diff --git a/NEWS b/NEWS
index 235b92f..a2eca85 100644
--- a/NEWS
+++ b/NEWS
@@ -1,5 +1,5 @@
 Version 0.1.3 (2016-mm-dd)
-  * Better support for ARM+SIMD and cross-compilation.
+  * Better support for clang, ARM+SIMD, and cross-compilation.
   * Other minor fixes/improvements to build/tests/documentation.
 
 Version 0.1.2 (2015-09-05)
diff --git a/cmake/Modules/FindOpenMP.cmake b/cmake/Modules/FindOpenMP.cmake
index eef8422..74e5bc5 100644
--- a/cmake/Modules/FindOpenMP.cmake
+++ b/cmake/Modules/FindOpenMP.cmake
@@ -50,6 +50,8 @@ include (FindPackageHandleStandardArgs)
 set (OpenMP_C_FLAG_CANDIDATES
   #Gnu
   "-fopenmp"
+  #Clang
+  "-fopenmp=libiomp5"
   #Microsoft Visual Studio
   "/openmp"
   #Intel windows
diff --git a/cmake/Modules/FindSIMD.cmake b/cmake/Modules/FindSIMD.cmake
index 2269b52..30dca70 100644
--- a/cmake/Modules/FindSIMD.cmake
+++ b/cmake/Modules/FindSIMD.cmake
@@ -62,6 +62,7 @@ else ()
 
   set (SIMD_C_FLAG_CANDIDATES
     # x64
+    "-Wno-cast-align"
     " "
     # Microsoft Visual Studio x86
     "/arch:SSE /fp:fast -D__SSE__"
diff --git a/src/filter.c b/src/filter.c
index a7809d1..482302e 100644
--- a/src/filter.c
+++ b/src/filter.c
@@ -101,7 +101,7 @@ double * lsx_make_lpf(
 
   if (h) for (i = 0; i <= m / 2; ++i) {
     double z = i - .5 * m, x = z * M_PI, y = z * mult1;
-    h[i] = x? sin(Fc * x) / x : Fc;
+    h[i] = x!=0? sin(Fc * x) / x : Fc;
     h[i] *= lsx_bessel_I_0(beta * sqrt(1 - y * y)) * mult;
     if (m - i != i)
       h[m - i] = h[i];
@@ -145,7 +145,7 @@ double * lsx_design_lpf(
 static double safe_log(double x)
 {
   assert(x >= 0);
-  if (x)
+  if (x!=0)
     return log(x);
   lsx_debug("log(0)");
   return -26;
@@ -222,7 +222,7 @@ void lsx_fir_to_phase(double * * h, int * len, int * post_len, double phase)
   while (peak && fabs(work[peak-1]) > fabs(work[peak]) && work[peak-1] * work[peak] > 0)
     --peak;
 
-  if (!phase1)
+  if (phase1==0)
     begin = 0;
   else if (phase1 == 1)
     begin = peak - *len / 2;
diff --git a/src/rate.h b/src/rate.h
index 1bc2ec7..f4466b7 100644
--- a/src/rate.h
+++ b/src/rate.h
@@ -416,7 +416,7 @@ static char const * rate_init(
   assert(passband_end + .005 < stopband_begin);
 
   p->factor = factor;
-  if (bits) while (!n++) {                               /* Determine stages: */
+  if (bits!=0) while (!n++) {                            /* Determine stages: */
     int try, L, M, x, maxL = interpolator > 0? 1 : mode? 2048 :
       (int)ceil((double)max_coefs_size * 1000. / (U100_l * sizeof(sample_t)));
     double d, epsilon = 0, frac;
@@ -425,9 +425,9 @@ static char const * rate_init(
     preM = upsample || (arbM > 1.5 && arbM < 2);
     postM = 1 + (arbM > 1 && preM), arbM /= postM;
     preL = 1 + (!preM && arbM < 2) + (upsample && mode), arbM *= preL;
-    if ((frac = arbM - (int)arbM))
+    if ((frac = arbM - (int)arbM)!=0)
       epsilon = fabs((uint32_t)(frac * MULT32 + .5) / (frac * MULT32) - 1);
-    for (i = 1, rational = !frac; i <= maxL && !rational; ++i) {
+    for (i = 1, rational = frac==0; i <= maxL && !rational; ++i) {
       d = frac * i, try = (int)(d + .5);
       if ((rational = fabs(try / d - 1) <= epsilon)) {    /* No long doubles! */
         if (try == i)
@@ -483,7 +483,7 @@ static char const * rate_init(
         log2_min_dft_size, log2_large_dft_size);
   }
 
-  if (!bits && have_arb_stage) {                  /* `Quick' cubic arb stage: */
+  if (bits==0 && have_arb_stage) {                  /* `Quick' cubic arb stage: */
     arb_stage.type = cubic_stage;
     arb_stage.fn = cubic_stage_fn;
     arb_stage.mult = multiplier, multiplier = 1;
@@ -516,7 +516,7 @@ static char const * rate_init(
         arbM /= arbL, arbL = 1, rational = false;
       phase_bits = (int)ceil(f1->scalar + log(mult)/log(2.));
       phases = !rational? (1 << phase_bits) : arbL;
-      if (!f->interp[0].scalar) {
+      if (f->interp[0].scalar==0) {
         int phases0 = max(phases, 19), n0 = 0;
         lsx_design_lpf(Fp, Fs, -Fn, attArb, &n0, phases0, f->beta);
         num_coefs = n0 / phases0 + 1, num_coefs += num_coefs & !preM;
diff --git a/src/soxr.c b/src/soxr.c
index 77b1bb0..d36891a 100644
--- a/src/soxr.c
+++ b/src/soxr.c
@@ -275,7 +275,7 @@ soxr_t soxr_create(
   soxr_quality_spec_t const * q_spec,
   soxr_runtime_spec_t const * runtime_spec)
 {
-  double io_ratio = output_rate? input_rate? input_rate / output_rate : -1 : input_rate? -1 : 0;
+  double io_ratio = output_rate!=0? input_rate!=0? input_rate / output_rate : -1 : input_rate!=0? -1 : 0;
   static const float datatype_full_scale[] = {1, 1, 65536.*32768, 32768};
   soxr_t p = 0;
   soxr_error_t error = 0;
@@ -343,7 +343,7 @@ soxr_t soxr_create(
     }
 #endif
 
-    if (p->num_channels && io_ratio)
+    if (p->num_channels && io_ratio!=0)
       error = soxr_set_io_ratio(p, io_ratio, 0);
   }
   if (error)
diff --git a/src/vr32.c b/src/vr32.c
index 000d340..9ad17d1 100644
--- a/src/vr32.c
+++ b/src/vr32.c
@@ -316,7 +316,7 @@ static void vr_init(rate_t * p, double default_io_ratio, int num_stages, double
   }
   fifo_create(&p->output_fifo, sizeof(float));
   p->default_io_ratio = default_io_ratio;
-  if (!fade_coefs[0]) {
+  if (fade_coefs[0]==0) {
     for (i = 0; i < iAL(fade_coefs); ++i)
       fade_coefs[i] = (float)(.5 * (1 + cos(M_PI * i / (AL(fade_coefs) - 1))));
     prepare_coefs(poly_fir_coefs_u, POLY_FIR_LEN_U, PHASES0_U, PHASES_U, coefs0_u, mult);
@@ -367,7 +367,7 @@ static void vr_set_io_ratio(rate_t * p, double io_ratio, size_t slew_len)
     }
   }
   else {
-    if (p->default_io_ratio) { /* Then this is the first call to this fn. */
+    if (p->default_io_ratio!=0) { /* Then this is the first call to this fn. */
       int octave = (int)floor(log(io_ratio) / M_LN2);
       p->current.stage_num = octave < 0? -1 : min(octave, p->num_stages0-1);
       enter_new_stage(p, 0);
@@ -375,7 +375,7 @@ static void vr_set_io_ratio(rate_t * p, double io_ratio, size_t slew_len)
     else if (p->fade_len)
       set_step(&p->fadeout, io_ratio);
     set_step(&p->current, io_ratio);
-    if (p->default_io_ratio) FRAC(p->current.at) = FRAC(p->current.step) >> 1;
+    if (p->default_io_ratio!=0) FRAC(p->current.at) = FRAC(p->current.step) >> 1;
     p->default_io_ratio = 0;
   }
 }
@@ -430,7 +430,7 @@ static bool do_input_stage(rate_t * p, int stage_num, int sign, int min_stage_nu
 static int vr_process(rate_t * p, int olen0)
 {
   assert(p->num_stages > 0);
-  if (p->default_io_ratio)
+  if (p->default_io_ratio!=0)
     vr_set_io_ratio(p, p->default_io_ratio, 0);
   {
     float * output = fifo_reserve(&p->output_fifo, olen0);
@@ -462,7 +462,7 @@ static int vr_process(rate_t * p, int olen0)
       olen = min(olen, (int)(AL(buf) >> 1));
       if (p->slew_len)
         olen = min(olen, p->slew_len);
-      else if (p->new_io_ratio) {
+      else if (p->new_io_ratio!=0) {
         set_step(&p->current, p->new_io_ratio);
         set_step(&p->fadeout, p->new_io_ratio);
         p->fadeout.step_step.all = p->current.step_step.all = 0;
diff --git a/tests/time-test b/tests/time-test
index a6ecf65..f253717 100755
--- a/tests/time-test
+++ b/tests/time-test
@@ -16,6 +16,7 @@ BASE=`basename $0`
 TIME=/tmp/$BASE-time-$$
 ERR=/tmp/$BASE-err-$$
 uname -m |grep -q ^arm && len=60 || len=600
+export OMP_NUM_THREADS=2
 
 for rate0 in $rates; do
 	rate1=44100