub.ftz.f32 %f130, %f129, %f260;
- add.ftz.f32 %f132, %f130, %f432;
- sub.ftz.f32 %f133, %f132, %f261;
- add.ftz.f32 %f135, %f133, %f433;
- sub.ftz.f32 %f136, %f135, %f274;
- max.f32 %f137, %f81, %f85;
- max.f32 %f138, %f89, %f93;
- max.f32 %f139, %f137, %f138;
- max.f32 %f140, %f97, %f101;
- max.f32 %f141, %f105, %f109;
- max.f32 %f142, %f140, %f141;
- max.f32 %f143, %f113, %f117;
- max.f32 %f144, %f121, %f124;
- max.f32 %f145, %f143, %f144;
- max.f32 %f146, %f127, %f130;
- max.f32 %f147, %f133, %f136;
- max.f32 %f148, %f146, %f147;
- max.f32 %f149, %f139, %f142;
- max.f32 %f150, %f145, %f148;
- max.f32 %f151, %f149, %f150;
- setp.ltu.ftz.f32 %p2, %f151, %f418;
- selp.f32 %f418, %f418, %f151, %p2;
- selp.b32 %r65, %r65, %r12, %p2;
- mov.f32 %f306, %f81;
- mov.f32 %f307, %f85;
- mov.f32 %f308, %f89;
- mov.f32 %f309, %f93;
- mov.f32 %f326, %f97;
- mov.f32 %f327, %f101;
- mov.f32 %f328, %f105;
- mov.f32 %f329, %f109;
- mov.f32 %f346, %f113;
- mov.f32 %f347, %f117;
- mov.f32 %f348, %f121;
- mov.f32 %f349, %f124;
- mov.f32 %f362, %f127;
- mov.f32 %f363, %f130;
- mov.f32 %f364, %f133;
- mov.f32 %f422, %f136;
- bar.sync 0;
- add.s32 %r64, %r64, 1;
- add.s32 %r21, %r8, 64;
- setp.ne.s32 %p3, %r64, 256;
- mov.u32 %r70, %r21;
- @%p3 bra BB0_1;
-
- shl.b32 %r50, %r2, 2;
- add.s32 %r22, %r35, %r50;
- st.shared.f32 [%r22], %f419;
- st.shared.f32 [%r22+1024], %f420;
- st.shared.f32 [%r22+2048], %f418;
- st.shared.u32 [%r22+3072], %r65;
- setp.lt.u32 %p4, %r2, 128;
- @%p4 bra BB0_5;
- bra.uni BB0_7;
-
-BB0_5:
- ld.shared.f32 %f152, [%r22+1536];
- ld.shared.f32 %f153, [%r22+512];
- sub.ftz.f32 %f154, %f153, %f419;
- add.ftz.f32 %f155, %f419, %f153;
- div.rn.ftz.f32 %f156, %f155, 0f40000000;
- st.shared.f32 [%r22], %f156;
- add.ftz.f32 %f157, %f420, %f152;
- mul.ftz.f32 %f158, %f154, %f154;
- mul.ftz.f32 %f159, %f158, %f421;
- div.rn.ftz.f32 %f160, %f159, 0f40000000;
- add.ftz.f32 %f161, %f157, %f160;
- st.shared.f32 [%r22+1024], %f161;
- add.ftz.f32 %f421, %f421, %f421;
- ld.shared.f32 %f13, [%r22+2560];
- setp.lt.ftz.f32 %p5, %f418, %f13;
- @%p5 bra BB0_6;
- bra.uni BB0_7;
-
-BB0_6:
- st.shared.f32 [%r22+2048], %f13;
- ld.shared.u32 %r52, [%r22+3584];
- st.shared.u32 [%r22+3072], %r52;
-
-BB0_7:
- bar.sync 0;
- setp.lt.u32 %p6, %r2, 64;
- @%p6 bra BB0_8;
- bra.uni BB0_10;
-
-BB0_8:
- ld.shared.f32 %f162, [%r22];
- ld.shared.f32 %f163, [%r22+1024];
- ld.shared.f32 %f164, [%r22+1280];
- ld.shared.f32 %f165, [%r22+256];
- sub.ftz.f32 %f166, %f165, %f162;
- add.ftz.f32 %f167, %f162, %f165;
- div.rn.ftz.f32 %f168, %f167, 0f40000000;
- st.shared.f32 [%r22], %f168;
- add.ftz.f32 %f169, %f163, %f164;
- mul.ftz.f32 %f170, %f166, %f166;
- mul.ftz.f32 %f171, %f170, %f421;
- div.rn.ftz.f32 %f172, %f171, 0f40000000;
- add.ftz.f32 %f173, %f169, %f172;
- st.shared.f32 [%r22+1024], %f173;
- add.ftz.f32 %f421, %f421, %f421;
- ld.shared.f32 %f16, [%r22+2304];
- ld.shared.f32 %f174, [%r22+2048];
- setp.lt.ftz.f32 %p7, %f174, %f16;
- @%p7 bra BB0_9;
- bra.uni BB0_10;
-
-BB0_9:
- st.shared.f32 [%r22+2048], %f16;
- ld.shared.u32 %r53, [%r22+3328];
- st.shared.u32 [%r22+3072], %r53;
-
-BB0_10:
- bar.sync 0;
- setp.lt.u32 %p8, %r2, 32;
- @%p8 bra BB0_11;
- bra.uni BB0_13;
-
-BB0_11:
- ld.shared.f32 %f175, [%r22];
- ld.shared.f32 %f176, [%r22+1024];
- ld.shared.f32 %f177, [%r22+1152];
- ld.shared.f32 %f178, [%r22+128];
- sub.ftz.f32 %f179, %f178, %f175;
- add.ftz.f32 %f180, %f175, %f178;
- div.rn.ftz.f32 %f181, %f180, 0f40000000;
- st.shared.f32 [%r22], %f181;
- add.ftz.f32 %f182, %f176, %f177;
- mul.ftz.f32 %f183, %f179, %f179;
- mul.ftz.f32 %f184, %f183, %f421;
- div.rn.ftz.f32 %f185, %f184, 0f40000000;
- add.ftz.f32 %f186, %f182, %f185;
- st.shared.f32 [%r22+1024], %f186;
- add.ftz.f32 %f421, %f421, %f421;
- ld.shared.f32 %f19, [%r22+2176];
- ld.shared.f32 %f187, [%r22+2048];
- setp.lt.ftz.f32 %p9, %f187, %f19;
- @%p9 bra BB0_12;
- bra.uni BB0_13;
-
-BB0_12:
- st.shared.f32 [%r22+2048], %f19;
- ld.shared.u32 %r54, [%r22+3200];
- st.shared.u32 [%r22+3072], %r54;
-
-BB0_13:
- bar.sync 0;
- setp.lt.u32 %p10, %r2, 16;
- @%p10 bra BB0_14;
- bra.uni BB0_16;
-
-BB0_14:
- ld.shared.f32 %f188, [%r22];
- ld.shared.f32 %f189, [%r22+1024];
- ld.shared.f32 %f190, [%r22+1088];
- ld.shared.f32 %f191, [%r22+64];
- sub.ftz.f32 %f192, %f191, %f188;
- add.ftz.f32 %f193, %f188, %f191;
- div.rn.ftz.f32 %f194, %f193, 0f40000000;
- st.shared.f32 [%r22], %f194;
- add.ftz.f32 %f195, %f189, %f190;
- mul.ftz.f32 %f196, %f192, %f192;
- mul.ftz.f32 %f197, %f196, %f421;
- div.rn.ftz.f32 %f198, %f197, 0f40000000;
- add.ftz.f32 %f199, %f195, %f198;
- st.shared.f32 [%r22+1024], %f199;
- add.ftz.f32 %f421, %f421, %f421;
- ld.shared.f32 %f22, [%r22+2112];
- ld.shared.f32 %f200, [%r22+2048];
- setp.lt.ftz.f32 %p11, %f200, %f22;
- @%p11 bra BB0_15;
- bra.uni BB0_16;
-
-BB0_15:
- st.shared.f32 [%r22+2048], %f22;
- ld.shared.u32 %r55, [%r22+3136];
- st.shared.u32 [%r22+3072], %r55;
-
-BB0_16:
- bar.sync 0;
- setp.lt.u32 %p12, %r2, 8;
- @%p12 bra BB0_17;
- bra.uni BB0_19;
-
-BB0_17:
- ld.shared.f32 %f201, [%r22];
- ld.shared.f32 %f202, [%r22+1024];
- ld.shared.f32 %f203, [%r22+1056];
- ld.shared.f32 %f204, [%r22+32];
- sub.ftz.f32 %f205, %f204, %f201;
- add.ftz.f32 %f206, %f201, %f204;
- div.rn.ftz.f32 %f207, %f206, 0f40000000;
- st.shared.f32 [%r22], %f207;
- add.ftz.f32 %f208, %f202, %f203;
- mul.ftz.f32 %f209, %f205, %f205;
- mul.ftz.f32 %f210, %f209, %f421;
- div.rn.ftz.f32 %f211, %f210, 0f40000000;
- add.ftz.f32 %f212, %f208, %f211;
- st.shared.f32 [%r22+1024], %f212;
- add.ftz.f32 %f421, %f421, %f421;
- ld.shared.f32 %f25, [%r22+2080];
- ld.shared.f32 %f213, [%r22+2048];
- setp.lt.ftz.f32 %p13, %f213, %f25;
- @%p13 bra BB0_18;
- bra.uni BB0_19;
-
-BB0_18:
- st.shared.f32 [%r22+2048], %f25;
- ld.shared.u32 %r56, [%r22+3104];
- st.shared.u32 [%r22+3072], %r56;
-
-BB0_19:
- bar.sync 0;
- setp.lt.u32 %p14, %r2, 4;
- @%p14 bra BB0_20;
- bra.uni BB0_22;
-
-BB0_20:
- ld.shared.f32 %f214, [%r22];
- ld.shared.f32 %f215, [%r22+1024];
- ld.shared.f32 %f216, [%r22+1040];
- ld.shared.f32 %f217, [%r22+16];
- sub.ftz.f32 %f218, %f217, %f214;
- add.ftz.f32 %f219, %f214, %f217;
- div.rn.ftz.f32 %f220, %f219, 0f40000000;
- st.shared.f32 [%r22], %f220;
- add.ftz.f32 %f221, %f215, %f216;
- mul.ftz.f32 %f222, %f218, %f218;
- mul.ftz.f32 %f223, %f222, %f421;
- div.rn.ftz.f32 %f224, %f223, 0f40000000;
- add.ftz.f32 %f225, %f221, %f224;
- st.shared.f32 [%r22+1024], %f225;
- add.ftz.f32 %f421, %f421, %f421;
- ld.shared.f32 %f28, [%r22+2064];
- ld.shared.f32 %f226, [%r22+2048];
- setp.lt.ftz.f32 %p15, %f226, %f28;
- @%p15 bra BB0_21;
- bra.uni BB0_22;
-
-BB0_21:
- st.shared.f32 [%r22+2048], %f28;
- ld.shared.u32 %r57, [%r22+3088];
- st.shared.u32 [%r22+3072], %r57;
-
-BB0_22:
- bar.sync 0;
- setp.lt.u32 %p16, %r2, 2;
- @%p16 bra BB0_23;
- bra.uni BB0_25;
-
-BB0_23:
- ld.shared.f32 %f227, [%r22];
- ld.shared.f32 %f228, [%r22+1024];
- ld.shared.f32 %f229, [%r22+1032];
- ld.shared.f32 %f230, [%r22+8];
- sub.ftz.f32 %f231, %f230, %f227;
- add.ftz.f32 %f232, %f227, %f230;
- div.rn.ftz.f32 %f233, %f232, 0f40000000;
- st.shared.f32 [%r22], %f233;
- add.ftz.f32 %f234, %f228, %f229;
- mul.ftz.f32 %f235, %f231, %f231;
- mul.ftz.f32 %f236, %f235, %f421;
- div.rn.ftz.f32 %f237, %f236, 0f40000000;
- add.ftz.f32 %f238, %f234, %f237;
- st.shared.f32 [%r22+1024], %f238;
- add.ftz.f32 %f421, %f421, %f421;
- ld.shared.f32 %f31, [%r22+2056];
- ld.shared.f32 %f239, [%r22+2048];
- setp.lt.ftz.f32 %p17, %f239, %f31;
- @%p17 bra BB0_24;
- bra.uni BB0_25;
-
-BB0_24:
- st.shared.f32 [%r22+2048], %f31;
- ld.shared.u32 %r58, [%r22+3080];
- st.shared.u32 [%r22+3072], %r58;
-
-BB0_25:
- bar.sync 0;
- setp.eq.s32 %p18, %r2, 0;
- @%p18 bra BB0_26;
- bra.uni BB0_28;
-
-BB0_26:
- ld.shared.f32 %f240, [%r22];
- ld.shared.f32 %f241, [%r22+1024];
- ld.shared.f32 %f242, [%r22+1028];
- ld.shared.f32 %f243, [%r22+4];
- sub.ftz.f32 %f244, %f243, %f240;
- add.ftz.f32 %f245, %f240, %f243;
- div.rn.ftz.f32 %f246, %f245, 0f40000000;
- st.shared.f32 [%r22], %f246;
- add.ftz.f32 %f247, %f241, %f242;
- mul.ftz.f32 %f248, %f244, %f244;
- mul.ftz.f32 %f249, %f248, %f421;
- div.rn.ftz.f32 %f250, %f249, 0f40000000;
- add.ftz.f32 %f251, %f247, %f250;
- st.shared.f32 [%r22+1024], %f251;
- add.ftz.f32 %f421, %f421, %f421;
- ld.shared.f32 %f34, [%r22+2052];
- ld.shared.f32 %f252, [%r22+2048];
- setp.lt.ftz.f32 %p19, %f252, %f34;
- @%p19 bra BB0_27;
- bra.uni BB0_28;
-
-BB0_27:
- st.shared.f32 [%r22+2048], %f34;
- ld.shared.u32 %r59, [%r22+3076];
- st.shared.u32 [%r22+3072], %r59;
-
-BB0_28:
- bar.sync 0;
- @%p18 bra BB0_30;
-
- ret;
-
-BB0_30:
- shl.b32 %r60, %r3, 4;
- ld.param.u32 %r63, [trigger_param_0];
- add.s32 %r61, %r63, %r60;
- ld.shared.f32 %f253, [shr_4_tmp];
- st.global.f32 [%r61], %f253;
- add.ftz.f32 %f254, %f421, 0fBF800000;
- ld.shared.f32 %f255, [shr_4_tmp+1024];
- div.approx.ftz.f32 %f256, %f255, %f254;
- st.global.f32 [%r61+4], %f256;
- ld.shared.f32 %f257, [shr_4_tmp+2048];
- st.global.f32 [%r61+8], %f257;
- ld.shared.u32 %r62, [shr_4_tmp+3072];
- st.global.u32 [%r61+12], %r62;
- ret;
-}
-
-
diff --git a/RTCP/GPUProc/src/octave-core b/RTCP/GPUProc/src/octave-core
deleted file mode 100644
index 608aba01c896492a0c4befa1f98dc1416cc6cf6c..0000000000000000000000000000000000000000
Binary files a/RTCP/GPUProc/src/octave-core and /dev/null differ