Remove unroll toggle; make throughput loop non-unrolled unconditionally

Author: leandrolcampos

libc/benchmarks/gpu/CMakeLists.txt

@@ -2,8 +2,6 @@ add_subdirectory(timing)
 
 add_custom_target(gpu-benchmark)
 
-option(LIBC_GPU_BENCHMARKS_ALLOW_UNROLL "Allow compiler loop unrolling in throughput loops" OFF)
-
 function(add_benchmark benchmark_name)
   cmake_parse_arguments(
     "BENCHMARK"
@@ -16,12 +14,6 @@ function(add_benchmark benchmark_name)
   if(NOT libc.src.time.clock IN_LIST TARGET_LLVMLIBC_ENTRYPOINTS)
     message(FATAL_ERROR "target does not support clock")
   endif()
-
-  set(benchmark_extra_flags "")
-  if(NOT LIBC_GPU_BENCHMARKS_ALLOW_UNROLL)
-    list(APPEND benchmark_extra_flags "-DLIBC_GPU_BENCHMARKS_DISABLE_UNROLL=1")
-  endif()
-
   add_libc_hermetic(
     ${benchmark_name}
     IS_GPU_BENCHMARK
@@ -34,7 +26,6 @@ function(add_benchmark benchmark_name)
     ${BENCHMARK_UNPARSED_ARGUMENTS}
     COMPILE_OPTIONS
       -flto
-      ${benchmark_extra_flags}
   )
   get_fq_target_name(${benchmark_name} fq_target_name)
   set(fq_build_target_name ${fq_target_name}.__build__)

libc/benchmarks/gpu/timing/amdgpu/timing.h

@@ -118,9 +118,7 @@ throughput_baseline(const cpp::array<T, N> &inputs) {
 
   T result{};
 
-  #if defined(LIBC_GPU_BENCHMARKS_DISABLE_UNROLL)
-  #pragma clang loop unroll(disable)
-  #endif
+#pragma clang loop unroll(disable)
   for (auto input : inputs) {
     asm("" ::"v"(input));
     result = input;
@@ -151,9 +149,7 @@ static LIBC_INLINE uint64_t throughput(F f, const cpp::array<T, N> &inputs) {
 
   T result{};
 
-  #if defined(LIBC_GPU_BENCHMARKS_DISABLE_UNROLL)
-  #pragma clang loop unroll(disable)
-  #endif
+#pragma clang loop unroll(disable)
   for (auto input : inputs) {
     asm("" ::"v"(input));
     result = f(input);
@@ -183,9 +179,7 @@ static LIBC_INLINE uint64_t throughput_baseline(
 
   T result{};
 
-  #if defined(LIBC_GPU_BENCHMARKS_DISABLE_UNROLL)
-  #pragma clang loop unroll(disable)
-  #endif
+#pragma clang loop unroll(disable)
   for (size_t i = 0; i < N; i++) {
     T x = inputs1[i];
     T y = inputs2[i];
@@ -219,9 +213,7 @@ static LIBC_INLINE uint64_t throughput(F f, const cpp::array<T, N> &inputs1,
 
   T result{};
 
-  #if defined(LIBC_GPU_BENCHMARKS_DISABLE_UNROLL)
-  #pragma clang loop unroll(disable)
-  #endif
+#pragma clang loop unroll(disable)
   for (size_t i = 0; i < N; i++) {
     T x = inputs1[i];
     T y = inputs2[i];

libc/benchmarks/gpu/timing/nvptx/timing.h

@@ -107,9 +107,7 @@ throughput_baseline(const cpp::array<T, N> &inputs) {
 
   T result{};
 
-  #if defined(LIBC_GPU_BENCHMARKS_DISABLE_UNROLL)
-  #pragma clang loop unroll(disable)
-  #endif
+#pragma clang loop unroll(disable)
   for (auto input : inputs) {
     asm("" ::"r"(input));
     result = input;
@@ -140,9 +138,7 @@ static LIBC_INLINE uint64_t throughput(F f, const cpp::array<T, N> &inputs) {
 
   T result{};
 
-  #if defined(LIBC_GPU_BENCHMARKS_DISABLE_UNROLL)
-  #pragma clang loop unroll(disable)
-  #endif
+#pragma clang loop unroll(disable)
   for (auto input : inputs) {
     asm("" ::"r"(input));
     result = f(input);
@@ -172,9 +168,7 @@ static LIBC_INLINE uint64_t throughput_baseline(
 
   T result{};
 
-  #if defined(LIBC_GPU_BENCHMARKS_DISABLE_UNROLL)
-  #pragma clang loop unroll(disable)
-  #endif
+#pragma clang loop unroll(disable)
   for (size_t i = 0; i < N; i++) {
     T x = inputs1[i];
     T y = inputs2[i];
@@ -208,9 +202,7 @@ static LIBC_INLINE uint64_t throughput(F f, const cpp::array<T, N> &inputs1,
 
   T result{};
 
-  #if defined(LIBC_GPU_BENCHMARKS_DISABLE_UNROLL)
-  #pragma clang loop unroll(disable)
-  #endif
+#pragma clang loop unroll(disable)
   for (size_t i = 0; i < N; i++) {
     T x = inputs1[i];
     T y = inputs2[i];