target/arm: Vectorize USHL and SSHL

These instructions shift left or right depending on the sign
of the input, and 7 bits are significant to the shift.  This
requires several masks and selects in addition to the actual
shifts to form the complete answer.

That said, the operation is still a small improvement even for
two 64-bit elements -- 13 vector operations instead of 2 * 7
integer operations.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20200216214232.4230-2-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

1 files changed, 88 insertions, 0 deletions

diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c
index dedef62403..fcb3663903 100644
--- a/target/arm/vec_helper.c
+++ b/target/arm/vec_helper.c
@@ -1046,3 +1046,91 @@ void HELPER(gvec_fmlal_idx_a64)(void *vd, void *vn, void *vm,
     do_fmlal_idx(vd, vn, vm, &env->vfp.fp_status, desc,
                  get_flush_inputs_to_zero(&env->vfp.fp_status_f16));
 }
+
+void HELPER(gvec_sshl_b)(void *vd, void *vn, void *vm, uint32_t desc)
+{
+    intptr_t i, opr_sz = simd_oprsz(desc);
+    int8_t *d = vd, *n = vn, *m = vm;
+
+    for (i = 0; i < opr_sz; ++i) {
+        int8_t mm = m[i];
+        int8_t nn = n[i];
+        int8_t res = 0;
+        if (mm >= 0) {
+            if (mm < 8) {
+                res = nn << mm;
+            }
+        } else {
+            res = nn >> (mm > -8 ? -mm : 7);
+        }
+        d[i] = res;
+    }
+    clear_tail(d, opr_sz, simd_maxsz(desc));
+}
+
+void HELPER(gvec_sshl_h)(void *vd, void *vn, void *vm, uint32_t desc)
+{
+    intptr_t i, opr_sz = simd_oprsz(desc);
+    int16_t *d = vd, *n = vn, *m = vm;
+
+    for (i = 0; i < opr_sz / 2; ++i) {
+        int8_t mm = m[i];   /* only 8 bits of shift are significant */
+        int16_t nn = n[i];
+        int16_t res = 0;
+        if (mm >= 0) {
+            if (mm < 16) {
+                res = nn << mm;
+            }
+        } else {
+            res = nn >> (mm > -16 ? -mm : 15);
+        }
+        d[i] = res;
+    }
+    clear_tail(d, opr_sz, simd_maxsz(desc));
+}
+
+void HELPER(gvec_ushl_b)(void *vd, void *vn, void *vm, uint32_t desc)
+{
+    intptr_t i, opr_sz = simd_oprsz(desc);
+    uint8_t *d = vd, *n = vn, *m = vm;
+
+    for (i = 0; i < opr_sz; ++i) {
+        int8_t mm = m[i];
+        uint8_t nn = n[i];
+        uint8_t res = 0;
+        if (mm >= 0) {
+            if (mm < 8) {
+                res = nn << mm;
+            }
+        } else {
+            if (mm > -8) {
+                res = nn >> -mm;
+            }
+        }
+        d[i] = res;
+    }
+    clear_tail(d, opr_sz, simd_maxsz(desc));
+}
+
+void HELPER(gvec_ushl_h)(void *vd, void *vn, void *vm, uint32_t desc)
+{
+    intptr_t i, opr_sz = simd_oprsz(desc);
+    uint16_t *d = vd, *n = vn, *m = vm;
+
+    for (i = 0; i < opr_sz / 2; ++i) {
+        int8_t mm = m[i];   /* only 8 bits of shift are significant */
+        uint16_t nn = n[i];
+        uint16_t res = 0;
+        if (mm >= 0) {
+            if (mm < 16) {
+                res = nn << mm;
+            }
+        } else {
+            if (mm > -16) {
+                res = nn >> -mm;
+            }
+        }
+        d[i] = res;
+    }
+    clear_tail(d, opr_sz, simd_maxsz(desc));
+}