diff options
-rw-r--r-- | data/diffraction.cl | 27 | ||||
-rw-r--r-- | src/diffraction-gpu.c | 24 | ||||
-rw-r--r-- | tests/gpu_sim_check.c | 36 |
3 files changed, 46 insertions, 41 deletions
diff --git a/data/diffraction.cl b/data/diffraction.cl index 994e27db..3b0779b5 100644 --- a/data/diffraction.cl +++ b/data/diffraction.cl @@ -51,29 +51,27 @@ const sampler_t sampler_c = CLK_NORMALIZED_COORDS_TRUE float4 get_q(float fs, float ss, float res, float clen, float k, float corner_x, float corner_y, - float fsx, float fsy, float ssx, float ssy) + float fsx, float fsy, float fsz, float ssx, float ssy, float ssz) { - float rx, ry, r; + float rx, ry, rz; float az, tt; float4 q; float xs, ys; float kx, ky, kz; + float ctt; - xs = fs*fsx + ss*ssx; - ys = fs*fsy + ss*ssy; - - rx = (xs + corner_x) / res; - ry = (ys + corner_y) / res; - - r = sqrt(pow(rx, 2.0f) + pow(ry, 2.0f)); - - tt = atan2(r, clen); + /* Calculate 3D position of given position, in m */ + rx = (corner_x + fs*fsx + ss*ssx) / res; + ry = (corner_y + fs*fsy + ss*ssy) / res; + rz = clen + (fs*fsz + ss*ssz)/res; + ctt = rz / sqrt(rx*rx + ry*ry + rz*rz); /* cos(2theta) */ + tt = acos(ctt); az = atan2(ry, rx); kx = k*native_sin(tt)*native_cos(az); ky = k*native_sin(tt)*native_sin(az); - kz = k*(native_cos(tt)-1.0); + kz = k*(ctt - 1.0); q = (float4)(kx, ky, kz, 0.0); @@ -159,7 +157,8 @@ float molecule_factor(global float *intensities, global float *flags, kernel void diffraction(global float *diff, float k, float weight, int w, float corner_x, float corner_y, - float fsx, float fsy, float ssx, float ssy, + float fsx, float fsy, float fsz, + float ssx, float ssy, float ssz, float res, float clen, float16 cell, global float *intensities, global float *flags, read_only image2d_t func_a, @@ -183,7 +182,7 @@ kernel void diffraction(global float *diff, float k, float weight, /* Get the scattering vector */ q = get_q(fs, ss, res, clen, k, - corner_x, corner_y, fsx, fsy, ssx, ssy); + corner_x, corner_y, fsx, fsy, fsz, ssx, ssy, ssz); /* Calculate the diffraction */ f_lattice = lattice_factor(cell, q, func_a, func_b, func_c); diff --git a/src/diffraction-gpu.c b/src/diffraction-gpu.c index 70f3ff58..331170ae 100644 --- a/src/diffraction-gpu.c +++ b/src/diffraction-gpu.c @@ -212,10 +212,12 @@ static int do_panels(struct gpu_context *gctx, struct image *image, if ( set_arg_float(gctx, 5, p->cny) ) return 1; if ( set_arg_float(gctx, 6, p->fsx) ) return 1; if ( set_arg_float(gctx, 7, p->fsy) ) return 1; - if ( set_arg_float(gctx, 8, p->ssx) ) return 1; - if ( set_arg_float(gctx, 9, p->ssy) ) return 1; - if ( set_arg_float(gctx, 10, p->res) ) return 1; - if ( set_arg_float(gctx, 11, p->clen) ) return 1; + if ( set_arg_float(gctx, 8, p->fsz) ) return 1; + if ( set_arg_float(gctx, 9, p->ssx) ) return 1; + if ( set_arg_float(gctx, 10, p->ssy) ) return 1; + if ( set_arg_float(gctx, 11, p->ssz) ) return 1; + if ( set_arg_float(gctx, 12, p->res) ) return 1; + if ( set_arg_float(gctx, 13, p->clen) ) return 1; dims[0] = p->w * sampling; dims[1] = p->h * sampling; @@ -223,7 +225,7 @@ static int do_panels(struct gpu_context *gctx, struct image *image, ldims[0] = sampling; ldims[1] = sampling; - err = clSetKernelArg(gctx->kern, 18, + err = clSetKernelArg(gctx->kern, 20, sampling*sampling*sizeof(cl_float), NULL); if ( err != CL_SUCCESS ) { ERROR("Couldn't set local memory: %s\n", clError(err)); @@ -305,17 +307,17 @@ int get_diffraction_gpu(struct gpu_context *gctx, struct image *image, cell.s[3] = bx; cell.s[4] = by; cell.s[5] = bz; cell.s[6] = cx; cell.s[7] = cy; cell.s[8] = cz; - err = clSetKernelArg(gctx->kern, 12, sizeof(cl_float16), &cell); + err = clSetKernelArg(gctx->kern, 14, sizeof(cl_float16), &cell); if ( err != CL_SUCCESS ) { ERROR("Couldn't set unit cell: %s\n", clError(err)); return 1; } - if ( set_arg_mem(gctx, 13, gctx->intensities) ) return 1; - if ( set_arg_mem(gctx, 14, gctx->flags) ) return 1; - if ( set_arg_mem(gctx, 15, gctx->sinc_luts[na-1]) ) return 1; - if ( set_arg_mem(gctx, 16, gctx->sinc_luts[nb-1]) ) return 1; - if ( set_arg_mem(gctx, 17, gctx->sinc_luts[nc-1]) ) return 1; + if ( set_arg_mem(gctx, 15, gctx->intensities) ) return 1; + if ( set_arg_mem(gctx, 16, gctx->flags) ) return 1; + if ( set_arg_mem(gctx, 17, gctx->sinc_luts[na-1]) ) return 1; + if ( set_arg_mem(gctx, 18, gctx->sinc_luts[nb-1]) ) return 1; + if ( set_arg_mem(gctx, 19, gctx->sinc_luts[nc-1]) ) return 1; /* Allocate memory for the result */ image->dp = malloc(image->det->n_panels * sizeof(float *)); diff --git a/tests/gpu_sim_check.c b/tests/gpu_sim_check.c index c471fb0f..ba51b8db 100644 --- a/tests/gpu_sim_check.c +++ b/tests/gpu_sim_check.c @@ -113,14 +113,16 @@ int main(int argc, char *argv[]) det->panels[0].orig_max_ss = 511; det->panels[0].w = 1024; det->panels[0].h = 512; - det->panels[0].fsx = 1; - det->panels[0].fsy = 0; - det->panels[0].ssx = 0; - det->panels[0].ssy = 1; - det->panels[0].xfs = 1; - det->panels[0].yfs = 0; - det->panels[0].xss = 0; - det->panels[0].yss = 1; + det->panels[0].fsx = 1.0; + det->panels[0].fsy = 0.0; + det->panels[0].fsz = 0.4; + det->panels[0].ssx = 0.0; + det->panels[0].ssy = 1.0; + det->panels[0].ssz = 0.0; + det->panels[0].xfs = 1.0; + det->panels[0].yfs = 0.0; + det->panels[0].xss = 0.0; + det->panels[0].yss = 1.0; det->panels[0].cnx = -512.0; det->panels[0].cny = -512.0-sep; det->panels[0].clen = 100.0e-3; @@ -134,14 +136,16 @@ int main(int argc, char *argv[]) det->panels[1].orig_max_ss = 1023; det->panels[1].w = 1024; det->panels[1].h = 512; - det->panels[1].fsx = 1; - det->panels[1].fsy = 0; - det->panels[1].ssx = 0; - det->panels[1].ssy = 1; - det->panels[1].xfs = 1; - det->panels[1].yfs = 0; - det->panels[1].xss = 0; - det->panels[1].yss = 1; + det->panels[1].fsx = 1.0; + det->panels[1].fsy = 0.0; + det->panels[1].fsz = 0.0; + det->panels[1].ssx = 0.0; + det->panels[1].ssy = 1.0; + det->panels[1].ssz = 1.4; + det->panels[1].xfs = 1.0; + det->panels[1].yfs = 0.0; + det->panels[1].xss = 0.0; + det->panels[1].yss = 1.0; det->panels[1].cnx = -512.0; det->panels[1].cny = sep; det->panels[1].clen = 100.0e-3; |