@@ -22,11 +22,51 @@
#include <stdint.h>
#include "config.h"
#include "libavutil/attributes.h"
+#include "libavutil/riscv/cpu.h"
#if defined (__GNUC__) || defined (__clang__)
#define av_bswap16 __builtin_bswap16
-#define av_bswap32 __builtin_bswap32
-#define av_bswap64 __builtin_bswap64
+
+static av_always_inline av_const uint32_t av_bswap32_rv(uint32_t x)
+{
+#if HAVE_RV && !defined(__riscv_zbb)
+ if (!__builtin_constant_p(x) &&
+ __builtin_expect(ff_rv_zbb_support(), 1)) {
+ uintptr_t y;
+
+ __asm__ (
+ ".option push\n"
+ ".option arch, +zbb\n"
+ "rev8 %0, %1\n"
+ ".option pop" : "=r" (y) : "r" (x));
+ return y >> (__riscv_xlen - 32);
+ }
+#endif
+ return __builtin_bswap32(x);
+}
+#define av_bswap32 av_bswap32_rv
+
+#if __riscv_xlen >= 64
+static av_always_inline av_const uint64_t av_bswap64_rv(uint64_t x)
+{
+#if HAVE_RV && !defined(__riscv_zbb)
+ if (!__builtin_constant_p(x) &&
+ __builtin_expect(ff_rv_zbb_support(), 1)) {
+ uintptr_t y;
+
+ __asm__ (
+ ".option push\n"
+ ".option arch, +zbb\n"
+ "rev8 %0, %1\n"
+ ".option pop" : "=r" (y) : "r" (x));
+ return y >> (__riscv_xlen - 64);
+ }
+#endif
+ return __builtin_bswap64(x);
+}
+#define av_bswap64 av_bswap64_rv
+#endif
+
#endif
#endif /* AVUTIL_RISCV_BSWAP_H */
This adds runtime support to use Zbb REV8 for 32- and 64-bit byte-wise swaps. The result is about five times slower than if targetting Zbb statically, but still a lot faster than the default bespoke C code or a call to GCC run-time functions. For 16-bit swap, this is however unsurprisingly a lot worse, and so this sticks to the baseline. In fact, even using REV8 statically does not seem to be beneficial in that case. Zbb static Zbb dynamic I baseline bswap16: 0.668184765 3.340764069 0.668029012 bswap32: 0.668174014 3.340763319 9.353855435 bswap64: 0.668221765 3.340496313 14.698672283 (seconds for 1 billion iterations on a SiFive-U74 core) --- libavutil/riscv/bswap.h | 44 +++++++++++++++++++++++++++++++++++++++-- 1 file changed, 42 insertions(+), 2 deletions(-)