diff --git a/example/wiffle_screen_space_store_queue.cpp b/example/wiffle_screen_space_store_queue.cpp
index cb2967a..d308c71 100644
--- a/example/wiffle_screen_space_store_queue.cpp
+++ b/example/wiffle_screen_space_store_queue.cpp
@@ -272,7 +272,7 @@ void make_temp()
     sum += n & 0xff;
     n >>= 8;
     sum += n & 0xff;
-    temp[i] = (float)(sum * 0.25);
+    temp[i] = (float)(sum);
   }
 }
 
@@ -425,5 +425,8 @@ void main()
     //serial::string("convolve end\n");
 
     theta += half_degree;
+    frame_ix += 1;
+    if (frame_ix > 20)
+      break;
   }
 }
diff --git a/sobel_fipr.s b/sobel_fipr.s
index 0909ce1..3898d71 100644
--- a/sobel_fipr.s
+++ b/sobel_fipr.s
@@ -28,7 +28,9 @@ __setup:
         /* constants */
         mova _const_100f,r0 /* r11 as temporary */
         fmov.s @r0,fr0
+        fschg
         fmov dr0,xd0
+        fschg
 
         /* save C arguments */
         mov r4,r0 /* r4 saved as r0 */
@@ -118,7 +120,9 @@ _loop_width:
         fsts FPUL,fr3
         fadd fr3,fr7
 
-        fmov dr0,xd0 /* load 100.f constant */
+        fschg
+        fmov xd0,dr0 /* load 100.f constant */
+        fschg
 
         add #4,r0 /* next pixel */
 
diff --git a/sobel_fipr_inner.s b/sobel_fipr_inner.s
index 82a8d5b..c8f5f14 100644
--- a/sobel_fipr_inner.s
+++ b/sobel_fipr_inner.s
@@ -1,13 +1,15 @@
         /* y multiplication */
+        mov #4,r1           /* r1 : temporary */
         fmov.s @r0,fr0      /* 0 */
+        mov #8,r2           /* r2 : temporary */
         fmov.s @(r0,r1),fr1 /* 1 */
         fmov.s @(r0,r2),fr2 /* 2 */
         fldi0 fr3
         fipr fv8,fv0
 
-        fmov.s @(r0,r5),fr4 /* 1280 */
-        fmov.s @(r0,r6),fr5 /* 1281 */
-        fmov.s @(r0,r7),fr6 /* 1282 */
+        fmov.s @(r0,r5),fr4 /* r0 + 1280 */
+        fmov.s @(r0,r6),fr5 /* r0 + 1281 */
+        fmov.s @(r0,r7),fr6 /* r0 + 1282 */
         fldi0 fr7
         fipr fv12,fv4
 
@@ -34,10 +36,10 @@
         fmov fr2,fr4
         fmov fr3,fr2
         /* load fr1,fr5 */
-        fmov.s @(r0,r3),fr1 /* 640 */
+        fmov.s @(r0,r3),fr1 /* r0 + 640 */
         fldi0 fr3
         fipr fv8,fv0
-        fmov.s @(r0,r4),fr5 /* 642 */
+        fmov.s @(r0,r4),fr5 /* r0 + 642 */
         fldi0 fr7
         fipr fv12,fv4
 
@@ -47,9 +49,10 @@
         fsts FPUL,fr3
         fadd fr3,fr7
 
-        fmov dr0,xd0 /* load 100.f constant */
-
+        fschg
+        fmov xd0,dr0 /* load 100.f constant */
         fcmp/gt fr0,fr7
+        fschg
         movt r9
         add #-1,r9
 
diff --git a/sobel_fipr_store_queue.s b/sobel_fipr_store_queue.s
index 06f1426..5d4c734 100644
--- a/sobel_fipr_store_queue.s
+++ b/sobel_fipr_store_queue.s
@@ -2,8 +2,8 @@
         .global _sobel_fipr_store_queue
 _sobel_fipr_store_queue:
         /* r0:  var   (input address) */
-        /* r1:  const 1     4         */
-        /* r2:  const 2     4         */
+        /* r1:  (temporary)         */
+        /* r2:  (temporary)         */
         /* r3:  const 640   4         */
         /* r4:  const 642   4         */
         /* r5:  const 1280  4         */
@@ -45,6 +45,9 @@ __setup:
         /* constants */
         mova _const_100f,r0    /* use r0 as temporary */
         fmov.s @r0,fr0
+        fschg
+        fmov dr0,xd0
+        fschg
 
         /* set qacr0 */
         mov r5,r0              /* r5: C argument */
@@ -66,8 +69,6 @@ __setup:
         mov r4,r0 /* r4 saved as r0 */
 
         /* offsets */
-        mov #(1 * 4),r1
-        mov #(2 * 4),r2
         mov.w _const_640,r3
         mov.w _const_642,r4
         mov.w _const_1280,r5
@@ -81,7 +82,7 @@ __setup:
         nop
 
         .align 4
-_const_100f:    .float 100
+_const_100f:    .float 1900
 
 _const_store_queue:             .long 0xe0000000
 _const_store_queue_mask:        .long 0x03ffffc0 /* (0xffffffff & (~0b111111)) & (~(0b111111 << 26)) */