Changeset 104185 in vbox

Timestamp:

Apr 5, 2024 12:58:11 PM (8 months ago)

Author:

vboxsync

Message:

VMM/IEM: Get rid of IEM_MC_IF_MXCSR_XCPT_PENDING() and IEM_MC_RAISE_SSE_AVX_SIMD_FP_OR_UD_XCPT() and replace with IEM_MC_MAYBE_RAISE_SSE_AVX_SIMD_FP_OR_UD_XCPT(). This makes the microcode blocks leaner and the recompiler doesn't has to do as much state keeping. Furthermore there is now exactly one way of doing SSE/AVX floating point operations when it comes to exception handling, bugref:10641 [fix]

File:

• trunk/src/VBox/VMM/VMMAll/IEMAllInstTwoByte0f.cpp.h (modified) (2 diffs)

trunk/src/VBox/VMM/VMMAll/IEMAllInstTwoByte0f.cpp.h

@@ -3751 +3751 @@
             IEM_MC_REF_GREG_I32_CONST(pi32Src, IEM_GET_MODRM_RM(pVCpu, bRm));
             IEM_MC_CALL_SSE_AIMPL_2(iemAImpl_cvtsi2sd_r64_i32, pr64Dst, pi32Src);
+            IEM_MC_MAYBE_RAISE_SSE_AVX_SIMD_FP_OR_UD_XCPT();
             IEM_MC_STORE_XREG_R64(IEM_GET_MODRM_REG(pVCpu, bRm), r64Dst);
 
@@ -3773 +3774 @@
             IEM_MC_FETCH_MEM_I32(i32Src, pVCpu->iem.s.iEffSeg, GCPtrEffSrc);
             IEM_MC_CALL_SSE_AIMPL_2(iemAImpl_cvtsi2sd_r64_i32, pr64Dst, pi32Src);
+            IEM_MC_MAYBE_RAISE_SSE_AVX_SIMD_FP_OR_UD_XCPT();
             IEM_MC_STORE_XREG_R64(IEM_GET_MODRM_REG(pVCpu, bRm), r64Dst);