hans added a comment. > $ bin/clang -m32 -O0 /tmp/a.c && ./a.out > -nan > > Before your change, it prints 3.140000.
I looked through the Intel manual to understand what's happening in detail: When we return from f() with the new ABI, we write to the %mm0 register, and as a side effect: (9.5.1) After each MMX instruction, the entire x87 FPU tag word is set to valid (00B). What does that mean? (8.1.7) "The x87 FPU uses the tag values to detect stack overflow and underflow conditions (see Section 8.5.1.1)" (8.5.1.1) "Stack overflow — An instruction attempts to load a non-empty x87 FPU register from memory. A non-empty register is defined as a register containing a zero (tag value of 01), a valid value (tag value of 00), or a special value (tag value of 10). When the x87 FPU detects stack overflow or underflow, it sets the IE flag (bit 0) and the SF flag (bit 6) in the x87 FPU status word to 1. It then sets condition-code flag C1 (bit 9) in the x87 FPU status word to 1 if stack overflow occurred or to 0 if stack underflow occurred. If the invalid-operation exception is masked, the x87 FPU returns the floating point, integer, or packed decimal integer indefinite value to the destination operand, depending on the instruction being executed. This value over- writes the destination register or memory location specified by the instruction." Okay, so essentially any MMX instruction marks the x87 register stack as full, and when we try to store into it in d() with "fldl" we get a stack overflow, and because the exception is masked, it stores "the floating point indefinite value" into the register, which is what we end up printing. At least I finally think I understand what's going on :-) Repository: rL LLVM CHANGES SINCE LAST ACTION https://reviews.llvm.org/D59744/new/ https://reviews.llvm.org/D59744 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
