On 14 August 2017 at 08:28, Dibyendu Majumdar <mobile@xxxxxxxxxxxxxxx> wrote:
> Hi,
>
> There appears to be a problem handling following:
>
> typedef struct lua_TValue {
> double value;
> int tt;
> } TValue;
> typedef TValue *StkId;
> struct lua_State {
> StkId top;
> StkId base;
> };
> static void lua_settop(struct lua_State *L, int idx)
> {
> if (idx >= 0) {
> { (void)L; };
> while (L->top < L->base + idx)
> ((L->top++)->tt = 0);
> L->top = L->base + idx;
> } else {
> { (void)L; };
> L->top += idx + 1;
> }
> }
>
> This generates:
>
> lua_settop:
> .L0:
> <entry-point>
> setge.32 %r2 <- %arg2, $0
> cbr %r2, .L6, .L2
>
> .L6:
> load.64 %r5 <- 0[VOID]
> load.64 %r7 <- 8[VOID]
> scast.64 %r9 <- (32) %arg2
> muls.64 %r10 <- %r9, $16
> add.64 %r11 <- %r7, %r10
> setb.32 %r12 <- %r5, %r11
> cbr %r12, .L3, .L5
>
> .L3:
> add.64 %r15 <- %r5, $16
> store.64 %r15 -> 0[VOID]
> store.32 $0 -> 8[%r5]
> br .L6
>
> .L5:
> store.64 %r11 -> 0[VOID]
> br .L7
>
> .L2:
> add.32 %r25 <- %arg2, $1
> scast.64 %r26 <- (32) %r25
> muls.64 %r27 <- %r26, $16
> load.64 %r29 <- 0[%arg1]
> cast.64 %r30 <- (64) %r29
> add.64 %r31 <- %r30, %r27
> ptrcast.64 %r32 <- (64) %r31
> store.64 %r32 -> 0[%arg1]
> br .L7
>
> .L7:
> ret
Interestingly with single store shortcut enabled the output (from dmrC
this time) is:
lua_settop:
.L0:
<entry-point>
setge.32 %r2 <- %arg2, $0
cbr %r2, .L6, .L2
.L6:
load.64* %r5 <- 0[%arg1]
load.64* %r7 <- 8[%arg1]
scast.64 %r9 <- (32) %arg2
muls.64 %r10 <- %r9, $16
add.64* %r11 <- %r7, %r10
setb.32 %r12 <- %r5, %r11
cbr %r12, .L3, .L5
.L3:
add.64* %r15 <- %r5, $16
store.64* %r15 -> 0[%arg1]
store.32 $0 -> 8[%r5]
br .L6
.L5:
store.64* %r11 -> 0[%arg1]
br .L7
.L2:
add.32 %r25 <- %arg2, $1
scast.64 %r26 <- (32) %r25
muls.64 %r27 <- %r26, $16
load.64* %r29 <- 0[%arg1]
cast.64 %r30 <- (64) %r29
add.64 %r31 <- %r30, %r27
ptrcast.64* %r32 <- (64) %r31
store.64* %r32 -> 0[%arg1]
br .L7
.L7:
ret
What this tells me is that potentially there is a bug in the way
single store shortcut has been removed - i.e. it is having a different
effect here.
Regards
Dibyendu
--
To unsubscribe from this list: send the line "unsubscribe linux-sparse" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
