target-i386: add pclmulqdq instruction

Reviewed-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

4 files changed, 41 insertions, 10 deletions

diff --git a/target-i386/cpu.c b/target-i386/cpu.c
index 356378c110..f334b54414 100644
--- a/target-i386/cpu.c
+++ b/target-i386/cpu.c
@@ -387,17 +387,16 @@ typedef struct x86_def_t {
           CPUID_PSE36 (needed for Solaris) */
           /* missing:
           CPUID_VME, CPUID_DTS, CPUID_SS, CPUID_HT, CPUID_TM, CPUID_PBE */
-#define TCG_EXT_FEATURES (CPUID_EXT_SSE3 | CPUID_EXT_MONITOR | \
-          CPUID_EXT_SSSE3 | CPUID_EXT_CX16 | CPUID_EXT_SSE41 | \
-          CPUID_EXT_SSE42 | CPUID_EXT_POPCNT | CPUID_EXT_MOVBE | \
-          CPUID_EXT_HYPERVISOR)
+#define TCG_EXT_FEATURES (CPUID_EXT_SSE3 | CPUID_EXT_PCLMULQDQ | \
+          CPUID_EXT_MONITOR | CPUID_EXT_SSSE3 | CPUID_EXT_CX16 | \
+          CPUID_EXT_SSE41 | CPUID_EXT_SSE42 | CPUID_EXT_POPCNT | \
+          CPUID_EXT_MOVBE | CPUID_EXT_HYPERVISOR)
           /* missing:
-          CPUID_EXT_PCLMULQDQ, CPUID_EXT_DTES64, CPUID_EXT_DSCPL,
-          CPUID_EXT_VMX, CPUID_EXT_SMX, CPUID_EXT_EST, CPUID_EXT_TM2,
-          CPUID_EXT_CID, CPUID_EXT_FMA, CPUID_EXT_XTPR, CPUID_EXT_PDCM,
-          CPUID_EXT_PCID, CPUID_EXT_DCA, CPUID_EXT_X2APIC,
-          CPUID_EXT_TSC_DEADLINE_TIMER, CPUID_EXT_AES, CPUID_EXT_XSAVE,
-          CPUID_EXT_OSXSAVE, CPUID_EXT_AVX, CPUID_EXT_F16C,
+          CPUID_EXT_DTES64, CPUID_EXT_DSCPL, CPUID_EXT_VMX, CPUID_EXT_SMX,
+          CPUID_EXT_EST, CPUID_EXT_TM2, CPUID_EXT_CID, CPUID_EXT_FMA,
+          CPUID_EXT_XTPR, CPUID_EXT_PDCM, CPUID_EXT_PCID, CPUID_EXT_DCA,
+          CPUID_EXT_X2APIC, CPUID_EXT_TSC_DEADLINE_TIMER, CPUID_EXT_AES,
+          CPUID_EXT_XSAVE, CPUID_EXT_OSXSAVE, CPUID_EXT_AVX, CPUID_EXT_F16C,
           CPUID_EXT_RDRAND */
 #define TCG_EXT2_FEATURES ((TCG_FEATURES & CPUID_EXT2_AMD_ALIASES) | \
           CPUID_EXT2_NX | CPUID_EXT2_MMXEXT | CPUID_EXT2_RDTSCP | \
diff --git a/target-i386/ops_sse.h b/target-i386/ops_sse.h
index a11dba1566..2ee5b8d9dd 100644
--- a/target-i386/ops_sse.h
+++ b/target-i386/ops_sse.h
@@ -2179,6 +2179,30 @@ target_ulong helper_popcnt(CPUX86State *env, target_ulong n, uint32_t type)
     return POPCOUNT(n, 5);
 #endif
 }
+
+void glue(helper_pclmulqdq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
+                                    uint32_t ctrl)
+{
+    uint64_t ah, al, b, resh, resl;
+
+    ah = 0;
+    al = d->Q((ctrl & 1) != 0);
+    b = s->Q((ctrl & 16) != 0);
+    resh = resl = 0;
+
+    while (b) {
+        if (b & 1) {
+            resl ^= al;
+            resh ^= ah;
+        }
+        ah = (ah << 1) | (al >> 63);
+        al <<= 1;
+        b >>= 1;
+    }
+
+    d->Q(0) = resl;
+    d->Q(1) = resh;
+}
 #endif
 
 #undef SHIFT
diff --git a/target-i386/ops_sse_header.h b/target-i386/ops_sse_header.h
index 401eac6f00..2842233943 100644
--- a/target-i386/ops_sse_header.h
+++ b/target-i386/ops_sse_header.h
@@ -336,6 +336,11 @@ DEF_HELPER_3(crc32, tl, i32, tl, i32)
 DEF_HELPER_3(popcnt, tl, env, tl, i32)
 #endif
 
+/* AES-NI op helpers */
+#if SHIFT == 1
+DEF_HELPER_4(glue(pclmulqdq, SUFFIX), void, env, Reg, Reg, i32)
+#endif
+
 #undef SHIFT
 #undef Reg
 #undef SUFFIX
diff --git a/target-i386/translate.c b/target-i386/translate.c
index 7596a90dc4..d649e99b3a 100644
--- a/target-i386/translate.c
+++ b/target-i386/translate.c
@@ -3147,6 +3147,8 @@ struct SSEOpHelper_eppi {
 #define SSE41_OP(x) { { NULL, gen_helper_ ## x ## _xmm }, CPUID_EXT_SSE41 }
 #define SSE42_OP(x) { { NULL, gen_helper_ ## x ## _xmm }, CPUID_EXT_SSE42 }
 #define SSE41_SPECIAL { { NULL, SSE_SPECIAL }, CPUID_EXT_SSE41 }
+#define PCLMULQDQ_OP(x) { { NULL, gen_helper_ ## x ## _xmm }, \
+        CPUID_EXT_PCLMULQDQ }
 
 static const struct SSEOpHelper_epp sse_op_table6[256] = {
     [0x00] = SSSE3_OP(pshufb),
@@ -3216,6 +3218,7 @@ static const struct SSEOpHelper_eppi sse_op_table7[256] = {
     [0x40] = SSE41_OP(dpps),
     [0x41] = SSE41_OP(dppd),
     [0x42] = SSE41_OP(mpsadbw),
+    [0x44] = PCLMULQDQ_OP(pclmulqdq),
     [0x60] = SSE42_OP(pcmpestrm),
     [0x61] = SSE42_OP(pcmpestri),
     [0x62] = SSE42_OP(pcmpistrm),