Thom Johansen's first EMAC optimisation for the Coldfire - about a 3%-4% speedup

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@6024 a1c6a512-1295-4272-9138-f99709370657

1 files changed, 165 insertions, 0 deletions

diff --git a/apps/codecs/libFLAC/coldfire.c b/apps/codecs/libFLAC/coldfire.c
new file mode 100644
index 0000000..c763c20
--- /dev/null
+++ b/apps/codecs/libFLAC/coldfire.c
@@ -0,0 +1,165 @@
+#ifndef SIMULATOR
+#include <private/coldfire.h>
+
+void FLAC__lpc_restore_signal_order8_mac(const FLAC__int32 residual[], unsigned data_len, const FLAC__int32 qlp_coeff[], unsigned order, int lp_quantization, FLAC__int32 data[])
+{
+	register const FLAC__int32 *qlp0 = &qlp_coeff[(order-1)];
+	register FLAC__int32 sum;
+	register const FLAC__int32 *history;
+	
+	SET_MACSR(0);
+	history = &data[(-order)];
+	SET_ACC(0, acc0);
+	
+	switch (order) {
+	case 8:
+		for( ; data_len != 0; --data_len) {
+			asm volatile(
+				"mov.l (%1), %%d0\n\t"
+				"mov.l (%2), %%d1\n\t"
+				"mac.l %%d0, %%d1, 4(%2), %%d1, %%acc0\n\t"
+				"mov.l -4(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 8(%2), %%d1, %%acc0\n\t"
+				"mov.l -8(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 12(%2), %%d1, %%acc0\n\t"
+				"mov.l -12(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 16(%2), %%d1, %%acc0\n\t"
+				"mov.l -16(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 20(%2), %%d1, %%acc0\n\t"
+				"mov.l -20(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 24(%2), %%d1, %%acc0\n\t"
+				"mov.l -24(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 28(%2), %%d1, %%acc0\n\t"
+				"mov.l -28(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, %%acc0\n\t"
+				"movclr.l %%acc0, %0"
+				: "=ad" (sum) : "a" (qlp0), "a" (history) : "d0", "d1");
+			++history;
+			*(data++) = *(residual++) + (sum >> lp_quantization);
+		}
+		return;
+	case 7:
+		for( ; data_len != 0; --data_len) {
+			asm volatile(
+				"mov.l (%1), %%d0\n\t"
+				"mov.l (%2), %%d1\n\t"
+				"mac.l %%d0, %%d1, 4(%2), %%d1, %%acc0\n\t"
+				"mov.l -4(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 8(%2), %%d1, %%acc0\n\t"
+				"mov.l -8(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 12(%2), %%d1, %%acc0\n\t"
+				"mov.l -12(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 16(%2), %%d1, %%acc0\n\t"
+				"mov.l -16(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 20(%2), %%d1, %%acc0\n\t"
+				"mov.l -20(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 24(%2), %%d1, %%acc0\n\t"
+				"mov.l -24(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, %%acc0\n\t"
+				"movclr.l %%acc0, %0"
+				: "=ad" (sum) : "a" (qlp0), "a" (history) : "d0", "d1");
+			++history;
+			*(data++) = *(residual++) + (sum >> lp_quantization);
+		} 
+		return;
+	case 6:
+		for( ; data_len != 0; --data_len) {
+			asm volatile(
+				"mov.l (%1), %%d0\n\t"
+				"mov.l (%2), %%d1\n\t"
+				"mac.l %%d0, %%d1, 4(%2), %%d1, %%acc0\n\t"
+				"mov.l -4(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 8(%2), %%d1, %%acc0\n\t"
+				"mov.l -8(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 12(%2), %%d1, %%acc0\n\t"
+				"mov.l -12(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 16(%2), %%d1, %%acc0\n\t"
+				"mov.l -16(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 20(%2), %%d1, %%acc0\n\t"
+				"mov.l -20(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, %%acc0\n\t"
+				"movclr.l %%acc0, %0"
+				: "=ad" (sum) : "a" (qlp0), "a" (history) : "d0", "d1");
+			++history;
+			*(data++) = *(residual++) + (sum >> lp_quantization);
+		} 
+		return;
+	case 5:
+		for( ; data_len != 0; --data_len) {
+			asm volatile(
+				"mov.l (%1), %%d0\n\t"
+				"mov.l (%2), %%d1\n\t"
+				"mac.l %%d0, %%d1, 4(%2), %%d1, %%acc0\n\t"
+				"mov.l -4(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 8(%2), %%d1, %%acc0\n\t"
+				"mov.l -8(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 12(%2), %%d1, %%acc0\n\t"
+				"mov.l -12(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 16(%2), %%d1, %%acc0\n\t"
+				"mov.l -16(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, %%acc0\n\t"
+				"movclr.l %%acc0, %0"
+				: "=ad" (sum) : "a" (qlp0), "a" (history) : "d0", "d1");
+			++history;
+			*(data++) = *(residual++) + (sum >> lp_quantization);
+		} 
+		return;
+	case 4:
+		for( ; data_len != 0; --data_len) {
+			asm volatile(
+				"mov.l (%1), %%d0\n\t"
+				"mov.l (%2), %%d1\n\t"
+				"mac.l %%d0, %%d1, 4(%2), %%d1, %%acc0\n\t"
+				"mov.l -4(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 8(%2), %%d1, %%acc0\n\t"
+				"mov.l -8(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 12(%2), %%d1, %%acc0\n\t"
+				"mov.l -12(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, %%acc0\n\t"
+				"movclr.l %%acc0, %0"
+				: "=ad" (sum) : "a" (qlp0), "a" (history) : "d0", "d1");
+			++history;
+			*(data++) = *(residual++) + (sum >> lp_quantization);
+		} 
+		return;
+	case 3:
+		for( ; data_len != 0; --data_len) {
+			asm volatile(
+				"mov.l (%1), %%d0\n\t"
+				"mov.l (%2), %%d1\n\t"
+				"mac.l %%d0, %%d1, 4(%2), %%d1, %%acc0\n\t"
+				"mov.l -4(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, 8(%2), %%d1, %%acc0\n\t"
+				"mov.l -8(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, %%acc0\n\t"
+				"movclr.l %%acc0, %0"
+				: "=ad" (sum) : "a" (qlp0), "a" (history) : "d0", "d1");
+			++history;
+			*(data++) = *(residual++) + (sum >> lp_quantization);
+		}
+		return;
+	case 2:
+		for( ; data_len != 0; --data_len) {
+			asm volatile(
+				"mov.l (%1), %%d0\n\t"
+				"mov.l (%2), %%d1\n\t"
+				"mac.l %%d0, %%d1, 4(%2), %%d1, %%acc0\n\t"
+				"mov.l -4(%1), %%d0\n\t"
+				"mac.l %%d0, %%d1, %%acc0\n\t"
+				"movclr.l %%acc0, %0"
+				: "=ad" (sum) : "a" (qlp0), "a" (history) : "d0", "d1");
+			++history;
+			*(data++) = *(residual++) + (sum >> lp_quantization);
+		}
+		return;
+	case 1:
+		// won't gain anything by using mac here. 
+		for( ; data_len != 0; --data_len) {
+			sum = (qlp0[0] * (*(history++)));
+			*(data++) = *(residual++) + (sum >> lp_quantization);
+		} 
+		return;
+	}
+}
+
+#endif