From 874a856e34b87cce714be4110ab697725017f332 Mon Sep 17 00:00:00 2001
From: Dr-Noob <peibolms@gmail.com>
Date: Mon, 27 Dec 2021 12:42:03 +0100
Subject: [PATCH] [v1.01] Run multiple independent instructions in the pipeline
 for AVX freq too. Fixes incorrect frequency measures under certain CPUs

---
 Makefile                   |  6 +++---
 src/x86/freq/freq.h        |  2 +-
 src/x86/freq/freq_avx.c    | 21 +++++++++++++++++----
 src/x86/freq/freq_avx512.c |  6 ------
 4 files changed, 21 insertions(+), 14 deletions(-)

diff --git a/Makefile b/Makefile
index a24e774..c689e69 100644
--- a/Makefile
+++ b/Makefile
@@ -70,13 +70,13 @@ static: $(OUTPUT)
 strict: CFLAGS += -O2 -Werror -fsanitize=undefined -D_FORTIFY_SOURCE=2
 strict: $(OUTPUT)
 
-freq_nov.o: Makefile $(SRC_DIR)freq/freq_nov.c $(SRC_DIR)freq/freq_nov.h
+freq_nov.o: Makefile $(SRC_DIR)freq/freq_nov.c $(SRC_DIR)freq/freq_nov.h $(SRC_FREQ)/freq.h $(SRC_DIR)freq/freq.h
 	$(CC) $(CFLAGS) $(SANITY_FLAGS) -c -pthread $(SRC_DIR)freq/freq_nov.c -o $@
 
-freq_avx.o: Makefile $(SRC_DIR)freq/freq_avx.c $(SRC_DIR)freq/freq_avx.h
+freq_avx.o: Makefile $(SRC_DIR)freq/freq_avx.c $(SRC_DIR)freq/freq_avx.h $(SRC_DIR)freq/freq.h
 	$(CC) $(CFLAGS) $(SANITY_FLAGS) -c -mavx -pthread $(SRC_DIR)freq/freq_avx.c -o $@
 
-freq_avx512.o: Makefile $(SRC_DIR)freq/freq_avx512.c $(SRC_DIR)freq/freq_avx512.h
+freq_avx512.o: Makefile $(SRC_DIR)freq/freq_avx512.c $(SRC_DIR)freq/freq_avx512.h $(SRC_DIR)freq/freq.h
 	$(CC) $(CFLAGS) $(SANITY_FLAGS) -c -mavx512f -pthread $(SRC_DIR)freq/freq_avx512.c -o $@
 
 $(OUTPUT): Makefile $(SOURCE) $(HEADERS)
diff --git a/src/x86/freq/freq.h b/src/x86/freq/freq.h
index d757e63..80f99d2 100644
--- a/src/x86/freq/freq.h
+++ b/src/x86/freq/freq.h
@@ -5,7 +5,7 @@
 #include "../../common/cpu.h"
 
 #define MEASURE_TIME_SECONDS         5
-#define LOOP_ITERS          1000000000
+#define LOOP_ITERS           100000000
 
 int64_t measure_frequency(struct cpuInfo* cpu);
 
diff --git a/src/x86/freq/freq_avx.c b/src/x86/freq/freq_avx.c
index 4935972..29bf667 100644
--- a/src/x86/freq/freq_avx.c
+++ b/src/x86/freq/freq_avx.c
@@ -15,13 +15,25 @@ void* compute_avx() {
 
   struct timeval begin, now;
 
-  __m256 a = _mm256_set1_ps(1.5);
-  __m256 b = _mm256_set1_ps(1.2);
+  __m256 a[8];
+  __m256 b[8];
+
+  for(int i=0; i < 8; i++) {
+    a[i] = _mm256_set1_ps(1.5);
+    b[i] = _mm256_set1_ps(1.2);
+  }
 
   gettimeofday(&begin, NULL);
   while(!end) {
     for(uint64_t i=0; i < LOOP_ITERS; i++) {
-      a = _mm256_add_ps(a, b);
+      a[0] = _mm256_add_ps(a[0], b[0]);
+      a[1] = _mm256_add_ps(a[1], b[1]);
+      a[2] = _mm256_add_ps(a[2], b[2]);
+      a[3] = _mm256_add_ps(a[3], b[3]);
+      a[4] = _mm256_add_ps(a[4], b[4]);
+      a[5] = _mm256_add_ps(a[5], b[5]);
+      a[6] = _mm256_add_ps(a[6], b[6]);
+      a[7] = _mm256_add_ps(a[7], b[7]);
     }
 
     gettimeofday(&now, NULL);
@@ -34,7 +46,8 @@ void* compute_avx() {
     printf("fopen: %s", strerror(errno));
   }
   else {
-    fprintf(fp, "%f", a[0]);
+    for(int i=0; i < 8; i++)
+      fprintf(fp, "%f", a[i][0]);
     fclose(fp);
   }
 
diff --git a/src/x86/freq/freq_avx512.c b/src/x86/freq/freq_avx512.c
index b546574..df3ace1 100644
--- a/src/x86/freq/freq_avx512.c
+++ b/src/x86/freq/freq_avx512.c
@@ -10,12 +10,6 @@
 #include <stdint.h>
 #include "freq.h"
 
-/*
- * For AVX512, it seems that multiple independent
- * instructions are needed to force the CPU to
- * use AVX512 frequency, since with only one instruction
- * (as the AVX implementaion) it still uses AVX frequency
- */
 void* compute_avx512() {
   bool end = false;