Make COMB_BLOCKS and COMB_TEETH configurable, and test in Cirrus

Author: sipa

.cirrus.yml

@@ -2,7 +2,8 @@ env:
   WIDEMUL: auto
   BIGNUM: auto
   STATICPRECOMPUTATION: yes
-  ECMULTGENPRECISION: auto
+  ECMULTGENBLOCKS: auto
+  ECMULTGENTEETH: auto
   ASM: no
   BUILD: check
   WITH_VALGRIND: yes
@@ -64,8 +65,9 @@ task:
         EXPERIMENTAL: yes
         SCHNORRSIG: yes
         CTIMETEST: no
-    - env: { ECMULTGENPRECISION: 2 }
-    - env: { ECMULTGENPRECISION: 8 }
+    - env: { ECMULTGENBLOCKS: 256, ECMULTGENTEETH: 1 }
+    - env: { ECMULTGENBLOCKS: 43, ECMULTGENTEETH: 6, STATICPRECOMPUTATION: no }
+    - env: { ECMULTGENBLOCKS: 1, ECMULTGENTEETH: 1, STATICPRECOMPUTATION: no }
     - env:
         RUN_VALGRIND: yes
         BIGNUM: no

ci/cirrus.sh

@@ -19,7 +19,7 @@ sed -i'' -e 's@/usr/bin/file@$(which file)@g' configure
 ./configure \
     --enable-experimental="$EXPERIMENTAL" \
     --with-test-override-wide-multiply="$WIDEMUL" --with-bignum="$BIGNUM" --with-asm="$ASM" \
-    --enable-ecmult-static-precomputation="$STATICPRECOMPUTATION" --with-ecmult-gen-precision="$ECMULTGENPRECISION" \
+    --enable-ecmult-static-precomputation="$STATICPRECOMPUTATION" --with-ecmult-gen-blocks="$ECMULTGENBLOCKS" --with-ecmult-gen-teeth="$ECMULTGENTEETH" \
     --enable-module-ecdh="$ECDH" --enable-module-recovery="$RECOVERY" \
     --enable-module-schnorrsig="$SCHNORRSIG" \
     --with-valgrind="$WITH_VALGRIND" \

configure.ac

@@ -170,13 +170,21 @@ AC_ARG_WITH([ecmult-window], [AS_HELP_STRING([--with-ecmult-window=SIZE|auto],
 )],
 [req_ecmult_window=$withval], [req_ecmult_window=auto])
 
-AC_ARG_WITH([ecmult-gen-precision], [AS_HELP_STRING([--with-ecmult-gen-precision=2|4|8|auto],
-[Precision bits to tune the precomputed table size for signing.]
-[The size of the table is 32kB for 2 bits, 64kB for 4 bits, 512kB for 8 bits of precision.]
-[A larger table size usually results in possible faster signing.]
+AC_ARG_WITH([ecmult-gen-blocks], [AS_HELP_STRING([--with-ecmult-gen-blocks=BLOCKS|auto],
+[The number of blocks to use in the multi-comb multiplication algorithm, in the range [1..256].]
+[Larger values result in possibly better performance at the cost of a linearly larger precomputed table.]
+[There must exist a multiple of BLOCKS*TEETH that is between 256 and 288, inclusive.]
 ["auto" is a reasonable setting for desktop machines (currently 4). [default=auto]]
 )],
-[req_ecmult_gen_precision=$withval], [req_ecmult_gen_precision=auto])
+[req_ecmult_gen_blocks=$withval], [req_ecmult_gen_blocks=auto])
+
+AC_ARG_WITH([ecmult-gen-teeth], [AS_HELP_STRING([--with-ecmult-gen-teeth=TEETH|auto],
+[The number of teeth to use in the multi-comb multiplication algorithm, in the range [1..8].]
+[Larger values result in possibly better performance at the cost of an exponentially larger precomputed table.]
+[There must exist a multiple of BLOCKS*TEETH that is between 256 and 288, inclusive.]
+["auto" is a reasonable setting for desktop machines (currently 5). [default=auto]]
+)],
+[req_ecmult_gen_teeth=$withval], [req_ecmult_gen_teeth=auto])
 
 AC_ARG_WITH([valgrind], [AS_HELP_STRING([--with-valgrind=yes|no|auto],
 [Build with extra checks for running inside Valgrind [default=auto]]
@@ -340,19 +348,45 @@ case $set_ecmult_window in
   ;;
 esac
 
-# Set ecmult gen precision
-if test x"$req_ecmult_gen_precision" = x"auto"; then
-  set_ecmult_gen_precision=4
+# Set ecmult gen blocks
+if test x"$req_ecmult_gen_blocks" = x"auto"; then
+  set_ecmult_gen_blocks=4
 else
-  set_ecmult_gen_precision=$req_ecmult_gen_precision
+  set_ecmult_gen_blocks=$req_ecmult_gen_blocks
 fi
+error_gen_blocks=['option to --with-ecmult-gen-blocks not an integer in range [1..256] or "auto"']
+case $set_ecmult_gen_blocks in
+''|*[[!0-9]]*)
+  # no valid integer
+  AC_MSG_ERROR($error_gen_blocks)
+  ;;
+*)
+  if test "$set_ecmult_gen_blocks" -lt 1 -o "$set_ecmult_gen_blocks" -gt 256 ; then
+    # not in range
+    AC_MSG_ERROR($error_gen_blocks)
+  fi
+  AC_DEFINE_UNQUOTED(COMB_BLOCKS, $set_ecmult_gen_blocks, [Set number of blocks in ecmult_gen precomputation])
+  ;;
+esac
 
-case $set_ecmult_gen_precision in
-2|4|8)
-  AC_DEFINE_UNQUOTED(ECMULT_GEN_PREC_BITS, $set_ecmult_gen_precision, [Set ecmult gen precision bits])
+#set ecmult gen teeth
+if test x"$req_ecmult_gen_teeth" = x"auto"; then
+  set_ecmult_gen_teeth=5
+else
+  set_ecmult_gen_teeth=$req_ecmult_gen_teeth
+fi
+error_gen_teeth=['option to --with-ecmult-gen-teeth not an integer in range [1..8] or "auto"']
+case $set_ecmult_gen_teeth in
+''|*[[!0-9]]*)
+  # no valid integer
+  AC_MSG_ERROR($error_gen_teeth)
   ;;
 *)
-  AC_MSG_ERROR(['ecmult gen precision not 2, 4, 8 or "auto"'])
+  if test "$set_ecmult_gen_teeth" -lt 1 -o "$set_ecmult_gen_teeth" -gt 8 ; then
+    # not in range
+    AC_MSG_ERROR($error_gen_teeth)
+  fi
+  AC_DEFINE_UNQUOTED(COMB_TEETH, $set_ecmult_gen_teeth, [Set number of teeth in ecmult_gen precomputation])
   ;;
 esac
 
@@ -565,7 +599,8 @@ echo
 echo "  asm                     = $set_asm"
 echo "  bignum                  = $set_bignum"
 echo "  ecmult window size      = $set_ecmult_window"
-echo "  ecmult gen prec. bits   = $set_ecmult_gen_precision"
+echo "  ecmult gen blocks       = $set_ecmult_gen_blocks"
+echo "  ecmult gen teeth        = $set_ecmult_gen_teeth"
 # Hide test-only options unless they're used.
 if test x"$set_widemul" != xauto; then
 echo "  wide multiplication     = $set_widemul"

src/ecmult_gen.h

@@ -10,11 +10,19 @@
 #include "scalar.h"
 #include "group.h"
 
+#if defined HAVE_CONFIG_H
+#include "libsecp256k1-config.h"
+#endif
+
 #if defined(EXHAUSTIVE_TEST_ORDER)
 
   /* We need to control these values for exhaustive tests because
    * the tables cannot have infinities in them (secp256k1_ge_storage
    * doesn't support infinities) */
+#undef COMB_BLOCKS
+#undef COMB_TEETH
+#undef COMB_SPACING
+
 #  if EXHAUSTIVE_TEST_ORDER > 32
 #    define COMB_BLOCKS 52
 #    define COMB_TEETH 5
@@ -39,9 +47,15 @@
   /* COMB_BLOCKS, COMB_TEETH, COMB_SPACING must all be positive and the product of the three (COMB_BITS)
    * must evaluate to a value in the range [256, 288]. The resulting memory usage for precomputation
    * will be COMB_POINTS_TOTAL * sizeof(secp256k1_ge_storage). */
-  #define COMB_BLOCKS 4
-  #define COMB_TEETH 5
-  #define COMB_SPACING 13
+#  ifndef COMB_BLOCKS
+#    define COMB_BLOCKS 4
+#  endif
+#  ifndef COMB_TEETH
+#    define COMB_TEETH 5
+#  endif
+#  ifndef COMB_SPACING
+#    define COMB_SPACING ((COMB_BLOCKS * COMB_TEETH + 255) / (COMB_BLOCKS * COMB_TEETH))
+#  endif
 
 #endif