tomsfastmath: prefer size over speed

we can disable highly optimized mul/sqr operations for about 30% speed
decrease but saving a lot in binary size.
only build the files necessary by including an explicit list of filenames
rather than doing a wildcard over tfm/*.c.

Makefile

@@ -18,7 +18,9 @@ INTFLAGS += -DUSE_OPENSSL
 endif
 
 TARGET = pixiewps
-TFMSRC = $(sort $(wildcard $(SRCDIR)/crypto/tfm/*.c))
+
+include $(SRCDIR)/crypto/tfm/sources.mak
+TFMSRC = $(patsubst ./%,$(SRCDIR)/crypto/tfm/%,$(TFM_SRCS))
 TFMOBJS = $(TFMSRC:.c=.o)
 
 SOURCE = $(SRCDIR)/pixiewps.c

src/crypto/tfm/sources.mak

@@ -0,0 +1,37 @@
+TFM_SRCS = \
+./fp_2expt.c \
+./fp_add.c \
+./fp_cmp.c \
+./fp_cmp_d.c \
+./fp_cmp_mag.c \
+./fp_count_bits.c \
+./fp_div.c \
+./fp_div_2.c \
+./fp_div_2d.c \
+./fp_exptmod.c \
+./fp_invmod.c \
+./fp_lshd.c \
+./fp_mod.c \
+./fp_mod_2d.c \
+./fp_montgomery_reduce.c \
+./fp_montgomery_setup.c \
+./fp_montgomery_calc_normalization.c \
+./fp_mul.c \
+./fp_mul_2.c \
+./fp_mul_2d.c \
+./fp_mul_comba.c \
+./fp_mul_comba_small_set.c \
+./fp_mul_d.c \
+./fp_mulmod.c \
+./fp_read_unsigned_bin.c \
+./fp_reverse.c \
+./fp_rshd.c \
+./fp_set.c \
+./fp_sqr.c \
+./fp_sqr_comba_generic.c \
+./fp_sqr_comba_small_set.c \
+./fp_sub.c \
+./fp_to_unsigned_bin.c \
+./fp_unsigned_bin_size.c \
+./s_fp_add.c \
+./s_fp_sub.c

src/crypto/tfm/tfm.h

@@ -59,13 +59,13 @@
 #define TFM_MUL9
 #define TFM_MUL12
 #define TFM_MUL17
-#endif
 #define TFM_MUL20
 #define TFM_MUL24
 #define TFM_MUL28
 #define TFM_MUL32
 #define TFM_MUL48
 #define TFM_MUL64
+#endif
 
 #if 0
 #define TFM_SQR3
@@ -76,13 +76,13 @@
 #define TFM_SQR9
 #define TFM_SQR12
 #define TFM_SQR17
-#endif
 #define TFM_SQR20
 #define TFM_SQR24
 #define TFM_SQR28
 #define TFM_SQR32
 #define TFM_SQR48
 #define TFM_SQR64
+#endif
 
 /* do we want some overflow checks
    Not required if you make sure your numbers are within range (e.g. by default a modulus for fp_exptmod() can only be upto 2048 bits long)