Remove 1L and 1UL

Summary: 1L and 1UL are problematic because sizeof(long) depends on the platform

Reviewed By: mlomeli1

Differential Revision: D49911901

fbshipit-source-id: d4e4cb1f0283a33330bf1b8ca6b7f7bf41bc6ff4

Author: mdouze

faiss/Clustering.cpp

@@ -576,7 +576,7 @@ float kmeans_clustering(
         const float* x,
         float* centroids) {
     Clustering clus(d, k);
-    clus.verbose = d * n * k > (1L << 30);
+    clus.verbose = d * n * k > (size_t(1) << 30);
     // display logs if > 1Gflop per iteration
     IndexFlatL2 index(d);
     clus.train(n, x, index);

faiss/Index2Layer.cpp

@@ -47,7 +47,7 @@ Index2Layer::Index2Layer(
           pq(quantizer->d, M, nbit) {
     is_trained = false;
     for (int nbyte = 0; nbyte < 7; nbyte++) {
-        if ((1L << (8 * nbyte)) >= nlist) {
+        if (((size_t)1 << (8 * nbyte)) >= nlist) {
             code_size_1 = nbyte;
             break;
         }
@@ -179,7 +179,7 @@ struct DistanceXPQ4 : Distance2Level {
     float operator()(idx_t i) override {
 #ifdef __SSE3__
         const uint8_t* code = storage.codes.data() + i * storage.code_size;
-        long key = 0;
+        idx_t key = 0;
         memcpy(&key, code, storage.code_size_1);
         code += storage.code_size_1;
 
@@ -225,7 +225,7 @@ struct Distance2xXPQ4 : Distance2Level {
 
     float operator()(idx_t i) override {
         const uint8_t* code = storage.codes.data() + i * storage.code_size;
-        long key01 = 0;
+        int64_t key01 = 0;
         memcpy(&key01, code, storage.code_size_1);
         code += storage.code_size_1;
 #ifdef __SSE3__
@@ -237,7 +237,7 @@ struct Distance2xXPQ4 : Distance2Level {
         __m128 accu = _mm_setzero_ps();
 
         for (int mi_m = 0; mi_m < 2; mi_m++) {
-            long l1_idx = key01 & ((1L << mi_nbits) - 1);
+            int64_t l1_idx = key01 & (((int64_t)1 << mi_nbits) - 1);
             const __m128* pq_l1 = pq_l1_t + M_2 * l1_idx;
 
             for (int m = 0; m < M_2; m++) {

faiss/IndexPQ.cpp

@@ -818,7 +818,7 @@ struct MinSumK {
             // enqueue followers
             int64_t ii = ti;
             for (int m = 0; m < M; m++) {
-                int64_t n = ii & ((1L << nbit) - 1);
+                int64_t n = ii & (((int64_t)1 << nbit) - 1);
                 ii >>= nbit;
                 if (n + 1 >= N)
                     continue;
@@ -842,7 +842,7 @@ struct MinSumK {
             }
             int64_t ti = 0;
             for (int m = 0; m < M; m++) {
-                int64_t n = ii & ((1L << nbit) - 1);
+                int64_t n = ii & (((int64_t)1 << nbit) - 1);
                 ti += int64_t(ssx[m].get_ord(n)) << (nbit * m);
                 ii >>= nbit;
             }
@@ -966,7 +966,7 @@ void MultiIndexQuantizer::search(
 void MultiIndexQuantizer::reconstruct(idx_t key, float* recons) const {
     int64_t jj = key;
     for (int m = 0; m < pq.M; m++) {
-        int64_t n = jj & ((1L << pq.nbits) - 1);
+        int64_t n = jj & (((int64_t)1 << pq.nbits) - 1);
         jj >>= pq.nbits;
         memcpy(recons, pq.get_centroids(m, n), sizeof(recons[0]) * pq.dsub);
         recons += pq.dsub;
@@ -1098,7 +1098,7 @@ void MultiIndexQuantizer2::search(
 
                 const idx_t* idmap0 = sub_ids.data() + i * k2;
                 int64_t ld_idmap = k2 * n;
-                int64_t mask1 = ksub - 1L;
+                int64_t mask1 = ksub - (int64_t)1;
 
                 for (int k = 0; k < K; k++) {
                     const idx_t* idmap = idmap0;

faiss/gpu/perf/PerfClustering.cpp

@@ -42,7 +42,7 @@ int main(int argc, char** argv) {
 
     cudaProfilerStop();
 
-    auto seed = FLAGS_seed != -1L ? FLAGS_seed : time(nullptr);
+    auto seed = FLAGS_seed != -1 ? FLAGS_seed : time(nullptr);
     printf("using seed %ld\n", seed);
 
     std::vector<float> vecs((size_t)FLAGS_num * FLAGS_dim);

faiss/impl/AuxIndexStructures.cpp

@@ -230,7 +230,7 @@ bool InterruptCallback::is_interrupted() {
 
 size_t InterruptCallback::get_period_hint(size_t flops) {
     if (!instance.get()) {
-        return 1L << 30; // never check
+        return (size_t)1 << 30; // never check
     }
     // for 10M flops, it is reasonable to check once every 10 iterations
     return std::max((size_t)10 * 10 * 1000 * 1000 / (flops + 1), (size_t)1);

faiss/impl/lattice_Zn.cpp

@@ -455,7 +455,7 @@ void ZnSphereCodec::decode(uint64_t code, float* c) const {
     int nnz = 0;
     for (int i = 0; i < dim; i++) {
         if (c[i] != 0) {
-            if (signs & (1UL << nnz)) {
+            if (signs & (uint64_t(1) << nnz)) {
                 c[i] = -c[i];
             }
             nnz++;