Fix a bug in memcpy, implement memset, and remove model_thread

[c11tester.git] / mymemory.cc

diff --git a/mymemory.cc b/mymemory.cc
index e82f8d2ef997f31f0d2d38871c74dd6cf17a8cbd..cda63dcc4bd7d20928bf16a089afaa318a0186b0 100644 (file)
--- a/mymemory.cc
+++ b/mymemory.cc
@@ -137,31 +137,32 @@ void init_memory_ops()
 }
 
 void * memcpy(void * dst, const void * src, size_t n) {
-       if (false && model && !inside_model) {
+       if (model && !inside_model) {
+               //model_print("memcpy intercepted\n");
                thread_id_t tid = thread_current_id();
                if (((uintptr_t)src&7) == 0 && ((uintptr_t)dst&7) == 0 && (n&7) == 0) {
                        for (uint i = 0; i < (n>>3); i++) {
-                               raceCheckRead64(tid, (void *)(((char *)src) + i));
+                               raceCheckRead64(tid, (void *)(((uint64_t *)src) + i));
                                ((volatile uint64_t *)dst)[i] = ((uint64_t *)src)[i];
-                               raceCheckWrite64(tid, (void *)(((char *)src) + i));
+                               raceCheckWrite64(tid, (void *)(((uint64_t *)dst) + i));
                        }
                } else if (((uintptr_t)src&3) == 0 && ((uintptr_t)dst&3) == 0 && (n&3) == 0) {
                        for (uint i = 0; i < (n>>2); i++) {
-                               raceCheckRead32(tid, (void *)(((char *)src) + i));
+                               raceCheckRead32(tid, (void *)(((uint32_t *)src) + i));
                                ((volatile uint32_t *)dst)[i] = ((uint32_t *)src)[i];
-                               raceCheckWrite32(tid, (void *)(((char *)src) + i));
+                               raceCheckWrite32(tid, (void *)(((uint32_t *)dst) + i));
                        }
                } else if (((uintptr_t)src&1) == 0 && ((uintptr_t)dst&1) == 0 && (n&1) == 0) {
                        for (uint i = 0; i < (n>>1); i++) {
-                               raceCheckRead16(tid, (void *)(((char *)src) + i));
+                               raceCheckRead16(tid, (void *)(((uint16_t *)src) + i));
                                ((volatile uint16_t *)dst)[i] = ((uint16_t *)src)[i];
-                               raceCheckWrite16(tid, (void *)(((char *)src) + i));
+                               raceCheckWrite16(tid, (void *)(((uint16_t *)dst) + i));
                        }
                } else {
                        for(uint i=0;i<n;i++) {
                                raceCheckRead8(tid, (void *)(((char *)src) + i));
                                ((volatile char *)dst)[i] = ((char *)src)[i];
-                               raceCheckWrite8(tid, (void *)(((char *)src) + i));
+                               raceCheckWrite8(tid, (void *)(((char *)dst) + i));
                        }
                }
        } else {
@@ -176,3 +177,47 @@ void * memcpy(void * dst, const void * src, size_t n) {
        }
        return dst;
 }
+
+void * memset(void *dst, int c, size_t n) {
+       if (model && !inside_model) {
+               thread_id_t tid = thread_current_id();
+               uint8_t cs = c&0xff;
+               if (((uintptr_t)dst&7) == 0 && (n&7) == 0) {
+                       for (uint i = 0; i < (n>>3); i++) {
+                   uint16_t cs2 = cs << 8 | cs;
+                   uint64_t cs3 = cs2 << 16 | cs2;
+                   uint64_t cs4 = cs3 << 32 | cs3;
+                               ((volatile uint64_t *)dst)[i] = cs4;
+                               raceCheckWrite64(tid, (void *)(((uint64_t *)dst) + i));
+                       }
+               } else if (((uintptr_t)dst&3) == 0 && (n&3) == 0) {
+                       for (uint i = 0; i < (n>>2); i++) {
+                   uint16_t cs2 = cs << 8 | cs;
+                   uint32_t cs3 = cs2 << 16 | cs2;
+                               ((volatile uint32_t *)dst)[i] = cs3;
+                               raceCheckWrite32(tid, (void *)(((uint32_t *)dst) + i));
+                       }
+               } else if (((uintptr_t)dst&1) == 0 && (n&1) == 0) {
+                       for (uint i = 0; i < (n>>1); i++) {
+                   uint16_t cs2 = cs << 8 | cs;
+                               ((volatile uint16_t *)dst)[i] = cs2;
+                               raceCheckWrite16(tid, (void *)(((uint16_t *)dst) + i));
+                       }
+               } else {
+                       for (uint i=0;i<n;i++) {
+                               ((volatile char *)dst)[i] = cs;
+                               raceCheckWrite8(tid, (void *)(((char *)dst) + i));
+                       }
+               }
+       } else {
+               if (((uintptr_t)real_memset) < 2) {
+                       //stuck in dynamic linker alloc cycle...
+                       for(size_t s=0;s<n;s++) {
+                               ((volatile char *)dst)[s] = (char) c;
+                       }
+                       return dst;
+               }
+               return real_memset(dst, c, n);
+       }
+       return dst;
+}