feat: add pthread parallel executionHEADmain

1 files changed, 33 insertions, 1 deletions

diff --git a/src/maths/simd.c b/src/maths/simd.c
index fc25de8..5daebbb 100644
--- a/src/maths/simd.c
+++ b/src/maths/simd.c
@@ -3,6 +3,8 @@
 #include <assert.h>
 #include <stdlib.h>
 
+#include "parallel.h"
+
 #if defined(__x86_64__) || defined(__i386__)
 #include <immintrin.h>
 #elif defined(__aarch64__)
@@ -316,9 +318,39 @@ void psi_apply_single_qubit_gate_simd(struct PsiComplex* state, const struct Psi
     apply_scalar(state, gate, target, num_qubits);
 }
 
+struct PsiSimdPairs
+{
+    struct PsiComplex* state;
+    size_t (*pairs)[2];
+    struct PsiComplex g00;
+    struct PsiComplex g01;
+    struct PsiComplex g10;
+    struct PsiComplex g11;
+};
+
+static void simd_pair_range(size_t start, size_t end, void* vctx)
+{
+    struct PsiSimdPairs* c = vctx;
+
+    for (size_t p = start; p < end; p++)
+        apply_pair(c->state, c->pairs[p][0], c->pairs[p][1], c->g00, c->g01, c->g10, c->g11);
+}
+
 void psi_apply_single_qubit_gate_simd_parallel(struct PsiComplex* state,
                                                const struct PsiComplex gate[2][2], size_t target,
                                                size_t num_qubits)
 {
-    apply_scalar(state, gate, target, num_qubits);
+    size_t target_bit = num_qubits - 1 - target;
+    size_t step = (size_t)1 << target_bit;
+    size_t dim = (size_t)1 << num_qubits;
+
+    size_t np;
+    size_t (*pairs)[2] = build_pairs(dim, target_bit, step, &np);
+
+    struct PsiSimdPairs ctx = {
+        state, pairs, gate[0][0], gate[0][1], gate[1][0], gate[1][1],
+    };
+    psi_parallel_for(np, simd_pair_range, &ctx);
+
+    free(pairs);
 }