target/i386: Clear OF, SF, and AF for fcomi/fucomi
PATCH v2 1/1 target/i386: Clear OF, SF, and AF for fcomi/fucomi
helper_fcomi_ST0_FT0() and helper_fucomi_ST0_FT0() only cleared CC_Z, CC_P, and CC_C before merging in the comparison result, leaving CC_O, CC_S, and CC_A untouched from whatever they were set to beforehand.
The Intel SDM documents FCOMI/FCOMIP/FUCOMI/FUCOMIP as setting OF, SF, and AF to 0 unconditionally. The AMD manual doesn't mention them at all. However, testing on multiple real Intel and AMD systems confirms all three are unconditionally cleared regardless of the comparison result or their prior value.
Since fcomi_ccval[] only ever contains CC_C, CC_Z, 0, or CC_Z|CC_P|CC_C, and CC_O|CC_S|CC_Z|CC_A|CC_P|CC_C already covers every flag bit, CC_SRC can be assigned from fcomi_ccval[ret + 1] directly instead of ORing it into a masked cpu_cc_compute_all() result.
Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/4133
---
v2: Simplify per Richard Henderson's review: CC_O|CC_S|CC_A|CC_Z|CC_P|CC_C already covers
every flag, so drop the now-pointless cpu_cc_compute_all() call and assign CC_SRC from fcomi_ccval[] directly.
target/i386/tcg/fpu_helper.c | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-)
target/i386/tcg/fpu_helper.c +4 -6
@@ -550,12 +550,11 @@ static const int fcomi_ccval[4] = {CC_C, CC_Z, 0, CC_Z | CC_P | CC_C};void helper_fcomi_ST0_FT0 (CPUX86State *env) {int old_flags =save_exception_flags (env); -int eflags;FloatRelation ret; ret =floatx80_compare (ST0 ,FT0 , &env->fp_status); - eflags =cpu_cc_compute_all (env) & ~(CC_Z |CC_P |CC_C ); -CC_SRC = eflags | fcomi_ccval[ret +1 ]; +/* OF, SF, and AF are unconditionally cleared to 0 */ +CC_SRC = fcomi_ccval[ret +1 ];CC_OP =CC_OP_EFLAGS ;merge_exception_flags (env, old_flags); }
@@ -563,12 +562,11 @@ void helper_fcomi_ST0_FT0(CPUX86State *env)void helper_fucomi_ST0_FT0 (CPUX86State *env) {int old_flags =save_exception_flags (env); -int eflags;FloatRelation ret; ret =floatx80_compare_quiet (ST0 ,FT0 , &env->fp_status); - eflags =cpu_cc_compute_all (env) & ~(CC_Z |CC_P |CC_C ); -CC_SRC = eflags | fcomi_ccval[ret +1 ]; +/* OF, SF, and AF are unconditionally cleared to 0 */ +CC_SRC = fcomi_ccval[ret +1 ];CC_OP =CC_OP_EFLAGS ;merge_exception_flags (env, old_flags); }
Signature · 1 line
2.53.0
Paolo Bonzini
AppliedQueued, thanks.
Paolo
Quoted 17 lines · Simon Scherer
helper_fcomi_ST0_FT0()andhelper_fucomi_ST0_FT0()only clearedCC_Z,CC_P, andCC_Cbefore merging in the comparison result, leavingCC_O,CC_S, andCC_Auntouched from whatever they were set to beforehand.The Intel SDM documents FCOMI/FCOMIP/FUCOMI/FUCOMIP as setting OF, SF, and AF to 0 unconditionally. The AMD manual doesn't mention them at all. However, testing on multiple real Intel and AMD systems confirms all three are unconditionally cleared regardless of the comparison result or their prior value.
Since fcomi_ccval[] only ever contains
CC_C,CC_Z, 0, or CC_Z|CC_P|CC_C, and CC_O|CC_S|CC_Z|CC_A|CC_P|CC_C already covers every flag bit,CC_SRCcan be assigned from fcomi_ccval[ret + 1] directly instead of ORing it into a maskedcpu_cc_compute_all()result.Resolves:https://gitlab.com/qemu-project/qemu/-/work_items/4133
Quoted 8 lines
Signed-off-by Simon Scherer<[email protected]>---
v2: Simplify per Richard Henderson's review: CC_O|CC_S|CC_A|CC_Z|CC_P|CC_C already covers
every flag, so drop the now-pointlesscpu_cc_compute_all()call and assignCC_SRCfrom fcomi_ccval[] directly.target/i386/tcg/fpu_helper.c | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-)
r~