1 files changed, 59 insertions, 67 deletions

diff --git a/src/cs16-fftr/gen/scalar-x4.c b/src/cs16-fftr/gen/scalar-x4.c
index 409920e86..be23e44c1 100644
--- a/src/cs16-fftr/gen/scalar-x4.c
+++ b/src/cs16-fftr/gen/scalar-x4.c
@@ -4,8 +4,8 @@
 //
 // Copyright 2022 Google LLC
 //
-// This source code is licensed under the BSD-style license found il the
-// LICENSE file il the root directory of this source tree.
+// This source code is licensed under the BSD-style license found in the
+// LICENSE file in the root directory of this source tree.
 
 #include <assert.h>
 #include <stddef.h>
@@ -17,53 +17,49 @@
 
 void xnn_cs16_fftr_ukernel__scalar_x4(
     size_t samples,
-    const int16_t* input,
-    int16_t* output,
+    int16_t* data,
     const int16_t* twiddle) {
 
   assert(samples >= 2);
   assert(samples % 2 == 0);
-  assert(input != NULL);
-  assert(output != NULL);
+  assert(data != NULL);
+  assert(data != NULL);
   assert(twiddle != NULL);
 
-  const int16_t* il = input;
-  const int16_t* ir = input + samples * 2;
-  int32_t vdcr = (int32_t) il[0];
-  int32_t vdci = (int32_t) il[1];
-  il += 2;
+  int16_t* dl = data;
+  int16_t* dr = data + samples * 2;
+  int32_t vdcr = (int32_t) dl[0];
+  int32_t vdci = (int32_t) dl[1];
+
   vdcr = math_asr_s32(vdcr * 16383 + 16384, 15);
   vdci = math_asr_s32(vdci * 16383 + 16384, 15);
 
-  int16_t* ol  = output;
-  int16_t* or = output + samples * 2;
-  ol[0] = vdcr + vdci;
-  ol[1] = 0;
-  ol += 2;
-  or[0] = vdcr - vdci;
-  or[1] = 0;
+  dl[0] = vdcr + vdci;
+  dl[1] = 0;
+  dl += 2;
+  dr[0] = vdcr - vdci;
+  dr[1] = 0;
 
   samples >>= 1;
 
   for (; samples >= 4; samples -= 4) {
-    int32_t vilr0 = il[0];
-    int32_t vili0 = il[1];
-    int32_t vilr1 = il[2];
-    int32_t vili1 = il[3];
-    int32_t vilr2 = il[4];
-    int32_t vili2 = il[5];
-    int32_t vilr3 = il[6];
-    int32_t vili3 = il[7];
-    il += 4 * 2;
-    ir -= 4 * 2;
-    int32_t virr0 =  (int32_t) ir[6];
-    int32_t viri0 = -(int32_t) ir[7];
-    int32_t virr1 =  (int32_t) ir[4];
-    int32_t viri1 = -(int32_t) ir[5];
-    int32_t virr2 =  (int32_t) ir[2];
-    int32_t viri2 = -(int32_t) ir[3];
-    int32_t virr3 =  (int32_t) ir[0];
-    int32_t viri3 = -(int32_t) ir[1];
+    dr -= 4 * 2;
+    int32_t vilr0 = dl[0];
+    int32_t vili0 = dl[1];
+    int32_t vilr1 = dl[2];
+    int32_t vili1 = dl[3];
+    int32_t vilr2 = dl[4];
+    int32_t vili2 = dl[5];
+    int32_t vilr3 = dl[6];
+    int32_t vili3 = dl[7];
+    int32_t virr0 =  (int32_t) dr[6];
+    int32_t viri0 = -(int32_t) dr[7];
+    int32_t virr1 =  (int32_t) dr[4];
+    int32_t viri1 = -(int32_t) dr[5];
+    int32_t virr2 =  (int32_t) dr[2];
+    int32_t viri2 = -(int32_t) dr[3];
+    int32_t virr3 =  (int32_t) dr[0];
+    int32_t viri3 = -(int32_t) dr[1];
     const int32_t vtwr0 = twiddle[0];
     const int32_t vtwi0 = twiddle[1];
     const int32_t vtwr1 = twiddle[2];
@@ -116,34 +112,32 @@ void xnn_cs16_fftr_ukernel__scalar_x4(
     const int32_t twi2 = math_asr_s32(vacc2r2 * vtwi2 + vacc2i2 * vtwr2 + 16384, 15);
     const int32_t twi3 = math_asr_s32(vacc2r3 * vtwi3 + vacc2i3 * vtwr3 + 16384, 15);
 
-    ol[0] = math_asr_s32(vacc1r0 + twr0, 1);
-    ol[1] = math_asr_s32(vacc1i0 + twi0, 1);
-    ol[2] = math_asr_s32(vacc1r1 + twr1, 1);
-    ol[3] = math_asr_s32(vacc1i1 + twi1, 1);
-    ol[4] = math_asr_s32(vacc1r2 + twr2, 1);
-    ol[5] = math_asr_s32(vacc1i2 + twi2, 1);
-    ol[6] = math_asr_s32(vacc1r3 + twr3, 1);
-    ol[7] = math_asr_s32(vacc1i3 + twi3, 1);
-    ol += 4 * 2;
-    or -= 4 * 2;
-    or[6] = math_asr_s32(vacc1r0 - twr0, 1);
-    or[7] = math_asr_s32(twi0 - vacc1i0, 1);
-    or[4] = math_asr_s32(vacc1r1 - twr1, 1);
-    or[5] = math_asr_s32(twi1 - vacc1i1, 1);
-    or[2] = math_asr_s32(vacc1r2 - twr2, 1);
-    or[3] = math_asr_s32(twi2 - vacc1i2, 1);
-    or[0] = math_asr_s32(vacc1r3 - twr3, 1);
-    or[1] = math_asr_s32(twi3 - vacc1i3, 1);
+    dl[0] = math_asr_s32(vacc1r0 + twr0, 1);
+    dl[1] = math_asr_s32(vacc1i0 + twi0, 1);
+    dl[2] = math_asr_s32(vacc1r1 + twr1, 1);
+    dl[3] = math_asr_s32(vacc1i1 + twi1, 1);
+    dl[4] = math_asr_s32(vacc1r2 + twr2, 1);
+    dl[5] = math_asr_s32(vacc1i2 + twi2, 1);
+    dl[6] = math_asr_s32(vacc1r3 + twr3, 1);
+    dl[7] = math_asr_s32(vacc1i3 + twi3, 1);
+    dr[6] = math_asr_s32(vacc1r0 - twr0, 1);
+    dr[7] = math_asr_s32(twi0 - vacc1i0, 1);
+    dr[4] = math_asr_s32(vacc1r1 - twr1, 1);
+    dr[5] = math_asr_s32(twi1 - vacc1i1, 1);
+    dr[2] = math_asr_s32(vacc1r2 - twr2, 1);
+    dr[3] = math_asr_s32(twi2 - vacc1i2, 1);
+    dr[0] = math_asr_s32(vacc1r3 - twr3, 1);
+    dr[1] = math_asr_s32(twi3 - vacc1i3, 1);
+    dl += 4 * 2;
   }
 
   if XNN_UNLIKELY(samples != 0) {
     do {
-      int32_t vilr = il[0];
-      int32_t vili = il[1];
-      il += 2;
-      ir -= 2;
-      int32_t virr =  (int32_t) ir[0];
-      int32_t viri = -(int32_t) ir[1];
+      dr -= 2;
+      int32_t vilr = dl[0];
+      int32_t vili = dl[1];
+      int32_t virr =  (int32_t) dr[0];
+      int32_t viri = -(int32_t) dr[1];
       const int32_t vtwr = twiddle[0];
       const int32_t vtwi = twiddle[1];
       twiddle += 2;
@@ -160,13 +154,11 @@ void xnn_cs16_fftr_ukernel__scalar_x4(
       const int32_t twr = math_asr_s32(vacc2r * vtwr - vacc2i * vtwi + 16384, 15);
       const int32_t twi = math_asr_s32(vacc2r * vtwi + vacc2i * vtwr + 16384, 15);
 
-      ol[0] = math_asr_s32(vacc1r + twr, 1);
-      ol[1] = math_asr_s32(vacc1i + twi, 1);
-      ol += 2;
-      or -= 2;
-      or[0] = math_asr_s32(vacc1r - twr, 1);
-      or[1] = math_asr_s32(twi - vacc1i, 1);
-
+      dl[0] = math_asr_s32(vacc1r + twr, 1);
+      dl[1] = math_asr_s32(vacc1i + twi, 1);
+      dr[0] = math_asr_s32(vacc1r - twr, 1);
+      dr[1] = math_asr_s32(twi - vacc1i, 1);
+      dl += 2;
     } while (--samples != 0);
   }
 }