[mips] Fix the usage of kDoubleRegZero in RoundFloat/RoundDouble

kDoubleRegZero register is not hardwired to 0.0, we need to check
it's value before using it.

Change-Id: I8a64ac55294f51a9a5f96a39a3b344f7e34d508a
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2479223Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Zhi An Ng <zhin@chromium.org>
Commit-Queue: Zhao Jiazhong <zhaojiazhong-hf@loongson.cn>
Cr-Commit-Position: refs/heads/master@{#70595}

src/codegen/mips/macro-assembler-mips.cc

@@ -2087,6 +2087,9 @@ void TurboAssembler::RoundFloat(FPURegister dst, FPURegister src,
     int32_t kFloat32MantissaBits = 23;
     int32_t kFloat32ExponentBits = 8;
     Label done;
+    if (!IsDoubleZeroRegSet()) {
+      Move(kDoubleRegZero, 0.0);
+    }
     mfc1(scratch, src);
     Ext(at, scratch, kFloat32MantissaBits, kFloat32ExponentBits);
     Branch(USE_DELAY_SLOT, &done, hs, at,

src/codegen/mips64/macro-assembler-mips64.cc

@@ -2509,6 +2509,9 @@ void TurboAssembler::RoundDouble(FPURegister dst, FPURegister src,
     ctc1(scratch, FCSR);
   } else {
     Label done;
+    if (!IsDoubleZeroRegSet()) {
+      Move(kDoubleRegZero, 0.0);
+    }
     mfhc1(scratch, src);
     Ext(at, scratch, HeapNumber::kExponentShift, HeapNumber::kExponentBits);
     Branch(USE_DELAY_SLOT, &done, hs, at,
@@ -2570,6 +2573,9 @@ void TurboAssembler::RoundFloat(FPURegister dst, FPURegister src,
     int32_t kFloat32MantissaBits = 23;
     int32_t kFloat32ExponentBits = 8;
     Label done;
+    if (!IsDoubleZeroRegSet()) {
+      Move(kDoubleRegZero, 0.0);
+    }
     mfc1(scratch, src);
     Ext(at, scratch, kFloat32MantissaBits, kFloat32ExponentBits);
     Branch(USE_DELAY_SLOT, &done, hs, at,