[OPUS]: arch-guard fp8/bf8 packed-cvt builtins for RDNA3/3.5 (gfx1151)

csrc/include/aiter_opus_plus.h

@@ -51,13 +51,19 @@ OPUS_D decltype(auto) fp32_to_fp8_scaled_x2(const S& s, float inverted_scale)
     constexpr float hi = 448.0f, lo = -448.0f;
 #endif
     float a = tmp[0], b = tmp[1];
+#if defined(__gfx942__) || defined(__gfx950__)
     int w;
     asm volatile("v_med3_f32 %1, %1, %3, %4\n"
                  "v_med3_f32 %2, %2, %3, %4\n"
                  "v_cvt_pk_fp8_f32 %0, %1, %2"
                  : "=v"(w), "+v"(a), "+v"(b)
                  : "v"(lo), "v"(hi));
     return __builtin_bit_cast(fp8x2_t, static_cast<int16_t>(w));
+#else
+    // Arches without packed fp8-cvt (RDNA3/3.5, host): compile-only stub.
+    // fp8 KV-cache is unused on these arches; never executed at runtime.
+    (void)a; (void)b; (void)lo; (void)hi; return fp8x2_t{};
+#endif
 }
 
 template <typename S, std::enable_if_t<std::is_same_v<S, fp32x4_t>, bool> = true>
@@ -76,13 +82,17 @@ OPUS_D decltype(auto) fp32_to_bf8_scaled_x2(const S& s, float inverted_scale)
     fp32x2_t tmp       = pk_mul_f32(s, fp32x2_t{inverted_scale, inverted_scale});
     constexpr float hi = 57344.0f, lo = -57344.0f;
     float a = tmp[0], b = tmp[1];
+#if defined(__gfx942__) || defined(__gfx950__)
     int w;
     asm volatile("v_med3_f32 %1, %1, %3, %4\n"
                  "v_med3_f32 %2, %2, %3, %4\n"
                  "v_cvt_pk_bf8_f32 %0, %1, %2"
                  : "=v"(w), "+v"(a), "+v"(b)
                  : "v"(lo), "v"(hi));
     return __builtin_bit_cast(bf8x2_t, static_cast<int16_t>(w));
+#else
+    (void)a; (void)b; (void)lo; (void)hi; return bf8x2_t{};
+#endif
 }
 
 template <typename S, std::enable_if_t<std::is_same_v<S, fp32x4_t>, bool> = true>

csrc/include/opus/opus.hpp

@@ -1125,12 +1125,22 @@ OPUS_D constexpr auto fp32_to_bf16(const fp32_t& x, number<rm> = {}) {
 // Template constexpr (packed variants, OPUS_CAST_DEFINE) survives because the check is deferred to instantiation.
 // TODO: we may remove constexpr from cast in the future
 OPUS_D auto fp32_to_fp8(const fp32_t& x) {
+#if defined(__HIP_DEVICE_COMPILE__) && !(defined(__gfx942__) || defined(__gfx950__) || defined(__gfx1200__) || defined(__gfx1201__) || defined(__gfx1250__))
+    // RDNA3/3.5 (gfx1100/gfx115x) lack fp8-conversion-insts; compile-only
+    // stub so headers build. BF16 code paths never invoke fp8 conversion.
+    (void)x; return __builtin_bit_cast(fp8_t, static_cast<signed char>(0));
+#else
     int w; w = __builtin_amdgcn_cvt_pk_fp8_f32(x, 0.0f, w, /*sel=lo*/0);
     return __builtin_bit_cast(fp8_t, static_cast<signed char>(w));
+#endif
 }
 OPUS_D auto fp8_to_fp32(const fp8_t& x) {
+#if defined(__HIP_DEVICE_COMPILE__) && !(defined(__gfx942__) || defined(__gfx950__) || defined(__gfx1200__) || defined(__gfx1201__) || defined(__gfx1250__))
+    (void)x; return fp32_t(0.0f);
+#else
     int w = static_cast<int>(__builtin_bit_cast(unsigned char, x));
     return __builtin_amdgcn_cvt_f32_fp8(w, /*byte=*/0);
+#endif
 }
 OPUS_D constexpr auto fp32_to_fp32(const fp32_t& x) { return x; }
 OPUS_D constexpr auto fp32_to_i8(const fp32_t& x) { return static_cast<i8_t>(x); }