Merge pull request #1270 from ben-xo/feature/avr-clockless-trinket-negative-advanceBy

Feature/avr clockless trinket negative advance by

2 files changed, 12 insertions, 6 deletions

diff --git a/src/controller.h b/src/controller.h
index fe32d70d..7b7a7cf9 100644
--- a/src/controller.h
+++ b/src/controller.h
@@ -404,7 +404,11 @@ protected:
     ///@param nLeds the number of leds being written out
     ///@param scale the rgb scaling to apply to each led before writing it out
     virtual void show(const struct CRGB *data, int nLeds, CRGB scale) {
-        PixelController<RGB_ORDER, LANES, MASK> pixels(data, nLeds, scale, getDither());
+        PixelController<RGB_ORDER, LANES, MASK> pixels(data, nLeds < 0 ? -nLeds : nLeds, scale, getDither());
+        if(nLeds < 0) {
+            // nLeds < 0 implies that we want to show them in reverse
+            pixels.mAdvance = -pixels.mAdvance;
+        }
         showPixels(pixels);
     }
 
diff --git a/src/platforms/avr/clockless_trinket.h b/src/platforms/avr/clockless_trinket.h
index 45143d91..49de3b91 100644
--- a/src/platforms/avr/clockless_trinket.h
+++ b/src/platforms/avr/clockless_trinket.h
@@ -248,13 +248,13 @@ protected:
 #define PRESCALEB4(D) asm __volatile__("brcc L_%=\n\tldi %[scale_base], 0xFF\n\tL_%=:\n\tneg %[" #D "]\n\tCLC" ASM_VARS);
 
 // Clamp for prescale, increment data, since we won't ever wrap 65k, this also effectively clears carry for us
-#define PSBIDATA4(D) asm __volatile__("brcc L_%=\n\tldi %[scale_base], 0xFF\n\tL_%=:\n\tadd %A[data], %[ADV]\n\tadc %B[data], __zero_reg__\n\t" ASM_VARS);
+#define PSBIDATA4(D) asm __volatile__("brcc L_%=\n\tldi %[scale_base], 0xFF\n\tL_%=:\n\tadd %A[data], %A[ADV]\n\tadc %B[data], %B[ADV]\n\t" ASM_VARS);
 
 #else
 #define PRESCALE4(D) _dc<4>(loopvar);
 #define PRESCALEA2(D) _dc<2>(loopvar);
 #define PRESCALEB4(D) _dc<4>(loopvar);
-#define PSBIDATA4(D) asm __volatile__( "add %A[data], %[ADV]\n\tadc %B[data], __zero_reg__\n\trjmp .+0\n\t" ASM_VARS );
+#define PSBIDATA4(D) asm __volatile__( "add %A[data], %A[ADV]\n\tadc %B[data], %B[ADV]\n\trjmp .+0\n\t" ASM_VARS );
 #endif
 
 // 2 cycles - perform one step of the scaling (if a given bit is set in scale, add scale-base to the scratch space)
@@ -306,8 +306,8 @@ protected:
 #define DONE asm __volatile__("2:" ASM_VARS );
 
 // 2 cycles - increment the data pointer
-#define IDATA2 asm __volatile__("add %A[data], %[ADV]\n\tadc %B[data], __zero_reg__\n\t"  ASM_VARS );
-#define IDATACLC3 asm __volatile__("add %A[data], %[ADV]\n\tadc %B[data], __zero_reg__\n\t" _CLC1  ASM_VARS );
+#define IDATA2 asm __volatile__("add %A[data], %A[ADV]\n\tadc %B[data], %B[ADV]\n\t"  ASM_VARS );
+#define IDATACLC3 asm __volatile__("add %A[data], %A[ADV]\n\tadc %B[data], %B[ADV]\n\t" _CLC1  ASM_VARS );
 
 // 1 cycle mov
 #define _MOV1(B1, B2) "mov %[" #B1 "], %[" #B2 "]\n\t"
@@ -375,7 +375,9 @@ protected:
 		pixels.preStepFirstByteDithering();
 
 		// pull the dithering/adjustment values out of the pixels object for direct asm access
-		uint8_t advanceBy = pixels.advanceBy();
+
+		// even though advanceBy is only an int8, we cast it to int16 for sign extension in case it's negative.
+		int16_t advanceBy = pixels.advanceBy();
 		uint16_t count = pixels.mLen;
 
 		uint8_t s0 = pixels.mScale.raw[RO(0)];