diff options
Diffstat (limited to 'camera.py')
| -rwxr-xr-x | camera.py | 34 |
1 files changed, 17 insertions, 17 deletions
@@ -136,8 +136,6 @@ class Camera(Thread): self.gpu = GPU(self.device_id) self.gpu.load_geometry(geometry) - self.kernels = CUDAFuncs(self.gpu.module, ['ray_trace', 'ray_trace_alpha', 'rotate', 'rotate_around_point', 'translate', 'update_xyz_lookup', 'update_xyz_image', 'process_image', 'init_rng']) - self.width, self.height = size pygame.init() @@ -180,7 +178,7 @@ class Camera(Thread): @timeit def initialize_render(self): self.rng_states_gpu = cuda.mem_alloc(self.width*self.height*sizeof('curandStateXORWOW', '#include <curand_kernel.h>')) - self.kernels.init_rng(np.int32(self.width*self.height), self.rng_states_gpu, np.int32(0), np.int32(0), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) + self.gpu.kernels.init_rng(np.int32(self.width*self.height), self.rng_states_gpu, np.int32(0), np.int32(0), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) self.xyz_lookup1_gpu = gpuarray.zeros(len(self.geometry.mesh.triangles), dtype=gpuarray.vec.float3) self.xyz_lookup2_gpu = gpuarray.zeros(len(self.geometry.mesh.triangles), dtype=gpuarray.vec.float3) self.image_gpu = gpuarray.zeros(self.width*self.height, dtype=gpuarray.vec.float3) @@ -200,13 +198,13 @@ class Camera(Thread): def update_xyz_lookup(self, source_position): for i in range(self.xyz_lookup1_gpu.size//(self.width*self.height)+1): - self.kernels.update_xyz_lookup(np.int32(self.width*self.height), np.int32(self.xyz_lookup1_gpu.size), np.int32(i*self.width*self.height), gpuarray.vec.make_float3(*source_position), self.rng_states_gpu, np.float32(685.0), gpuarray.vec.make_float3(1.0,0.0,0.0), self.xyz_lookup1_gpu, self.xyz_lookup2_gpu, np.int32(self.max_steps), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) + self.gpu.kernels.update_xyz_lookup(np.int32(self.width*self.height), np.int32(self.xyz_lookup1_gpu.size), np.int32(i*self.width*self.height), gpuarray.vec.make_float3(*source_position), self.rng_states_gpu, np.float32(685.0), gpuarray.vec.make_float3(1.0,0.0,0.0), self.xyz_lookup1_gpu, self.xyz_lookup2_gpu, np.int32(self.max_steps), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) for i in range(self.xyz_lookup1_gpu.size//(self.width*self.height)+1): - self.kernels.update_xyz_lookup(np.int32(self.width*self.height), np.int32(self.xyz_lookup1_gpu.size), np.int32(i*self.width*self.height), gpuarray.vec.make_float3(*source_position), self.rng_states_gpu, np.float32(545.0), gpuarray.vec.make_float3(0.0,1.0,0.0), self.xyz_lookup1_gpu, self.xyz_lookup2_gpu, np.int32(self.max_steps), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) + self.gpu.kernels.update_xyz_lookup(np.int32(self.width*self.height), np.int32(self.xyz_lookup1_gpu.size), np.int32(i*self.width*self.height), gpuarray.vec.make_float3(*source_position), self.rng_states_gpu, np.float32(545.0), gpuarray.vec.make_float3(0.0,1.0,0.0), self.xyz_lookup1_gpu, self.xyz_lookup2_gpu, np.int32(self.max_steps), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) for i in range(self.xyz_lookup1_gpu.size//(self.width*self.height)+1): - self.kernels.update_xyz_lookup(np.int32(self.width*self.height), np.int32(self.xyz_lookup1_gpu.size), np.int32(i*self.width*self.height), gpuarray.vec.make_float3(*source_position), self.rng_states_gpu, np.float32(445.0), gpuarray.vec.make_float3(0.0,0.0,1.0), self.xyz_lookup1_gpu, self.xyz_lookup2_gpu, np.int32(self.max_steps), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) + self.gpu.kernels.update_xyz_lookup(np.int32(self.width*self.height), np.int32(self.xyz_lookup1_gpu.size), np.int32(i*self.width*self.height), gpuarray.vec.make_float3(*source_position), self.rng_states_gpu, np.float32(445.0), gpuarray.vec.make_float3(0.0,0.0,1.0), self.xyz_lookup1_gpu, self.xyz_lookup2_gpu, np.int32(self.max_steps), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) self.nlookup_calls += 1 @@ -216,16 +214,16 @@ class Camera(Thread): self.nimages = 0 def update_image(self): - self.kernels.update_xyz_image(np.int32(self.width*self.height), self.rng_states_gpu, self.origins_gpu, self.directions_gpu, np.float32(685.0), gpuarray.vec.make_float3(1.0,0.0,0.0), self.xyz_lookup1_gpu, self.xyz_lookup2_gpu, self.image_gpu, np.int32(self.nlookup_calls), np.int32(self.max_steps), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) + self.gpu.kernels.update_xyz_image(np.int32(self.width*self.height), self.rng_states_gpu, self.origins_gpu, self.directions_gpu, np.float32(685.0), gpuarray.vec.make_float3(1.0,0.0,0.0), self.xyz_lookup1_gpu, self.xyz_lookup2_gpu, self.image_gpu, np.int32(self.nlookup_calls), np.int32(self.max_steps), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) - self.kernels.update_xyz_image(np.int32(self.width*self.height), self.rng_states_gpu, self.origins_gpu, self.directions_gpu, np.float32(545.0), gpuarray.vec.make_float3(0.0,1.0,0.0), self.xyz_lookup1_gpu, self.xyz_lookup2_gpu, self.image_gpu, np.int32(self.nlookup_calls), np.int32(self.max_steps), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) + self.gpu.kernels.update_xyz_image(np.int32(self.width*self.height), self.rng_states_gpu, self.origins_gpu, self.directions_gpu, np.float32(545.0), gpuarray.vec.make_float3(0.0,1.0,0.0), self.xyz_lookup1_gpu, self.xyz_lookup2_gpu, self.image_gpu, np.int32(self.nlookup_calls), np.int32(self.max_steps), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) - self.kernels.update_xyz_image(np.int32(self.width*self.height), self.rng_states_gpu, self.origins_gpu, self.directions_gpu, np.float32(445.0), gpuarray.vec.make_float3(0.0,0.0,1.0), self.xyz_lookup1_gpu, self.xyz_lookup2_gpu, self.image_gpu, np.int32(self.nlookup_calls), np.int32(self.max_steps), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) + self.gpu.kernels.update_xyz_image(np.int32(self.width*self.height), self.rng_states_gpu, self.origins_gpu, self.directions_gpu, np.float32(445.0), gpuarray.vec.make_float3(0.0,0.0,1.0), self.xyz_lookup1_gpu, self.xyz_lookup2_gpu, self.image_gpu, np.int32(self.nlookup_calls), np.int32(self.max_steps), block=(self.nblocks,1,1), grid=(self.width*self.height//self.nblocks+1,1)) self.nimages += 1 def process_image(self): - self.kernels.process_image(np.int32(self.width*self.height), self.image_gpu, self.pixels_gpu, np.int32(self.nimages), block=(self.nblocks,1,1), grid=((self.width*self.height)//self.nblocks+1,1)) + self.gpu.kernels.process_image(np.int32(self.width*self.height), self.image_gpu, self.pixels_gpu, np.int32(self.nimages), block=(self.nblocks,1,1), grid=((self.width*self.height)//self.nblocks+1,1)) def screenshot(self, dir='', start=0): root, ext = 'screenshot', 'png' @@ -240,8 +238,8 @@ class Camera(Thread): print 'image saved to %s' % filename def rotate(self, phi, n): - self.kernels.rotate(np.int32(self.pixels_gpu.size), self.origins_gpu, np.float32(phi), gpuarray.vec.make_float3(*n), block=(self.nblocks,1,1), grid=(self.pixels_gpu.size//self.nblocks+1,1)) - self.kernels.rotate(np.int32(self.pixels_gpu.size), self.directions_gpu, np.float32(phi), gpuarray.vec.make_float3(*n), block=(self.nblocks,1,1), grid=(self.pixels_gpu.size//self.nblocks+1,1)) + self.gpu.kernels.rotate(np.int32(self.pixels_gpu.size), self.origins_gpu, np.float32(phi), gpuarray.vec.make_float3(*n), block=(self.nblocks,1,1), grid=(self.pixels_gpu.size//self.nblocks+1,1)) + self.gpu.kernels.rotate(np.int32(self.pixels_gpu.size), self.directions_gpu, np.float32(phi), gpuarray.vec.make_float3(*n), block=(self.nblocks,1,1), grid=(self.pixels_gpu.size//self.nblocks+1,1)) self.point = rotate(self.point, phi, n) self.axis1 = rotate(self.axis1, phi, n) @@ -253,8 +251,8 @@ class Camera(Thread): self.update() def rotate_around_point(self, phi, n, point, redraw=True): - self.kernels.rotate_around_point(np.int32(self.origins_gpu.size), self.origins_gpu, np.float32(phi), gpuarray.vec.make_float3(*n), gpuarray.vec.make_float3(*point), block=(self.nblocks,1,1), grid=(self.origins_gpu.size//self.nblocks+1,1)) - self.kernels.rotate(np.int32(self.directions_gpu.size), self.directions_gpu, np.float32(phi), gpuarray.vec.make_float3(*n), block=(self.nblocks,1,1), grid=(self.directions_gpu.size//self.nblocks+1,1)) + self.gpu.kernels.rotate_around_point(np.int32(self.origins_gpu.size), self.origins_gpu, np.float32(phi), gpuarray.vec.make_float3(*n), gpuarray.vec.make_float3(*point), block=(self.nblocks,1,1), grid=(self.origins_gpu.size//self.nblocks+1,1)) + self.gpu.kernels.rotate(np.int32(self.directions_gpu.size), self.directions_gpu, np.float32(phi), gpuarray.vec.make_float3(*n), block=(self.nblocks,1,1), grid=(self.directions_gpu.size//self.nblocks+1,1)) self.axis1 = rotate(self.axis1, phi, n) self.axis2 = rotate(self.axis2, phi, n) @@ -266,7 +264,7 @@ class Camera(Thread): self.update() def translate(self, v, redraw=True): - self.kernels.translate(np.int32(self.pixels_gpu.size), self.origins_gpu, gpuarray.vec.make_float3(*v), block=(self.nblocks,1,1), grid=(self.pixels_gpu.size//self.nblocks,1)) + self.gpu.kernels.translate(np.int32(self.pixels_gpu.size), self.origins_gpu, gpuarray.vec.make_float3(*v), block=(self.nblocks,1,1), grid=(self.pixels_gpu.size//self.nblocks,1)) self.point += v @@ -284,9 +282,9 @@ class Camera(Thread): self.process_image() else: if self.alpha: - self.kernels.ray_trace_alpha(np.int32(self.pixels_gpu.size), self.origins_gpu, self.directions_gpu, self.pixels_gpu, block=(self.nblocks,1,1), grid=(self.pixels_gpu.size//self.nblocks+1,1)) + self.gpu.kernels.ray_trace_alpha(np.int32(self.pixels_gpu.size), self.origins_gpu, self.directions_gpu, self.pixels_gpu, block=(self.nblocks,1,1), grid=(self.pixels_gpu.size//self.nblocks+1,1)) else: - self.kernels.ray_trace(np.int32(self.pixels_gpu.size), self.origins_gpu, self.directions_gpu, self.pixels_gpu, block=(self.nblocks,1,1), grid=(self.pixels_gpu.size//self.nblocks+1,1)) + self.gpu.kernels.ray_trace(np.int32(self.pixels_gpu.size), self.origins_gpu, self.directions_gpu, self.pixels_gpu, block=(self.nblocks,1,1), grid=(self.pixels_gpu.size//self.nblocks+1,1)) pygame.surfarray.blit_array(self.screen, self.pixels_gpu.get().reshape(self.size)) if self.doom_mode: @@ -624,4 +622,6 @@ if __name__ == '__main__': camera = EventViewer(geometry, options.io_file, size=size) else: camera = Camera(geometry, size) + camera.start() + camera.join() |