Reduce the number of registers used in MIPS optimizations.
This change is needed by ChromeOS as it introduces -fno-omit-frame-pointer
flag (see code.google.com/p/chromium/issues/detail?id=477749). This causes
compile error for MIPS, as some MIPS optimization blocks use maximum possible
number of available registers.
Also, this change contains minor GN build fix for MIPS platform regarding the
pitch_filter_mips.c / pitch_filter_c.c file inclusion.
BUG=477749
R=andrew@webrtc.org, djordje.pesut@imgtec.com, tina.legrand@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/48139004
Patch from Ljubomir Papuga <lpapuga@mips.com>.
Cr-Commit-Position: refs/heads/master@{#9047}
diff --git a/webrtc/modules/audio_coding/BUILD.gn b/webrtc/modules/audio_coding/BUILD.gn
index 8db2930..790b1a9 100644
--- a/webrtc/modules/audio_coding/BUILD.gn
+++ b/webrtc/modules/audio_coding/BUILD.gn
@@ -498,7 +498,9 @@
"codecs/isac/fix/source/lpc_tables.h",
"codecs/isac/fix/source/pitch_estimator.c",
"codecs/isac/fix/source/pitch_estimator.h",
+ "codecs/isac/fix/source/pitch_estimator_c.c",
"codecs/isac/fix/source/pitch_filter.c",
+ "codecs/isac/fix/source/pitch_filter_c.c",
"codecs/isac/fix/source/pitch_gain_tables.c",
"codecs/isac/fix/source/pitch_gain_tables.h",
"codecs/isac/fix/source/pitch_lag_tables.c",
@@ -546,8 +548,9 @@
"codecs/isac/fix/source/lattice_armv7.S",
"codecs/isac/fix/source/pitch_filter_armv6.S",
]
- } else {
- sources += [ "codecs/isac/fix/source/pitch_filter_c.c" ]
+ sources -= [
+ "codecs/isac/fix/source/pitch_filter_c.c",
+ ]
}
if (current_cpu == "mipsel") {
@@ -558,19 +561,23 @@
"codecs/isac/fix/source/pitch_estimator_mips.c",
"codecs/isac/fix/source/transform_mips.c",
]
+ sources -= [
+ "codecs/isac/fix/source/pitch_estimator_c.c"
+ ]
if (mips_dsp_rev > 0) {
- sources += [ "codecs/isac/fix/source/filterbanks_mips.c" ]
+ sources += [
+ "codecs/isac/fix/source/filterbanks_mips.c"
+ ]
}
if (mips_dsp_rev > 1) {
sources += [
"codecs/isac/fix/source/lpc_masking_model_mips.c",
"codecs/isac/fix/source/pitch_filter_mips.c",
]
- } else {
- sources += [ "codecs/isac/fix/source/pitch_filter_c.c" ]
+ sources -= [
+ "codecs/isac/fix/source/pitch_filter_c.c"
+ ]
}
- } else {
- sources += [ "codecs/isac/fix/source/pitch_estimator_c.c" ]
}
if (!rtc_build_armv7_neon && current_cpu != "mipsel") {
diff --git a/webrtc/modules/audio_coding/codecs/isac/fix/source/transform_mips.c b/webrtc/modules/audio_coding/codecs/isac/fix/source/transform_mips.c
index 656a77e..e5d35f2 100644
--- a/webrtc/modules/audio_coding/codecs/isac/fix/source/transform_mips.c
+++ b/webrtc/modules/audio_coding/codecs/isac/fix/source/transform_mips.c
@@ -606,7 +606,7 @@
int32_t* outre2;
int16_t* cosptr = (int16_t*)WebRtcIsacfix_kCosTab2;
int16_t* sinptr = (int16_t*)WebRtcIsacfix_kSinTab2;
- int32_t r0, r1, r2, r3, r4, r5, r6, r7, r8, r9, r10, max, max1;
+ int32_t r0, r1, r2, r3, r4, r5, r6, r7, r8, r9, max, max1;
#if defined(MIPS_DSP_R1_LE)
int32_t offset = FRAMESAMPLES - 4;
#else // #if defined(MIPS_DSP_R1_LE)
@@ -658,14 +658,14 @@
"subu %[r9], %[r9], %[r8] \n\t"
"subu %[r7], %[r6], %[r9] \n\t"
"addu %[r6], %[r6], %[r9] \n\t"
- "sll %[r10], %[offset], 1 \n\t"
- "addu %[r10], %[outre1], %[r10] \n\t"
"sw %[r7], 0(%[outre1]) \n\t"
"absq_s.w %[r7], %[r7] \n\t"
- "sw %[r6], 4(%[r10]) \n\t"
- "absq_s.w %[r6], %[r6] \n\t"
"slt %[r8], %[max], %[r7] \n\t"
"movn %[max], %[r7], %[r8] \n\t"
+ "sll %[r7], %[offset], 1 \n\t"
+ "addu %[r7], %[outre1], %[r7] \n\t"
+ "sw %[r6], 4(%[r7]) \n\t"
+ "absq_s.w %[r6], %[r6] \n\t"
"slt %[r8], %[max], %[r6] \n\t"
"movn %[max], %[r6], %[r8] \n\t"
"muleq_s.w.phl %[r6], %[r0], %[r2] \n\t"
@@ -682,10 +682,12 @@
"addu %[r6], %[r6], %[r9] \n\t"
"sw %[r7], 4(%[outre1]) \n\t"
"absq_s.w %[r7], %[r7] \n\t"
- "sw %[r6], 0(%[r10]) \n\t"
- "absq_s.w %[r6], %[r6] \n\t"
"slt %[r8], %[max], %[r7] \n\t"
"movn %[max], %[r7], %[r8] \n\t"
+ "sll %[r7], %[offset], 1 \n\t"
+ "addu %[r7], %[outre1], %[r7] \n\t"
+ "sw %[r6], 0(%[r7]) \n\t"
+ "absq_s.w %[r6], %[r6] \n\t"
"slt %[r8], %[max], %[r6] \n\t"
"movn %[max], %[r6], %[r8] \n\t"
"muleq_s.w.phr %[r6], %[r1], %[r2] \n\t"
@@ -702,14 +704,14 @@
"subu %[r7], %[r6], %[r9] \n\t"
"addu %[r6], %[r9], %[r6] \n\t"
"negu %[r6], %[r6] \n\t"
- "sll %[r10], %[offset], 1 \n\t"
- "addu %[r10], %[outre2], %[r10] \n\t"
"sw %[r7], 0(%[outre2]) \n\t"
"absq_s.w %[r7], %[r7] \n\t"
- "sw %[r6], 4(%[r10]) \n\t"
- "absq_s.w %[r6], %[r6] \n\t"
"slt %[r8], %[max], %[r7] \n\t"
"movn %[max], %[r7], %[r8] \n\t"
+ "sll %[r7], %[offset], 1 \n\t"
+ "addu %[r7], %[outre2], %[r7] \n\t"
+ "sw %[r6], 4(%[r7]) \n\t"
+ "absq_s.w %[r6], %[r6] \n\t"
"slt %[r8], %[max], %[r6] \n\t"
"movn %[max], %[r6], %[r8] \n\t"
"muleq_s.w.phl %[r6], %[r1], %[r2] \n\t"
@@ -728,10 +730,12 @@
"negu %[r6], %[r6] \n\t"
"sw %[r7], 4(%[outre2]) \n\t"
"absq_s.w %[r7], %[r7] \n\t"
- "sw %[r6], 0(%[r10]) \n\t"
- "absq_s.w %[r6], %[r6] \n\t"
"slt %[r8], %[max], %[r7] \n\t"
"movn %[max], %[r7], %[r8] \n\t"
+ "sll %[r7], %[offset], 1 \n\t"
+ "addu %[r7], %[outre2], %[r7] \n\t"
+ "sw %[r6], 0(%[r7]) \n\t"
+ "absq_s.w %[r6], %[r6] \n\t"
"slt %[r8], %[max], %[r6] \n\t"
"movn %[max], %[r6], %[r8] \n\t"
"bgtz %[k], 1b \n\t"
@@ -824,8 +828,8 @@
[offset] "+r" (offset), [k] "+r" (k), [r0] "=&r" (r0),
[r1] "=&r" (r1), [r2] "=&r" (r2), [r3] "=&r" (r3),
[r4] "=&r" (r4), [r5] "=&r" (r5), [r6] "=&r" (r6),
- [r7] "=&r" (r7), [r10] "=&r" (r10),
- [r8] "=&r" (r8), [r9] "=&r" (r9), [max] "=&r" (max)
+ [r7] "=&r" (r7), [r8] "=&r" (r8), [r9] "=&r" (r9),
+ [max] "=&r" (max)
: [inreQ7] "r" (inreQ7), [inimQ7] "r" (inimQ7),
[cosptr] "r" (cosptr), [sinptr] "r" (sinptr),
[outre1Q16] "r" (outre1Q16), [outre2Q16] "r" (outre2Q16)
