Rotate bug

rotate_bug/JULIA_ONEMKL_RUN.txt

@@ -0,0 +1,55 @@
+kali@sdp:~/Kali/2023/2705_rotbug/oneAPI.jl$ $JULIA --project -L test/setup.jl test/rotbug.jl
+x = ComplexF64[1.0 + 0.0im]
+y = ComplexF64[1.0 + 0.0im]
+x = ComplexF64[1.3928964429210167 + 0.0im]
+Array(d_x) = ComplexF64[0.9592049927584356 + 0.0im]
+c = 0.8187587885612081
+s = 0.5741376543598085 + 0.0im
+ERROR: LoadError: AssertionError: x == Array(d_x)
+Stacktrace:
+ [1] main(; n::Int64, T::Type{ComplexF64})
+   @ Main ~/Kali/2023/2705_rotbug/oneAPI.jl/test/rotbug.jl:36
+ [2] main()
+   @ Main ~/Kali/2023/2705_rotbug/oneAPI.jl/test/rotbug.jl:4
+ [3] top-level scope
+   @ ~/Kali/2023/2705_rotbug/oneAPI.jl/test/rotbug.jl:39
+in expression starting at /home/kali/Kali/2023/2705_rotbug/oneAPI.jl/test/rotbug.jl:39
+kali@sdp:~/Kali/2023/2705_rotbug/oneAPI.jl$ vim test/rotbug.jl
+kali@sdp:~/Kali/2023/2705_rotbug/oneAPI.jl$ $JULIA --project -L test/setup.jl test/rotbug.jl
+x = ComplexF64[1.0 + 0.0im]
+y = ComplexF64[1.0 + 0.0im]
+x = ComplexF64[1.3928964429210167 + 0.0im]
+Array(d_x) = ComplexF64[1.3928964429210167 + 0.0im]
+c = 0.8187587885612081
+s = 0.5741376543598085 + 0.0im
+kali@sdp:~/Kali/2023/2705_rotbug/oneAPI.jl$ $JULIA --project -L test/setup.jl test/rotbug.jl
+x = ComplexF64[1.0 + 0.0im]
+y = ComplexF64[1.0 + 0.0im]
+x = ComplexF64[1.3928964429210167 + 0.0im]
+Array(d_x) = ComplexF64[0.9592049927584356 + 0.0im]
+c = 0.8187587885612081
+s = 0.5741376543598085 + 0.0im
+ERROR: LoadError: AssertionError: x == Array(d_x)
+Stacktrace:
+ [1] main(; n::Int64, T::Type{ComplexF64})
+   @ Main ~/Kali/2023/2705_rotbug/oneAPI.jl/test/rotbug.jl:37
+ [2] main()
+   @ Main ~/Kali/2023/2705_rotbug/oneAPI.jl/test/rotbug.jl:4
+ [3] top-level scope
+   @ ~/Kali/2023/2705_rotbug/oneAPI.jl/test/rotbug.jl:40
+in expression starting at /home/kali/Kali/2023/2705_rotbug/oneAPI.jl/test/rotbug.jl:40
+kali@sdp:~/Kali/2023/2705_rotbug/oneAPI.jl$ $JULIA --project -L test/setup.jl test/rotbug.jl
+x = ComplexF64[1.0 + 0.0im]
+y = ComplexF64[1.0 + 0.0im]
+x = ComplexF64[1.3928964429210167 + 0.0im]
+Array(d_x) = ComplexF64[1.3928964429210167 + 0.0im]
+c = 0.8187587885612081
+s = 0.5741376543598085 + 0.0im
+kali@sdp:~/Kali/2023/2705_rotbug/oneAPI.jl$ $JULIA --project -L test/setup.jl test/rotbug.jl
+x = ComplexF64[1.0 + 0.0im]
+y = ComplexF64[1.0 + 0.0im]
+x = ComplexF64[1.3928964429210167 + 0.0im]
+Array(d_x) = ComplexF64[1.3928964429210167 + 0.0im]
+c = 0.8187587885612081
+s = 0.5741376543598085 + 0.0im
+kali@sdp:~/Kali/2023/2705_rotbug/oneAPI.jl$

rotate_bug/README.log

@@ -0,0 +1,3 @@
+Compile Command: dpcpp -g -fsycl -I/opt/intel/oneapi/mkl/2023.1.0/include/ -L/opt/intel/oneapi/mkl/2023.1.0/lib/ -lmkl_intel_lp64 -lmkl_sequential -lmkl_core -lpthread -lm -ldl -lmkl_sycl rotate_gpu_usm.cpp -o rotgpu_usm.exe   
+
+Run Julia: $JULIA --project -L test/setup.jl test/rotbug.jl

rotate_bug/SYCL_ONEMKL_RUN.txt

@@ -0,0 +1,121 @@
+kali@sdp:~/Kali/2023/2705_rotbug/onemkl_reproducer/2023.1.0/examples/dpcpp/blas/source$ ./rotate.exe 
+c: 0.818759s: (0.574138,0)
+(1,0)
+GPU Results: 
+1
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+Done with GPU Rotate !!, Starting CPU rotate
+CPU Results: 
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+kali@sdp:~/Kali/2023/2705_rotbug/onemkl_reproducer/2023.1.0/examples/dpcpp/blas/source$ ./rotate.exe 
+c: 0.818759s: (0.574138,0)
+(1,0)
+GPU Results: 
+1
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+Done with GPU Rotate !!, Starting CPU rotate
+CPU Results: 
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+kali@sdp:~/Kali/2023/2705_rotbug/onemkl_reproducer/2023.1.0/examples/dpcpp/blas/source$ ./rotate.exe 
+c: 0.818759s: (0.574138,0)
+(1,0)
+GPU Results: 
+1
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+Done with GPU Rotate !!, Starting CPU rotate
+CPU Results: 
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+kali@sdp:~/Kali/2023/2705_rotbug/onemkl_reproducer/2023.1.0/examples/dpcpp/blas/source$ ./rotate.exe 
+c: 0.818759s: (0.574138,0)
+(1,0)
+GPU Results: 
+1
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+Done with GPU Rotate !!, Starting CPU rotate
+CPU Results: 
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+kali@sdp:~/Kali/2023/2705_rotbug/onemkl_reproducer/2023.1.0/examples/dpcpp/blas/source$ ./rotate.exe 
+c: 0.818759s: (0.574138,0)
+(1,0)
+GPU Results: 
+1
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+Done with GPU Rotate !!, Starting CPU rotate
+CPU Results: 
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+kali@sdp:~/Kali/2023/2705_rotbug/onemkl_reproducer/2023.1.0/examples/dpcpp/blas/source$ ./rotate.exe 
+c: 0.818759s: (0.574138,0)
+(1,0)
+GPU Results: 
+1
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+Done with GPU Rotate !!, Starting CPU rotate
+CPU Results: 
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+kali@sdp:~/Kali/2023/2705_rotbug/onemkl_reproducer/2023.1.0/examples/dpcpp/blas/source$ ./rotate.exe 
+c: 0.818759s: (0.574138,0)
+(1,0)
+GPU Results: 
+1
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+Done with GPU Rotate !!, Starting CPU rotate
+CPU Results: 
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+kali@sdp:~/Kali/2023/2705_rotbug/onemkl_reproducer/2023.1.0/examples/dpcpp/blas/source$ ./rotate.exe 
+c: 0.818759s: (0.574138,0)
+(1,0)
+GPU Results: 
+1
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+Done with GPU Rotate !!, Starting CPU rotate
+CPU Results: 
+X - array
+(1.3929,0) 
+Y - array
+(0.244621,0) 
+kali@sdp:~/Kali/2023/2705_rotbug/onemkl_reproducer/2023.1.0/examples/dpcpp/blas/source$ 

rotate_bug/rotate_gpu_usm.cpp

@@ -0,0 +1,96 @@
+#include <CL/sycl.hpp>
+#include <oneapi/mkl/blas.hpp>
+#include <complex>
+#include <mkl.h>
+using namespace sycl;
+
+int main() {
+//    double c = 0.5550150321048569;
+//    std::complex<double> s(0.83184031769183,0.0); 
+//      double c = 0.40516847542537016;
+//      std::complex<double> s(0.9142420393536284,0.0);
+
+	// Occasional failure
+      double c = 0.8187587885612081;
+      std::complex<double> s(0.5741376543598085,0.0);
+    std::cout << "c: " << c << "s: " << s << std::endl; 
+    try {
+        // Create a SYCL queue
+    cl::sycl::queue main_queue(cl::sycl::gpu_selector{});
+	auto cxt = main_queue.get_context();
+	auto dev = main_queue.get_device();
+#if 0
+	auto ua = usm_allocator<std::complex<double>, usm::alloc::shared, 64>(cxt, dev);
+        std::vector<std::complex<double>, decltype(ua)> x(ua), y(ua);
+	for (int i = 0; i < 1; i++) {
+		x.push_back({1.0, 0.0});
+		y.push_back({1.0, 0.0});
+	}
+#endif
+	std::complex<double> *x = (std::complex<double> *) malloc_shared(10 * sizeof(std::complex<double>), dev, cxt); 
+	std::complex<double> *y = (std::complex<double> *) malloc_shared(10 * sizeof(std::complex<double>), dev, cxt); 
+
+	for (int i = 0; i < 1; i++) {
+		x[i] = {1,0};
+		y[i] = {1,0};
+	}
+
+	for (int i = 0; i < 1; i++) {
+		std::cout << x[i] << std::endl;
+	}
+
+	// Perform the Givens rotation on the vectors
+        auto status = oneapi::mkl::blas::column_major::rot(main_queue, 1, x, 1, 
+							y, 1, c, s);
+	sycl::get_native<sycl::backend::ext_oneapi_level_zero>(status);
+	main_queue.wait_and_throw();
+	std::cout << "GPU Results: "<< std::endl; 
+	    int n = 1;
+	    std::cout << n << std::endl;
+	    std::cout << "X - array\n";
+	    for(int i = 0; i < n; i++) {
+		std::cout << x[i] << " ";
+	    }
+	    std::cout << std::endl;
+
+	    std::cout << "Y - array\n";
+	    for(int i = 0; i < n; i++) {
+		std::cout << y[i] << " ";
+	    }
+	    std::cout << std::endl;
+
+
+    } catch (cl::sycl::exception& e) {
+        std::cout << "SYCL exception encountered: " << e.what() << std::endl;
+        return 1;
+    }
+
+    std::cout << "Done with GPU Rotate !!, Starting CPU rotate" << std::endl;
+
+    std::vector<std::complex<double> > x;
+    std::vector<std::complex<double> >y;
+    for (int i = 0; i < 1; i++) {
+	x.push_back({1.0,0.0});
+	y.push_back({1.0,0.0});
+    }
+    int n = sizeof(x) / sizeof(x[0]);
+
+    cblas_zrot(n, x.data(), 1, y.data(), 1, c, &s);
+    std::cout << "CPU Results: "<< std::endl; 
+
+    std::cout << "X - array\n";
+    for(int i = 0; i < n; i++) {
+    	std::cout << x[i] << " ";
+    }
+    std::cout << std::endl;
+
+    std::cout << "Y - array\n";
+    for(int i = 0; i < n; i++) {
+    	std::cout << y[i] << " ";
+    }
+    std::cout << std::endl;
+
+
+    return 0;
+}
+

rotate_bug/rotbug.jl

@@ -0,0 +1,40 @@
+using oneAPI
+using MKL
+using LinearAlgebra
+function main(; n = 1, T = ComplexF64)
+    x = ones(T, n)
+    y = ones(T, n)
+    #c = rand(real(T))
+    #s = T(sqrt(1 - c^2)) 
+    c = 0.8187587885612081
+    s = T(sqrt(1-c^2))
+    @show x
+    @show y
+
+    @test c*c + s*conj(s) ≈ 1
+    d_x = oneArray(x)
+    d_y = oneArray(y)
+
+    incx = stride(x, 1)
+    incy = stride(y, 1)
+    BLAS.rot!(n, x, incx, y, incy, c, s)
+
+    queue = global_queue(context(d_x), device(d_x))
+    incx = stride(d_x, 1)
+    incy = stride(d_y, 1)
+    @assert length(d_x) >= 1 + (n - 1)*abs(incx)
+    @assert length(d_y) >= 1 + (n - 1)*abs(incy)
+
+    (T == Float32)    && oneMKL.onemklSrot(sycl_queue(queue), n, d_x, incx, d_y, incy, c, s)
+    (T == Float64)    && oneMKL.onemklDrot(sycl_queue(queue), n, d_x, incx, d_y, incy, c, s)
+    (T == ComplexF32) && oneMKL.onemklCrot(sycl_queue(queue), n, d_x, incx, d_y, incy, c, s)
+    (T == ComplexF64) && oneMKL.onemklZrot(sycl_queue(queue), n, d_x, incx, d_y, incy, c, s)
+
+    @show x
+    @show Array(d_x)
+    @show c
+    @show s
+    @assert x == Array(d_x)
+end
+
+main()