Use thread time instead of process time

Fixes problem with measuring performance on Arm platform for
reasons unknown.

dmath.h

@@ -72,7 +72,7 @@ static inline constexpr void lfsr_next(uint64_t& state, uint32_t tap) { const ui
 static inline constexpr __attribute__((const))  uint64_t lfsr_init(uint64_t state, uint32_t bits) { return fastrange(splitmix64(state), 1, 1 << bits); }
 
 /// CPU time measurement
-static inline uint64_t cpu_gettime() { struct timespec t; clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &t); return ((uint64_t)t.tv_sec * 1000000000ull + (uint64_t)t.tv_nsec); }
+static inline uint64_t cpu_gettime() { struct timespec t; clock_gettime(CLOCK_THREAD_CPUTIME_ID, &t); return ((uint64_t)t.tv_sec * 1000000000ull + (uint64_t)t.tv_nsec); }
 
 // -- 2^10 fractional math --
 // Numbers based around 100 are not good to do fast math on, so we can use to a 2^10 system instead. We call these "perten" values. Math using

tests/perf_table_rolls.cpp

@@ -30,27 +30,27 @@ int main(int argc, char **argv)
 	printf("Size of 5000 entry LRT=%u LS=%u\n", (unsigned)(sizeof(lrt5k) + lrt5k.table.size() * sizeof(*lrt5k.table.data())), (unsigned)sizeof(ls5k));
 
 	t1 = cpu_gettime();
-	for (int i = 0; i < 4000; i++) lrt5k.roll();
+	for (int i = 0; i < 40000; i++) lrt5k.roll();
 	t2 = cpu_gettime();
-	printf("%-30s %'12" PRIu64 "\n", "LRT 4k rolls 5k alloc", t2 - t1);
+	printf("%-30s %'12" PRIu64 "\n", "LRT 40k rolls 5k alloc", t2 - t1);
 
 	t1 = cpu_gettime();
-	for (int i = 0; i < 4000; i++) ls5k.roll();
+	for (int i = 0; i < 40000; i++) ls5k.roll();
 	t2 = cpu_gettime();
-	printf("%-30s %'12" PRIu64 "\n", "LS 4k rolls 5k alloc", t2 - t1);
+	printf("%-30s %'12" PRIu64 "\n", "LS 40k rolls 5k alloc", t2 - t1);
 
 	const std::vector<int> w5k(5000, 1);
 	roll_table rt5k(s, w5k);
 	t1 = cpu_gettime();
-	for (int i = 0; i < 4000; i++) rt5k.roll();
+	for (int i = 0; i < 40000; i++) rt5k.roll();
 	t2 = cpu_gettime();
-	printf("%-30s %'12" PRIu64 "\n", "RT 4k rolls 5k alloc", t2 - t1);
+	printf("%-30s %'12" PRIu64 "\n", "RT 40k rolls 5k alloc", t2 - t1);
 
 	linear_series ls4k(s, 4096-1);
 	t1 = cpu_gettime();
-	for (int i = 0; i < 4000; i++) ls4k.roll();
+	for (int i = 0; i < 40000; i++) ls4k.roll();
 	t2 = cpu_gettime();
-	printf("%-30s %'12" PRIu64 "\n", "LS 4k rolls 4k-1 alloc", t2 - t1);
+	printf("%-30s %'12" PRIu64 "\n", "LS 40k rolls 4k-1 alloc", t2 - t1);
 
 	std::vector<int> weights(200);
 	for (int i = 0; i < 200; i++) weights[i] = 100 + i*50;