no pidx

msm/pippenger.cuh

@@ -479,7 +479,7 @@ class msm_t {
         size_t digits_sz = nwins * stride * sizeof(uint32_t);
         size_t pidx_sz = stride * sizeof(uint32_t);
 
-        size_t d_blob_sz = d_buckets_sz + d_hist_sz + temp_sz + pidx_sz + digits_sz;
+        size_t d_blob_sz = d_buckets_sz + d_hist_sz + temp_sz + digits_sz + pidx_sz;
 
         d_total_blob = reinterpret_cast<char *>(gpu.Dmalloc(d_blob_sz));
         size_t offset = 0;
@@ -491,14 +491,15 @@ class msm_t {
         d_temps = vec2d_t<uint2>((uint2 *)&d_total_blob[offset], stride);
         d_scalars = (scalar_t *)&d_total_blob[offset];
         offset += temp_sz;
-        d_pidx = (uint32_t *)&d_total_blob[offset];
-        offset += pidx_sz;
         d_digits = vec2d_t<uint32_t>((uint32_t *)&d_total_blob[offset], stride);
+        offset += digits_sz;
+        d_pidx = (uint32_t *)&d_total_blob[offset];
     }
 
     RustError invoke(point_t &out, const scalar_t *scalars, size_t nscalars, uint32_t pidx[], bool mont = true) {
         // assert(this->nscalars <= nscalars);
         setup_scratch(nscalars);
+        printf("pidx is null?: %p", pidx);
 
         std::vector<result_t> res(nwins);
         std::vector<bucket_t> ones(gpu.sm_count() * BATCH_ADD_BLOCK_SIZE / WARP_SZ);