Release v4.1.3

Fix memory address int32 overflow for heavily contracted basis
sunqm · Apr 13, 2021 · 38fa7cd · 38fa7cd
1 parent faaba0b
commit 38fa7cd
Show file tree

Hide file tree

Showing 7 changed files with 79 additions and 53 deletions.
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -2,7 +2,7 @@ cmake_minimum_required (VERSION 3.5)
 project (qcint C)
 set(qcint_VERSION_MAJOR "4")
 set(qcint_VERSION_MINOR "1")
-set(qcint_VERSION_PATCH "2")
+set(qcint_VERSION_PATCH "3")
 set(qcint_VERSION_TWEAK "0")
 set(qcint_VERSION "${qcint_VERSION_MAJOR}.${qcint_VERSION_MINOR}.${qcint_VERSION_PATCH}")
 set(qcint_SOVERSION "${qcint_VERSION_MAJOR}")

diff --git a/ChangeLog b/ChangeLog
@@ -1,3 +1,5 @@
+Version 4.1.3  (2021-04-13):
+	* Fix memory address int32 overflow for heavily contracted basis
 Version 4.1.2  (2021-04-10):
 	* Fix a bug due to significant digits of float128 in the core Fmt integrals
 	* Fix compiling error for stg, coulomb_erf integral code

diff --git a/README.md b/README.md
@@ -3,9 +3,9 @@ qcint (quick libcint)
 
 An optimized libcint branch for X86 platform
 
-version 4.1.2
+version 4.1.3
 
-2021-04-10
+2021-04-13
 
 
 What is qcint

diff --git a/src/cart2sph.c b/src/cart2sph.c
@@ -7343,8 +7343,8 @@ void c2s_dset0(double *out, int *dims, int *counts)
         int ni = dims[0];
         int nj = dims[1];
         int nk = dims[2];
-        int nij = ni * nj;
-        int nijk = nij * nk;
+        size_t nij = ni * nj;
+        size_t nijk = nij * nk;
         int i, j, k, l;
         if (dims == counts) {
                 for (i = 0; i < nijk * counts[3]; i++) {
@@ -7373,8 +7373,8 @@ void c2s_zset0(double complex *out, int *dims, int *counts)
         int ni = dims[0];
         int nj = dims[1];
         int nk = dims[2];
-        int nij = ni * nj;
-        int nijk = nij * nk;
+        size_t nij = ni * nj;
+        size_t nijk = nij * nk;
         int i, j, k, l;
         if (dims == counts) {
                 for (i = 0; i < nijk * counts[3]; i++) {

diff --git a/src/cint2e.c b/src/cint2e.c
@@ -19,7 +19,8 @@
  */
 
 #include <stdlib.h>
-#include <string.h>
+#include <stdio.h>
+#include <stdint.h>
 #include <math.h>
 #include "cint_bas.h"
 #include "misc.h"
@@ -218,7 +219,7 @@ int CINT2e_loop_nopt(double *out, CINTEnvVars *envs, double *cache)
         double *rk = envs->rk;
         double *rl = envs->rl;
         int n_comp = envs->ncomp_e1 * envs->ncomp_e2 * envs->ncomp_tensor;
-        int nf = envs->nf;
+        size_t nf = envs->nf;
         double fac1i, fac1j, fac1k, fac1l;
         int ip, jp, kp, lp, i, it, im;
         int empty[4] = {1, 1, 1, 1};
@@ -251,18 +252,18 @@ int CINT2e_loop_nopt(double *out, CINTEnvVars *envs, double *cache)
                 * CINTcommon_fac_sp(envs->i_l) * CINTcommon_fac_sp(envs->j_l)
                 * CINTcommon_fac_sp(envs->k_l) * CINTcommon_fac_sp(envs->l_l);
 
-        int ngp[4];
+        size_t ngp[4];
         ngp[0] = nf * n_comp;
         ngp[1] = ngp[0] * i_ctr;
         ngp[2] = ngp[1] * j_ctr;
         ngp[3] = ngp[2] * k_ctr;
         // (irys,i,j,k,l,coord,0:1); +1 for nabla-r12
         int leng = envs->g_size * 3 * ((1<<envs->gbits)+1) * SIMDD;
-        int len0 = ngp[0] * SIMDD;
-        int leni = ALIGN_UP(ngp[1], SIMDD);
-        int lenj = ALIGN_UP(ngp[2], SIMDD);
-        int lenk = ALIGN_UP(ngp[3], SIMDD);
-        int lenl = ALIGN_UP(ngp[3] * l_ctr, SIMDD);
+        size_t len0 = ngp[0] * SIMDD;
+        size_t leni = ALIGN_UP(ngp[1], SIMDD);
+        size_t lenj = ALIGN_UP(ngp[2], SIMDD);
+        size_t lenk = ALIGN_UP(ngp[3], SIMDD);
+        size_t lenl = ALIGN_UP(ngp[3] * l_ctr, SIMDD);
 
         double *gout, *g, *g1;
         double *gctr[4];
@@ -402,7 +403,7 @@ int CINT2e_loop(double *out, CINTEnvVars *envs, CINTOpt *opt, double *cache)
         }
 
         int n_comp = envs->ncomp_e1 * envs->ncomp_e2 * envs->ncomp_tensor;
-        int nf = envs->nf;
+        size_t nf = envs->nf;
         double fac1i, fac1j, fac1k, fac1l;
         int ip, jp, kp, lp, i, it, im;
         int empty[4] = {1, 1, 1, 1};
@@ -420,18 +421,18 @@ int CINT2e_loop(double *out, CINTEnvVars *envs, CINTOpt *opt, double *cache)
                 CINTg4c_index_xyz(idx, envs);
         }
 
-        int ngp[4];
+        size_t ngp[4];
         ngp[0] = nf * n_comp;
         ngp[1] = ngp[0] * i_ctr;
         ngp[2] = ngp[1] * j_ctr;
         ngp[3] = ngp[2] * k_ctr;
         // (irys,i,j,k,l,coord,0:1); +1 for nabla-r12
         int leng = envs->g_size * 3 * ((1<<envs->gbits)+1) * SIMDD;
-        int len0 = ngp[0] * SIMDD;
-        int leni = ALIGN_UP(ngp[1], SIMDD);
-        int lenj = ALIGN_UP(ngp[2], SIMDD);
-        int lenk = ALIGN_UP(ngp[3], SIMDD);
-        int lenl = ALIGN_UP(ngp[3] * l_ctr, SIMDD);
+        size_t len0 = ngp[0] * SIMDD;
+        size_t leni = ALIGN_UP(ngp[1], SIMDD);
+        size_t lenj = ALIGN_UP(ngp[2], SIMDD);
+        size_t lenk = ALIGN_UP(ngp[3], SIMDD);
+        size_t lenl = ALIGN_UP(ngp[3] * l_ctr, SIMDD);
         double *gout, *g, *g1;
         double *gctr[4];
         double *bufctr[4];
@@ -521,32 +522,39 @@ k_contracted: ;
                 int j_prim = bas(NPRIM_OF, shls[1]); \
                 int k_prim = bas(NPRIM_OF, shls[2]); \
                 int l_prim = bas(NPRIM_OF, shls[3]); \
-                int ps = ((i_prim*j_prim + k_prim*l_prim) * 5 \
+                size_t ps = ((i_prim*j_prim + k_prim*l_prim) * 5 \
                           + i_prim * x_ctr[0] \
                           + j_prim * x_ctr[1] \
                           + k_prim * x_ctr[2] \
                           + l_prim * x_ctr[3] \
-                          +(i_prim+j_prim+k_prim+l_prim)*2 + envs->nf*3);
+                          +(i_prim+j_prim+k_prim+l_prim)*2 + nf*3);
 
 int CINT2e_cart_drv(double *out, int *dims, CINTEnvVars *envs, CINTOpt *opt,
                     double *cache)
 {
         int *x_ctr = envs->x_ctr;
-        int nc = envs->nf * x_ctr[0] * x_ctr[1] * x_ctr[2] * x_ctr[3];
+        size_t nf = envs->nf;
+        size_t nc = nf * x_ctr[0] * x_ctr[1] * x_ctr[2] * x_ctr[3];
         int n_comp = envs->ncomp_e1 * envs->ncomp_e2 * envs->ncomp_tensor;
         if (out == NULL) {
                 PAIRDATA_NON0IDX_SIZE(pdata_size);
                 int leng = envs->g_size*3*((1<<envs->gbits)+1)*SIMDD;
-                int len0 = envs->nf*n_comp * SIMDD;
-                int cache_size = leng + len0 + nc*n_comp*2 + SIMDD*4 + pdata_size;
+                size_t len0 = nf*n_comp * SIMDD;
+                size_t cache_size = leng + len0 + nc*n_comp*2 + SIMDD*4 + pdata_size;
+                if (cache_size >= INT32_MAX) {
+                        fprintf(stderr, "CINT2e_cart_drv cache_size overflow: "
+                                "cache_size %ld > %d, nf %ld, nc %ld, n_comp %d\n",
+                                cache_size, INT32_MAX, nf, nc, n_comp);
+                        cache_size = 0;
+                }
                 return cache_size;
         }
         double *stack = NULL;
         if (cache == NULL) {
                 PAIRDATA_NON0IDX_SIZE(pdata_size);
                 int leng = envs->g_size*3*((1<<envs->gbits)+1)*SIMDD;
-                int len0 = envs->nf*n_comp * SIMDD;
-                int cache_size = leng + len0 + nc*n_comp*2 + SIMDD*4 + pdata_size;
+                size_t len0 = nf*n_comp * SIMDD;
+                size_t cache_size = leng + len0 + nc*n_comp*2 + SIMDD*4 + pdata_size;
                 stack = _mm_malloc(sizeof(double)*cache_size, sizeof(double)*SIMDD);
                 cache = stack;
         }
@@ -587,23 +595,30 @@ int CINT2e_spheric_drv(double *out, int *dims, CINTEnvVars *envs, CINTOpt *opt,
                        double *cache)
 {
         int *x_ctr = envs->x_ctr;
-        int nc = envs->nf * x_ctr[0] * x_ctr[1] * x_ctr[2] * x_ctr[3];
+        size_t nf = envs->nf;
+        size_t nc = nf * x_ctr[0] * x_ctr[1] * x_ctr[2] * x_ctr[3];
         int n_comp = envs->ncomp_e1 * envs->ncomp_e2 * envs->ncomp_tensor;
         if (out == NULL) {
                 PAIRDATA_NON0IDX_SIZE(pdata_size);
                 int leng = envs->g_size*3*((1<<envs->gbits)+1)*SIMDD;
-                int len0 = envs->nf*n_comp * SIMDD;
-                int cache_size = MAX(leng+len0+nc*n_comp*2 + pdata_size,
-                                     nc*n_comp+envs->nf*4) + SIMDD*4;
+                size_t len0 = nf*n_comp * SIMDD;
+                size_t cache_size = MAX(leng+len0+nc*n_comp*2 + pdata_size,
+                                     nc*n_comp+nf*4) + SIMDD*4;
+                if (cache_size >= INT32_MAX) {
+                        fprintf(stderr, "CINT2e_spinor_drv cache_size overflow: "
+                                "cache_size %ld > %d, nf %ld, nc %ld, n_comp %d\n",
+                                cache_size, INT32_MAX, nf, nc, n_comp);
+                        cache_size = 0;
+                }
                 return cache_size;
         }
         double *stack = NULL;
         if (cache == NULL) {
                 PAIRDATA_NON0IDX_SIZE(pdata_size);
                 int leng = envs->g_size*3*((1<<envs->gbits)+1)*SIMDD;
-                int len0 = envs->nf*n_comp * SIMDD;
-                int cache_size = MAX(leng+len0+nc*n_comp*2 + pdata_size,
-                                     nc*n_comp+envs->nf*4) + SIMDD*4;
+                size_t len0 = nf*n_comp * SIMDD;
+                size_t cache_size = MAX(leng+len0+nc*n_comp*2 + pdata_size,
+                                     nc*n_comp+nf*4) + SIMDD*4;
                 stack = _mm_malloc(sizeof(double)*cache_size, sizeof(double)*SIMDD);
                 cache = stack;
         }
@@ -651,27 +666,34 @@ int CINT2e_spinor_drv(double complex *out, int *dims, CINTEnvVars *envs, CINTOpt
         counts[2] = CINTcgto_spinor(shls[2], bas);
         counts[3] = CINTcgto_spinor(shls[3], bas);
         int *x_ctr = envs->x_ctr;
-        int nc = envs->nf * x_ctr[0] * x_ctr[1] * x_ctr[2] * x_ctr[3];
+        size_t nf = envs->nf;
+        size_t nc = nf * x_ctr[0] * x_ctr[1] * x_ctr[2] * x_ctr[3];
         int n_comp = envs->ncomp_e1 * envs->ncomp_e2 * envs->ncomp_tensor;
         int n1 = counts[0] * envs->nfk * x_ctr[2]
                            * envs->nfl * x_ctr[3] * counts[1];
         if (out == NULL) {
                 PAIRDATA_NON0IDX_SIZE(pdata_size);
                 int leng = envs->g_size*3*((1<<envs->gbits)+1)*SIMDD;
-                int len0 = envs->nf*n_comp * SIMDD;
-                int cache_size = MAX(leng+len0+nc*n_comp*2 + pdata_size,
+                size_t len0 = nf*n_comp * SIMDD;
+                size_t cache_size = MAX(leng+len0+nc*n_comp*2 + pdata_size,
                                      nc*n_comp + n1*envs->ncomp_e2*OF_CMPLX
-                                     + envs->nf*32*OF_CMPLX) + SIMDD*4;
+                                     + nf*32*OF_CMPLX) + SIMDD*4;
+                if (cache_size >= INT32_MAX) {
+                        fprintf(stderr, "CINT2e_spheric_drv cache_size overflow: "
+                                "cache_size %ld > %d, nf %ld, nc %ld, n_comp %d\n",
+                                cache_size, INT32_MAX, nf, nc, n_comp);
+                        cache_size = 0;
+                }
                 return cache_size;
         }
         double *stack = NULL;
         if (cache == NULL) {
                 PAIRDATA_NON0IDX_SIZE(pdata_size);
                 int leng = envs->g_size*3*((1<<envs->gbits)+1)*SIMDD;
-                int len0 = envs->nf*n_comp * SIMDD;
-                int cache_size = MAX(leng+len0+nc*n_comp*2 + pdata_size,
+                size_t len0 = nf*n_comp * SIMDD;
+                size_t cache_size = MAX(leng+len0+nc*n_comp*2 + pdata_size,
                                      nc*n_comp + n1*envs->ncomp_e2*OF_CMPLX
-                                     + envs->nf*32*OF_CMPLX) + SIMDD*4;
+                                     + nf*32*OF_CMPLX) + SIMDD*4;
                 stack = _mm_malloc(sizeof(double)*cache_size, sizeof(double)*SIMDD);
                 cache = stack;
         }

diff --git a/src/g1e.c b/src/g1e.c
@@ -609,10 +609,11 @@ MM_STORE(fz+n*SIMDD, MM_FMA(r2, MM_LOAD(gz+n*SIMDD), MM_LOAD(p1z+n*SIMDD)));
 }
 
 
-void CINTprim_to_ctr_0(double *RESTRICT gc, double *RESTRICT gp, double *RESTRICT coeff, int nf,
+void CINTprim_to_ctr_0(double *RESTRICT gc, double *RESTRICT gp, double *RESTRICT coeff, size_t nf,
                        int nprim, int nctr, int non0ctr, int *sortedidx)
 {
-        int n, i;
+        int i;
+        size_t n;
         double c0, c1, c2, c3;
         double *RESTRICT p0;
         double *RESTRICT p1;
@@ -740,10 +741,11 @@ void CINTprim_to_ctr_0(double *RESTRICT gc, double *RESTRICT gp, double *RESTRIC
         }
 }
 
-void CINTprim_to_ctr_1(double *RESTRICT gc, double *RESTRICT gp, double *RESTRICT coeff, int nf,
+void CINTprim_to_ctr_1(double *RESTRICT gc, double *RESTRICT gp, double *RESTRICT coeff, size_t nf,
                        int nprim, int nctr, int non0ctr, int *sortedidx)
 {
-        int n, i;
+        int i;
+        size_t n;
         double c0, c1, c2, c3;
         double *RESTRICT p0;
         double *RESTRICT p1;
@@ -877,7 +879,7 @@ double CINTcommon_fac_sp(int l)
 }
 
 
-void CINTiprim_to_ctr_0(double *RESTRICT gc, double *RESTRICT gp, double *RESTRICT coeff, int nf,
+void CINTiprim_to_ctr_0(double *RESTRICT gc, double *RESTRICT gp, double *RESTRICT coeff, size_t nf,
                         int nprim, int nctr, int non0ctr, int *sortedidx)
 {
         int i;
@@ -927,7 +929,7 @@ void CINTiprim_to_ctr_0(double *RESTRICT gc, double *RESTRICT gp, double *RESTRI
         }
 }
 
-void CINTiprim_to_ctr_1(double *RESTRICT gc, double *RESTRICT gp, double *RESTRICT coeff, int nf,
+void CINTiprim_to_ctr_1(double *RESTRICT gc, double *RESTRICT gp, double *RESTRICT coeff, size_t nf,
                         int nprim, int nctr, int non0ctr, int *sortedidx)
 {
         int i;

diff --git a/src/g1e.h b/src/g1e.h
@@ -120,13 +120,13 @@ void CINTx1k_1e(double *f, double *g, double *rk,
 
 void CINTprim_to_ctr(double *gc, int nf, double *gp,
                      int inc, int nprim, int nctr, double *pcoeff);
-void CINTprim_to_ctr_0(double *gc, double *gp, double *coeff, int nf,
+void CINTprim_to_ctr_0(double *gc, double *gp, double *coeff, size_t nf,
                        int nprim, int nctr, int non0ctr, int *sortedidx);
-void CINTprim_to_ctr_1(double *gc, double *gp, double *coeff, int nf,
+void CINTprim_to_ctr_1(double *gc, double *gp, double *coeff, size_t nf,
                        int nprim, int nctr, int non0ctr, int *sortedidx);
-void CINTiprim_to_ctr_0(double *gc, double *gp, double *coeff, int nf,
+void CINTiprim_to_ctr_0(double *gc, double *gp, double *coeff, size_t nf,
                         int nprim, int nctr, int non0ctr, int *sortedidx);
-void CINTiprim_to_ctr_1(double *gc, double *gp, double *coeff, int nf,
+void CINTiprim_to_ctr_1(double *gc, double *gp, double *coeff, size_t nf,
                         int nprim, int nctr, int non0ctr, int *sortedidx);
 void CINTsort_gout(double *sout, double *gout, int nf, int count);