From e8ab4ce24dc1b25a56b5f1b69488d75c5a44e262 Mon Sep 17 00:00:00 2001
From: Ismael Gomez <ismagom@gmail.com>
Date: Wed, 8 Mar 2017 11:53:55 +0100
Subject: [PATCH] workaround for simd lut in debug mode

---
 CMakeLists.txt                 | 2 +-
 srslte/lib/fec/rm_turbo.c      | 5 +++++
 srslte/lib/utils/vector.c      | 5 +++++
 srslte/lib/utils/vector_simd.c | 2 ++
 4 files changed, 13 insertions(+), 1 deletion(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index faf11be37..8453f7b56 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -100,7 +100,7 @@ if(CMAKE_COMPILER_IS_GNUCC)
 
   if(${CMAKE_BUILD_TYPE} STREQUAL "Debug")
     find_package(SSE)
-    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -O0")
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -O0 -DDEBUG_MODE")
     if(HAVE_AVX)
       set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=native -mfpmath=sse -mavx -DLV_HAVE_AVX -DLV_HAVE_SSE")      
     elseif(HAVE_SSE)
diff --git a/srslte/lib/fec/rm_turbo.c b/srslte/lib/fec/rm_turbo.c
index 5856a4e46..1d250ff43 100644
--- a/srslte/lib/fec/rm_turbo.c
+++ b/srslte/lib/fec/rm_turbo.c
@@ -36,6 +36,11 @@
 #include "srslte/utils/vector.h"
 #include "srslte/fec/cbsegm.h"
 
+#ifdef DEBUG_MODE
+#warning FIXME: Disabling SSE/AVX turbo rate matching 
+#undef LV_HAVE_SSE
+#undef LV_HAVE_AVX
+#endif
 
 #ifdef LV_HAVE_SSE
 #include <smmintrin.h>
diff --git a/srslte/lib/utils/vector.c b/srslte/lib/utils/vector.c
index 469320177..f42c75b0e 100644
--- a/srslte/lib/utils/vector.c
+++ b/srslte/lib/utils/vector.c
@@ -285,6 +285,10 @@ void srslte_vec_lut_fuf(float *x, uint32_t *lut, float *y, uint32_t len) {
 }
 
 void srslte_vec_lut_sss(short *x, unsigned short *lut, short *y, uint32_t len) {
+#ifdef DEBUG_MODE
+#warning FIXME: Disabling SSE/AVX in srslte_vec_lut_sss
+  srslte_vec_lut_sss_simd(x, lut, y, len);
+#else
 #ifndef LV_HAVE_SSE
   for (int i=0;i<len;i++) {
     y[lut[i]] = x[i];
@@ -292,6 +296,7 @@ void srslte_vec_lut_sss(short *x, unsigned short *lut, short *y, uint32_t len) {
 #else
   srslte_vec_lut_sss_simd(x, lut, y, len);
 #endif
+#endif
 }
 
 void srslte_vec_interleave_cf(float *real, float *imag, cf_t *x, uint32_t len) {
diff --git a/srslte/lib/utils/vector_simd.c b/srslte/lib/utils/vector_simd.c
index 8fee71f21..f7dfa0574 100644
--- a/srslte/lib/utils/vector_simd.c
+++ b/srslte/lib/utils/vector_simd.c
@@ -213,6 +213,7 @@ void srslte_vec_sc_div2_sss_simd(short *x, int k, short *z, uint32_t len)
 /* No improvement with AVX */
 void srslte_vec_lut_sss_simd(short *x, unsigned short *lut, short *y, uint32_t len)
 {
+#ifndef DEBUG_MODE
 #ifdef LV_HAVE_SSE
   unsigned int number = 0;
   const unsigned int points = len / 8;
@@ -240,6 +241,7 @@ void srslte_vec_lut_sss_simd(short *x, unsigned short *lut, short *y, uint32_t l
     y[lut[number]] = x[number];
   }
 #endif  
+#endif
 }
 
 /* Modified from volk_32f_s32f_convert_16i_a_simd2. Removed clipping */