[Qemu-devel] [PATCH 27/28] target-ppc: Altivec 2.07: Vector SHA Sigma Instructions

[Tom Musta <tommusta@gmail.com>]

This patch adds the Vector SHA Sigma instructions introduced in Power
ISA Version 2.07:

  - Vector SHA-512 Sigma Doubleword (vshasigmad)
  - Vector SHA-256 Sigma Word (vshasigmaw)

Signed-off-by: Tom Musta <tommusta@gmail.com>
---
 target-ppc/helper.h     |    2 +
 target-ppc/int_helper.c |   82 +++++++++++++++++++++++++++++++++++++++++++++++
 target-ppc/translate.c  |   24 ++++++++++++++
 3 files changed, 108 insertions(+), 0 deletions(-)

diff --git a/target-ppc/helper.h b/target-ppc/helper.h
index 93e549e..dc0527b 100644
--- a/target-ppc/helper.h
+++ b/target-ppc/helper.h
@@ -321,6 +321,8 @@ DEF_HELPER_3(vcipher, void, avr, avr, avr)
 DEF_HELPER_3(vcipherlast, void, avr, avr, avr)
 DEF_HELPER_3(vncipher, void, avr, avr, avr)
 DEF_HELPER_3(vncipherlast, void, avr, avr, avr)
+DEF_HELPER_3(vshasigmaw, void, avr, avr, i32)
+DEF_HELPER_3(vshasigmad, void, avr, avr, i32)
 
 DEF_HELPER_4(bcdadd, i32, avr, avr, avr, i32)
 DEF_HELPER_4(bcdsub, i32, avr, avr, avr, i32)
diff --git a/target-ppc/int_helper.c b/target-ppc/int_helper.c
index cd04e8a..e6a7ad0 100644
--- a/target-ppc/int_helper.c
+++ b/target-ppc/int_helper.c
@@ -2618,6 +2618,88 @@ void helper_vncipherlast(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
     r->u64[1] = vtemp2.u64[1] ^ b->u64[1];
 }
 
+#define ROTRu32(v, n) (((v) >> (n)) | ((v) << (32-n)))
+#if defined(HOST_WORDS_BIGENDIAN)
+#define EL_IDX(i) (i)
+#else
+#define EL_IDX(i) (3 - (i))
+#endif
+
+void helper_vshasigmaw(ppc_avr_t *r,  ppc_avr_t *a, uint32_t st_six)
+{
+    int st = (st_six & 0x10) != 0;
+    int six = st_six & 0xF;
+    int i;
+
+    VECTOR_FOR_INORDER_I(i, u32) {
+        if (st == 0) {
+            if ((six & (0x8 >> i)) == 0) {
+                r->u32[EL_IDX(i)] = ROTRu32(a->u32[EL_IDX(i)], 7) ^
+                                    ROTRu32(a->u32[EL_IDX(i)], 18) ^
+                                    (a->u32[EL_IDX(i)] >> 3);
+            } else { /* six.bit[i] == 1 */
+                r->u32[EL_IDX(i)] = ROTRu32(a->u32[EL_IDX(i)], 17) ^
+                                    ROTRu32(a->u32[EL_IDX(i)], 19) ^
+                                    (a->u32[EL_IDX(i)] >> 10);
+            }
+        } else { /* st == 1 */
+            if ((six & (0x8 >> i)) == 0) {
+                r->u32[EL_IDX(i)] = ROTRu32(a->u32[EL_IDX(i)], 2) ^
+                                    ROTRu32(a->u32[EL_IDX(i)], 13) ^
+                                    ROTRu32(a->u32[EL_IDX(i)], 22);
+            } else { /* six.bit[i] == 1 */
+                r->u32[EL_IDX(i)] = ROTRu32(a->u32[EL_IDX(i)], 6) ^
+                                    ROTRu32(a->u32[EL_IDX(i)], 11) ^
+                                    ROTRu32(a->u32[EL_IDX(i)], 25);
+            }
+        }
+    }
+}
+
+#undef ROTRu32
+#undef EL_IDX
+
+#define ROTRu64(v, n) (((v) >> (n)) | ((v) << (64-n)))
+#if defined(HOST_WORDS_BIGENDIAN)
+#define EL_IDX(i) (i)
+#else
+#define EL_IDX(i) (1 - (i))
+#endif
+
+void helper_vshasigmad(ppc_avr_t *r,  ppc_avr_t *a, uint32_t st_six)
+{
+    int st = (st_six & 0x10) != 0;
+    int six = st_six & 0xF;
+    int i;
+
+    VECTOR_FOR_INORDER_I(i, u64) {
+        if (st == 0) {
+            if ((six & (0x8 >> (2*i))) == 0) {
+                r->u64[EL_IDX(i)] = ROTRu64(a->u64[EL_IDX(i)], 1) ^
+                                    ROTRu64(a->u64[EL_IDX(i)], 8) ^
+                                    (a->u64[EL_IDX(i)] >> 7);
+            } else { /* six.bit[2*i] == 1 */
+                r->u64[EL_IDX(i)] = ROTRu64(a->u64[EL_IDX(i)], 19) ^
+                                    ROTRu64(a->u64[EL_IDX(i)], 61) ^
+                                    (a->u64[EL_IDX(i)] >> 6);
+            }
+        } else { /* st == 1 */
+            if ((six & (0x8 >> (2*i))) == 0) {
+                r->u64[EL_IDX(i)] = ROTRu64(a->u64[EL_IDX(i)], 28) ^
+                                    ROTRu64(a->u64[EL_IDX(i)], 34) ^
+                                    ROTRu64(a->u64[EL_IDX(i)], 39);
+            } else { /* six.bit[2*i] == 1 */
+                r->u64[EL_IDX(i)] = ROTRu64(a->u64[EL_IDX(i)], 14) ^
+                                    ROTRu64(a->u64[EL_IDX(i)], 18) ^
+                                    ROTRu64(a->u64[EL_IDX(i)], 41);
+            }
+        }
+    }
+}
+
+#undef ROTRu64
+#undef EL_IDX
+
 #undef VECTOR_FOR_INORDER_I
 #undef HI_IDX
 #undef LO_IDX
diff --git a/target-ppc/translate.c b/target-ppc/translate.c
index fe98367..e48d756 100644
--- a/target-ppc/translate.c
+++ b/target-ppc/translate.c
@@ -7436,6 +7436,27 @@ GEN_VXFORM_DUAL(vcipher, PPC_NONE, PPC2_ALTIVEC_207,
 GEN_VXFORM_DUAL(vncipher, PPC_NONE, PPC2_ALTIVEC_207,
                 vncipherlast, PPC_NONE, PPC2_ALTIVEC_207)
 
+#define VSHASIGMA(op)                         \
+static void gen_##op(DisasContext *ctx)       \
+{                                             \
+    TCGv_ptr ra, rd;                          \
+    TCGv_i32 st_six;                          \
+    if (unlikely(!ctx->altivec_enabled)) {    \
+        gen_exception(ctx, POWERPC_EXCP_VPU); \
+        return;                               \
+    }                                         \
+    ra = gen_avr_ptr(rA(ctx->opcode));        \
+    rd = gen_avr_ptr(rD(ctx->opcode));        \
+    st_six = tcg_const_i32(rB(ctx->opcode));  \
+    gen_helper_##op(rd, ra, st_six);          \
+    tcg_temp_free_ptr(ra);                    \
+    tcg_temp_free_ptr(rd);                    \
+    tcg_temp_free_i32(st_six);                \
+}
+
+VSHASIGMA(vshasigmaw)
+VSHASIGMA(vshasigmad)
+
 /***                           VSX extension                               ***/
 
 static inline TCGv_i64 cpu_vsrh(int n)
@@ -10671,6 +10692,9 @@ GEN_VXFORM_207(vsbox, 4, 23),
 GEN_VXFORM_DUAL(vcipher, vcipherlast, 4, 20, PPC_NONE, PPC2_ALTIVEC_207),
 GEN_VXFORM_DUAL(vncipher, vncipherlast, 4, 21, PPC_NONE, PPC2_ALTIVEC_207),
 
+GEN_VXFORM_207(vshasigmaw, 1, 26),
+GEN_VXFORM_207(vshasigmad, 1, 27),
+
 GEN_HANDLER_E(lxsdx, 0x1F, 0x0C, 0x12, 0, PPC_NONE, PPC2_VSX),
 GEN_HANDLER_E(lxsiwax, 0x1F, 0x0C, 0x02, 0, PPC_NONE, PPC2_VSX207),
 GEN_HANDLER_E(lxsiwzx, 0x1F, 0x0C, 0x00, 0, PPC_NONE, PPC2_VSX207),
-- 
1.7.1